diff options
Diffstat (limited to 'src/gallium')
301 files changed, 14554 insertions, 6300 deletions
diff --git a/src/gallium/SConscript b/src/gallium/SConscript index 0c632ac2b8..32b7c2980a 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -22,6 +22,7 @@ SConscript([ 'auxiliary/draw/SConscript', 'auxiliary/pipebuffer/SConscript', 'auxiliary/indices/SConscript', + 'auxiliary/rbug/SConscript', ]) for driver in env['drivers']: @@ -29,6 +30,7 @@ for driver in env['drivers']: SConscript('state_trackers/python/SConscript') SConscript('state_trackers/glx/xlib/SConscript') +SConscript('state_trackers/dri/SConscript') if platform == 'windows': SConscript('state_trackers/wgl/SConscript') diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h index dbad8f98ac..479250729f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.h +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -57,6 +57,7 @@ struct draw_stage struct draw_context *draw; /**< parent context */ struct draw_stage *next; /**< next stage in pipeline */ + const char *name; /**< for debugging */ struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ unsigned nr_tmps; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index ca69f0b95e..9f956715a2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -60,8 +60,6 @@ struct aaline_fragment_shader struct pipe_shader_state state; void *driver_fs; void *aaline_fs; - void *aapoint_fs; /* not yet */ - void *sprite_fs; /* not yet */ uint sampler_unit; int generic_attrib; /**< texcoord/generic used for texture */ }; @@ -373,10 +371,15 @@ generate_aaline_fs(struct aaline_stage *aaline) aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); if (aaline->fs->aaline_fs == NULL) - return FALSE; + goto fail; aaline->fs->generic_attrib = transform.maxGeneric + 1; + FREE((void *)aaline_fs.tokens); return TRUE; + +fail: + FREE((void *)aaline_fs.tokens); + return FALSE; } @@ -746,6 +749,7 @@ draw_aaline_stage(struct draw_context *draw) goto fail; aaline->stage.draw = draw; + aaline->stage.name = "aaline"; aaline->stage.next = NULL; aaline->stage.point = draw_pipe_passthrough_point; aaline->stage.line = aaline_first_line; @@ -815,6 +819,10 @@ aaline_delete_fs_state(struct pipe_context *pipe, void *fs) struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; /* pass-through */ aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs); + + if (aafs->aaline_fs) + aaline->driver_delete_fs_state(aaline->pipe, aafs->aaline_fs); + FREE(aafs); } diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 3133abe5dc..ae1712fe12 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -523,11 +523,15 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) aapoint->fs->aapoint_fs = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); if (aapoint->fs->aapoint_fs == NULL) - return FALSE; + goto fail; aapoint->fs->generic_attrib = transform.maxGeneric + 1; - + FREE((void *)aapoint_fs.tokens); return TRUE; + +fail: + FREE((void *)aapoint_fs.tokens); + return FALSE; } @@ -757,6 +761,7 @@ draw_aapoint_stage(struct draw_context *draw) goto fail; aapoint->stage.draw = draw; + aapoint->stage.name = "aapoint"; aapoint->stage.next = NULL; aapoint->stage.point = aapoint_first_point; aapoint->stage.line = draw_pipe_passthrough_line; @@ -824,8 +829,13 @@ aapoint_delete_fs_state(struct pipe_context *pipe, void *fs) { struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; + /* pass-through */ aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs); + + if (aafs->aapoint_fs) + aapoint->driver_delete_fs_state(aapoint->pipe, aafs->aapoint_fs); + FREE(aafs); } diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 3265dcd154..0670268a19 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -496,6 +496,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) goto fail; clipper->stage.draw = draw; + clipper->stage.name = "clipper"; clipper->stage.point = clip_point; clipper->stage.line = clip_first_line; clipper->stage.tri = clip_first_tri; diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index 053be5f050..0a70483858 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -129,6 +129,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) goto fail; cull->stage.draw = draw; + cull->stage.name = "cull"; cull->stage.next = NULL; cull->stage.point = draw_pipe_passthrough_point; cull->stage.line = draw_pipe_passthrough_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 43d1fecc4d..34afb1a0b6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -261,6 +261,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) goto fail; flatshade->stage.draw = draw; + flatshade->stage.name = "flatshade"; flatshade->stage.next = NULL; flatshade->stage.point = draw_pipe_passthrough_point; flatshade->stage.line = flatshade_first_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index 62c31532d0..40798a5d6e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -166,6 +166,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) draw_alloc_temp_verts( &offset->stage, 3 ); offset->stage.draw = draw; + offset->stage.name = "offset"; offset->stage.next = NULL; offset->stage.point = draw_pipe_passthrough_point; offset->stage.line = draw_pipe_passthrough_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 04e59152c5..30a6d2919d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -358,6 +358,7 @@ generate_pstip_fs(struct pstip_stage *pstip) pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); + FREE((void *)pstip_fs.tokens); return TRUE; } @@ -586,6 +587,7 @@ draw_pstip_stage(struct draw_context *draw) draw_alloc_temp_verts( &pstip->stage, 8 ); pstip->stage.draw = draw; + pstip->stage.name = "pstip"; pstip->stage.next = NULL; pstip->stage.point = draw_pipe_passthrough_point; pstip->stage.line = draw_pipe_passthrough_line; @@ -648,6 +650,10 @@ pstip_delete_fs_state(struct pipe_context *pipe, void *fs) struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; /* pass-through */ pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs); + + if (aafs->pstip_fs) + pstip->driver_delete_fs_state(pstip->pipe, aafs->pstip_fs); + FREE(aafs); } diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index b65e2aa102..6e921bac27 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -238,6 +238,7 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) draw_alloc_temp_verts( &stipple->stage, 2 ); stipple->stage.draw = draw; + stipple->stage.name = "stipple"; stipple->stage.next = NULL; stipple->stage.point = stipple_reset_point; stipple->stage.line = stipple_first_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index c329d92339..eef0238b15 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -181,6 +181,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) goto fail; twoside->stage.draw = draw; + twoside->stage.name = "twoside"; twoside->stage.next = NULL; twoside->stage.point = draw_pipe_passthrough_point; twoside->stage.line = draw_pipe_passthrough_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index 68835fd1a5..03bb842e20 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -184,6 +184,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) goto fail; unfilled->stage.draw = draw; + unfilled->stage.name = "unfilled"; unfilled->stage.next = NULL; unfilled->stage.tmp = NULL; unfilled->stage.point = draw_pipe_passthrough_point; diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c index 03e842ce08..bea90e50d3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_validate.c +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -262,7 +262,15 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) draw->pipeline.first = next; - return next; + + if (0) { + debug_printf("draw pipeline:\n"); + for (next = draw->pipeline.first; next ; next = next->next ) + debug_printf(" %s\n", next->name); + debug_printf("\n"); + } + + return draw->pipeline.first; } static void validate_tri( struct draw_stage *stage, @@ -318,6 +326,7 @@ struct draw_stage *draw_validate_stage( struct draw_context *draw ) return NULL; stage->draw = draw; + stage->name = "validate"; stage->next = NULL; stage->point = validate_point; stage->line = validate_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 12325d30d6..a5d840b96e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -446,6 +446,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, goto fail; vbuf->stage.draw = draw; + vbuf->stage.name = "vbuf"; vbuf->stage.point = vbuf_first_point; vbuf->stage.line = vbuf_first_line; vbuf->stage.tri = vbuf_first_tri; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 184e363594..f32cbef983 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -168,6 +168,7 @@ struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) draw_alloc_temp_verts( &wide->stage, 4 ); wide->stage.draw = draw; + wide->stage.name = "wide-line"; wide->stage.next = NULL; wide->stage.point = draw_pipe_passthrough_point; wide->stage.line = wideline_line; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index e1af9e56a2..d84bab9eaa 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -253,6 +253,7 @@ static void widepoint_flush( struct draw_stage *stage, unsigned flags ) { stage->point = widepoint_first_point; stage->next->flush( stage->next, flags ); + stage->draw->extra_vp_outputs.slot = 0; } @@ -279,6 +280,7 @@ struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) goto fail; wide->stage.draw = draw; + wide->stage.name = "wide-point"; wide->stage.next = NULL; wide->stage.point = widepoint_first_point; wide->stage.line = draw_pipe_passthrough_line; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 6f3e1e0289..8ef0ea8011 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -187,6 +187,7 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); struct pt_fetch; void draw_pt_fetch_prepare( struct pt_fetch *fetch, + unsigned vertex_input_count, unsigned vertex_size ); void draw_pt_fetch_run( struct pt_fetch *fetch, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 058caf7dcc..65c3a34c34 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "util/u_memory.h" +#include "util/u_math.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" @@ -56,9 +57,11 @@ struct pt_fetch { * */ void draw_pt_fetch_prepare( struct pt_fetch *fetch, + unsigned vs_input_count, unsigned vertex_size ) { struct draw_context *draw = fetch->draw; + unsigned nr_inputs; unsigned i, nr = 0; unsigned dst_offset = 0; struct translate_key key; @@ -89,8 +92,11 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, dst_offset += 4 * sizeof(float); } + assert( draw->pt.nr_vertex_elements >= vs_input_count ); - for (i = 0; i < draw->pt.nr_vertex_elements; i++) { + nr_inputs = MIN2( vs_input_count, draw->pt.nr_vertex_elements ); + + for (i = 0; i < nr_inputs; i++) { key.element[nr].input_format = draw->pt.vertex_element[i].src_format; key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index; key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset; @@ -114,7 +120,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_cache_find(fetch->cache, &key); { - static struct vertex_header vh = { 0, 1, 0, 0xffff }; + static struct vertex_header vh = { 0, 1, 0, UNDEFINED_VERTEX_ID, { .0f, .0f, .0f, .0f } }; fetch->translate->set_buffer(fetch->translate, draw->pt.nr_vertex_buffers, &vh, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 6b7d02a19b..e7fe6b3b76 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -130,6 +130,10 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, unsigned output_format; switch (vinfo->attrib[i].emit) { + case EMIT_4UB: + output_format = PIPE_FORMAT_R8G8B8A8_UNORM; + emit_sz = 4 * sizeof(unsigned char); + break; case EMIT_4F: output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit_sz = 4 * sizeof(float); @@ -153,6 +157,8 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); break; + case EMIT_OMIT: + continue; default: assert(0); output_format = PIPE_FORMAT_NONE; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 11ac90fc56..df6c265b7e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -77,8 +77,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, draw_pt_fetch_prepare( fpme->fetch, + vs->info.num_inputs, fpme->vertex_size ); - /* XXX: it's not really gl rasterization rules we care about here, * but gl vs dx9 clip spaces. */ diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index 55a8e6521d..010c7a18a7 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -9,7 +9,7 @@ static void FUNC(struct draw_pt_front_end *frontend, unsigned count) { struct varray_frontend *varray = (struct varray_frontend *)frontend; - unsigned start = (unsigned)elts; + unsigned start = (unsigned) ((char *) elts - (char *) NULL); unsigned j; unsigned first, incr; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index b358bd2df4..3240e3745d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -74,7 +74,7 @@ void PIPE_CDECL aos_do_lit( struct aos_machine *machine, { result[0] = 1.0F; result[1] = in[0]; - result[2] = 1.0; + result[2] = 0.0F; result[3] = 1.0F; } else @@ -108,7 +108,7 @@ static void PIPE_CDECL do_lit_lut( struct aos_machine *machine, { result[0] = 1.0F; result[1] = in[0]; - result[2] = 1.0; + result[2] = 0.0F; result[3] = 1.0F; return; } diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index dbbc33fffa..f2368dde5c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -114,6 +114,12 @@ vs_exec_run_linear( struct draw_vertex_shader *shader, #endif for (slot = 0; slot < shader->info.num_inputs; slot++) { +#if 0 + assert(!util_is_inf_or_nan(input[slot][0])); + assert(!util_is_inf_or_nan(input[slot][1])); + assert(!util_is_inf_or_nan(input[slot][2])); + assert(!util_is_inf_or_nan(input[slot][3])); +#endif machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 925e948763..2d2af3085e 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -24,6 +24,8 @@ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ + +#include <cstdio> #include "instructionssoa.h" #include "storagesoa.h" diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.c b/src/gallium/auxiliary/indices/u_unfilled_gen.c index fd082ebbb3..93897c98de 100644 --- a/src/gallium/auxiliary/indices/u_unfilled_gen.c +++ b/src/gallium/auxiliary/indices/u_unfilled_gen.c @@ -71,13 +71,10 @@ static void generate_tris_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (ushort)(i); (out+j)[1] = (ushort)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)(i+1); (out+j+2)[1] = (ushort)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (ushort)(i+2); (out+j+4)[1] = (ushort)(i); } @@ -90,13 +87,10 @@ static void generate_tristrip_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (ushort)(i); (out+j)[1] = (ushort)(i+1/*+(i&1)*/); - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (ushort)(i+1/*+(i&1)*/); (out+j+2)[1] = (ushort)(i+2/*-(i&1)*/); - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (ushort)(i+2/*-(i&1)*/); (out+j+4)[1] = (ushort)(i); } @@ -109,13 +103,10 @@ static void generate_trifan_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)(0); (out+j)[1] = (ushort)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)(i+1); (out+j+2)[1] = (ushort)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)(i+2); (out+j+4)[1] = (ushort)(0); } @@ -128,16 +119,12 @@ static void generate_quads_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (ushort)(i+0); (out+j)[1] = (ushort)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)(i+1); (out+j+2)[1] = (ushort)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (ushort)(i+2); (out+j+4)[1] = (ushort)(i+3); - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (ushort)(i+3); (out+j+6)[1] = (ushort)(i+0); } @@ -150,16 +137,12 @@ static void generate_quadstrip_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (ushort)(i+2); (out+j)[1] = (ushort)(i+0); - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (ushort)(i+0); (out+j+2)[1] = (ushort)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (ushort)(i+1); (out+j+4)[1] = (ushort)(i+3); - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (ushort)(i+3); (out+j+6)[1] = (ushort)(i+2); } @@ -172,13 +155,10 @@ static void generate_polygon_ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)(0); (out+j)[1] = (ushort)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)(i+1); (out+j+2)[1] = (ushort)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)(i+2); (out+j+4)[1] = (ushort)(0); } @@ -191,13 +171,10 @@ static void generate_tris_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (uint)(i); (out+j)[1] = (uint)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)(i+1); (out+j+2)[1] = (uint)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (uint)(i+2); (out+j+4)[1] = (uint)(i); } @@ -210,13 +187,10 @@ static void generate_tristrip_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (uint)(i); (out+j)[1] = (uint)(i+1/*+(i&1)*/); - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (uint)(i+1/*+(i&1)*/); (out+j+2)[1] = (uint)(i+2/*-(i&1)*/); - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (uint)(i+2/*-(i&1)*/); (out+j+4)[1] = (uint)(i); } @@ -229,13 +203,10 @@ static void generate_trifan_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)(0); (out+j)[1] = (uint)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)(i+1); (out+j+2)[1] = (uint)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)(i+2); (out+j+4)[1] = (uint)(0); } @@ -248,16 +219,12 @@ static void generate_quads_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (uint)(i+0); (out+j)[1] = (uint)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)(i+1); (out+j+2)[1] = (uint)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (uint)(i+2); (out+j+4)[1] = (uint)(i+3); - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (uint)(i+3); (out+j+6)[1] = (uint)(i+0); } @@ -270,16 +237,12 @@ static void generate_quadstrip_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (uint)(i+2); (out+j)[1] = (uint)(i+0); - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (uint)(i+0); (out+j+2)[1] = (uint)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (uint)(i+1); (out+j+4)[1] = (uint)(i+3); - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (uint)(i+3); (out+j+6)[1] = (uint)(i+2); } @@ -292,13 +255,10 @@ static void generate_polygon_uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)(0); (out+j)[1] = (uint)(i+1); - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)(i+1); (out+j+2)[1] = (uint)(i+2); - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)(i+2); (out+j+4)[1] = (uint)(0); } @@ -313,13 +273,10 @@ static void translate_tris_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i]; } @@ -334,13 +291,10 @@ static void translate_tristrip_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (ushort)in[i]; } @@ -355,13 +309,10 @@ static void translate_trifan_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -376,16 +327,12 @@ static void translate_quads_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (ushort)in[i+0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+0]; } @@ -400,16 +347,12 @@ static void translate_quadstrip_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (ushort)in[i+2]; (out+j)[1] = (ushort)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (ushort)in[i+0]; (out+j+2)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (ushort)in[i+1]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+2]; } @@ -424,13 +367,10 @@ static void translate_polygon_ubyte2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -445,13 +385,10 @@ static void translate_tris_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i]; } @@ -466,13 +403,10 @@ static void translate_tristrip_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (uint)in[i]; } @@ -487,13 +421,10 @@ static void translate_trifan_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } @@ -508,16 +439,12 @@ static void translate_quads_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (uint)in[i+0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+0]; } @@ -532,16 +459,12 @@ static void translate_quadstrip_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (uint)in[i+2]; (out+j)[1] = (uint)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (uint)in[i+0]; (out+j+2)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (uint)in[i+1]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+2]; } @@ -556,13 +479,10 @@ static void translate_polygon_ubyte2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } @@ -577,13 +497,10 @@ static void translate_tris_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i]; } @@ -598,13 +515,10 @@ static void translate_tristrip_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (ushort)in[i]; } @@ -619,13 +533,10 @@ static void translate_trifan_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -640,16 +551,12 @@ static void translate_quads_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (ushort)in[i+0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+0]; } @@ -664,16 +571,12 @@ static void translate_quadstrip_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (ushort)in[i+2]; (out+j)[1] = (ushort)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (ushort)in[i+0]; (out+j+2)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (ushort)in[i+1]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+2]; } @@ -688,13 +591,10 @@ static void translate_polygon_ushort2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -709,13 +609,10 @@ static void translate_tris_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i]; } @@ -730,13 +627,10 @@ static void translate_tristrip_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (uint)in[i]; } @@ -751,13 +645,10 @@ static void translate_trifan_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } @@ -772,16 +663,12 @@ static void translate_quads_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (uint)in[i+0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+0]; } @@ -796,16 +683,12 @@ static void translate_quadstrip_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (uint)in[i+2]; (out+j)[1] = (uint)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (uint)in[i+0]; (out+j+2)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (uint)in[i+1]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+2]; } @@ -820,13 +703,10 @@ static void translate_polygon_ushort2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } @@ -841,13 +721,10 @@ static void translate_tris_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i]; } @@ -862,13 +739,10 @@ static void translate_tristrip_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (ushort)in[i]; (out+j)[1] = (ushort)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (ushort)in[i]; } @@ -883,13 +757,10 @@ static void translate_trifan_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -904,16 +775,12 @@ static void translate_quads_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (ushort)in[i+0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+0]; } @@ -928,16 +795,12 @@ static void translate_quadstrip_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (ushort)in[i+2]; (out+j)[1] = (ushort)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (ushort)in[i+0]; (out+j+2)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (ushort)in[i+1]; (out+j+4)[1] = (ushort)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (ushort)in[i+3]; (out+j+6)[1] = (ushort)in[i+2]; } @@ -952,13 +815,10 @@ static void translate_polygon_uint2ushort( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (ushort)in[0]; (out+j)[1] = (ushort)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (ushort)in[i+1]; (out+j+2)[1] = (ushort)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (ushort)in[i+2]; (out+j+4)[1] = (ushort)in[0]; } @@ -973,13 +833,10 @@ static void translate_tris_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i+=3) { - debug_printf(" line %d %d\n", (int)i, (int)i+1); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i]; } @@ -994,13 +851,10 @@ static void translate_tristrip_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)i, (int)i+1/*+(i&1)*/); (out+j)[0] = (uint)in[i]; (out+j)[1] = (uint)in[i+1/*+(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/); (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/]; (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/]; - debug_printf(" line %d %d\n", (int)i+2/*-(i&1)*/, (int)i); (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/]; (out+j+4)[1] = (uint)in[i]; } @@ -1015,13 +869,10 @@ static void translate_trifan_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } @@ -1036,16 +887,12 @@ static void translate_quads_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=4) { - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j)[0] = (uint)in[i+0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)i+3); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+0); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+0]; } @@ -1060,16 +907,12 @@ static void translate_quadstrip_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=8, i+=2) { - debug_printf(" line %d %d\n", (int)i+2, (int)i+0); (out+j)[0] = (uint)in[i+2]; (out+j)[1] = (uint)in[i+0]; - debug_printf(" line %d %d\n", (int)i+0, (int)i+1); (out+j+2)[0] = (uint)in[i+0]; (out+j+2)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+3); (out+j+4)[0] = (uint)in[i+1]; (out+j+4)[1] = (uint)in[i+3]; - debug_printf(" line %d %d\n", (int)i+3, (int)i+2); (out+j+6)[0] = (uint)in[i+3]; (out+j+6)[1] = (uint)in[i+2]; } @@ -1084,13 +927,10 @@ static void translate_polygon_uint2uint( unsigned i, j; (void)j; for (j = i = 0; j < nr; j+=6, i++) { - debug_printf(" line %d %d\n", (int)0, (int)i+1); (out+j)[0] = (uint)in[0]; (out+j)[1] = (uint)in[i+1]; - debug_printf(" line %d %d\n", (int)i+1, (int)i+2); (out+j+2)[0] = (uint)in[i+1]; (out+j+2)[1] = (uint)in[i+2]; - debug_printf(" line %d %d\n", (int)i+2, (int)0); (out+j+4)[0] = (uint)in[i+2]; (out+j+4)[1] = (uint)in[0]; } diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.py b/src/gallium/auxiliary/indices/u_unfilled_gen.py index d0344fe313..36896ce605 100644 --- a/src/gallium/auxiliary/indices/u_unfilled_gen.py +++ b/src/gallium/auxiliary/indices/u_unfilled_gen.py @@ -99,7 +99,6 @@ def vert( intype, outtype, v0 ): return '(' + outtype + ')in[' + v0 + ']' def line( intype, outtype, ptr, v0, v1 ): - print ' debug_printf(" line %d %d\\n", (int)' + v0 + ', (int)' + v1 + ');' print ' (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';' print ' (' + ptr + ')[1] = ' + vert( intype, outtype, v1 ) + ';' diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 2cd0b8a8cd..4698efa69c 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -365,21 +365,22 @@ fenced_buffer_validate(struct pb_buffer *buf, if(fenced_buf->vl && fenced_buf->vl != vl) return PIPE_ERROR_RETRY; +#if 0 /* Do not validate if buffer is still mapped */ if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { /* TODO: wait for the thread that mapped the buffer to unmap it */ return PIPE_ERROR_RETRY; } + /* Final sanity checking */ + assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE)); + assert(!fenced_buf->mapcount); +#endif if(fenced_buf->vl == vl && (fenced_buf->validation_flags & flags) == flags) { /* Nothing to do -- buffer already validated */ return PIPE_OK; } - - /* Final sanity checking */ - assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE)); - assert(!fenced_buf->mapcount); ret = pb_validate(fenced_buf->buffer, vl, flags); if (ret != PIPE_OK) @@ -530,16 +531,17 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) pipe_mutex_lock(fenced_list->mutex); - debug_printf("%10s %7s %10s %s\n", - "buffer", "reference.count", "fence", "signalled"); + debug_printf("%10s %7s %7s %10s %s\n", + "buffer", "size", "refcount", "fence", "signalled"); curr = fenced_list->unfenced.next; next = curr->next; while(curr != &fenced_list->unfenced) { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); assert(!fenced_buf->fence); - debug_printf("%10p %7u\n", + debug_printf("%10p %7u %7u\n", fenced_buf, + fenced_buf->base.base.size, fenced_buf->base.base.reference.count); curr = next; next = curr->next; @@ -551,8 +553,9 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) int signaled; fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); - debug_printf("%10p %7u %10p %s\n", + debug_printf("%10p %7u %7u %10p %s\n", fenced_buf, + fenced_buf->base.base.size, fenced_buf->base.base.reference.count, fenced_buf->fence, signaled == 0 ? "y" : "n"); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index f1a05be46e..cedf745bda 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -40,6 +40,7 @@ #include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_time.h" +#include "util/u_debug_stack.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -48,6 +49,10 @@ #ifdef DEBUG +#define PB_DEBUG_CREATE_BACKTRACE 8 +#define PB_DEBUG_MAP_BACKTRACE 8 + + /** * Convenience macro (type safe). */ @@ -69,6 +74,14 @@ struct pb_debug_buffer size_t underflow_size; size_t overflow_size; + + struct debug_stack_frame create_backtrace[PB_DEBUG_CREATE_BACKTRACE]; + + pipe_mutex mutex; + unsigned map_count; + struct debug_stack_frame map_backtrace[PB_DEBUG_MAP_BACKTRACE]; + + struct list_head head; }; @@ -80,6 +93,9 @@ struct pb_debug_manager size_t underflow_size; size_t overflow_size; + + pipe_mutex mutex; + struct list_head list; }; @@ -189,6 +205,9 @@ pb_debug_buffer_check(struct pb_debug_buffer *buf) max_ofs == buf->overflow_size - 1 ? "+" : ""); } + if(underflow || overflow) + debug_backtrace_dump(buf->create_backtrace, PB_DEBUG_CREATE_BACKTRACE); + debug_assert(!underflow && !overflow); /* re-fill if not aborted */ @@ -207,11 +226,18 @@ static void pb_debug_buffer_destroy(struct pb_buffer *_buf) { struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + struct pb_debug_manager *mgr = buf->mgr; assert(!pipe_is_referenced(&buf->base.base.reference)); pb_debug_buffer_check(buf); + pipe_mutex_lock(mgr->mutex); + LIST_DEL(&buf->head); + pipe_mutex_unlock(mgr->mutex); + + pipe_mutex_destroy(buf->mutex); + pb_reference(&buf->buffer, NULL); FREE(buf); } @@ -230,6 +256,13 @@ pb_debug_buffer_map(struct pb_buffer *_buf, if(!map) return NULL; + if(map) { + pipe_mutex_lock(buf->mutex); + ++buf->map_count; + debug_backtrace_capture(buf->map_backtrace, 1, PB_DEBUG_MAP_BACKTRACE); + pipe_mutex_unlock(buf->mutex); + } + return (uint8_t *)map + buf->underflow_size; } @@ -238,6 +271,13 @@ static void pb_debug_buffer_unmap(struct pb_buffer *_buf) { struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + + pipe_mutex_lock(buf->mutex); + assert(buf->map_count); + if(buf->map_count) + --buf->map_count; + pipe_mutex_unlock(buf->mutex); + pb_unmap(buf->buffer); pb_debug_buffer_check(buf); @@ -262,6 +302,14 @@ pb_debug_buffer_validate(struct pb_buffer *_buf, { struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + pipe_mutex_lock(buf->mutex); + if(buf->map_count) { + debug_printf("%s: attempting to validate a mapped buffer\n", __FUNCTION__); + debug_printf("last map backtrace is\n"); + debug_backtrace_dump(buf->map_backtrace, PB_DEBUG_MAP_BACKTRACE); + } + pipe_mutex_unlock(buf->mutex); + pb_debug_buffer_check(buf); return pb_validate(buf->buffer, vl, flags); @@ -288,6 +336,31 @@ pb_debug_buffer_vtbl = { }; +static void +pb_debug_manager_dump(struct pb_debug_manager *mgr) +{ + struct list_head *curr, *next; + struct pb_debug_buffer *buf; + + pipe_mutex_lock(mgr->mutex); + + curr = mgr->list.next; + next = curr->next; + while(curr != &mgr->list) { + buf = LIST_ENTRY(struct pb_debug_buffer, curr, head); + + debug_printf("buffer = %p\n", buf); + debug_printf(" .size = %p\n", buf->base.base.size); + debug_backtrace_dump(buf->create_backtrace, PB_DEBUG_CREATE_BACKTRACE); + + curr = next; + next = curr->next; + } + + pipe_mutex_unlock(mgr->mutex); +} + + static struct pb_buffer * pb_debug_manager_create_buffer(struct pb_manager *_mgr, size_t size, @@ -312,6 +385,13 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr, &real_desc); if(!buf->buffer) { FREE(buf); +#if 0 + pipe_mutex_lock(mgr->mutex); + debug_printf("%s: failed to create buffer\n", __FUNCTION__); + if(!LIST_IS_EMPTY(&mgr->list)) + pb_debug_manager_dump(mgr); + pipe_mutex_unlock(mgr->mutex); +#endif return NULL; } @@ -331,8 +411,16 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr, buf->underflow_size = mgr->underflow_size; buf->overflow_size = buf->buffer->base.size - buf->underflow_size - size; + debug_backtrace_capture(buf->create_backtrace, 1, PB_DEBUG_CREATE_BACKTRACE); + pb_debug_buffer_fill(buf); + pipe_mutex_init(buf->mutex); + + pipe_mutex_lock(mgr->mutex); + LIST_ADDTAIL(&buf->head, &mgr->list); + pipe_mutex_unlock(mgr->mutex); + return &buf->base; } @@ -351,6 +439,15 @@ static void pb_debug_manager_destroy(struct pb_manager *_mgr) { struct pb_debug_manager *mgr = pb_debug_manager(_mgr); + + pipe_mutex_lock(mgr->mutex); + if(!LIST_IS_EMPTY(&mgr->list)) { + debug_printf("%s: unfreed buffers\n", __FUNCTION__); + pb_debug_manager_dump(mgr); + } + pipe_mutex_unlock(mgr->mutex); + + pipe_mutex_destroy(mgr->mutex); mgr->provider->destroy(mgr->provider); FREE(mgr); } @@ -375,7 +472,10 @@ pb_debug_manager_create(struct pb_manager *provider, mgr->provider = provider; mgr->underflow_size = underflow_size; mgr->overflow_size = overflow_size; - + + pipe_mutex_init(mgr->mutex); + LIST_INITHEAD(&mgr->list); + return &mgr->base; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 5a342fbf3b..d8decba49d 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -200,8 +200,8 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, mm_buf->block = u_mmAllocMem(mm->heap, size, mm->align2, 0); if(!mm_buf->block) { - debug_printf("warning: heap full\n"); #if 0 + debug_printf("warning: heap full\n"); mmDumpMemInfo(mm->heap); #endif FREE(mm_buf); diff --git a/src/gallium/auxiliary/rbug/Makefile b/src/gallium/auxiliary/rbug/Makefile new file mode 100644 index 0000000000..cd12e8468f --- /dev/null +++ b/src/gallium/auxiliary/rbug/Makefile @@ -0,0 +1,14 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = rbug + +C_SOURCES = \ + rbug_connection.c \ + rbug_core.c \ + rbug_texture.c \ + rbug_context.c \ + rbug_shader.c \ + rbug_demarshal.c + +include ../../Makefile.template diff --git a/src/gallium/auxiliary/rbug/README b/src/gallium/auxiliary/rbug/README new file mode 100644 index 0000000000..33d76371de --- /dev/null +++ b/src/gallium/auxiliary/rbug/README @@ -0,0 +1,21 @@ + GALLIUM REMOTE DEBUGGING COMMON CODE + += About = + +This directory contains the common code for the Gallium 3D remote debugging +driver and clients. The code is two parts the connection managment code and +the (de)marsheller. + +The code currently uses tcp and ip4v for connections. + +Information about driver integration can be found in: + +src/gallium/drivers/trace/README + +for information about applications look in: + +progs/rbug/README + + +-- +Jakob Bornecrantz <jakob@vmware.com> diff --git a/src/gallium/auxiliary/rbug/SConscript b/src/gallium/auxiliary/rbug/SConscript new file mode 100644 index 0000000000..4a9afb45d3 --- /dev/null +++ b/src/gallium/auxiliary/rbug/SConscript @@ -0,0 +1,14 @@ +Import('*') + +rbug = env.ConvenienceLibrary( + target = 'rbug', + source = [ + 'rbug_core.c', + 'rbug_shader.c', + 'rbug_context.c', + 'rbug_texture.c', + 'rbug_demarshal.c', + 'rbug_connection.c', + ]) + +auxiliaries.insert(0, rbug) diff --git a/src/gallium/auxiliary/rbug/rbug.h b/src/gallium/auxiliary/rbug/rbug.h new file mode 100644 index 0000000000..259bfc6c79 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug.h @@ -0,0 +1,33 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Include all for users the remote debugger protocol code. + */ + +#include "rbug/rbug_core.h" +#include "rbug/rbug_shader.h" +#include "rbug/rbug_context.h" +#include "rbug/rbug_texture.h" +#include "rbug/rbug_connection.h" diff --git a/src/gallium/auxiliary/rbug/rbug_connection.c b/src/gallium/auxiliary/rbug/rbug_connection.c new file mode 100644 index 0000000000..52acb700af --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_connection.c @@ -0,0 +1,167 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "rbug/rbug.h" +#include "rbug/rbug_internal.h" + +#include "util/u_network.h" + +struct rbug_connection +{ + int socket; + uint32_t send_serial; + uint32_t recv_serial; + enum rbug_opcode opcode; +}; + +/** + * Create a rbug connection from a socket created with u_socket. + * + * Result: + * A new allocated connection using socket as communication path + */ +struct rbug_connection * +rbug_from_socket(int socket) +{ + struct rbug_connection *c = CALLOC_STRUCT(rbug_connection); + c->socket = socket; + return c; +} + +/** + * Free a connection, also closes socket. + */ +void +rbug_disconnect(struct rbug_connection *c) +{ + u_socket_close(c->socket); + FREE(c); +} + +/** + * Waits for a message to be fully received. + * Also returns the serial for the message, serial is not touched for replys. + * + * Result: + * demarshaled message on success, NULL on connection error + */ +struct rbug_header * +rbug_get_message(struct rbug_connection *c, uint32_t *serial) +{ + struct rbug_proto_header header; + struct rbug_header *out; + struct rbug_proto_header *data; + size_t length = 0; + size_t read = 0; + int ret; + + + ret = u_socket_peek(c->socket, &header, sizeof(header)); + if (ret <= 0) { + return NULL; + } + + length = (size_t)header.length * 4; + data = MALLOC(length); + if (!data) { + return NULL; + } + + do { + uint8_t *ptr = ((uint8_t*)data) + read; + ret = u_socket_recv(c->socket, ptr, length - read); + + if (ret <= 0) { + FREE(data); + return NULL; + } + + read += ret; + } while(read < length); + + out = rbug_demarshal(data); + if (!out) + FREE(data); + else if (serial) + *serial = c->recv_serial++; + else + c->recv_serial++; + + return out; +} + +/** + * Frees a message and associated data. + */ +void +rbug_free_header(struct rbug_header *header) +{ + if (!header) + return; + + FREE(header->__message); + FREE(header); +} + +/** + * Internal function used by rbug_send_* functions. + * + * Start sending a message. + */ +int +rbug_connection_send_start(struct rbug_connection *c, enum rbug_opcode opcode, uint32_t length) +{ + c->opcode = opcode; + return 0; +} + +/** + * Internal function used by rbug_send_* functions. + * + * Write data to the socket. + */ +int +rbug_connection_write(struct rbug_connection *c, void *to, uint32_t size) +{ + int ret = u_socket_send(c->socket, to, size); + return ret; +} + +/** + * Internal function used by rbug_send_* functions. + * + * Finish writeing data to the socket. + * Ups the send_serial and sets the serial argument if supplied. + */ +int rbug_connection_send_finish(struct rbug_connection *c, uint32_t *serial) +{ + if (c->opcode < 0) + return 0; + else if (serial) + *serial = c->send_serial++; + else + c->send_serial++; + + return 0; +} diff --git a/src/gallium/auxiliary/rbug/rbug_connection.h b/src/gallium/auxiliary/rbug/rbug_connection.h new file mode 100644 index 0000000000..1f2c9ff347 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_connection.h @@ -0,0 +1,45 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file contains the function defentions for connection see c file for + * more comments covering function use. + */ + +#ifndef _RBUG_CONNECTION_H_ +#define _RBUG_CONNECTION_H_ + +#include "rbug/rbug_proto.h" + +struct rbug_connection * rbug_from_socket(int socket); + +void rbug_disconnect(struct rbug_connection *c); + +struct rbug_header * rbug_get_message(struct rbug_connection *c, uint32_t *serial); + +void rbug_free_header(struct rbug_header *header); + +struct rbug_header * rbug_demarshal(struct rbug_proto_header *header); + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_context.c b/src/gallium/auxiliary/rbug/rbug_context.c new file mode 100644 index 0000000000..1832425658 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_context.c @@ -0,0 +1,759 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds the function implementation for one of the rbug extensions. + * Prototypes and declerations of functions and structs is in the same folder + * in the header file matching this file's name. + * + * The functions starting rbug_send_* encodes a call to the write format and + * sends that to the supplied connection, while functions starting with + * rbug_demarshal_* demarshal data in the wire protocol. + * + * Functions ending with _reply are replies to requests. + */ + +#include "rbug_internal.h" +#include "rbug/rbug_context.h" + +int rbug_send_context_list(struct rbug_connection *__con, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_info(struct rbug_connection *__con, + rbug_context_t context, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_draw_block(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t block, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(4); /* block */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCK)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(4, rbug_block_t, block); /* block */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCK, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_draw_step(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t step, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(4); /* step */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_STEP)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(4, rbug_block_t, step); /* step */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_STEP, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_draw_unblock(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t unblock, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(4); /* unblock */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(4, rbug_block_t, unblock); /* unblock */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_UNBLOCK, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_draw_rule(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t vertex, + rbug_shader_t fragment, + rbug_texture_t texture, + rbug_texture_t surface, + rbug_block_t block, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(8); /* vertex */ + LEN(8); /* fragment */ + LEN(8); /* texture */ + LEN(8); /* surface */ + LEN(4); /* block */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_RULE)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(8, rbug_shader_t, vertex); /* vertex */ + WRITE(8, rbug_shader_t, fragment); /* fragment */ + WRITE(8, rbug_texture_t, texture); /* texture */ + WRITE(8, rbug_texture_t, surface); /* surface */ + WRITE(4, rbug_block_t, block); /* block */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_RULE, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_flush(struct rbug_connection *__con, + rbug_context_t context, + int32_t flags, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(4); /* flags */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_FLUSH)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(4, int32_t, flags); /* flags */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_FLUSH, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_context_t *contexts, + uint32_t contexts_len, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN_ARRAY(8, contexts); /* contexts */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE_ARRAY(8, rbug_context_t, contexts); /* contexts */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_info_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_shader_t vertex, + rbug_shader_t fragment, + rbug_texture_t *texs, + uint32_t texs_len, + rbug_texture_t *cbufs, + uint32_t cbufs_len, + rbug_texture_t zsbuf, + rbug_block_t blocker, + rbug_block_t blocked, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN(8); /* vertex */ + LEN(8); /* fragment */ + LEN_ARRAY(8, texs); /* texs */ + LEN_ARRAY(8, cbufs); /* cbufs */ + LEN(8); /* zsbuf */ + LEN(4); /* blocker */ + LEN(4); /* blocked */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE(8, rbug_shader_t, vertex); /* vertex */ + WRITE(8, rbug_shader_t, fragment); /* fragment */ + WRITE_ARRAY(8, rbug_texture_t, texs); /* texs */ + WRITE_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */ + WRITE(8, rbug_texture_t, zsbuf); /* zsbuf */ + WRITE(4, rbug_block_t, blocker); /* blocker */ + WRITE(4, rbug_block_t, blocked); /* blocked */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_context_draw_blocked(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t block, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(4); /* block */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCKED)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(4, rbug_block_t, block); /* block */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCKED, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_list *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + + return ret; +} + +struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_info *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + + return ret; +} + +struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_draw_block *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCK) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(4, rbug_block_t, block); /* block */ + + return ret; +} + +struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_draw_step *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_STEP) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(4, rbug_block_t, step); /* step */ + + return ret; +} + +struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_draw_unblock *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(4, rbug_block_t, unblock); /* unblock */ + + return ret; +} + +struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_draw_rule *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_RULE) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(8, rbug_shader_t, vertex); /* vertex */ + READ(8, rbug_shader_t, fragment); /* fragment */ + READ(8, rbug_texture_t, texture); /* texture */ + READ(8, rbug_texture_t, surface); /* surface */ + READ(4, rbug_block_t, block); /* block */ + + return ret; +} + +struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_flush *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_FLUSH) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(4, int32_t, flags); /* flags */ + + return ret; +} + +struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_list_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ_ARRAY(8, rbug_context_t, contexts); /* contexts */ + + return ret; +} + +struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_info_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ(8, rbug_shader_t, vertex); /* vertex */ + READ(8, rbug_shader_t, fragment); /* fragment */ + READ_ARRAY(8, rbug_texture_t, texs); /* texs */ + READ_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */ + READ(8, rbug_texture_t, zsbuf); /* zsbuf */ + READ(4, rbug_block_t, blocker); /* blocker */ + READ(4, rbug_block_t, blocked); /* blocked */ + + return ret; +} + +struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_context_draw_blocked *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCKED) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(4, rbug_block_t, block); /* block */ + + return ret; +} diff --git a/src/gallium/auxiliary/rbug/rbug_context.h b/src/gallium/auxiliary/rbug/rbug_context.h new file mode 100644 index 0000000000..da61c2365b --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_context.h @@ -0,0 +1,212 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds structs decelerations and function prototypes for one of + * the rbug extensions. Implementation of the functions is in the same folder + * in the c file matching this file's name. + * + * The structs what is returned from the demarshal functions. The functions + * starting rbug_send_* encodes a call to the write format and sends that to + * the supplied connection, while functions starting with rbug_demarshal_* + * demarshal data from the wire protocol. + * + * Structs and functions ending with _reply are replies to requests. + */ + +#ifndef _RBUG_PROTO_CONTEXT_H_ +#define _RBUG_PROTO_CONTEXT_H_ + +#include "rbug/rbug_proto.h" +#include "rbug/rbug_core.h" + +typedef enum +{ + RBUG_BLOCK_BEFORE = 1, + RBUG_BLOCK_AFTER = 2, + RBUG_BLOCK_RULE = 4, + RBUG_BLOCK_MASK = 7, +} rbug_block_t; + +struct rbug_proto_context_list +{ + struct rbug_header header; +}; + +struct rbug_proto_context_info +{ + struct rbug_header header; + rbug_context_t context; +}; + +struct rbug_proto_context_draw_block +{ + struct rbug_header header; + rbug_context_t context; + rbug_block_t block; +}; + +struct rbug_proto_context_draw_step +{ + struct rbug_header header; + rbug_context_t context; + rbug_block_t step; +}; + +struct rbug_proto_context_draw_unblock +{ + struct rbug_header header; + rbug_context_t context; + rbug_block_t unblock; +}; + +struct rbug_proto_context_draw_rule +{ + struct rbug_header header; + rbug_context_t context; + rbug_shader_t vertex; + rbug_shader_t fragment; + rbug_texture_t texture; + rbug_texture_t surface; + rbug_block_t block; +}; + +struct rbug_proto_context_flush +{ + struct rbug_header header; + rbug_context_t context; + int32_t flags; +}; + +struct rbug_proto_context_list_reply +{ + struct rbug_header header; + uint32_t serial; + rbug_context_t *contexts; + uint32_t contexts_len; +}; + +struct rbug_proto_context_info_reply +{ + struct rbug_header header; + uint32_t serial; + rbug_shader_t vertex; + rbug_shader_t fragment; + rbug_texture_t *texs; + uint32_t texs_len; + rbug_texture_t *cbufs; + uint32_t cbufs_len; + rbug_texture_t zsbuf; + rbug_block_t blocker; + rbug_block_t blocked; +}; + +struct rbug_proto_context_draw_blocked +{ + struct rbug_header header; + rbug_context_t context; + rbug_block_t block; +}; + +int rbug_send_context_list(struct rbug_connection *__con, + uint32_t *__serial); + +int rbug_send_context_info(struct rbug_connection *__con, + rbug_context_t context, + uint32_t *__serial); + +int rbug_send_context_draw_block(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t block, + uint32_t *__serial); + +int rbug_send_context_draw_step(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t step, + uint32_t *__serial); + +int rbug_send_context_draw_unblock(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t unblock, + uint32_t *__serial); + +int rbug_send_context_draw_rule(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t vertex, + rbug_shader_t fragment, + rbug_texture_t texture, + rbug_texture_t surface, + rbug_block_t block, + uint32_t *__serial); + +int rbug_send_context_flush(struct rbug_connection *__con, + rbug_context_t context, + int32_t flags, + uint32_t *__serial); + +int rbug_send_context_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_context_t *contexts, + uint32_t contexts_len, + uint32_t *__serial); + +int rbug_send_context_info_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_shader_t vertex, + rbug_shader_t fragment, + rbug_texture_t *texs, + uint32_t texs_len, + rbug_texture_t *cbufs, + uint32_t cbufs_len, + rbug_texture_t zsbuf, + rbug_block_t blocker, + rbug_block_t blocked, + uint32_t *__serial); + +int rbug_send_context_draw_blocked(struct rbug_connection *__con, + rbug_context_t context, + rbug_block_t block, + uint32_t *__serial); + +struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header); + +struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header); + +struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header); + +struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header); + +struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header); + +struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header); + +struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header); + +struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header); + +struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header); + +struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header); + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_core.c b/src/gallium/auxiliary/rbug/rbug_core.c new file mode 100644 index 0000000000..876ae5a0ce --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_core.c @@ -0,0 +1,359 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds the function implementation for one of the rbug extensions. + * Prototypes and declerations of functions and structs is in the same folder + * in the header file matching this file's name. + * + * The functions starting rbug_send_* encodes a call to the write format and + * sends that to the supplied connection, while functions starting with + * rbug_demarshal_* demarshal data in the wire protocol. + * + * Functions ending with _reply are replies to requests. + */ + +#include "rbug_internal.h" +#include "rbug/rbug_core.h" + +int rbug_send_noop(struct rbug_connection *__con, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_NOOP)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_NOOP, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_ping(struct rbug_connection *__con, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_PING)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_PING, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_error(struct rbug_connection *__con, + uint32_t error, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* error */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, error); /* error */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_ERROR, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_ping_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_PING_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_PING_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_error_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t error, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN(4); /* error */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE(4, uint32_t, error); /* error */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_ERROR_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_noop *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_NOOP) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + + return ret; +} + +struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_ping *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_PING) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + + return ret; +} + +struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_error *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_ERROR) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, error); /* error */ + + return ret; +} + +struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_ping_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_PING_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + + return ret; +} + +struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_error_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_ERROR_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ(4, uint32_t, error); /* error */ + + return ret; +} diff --git a/src/gallium/auxiliary/rbug/rbug_core.h b/src/gallium/auxiliary/rbug/rbug_core.h new file mode 100644 index 0000000000..99a36a0163 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_core.h @@ -0,0 +1,105 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds structs decelerations and function prototypes for one of + * the rbug extensions. Implementation of the functions is in the same folder + * in the c file matching this file's name. + * + * The structs what is returned from the demarshal functions. The functions + * starting rbug_send_* encodes a call to the write format and sends that to + * the supplied connection, while functions starting with rbug_demarshal_* + * demarshal data from the wire protocol. + * + * Structs and functions ending with _reply are replies to requests. + */ + +#ifndef _RBUG_PROTO_CORE_H_ +#define _RBUG_PROTO_CORE_H_ + +#include "rbug/rbug_proto.h" + +typedef uint64_t rbug_shader_t; +typedef uint64_t rbug_context_t; +typedef uint64_t rbug_texture_t; + +struct rbug_proto_noop +{ + struct rbug_header header; +}; + +struct rbug_proto_ping +{ + struct rbug_header header; +}; + +struct rbug_proto_error +{ + struct rbug_header header; + uint32_t error; +}; + +struct rbug_proto_ping_reply +{ + struct rbug_header header; + uint32_t serial; +}; + +struct rbug_proto_error_reply +{ + struct rbug_header header; + uint32_t serial; + uint32_t error; +}; + +int rbug_send_noop(struct rbug_connection *__con, + uint32_t *__serial); + +int rbug_send_ping(struct rbug_connection *__con, + uint32_t *__serial); + +int rbug_send_error(struct rbug_connection *__con, + uint32_t error, + uint32_t *__serial); + +int rbug_send_ping_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t *__serial); + +int rbug_send_error_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t error, + uint32_t *__serial); + +struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header); + +struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header); + +struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header); + +struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header); + +struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header); + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_demarshal.c b/src/gallium/auxiliary/rbug/rbug_demarshal.c new file mode 100644 index 0000000000..47390fbcee --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_demarshal.c @@ -0,0 +1,93 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "rbug.h" + +/** + * Small function that looks at the proto_header and selects the correct + * demarshal functions and return the result. + */ +struct rbug_header * rbug_demarshal(struct rbug_proto_header *header) +{ + switch(header->opcode) { + case RBUG_OP_NOOP: + return (struct rbug_header *)rbug_demarshal_noop(header); + case RBUG_OP_PING: + return (struct rbug_header *)rbug_demarshal_ping(header); + case RBUG_OP_ERROR: + return (struct rbug_header *)rbug_demarshal_error(header); + case RBUG_OP_PING_REPLY: + return (struct rbug_header *)rbug_demarshal_ping_reply(header); + case RBUG_OP_ERROR_REPLY: + return (struct rbug_header *)rbug_demarshal_error_reply(header); + case RBUG_OP_TEXTURE_LIST: + return (struct rbug_header *)rbug_demarshal_texture_list(header); + case RBUG_OP_TEXTURE_INFO: + return (struct rbug_header *)rbug_demarshal_texture_info(header); + case RBUG_OP_TEXTURE_WRITE: + return (struct rbug_header *)rbug_demarshal_texture_write(header); + case RBUG_OP_TEXTURE_READ: + return (struct rbug_header *)rbug_demarshal_texture_read(header); + case RBUG_OP_TEXTURE_LIST_REPLY: + return (struct rbug_header *)rbug_demarshal_texture_list_reply(header); + case RBUG_OP_TEXTURE_INFO_REPLY: + return (struct rbug_header *)rbug_demarshal_texture_info_reply(header); + case RBUG_OP_TEXTURE_READ_REPLY: + return (struct rbug_header *)rbug_demarshal_texture_read_reply(header); + case RBUG_OP_CONTEXT_LIST: + return (struct rbug_header *)rbug_demarshal_context_list(header); + case RBUG_OP_CONTEXT_INFO: + return (struct rbug_header *)rbug_demarshal_context_info(header); + case RBUG_OP_CONTEXT_DRAW_BLOCK: + return (struct rbug_header *)rbug_demarshal_context_draw_block(header); + case RBUG_OP_CONTEXT_DRAW_STEP: + return (struct rbug_header *)rbug_demarshal_context_draw_step(header); + case RBUG_OP_CONTEXT_DRAW_UNBLOCK: + return (struct rbug_header *)rbug_demarshal_context_draw_unblock(header); + case RBUG_OP_CONTEXT_DRAW_RULE: + return (struct rbug_header *)rbug_demarshal_context_draw_rule(header); + case RBUG_OP_CONTEXT_FLUSH: + return (struct rbug_header *)rbug_demarshal_context_flush(header); + case RBUG_OP_CONTEXT_LIST_REPLY: + return (struct rbug_header *)rbug_demarshal_context_list_reply(header); + case RBUG_OP_CONTEXT_INFO_REPLY: + return (struct rbug_header *)rbug_demarshal_context_info_reply(header); + case RBUG_OP_CONTEXT_DRAW_BLOCKED: + return (struct rbug_header *)rbug_demarshal_context_draw_blocked(header); + case RBUG_OP_SHADER_LIST: + return (struct rbug_header *)rbug_demarshal_shader_list(header); + case RBUG_OP_SHADER_INFO: + return (struct rbug_header *)rbug_demarshal_shader_info(header); + case RBUG_OP_SHADER_DISABLE: + return (struct rbug_header *)rbug_demarshal_shader_disable(header); + case RBUG_OP_SHADER_REPLACE: + return (struct rbug_header *)rbug_demarshal_shader_replace(header); + case RBUG_OP_SHADER_LIST_REPLY: + return (struct rbug_header *)rbug_demarshal_shader_list_reply(header); + case RBUG_OP_SHADER_INFO_REPLY: + return (struct rbug_header *)rbug_demarshal_shader_info_reply(header); + default: + return NULL; + } +} diff --git a/src/gallium/auxiliary/rbug/rbug_internal.h b/src/gallium/auxiliary/rbug/rbug_internal.h new file mode 100644 index 0000000000..4aba1a810f --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_internal.h @@ -0,0 +1,100 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file is internal to the rbug protocol code, and contains asorted + * features needed by the code. + */ + +#ifndef _RBUG_INTERNAL_H_ +#define _RBUG_INTERNAL_H_ + +#include "rbug/rbug_proto.h" + +#include "util/u_memory.h" +#include "util/u_debug.h" +#include <errno.h> + +int rbug_connection_send_start(struct rbug_connection *con, enum rbug_opcode opcode, uint32_t length); +int rbug_connection_write(struct rbug_connection *con, void *data, uint32_t size); +int rbug_connection_send_finish(struct rbug_connection *con, uint32_t *c); + +/** + * Only works with multiples of 2 + */ +#define PAD(from, to) \ +do { \ + from = (from + to - 1) & ~(to - 1); \ +} while(0) + +#define LEN(size) \ +do { \ + PAD(__len, size); \ + __len += size; \ +} while(0) + +#define LEN_ARRAY(size, name) \ +do { \ + LEN(4); \ + PAD(__len, size); \ + __len += size * name##_len; \ +} while(0) + +#define WRITE(size, type, name) \ +do { \ + PAD(__pos, size); \ + *((type *)(&__data[__pos])) = name; \ + __pos += size; \ +} while(0) + +#define WRITE_ARRAY(size, type, name) \ +do { \ + WRITE(4, uint32_t, name##_len); \ + PAD(__pos, size); \ + memcpy(&__data[__pos], name, size * name##_len); \ + __pos += size * name##_len; \ +} while(0) + +#define READ(size, type, name) \ +do { \ + PAD(pos, size); \ + pos += size; \ + if (pos > len) \ + break; \ + ret->name = *((type *)(&data[pos - size])); \ +} while(0) + +#define READ_ARRAY(size, type, name) \ +do { \ + READ(4, uint32_t, name##_len); \ + if (pos > len) \ + break; \ + PAD(pos, size); \ + pos += size * ret->name##_len; \ + if (pos > len) \ + break; \ + ret->name = (type *)&data[pos - size * ret->name##_len]; \ +} while(0) + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_proto.h b/src/gallium/auxiliary/rbug/rbug_proto.h new file mode 100644 index 0000000000..d273be0166 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_proto.h @@ -0,0 +1,94 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds common definitions of the gallium remote debugging protocol. + */ + +#ifndef _RBUG_PROTO_H_ +#define _RBUG_PROTO_H_ + +#include "pipe/p_compiler.h" + +/** + * Uniqe indentifier for each command. + * + * Replys are designated by negative. + */ +enum rbug_opcode +{ + RBUG_OP_NOOP = 0, + RBUG_OP_PING = 1, + RBUG_OP_ERROR = 2, + RBUG_OP_PING_REPLY = -1, + RBUG_OP_ERROR_REPLY = -2, + RBUG_OP_TEXTURE_LIST = 256, + RBUG_OP_TEXTURE_INFO = 257, + RBUG_OP_TEXTURE_WRITE = 258, + RBUG_OP_TEXTURE_READ = 259, + RBUG_OP_TEXTURE_LIST_REPLY = -256, + RBUG_OP_TEXTURE_INFO_REPLY = -257, + RBUG_OP_TEXTURE_READ_REPLY = -259, + RBUG_OP_CONTEXT_LIST = 512, + RBUG_OP_CONTEXT_INFO = 513, + RBUG_OP_CONTEXT_DRAW_BLOCK = 514, + RBUG_OP_CONTEXT_DRAW_STEP = 515, + RBUG_OP_CONTEXT_DRAW_UNBLOCK = 516, + RBUG_OP_CONTEXT_DRAW_RULE = 518, + RBUG_OP_CONTEXT_FLUSH = 519, + RBUG_OP_CONTEXT_LIST_REPLY = -512, + RBUG_OP_CONTEXT_INFO_REPLY = -513, + RBUG_OP_CONTEXT_DRAW_BLOCKED = 517, + RBUG_OP_SHADER_LIST = 768, + RBUG_OP_SHADER_INFO = 769, + RBUG_OP_SHADER_DISABLE = 770, + RBUG_OP_SHADER_REPLACE = 771, + RBUG_OP_SHADER_LIST_REPLY = -768, + RBUG_OP_SHADER_INFO_REPLY = -769, +}; + +/** + * Header for demarshaled message. + */ +struct rbug_header +{ + enum rbug_opcode opcode; + void *__message; +}; + +/** + * Header for a message in wire format. + */ +struct rbug_proto_header +{ + int32_t opcode; + uint32_t length; +}; + +/** + * Forward declare connection here, as this file is included by all users. + */ +struct rbug_connection; + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_shader.c b/src/gallium/auxiliary/rbug/rbug_shader.c new file mode 100644 index 0000000000..fccd2f55ef --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_shader.c @@ -0,0 +1,468 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds the function implementation for one of the rbug extensions. + * Prototypes and declerations of functions and structs is in the same folder + * in the header file matching this file's name. + * + * The functions starting rbug_send_* encodes a call to the write format and + * sends that to the supplied connection, while functions starting with + * rbug_demarshal_* demarshal data in the wire protocol. + * + * Functions ending with _reply are replies to requests. + */ + +#include "rbug_internal.h" +#include "rbug/rbug_shader.h" + +int rbug_send_shader_list(struct rbug_connection *__con, + rbug_context_t context, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_shader_info(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(8); /* shader */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(8, rbug_shader_t, shader); /* shader */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_shader_disable(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint8_t disable, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(8); /* shader */ + LEN(1); /* disable */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_DISABLE)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(8, rbug_shader_t, shader); /* shader */ + WRITE(1, uint8_t, disable); /* disable */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_DISABLE, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_shader_replace(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint32_t *tokens, + uint32_t tokens_len, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* context */ + LEN(8); /* shader */ + LEN_ARRAY(4, tokens); /* tokens */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_REPLACE)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_context_t, context); /* context */ + WRITE(8, rbug_shader_t, shader); /* shader */ + WRITE_ARRAY(4, uint32_t, tokens); /* tokens */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_REPLACE, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_shader_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_shader_t *shaders, + uint32_t shaders_len, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN_ARRAY(8, shaders); /* shaders */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE_ARRAY(8, rbug_shader_t, shaders); /* shaders */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_shader_info_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t *original, + uint32_t original_len, + uint32_t *replaced, + uint32_t replaced_len, + uint8_t disabled, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN_ARRAY(4, original); /* original */ + LEN_ARRAY(4, replaced); /* replaced */ + LEN(1); /* disabled */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE_ARRAY(4, uint32_t, original); /* original */ + WRITE_ARRAY(4, uint32_t, replaced); /* replaced */ + WRITE(1, uint8_t, disabled); /* disabled */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_list *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + + return ret; +} + +struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_info *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(8, rbug_shader_t, shader); /* shader */ + + return ret; +} + +struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_disable *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_DISABLE) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(8, rbug_shader_t, shader); /* shader */ + READ(1, uint8_t, disable); /* disable */ + + return ret; +} + +struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_replace *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_REPLACE) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_context_t, context); /* context */ + READ(8, rbug_shader_t, shader); /* shader */ + READ_ARRAY(4, uint32_t, tokens); /* tokens */ + + return ret; +} + +struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_list_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ_ARRAY(8, rbug_shader_t, shaders); /* shaders */ + + return ret; +} + +struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_shader_info_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ_ARRAY(4, uint32_t, original); /* original */ + READ_ARRAY(4, uint32_t, replaced); /* replaced */ + READ(1, uint8_t, disabled); /* disabled */ + + return ret; +} diff --git a/src/gallium/auxiliary/rbug/rbug_shader.h b/src/gallium/auxiliary/rbug/rbug_shader.h new file mode 100644 index 0000000000..b5d886781d --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_shader.h @@ -0,0 +1,142 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds structs decelerations and function prototypes for one of + * the rbug extensions. Implementation of the functions is in the same folder + * in the c file matching this file's name. + * + * The structs what is returned from the demarshal functions. The functions + * starting rbug_send_* encodes a call to the write format and sends that to + * the supplied connection, while functions starting with rbug_demarshal_* + * demarshal data from the wire protocol. + * + * Structs and functions ending with _reply are replies to requests. + */ + +#ifndef _RBUG_PROTO_SHADER_H_ +#define _RBUG_PROTO_SHADER_H_ + +#include "rbug/rbug_proto.h" +#include "rbug/rbug_core.h" + +struct rbug_proto_shader_list +{ + struct rbug_header header; + rbug_context_t context; +}; + +struct rbug_proto_shader_info +{ + struct rbug_header header; + rbug_context_t context; + rbug_shader_t shader; +}; + +struct rbug_proto_shader_disable +{ + struct rbug_header header; + rbug_context_t context; + rbug_shader_t shader; + uint8_t disable; +}; + +struct rbug_proto_shader_replace +{ + struct rbug_header header; + rbug_context_t context; + rbug_shader_t shader; + uint32_t *tokens; + uint32_t tokens_len; +}; + +struct rbug_proto_shader_list_reply +{ + struct rbug_header header; + uint32_t serial; + rbug_shader_t *shaders; + uint32_t shaders_len; +}; + +struct rbug_proto_shader_info_reply +{ + struct rbug_header header; + uint32_t serial; + uint32_t *original; + uint32_t original_len; + uint32_t *replaced; + uint32_t replaced_len; + uint8_t disabled; +}; + +int rbug_send_shader_list(struct rbug_connection *__con, + rbug_context_t context, + uint32_t *__serial); + +int rbug_send_shader_info(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint32_t *__serial); + +int rbug_send_shader_disable(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint8_t disable, + uint32_t *__serial); + +int rbug_send_shader_replace(struct rbug_connection *__con, + rbug_context_t context, + rbug_shader_t shader, + uint32_t *tokens, + uint32_t tokens_len, + uint32_t *__serial); + +int rbug_send_shader_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_shader_t *shaders, + uint32_t shaders_len, + uint32_t *__serial); + +int rbug_send_shader_info_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t *original, + uint32_t original_len, + uint32_t *replaced, + uint32_t replaced_len, + uint8_t disabled, + uint32_t *__serial); + +struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header); + +struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header); + +struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header); + +struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header); + +struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header); + +struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header); + +#endif diff --git a/src/gallium/auxiliary/rbug/rbug_texture.c b/src/gallium/auxiliary/rbug/rbug_texture.c new file mode 100644 index 0000000000..5a918fe6bc --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_texture.c @@ -0,0 +1,631 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds the function implementation for one of the rbug extensions. + * Prototypes and declerations of functions and structs is in the same folder + * in the header file matching this file's name. + * + * The functions starting rbug_send_* encodes a call to the write format and + * sends that to the supplied connection, while functions starting with + * rbug_demarshal_* demarshal data in the wire protocol. + * + * Functions ending with _reply are replies to requests. + */ + +#include "rbug_internal.h" +#include "rbug/rbug_texture.h" + +int rbug_send_texture_list(struct rbug_connection *__con, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_info(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* texture */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_texture_t, texture); /* texture */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_write(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t face, + uint32_t level, + uint32_t zslice, + uint32_t x, + uint32_t y, + uint32_t w, + uint32_t h, + uint8_t *data, + uint32_t data_len, + uint32_t stride, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* texture */ + LEN(4); /* face */ + LEN(4); /* level */ + LEN(4); /* zslice */ + LEN(4); /* x */ + LEN(4); /* y */ + LEN(4); /* w */ + LEN(4); /* h */ + LEN_ARRAY(1, data); /* data */ + LEN(4); /* stride */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_WRITE)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_texture_t, texture); /* texture */ + WRITE(4, uint32_t, face); /* face */ + WRITE(4, uint32_t, level); /* level */ + WRITE(4, uint32_t, zslice); /* zslice */ + WRITE(4, uint32_t, x); /* x */ + WRITE(4, uint32_t, y); /* y */ + WRITE(4, uint32_t, w); /* w */ + WRITE(4, uint32_t, h); /* h */ + WRITE_ARRAY(1, uint8_t, data); /* data */ + WRITE(4, uint32_t, stride); /* stride */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_WRITE, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_read(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t face, + uint32_t level, + uint32_t zslice, + uint32_t x, + uint32_t y, + uint32_t w, + uint32_t h, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(8); /* texture */ + LEN(4); /* face */ + LEN(4); /* level */ + LEN(4); /* zslice */ + LEN(4); /* x */ + LEN(4); /* y */ + LEN(4); /* w */ + LEN(4); /* h */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(8, rbug_texture_t, texture); /* texture */ + WRITE(4, uint32_t, face); /* face */ + WRITE(4, uint32_t, level); /* level */ + WRITE(4, uint32_t, zslice); /* zslice */ + WRITE(4, uint32_t, x); /* x */ + WRITE(4, uint32_t, y); /* y */ + WRITE(4, uint32_t, w); /* w */ + WRITE(4, uint32_t, h); /* h */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_texture_t *textures, + uint32_t textures_len, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN_ARRAY(8, textures); /* textures */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE_ARRAY(8, rbug_texture_t, textures); /* textures */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_info_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t target, + uint32_t format, + uint32_t *width, + uint32_t width_len, + uint32_t *height, + uint32_t height_len, + uint32_t *depth, + uint32_t depth_len, + uint32_t blockw, + uint32_t blockh, + uint32_t blocksize, + uint32_t last_level, + uint32_t nr_samples, + uint32_t tex_usage, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN(4); /* target */ + LEN(4); /* format */ + LEN_ARRAY(4, width); /* width */ + LEN_ARRAY(4, height); /* height */ + LEN_ARRAY(4, depth); /* depth */ + LEN(4); /* blockw */ + LEN(4); /* blockh */ + LEN(4); /* blocksize */ + LEN(4); /* last_level */ + LEN(4); /* nr_samples */ + LEN(4); /* tex_usage */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE(4, uint32_t, target); /* target */ + WRITE(4, uint32_t, format); /* format */ + WRITE_ARRAY(4, uint32_t, width); /* width */ + WRITE_ARRAY(4, uint32_t, height); /* height */ + WRITE_ARRAY(4, uint32_t, depth); /* depth */ + WRITE(4, uint32_t, blockw); /* blockw */ + WRITE(4, uint32_t, blockh); /* blockh */ + WRITE(4, uint32_t, blocksize); /* blocksize */ + WRITE(4, uint32_t, last_level); /* last_level */ + WRITE(4, uint32_t, nr_samples); /* nr_samples */ + WRITE(4, uint32_t, tex_usage); /* tex_usage */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +int rbug_send_texture_read_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t format, + uint32_t blockw, + uint32_t blockh, + uint32_t blocksize, + uint8_t *data, + uint32_t data_len, + uint32_t stride, + uint32_t *__serial) +{ + uint32_t __len = 0; + uint32_t __pos = 0; + uint8_t *__data = NULL; + int __ret = 0; + + LEN(8); /* header */ + LEN(4); /* serial */ + LEN(4); /* format */ + LEN(4); /* blockw */ + LEN(4); /* blockh */ + LEN(4); /* blocksize */ + LEN_ARRAY(1, data); /* data */ + LEN(4); /* stride */ + + /* align */ + PAD(__len, 8); + + __data = (uint8_t*)MALLOC(__len); + if (!__data) + return -ENOMEM; + + WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ_REPLY)); + WRITE(4, uint32_t, ((uint32_t)(__len / 4))); + WRITE(4, uint32_t, serial); /* serial */ + WRITE(4, uint32_t, format); /* format */ + WRITE(4, uint32_t, blockw); /* blockw */ + WRITE(4, uint32_t, blockh); /* blockh */ + WRITE(4, uint32_t, blocksize); /* blocksize */ + WRITE_ARRAY(1, uint8_t, data); /* data */ + WRITE(4, uint32_t, stride); /* stride */ + + /* final pad */ + PAD(__pos, 8); + + if (__pos != __len) { + __ret = -EINVAL; + } else { + rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ_REPLY, __len); + rbug_connection_write(__con, __data, __len); + __ret = rbug_connection_send_finish(__con, __serial); + } + + FREE(__data); + return __ret; +} + +struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_list *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + + return ret; +} + +struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_info *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_texture_t, texture); /* texture */ + + return ret; +} + +struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_write *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_WRITE) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_texture_t, texture); /* texture */ + READ(4, uint32_t, face); /* face */ + READ(4, uint32_t, level); /* level */ + READ(4, uint32_t, zslice); /* zslice */ + READ(4, uint32_t, x); /* x */ + READ(4, uint32_t, y); /* y */ + READ(4, uint32_t, w); /* w */ + READ(4, uint32_t, h); /* h */ + READ_ARRAY(1, uint8_t, data); /* data */ + READ(4, uint32_t, stride); /* stride */ + + return ret; +} + +struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_read *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(8, rbug_texture_t, texture); /* texture */ + READ(4, uint32_t, face); /* face */ + READ(4, uint32_t, level); /* level */ + READ(4, uint32_t, zslice); /* zslice */ + READ(4, uint32_t, x); /* x */ + READ(4, uint32_t, y); /* y */ + READ(4, uint32_t, w); /* w */ + READ(4, uint32_t, h); /* h */ + + return ret; +} + +struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_list_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ_ARRAY(8, rbug_texture_t, textures); /* textures */ + + return ret; +} + +struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_info_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ(4, uint32_t, target); /* target */ + READ(4, uint32_t, format); /* format */ + READ_ARRAY(4, uint32_t, width); /* width */ + READ_ARRAY(4, uint32_t, height); /* height */ + READ_ARRAY(4, uint32_t, depth); /* depth */ + READ(4, uint32_t, blockw); /* blockw */ + READ(4, uint32_t, blockh); /* blockh */ + READ(4, uint32_t, blocksize); /* blocksize */ + READ(4, uint32_t, last_level); /* last_level */ + READ(4, uint32_t, nr_samples); /* nr_samples */ + READ(4, uint32_t, tex_usage); /* tex_usage */ + + return ret; +} + +struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header) +{ + uint32_t len = 0; + uint32_t pos = 0; + uint8_t *data = NULL; + struct rbug_proto_texture_read_reply *ret; + + if (!header) + return NULL; + if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ_REPLY) + return NULL; + + pos = 0; + len = header->length * 4; + data = (uint8_t*)&header[1]; + ret = MALLOC(sizeof(*ret)); + if (!ret) + return NULL; + + ret->header.__message = header; + ret->header.opcode = header->opcode; + + READ(4, uint32_t, serial); /* serial */ + READ(4, uint32_t, format); /* format */ + READ(4, uint32_t, blockw); /* blockw */ + READ(4, uint32_t, blockh); /* blockh */ + READ(4, uint32_t, blocksize); /* blocksize */ + READ_ARRAY(1, uint8_t, data); /* data */ + READ(4, uint32_t, stride); /* stride */ + + return ret; +} diff --git a/src/gallium/auxiliary/rbug/rbug_texture.h b/src/gallium/auxiliary/rbug/rbug_texture.h new file mode 100644 index 0000000000..fbb247e1d4 --- /dev/null +++ b/src/gallium/auxiliary/rbug/rbug_texture.h @@ -0,0 +1,206 @@ +/* + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * This file holds structs decelerations and function prototypes for one of + * the rbug extensions. Implementation of the functions is in the same folder + * in the c file matching this file's name. + * + * The structs what is returned from the demarshal functions. The functions + * starting rbug_send_* encodes a call to the write format and sends that to + * the supplied connection, while functions starting with rbug_demarshal_* + * demarshal data from the wire protocol. + * + * Structs and functions ending with _reply are replies to requests. + */ + +#ifndef _RBUG_PROTO_TEXTURE_H_ +#define _RBUG_PROTO_TEXTURE_H_ + +#include "rbug/rbug_proto.h" +#include "rbug/rbug_core.h" + +struct rbug_proto_texture_list +{ + struct rbug_header header; +}; + +struct rbug_proto_texture_info +{ + struct rbug_header header; + rbug_texture_t texture; +}; + +struct rbug_proto_texture_write +{ + struct rbug_header header; + rbug_texture_t texture; + uint32_t face; + uint32_t level; + uint32_t zslice; + uint32_t x; + uint32_t y; + uint32_t w; + uint32_t h; + uint8_t *data; + uint32_t data_len; + uint32_t stride; +}; + +struct rbug_proto_texture_read +{ + struct rbug_header header; + rbug_texture_t texture; + uint32_t face; + uint32_t level; + uint32_t zslice; + uint32_t x; + uint32_t y; + uint32_t w; + uint32_t h; +}; + +struct rbug_proto_texture_list_reply +{ + struct rbug_header header; + uint32_t serial; + rbug_texture_t *textures; + uint32_t textures_len; +}; + +struct rbug_proto_texture_info_reply +{ + struct rbug_header header; + uint32_t serial; + uint32_t target; + uint32_t format; + uint32_t *width; + uint32_t width_len; + uint32_t *height; + uint32_t height_len; + uint32_t *depth; + uint32_t depth_len; + uint32_t blockw; + uint32_t blockh; + uint32_t blocksize; + uint32_t last_level; + uint32_t nr_samples; + uint32_t tex_usage; +}; + +struct rbug_proto_texture_read_reply +{ + struct rbug_header header; + uint32_t serial; + uint32_t format; + uint32_t blockw; + uint32_t blockh; + uint32_t blocksize; + uint8_t *data; + uint32_t data_len; + uint32_t stride; +}; + +int rbug_send_texture_list(struct rbug_connection *__con, + uint32_t *__serial); + +int rbug_send_texture_info(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t *__serial); + +int rbug_send_texture_write(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t face, + uint32_t level, + uint32_t zslice, + uint32_t x, + uint32_t y, + uint32_t w, + uint32_t h, + uint8_t *data, + uint32_t data_len, + uint32_t stride, + uint32_t *__serial); + +int rbug_send_texture_read(struct rbug_connection *__con, + rbug_texture_t texture, + uint32_t face, + uint32_t level, + uint32_t zslice, + uint32_t x, + uint32_t y, + uint32_t w, + uint32_t h, + uint32_t *__serial); + +int rbug_send_texture_list_reply(struct rbug_connection *__con, + uint32_t serial, + rbug_texture_t *textures, + uint32_t textures_len, + uint32_t *__serial); + +int rbug_send_texture_info_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t target, + uint32_t format, + uint32_t *width, + uint32_t width_len, + uint32_t *height, + uint32_t height_len, + uint32_t *depth, + uint32_t depth_len, + uint32_t blockw, + uint32_t blockh, + uint32_t blocksize, + uint32_t last_level, + uint32_t nr_samples, + uint32_t tex_usage, + uint32_t *__serial); + +int rbug_send_texture_read_reply(struct rbug_connection *__con, + uint32_t serial, + uint32_t format, + uint32_t blockw, + uint32_t blockh, + uint32_t blocksize, + uint8_t *data, + uint32_t data_len, + uint32_t stride, + uint32_t *__serial); + +struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header); + +struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header); + +struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header); + +struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header); + +struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header); + +struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header); + +struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header); + +#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index 03bdd47238..2e15751e50 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -30,6 +30,7 @@ #include "rtasm_cpu.h" +#if defined(PIPE_ARCH_X86) static boolean rtasm_sse_enabled(void) { static boolean firsttime = 1; @@ -43,6 +44,7 @@ static boolean rtasm_sse_enabled(void) } return enabled; } +#endif int rtasm_cpu_has_sse(void) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index e8bd7cda3b..aba7a3f937 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -53,6 +53,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_dump.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_exec.h" @@ -169,6 +170,56 @@ print_temp(const struct tgsi_exec_machine *mach, uint index) #endif +/** + * Check if there's a potential src/dst register data dependency when + * using SOA execution. + * Example: + * MOV T, T.yxwz; + * This would expand into: + * MOV t0, t1; + * MOV t1, t0; + * MOV t2, t3; + * MOV t3, t2; + * The second instruction will have the wrong value for t0 if executed as-is. + */ +static boolean +tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst) +{ + uint i, chan; + + uint writemask = inst->FullDstRegisters[0].DstRegister.WriteMask; + if (writemask == TGSI_WRITEMASK_X || + writemask == TGSI_WRITEMASK_Y || + writemask == TGSI_WRITEMASK_Z || + writemask == TGSI_WRITEMASK_W || + writemask == TGSI_WRITEMASK_NONE) { + /* no chance of data dependency */ + return FALSE; + } + + /* loop over src regs */ + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + if ((inst->FullSrcRegisters[i].SrcRegister.File == + inst->FullDstRegisters[0].DstRegister.File) && + (inst->FullSrcRegisters[i].SrcRegister.Index == + inst->FullDstRegisters[0].DstRegister.Index)) { + /* loop over dest channels */ + uint channelsWritten = 0x0; + FOR_EACH_ENABLED_CHANNEL(*inst, chan) { + /* check if we're reading a channel that's been written */ + uint swizzle = tgsi_util_get_full_src_register_extswizzle(&inst->FullSrcRegisters[i], chan); + if (swizzle <= TGSI_SWIZZLE_W && + (channelsWritten & (1 << swizzle))) { + return TRUE; + } + + channelsWritten |= (1 << chan); + } + } + } + return FALSE; +} + /** * Initialize machine state by expanding tokens to full instructions, @@ -280,6 +331,17 @@ tgsi_exec_machine_bind_shader( memcpy(instructions + numInstructions, &parse.FullToken.FullInstruction, sizeof(instructions[0])); + +#if 0 + if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) { + debug_printf("SOA dependency in instruction:\n"); + tgsi_dump_instruction(&parse.FullToken.FullInstruction, + numInstructions); + } +#else + (void) tgsi_check_soa_dependencies; +#endif + numInstructions++; break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index 0b4b2a6fb6..da22baad3e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -34,7 +34,7 @@ extern "C" { #endif -#define MAX_LABELS 1024 +#define MAX_LABELS (4 * 1024) /**< basically, max instructions */ #define NUM_CHANNELS 4 /* R,G,B,A */ #define QUAD_SIZE 4 /* 4 pixel/quad */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index d70bcd03c5..ba2bfdef06 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1466,6 +1466,31 @@ emit_cmp( } } + +/** + * Check if inst src/dest regs use indirect addressing into temporary + * register file. + */ +static boolean +indirect_temp_reference(const struct tgsi_full_instruction *inst) +{ + uint i; + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[i]; + if (reg->SrcRegister.File == TGSI_FILE_TEMPORARY && + reg->SrcRegister.Indirect) + return TRUE; + } + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[i]; + if (reg->DstRegister.File == TGSI_FILE_TEMPORARY && + reg->DstRegister.Indirect) + return TRUE; + } + return FALSE; +} + + static int emit_instruction( struct x86_function *func, @@ -1473,10 +1498,15 @@ emit_instruction( { unsigned chan_index; + /* we can't handle indirect addressing into temp register file yet */ + if (indirect_temp_reference(inst)) + return FALSE; + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); + emit_flr(func, 0, 0); emit_f2it( func, 0 ); STORE( func, *inst, 0, 0, chan_index ); } @@ -1553,7 +1583,7 @@ emit_instruction( func, make_xmm( 2 ), make_xmm( 0 ), - cc_LessThanEqual ); + cc_LessThan ); sse_andps( func, make_xmm( 2 ), @@ -2177,32 +2207,83 @@ emit_instruction( /* 3 or 4-component normalization */ { uint dims = (inst->Instruction.Opcode == TGSI_OPCODE_NRM) ? 3 : 4; - /* note: cannot use xmm regs 2/3 here (see emit_rsqrt() above) */ - FETCH( func, *inst, 4, 0, CHAN_X ); /* xmm4 = src[0].x */ - FETCH( func, *inst, 5, 0, CHAN_Y ); /* xmm5 = src[0].y */ - FETCH( func, *inst, 6, 0, CHAN_Z ); /* xmm6 = src[0].z */ - if (dims == 4) { - FETCH( func, *inst, 7, 0, CHAN_W ); /* xmm7 = src[0].w */ - } - emit_MOV( func, 0, 4 ); /* xmm0 = xmm3 */ - emit_mul( func, 0, 4 ); /* xmm0 *= xmm3 */ - emit_MOV( func, 1, 5 ); /* xmm1 = xmm4 */ - emit_mul( func, 1, 5 ); /* xmm1 *= xmm4 */ - emit_add( func, 0, 1 ); /* xmm0 += xmm1 */ - emit_MOV( func, 1, 6 ); /* xmm1 = xmm5 */ - emit_mul( func, 1, 6 ); /* xmm1 *= xmm5 */ - emit_add( func, 0, 1 ); /* xmm0 += xmm1 */ - if (dims == 4) { - emit_MOV( func, 1, 7 ); /* xmm1 = xmm7 */ - emit_mul( func, 1, 7 ); /* xmm1 *= xmm7 */ - emit_add( func, 0, 0 ); /* xmm0 += xmm1 */ - } - emit_rsqrt( func, 1, 0 ); /* xmm1 = 1/sqrt(xmm0) */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - if (chan_index < dims) { - emit_mul( func, 4+chan_index, 1); /* xmm[4+ch] *= xmm1 */ - STORE( func, *inst, 4+chan_index, 0, chan_index ); + + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) || + IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y) || + IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z) || + (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W) && dims == 4)) { + + /* NOTE: Cannot use xmm regs 2/3 here (see emit_rsqrt() above). */ + + /* xmm4 = src.x */ + /* xmm0 = src.x * src.x */ + FETCH(func, *inst, 0, 0, CHAN_X); + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X)) { + emit_MOV(func, 4, 0); + } + emit_mul(func, 0, 0); + + /* xmm5 = src.y */ + /* xmm0 = xmm0 + src.y * src.y */ + FETCH(func, *inst, 1, 0, CHAN_Y); + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) { + emit_MOV(func, 5, 1); + } + emit_mul(func, 1, 1); + emit_add(func, 0, 1); + + /* xmm6 = src.z */ + /* xmm0 = xmm0 + src.z * src.z */ + FETCH(func, *inst, 1, 0, CHAN_Z); + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) { + emit_MOV(func, 6, 1); + } + emit_mul(func, 1, 1); + emit_add(func, 0, 1); + + if (dims == 4) { + /* xmm7 = src.w */ + /* xmm0 = xmm0 + src.w * src.w */ + FETCH(func, *inst, 1, 0, CHAN_W); + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W)) { + emit_MOV(func, 7, 1); + } + emit_mul(func, 1, 1); + emit_add(func, 0, 1); } + + /* xmm1 = 1 / sqrt(xmm0) */ + emit_rsqrt(func, 1, 0); + + /* dst.x = xmm1 * src.x */ + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X)) { + emit_mul(func, 4, 1); + STORE(func, *inst, 4, 0, CHAN_X); + } + + /* dst.y = xmm1 * src.y */ + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) { + emit_mul(func, 5, 1); + STORE(func, *inst, 5, 0, CHAN_Y); + } + + /* dst.z = xmm1 * src.z */ + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) { + emit_mul(func, 6, 1); + STORE(func, *inst, 6, 0, CHAN_Z); + } + + /* dst.w = xmm1 * src.w */ + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) && dims == 4) { + emit_mul(func, 7, 1); + STORE(func, *inst, 7, 0, CHAN_W); + } + } + + /* dst0.w = 1.0 */ + if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W) && dims == 3) { + emit_tempf(func, 0, TEMP_ONE_I, TEMP_ONE_C); + STORE(func, *inst, 0, 0, CHAN_W); } } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c index 062c1be938..bc9c18fd4a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.c +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c @@ -198,3 +198,30 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in, return ctx->ti; } + + +#include "tgsi_text.h" + +extern int tgsi_transform_foo( struct tgsi_token *tokens_out, + uint max_tokens_out ); + +/* This function exists only so that tgsi_text_translate() doesn't get + * magic-ed out of the libtgsi.a archive by the build system. Don't + * remove unless you know this has been fixed - check on mingw/scons + * builds as well. + */ +int +tgsi_transform_foo( struct tgsi_token *tokens_out, + uint max_tokens_out ) +{ + const char *text = + "FRAG1.1\n" + "DCL IN[0], COLOR, CONSTANT\n" + "DCL OUT[0], COLOR\n" + " 0: MOV OUT[0], IN[0]\n" + " 1: END"; + + return tgsi_text_translate( text, + tokens_out, + max_tokens_out ); +} diff --git a/src/gallium/auxiliary/trace/trace_drm.h b/src/gallium/auxiliary/trace/trace_drm.h new file mode 100644 index 0000000000..892bd9860c --- /dev/null +++ b/src/gallium/auxiliary/trace/trace_drm.h @@ -0,0 +1,165 @@ +/* + * Copyright 2009 Jakob Bornecrantz <jakob@vmware.com> + * Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef TRACE_DRM_H +#define TRACE_DRM_H + +#include "state_tracker/drm_api.h" + +#include "trace/tr_buffer.h" +#include "trace/tr_context.h" +#include "trace/tr_screen.h" +#include "trace/tr_texture.h" + +struct drm_api hooks; + +static struct pipe_screen * +trace_drm_create_screen(int fd, struct drm_create_screen_arg *arg) +{ + struct pipe_screen *screen; + + if (arg && arg->mode != DRM_CREATE_NORMAL) + return NULL; + + screen = hooks.create_screen(fd, arg); + + return trace_screen_create(screen); +}; + +static struct pipe_context * +trace_drm_create_context(struct pipe_screen *_screen) +{ + struct pipe_screen *screen; + struct pipe_context *pipe; + + if (trace_enabled()) + screen = trace_screen(_screen)->screen; + else + screen = _screen; + + pipe = hooks.create_context(screen); + + if (trace_enabled()) + pipe = trace_context_create(_screen, pipe); + + return pipe; +}; + +static boolean +trace_drm_buffer_from_texture(struct pipe_texture *_texture, + struct pipe_buffer **_buffer, + unsigned *stride) +{ + struct pipe_texture *texture; + struct pipe_buffer *buffer = NULL; + boolean result; + + if (trace_enabled()) + texture = trace_texture(_texture)->texture; + else + texture = _texture; + + result = hooks.buffer_from_texture(texture, &buffer, stride); + + if (result && _buffer) + buffer = trace_buffer_create(trace_screen(texture->screen), buffer); + + if (_buffer) + *_buffer = buffer; + else + pipe_buffer_reference(&buffer, NULL); + + return result; +} + +static struct pipe_buffer * +trace_drm_buffer_from_handle(struct pipe_screen *_screen, + const char *name, + unsigned handle) +{ + struct pipe_screen *screen; + struct pipe_buffer *result; + + if (trace_enabled()) + screen = trace_screen(_screen)->screen; + else + screen = _screen; + + result = hooks.buffer_from_handle(screen, name, handle); + + if (trace_enabled()) + result = trace_buffer_create(trace_screen(_screen), result); + + return result; +} + +static boolean +trace_drm_handle_from_buffer(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct pipe_screen *screen; + struct pipe_buffer *buffer; + + if (trace_enabled()) { + screen = trace_screen(_screen)->screen; + buffer = trace_buffer(_buffer)->buffer; + } else { + screen = _screen; + buffer = _buffer; + } + + return hooks.handle_from_buffer(screen, buffer, handle); +} + +static boolean +trace_drm_global_handle_from_buffer(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct pipe_screen *screen; + struct pipe_buffer *buffer; + + if (trace_enabled()) { + screen = trace_screen(_screen)->screen; + buffer = trace_buffer(_buffer)->buffer; + } else { + screen = _screen; + buffer = _buffer; + } + + return hooks.global_handle_from_buffer(screen, buffer, handle); +} + +struct drm_api drm_api_hooks = +{ + .create_screen = trace_drm_create_screen, + .create_context = trace_drm_create_context, + + .buffer_from_texture = trace_drm_buffer_from_texture, + .buffer_from_handle = trace_drm_buffer_from_handle, + .handle_from_buffer = trace_drm_handle_from_buffer, + .global_handle_from_buffer = trace_drm_global_handle_from_buffer, +}; + +#endif /* TRACE_DRM_H */ diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 5035e9cc13..6a8eb73e84 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -16,6 +16,7 @@ C_SOURCES = \ u_hash.c \ u_keymap.c \ u_linear.c \ + u_network.c \ u_math.c \ u_mm.c \ u_rect.c \ @@ -23,6 +24,7 @@ C_SOURCES = \ u_snprintf.c \ u_stream_stdc.c \ u_stream_wd.c \ + u_surface.c \ u_tile.c \ u_time.c \ u_timed_winsys.c \ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 8317263bb8..fb142eebca 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -17,6 +17,7 @@ util = env.ConvenienceLibrary( 'u_hash.c', 'u_hash_table.c', 'u_keymap.c', + 'u_network.c', 'u_math.c', 'u_mm.c', 'u_rect.c', @@ -24,6 +25,7 @@ util = env.ConvenienceLibrary( 'u_snprintf.c', 'u_stream_stdc.c', 'u_stream_wd.c', + 'u_surface.c', 'u_tile.c', 'u_time.c', 'u_timed_winsys.c', diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index deb68c43a6..414cf91025 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -89,10 +89,6 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) /* disabled blending/masking */ memset(&ctx->blend, 0, sizeof(ctx->blend)); - ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; - ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; ctx->blend.colormask = PIPE_MASK_RGBA; /* no-op depth/stencil/alpha */ @@ -337,7 +333,6 @@ util_blit_pixels(struct blit_state *ctx, texTemp.width[0] = srcW; texTemp.height[0] = srcH; texTemp.depth[0] = 1; - texTemp.compressed = 0; pf_get_block(src->format, &texTemp.block); tex = screen->texture_create(screen, &texTemp); diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c index 96a2222f9b..18597ef839 100644 --- a/src/gallium/auxiliary/util/u_debug.c +++ b/src/gallium/auxiliary/util/u_debug.c @@ -109,6 +109,7 @@ void _debug_vprintf(const char *format, va_list ap) } if(GetConsoleWindow() && !IsDebuggerPresent()) { + fflush(stdout); vfprintf(stderr, format, ap); fflush(stderr); } @@ -145,6 +146,7 @@ void _debug_vprintf(const char *format, va_list ap) /* TODO */ #else /* !PIPE_SUBSYSTEM_WINDOWS */ #ifdef DEBUG + fflush(stdout); vfprintf(stderr, format, ap); #endif #endif @@ -719,6 +721,7 @@ void debug_dump_surface_bmp(const char *filename, struct pipe_surface *surface) { +#ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT struct pipe_transfer *transfer; struct pipe_texture *texture = surface->texture; struct pipe_screen *screen = texture->screen; @@ -731,6 +734,7 @@ debug_dump_surface_bmp(const char *filename, debug_dump_transfer_bmp(filename, transfer); screen->tex_transfer_destroy(transfer); +#endif } void @@ -738,11 +742,7 @@ debug_dump_transfer_bmp(const char *filename, struct pipe_transfer *transfer) { #ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT - struct util_stream *stream; - struct bmp_file_header bmfh; - struct bmp_info_header bmih; float *rgba; - unsigned x, y; if (!transfer) goto error1; @@ -751,19 +751,47 @@ debug_dump_transfer_bmp(const char *filename, if(!rgba) goto error1; + pipe_get_tile_rgba(transfer, 0, 0, + transfer->width, transfer->height, + rgba); + + debug_dump_float_rgba_bmp(filename, + transfer->width, transfer->height, + rgba, transfer->width); + + FREE(rgba); +error1: + ; +#endif +} + +void +debug_dump_float_rgba_bmp(const char *filename, + unsigned width, unsigned height, + float *rgba, unsigned stride) +{ +#ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT + struct util_stream *stream; + struct bmp_file_header bmfh; + struct bmp_info_header bmih; + unsigned x, y; + + if(!rgba) + goto error1; + bmfh.bfType = 0x4d42; - bmfh.bfSize = 14 + 40 + transfer->height*transfer->width*4; + bmfh.bfSize = 14 + 40 + height*width*4; bmfh.bfReserved1 = 0; bmfh.bfReserved2 = 0; bmfh.bfOffBits = 14 + 40; bmih.biSize = 40; - bmih.biWidth = transfer->width; - bmih.biHeight = transfer->height; + bmih.biWidth = width; + bmih.biHeight = height; bmih.biPlanes = 1; bmih.biBitCount = 32; bmih.biCompression = 0; - bmih.biSizeImage = transfer->height*transfer->width*4; + bmih.biSizeImage = height*width*4; bmih.biXPelsPerMeter = 0; bmih.biYPelsPerMeter = 0; bmih.biClrUsed = 0; @@ -771,19 +799,15 @@ debug_dump_transfer_bmp(const char *filename, stream = util_stream_create(filename, bmfh.bfSize); if(!stream) - goto error2; + goto error1; util_stream_write(stream, &bmfh, 14); util_stream_write(stream, &bmih, 40); - pipe_get_tile_rgba(transfer, 0, 0, - transfer->width, transfer->height, - rgba); - - y = transfer->height; + y = height; while(y--) { - float *ptr = rgba + (transfer->width * y * 4); - for(x = 0; x < transfer->width; ++x) + float *ptr = rgba + (stride * y * 4); + for(x = 0; x < width; ++x) { struct bmp_rgb_quad pixel; pixel.rgbRed = float_to_ubyte(ptr[x*4 + 0]); @@ -795,8 +819,6 @@ debug_dump_transfer_bmp(const char *filename, } util_stream_close(stream); -error2: - FREE(rgba); error1: ; #endif diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h index 8d703e47fc..d42b65ce28 100644 --- a/src/gallium/auxiliary/util/u_debug.h +++ b/src/gallium/auxiliary/util/u_debug.h @@ -102,6 +102,22 @@ debug_printf(const char *format, ...) } +/* + * ... isn't portable so we need to pass arguments in parentheses. + * + * usage: + * debug_printf_once(("awnser: %i\n", 42)); + */ +#define debug_printf_once(args) \ + do { \ + static boolean once = TRUE; \ + if (once) { \ + once = FALSE; \ + debug_printf args; \ + } \ + } while (0) + + #ifdef DEBUG #define debug_vprintf(_format, _ap) _debug_vprintf(_format, _ap) #else @@ -347,10 +363,15 @@ void debug_dump_surface_bmp(const char *filename, struct pipe_surface *surface); void debug_dump_transfer_bmp(const char *filename, struct pipe_transfer *transfer); +void debug_dump_float_rgba_bmp(const char *filename, + unsigned width, unsigned height, + float *rgba, unsigned stride); #else #define debug_dump_image(prefix, format, cpp, width, height, stride, data) ((void)0) #define debug_dump_surface(prefix, surface) ((void)0) #define debug_dump_surface_bmp(filename, surface) ((void)0) +#define debug_dump_transfer_bmp(filename, transfer) ((void)0) +#define debug_dump_rgba_float_bmp(filename, width, height, rgba, stride) ((void)0) #endif diff --git a/src/gallium/auxiliary/util/u_debug_stack.c b/src/gallium/auxiliary/util/u_debug_stack.c index e9891fde8a..528a1c394b 100644 --- a/src/gallium/auxiliary/util/u_debug_stack.c +++ b/src/gallium/auxiliary/util/u_debug_stack.c @@ -62,6 +62,8 @@ debug_backtrace_capture(struct debug_stack_frame *backtrace, #ifdef PIPE_ARCH_X86 while(nr_frames) { + const void **next_frame_pointer; + if(!frame_pointer) break; @@ -72,7 +74,14 @@ debug_backtrace_capture(struct debug_stack_frame *backtrace, --nr_frames; } - frame_pointer = (const void **)frame_pointer[0]; + next_frame_pointer = (const void **)frame_pointer[0]; + + /* Limit the stack walk to avoid referencing undefined memory */ + if((uintptr_t)next_frame_pointer <= (uintptr_t)frame_pointer || + (uintptr_t)next_frame_pointer > (uintptr_t)frame_pointer + 64*1024) + break; + + frame_pointer = next_frame_pointer; } #endif diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index b32ad1cbe9..ca797486a0 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -1278,10 +1278,6 @@ util_create_gen_mipmap(struct pipe_context *pipe, /* disabled blending/masking */ memset(&ctx->blend, 0, sizeof(ctx->blend)); - ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; - ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; ctx->blend.colormask = PIPE_MASK_RGBA; /* no-op depth/stencil/alpha */ @@ -1320,7 +1316,6 @@ util_create_gen_mipmap(struct pipe_context *pipe, for (i = 0; i < 4; i++) { ctx->vertices[i][0][2] = 0.0f; /* z */ ctx->vertices[i][0][3] = 1.0f; /* w */ - ctx->vertices[i][1][2] = 0.0f; /* r */ ctx->vertices[i][1][3] = 1.0f; /* q */ } @@ -1354,29 +1349,104 @@ get_next_slot(struct gen_mipmap_state *ctx) static unsigned -set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) +set_vertex_data(struct gen_mipmap_state *ctx, + enum pipe_texture_target tex_target, + uint face, float width, float height) { unsigned offset; + /* vert[0].position */ ctx->vertices[0][0][0] = 0.0f; /*x*/ ctx->vertices[0][0][1] = 0.0f; /*y*/ - ctx->vertices[0][1][0] = 0.0f; /*s*/ - ctx->vertices[0][1][1] = 0.0f; /*t*/ + /* vert[1].position */ ctx->vertices[1][0][0] = width; ctx->vertices[1][0][1] = 0.0f; - ctx->vertices[1][1][0] = 1.0f; - ctx->vertices[1][1][1] = 0.0f; + /* vert[2].position */ ctx->vertices[2][0][0] = width; ctx->vertices[2][0][1] = height; - ctx->vertices[2][1][0] = 1.0f; - ctx->vertices[2][1][1] = 1.0f; + /* vert[3].position */ ctx->vertices[3][0][0] = 0.0f; ctx->vertices[3][0][1] = height; - ctx->vertices[3][1][0] = 0.0f; - ctx->vertices[3][1][1] = 1.0f; + + /* Setup vertex texcoords. This is a little tricky for cube maps. */ + if (tex_target == PIPE_TEXTURE_CUBE) { + static const float st[4][2] = { + {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f} + }; + float rx, ry, rz; + uint i; + + /* loop over quad verts */ + for (i = 0; i < 4; i++) { + /* Compute sc = +/-scale and tc = +/-scale. + * Not +/-1 to avoid cube face selection ambiguity near the edges, + * though that can still sometimes happen with this scale factor... + */ + const float scale = 0.9999f; + const float sc = (2.0f * st[i][0] - 1.0f) * scale; + const float tc = (2.0f * st[i][1] - 1.0f) * scale; + + switch (face) { + case PIPE_TEX_FACE_POS_X: + rx = 1.0f; + ry = -tc; + rz = -sc; + break; + case PIPE_TEX_FACE_NEG_X: + rx = -1.0f; + ry = -tc; + rz = sc; + break; + case PIPE_TEX_FACE_POS_Y: + rx = sc; + ry = 1.0f; + rz = tc; + break; + case PIPE_TEX_FACE_NEG_Y: + rx = sc; + ry = -1.0f; + rz = -tc; + break; + case PIPE_TEX_FACE_POS_Z: + rx = sc; + ry = -tc; + rz = 1.0f; + break; + case PIPE_TEX_FACE_NEG_Z: + rx = -sc; + ry = -tc; + rz = -1.0f; + break; + default: + assert(0); + } + + ctx->vertices[i][1][0] = rx; /*s*/ + ctx->vertices[i][1][1] = ry; /*t*/ + ctx->vertices[i][1][2] = rz; /*r*/ + } + } + else { + /* 1D/2D */ + ctx->vertices[0][1][0] = 0.0f; /*s*/ + ctx->vertices[0][1][1] = 0.0f; /*t*/ + ctx->vertices[0][1][2] = 0.0f; /*r*/ + + ctx->vertices[1][1][0] = 1.0f; + ctx->vertices[1][1][1] = 0.0f; + ctx->vertices[1][1][2] = 0.0f; + + ctx->vertices[2][1][0] = 1.0f; + ctx->vertices[2][1][1] = 1.0f; + ctx->vertices[2][1][2] = 0.0f; + + ctx->vertices[3][1][0] = 0.0f; + ctx->vertices[3][1][1] = 1.0f; + ctx->vertices[3][1][2] = 0.0f; + } offset = get_next_slot( ctx ); @@ -1507,6 +1577,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, /* quad coords in window coords (bypassing vs, clip and viewport) */ offset = set_vertex_data(ctx, + pt->target, + face, (float) pt->width[dstLevel], (float) pt->height[dstLevel]); diff --git a/src/gallium/auxiliary/util/u_linear.c b/src/gallium/auxiliary/util/u_linear.c index 6be365e53b..a1dce3f5cf 100644 --- a/src/gallium/auxiliary/util/u_linear.c +++ b/src/gallium/auxiliary/util/u_linear.c @@ -1,3 +1,34 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Functions for converting tiled data to linear and vice versa. + */ + #include "util/u_debug.h" #include "u_linear.h" diff --git a/src/gallium/auxiliary/util/u_linear.h b/src/gallium/auxiliary/util/u_linear.h index 1589f029bc..b74308ffa3 100644 --- a/src/gallium/auxiliary/util/u_linear.h +++ b/src/gallium/auxiliary/util/u_linear.h @@ -1,3 +1,34 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Functions for converting tiled data to linear and vice versa. + */ + #ifndef U_LINEAR_H #define U_LINEAR_H diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c new file mode 100644 index 0000000000..07d804ecdb --- /dev/null +++ b/src/gallium/auxiliary/util/u_network.c @@ -0,0 +1,188 @@ + +#include "pipe/p_compiler.h" +#include "util/u_network.h" +#include "util/u_debug.h" + +#if defined(PIPE_SUBSYSTEM_WINDOWS_USER) +# include <winsock2.h> +# include <windows.h> +#elif defined(PIPE_OS_LINUX) +# include <sys/socket.h> +# include <netinet/in.h> +# include <unistd.h> +# include <fcntl.h> +# include <netdb.h> +#else +# warning "No socket implementation" +#endif + +boolean +u_socket_init() +{ +#if defined(PIPE_SUBSYSTEM_WINDOWS_USER) + WORD wVersionRequested; + WSADATA wsaData; + int err; + + /* Use the MAKEWORD(lowbyte, highbyte) macro declared in Windef.h */ + wVersionRequested = MAKEWORD(1, 1); + + err = WSAStartup(wVersionRequested, &wsaData); + if (err != 0) { + debug_printf("WSAStartup failed with error: %d\n", err); + return FALSE; + } + return TRUE; +#elif defined(PIPE_HAVE_SOCKETS) + return TRUE; +#else + return FALSE; +#endif +} + +void +u_socket_stop() +{ +#if defined(PIPE_SUBSYSTEM_WINDOWS_USER) + WSACleanup(); +#endif +} + +void +u_socket_close(int s) +{ + if (s < 0) + return; + +#if defined(PIPE_OS_LINUX) + shutdown(s, SHUT_RDWR); + close(s); +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) + shutdown(s, SD_BOTH); + closesocket(s); +#else + assert(0); +#endif +} + +int u_socket_accept(int s) +{ +#if defined(PIPE_HAVE_SOCKETS) + return accept(s, NULL, NULL); +#else + return -1; +#endif +} + +int +u_socket_send(int s, void *data, size_t size) +{ +#if defined(PIPE_HAVE_SOCKETS) + return send(s, data, size, 0); +#else + return -1; +#endif +} + +int +u_socket_peek(int s, void *data, size_t size) +{ +#if defined(PIPE_HAVE_SOCKETS) + return recv(s, data, size, MSG_PEEK); +#else + return -1; +#endif +} + +int +u_socket_recv(int s, void *data, size_t size) +{ +#if defined(PIPE_HAVE_SOCKETS) + return recv(s, data, size, 0); +#else + return -1; +#endif +} + +int +u_socket_connect(const char *hostname, uint16_t port) +{ +#if defined(PIPE_HAVE_SOCKETS) + int s; + struct sockaddr_in sa; + struct hostent *host = NULL; + + memset(&sa, 0, sizeof(struct sockaddr_in)); + host = gethostbyname(hostname); + if (!host) + return -1; + + memcpy((char *)&sa.sin_addr,host->h_addr,host->h_length); + sa.sin_family= host->h_addrtype; + sa.sin_port = htons(port); + + s = socket(host->h_addrtype, SOCK_STREAM, IPPROTO_TCP); + if (s < 0) + return -1; + + if (connect(s, (struct sockaddr *)&sa, sizeof(sa))) { + u_socket_close(s); + return -1; + } + + return s; +#else + assert(0); + return -1; +#endif +} + +int +u_socket_listen_on_port(uint16_t portnum) +{ +#if defined(PIPE_HAVE_SOCKETS) + int s; + struct sockaddr_in sa; + memset(&sa, 0, sizeof(struct sockaddr_in)); + + sa.sin_family = AF_INET; + sa.sin_port = htons(portnum); + + s = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP); + if (s < 0) + return -1; + + if (bind(s, (struct sockaddr *)&sa, sizeof(struct sockaddr_in)) == -1) { + u_socket_close(s); + return -1; + } + + listen(s, 0); + + return s; +#else + assert(0); + return -1; +#endif +} + +void +u_socket_block(int s, boolean block) +{ +#if defined(PIPE_OS_LINUX) + int old = fcntl(s, F_GETFL, 0); + if (old == -1) + return; + + /* TODO obey block */ + if (block) + fcntl(s, F_SETFL, old & ~O_NONBLOCK); + else + fcntl(s, F_SETFL, old | O_NONBLOCK); +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) + u_long iMode = block ? 0 : 1; + ioctlsocket(s, FIONBIO, &iMode); +#else + assert(0); +#endif +} diff --git a/src/gallium/auxiliary/util/u_network.h b/src/gallium/auxiliary/util/u_network.h new file mode 100644 index 0000000000..14d3884427 --- /dev/null +++ b/src/gallium/auxiliary/util/u_network.h @@ -0,0 +1,24 @@ + +#ifndef _U_NETWORK_H_ +#define _U_NETWORK_H_ + +#include "pipe/p_compiler.h" + +#if defined(PIPE_SUBSYSTEM_WINDOWS_USER) +# define PIPE_HAVE_SOCKETS +#elif defined(PIPE_OS_LINUX) +# define PIPE_HAVE_SOCKETS +#endif + +boolean u_socket_init(void); +void u_socket_stop(void); +void u_socket_close(int s); +int u_socket_listen_on_port(uint16_t portnum); +int u_socket_accept(int s); +int u_socket_connect(const char *host, uint16_t port); +int u_socket_send(int s, void *data, size_t size); +int u_socket_peek(int s, void *data, size_t size); +int u_socket_recv(int s, void *data, size_t size); +void u_socket_block(int s, boolean block); + +#endif diff --git a/src/gallium/auxiliary/util/u_surface.c b/src/gallium/auxiliary/util/u_surface.c new file mode 100644 index 0000000000..85e443204e --- /dev/null +++ b/src/gallium/auxiliary/util/u_surface.c @@ -0,0 +1,113 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Surface utility functions. + * + * @author Brian Paul + */ + + +#include "pipe/p_screen.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" + +#include "util/u_surface.h" + + +/** + * Helper to quickly create an RGBA rendering surface of a certain size. + * \param textureOut returns the new texture + * \param surfaceOut returns the new surface + * \return TRUE for success, FALSE if failure + */ +boolean +util_create_rgba_surface(struct pipe_screen *screen, + uint width, uint height, + struct pipe_texture **textureOut, + struct pipe_surface **surfaceOut) +{ + static const enum pipe_format rgbaFormats[] = { + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_NONE + }; + const uint target = PIPE_TEXTURE_2D; + const uint usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + enum pipe_format format = PIPE_FORMAT_NONE; + struct pipe_texture templ; + uint i; + + /* Choose surface format */ + for (i = 0; rgbaFormats[i]; i++) { + if (screen->is_format_supported(screen, rgbaFormats[i], + target, usage, 0)) { + format = rgbaFormats[i]; + break; + } + } + if (format == PIPE_FORMAT_NONE) + return FALSE; /* unable to get an rgba format!?! */ + + /* create texture */ + memset(&templ, 0, sizeof(templ)); + templ.target = target; + templ.format = format; + templ.last_level = 0; + templ.width[0] = width; + templ.height[0] = height; + templ.depth[0] = 1; + pf_get_block(format, &templ.block); + templ.tex_usage = usage; + + *textureOut = screen->texture_create(screen, &templ); + if (!*textureOut) + return FALSE; + + /* create surface / view into texture */ + *surfaceOut = screen->get_tex_surface(screen, *textureOut, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE); + if (!*surfaceOut) { + pipe_texture_reference(textureOut, NULL); + return FALSE; + } + + return TRUE; +} + + +/** + * Release the surface and texture from util_create_rgba_surface(). + */ +void +util_destroy_rgba_surface(struct pipe_texture *texture, + struct pipe_surface *surface) +{ + pipe_surface_reference(&surface, NULL); + pipe_texture_reference(&texture, NULL); +} + diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.h b/src/gallium/auxiliary/util/u_surface.h index a0e4c5d98e..a5b73cfc20 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.h +++ b/src/gallium/auxiliary/util/u_surface.h @@ -1,8 +1,7 @@ /************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,26 +9,44 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ -#ifndef WGL_ARBEXTENSIONSSTRING_H -#define WGL_ARBEXTENSIONSSTRING_H -WINGDIAPI const char * APIENTRY -wglGetExtensionsStringARB( - HDC hdc ); +#ifndef U_SURFACE_H +#define U_SURFACE_H -#endif /* WGL_ARBEXTENSIONSSTRING_H */ + +#include "pipe/p_compiler.h" + + +struct pipe_screen; +struct pipe_texture; +struct pipe_surface; + + +extern boolean +util_create_rgba_surface(struct pipe_screen *screen, + uint width, uint height, + struct pipe_texture **textureOut, + struct pipe_surface **surfaceOut); + + +extern void +util_destroy_rgba_surface(struct pipe_texture *texture, + struct pipe_surface *surface); + + +#endif /* U_SURFACE_H */ diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index f0a5a339eb..9747a55cbf 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -1126,7 +1126,22 @@ pipe_get_tile_z(struct pipe_transfer *pt, for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 24-bit Z to 32-bit Z */ - pDest[j] = (ptrc[j] << 8) | (ptrc[j] & 0xff); + pDest[j] = (ptrc[j] << 8) | ((ptrc[j] >> 16) & 0xff); + } + pDest += dstStride; + ptrc += pt->stride/4; + } + } + break; + case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: + { + const uint *ptrc + = (const uint *)(map + y * pt->stride + x*4); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 24-bit Z to 32-bit Z */ + pDest[j] = (ptrc[j] & 0xffffff00) | ((ptrc[j] >> 24) & 0xff); } pDest += dstStride; ptrc += pt->stride/4; diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 8afe4fccf7..5268cbf79c 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -217,4 +217,9 @@ void util_time_sleep(unsigned usecs) } while(start <= curr && curr < end || end < start && (curr < end || start <= curr)); } +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) +void util_time_sleep(unsigned usecs) +{ + Sleep((usecs + 999)/ 1000); +} #endif diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c index d9c0d7afa8..2eb98068c8 100644 --- a/src/gallium/auxiliary/util/u_upload_mgr.c +++ b/src/gallium/auxiliary/util/u_upload_mgr.c @@ -70,7 +70,7 @@ struct u_upload_mgr *u_upload_create( struct pipe_screen *screen, } -static INLINE void +static INLINE enum pipe_error my_buffer_write(struct pipe_screen *screen, struct pipe_buffer *buf, unsigned offset, unsigned size, unsigned dirty_size, @@ -84,12 +84,14 @@ my_buffer_write(struct pipe_screen *screen, assert(size); map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_WRITE); - assert(map); - if(map) { - memcpy(map + offset, data, size); - pipe_buffer_flush_mapped_range(screen, buf, offset, dirty_size); - pipe_buffer_unmap(screen, buf); - } + if (map == NULL) + return PIPE_ERROR_OUT_OF_MEMORY; + + memcpy(map + offset, data, size); + pipe_buffer_flush_mapped_range(screen, buf, offset, dirty_size); + pipe_buffer_unmap(screen, buf); + + return PIPE_OK; } /* Release old buffer. @@ -162,12 +164,14 @@ enum pipe_error u_upload_data( struct u_upload_mgr *upload, /* Copy the data, using map_range if available: */ - my_buffer_write( upload->screen, - upload->buffer, - upload->offset, - size, - alloc_size, - data ); + ret = my_buffer_write( upload->screen, + upload->buffer, + upload->offset, + size, + alloc_size, + data ); + if (ret) + return ret; /* Emit the return values: */ diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 808be589bd..ebb7a7acc4 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -99,6 +99,28 @@ static const struct debug_named_value cell_debug_flags[] = { {NULL, 0} }; +static unsigned int +cell_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +cell_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} struct pipe_context * cell_create_context(struct pipe_screen *screen, @@ -122,6 +144,9 @@ cell_create_context(struct pipe_screen *screen, cell->pipe.clear = cell_clear; cell->pipe.flush = cell_flush; + cell->pipe.is_texture_referenced = cell_is_texture_referenced; + cell->pipe.is_buffer_referenced = cell_is_buffer_referenced; + #if 0 cell->pipe.begin_query = cell_begin_query; cell->pipe.end_query = cell_end_query; diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c index 7c225e2f27..5ffb7073ab 100644 --- a/src/gallium/drivers/cell/spu/spu_render.c +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -32,6 +32,7 @@ #include "spu_main.h" #include "spu_render.h" +#include "spu_shuffle.h" #include "spu_tri.h" #include "spu_tile.h" #include "cell/common.h" @@ -267,15 +268,75 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) uint drawn = 0; - /* loop over tris */ - for (j = 0; j < render->num_indexes; j += 3) { - const float *v0, *v1, *v2; - - v0 = (const float *) (vertices + indexes[j+0] * vertex_size); - v1 = (const float *) (vertices + indexes[j+1] * vertex_size); - v2 = (const float *) (vertices + indexes[j+2] * vertex_size); - - drawn += tri_draw(v0, v1, v2, tx, ty); + const qword vertex_sizes = (qword)spu_splats(vertex_size); + const qword verticess = (qword)spu_splats((uint)vertices); + + ASSERT_ALIGN16(&indexes[0]); + + const uint num_indexes = render->num_indexes; + + /* loop over tris + * &indexes[0] will be 16 byte aligned. This loop is heavily unrolled + * avoiding variable rotates when extracting vertex indices. + */ + for (j = 0; j < num_indexes; j += 24) { + /* Load three vectors, containing 24 ushort indices */ + const qword* lower_qword = (qword*)&indexes[j]; + const qword indices0 = lower_qword[0]; + const qword indices1 = lower_qword[1]; + const qword indices2 = lower_qword[2]; + + /* stores three indices for each tri n in slots 0, 1 and 2 of vsn */ + /* Straightforward rotates for these */ + qword vs0 = indices0; + qword vs1 = si_shlqbyi(indices0, 6); + qword vs3 = si_shlqbyi(indices1, 2); + qword vs4 = si_shlqbyi(indices1, 8); + qword vs6 = si_shlqbyi(indices2, 4); + qword vs7 = si_shlqbyi(indices2, 10); + + /* For tri 2 and 5, the three indices are split across two machine + * words - rotate and combine */ + const qword tmp2a = si_shlqbyi(indices0, 12); + const qword tmp2b = si_rotqmbyi(indices1, 12|16); + qword vs2 = si_selb(tmp2a, tmp2b, si_fsmh(si_from_uint(0x20))); + + const qword tmp5a = si_shlqbyi(indices1, 14); + const qword tmp5b = si_rotqmbyi(indices2, 14|16); + qword vs5 = si_selb(tmp5a, tmp5b, si_fsmh(si_from_uint(0x60))); + + /* unpack indices from halfword slots to word slots */ + vs0 = si_shufb(vs0, vs0, SHUFB8(0,A,0,B,0,C,0,0)); + vs1 = si_shufb(vs1, vs1, SHUFB8(0,A,0,B,0,C,0,0)); + vs2 = si_shufb(vs2, vs2, SHUFB8(0,A,0,B,0,C,0,0)); + vs3 = si_shufb(vs3, vs3, SHUFB8(0,A,0,B,0,C,0,0)); + vs4 = si_shufb(vs4, vs4, SHUFB8(0,A,0,B,0,C,0,0)); + vs5 = si_shufb(vs5, vs5, SHUFB8(0,A,0,B,0,C,0,0)); + vs6 = si_shufb(vs6, vs6, SHUFB8(0,A,0,B,0,C,0,0)); + vs7 = si_shufb(vs7, vs7, SHUFB8(0,A,0,B,0,C,0,0)); + + /* Calculate address of vertex in vertices[] */ + vs0 = si_mpya(vs0, vertex_sizes, verticess); + vs1 = si_mpya(vs1, vertex_sizes, verticess); + vs2 = si_mpya(vs2, vertex_sizes, verticess); + vs3 = si_mpya(vs3, vertex_sizes, verticess); + vs4 = si_mpya(vs4, vertex_sizes, verticess); + vs5 = si_mpya(vs5, vertex_sizes, verticess); + vs6 = si_mpya(vs6, vertex_sizes, verticess); + vs7 = si_mpya(vs7, vertex_sizes, verticess); + + /* Select the appropriate call based on the number of vertices + * remaining */ + switch(num_indexes - j) { + default: drawn += tri_draw(vs7, tx, ty); + case 21: drawn += tri_draw(vs6, tx, ty); + case 18: drawn += tri_draw(vs5, tx, ty); + case 15: drawn += tri_draw(vs4, tx, ty); + case 12: drawn += tri_draw(vs3, tx, ty); + case 9: drawn += tri_draw(vs2, tx, ty); + case 6: drawn += tri_draw(vs1, tx, ty); + case 3: drawn += tri_draw(vs0, tx, ty); + } } //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3); diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index d727268475..58be001be4 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -133,7 +133,15 @@ struct setup_stage { uint tx, ty; /**< position of current tile (x, y) */ - int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; + union { + struct { + int cliprect_minx; + int cliprect_miny; + int cliprect_maxx; + int cliprect_maxy; + }; + qword cliprect; + }; struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; @@ -432,6 +440,41 @@ print_vertex(const struct vertex_header *v) } #endif +/* Returns the minimum of each slot of two vec_float4s as qwords. + * i.e. return[n] = min(q0[n],q1[n]); + */ +static qword +minfq(qword q0, qword q1) +{ + const qword q0q1m = si_fcgt(q0, q1); + return si_selb(q0, q1, q0q1m); +} + +/* Returns the minimum of each slot of three vec_float4s as qwords. + * i.e. return[n] = min(q0[n],q1[n],q2[n]); + */ +static qword +min3fq(qword q0, qword q1, qword q2) +{ + return minfq(minfq(q0, q1), q2); +} + +/* Returns the maximum of each slot of two vec_float4s as qwords. + * i.e. return[n] = min(q0[n],q1[n],q2[n]); + */ +static qword +maxfq(qword q0, qword q1) { + const qword q0q1m = si_fcgt(q0, q1); + return si_selb(q1, q0, q0q1m); +} + +/* Returns the maximum of each slot of three vec_float4s as qwords. + * i.e. return[n] = min(q0[n],q1[n],q2[n]); + */ +static qword +max3fq(qword q0, qword q1, qword q2) { + return maxfq(maxfq(q0, q1), q2); +} /** * Sort vertices from top to bottom. @@ -440,9 +483,7 @@ print_vertex(const struct vertex_header *v) * \return FALSE if tri is totally outside tile, TRUE otherwise */ static boolean -setup_sort_vertices(const struct vertex_header *v0, - const struct vertex_header *v1, - const struct vertex_header *v2) +setup_sort_vertices(const qword vs) { float area, sign; @@ -455,57 +496,57 @@ setup_sort_vertices(const struct vertex_header *v0, } #endif - /* determine bottom to top order of vertices */ { + /* Load the float values for various processing... */ + const qword f0 = (qword)(((const struct vertex_header*)si_to_ptr(vs))->data[0]); + const qword f1 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 4)))->data[0]); + const qword f2 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 8)))->data[0]); + + /* Check if triangle is completely outside the tile bounds + * Find the min and max x and y positions of the three poits */ + const qword minf = min3fq(f0, f1, f2); + const qword maxf = max3fq(f0, f1, f2); + + /* Compare min and max against cliprect vals */ + const qword maxsmins = si_shufb(maxf, minf, SHUFB4(A,B,a,b)); + const qword outside = si_fcgt(maxsmins, si_csflt(setup.cliprect, 0)); + + /* Use a little magic to work out of the tri is visible or not */ + if(si_to_uint(si_xori(si_gb(outside), 0xc))) return FALSE; + + /* determine bottom to top order of vertices */ /* A table of shuffle patterns for putting vertex_header pointers into correct order. Quite magical. */ - const vec_uchar16 sort_order_patterns[] = { - SHUFFLE4(A,B,C,C), - SHUFFLE4(C,A,B,C), - SHUFFLE4(A,C,B,C), - SHUFFLE4(B,C,A,C), - SHUFFLE4(B,A,C,C), - SHUFFLE4(C,B,A,C) }; - - /* The vertex_header pointers, packed for easy shuffling later */ - const vec_uint4 vs = {(unsigned)v0, (unsigned)v1, (unsigned)v2}; + const qword sort_order_patterns[] = { + SHUFB4(A,B,C,C), + SHUFB4(C,A,B,C), + SHUFB4(A,C,B,C), + SHUFB4(B,C,A,C), + SHUFB4(B,A,C,C), + SHUFB4(C,B,A,C) }; /* Collate y values into two vectors for comparison. Using only one shuffle constant! ;) */ - const vec_float4 y_02_ = spu_shuffle(v0->data[0], v2->data[0], SHUFFLE4(0,B,b,C)); - const vec_float4 y_10_ = spu_shuffle(v1->data[0], v0->data[0], SHUFFLE4(0,B,b,C)); - const vec_float4 y_012 = spu_shuffle(y_02_, v1->data[0], SHUFFLE4(0,B,b,C)); - const vec_float4 y_120 = spu_shuffle(y_10_, v2->data[0], SHUFFLE4(0,B,b,C)); + const qword y_02_ = si_shufb(f0, f2, SHUFB4(0,B,b,C)); + const qword y_10_ = si_shufb(f1, f0, SHUFB4(0,B,b,C)); + const qword y_012 = si_shufb(y_02_, f1, SHUFB4(0,B,b,C)); + const qword y_120 = si_shufb(y_10_, f2, SHUFB4(0,B,b,C)); /* Perform comparison: {y0,y1,y2} > {y1,y2,y0} */ - const vec_uint4 compare = spu_cmpgt(y_012, y_120); + const qword compare = si_fcgt(y_012, y_120); /* Compress the result of the comparison into 4 bits */ - const vec_uint4 gather = spu_gather(compare); + const qword gather = si_gb(compare); /* Subtract one to attain the index into the LUT. Magical. */ - const unsigned int index = spu_extract(gather, 0) - 1; + const unsigned int index = si_to_uint(gather) - 1; /* Load the appropriate pattern and construct the desired vector. */ - setup.vertex_headers = (qword)spu_shuffle(vs, vs, sort_order_patterns[index]); + setup.vertex_headers = si_shufb(vs, vs, sort_order_patterns[index]); /* Using the result of the comparison, set sign. Very magical. */ - sign = ((si_to_uint(si_cntb((qword)gather)) == 2) ? 1.0f : -1.0f); + sign = ((si_to_uint(si_cntb(gather)) == 2) ? 1.0f : -1.0f); } - /* Check if triangle is completely outside the tile bounds */ - if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy) - return FALSE; - if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny) - return FALSE; - if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx && - spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx && - spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx) - return FALSE; - if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx && - spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx && - spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx) - return FALSE; - setup.ebot.ds = spu_sub(setup.vmid->data[0], setup.vmin->data[0]); setup.emaj.ds = spu_sub(setup.vmax->data[0], setup.vmin->data[0]); setup.etop.ds = spu_sub(setup.vmax->data[0], setup.vmid->data[0]); @@ -761,21 +802,19 @@ subtriangle(struct edge *eleft, struct edge *eright, unsigned lines) * The tile data should have already been fetched. */ boolean -tri_draw(const float *v0, const float *v1, const float *v2, +tri_draw(const qword vs, uint tx, uint ty) { setup.tx = tx; setup.ty = ty; /* set clipping bounds to tile bounds */ - setup.cliprect_minx = tx * TILE_SIZE; - setup.cliprect_miny = ty * TILE_SIZE; - setup.cliprect_maxx = (tx + 1) * TILE_SIZE; - setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - - if (!setup_sort_vertices((struct vertex_header *) v0, - (struct vertex_header *) v1, - (struct vertex_header *) v2)) { + const qword clipbase = (qword)((vec_uint4){tx, ty}); + const qword clipmin = si_mpyui(clipbase, TILE_SIZE); + const qword clipmax = si_ai(clipmin, TILE_SIZE); + setup.cliprect = si_shufb(clipmin, clipmax, SHUFB4(A,B,a,b)); + + if(!setup_sort_vertices(vs)) { return FALSE; /* totally clipped */ } diff --git a/src/gallium/drivers/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h index aa694dd7c9..82e3b19ad7 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.h +++ b/src/gallium/drivers/cell/spu/spu_tri.h @@ -31,7 +31,7 @@ extern boolean -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); +tri_draw(const qword vs, uint tx, uint ty); #endif /* SPU_TRI_H */ diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index fcad717cf8..37184eac7b 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -105,7 +105,28 @@ static boolean failover_draw_arrays( struct pipe_context *pipe, return failover_draw_elements(pipe, NULL, 0, prim, start, count); } +static unsigned int +failover_is_texture_referenced( struct pipe_context *_pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + struct failover_context *failover = failover_context( _pipe ); + struct pipe_context *pipe = (failover->mode == FO_HW) ? + failover->hw : failover->sw; + + return pipe->is_texture_referenced(pipe, texture, face, level); +} +static unsigned int +failover_is_buffer_referenced( struct pipe_context *_pipe, + struct pipe_buffer *buf) +{ + struct failover_context *failover = failover_context( _pipe ); + struct pipe_context *pipe = (failover->mode == FO_HW) ? + failover->hw : failover->sw; + + return pipe->is_buffer_referenced(pipe, buf); +} struct pipe_context *failover_create( struct pipe_context *hw, struct pipe_context *sw ) @@ -151,6 +172,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, #endif failover->pipe.flush = hw->flush; + failover->pipe.is_texture_referenced = failover_is_texture_referenced; + failover->pipe.is_buffer_referenced = failover_is_buffer_referenced; failover->dirty = 0; diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 12821c5a76..8870b39866 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -8,7 +8,6 @@ C_SOURCES = \ i915_clear.c \ i915_flush.c \ i915_context.c \ - i915_context.c \ i915_debug.c \ i915_debug_fp.c \ i915_state.c \ diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 3e3a596884..ccf9bb31fb 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -136,6 +136,29 @@ static boolean i915_draw_arrays( struct pipe_context *pipe, } +static unsigned int +i915_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +i915_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context *i915_create_context( struct pipe_screen *screen, struct pipe_winsys *pipe_winsys, @@ -160,6 +183,9 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, i915->pipe.draw_elements = i915_draw_elements; i915->pipe.draw_range_elements = i915_draw_range_elements; + i915->pipe.is_texture_referenced = i915_is_texture_referenced; + i915->pipe.is_buffer_referenced = i915_is_buffer_referenced; + /* * Create drawing context and plug our rendering stage into it. */ diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index c74cbf8d73..9b33285bc7 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -73,6 +73,28 @@ static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, pipe->surface_fill(pipe, ps, x, y, w, h, clearValue); } +static unsigned int +brw_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +brw_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} struct pipe_context *brw_create(struct pipe_screen *screen, struct brw_winsys *brw_winsys, @@ -94,6 +116,9 @@ struct pipe_context *brw_create(struct pipe_screen *screen, brw->pipe.destroy = brw_destroy; brw->pipe.clear = brw_clear; + brw->pipe.is_texture_referenced = brw_is_texture_referenced; + brw->pipe.is_buffer_referenced = brw_is_buffer_referenced; + brw_init_surface_functions(brw); brw_init_texture_functions(brw); brw_init_state_functions(brw); diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index f44bd17451..8aea8c0558 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -240,7 +240,7 @@ static boolean brw_miptree_layout(struct brw_texture *tex) nblocksx = pf_get_nblocksx(&pt->block, width); nblocksy = pf_get_nblocksy(&pt->block, height); - if (pt->compressed) { + if (pf_is_compressed(pt->format)) { pack_y_pitch = (height + 3) / 4; if (pack_x_pitch > align(width, align_w)) { diff --git a/src/gallium/drivers/nouveau/Makefile b/src/gallium/drivers/nouveau/Makefile new file mode 100644 index 0000000000..dbe8a6e7bf --- /dev/null +++ b/src/gallium/drivers/nouveau/Makefile @@ -0,0 +1,8 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = nouveau + +C_SOURCES = nouveau_screen.c + +include ../../Makefile.template diff --git a/src/gallium/drivers/nouveau/nouveau_push.h b/src/gallium/drivers/nouveau/nouveau_push.h index 54ef1c1291..9c235080a5 100644 --- a/src/gallium/drivers/nouveau/nouveau_push.h +++ b/src/gallium/drivers/nouveau/nouveau_push.h @@ -9,13 +9,13 @@ #define OUT_RING(data) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - (*pc->nvws->channel->pushbuf->cur++) = (data); \ + (*pc->base.channel->pushbuf->cur++) = (data); \ } while(0) #define OUT_RINGp(src,size) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - memcpy(pc->nvws->channel->pushbuf->cur, (src), (size) * 4); \ - pc->nvws->channel->pushbuf->cur += (size); \ + memcpy(pc->base.channel->pushbuf->cur, (src), (size) * 4); \ + pc->base.channel->pushbuf->cur += (size); \ } while(0) #define OUT_RINGf(data) do { \ @@ -26,25 +26,35 @@ #define BEGIN_RING(obj,mthd,size) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - if (pc->nvws->channel->pushbuf->remaining < ((size) + 1)) \ - pc->nvws->push_flush(pc->nvws, ((size) + 1), NULL); \ + struct nouveau_channel *chan = pc->base.channel; \ + if (chan->pushbuf->remaining < ((size) + 1)) \ + nouveau_pushbuf_flush(chan, ((size) + 1)); \ OUT_RING((pc->obj->subc << 13) | ((size) << 18) | (mthd)); \ - pc->nvws->channel->pushbuf->remaining -= ((size) + 1); \ + chan->pushbuf->remaining -= ((size) + 1); \ } while(0) #define BEGIN_RING_NI(obj,mthd,size) do { \ BEGIN_RING(obj, (mthd) | 0x40000000, (size)); \ } while(0) +static inline void +DO_FIRE_RING(struct nouveau_channel *chan, struct pipe_fence_handle **fence) +{ + nouveau_pushbuf_flush(chan, 0); + if (fence) + *fence = NULL; +} + #define FIRE_RING(fence) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - pc->nvws->push_flush(pc->nvws, 0, fence); \ + DO_FIRE_RING(pc->base.channel, fence); \ } while(0) #define OUT_RELOC(bo,data,flags,vor,tor) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - pc->nvws->push_reloc(pc->nvws, pc->nvws->channel->pushbuf->cur++, \ - (bo), (data), (flags), (vor), (tor)); \ + struct nouveau_channel *chan = pc->base.channel; \ + nouveau_pushbuf_emit_reloc(chan, chan->pushbuf->cur++, nouveau_bo(bo), \ + (data), 0, (flags), (vor), (tor)); \ } while(0) /* Raw data + flags depending on FB/TT buffer */ @@ -55,8 +65,8 @@ /* FB/TT object handle */ #define OUT_RELOCo(bo,flags) do { \ OUT_RELOC((bo), 0, (flags) | NOUVEAU_BO_OR, \ - pc->nvws->channel->vram->handle, \ - pc->nvws->channel->gart->handle); \ + pc->base.channel->vram->handle, \ + pc->base.channel->gart->handle); \ } while(0) /* Low 32-bits of offset */ @@ -72,11 +82,12 @@ /* A reloc which'll recombine into a NV_DMA_METHOD packet header */ #define OUT_RELOCm(bo, flags, obj, mthd, size) do { \ NOUVEAU_PUSH_CONTEXT(pc); \ - if (pc->nvws->channel->pushbuf->remaining < ((size) + 1)) \ - pc->nvws->push_flush(pc->nvws->channel, ((size) + 1), NULL); \ + struct nouveau_channel *chan = pc->base.channel; \ + if (chan->pushbuf->remaining < ((size) + 1)) \ + nouveau_pushbuf_flush(chan, ((size) + 1)); \ OUT_RELOCd((bo), (pc->obj->subc << 13) | ((size) << 18) | (mthd), \ (flags), 0, 0); \ - pc->nvws->channel->pushbuf->remaining -= ((size) + 1); \ + chan->pushbuf->remaining -= ((size) + 1); \ } while(0) #endif diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c new file mode 100644 index 0000000000..832366e646 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_screen.c @@ -0,0 +1,240 @@ +#include <pipe/p_defines.h> +#include <pipe/p_screen.h> +#include <pipe/p_state.h> + +#include <util/u_memory.h> + +#include "nouveau/nouveau_bo.h" +#include "nouveau_winsys.h" +#include "nouveau_screen.h" + +static const char * +nouveau_screen_get_name(struct pipe_screen *pscreen) +{ + struct nouveau_device *dev = nouveau_screen(pscreen)->device; + static char buffer[128]; + + snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); + return buffer; +} + +static const char * +nouveau_screen_get_vendor(struct pipe_screen *pscreen) +{ + return "nouveau"; +} + +static struct pipe_buffer * +nouveau_screen_bo_skel(struct pipe_screen *pscreen, struct nouveau_bo *bo, + unsigned alignment, unsigned usage, unsigned size) +{ + struct pipe_buffer *pb; + + pb = CALLOC(1, sizeof(struct pipe_buffer)+sizeof(struct nouveau_bo *)); + if (!pb) { + nouveau_bo_ref(NULL, &bo); + return NULL; + } + + pipe_reference_init(&pb->reference, 1); + pb->screen = pscreen; + pb->alignment = alignment; + pb->usage = usage; + pb->size = size; + *(struct nouveau_bo **)(pb + 1) = bo; + return pb; +} + +static struct pipe_buffer * +nouveau_screen_bo_new(struct pipe_screen *pscreen, unsigned alignment, + unsigned usage, unsigned size) +{ + struct nouveau_device *dev = nouveau_screen(pscreen)->device; + struct nouveau_bo *bo = NULL; + uint32_t flags = NOUVEAU_BO_MAP; + int ret; + + if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER) + flags |= NOUVEAU_BO_GART; + else + if (usage & PIPE_BUFFER_USAGE_VERTEX) { + if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF)) + flags |= NOUVEAU_BO_GART; + } else + if (usage & PIPE_BUFFER_USAGE_INDEX) { + if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF)) + flags |= NOUVEAU_BO_GART; + } + + if (usage & PIPE_BUFFER_USAGE_PIXEL) { + if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE) + flags |= NOUVEAU_BO_GART; + if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE)) + flags |= NOUVEAU_BO_VRAM; + + if (dev->chipset == 0x50 || dev->chipset >= 0x80) { + flags |= NOUVEAU_BO_TILED; + if (usage & NOUVEAU_BUFFER_USAGE_ZETA) + flags |= NOUVEAU_BO_ZTILE; + } + } + + ret = nouveau_bo_new(dev, flags, alignment, size, &bo); + if (ret) + return NULL; + + return nouveau_screen_bo_skel(pscreen, bo, alignment, usage, size); +} + +static struct pipe_buffer * +nouveau_screen_bo_user(struct pipe_screen *pscreen, void *ptr, unsigned bytes) +{ + struct nouveau_device *dev = nouveau_screen(pscreen)->device; + struct nouveau_bo *bo = NULL; + int ret; + + ret = nouveau_bo_user(dev, ptr, bytes, &bo); + if (ret) + return NULL; + + return nouveau_screen_bo_skel(pscreen, bo, 0, 0, bytes); +} + +static inline uint32_t +nouveau_screen_map_flags(unsigned pipe) +{ + uint32_t flags = 0; + + if (pipe & PIPE_BUFFER_USAGE_CPU_READ) + flags |= NOUVEAU_BO_RD; + if (pipe & PIPE_BUFFER_USAGE_CPU_WRITE) + flags |= NOUVEAU_BO_WR; + if (pipe & PIPE_BUFFER_USAGE_DISCARD) + flags |= NOUVEAU_BO_INVAL; + if (pipe & PIPE_BUFFER_USAGE_DONTBLOCK) + flags |= NOUVEAU_BO_NOWAIT; + else + if (pipe & 0 /*PIPE_BUFFER_USAGE_UNSYNCHRONIZED*/) + flags |= NOUVEAU_BO_NOSYNC; + + return flags; +} + +static void * +nouveau_screen_bo_map(struct pipe_screen *pscreen, struct pipe_buffer *pb, + unsigned usage) +{ + struct nouveau_bo *bo = nouveau_bo(pb); + int ret; + + ret = nouveau_bo_map(bo, nouveau_screen_map_flags(usage)); + if (ret) { + debug_printf("map failed: %d\n", ret); + return NULL; + } + + return bo->map; +} + +static void * +nouveau_screen_bo_map_range(struct pipe_screen *pscreen, struct pipe_buffer *pb, + unsigned offset, unsigned length, unsigned usage) +{ + struct nouveau_bo *bo = nouveau_bo(pb); + int ret; + + ret = nouveau_bo_map_range(bo, offset, length, + nouveau_screen_map_flags(usage)); + if (ret) { + debug_printf("map_range failed: %d\n", ret); + return NULL; + } + + return (char *)bo->map - offset; /* why gallium? why? */ +} + +static void +nouveau_screen_bo_map_flush(struct pipe_screen *pscreen, struct pipe_buffer *pb, + unsigned offset, unsigned length) +{ + struct nouveau_bo *bo = nouveau_bo(pb); + + nouveau_bo_map_flush(bo, offset, length); +} + +static void +nouveau_screen_bo_unmap(struct pipe_screen *pscreen, struct pipe_buffer *pb) +{ + struct nouveau_bo *bo = nouveau_bo(pb); + + nouveau_bo_unmap(bo); +} + +static void +nouveau_screen_bo_del(struct pipe_buffer *pb) +{ + struct nouveau_bo *bo = nouveau_bo(pb); + + nouveau_bo_ref(NULL, &bo); + FREE(pb); +} + +static void +nouveau_screen_fence_ref(struct pipe_screen *pscreen, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *pfence) +{ + *ptr = pfence; +} + +static int +nouveau_screen_fence_signalled(struct pipe_screen *screen, + struct pipe_fence_handle *pfence, + unsigned flags) +{ + return 0; +} + +static int +nouveau_screen_fence_finish(struct pipe_screen *screen, + struct pipe_fence_handle *pfence, + unsigned flags) +{ + return 0; +} + +int +nouveau_screen_init(struct nouveau_screen *screen, struct nouveau_device *dev) +{ + struct pipe_screen *pscreen = &screen->base; + int ret; + + ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202, + &screen->channel); + if (ret) + return ret; + screen->device = dev; + + pscreen->get_name = nouveau_screen_get_name; + pscreen->get_vendor = nouveau_screen_get_vendor; + + pscreen->buffer_create = nouveau_screen_bo_new; + pscreen->user_buffer_create = nouveau_screen_bo_user; + pscreen->buffer_map = nouveau_screen_bo_map; + pscreen->buffer_map_range = nouveau_screen_bo_map_range; + pscreen->buffer_flush_mapped_range = nouveau_screen_bo_map_flush; + pscreen->buffer_unmap = nouveau_screen_bo_unmap; + pscreen->buffer_destroy = nouveau_screen_bo_del; + + pscreen->fence_reference = nouveau_screen_fence_ref; + pscreen->fence_signalled = nouveau_screen_fence_signalled; + pscreen->fence_finish = nouveau_screen_fence_finish; + + return 0; +} + +void +nouveau_screen_fini(struct nouveau_screen *screen) +{ +} + diff --git a/src/gallium/drivers/nouveau/nouveau_screen.h b/src/gallium/drivers/nouveau/nouveau_screen.h new file mode 100644 index 0000000000..9968b07896 --- /dev/null +++ b/src/gallium/drivers/nouveau/nouveau_screen.h @@ -0,0 +1,25 @@ +#ifndef __NOUVEAU_SCREEN_H__ +#define __NOUVEAU_SCREEN_H__ + +struct nouveau_screen { + struct pipe_screen base; + struct nouveau_device *device; + struct nouveau_channel *channel; +}; + +static inline struct nouveau_screen * +nouveau_screen(struct pipe_screen *pscreen) +{ + return (struct nouveau_screen *)pscreen; +} + +static inline struct nouveau_bo * +nouveau_bo(struct pipe_buffer *pb) +{ + return pb ? *(struct nouveau_bo **)(pb + 1) : NULL; +} + +int nouveau_screen_init(struct nouveau_screen *, struct nouveau_device *); +void nouveau_screen_fini(struct nouveau_screen *); + +#endif diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h index a54820e851..b595405357 100644 --- a/src/gallium/drivers/nouveau/nouveau_stateobj.h +++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h @@ -4,7 +4,7 @@ #include "util/u_debug.h" struct nouveau_stateobj_reloc { - struct pipe_buffer *bo; + struct nouveau_bo *bo; unsigned offset; unsigned packet; @@ -51,7 +51,7 @@ so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso) if (pipe_reference((struct pipe_reference**)pso, &ref->reference)) { free(so->push); for (i = 0; i < so->cur_reloc; i++) - pipe_buffer_reference(&so->reloc[i].bo, NULL); + nouveau_bo_ref(NULL, &so->reloc[i].bo); free(so->reloc); free(so); } @@ -81,13 +81,13 @@ so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr, } static INLINE void -so_reloc(struct nouveau_stateobj *so, struct pipe_buffer *bo, +so_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo, unsigned data, unsigned flags, unsigned vor, unsigned tor) { struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++]; r->bo = NULL; - pipe_buffer_reference(&r->bo, bo); + nouveau_bo_ref(bo, &r->bo); r->offset = so->cur - so->push; r->packet = so->cur_packet; r->data = data; @@ -107,50 +107,52 @@ so_dump(struct nouveau_stateobj *so) } static INLINE void -so_emit(struct nouveau_winsys *nvws, struct nouveau_stateobj *so) +so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so) { - struct nouveau_pushbuf *pb = nvws->channel->pushbuf; + struct nouveau_pushbuf *pb = chan->pushbuf; unsigned nr, i; nr = so->cur - so->push; if (pb->remaining < nr) - nvws->push_flush(nvws, nr, NULL); + nouveau_pushbuf_flush(chan, nr); pb->remaining -= nr; memcpy(pb->cur, so->push, nr * 4); for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - nvws->push_reloc(nvws, pb->cur + r->offset, r->bo, - r->data, r->flags, r->vor, r->tor); + nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset, + r->bo, r->data, 0, r->flags, + r->vor, r->tor); } pb->cur += nr; } static INLINE void -so_emit_reloc_markers(struct nouveau_winsys *nvws, struct nouveau_stateobj *so) +so_emit_reloc_markers(struct nouveau_channel *chan, struct nouveau_stateobj *so) { - struct nouveau_pushbuf *pb = nvws->channel->pushbuf; + struct nouveau_pushbuf *pb = chan->pushbuf; unsigned i; if (!so) return; i = so->cur_reloc << 1; - if (nvws->channel->pushbuf->remaining < i) - nvws->push_flush(nvws, i, NULL); - nvws->channel->pushbuf->remaining -= i; + if (pb->remaining < i) + nouveau_pushbuf_flush(chan, i); + pb->remaining -= i; for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - nvws->push_reloc(nvws, pb->cur++, r->bo, r->packet, - (r->flags & (NOUVEAU_BO_VRAM | - NOUVEAU_BO_GART | - NOUVEAU_BO_RDWR)) | - NOUVEAU_BO_DUMMY, 0, 0); - nvws->push_reloc(nvws, pb->cur++, r->bo, r->data, - r->flags | NOUVEAU_BO_DUMMY, r->vor, r->tor); + nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->packet, 0, + (r->flags & (NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART | + NOUVEAU_BO_RDWR)) | + NOUVEAU_BO_DUMMY, 0, 0); + nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->data, 0, + r->flags | NOUVEAU_BO_DUMMY, + r->vor, r->tor); } } diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h index ff7dd1c51c..42c77e5e77 100644 --- a/src/gallium/drivers/nouveau/nouveau_winsys.h +++ b/src/gallium/drivers/nouveau/nouveau_winsys.h @@ -23,77 +23,38 @@ #define NOUVEAU_BUFFER_USAGE_ZETA (1 << 17) #define NOUVEAU_BUFFER_USAGE_TRANSFER (1 << 18) -struct nouveau_winsys { - struct pipe_winsys *ws; - - struct nouveau_channel *channel; - - int (*res_init)(struct nouveau_resource **heap, unsigned start, - unsigned size); - int (*res_alloc)(struct nouveau_resource *heap, int size, void *priv, - struct nouveau_resource **); - void (*res_free)(struct nouveau_resource **); - - int (*push_reloc)(struct nouveau_winsys *, void *ptr, - struct pipe_buffer *, uint32_t data, - uint32_t flags, uint32_t vor, uint32_t tor); - int (*push_flush)(struct nouveau_winsys *, unsigned size, - struct pipe_fence_handle **fence); - - int (*grobj_alloc)(struct nouveau_winsys *, int grclass, - struct nouveau_grobj **); - void (*grobj_free)(struct nouveau_grobj **); - - int (*notifier_alloc)(struct nouveau_winsys *, int count, - struct nouveau_notifier **); - void (*notifier_free)(struct nouveau_notifier **); - void (*notifier_reset)(struct nouveau_notifier *, int id); - uint32_t (*notifier_status)(struct nouveau_notifier *, int id); - uint32_t (*notifier_retval)(struct nouveau_notifier *, int id); - int (*notifier_wait)(struct nouveau_notifier *, int id, - int status, double timeout); - - int (*surface_copy)(struct nouveau_winsys *, struct pipe_surface *, - unsigned, unsigned, struct pipe_surface *, - unsigned, unsigned, unsigned, unsigned); - int (*surface_fill)(struct nouveau_winsys *, struct pipe_surface *, - unsigned, unsigned, unsigned, unsigned, unsigned); - - struct nouveau_bo *(*get_bo)(struct pipe_buffer *); -}; - extern struct pipe_screen * -nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv04_create(struct pipe_screen *, unsigned pctx_id); extern struct pipe_screen * -nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv10_create(struct pipe_screen *, unsigned pctx_id); extern struct pipe_screen * -nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv20_create(struct pipe_screen *, unsigned pctx_id); extern struct pipe_screen * -nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv30_create(struct pipe_screen *, unsigned pctx_id); extern struct pipe_screen * -nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv40_create(struct pipe_screen *, unsigned pctx_id); extern struct pipe_screen * -nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *); +nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *); extern struct pipe_context * nv50_create(struct pipe_screen *, unsigned pctx_id); diff --git a/src/gallium/drivers/nv04/nv04_context.c b/src/gallium/drivers/nv04/nv04_context.c index d6710cd892..17166c9f51 100644 --- a/src/gallium/drivers/nv04/nv04_context.c +++ b/src/gallium/drivers/nv04/nv04_context.c @@ -64,6 +64,30 @@ nv04_init_hwctx(struct nv04_context *nv04) return TRUE; } +static unsigned int +nv04_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv04_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + + struct pipe_context * nv04_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -89,6 +113,9 @@ nv04_create(struct pipe_screen *pscreen, unsigned pctx_id) nv04->pipe.clear = nv04_clear; nv04->pipe.flush = nv04_flush; + nv04->pipe.is_texture_referenced = nv04_is_texture_referenced; + nv04->pipe.is_buffer_referenced = nv04_is_buffer_referenced; + nv04_init_surface_functions(nv04); nv04_init_state_functions(nv04); diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c index 4da833c25e..93f752faec 100644 --- a/src/gallium/drivers/nv04/nv04_miptree.c +++ b/src/gallium/drivers/nv04/nv04_miptree.c @@ -31,7 +31,8 @@ nv04_miptree_layout(struct nv04_miptree *nv04mt) for (l = 0; l <= pt->last_level; l++) { - nv04mt->level[l].image_offset = offset; + nv04mt->level[l].image_offset = + CALLOC(nr_faces, sizeof(unsigned)); offset += nv04mt->level[l].pitch * pt->height[l]; } diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c index f9f6d97426..4bbedfb4d6 100644 --- a/src/gallium/drivers/nv04/nv04_screen.c +++ b/src/gallium/drivers/nv04/nv04_screen.c @@ -1,27 +1,9 @@ #include "pipe/p_screen.h" #include "pipe/p_inlines.h" -#include "util/u_simple_screen.h" #include "nv04_context.h" #include "nv04_screen.h" -static const char * -nv04_screen_get_name(struct pipe_screen *screen) -{ - struct nv04_screen *nv04screen = nv04_screen(screen); - struct nouveau_device *dev = nv04screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv04_screen_get_vendor(struct pipe_screen *screen) -{ - return "nouveau"; -} - static int nv04_screen_get_param(struct pipe_screen *screen, int param) { @@ -123,10 +105,9 @@ static void nv04_screen_destroy(struct pipe_screen *pscreen) { struct nv04_screen *screen = nv04_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - nvws->notifier_free(&screen->sync); - nvws->grobj_free(&screen->fahrenheit); + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->fahrenheit); nv04_surface_2d_takedown(&screen->eng2d); FREE(pscreen); @@ -141,21 +122,38 @@ nv04_surface_buffer(struct pipe_surface *surf) } struct pipe_screen * -nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv04_screen *screen = CALLOC_STRUCT(nv04_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; unsigned fahrenheit_class = 0, sub3d_class = 0; - unsigned chipset = nvws->channel->device->chipset; int ret; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; + + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv04_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv04_screen_destroy; + pscreen->get_param = nv04_screen_get_param; + pscreen->get_paramf = nv04_screen_get_paramf; + pscreen->is_format_supported = nv04_screen_is_format_supported; - if (chipset>=0x20) { + nv04_screen_init_miptree_functions(pscreen); + nv04_screen_init_transfer_functions(pscreen); + + if (dev->chipset >= 0x20) { fahrenheit_class = 0; sub3d_class = 0; - } else if (chipset>=0x10) { + } else if (dev->chipset >= 0x10) { fahrenheit_class = NV10_DX5_TEXTURED_TRIANGLE; sub3d_class = NV10_CONTEXT_SURFACES_3D; } else { @@ -164,50 +162,40 @@ nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) } if (!fahrenheit_class) { - NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", chipset); + NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", dev->chipset); return NULL; } - /* 2D engine setup */ - screen->eng2d = nv04_surface_2d_init(nvws); - screen->eng2d->buf = nv04_surface_buffer; - /* 3D object */ - ret = nvws->grobj_alloc(nvws, fahrenheit_class, &screen->fahrenheit); + ret = nouveau_grobj_alloc(chan, 0xbeef0001, fahrenheit_class, + &screen->fahrenheit); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return NULL; } + BIND_RING(chan, screen->fahrenheit, 7); /* 3D surface object */ - ret = nvws->grobj_alloc(nvws, sub3d_class, &screen->context_surfaces_3d); + ret = nouveau_grobj_alloc(chan, 0xbeef0002, sub3d_class, + &screen->context_surfaces_3d); if (ret) { NOUVEAU_ERR("Error creating 3D surface object: %d\n", ret); return NULL; } + BIND_RING(chan, screen->context_surfaces_3d, 6); + + /* 2D engine setup */ + screen->eng2d = nv04_surface_2d_init(&screen->base); + screen->eng2d->buf = nv04_surface_buffer; /* Notifier for sync purposes */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv04_screen_destroy(&screen->pipe); + nv04_screen_destroy(pscreen); return NULL; } - screen->pipe.winsys = ws; - screen->pipe.destroy = nv04_screen_destroy; - - screen->pipe.get_name = nv04_screen_get_name; - screen->pipe.get_vendor = nv04_screen_get_vendor; - screen->pipe.get_param = nv04_screen_get_param; - screen->pipe.get_paramf = nv04_screen_get_paramf; - - screen->pipe.is_format_supported = nv04_screen_is_format_supported; - - nv04_screen_init_miptree_functions(&screen->pipe); - nv04_screen_init_transfer_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); - - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h index ee6fb6db44..11466b9442 100644 --- a/src/gallium/drivers/nv04/nv04_screen.h +++ b/src/gallium/drivers/nv04/nv04_screen.h @@ -1,11 +1,11 @@ #ifndef __NV04_SCREEN_H__ #define __NV04_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" #include "nv04_surface_2d.h" struct nv04_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; unsigned chipset; diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c index 87c635f962..d356ebd8b3 100644 --- a/src/gallium/drivers/nv04/nv04_state.c +++ b/src/gallium/drivers/nv04/nv04_state.c @@ -2,6 +2,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" #include "tgsi/tgsi_parse.h" @@ -334,7 +335,7 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf ) { struct nv04_context *nv04 = nv04_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; assert(shader < PIPE_SHADER_TYPES); assert(index == 0); @@ -342,12 +343,12 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, if (buf) { void *mapped; if (buf->buffer && buf->buffer->size && - (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) + (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) { memcpy(nv04->constbuf[shader], mapped, buf->buffer->size); nv04->constbuf_nr[shader] = buf->buffer->size / (4 * sizeof(float)); - ws->buffer_unmap(ws, buf->buffer); + pipe_buffer_unmap(pscreen, buf->buffer); } } } diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h index 0d51439e3f..399f750dbe 100644 --- a/src/gallium/drivers/nv04/nv04_state.h +++ b/src/gallium/drivers/nv04/nv04_state.h @@ -37,7 +37,7 @@ struct nv04_miptree { struct { uint pitch; - uint image_offset; + uint *image_offset; } level[PIPE_MAX_TEXTURE_LEVELS]; }; diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c index f3a8d7efee..5afd028ddd 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -4,6 +4,7 @@ #include "nouveau/nouveau_winsys.h" #include "nouveau/nouveau_util.h" +#include "nouveau/nouveau_screen.h" #include "nv04_surface_2d.h" static INLINE int @@ -96,11 +97,11 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, struct pipe_surface *src, int sx, int sy, int w, int h) { - struct nouveau_channel *chan = ctx->nvws->channel; + struct nouveau_channel *chan = ctx->swzsurf->channel; struct nouveau_grobj *swzsurf = ctx->swzsurf; struct nouveau_grobj *sifm = ctx->sifm; - struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); - struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src)); + struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst)); const unsigned src_pitch = ((struct nv04_surface *)src)->pitch; const unsigned max_w = 1024; const unsigned max_h = 1024; @@ -167,10 +168,10 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, struct pipe_surface *dst, int dx, int dy, struct pipe_surface *src, int sx, int sy, int w, int h) { - struct nouveau_channel *chan = ctx->nvws->channel; + struct nouveau_channel *chan = ctx->m2mf->channel; struct nouveau_grobj *m2mf = ctx->m2mf; - struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); - struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src)); + struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst)); unsigned src_pitch = ((struct nv04_surface *)src)->pitch; unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; unsigned dst_offset = dst->offset + dy * dst_pitch + @@ -213,11 +214,11 @@ nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst, int dx, int dy, struct pipe_surface *src, int sx, int sy, int w, int h) { - struct nouveau_channel *chan = ctx->nvws->channel; + struct nouveau_channel *chan = ctx->surf2d->channel; struct nouveau_grobj *surf2d = ctx->surf2d; struct nouveau_grobj *blit = ctx->blit; - struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src)); - struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src)); + struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst)); unsigned src_pitch = ((struct nv04_surface *)src)->pitch; unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; int format; @@ -279,10 +280,10 @@ static void nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst, int dx, int dy, int w, int h, unsigned value) { - struct nouveau_channel *chan = ctx->nvws->channel; + struct nouveau_channel *chan = ctx->surf2d->channel; struct nouveau_grobj *surf2d = ctx->surf2d; struct nouveau_grobj *rect = ctx->rect; - struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst)); + struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst)); unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; int cs2d_format, gdirect_format; @@ -334,10 +335,10 @@ nv04_surface_2d_takedown(struct nv04_surface_2d **pctx) } struct nv04_surface_2d * -nv04_surface_2d_init(struct nouveau_winsys *nvws) +nv04_surface_2d_init(struct nouveau_screen *screen) { struct nv04_surface_2d *ctx = CALLOC_STRUCT(nv04_surface_2d); - struct nouveau_channel *chan = nvws->channel; + struct nouveau_channel *chan = screen->channel; unsigned handle = 0x88000000, class; int ret; @@ -460,7 +461,6 @@ nv04_surface_2d_init(struct nouveau_winsys *nvws) return NULL; } - ctx->nvws = nvws; ctx->copy = nv04_surface_copy; ctx->fill = nv04_surface_fill; return ctx; diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.h b/src/gallium/drivers/nv04/nv04_surface_2d.h index 82ce7189c8..02b3f56ba8 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.h +++ b/src/gallium/drivers/nv04/nv04_surface_2d.h @@ -7,7 +7,6 @@ struct nv04_surface { }; struct nv04_surface_2d { - struct nouveau_winsys *nvws; struct nouveau_notifier *ntfy; struct nouveau_grobj *surf2d; struct nouveau_grobj *swzsurf; @@ -26,7 +25,7 @@ struct nv04_surface_2d { }; struct nv04_surface_2d * -nv04_surface_2d_init(struct nouveau_winsys *nvws); +nv04_surface_2d_init(struct nouveau_screen *screen); void nv04_surface_2d_takedown(struct nv04_surface_2d **); diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c index e925a44e29..854b855d64 100644 --- a/src/gallium/drivers/nv04/nv04_transfer.c +++ b/src/gallium/drivers/nv04/nv04_transfer.c @@ -43,7 +43,6 @@ nv04_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; template->last_level = 0; - template->compressed = pt->compressed; template->nr_samples = pt->nr_samples; template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c index d21a0e34f7..e3167814f2 100644 --- a/src/gallium/drivers/nv04/nv04_vbo.c +++ b/src/gallium/drivers/nv04/nv04_vbo.c @@ -1,6 +1,7 @@ #include "draw/draw_context.h" #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv04_context.h" #include "nv04_state.h" @@ -13,6 +14,7 @@ boolean nv04_draw_elements( struct pipe_context *pipe, unsigned indexSize, unsigned prim, unsigned start, unsigned count) { + struct pipe_screen *pscreen = pipe->screen; struct nv04_context *nv04 = nv04_context( pipe ); struct draw_context *draw = nv04->draw; unsigned i; @@ -25,17 +27,17 @@ boolean nv04_draw_elements( struct pipe_context *pipe, for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv04->vtxbuf[i].buffer) { void *buf - = pipe->winsys->buffer_map(pipe->winsys, - nv04->vtxbuf[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + = pipe_buffer_map(pscreen, + nv04->vtxbuf[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + = pipe_buffer_map(pscreen, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -55,12 +57,12 @@ boolean nv04_draw_elements( struct pipe_context *pipe, */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv04->vtxbuf[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, nv04->vtxbuf[i].buffer); + pipe_buffer_unmap(pscreen, nv04->vtxbuf[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); } } if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + pipe_buffer_unmap(pscreen, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } diff --git a/src/gallium/drivers/nv10/nv10_context.c b/src/gallium/drivers/nv10/nv10_context.c index ef2c0c5d9f..a127b134ec 100644 --- a/src/gallium/drivers/nv10/nv10_context.c +++ b/src/gallium/drivers/nv10/nv10_context.c @@ -30,18 +30,18 @@ nv10_destroy(struct pipe_context *pipe) static void nv10_init_hwctx(struct nv10_context *nv10) { struct nv10_screen *screen = nv10->screen; - struct nouveau_winsys *nvws = screen->nvws; + struct nouveau_channel *chan = screen->base.channel; int i; float projectionmatrix[16]; BEGIN_RING(celsius, NV10TCL_DMA_NOTIFY, 1); OUT_RING (screen->sync->handle); BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY0, 2); - OUT_RING (nvws->channel->vram->handle); - OUT_RING (nvws->channel->gart->handle); + OUT_RING (chan->vram->handle); + OUT_RING (chan->gart->handle); BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY2, 2); - OUT_RING (nvws->channel->vram->handle); - OUT_RING (nvws->channel->vram->handle); + OUT_RING (chan->vram->handle); + OUT_RING (chan->vram->handle); BEGIN_RING(celsius, NV10TCL_NOP, 1); OUT_RING (0); @@ -257,6 +257,29 @@ nv10_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) { } +static unsigned int +nv10_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv10_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context * nv10_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -282,6 +305,9 @@ nv10_create(struct pipe_screen *pscreen, unsigned pctx_id) nv10->pipe.clear = nv10_clear; nv10->pipe.flush = nv10_flush; + nv10->pipe.is_texture_referenced = nv10_is_texture_referenced; + nv10->pipe.is_buffer_referenced = nv10_is_buffer_referenced; + nv10_init_surface_functions(nv10); nv10_init_state_functions(nv10); diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c index 089c236302..1806d5f8cc 100644 --- a/src/gallium/drivers/nv10/nv10_prim_vbuf.c +++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c @@ -40,7 +40,6 @@ #include "util/u_debug.h" #include "pipe/p_inlines.h" -#include "pipe/internal/p_winsys_screen.h" #include "nv10_context.h" #include "nv10_state.h" @@ -124,11 +123,10 @@ nv10_vbuf_render_map_vertices( struct vbuf_render *render ) { struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); struct nv10_context *nv10 = nv10_render->nv10; - struct pipe_winsys *winsys = nv10->pipe.winsys; + struct pipe_screen *pscreen = nv10->pipe.screen; - return winsys->buffer_map(winsys, - nv10_render->buffer, - PIPE_BUFFER_USAGE_CPU_WRITE); + return pipe_buffer_map(pscreen, nv10_render->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); } static void @@ -138,10 +136,10 @@ nv10_vbuf_render_unmap_vertices( struct vbuf_render *render, { struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); struct nv10_context *nv10 = nv10_render->nv10; - struct pipe_winsys *winsys = nv10->pipe.winsys; + struct pipe_screen *pscreen = nv10->pipe.screen; assert(!nv10_render->buffer); - winsys->buffer_unmap(winsys, nv10_render->buffer); + pipe_buffer_unmap(pscreen, nv10_render->buffer); } static boolean @@ -202,8 +200,6 @@ static void nv10_vbuf_render_release_vertices( struct vbuf_render *render ) { struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render); - struct nv10_context *nv10 = nv10_render->nv10; - struct pipe_screen *pscreen = &nv10->screen->pipe; assert(nv10_render->buffer); pipe_buffer_reference(&nv10_render->buffer, NULL); diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c index 6532a93c7b..b03c291f9d 100644 --- a/src/gallium/drivers/nv10/nv10_screen.c +++ b/src/gallium/drivers/nv10/nv10_screen.c @@ -1,26 +1,8 @@ #include "pipe/p_screen.h" -#include "util/u_simple_screen.h" #include "nv10_context.h" #include "nv10_screen.h" -static const char * -nv10_screen_get_name(struct pipe_screen *screen) -{ - struct nv10_screen *nv10screen = nv10_screen(screen); - struct nouveau_device *dev = nv10screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv10_screen_get_vendor(struct pipe_screen *screen) -{ - return "nouveau"; -} - static int nv10_screen_get_param(struct pipe_screen *screen, int param) { @@ -120,10 +102,9 @@ static void nv10_screen_destroy(struct pipe_screen *pscreen) { struct nv10_screen *screen = nv10_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - nvws->notifier_free(&screen->sync); - nvws->grobj_free(&screen->celsius); + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->celsius); FREE(pscreen); } @@ -137,64 +118,69 @@ nv10_surface_buffer(struct pipe_surface *surf) } struct pipe_screen * -nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv10_screen *screen = CALLOC_STRUCT(nv10_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; unsigned celsius_class; - unsigned chipset = nvws->channel->device->chipset; int ret; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; - /* 2D engine setup */ - screen->eng2d = nv04_surface_2d_init(nvws); - screen->eng2d->buf = nv10_surface_buffer; + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv10_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv10_screen_destroy; + pscreen->get_param = nv10_screen_get_param; + pscreen->get_paramf = nv10_screen_get_paramf; + pscreen->is_format_supported = nv10_screen_is_format_supported; + + nv10_screen_init_miptree_functions(pscreen); + nv10_screen_init_transfer_functions(pscreen); /* 3D object */ - if (chipset>=0x20) - celsius_class=NV11TCL; - else if (chipset>=0x17) - celsius_class=NV17TCL; - else if (chipset>=0x11) - celsius_class=NV11TCL; + if (dev->chipset >= 0x20) + celsius_class = NV11TCL; + else if (dev->chipset >= 0x17) + celsius_class = NV17TCL; + else if (dev->chipset >= 0x11) + celsius_class = NV11TCL; else - celsius_class=NV10TCL; + celsius_class = NV10TCL; if (!celsius_class) { - NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", chipset); + NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", dev->chipset); return NULL; } - ret = nvws->grobj_alloc(nvws, celsius_class, &screen->celsius); + ret = nouveau_grobj_alloc(chan, 0xbeef0001, celsius_class, + &screen->celsius); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } + BIND_RING(chan, screen->celsius, 7); + + /* 2D engine setup */ + screen->eng2d = nv04_surface_2d_init(&screen->base); + screen->eng2d->buf = nv10_surface_buffer; /* Notifier for sync purposes */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv10_screen_destroy(&screen->pipe); + nv10_screen_destroy(pscreen); return NULL; } - screen->pipe.winsys = ws; - screen->pipe.destroy = nv10_screen_destroy; - - screen->pipe.get_name = nv10_screen_get_name; - screen->pipe.get_vendor = nv10_screen_get_vendor; - screen->pipe.get_param = nv10_screen_get_param; - screen->pipe.get_paramf = nv10_screen_get_paramf; - - screen->pipe.is_format_supported = nv10_screen_is_format_supported; - - nv10_screen_init_miptree_functions(&screen->pipe); - nv10_screen_init_transfer_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); - - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h index ad829ee3fd..86b6d8def5 100644 --- a/src/gallium/drivers/nv10/nv10_screen.h +++ b/src/gallium/drivers/nv10/nv10_screen.h @@ -1,11 +1,11 @@ #ifndef __NV10_SCREEN_H__ #define __NV10_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" #include "nv04/nv04_surface_2d.h" struct nv10_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c index 119af66dfd..9b38219b99 100644 --- a/src/gallium/drivers/nv10/nv10_state.c +++ b/src/gallium/drivers/nv10/nv10_state.c @@ -2,6 +2,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" #include "tgsi/tgsi_parse.h" @@ -460,7 +461,7 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf ) { struct nv10_context *nv10 = nv10_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; assert(shader < PIPE_SHADER_TYPES); assert(index == 0); @@ -468,12 +469,12 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, if (buf) { void *mapped; if (buf->buffer && buf->buffer->size && - (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) + (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) { memcpy(nv10->constbuf[shader], mapped, buf->buffer->size); nv10->constbuf_nr[shader] = buf->buffer->size / (4 * sizeof(float)); - ws->buffer_unmap(ws, buf->buffer); + pipe_buffer_unmap(pscreen, buf->buffer); } } } diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c index 5a99225409..c06b8d34c7 100644 --- a/src/gallium/drivers/nv10/nv10_transfer.c +++ b/src/gallium/drivers/nv10/nv10_transfer.c @@ -43,7 +43,6 @@ nv10_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; template->last_level = 0; - template->compressed = pt->compressed; template->nr_samples = pt->nr_samples; template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c index d0e788ac03..441a4f75f3 100644 --- a/src/gallium/drivers/nv10/nv10_vbo.c +++ b/src/gallium/drivers/nv10/nv10_vbo.c @@ -1,6 +1,7 @@ #include "draw/draw_context.h" #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv10_context.h" #include "nv10_state.h" @@ -15,6 +16,7 @@ boolean nv10_draw_elements( struct pipe_context *pipe, { struct nv10_context *nv10 = nv10_context( pipe ); struct draw_context *draw = nv10->draw; + struct pipe_screen *pscreen = pipe->screen; unsigned i; nv10_emit_hw_state(nv10); @@ -24,9 +26,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe, */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv10->vtxbuf[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - nv10->vtxbuf[i].buffer, + void *buf = + pipe_buffer_map(pscreen, nv10->vtxbuf[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } @@ -34,8 +35,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe, /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + = pipe_buffer_map(pscreen, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -55,12 +56,12 @@ boolean nv10_draw_elements( struct pipe_context *pipe, */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv10->vtxbuf[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, nv10->vtxbuf[i].buffer); + pipe_buffer_unmap(pscreen, nv10->vtxbuf[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); } } if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + pipe_buffer_unmap(pscreen, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } diff --git a/src/gallium/drivers/nv20/nv20_context.c b/src/gallium/drivers/nv20/nv20_context.c index 1659aec8fa..b32d0d83ba 100644 --- a/src/gallium/drivers/nv20/nv20_context.c +++ b/src/gallium/drivers/nv20/nv20_context.c @@ -30,7 +30,7 @@ nv20_destroy(struct pipe_context *pipe) static void nv20_init_hwctx(struct nv20_context *nv20) { struct nv20_screen *screen = nv20->screen; - struct nouveau_winsys *nvws = screen->nvws; + struct nouveau_channel *chan = screen->base.channel; int i; float projectionmatrix[16]; const boolean is_nv25tcl = (nv20->screen->kelvin->grclass == NV25TCL); @@ -38,11 +38,11 @@ static void nv20_init_hwctx(struct nv20_context *nv20) BEGIN_RING(kelvin, NV20TCL_DMA_NOTIFY, 1); OUT_RING (screen->sync->handle); BEGIN_RING(kelvin, NV20TCL_DMA_TEXTURE0, 2); - OUT_RING (nvws->channel->vram->handle); - OUT_RING (nvws->channel->gart->handle); /* TEXTURE1 */ + OUT_RING (chan->vram->handle); + OUT_RING (chan->gart->handle); /* TEXTURE1 */ BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 2); - OUT_RING (nvws->channel->vram->handle); - OUT_RING (nvws->channel->vram->handle); /* ZETA */ + OUT_RING (chan->vram->handle); + OUT_RING (chan->vram->handle); /* ZETA */ BEGIN_RING(kelvin, NV20TCL_DMA_QUERY, 1); OUT_RING (0); /* renouveau: beef0351, unique */ @@ -99,9 +99,9 @@ static void nv20_init_hwctx(struct nv20_context *nv20) OUT_RING (3); BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY9, 1); - OUT_RING (nvws->channel->vram->handle); + OUT_RING (chan->vram->handle); BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY8, 1); - OUT_RING (nvws->channel->vram->handle); + OUT_RING (chan->vram->handle); } BEGIN_RING(kelvin, NV20TCL_DMA_FENCE, 1); OUT_RING (0); /* renouveau: beef1e10 */ @@ -380,6 +380,30 @@ nv20_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) { } + +static unsigned int +nv20_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv20_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context * nv20_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -405,6 +429,9 @@ nv20_create(struct pipe_screen *pscreen, unsigned pctx_id) nv20->pipe.clear = nv20_clear; nv20->pipe.flush = nv20_flush; + nv20->pipe.is_texture_referenced = nv20_is_texture_referenced; + nv20->pipe.is_buffer_referenced = nv20_is_buffer_referenced; + nv20_init_surface_functions(nv20); nv20_init_state_functions(nv20); diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c index 8aa342cd2d..ddfcdb8057 100644 --- a/src/gallium/drivers/nv20/nv20_prim_vbuf.c +++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c @@ -152,12 +152,11 @@ static void * nv20_vbuf_render_map_vertices( struct vbuf_render *render ) { struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); - struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; + struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen; if (nv20_render->pbuffer) { - return winsys->buffer_map(winsys, - nv20_render->pbuffer, - PIPE_BUFFER_USAGE_CPU_WRITE); + return pipe_buffer_map(pscreen, nv20_render->pbuffer, + PIPE_BUFFER_USAGE_CPU_WRITE); } else if (nv20_render->mbuffer) { return nv20_render->mbuffer; } else @@ -173,10 +172,10 @@ nv20_vbuf_render_unmap_vertices( struct vbuf_render *render, ushort max_index ) { struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); - struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys; + struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen; if (nv20_render->pbuffer) - winsys->buffer_unmap(winsys, nv20_render->pbuffer); + pipe_buffer_unmap(pscreen, nv20_render->pbuffer); } static boolean @@ -358,7 +357,6 @@ nv20_vbuf_render_release_vertices( struct vbuf_render *render ) { struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render); struct nv20_context *nv20 = nv20_render->nv20; - struct pipe_screen *pscreen = &nv20->screen->pipe; if (nv20_render->pbuffer) { pipe_buffer_reference(&nv20_render->pbuffer, NULL); diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c index 7760ae27c0..024356ca74 100644 --- a/src/gallium/drivers/nv20/nv20_screen.c +++ b/src/gallium/drivers/nv20/nv20_screen.c @@ -1,26 +1,8 @@ #include "pipe/p_screen.h" -#include "util/u_simple_screen.h" #include "nv20_context.h" #include "nv20_screen.h" -static const char * -nv20_screen_get_name(struct pipe_screen *screen) -{ - struct nv20_screen *nv20screen = nv20_screen(screen); - struct nouveau_device *dev = nv20screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv20_screen_get_vendor(struct pipe_screen *screen) -{ - return "nouveau"; -} - static int nv20_screen_get_param(struct pipe_screen *screen, int param) { @@ -120,10 +102,9 @@ static void nv20_screen_destroy(struct pipe_screen *pscreen) { struct nv20_screen *screen = nv20_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - nvws->notifier_free(&screen->sync); - nvws->grobj_free(&screen->kelvin); + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->kelvin); FREE(pscreen); } @@ -137,60 +118,65 @@ nv20_surface_buffer(struct pipe_surface *surf) } struct pipe_screen * -nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv20_screen *screen = CALLOC_STRUCT(nv20_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; unsigned kelvin_class = 0; - unsigned chipset = nvws->channel->device->chipset; int ret; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; - /* 2D engine setup */ - screen->eng2d = nv04_surface_2d_init(nvws); - screen->eng2d->buf = nv20_surface_buffer; + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv20_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv20_screen_destroy; + pscreen->get_param = nv20_screen_get_param; + pscreen->get_paramf = nv20_screen_get_paramf; + pscreen->is_format_supported = nv20_screen_is_format_supported; + + nv20_screen_init_miptree_functions(pscreen); + nv20_screen_init_transfer_functions(pscreen); /* 3D object */ - if (chipset >= 0x25) + if (dev->chipset >= 0x25) kelvin_class = NV25TCL; - else if (chipset >= 0x20) + else if (dev->chipset >= 0x20) kelvin_class = NV20TCL; - if (!kelvin_class || chipset >= 0x30) { - NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", chipset); + if (!kelvin_class || dev->chipset >= 0x30) { + NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", dev->chipset); return NULL; } - ret = nvws->grobj_alloc(nvws, kelvin_class, &screen->kelvin); + ret = nouveau_grobj_alloc(chan, 0xbeef0097, kelvin_class, + &screen->kelvin); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } + BIND_RING(chan, screen->kelvin, 7); + + /* 2D engine setup */ + screen->eng2d = nv04_surface_2d_init(&screen->base); + screen->eng2d->buf = nv20_surface_buffer; /* Notifier for sync purposes */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv20_screen_destroy(&screen->pipe); + nv20_screen_destroy(pscreen); return NULL; } - screen->pipe.winsys = ws; - screen->pipe.destroy = nv20_screen_destroy; - - screen->pipe.get_name = nv20_screen_get_name; - screen->pipe.get_vendor = nv20_screen_get_vendor; - screen->pipe.get_param = nv20_screen_get_param; - screen->pipe.get_paramf = nv20_screen_get_paramf; - - screen->pipe.is_format_supported = nv20_screen_is_format_supported; - - nv20_screen_init_miptree_functions(&screen->pipe); - nv20_screen_init_transfer_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); - - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h index d9fce2bced..fc7bb05033 100644 --- a/src/gallium/drivers/nv20/nv20_screen.h +++ b/src/gallium/drivers/nv20/nv20_screen.h @@ -1,11 +1,11 @@ #ifndef __NV20_SCREEN_H__ #define __NV20_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" #include "nv04/nv04_surface_2d.h" struct nv20_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c index ecec4f49a0..ed4084980f 100644 --- a/src/gallium/drivers/nv20/nv20_state.c +++ b/src/gallium/drivers/nv20/nv20_state.c @@ -2,6 +2,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" #include "tgsi/tgsi_parse.h" @@ -453,7 +454,7 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf ) { struct nv20_context *nv20 = nv20_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; assert(shader < PIPE_SHADER_TYPES); assert(index == 0); @@ -461,12 +462,12 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, if (buf) { void *mapped; if (buf->buffer && buf->buffer->size && - (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) + (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) { memcpy(nv20->constbuf[shader], mapped, buf->buffer->size); nv20->constbuf_nr[shader] = buf->buffer->size / (4 * sizeof(float)); - ws->buffer_unmap(ws, buf->buffer); + pipe_buffer_unmap(pscreen, buf->buffer); } } } diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c index e5255296aa..5018995596 100644 --- a/src/gallium/drivers/nv20/nv20_transfer.c +++ b/src/gallium/drivers/nv20/nv20_transfer.c @@ -43,7 +43,6 @@ nv20_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; template->last_level = 0; - template->compressed = pt->compressed; template->nr_samples = pt->nr_samples; template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c index 24d8f4bef0..84d7db6c5e 100644 --- a/src/gallium/drivers/nv20/nv20_vbo.c +++ b/src/gallium/drivers/nv20/nv20_vbo.c @@ -1,6 +1,7 @@ #include "draw/draw_context.h" #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv20_context.h" #include "nv20_state.h" @@ -13,6 +14,7 @@ boolean nv20_draw_elements( struct pipe_context *pipe, unsigned indexSize, unsigned prim, unsigned start, unsigned count) { + struct pipe_screen *pscreen = pipe->screen; struct nv20_context *nv20 = nv20_context( pipe ); struct draw_context *draw = nv20->draw; unsigned i; @@ -25,17 +27,17 @@ boolean nv20_draw_elements( struct pipe_context *pipe, for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv20->vtxbuf[i].buffer) { void *buf - = pipe->winsys->buffer_map(pipe->winsys, - nv20->vtxbuf[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + = pipe_buffer_map(pscreen, + nv20->vtxbuf[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + = pipe_buffer_map(pscreen, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -55,12 +57,12 @@ boolean nv20_draw_elements( struct pipe_context *pipe, */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (nv20->vtxbuf[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, nv20->vtxbuf[i].buffer); + pipe_buffer_unmap(pscreen, nv20->vtxbuf[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); } } if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + pipe_buffer_unmap(pscreen, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c index 5db0e807ff..c1e588902b 100644 --- a/src/gallium/drivers/nv20/nv20_vertprog.c +++ b/src/gallium/drivers/nv20/nv20_vertprog.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" @@ -645,8 +646,8 @@ out_err: static boolean nv20_vertprog_validate(struct nv20_context *nv20) { + struct pipe_screen *pscreen = nv20->pipe.screen; struct nouveau_winsys *nvws = nv20->nvws; - struct pipe_winsys *ws = nv20->pipe.winsys; struct nouveau_grobj *rankine = nv20->screen->rankine; struct nv20_vertex_program *vp; struct pipe_buffer *constbuf; @@ -749,8 +750,8 @@ nv20_vertprog_validate(struct nv20_context *nv20) float *map = NULL; if (constbuf) { - map = ws->buffer_map(ws, constbuf, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); } for (i = 0; i < vp->nr_consts; i++) { @@ -770,9 +771,8 @@ nv20_vertprog_validate(struct nv20_context *nv20) OUT_RINGp ((uint32_t *)vpd->value, 4); } - if (constbuf) { - ws->buffer_unmap(ws, constbuf); - } + if (constbuf) + pipe_buffer_unmap(pscreen, constbuf); } /* Upload vtxprog */ diff --git a/src/gallium/drivers/nv30/nv30_context.c b/src/gallium/drivers/nv30/nv30_context.c index 61654f8756..f827bdc78b 100644 --- a/src/gallium/drivers/nv30/nv30_context.c +++ b/src/gallium/drivers/nv30/nv30_context.c @@ -31,6 +31,29 @@ nv30_destroy(struct pipe_context *pipe) FREE(nv30); } +static unsigned int +nv30_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv30_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context * nv30_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -55,6 +78,9 @@ nv30_create(struct pipe_screen *pscreen, unsigned pctx_id) nv30->pipe.clear = nv30_clear; nv30->pipe.flush = nv30_flush; + nv30->pipe.is_texture_referenced = nv30_is_texture_referenced; + nv30->pipe.is_buffer_referenced = nv30_is_buffer_referenced; + nv30_init_query_functions(nv30); nv30_init_surface_functions(nv30); nv30_init_state_functions(nv30); diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c index bdfe1425d2..1d1c556fb1 100644 --- a/src/gallium/drivers/nv30/nv30_fragprog.c +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" @@ -798,12 +799,12 @@ static void nv30_fragprog_upload(struct nv30_context *nv30, struct nv30_fragment_program *fp) { - struct pipe_winsys *ws = nv30->pipe.winsys; + struct pipe_screen *pscreen = nv30->pipe.screen; const uint32_t le = 1; uint32_t *map; int i; - map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); #if 0 for (i = 0; i < fp->insn_len; i++) { @@ -825,7 +826,7 @@ nv30_fragprog_upload(struct nv30_context *nv30, } } - ws->buffer_unmap(ws, fp->buffer); + pipe_buffer_unmap(pscreen, fp->buffer); } static boolean @@ -834,8 +835,7 @@ nv30_fragprog_validate(struct nv30_context *nv30) struct nv30_fragment_program *fp = nv30->fragprog; struct pipe_buffer *constbuf = nv30->constbuf[PIPE_SHADER_FRAGMENT]; - struct pipe_screen *screen = nv30->pipe.screen; - struct pipe_winsys *ws = nv30->pipe.winsys; + struct pipe_screen *pscreen = nv30->pipe.screen; struct nouveau_stateobj *so; boolean new_consts = FALSE; int i; @@ -850,14 +850,15 @@ nv30_fragprog_validate(struct nv30_context *nv30) return FALSE; } - fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4); + fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4); nv30_fragprog_upload(nv30, fp); so = so_new(8, 1); so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1); - so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | - NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, - NV34TCL_FP_ACTIVE_PROGRAM_DMA0, NV34TCL_FP_ACTIVE_PROGRAM_DMA1); + so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW | + NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0, + NV34TCL_FP_ACTIVE_PROGRAM_DMA1); so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1); so_data (so, fp->fp_control); so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1); @@ -871,7 +872,8 @@ update_constants: if (fp->nr_consts) { float *map; - map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); for (i = 0; i < fp->nr_consts; i++) { struct nv30_fragment_program_data *fpd = &fp->consts[i]; uint32_t *p = &fp->insn[fpd->offset]; @@ -882,7 +884,7 @@ update_constants: memcpy(p, cb, 4 * sizeof(float)); new_consts = TRUE; } - ws->buffer_unmap(ws, constbuf); + pipe_buffer_unmap(pscreen, constbuf); if (new_consts) nv30_fragprog_upload(nv30, fp); diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c index 8b6ab992d1..822e1d8def 100644 --- a/src/gallium/drivers/nv30/nv30_fragtex.c +++ b/src/gallium/drivers/nv30/nv30_fragtex.c @@ -61,6 +61,7 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit) struct nv30_sampler_state *ps = nv30->tex_sampler[unit]; struct nv30_miptree *nv30mt = nv30->tex_miptree[unit]; struct pipe_texture *pt = &nv30mt->base; + struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer); struct nv30_texture_format *tf; struct nouveau_stateobj *so; uint32_t txf, txs , txp; @@ -106,9 +107,9 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit) so = so_new(16, 2); so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8); - so_reloc (so, nv30mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); - so_reloc (so, nv30mt->buffer, txf, tex_flags | NOUVEAU_BO_OR, - NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1); + so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, + NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1); so_data (so, ps->wrap); so_data (so, NV34TCL_TX_ENABLE_ENABLE | ps->en); so_data (so, txs); diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c index 2f974cf5c4..1d1c8a484e 100644 --- a/src/gallium/drivers/nv30/nv30_query.c +++ b/src/gallium/drivers/nv30/nv30_query.c @@ -29,11 +29,10 @@ nv30_query_create(struct pipe_context *pipe, unsigned query_type) static void nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) { - struct nv30_context *nv30 = nv30_context(pipe); struct nv30_query *q = nv30_query(pq); if (q->object) - nv30->nvws->res_free(&q->object); + nouveau_resource_free(&q->object); FREE(q); } @@ -54,9 +53,9 @@ nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq) pipe->get_query_result(pipe, pq, 1, &tmp); } - if (nv30->nvws->res_alloc(nv30->screen->query_heap, 1, NULL, &q->object)) + if (nouveau_resource_alloc(nv30->screen->query_heap, 1, NULL, &q->object)) assert(0); - nv30->nvws->notifier_reset(nv30->screen->query, q->object->start); + nouveau_notifier_reset(nv30->screen->query, q->object->start); BEGIN_RING(rankine, NV34TCL_QUERY_RESET, 1); OUT_RING (1); @@ -84,27 +83,27 @@ nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq, { struct nv30_context *nv30 = nv30_context(pipe); struct nv30_query *q = nv30_query(pq); - struct nouveau_winsys *nvws = nv30->nvws; assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER); if (!q->ready) { unsigned status; - status = nvws->notifier_status(nv30->screen->query, - q->object->start); + status = nouveau_notifier_status(nv30->screen->query, + q->object->start); if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) { if (wait == FALSE) return FALSE; - nvws->notifier_wait(nv30->screen->query, q->object->start, - NV_NOTIFY_STATE_STATUS_COMPLETED, - 0); + + nouveau_notifier_wait_status(nv30->screen->query, + q->object->start, + NV_NOTIFY_STATE_STATUS_COMPLETED, 0); } - q->result = nvws->notifier_retval(nv30->screen->query, - q->object->start); + q->result = nouveau_notifier_return_val(nv30->screen->query, + q->object->start); q->ready = TRUE; - nvws->res_free(&q->object); + nouveau_resource_free(&q->object); } *result = q->result; diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c index d395c5e1b7..31bc1f3173 100644 --- a/src/gallium/drivers/nv30/nv30_screen.c +++ b/src/gallium/drivers/nv30/nv30_screen.c @@ -1,5 +1,7 @@ #include "pipe/p_screen.h" -#include "util/u_simple_screen.h" +#include "pipe/p_state.h" + +#include "nouveau/nouveau_screen.h" #include "nv30_context.h" #include "nv30_screen.h" @@ -8,23 +10,6 @@ #define NV34TCL_CHIPSET_3X_MASK 0x00000010 #define NV35TCL_CHIPSET_3X_MASK 0x000001e0 -static const char * -nv30_screen_get_name(struct pipe_screen *pscreen) -{ - struct nv30_screen *screen = nv30_screen(pscreen); - struct nouveau_device *dev = screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv30_screen_get_vendor(struct pipe_screen *pscreen) -{ - return "nouveau"; -} - static int nv30_screen_get_param(struct pipe_screen *pscreen, int param) { @@ -139,45 +124,57 @@ static void nv30_screen_destroy(struct pipe_screen *pscreen) { struct nv30_screen *screen = nv30_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - nvws->res_free(&screen->vp_exec_heap); - nvws->res_free(&screen->vp_data_heap); - nvws->res_free(&screen->query_heap); - nvws->notifier_free(&screen->query); - nvws->notifier_free(&screen->sync); - nvws->grobj_free(&screen->rankine); + nouveau_resource_free(&screen->vp_exec_heap); + nouveau_resource_free(&screen->vp_data_heap); + nouveau_resource_free(&screen->query_heap); + nouveau_notifier_free(&screen->query); + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->rankine); FREE(pscreen); } struct pipe_screen * -nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; struct nouveau_stateobj *so; unsigned rankine_class = 0; - unsigned chipset = nvws->channel->device->chipset; int ret, i; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; - /* 2D engine setup */ - screen->eng2d = nv04_surface_2d_init(nvws); - screen->eng2d->buf = nv30_surface_buffer; + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv30_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv30_screen_destroy; + pscreen->get_param = nv30_screen_get_param; + pscreen->get_paramf = nv30_screen_get_paramf; + pscreen->is_format_supported = nv30_screen_surface_format_supported; + + nv30_screen_init_miptree_functions(pscreen); + nv30_screen_init_transfer_functions(pscreen); /* 3D object */ - switch (chipset & 0xf0) { + switch (dev->chipset & 0xf0) { case 0x30: - if (NV30TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + if (NV30TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f))) rankine_class = 0x0397; else - if (NV34TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + if (NV34TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f))) rankine_class = 0x0697; else - if (NV35TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f))) + if (NV35TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f))) rankine_class = 0x0497; break; default: @@ -185,43 +182,49 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) } if (!rankine_class) { - NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", chipset); + NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", dev->chipset); return NULL; } - ret = nvws->grobj_alloc(nvws, rankine_class, &screen->rankine); + ret = nouveau_grobj_alloc(chan, 0xbeef3097, rankine_class, + &screen->rankine); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } + BIND_RING(chan, screen->rankine, 7); + + /* 2D engine setup */ + screen->eng2d = nv04_surface_2d_init(&screen->base); + screen->eng2d->buf = nv30_surface_buffer; /* Notifier for sync purposes */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv30_screen_destroy(&screen->pipe); + nv30_screen_destroy(pscreen); return NULL; } /* Query objects */ - ret = nvws->notifier_alloc(nvws, 32, &screen->query); + ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query); if (ret) { NOUVEAU_ERR("Error initialising query objects: %d\n", ret); - nv30_screen_destroy(&screen->pipe); + nv30_screen_destroy(pscreen); return NULL; } - ret = nvws->res_init(&screen->query_heap, 0, 32); + ret = nouveau_resource_init(&screen->query_heap, 0, 32); if (ret) { NOUVEAU_ERR("Error initialising query object heap: %d\n", ret); - nv30_screen_destroy(&screen->pipe); + nv30_screen_destroy(pscreen); return NULL; } /* Vtxprog resources */ - if (nvws->res_init(&screen->vp_exec_heap, 0, 256) || - nvws->res_init(&screen->vp_data_heap, 0, 256)) { - nv30_screen_destroy(&screen->pipe); + if (nouveau_resource_init(&screen->vp_exec_heap, 0, 256) || + nouveau_resource_init(&screen->vp_data_heap, 0, 256)) { + nv30_screen_destroy(pscreen); return NULL; } @@ -230,23 +233,23 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1); so_data (so, screen->sync->handle); so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->gart->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->gart->handle); so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->gart->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->gart->handle); /* so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2); so_data (so, 0); so_data (so, screen->query->handle);*/ so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); for (i=1; i<8; i++) { so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1); @@ -301,23 +304,9 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->rankine, 0x1e94, 1); so_data (so, 0x13); - so_emit(nvws, so); + so_emit(chan, so); so_ref(NULL, &so); - nvws->push_flush(nvws, 0, NULL); - - screen->pipe.winsys = ws; - screen->pipe.destroy = nv30_screen_destroy; - - screen->pipe.get_name = nv30_screen_get_name; - screen->pipe.get_vendor = nv30_screen_get_vendor; - screen->pipe.get_param = nv30_screen_get_param; - screen->pipe.get_paramf = nv30_screen_get_paramf; - - screen->pipe.is_format_supported = nv30_screen_surface_format_supported; - - nv30_screen_init_miptree_functions(&screen->pipe); - nv30_screen_init_transfer_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); + nouveau_pushbuf_flush(chan, 0); - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h index 8e36883975..5fbd998b53 100644 --- a/src/gallium/drivers/nv30/nv30_screen.h +++ b/src/gallium/drivers/nv30/nv30_screen.h @@ -1,11 +1,12 @@ #ifndef __NV30_SCREEN_H__ #define __NV30_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" + #include "nv04/nv04_surface_2d.h" struct nv30_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c index c18be20a32..621b8846c8 100644 --- a/src/gallium/drivers/nv30/nv30_state_emit.c +++ b/src/gallium/drivers/nv30/nv30_state_emit.c @@ -38,6 +38,7 @@ nv30_state_do_validate(struct nv30_context *nv30, void nv30_state_emit(struct nv30_context *nv30) { + struct nouveau_channel *chan = nv30->screen->base.channel; struct nv30_state *state = &nv30->state; struct nv30_screen *screen = nv30->screen; unsigned i, samplers; @@ -57,23 +58,23 @@ nv30_state_emit(struct nv30_context *nv30) continue; so_ref (state->hw[i], &nv30->screen->state[i]); if (state->hw[i]) - so_emit(nv30->nvws, nv30->screen->state[i]); + so_emit(chan, nv30->screen->state[i]); states &= ~(1ULL << i); } state->dirty = 0; - so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FB]); + so_emit_reloc_markers(chan, state->hw[NV30_STATE_FB]); for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { if (!(samplers & (1 << i))) continue; - so_emit_reloc_markers(nv30->nvws, + so_emit_reloc_markers(chan, state->hw[NV30_STATE_FRAGTEX0+i]); samplers &= ~(1ULL << i); } - so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FRAGPROG]); + so_emit_reloc_markers(chan, state->hw[NV30_STATE_FRAGPROG]); if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/) - so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_VTXBUF]); + so_emit_reloc_markers(chan, state->hw[NV30_STATE_VTXBUF]); } boolean diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c index fdc1cade90..cb1a260eaa 100644 --- a/src/gallium/drivers/nv30/nv30_state_fb.c +++ b/src/gallium/drivers/nv30/nv30_state_fb.c @@ -5,6 +5,8 @@ static boolean nv30_state_framebuffer_validate(struct nv30_context *nv30) { struct pipe_framebuffer_state *fb = &nv30->framebuffer; + struct nouveau_channel *chan = nv30->screen->base.channel; + struct nouveau_grobj *rankine = nv30->screen->rankine; struct nv04_surface *rt[2], *zeta = NULL; uint32_t rt_enable, rt_format; int i, colour_format = 0, zeta_format = 0; @@ -79,56 +81,53 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) } nv30mt = (struct nv30_miptree *)rt[0]->base.texture; - so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1); - so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, - nv30->nvws->channel->vram->handle, - nv30->nvws->channel->gart->handle); - so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2); + so_method(so, rankine, NV34TCL_DMA_COLOR0, 1); + so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, rankine, NV34TCL_COLOR0_PITCH, 2); so_data (so, pitch); - so_reloc (so, nv30mt->buffer, rt[0]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, nouveau_bo(nv30mt->buffer), rt[0]->base.offset, + rt_flags | NOUVEAU_BO_LOW, 0, 0); } if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) { nv30mt = (struct nv30_miptree *)rt[1]->base.texture; - so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1); - so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, - nv30->nvws->channel->vram->handle, - nv30->nvws->channel->gart->handle); - so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2); - so_reloc (so, nv30mt->buffer, rt[1]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); + so_method(so, rankine, NV34TCL_DMA_COLOR1, 1); + so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, rankine, NV34TCL_COLOR1_OFFSET, 2); + so_reloc (so, nouveau_bo(nv30mt->buffer), rt[1]->base.offset, + rt_flags | NOUVEAU_BO_LOW, 0, 0); so_data (so, rt[1]->pitch); } if (zeta_format) { nv30mt = (struct nv30_miptree *)zeta->base.texture; - so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1); - so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR, - nv30->nvws->channel->vram->handle, - nv30->nvws->channel->gart->handle); - so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1); - so_reloc (so, nv30mt->buffer, zeta->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); + so_method(so, rankine, NV34TCL_DMA_ZETA, 1); + so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, rankine, NV34TCL_ZETA_OFFSET, 1); + so_reloc (so, nouveau_bo(nv30mt->buffer), zeta->base.offset, + rt_flags | NOUVEAU_BO_LOW, 0, 0); /* TODO: allocate LMA depth buffer */ } - so_method(so, nv30->screen->rankine, NV34TCL_RT_ENABLE, 1); + so_method(so, rankine, NV34TCL_RT_ENABLE, 1); so_data (so, rt_enable); - so_method(so, nv30->screen->rankine, NV34TCL_RT_HORIZ, 3); + so_method(so, rankine, NV34TCL_RT_HORIZ, 3); so_data (so, (w << 16) | 0); so_data (so, (h << 16) | 0); so_data (so, rt_format); - so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_HORIZ, 2); + so_method(so, rankine, NV34TCL_VIEWPORT_HORIZ, 2); so_data (so, (w << 16) | 0); so_data (so, (h << 16) | 0); - so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2); + so_method(so, rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2); so_data (so, ((w - 1) << 16) | 0); so_data (so, ((h - 1) << 16) | 0); - so_method(so, nv30->screen->rankine, 0x1d88, 1); + so_method(so, rankine, 0x1d88, 1); so_data (so, (1 << 12) | h); /* Wonder why this is needed, context should all be set to zero on init */ - so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1); + so_method(so, rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1); so_data (so, 0); so_ref(so, &nv30->state.hw[NV30_STATE_FB]); diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c index 8b915b35bd..2367571878 100644 --- a/src/gallium/drivers/nv30/nv30_transfer.c +++ b/src/gallium/drivers/nv30/nv30_transfer.c @@ -43,7 +43,6 @@ nv30_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; template->last_level = 0; - template->compressed = pt->compressed; template->nr_samples = pt->nr_samples; template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c index 990a876382..189656ec81 100644 --- a/src/gallium/drivers/nv30/nv30_vbo.c +++ b/src/gallium/drivers/nv30/nv30_vbo.c @@ -1,5 +1,6 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv30_context.h" #include "nv30_state.h" @@ -70,7 +71,7 @@ static boolean nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib, unsigned ib_size) { - struct pipe_screen *pscreen = &nv30->screen->pipe; + struct pipe_screen *pscreen = &nv30->screen->base.base; unsigned type; if (!ib) { @@ -108,7 +109,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so, int attrib, struct pipe_vertex_element *ve, struct pipe_vertex_buffer *vb) { - struct pipe_winsys *ws = nv30->pipe.winsys; + struct pipe_screen *pscreen = nv30->pipe.screen; struct nouveau_grobj *rankine = nv30->screen->rankine; unsigned type, ncomp; void *map; @@ -116,7 +117,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so, if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp)) return FALSE; - map = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); map += vb->buffer_offset + ve->src_offset; switch (type) { @@ -148,18 +149,17 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so, so_data (so, fui(v[0])); break; default: - ws->buffer_unmap(ws, vb->buffer); + pipe_buffer_unmap(pscreen, vb->buffer); return FALSE; } } break; default: - ws->buffer_unmap(ws, vb->buffer); + pipe_buffer_unmap(pscreen, vb->buffer); return FALSE; } - ws->buffer_unmap(ws, vb->buffer); - + pipe_buffer_unmap(pscreen, vb->buffer); return TRUE; } @@ -168,7 +168,7 @@ nv30_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv30_context *nv30 = nv30_context(pipe); - struct nouveau_channel *chan = nv30->nvws->channel; + struct nouveau_channel *chan = nv30->screen->base.channel; unsigned restart = 0; nv30_vbo_set_idxbuf(nv30, NULL, 0); @@ -228,7 +228,7 @@ static INLINE void nv30_draw_elements_u08(struct nv30_context *nv30, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv30->nvws->channel; + struct nouveau_channel *chan = nv30->screen->base.channel; while (count) { uint8_t *elts = (uint8_t *)ib + start; @@ -277,7 +277,7 @@ static INLINE void nv30_draw_elements_u16(struct nv30_context *nv30, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv30->nvws->channel; + struct nouveau_channel *chan = nv30->screen->base.channel; while (count) { uint16_t *elts = (uint16_t *)ib + start; @@ -326,7 +326,7 @@ static INLINE void nv30_draw_elements_u32(struct nv30_context *nv30, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv30->nvws->channel; + struct nouveau_channel *chan = nv30->screen->base.channel; while (count) { uint32_t *elts = (uint32_t *)ib + start; @@ -368,10 +368,10 @@ nv30_draw_elements_inline(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv30_context *nv30 = nv30_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; void *map; - map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ); if (!ib) { NOUVEAU_ERR("failed mapping ib\n"); return FALSE; @@ -392,7 +392,7 @@ nv30_draw_elements_inline(struct pipe_context *pipe, break; } - ws->buffer_unmap(ws, ib); + pipe_buffer_unmap(pscreen, ib); return TRUE; } @@ -401,7 +401,7 @@ nv30_draw_elements_vbo(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv30_context *nv30 = nv30_context(pipe); - struct nouveau_channel *chan = nv30->nvws->channel; + struct nouveau_channel *chan = nv30->screen->base.channel; unsigned restart = 0; while (count) { @@ -521,18 +521,20 @@ nv30_vbo_validate(struct nv30_context *nv30) return FALSE; } - so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset, - vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, - 0, NV34TCL_VTXBUF_ADDRESS_DMA1); + so_reloc(vtxbuf, nouveau_bo(vb->buffer), vb->buffer_offset + + ve->src_offset, vb_flags | NOUVEAU_BO_LOW | + NOUVEAU_BO_OR, 0, NV34TCL_VTXBUF_ADDRESS_DMA1); so_data (vtxfmt, ((vb->stride << NV34TCL_VTXFMT_STRIDE_SHIFT) | (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type)); } if (ib) { + struct nouveau_bo *bo = nouveau_bo(ib); + so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2); - so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); - so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR, - 0, NV34TCL_IDXBUF_FORMAT_DMA1); + so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR, + 0, NV34TCL_IDXBUF_FORMAT_DMA1); } so_method(vtxbuf, rankine, 0x1710, 1); diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c index eaf543b8f7..c7514efcfe 100644 --- a/src/gallium/drivers/nv30/nv30_vertprog.c +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" @@ -645,8 +646,7 @@ out_err: static boolean nv30_vertprog_validate(struct nv30_context *nv30) { - struct nouveau_winsys *nvws = nv30->nvws; - struct pipe_winsys *ws = nv30->pipe.winsys; + struct pipe_screen *pscreen = nv30->pipe.screen; struct nouveau_grobj *rankine = nv30->screen->rankine; struct nv30_vertex_program *vp; struct pipe_buffer *constbuf; @@ -669,15 +669,15 @@ nv30_vertprog_validate(struct nv30_context *nv30) struct nouveau_stateobj *so; uint vplen = vp->nr_insns; - if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) { + if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) { while (heap->next && heap->size < vplen) { struct nv30_vertex_program *evict; evict = heap->next->priv; - nvws->res_free(&evict->exec); + nouveau_resource_free(&evict->exec); } - if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) + if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) assert(0); } @@ -694,15 +694,16 @@ nv30_vertprog_validate(struct nv30_context *nv30) if (vp->nr_consts && !vp->data) { struct nouveau_resource *heap = nv30->screen->vp_data_heap; - if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) { + if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) { while (heap->next && heap->size < vp->nr_consts) { struct nv30_vertex_program *evict; evict = heap->next->priv; - nvws->res_free(&evict->data); + nouveau_resource_free(&evict->data); } - if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) + if (nouveau_resource_alloc(heap, vp->nr_consts, vp, + &vp->data)) assert(0); } @@ -750,8 +751,8 @@ nv30_vertprog_validate(struct nv30_context *nv30) float *map = NULL; if (constbuf) { - map = ws->buffer_map(ws, constbuf, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); } for (i = 0; i < vp->nr_consts; i++) { @@ -771,9 +772,8 @@ nv30_vertprog_validate(struct nv30_context *nv30) OUT_RINGp ((uint32_t *)vpd->value, 4); } - if (constbuf) { - ws->buffer_unmap(ws, constbuf); - } + if (constbuf) + pipe_buffer_unmap(pscreen, constbuf); } /* Upload vtxprog */ @@ -804,8 +804,6 @@ nv30_vertprog_validate(struct nv30_context *nv30) void nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp) { - struct nouveau_winsys *nvws = nv30->screen->nvws; - vp->translated = FALSE; if (vp->nr_insns) { @@ -820,9 +818,9 @@ nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp) vp->nr_consts = 0; } - nvws->res_free(&vp->exec); + nouveau_resource_free(&vp->exec); vp->exec_start = 0; - nvws->res_free(&vp->data); + nouveau_resource_free(&vp->data); vp->data_start = 0; vp->data_start_min = 0; diff --git a/src/gallium/drivers/nv40/nv40_context.c b/src/gallium/drivers/nv40/nv40_context.c index 5d325f5067..8eba6a43ef 100644 --- a/src/gallium/drivers/nv40/nv40_context.c +++ b/src/gallium/drivers/nv40/nv40_context.c @@ -31,6 +31,29 @@ nv40_destroy(struct pipe_context *pipe) FREE(nv40); } +static unsigned int +nv40_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv40_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context * nv40_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -55,6 +78,9 @@ nv40_create(struct pipe_screen *pscreen, unsigned pctx_id) nv40->pipe.clear = nv40_clear; nv40->pipe.flush = nv40_flush; + nv40->pipe.is_texture_referenced = nv40_is_texture_referenced; + nv40->pipe.is_buffer_referenced = nv40_is_buffer_referenced; + nv40_init_query_functions(nv40); nv40_init_surface_functions(nv40); nv40_init_state_functions(nv40); diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c index c83ff91d7e..b2f19ecb69 100644 --- a/src/gallium/drivers/nv40/nv40_draw.c +++ b/src/gallium/drivers/nv40/nv40_draw.c @@ -1,4 +1,5 @@ #include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" #include "util/u_pack_color.h" @@ -81,7 +82,7 @@ nv40_render_prim(struct draw_stage *stage, struct prim_header *prim, { struct nv40_render_stage *rs = nv40_render_stage(stage); struct nv40_context *nv40 = rs->nv40; - struct nouveau_pushbuf *pb = nv40->nvws->channel->pushbuf; + struct nouveau_pushbuf *pb = nv40->screen->base.channel->pushbuf; unsigned i; /* Ensure there's room for 4xfloat32 + potentially 3 begin/end */ @@ -231,7 +232,7 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv40_context *nv40 = nv40_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; unsigned i; void *map; @@ -241,13 +242,14 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe, nv40_state_emit(nv40); for (i = 0; i < nv40->vtxbuf_nr; i++) { - map = ws->buffer_map(ws, nv40->vtxbuf[i].buffer, + map = pipe_buffer_map(pscreen, nv40->vtxbuf[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(nv40->draw, i, map); } if (idxbuf) { - map = ws->buffer_map(ws, idxbuf, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, idxbuf, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map); } else { draw_set_mapped_element_buffer(nv40->draw, 0, NULL); @@ -256,21 +258,22 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe, if (nv40->constbuf[PIPE_SHADER_VERTEX]) { const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX]; - map = ws->buffer_map(ws, nv40->constbuf[PIPE_SHADER_VERTEX], - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, + nv40->constbuf[PIPE_SHADER_VERTEX], + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_constant_buffer(nv40->draw, map, nr); } draw_arrays(nv40->draw, mode, start, count); for (i = 0; i < nv40->vtxbuf_nr; i++) - ws->buffer_unmap(ws, nv40->vtxbuf[i].buffer); + pipe_buffer_unmap(pscreen, nv40->vtxbuf[i].buffer); if (idxbuf) - ws->buffer_unmap(ws, idxbuf); + pipe_buffer_unmap(pscreen, idxbuf); if (nv40->constbuf[PIPE_SHADER_VERTEX]) - ws->buffer_unmap(ws, nv40->constbuf[PIPE_SHADER_VERTEX]); + pipe_buffer_unmap(pscreen, nv40->constbuf[PIPE_SHADER_VERTEX]); draw_flush(nv40->draw); pipe->flush(pipe, 0, NULL); diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c index 16e40889ec..680976da56 100644 --- a/src/gallium/drivers/nv40/nv40_fragprog.c +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" @@ -881,12 +882,12 @@ static void nv40_fragprog_upload(struct nv40_context *nv40, struct nv40_fragment_program *fp) { - struct pipe_winsys *ws = nv40->pipe.winsys; + struct pipe_screen *pscreen = nv40->pipe.screen; const uint32_t le = 1; uint32_t *map; int i; - map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); #if 0 for (i = 0; i < fp->insn_len; i++) { @@ -908,7 +909,7 @@ nv40_fragprog_upload(struct nv40_context *nv40, } } - ws->buffer_unmap(ws, fp->buffer); + pipe_buffer_unmap(pscreen, fp->buffer); } static boolean @@ -917,8 +918,7 @@ nv40_fragprog_validate(struct nv40_context *nv40) struct nv40_fragment_program *fp = nv40->fragprog; struct pipe_buffer *constbuf = nv40->constbuf[PIPE_SHADER_FRAGMENT]; - struct pipe_screen *screen = nv40->pipe.screen; - struct pipe_winsys *ws = nv40->pipe.winsys; + struct pipe_screen *pscreen = nv40->pipe.screen; struct nouveau_stateobj *so; boolean new_consts = FALSE; int i; @@ -933,14 +933,15 @@ nv40_fragprog_validate(struct nv40_context *nv40) return FALSE; } - fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4); + fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4); nv40_fragprog_upload(nv40, fp); so = so_new(4, 1); so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1); - so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | - NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, - NV40TCL_FP_ADDRESS_DMA0, NV40TCL_FP_ADDRESS_DMA1); + so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW | + NOUVEAU_BO_OR, NV40TCL_FP_ADDRESS_DMA0, + NV40TCL_FP_ADDRESS_DMA1); so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1); so_data (so, fp->fp_control); so_ref(so, &fp->so); @@ -950,7 +951,8 @@ update_constants: if (fp->nr_consts) { float *map; - map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); for (i = 0; i < fp->nr_consts; i++) { struct nv40_fragment_program_data *fpd = &fp->consts[i]; uint32_t *p = &fp->insn[fpd->offset]; @@ -961,7 +963,7 @@ update_constants: memcpy(p, cb, 4 * sizeof(float)); new_consts = TRUE; } - ws->buffer_unmap(ws, constbuf); + pipe_buffer_unmap(pscreen, constbuf); if (new_consts) nv40_fragprog_upload(nv40, fp); diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c index eb3002dc05..f6cdf31dfe 100644 --- a/src/gallium/drivers/nv40/nv40_fragtex.c +++ b/src/gallium/drivers/nv40/nv40_fragtex.c @@ -62,6 +62,7 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit) { struct nv40_sampler_state *ps = nv40->tex_sampler[unit]; struct nv40_miptree *nv40mt = nv40->tex_miptree[unit]; + struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer); struct pipe_texture *pt = &nv40mt->base; struct nv40_texture_format *tf; struct nouveau_stateobj *so; @@ -108,9 +109,9 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit) so = so_new(16, 2); so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8); - so_reloc (so, nv40mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); - so_reloc (so, nv40mt->buffer, txf, tex_flags | NOUVEAU_BO_OR, - NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1); + so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, + NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1); so_data (so, ps->wrap); so_data (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en); so_data (so, txs); diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c index 9b9a43f49d..7874aedd42 100644 --- a/src/gallium/drivers/nv40/nv40_query.c +++ b/src/gallium/drivers/nv40/nv40_query.c @@ -29,11 +29,10 @@ nv40_query_create(struct pipe_context *pipe, unsigned query_type) static void nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) { - struct nv40_context *nv40 = nv40_context(pipe); struct nv40_query *q = nv40_query(pq); if (q->object) - nv40->nvws->res_free(&q->object); + nouveau_resource_free(&q->object); FREE(q); } @@ -54,9 +53,9 @@ nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq) pipe->get_query_result(pipe, pq, 1, &tmp); } - if (nv40->nvws->res_alloc(nv40->screen->query_heap, 1, NULL, &q->object)) + if (nouveau_resource_alloc(nv40->screen->query_heap, 1, NULL, &q->object)) assert(0); - nv40->nvws->notifier_reset(nv40->screen->query, q->object->start); + nouveau_notifier_reset(nv40->screen->query, q->object->start); BEGIN_RING(curie, NV40TCL_QUERY_RESET, 1); OUT_RING (1); @@ -84,27 +83,27 @@ nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq, { struct nv40_context *nv40 = nv40_context(pipe); struct nv40_query *q = nv40_query(pq); - struct nouveau_winsys *nvws = nv40->nvws; assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER); if (!q->ready) { unsigned status; - status = nvws->notifier_status(nv40->screen->query, - q->object->start); + status = nouveau_notifier_status(nv40->screen->query, + q->object->start); if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) { if (wait == FALSE) return FALSE; - nvws->notifier_wait(nv40->screen->query, q->object->start, - NV_NOTIFY_STATE_STATUS_COMPLETED, - 0); + nouveau_notifier_wait_status(nv40->screen->query, + q->object->start, + NV_NOTIFY_STATE_STATUS_COMPLETED, + 0); } - q->result = nvws->notifier_retval(nv40->screen->query, - q->object->start); + q->result = nouveau_notifier_return_val(nv40->screen->query, + q->object->start); q->ready = TRUE; - nvws->res_free(&q->object); + nouveau_resource_free(&q->object); } *result = q->result; diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c index 0d4baefaea..b8b2af482a 100644 --- a/src/gallium/drivers/nv40/nv40_screen.c +++ b/src/gallium/drivers/nv40/nv40_screen.c @@ -1,5 +1,4 @@ #include "pipe/p_screen.h" -#include "util/u_simple_screen.h" #include "nv40_context.h" #include "nv40_screen.h" @@ -8,23 +7,6 @@ #define NV4X_GRCLASS4497_CHIPSETS 0x00005450 #define NV6X_GRCLASS4497_CHIPSETS 0x00000088 -static const char * -nv40_screen_get_name(struct pipe_screen *pscreen) -{ - struct nv40_screen *screen = nv40_screen(pscreen); - struct nouveau_device *dev = screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv40_screen_get_vendor(struct pipe_screen *pscreen) -{ - return "nouveau"; -} - static int nv40_screen_get_param(struct pipe_screen *pscreen, int param) { @@ -148,88 +130,107 @@ static void nv40_screen_destroy(struct pipe_screen *pscreen) { struct nv40_screen *screen = nv40_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - nvws->res_free(&screen->vp_exec_heap); - nvws->res_free(&screen->vp_data_heap); - nvws->res_free(&screen->query_heap); - nvws->notifier_free(&screen->query); - nvws->notifier_free(&screen->sync); - nvws->grobj_free(&screen->curie); + nouveau_resource_free(&screen->vp_exec_heap); + nouveau_resource_free(&screen->vp_data_heap); + nouveau_resource_free(&screen->query_heap); + nouveau_notifier_free(&screen->query); + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->curie); + + nouveau_screen_fini(&screen->base); FREE(pscreen); } struct pipe_screen * -nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; struct nouveau_stateobj *so; unsigned curie_class = 0; - unsigned chipset = nvws->channel->device->chipset; int ret; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; - /* 2D engine setup */ - screen->eng2d = nv04_surface_2d_init(nvws); - screen->eng2d->buf = nv40_surface_buffer; + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv40_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv40_screen_destroy; + pscreen->get_param = nv40_screen_get_param; + pscreen->get_paramf = nv40_screen_get_paramf; + pscreen->is_format_supported = nv40_screen_surface_format_supported; + + nv40_screen_init_miptree_functions(pscreen); + nv40_screen_init_transfer_functions(pscreen); /* 3D object */ - switch (chipset & 0xf0) { + switch (dev->chipset & 0xf0) { case 0x40: - if (NV4X_GRCLASS4097_CHIPSETS & (1 << (chipset & 0x0f))) + if (NV4X_GRCLASS4097_CHIPSETS & (1 << (dev->chipset & 0x0f))) curie_class = NV40TCL; else - if (NV4X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f))) + if (NV4X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f))) curie_class = NV44TCL; break; case 0x60: - if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f))) + if (NV6X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f))) curie_class = NV44TCL; break; } if (!curie_class) { - NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", chipset); + NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", dev->chipset); return NULL; } - ret = nvws->grobj_alloc(nvws, curie_class, &screen->curie); + ret = nouveau_grobj_alloc(chan, 0xbeef3097, curie_class, &screen->curie); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } + BIND_RING(chan, screen->curie, 7); + + /* 2D engine setup */ + screen->eng2d = nv04_surface_2d_init(&screen->base); + screen->eng2d->buf = nv40_surface_buffer; /* Notifier for sync purposes */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv40_screen_destroy(&screen->pipe); + nv40_screen_destroy(pscreen); return NULL; } /* Query objects */ - ret = nvws->notifier_alloc(nvws, 32, &screen->query); + ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query); if (ret) { NOUVEAU_ERR("Error initialising query objects: %d\n", ret); - nv40_screen_destroy(&screen->pipe); + nv40_screen_destroy(pscreen); return NULL; } - ret = nvws->res_init(&screen->query_heap, 0, 32); + nouveau_resource_init(&screen->query_heap, 0, 32); if (ret) { NOUVEAU_ERR("Error initialising query object heap: %d\n", ret); - nv40_screen_destroy(&screen->pipe); + nv40_screen_destroy(pscreen); return NULL; } /* Vtxprog resources */ - if (nvws->res_init(&screen->vp_exec_heap, 0, 512) || - nvws->res_init(&screen->vp_data_heap, 0, 256)) { - nv40_screen_destroy(&screen->pipe); + if (nouveau_resource_init(&screen->vp_exec_heap, 0, 512) || + nouveau_resource_init(&screen->vp_data_heap, 0, 256)) { + nv40_screen_destroy(pscreen); return NULL; } @@ -238,25 +239,25 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1); so_data (so, screen->sync->handle); so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->gart->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->gart->handle); so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->gart->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->gart->handle); so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2); so_data (so, 0); so_data (so, screen->query->handle); so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2); - so_data (so, nvws->channel->vram->handle); - so_data (so, nvws->channel->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->curie, 0x1ea4, 3); so_data (so, 0x00000010); @@ -281,24 +282,10 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->curie, 0x1e94, 1); so_data (so, 0x00000001); - so_emit(nvws, so); + so_emit(chan, so); so_ref(NULL, &so); - nvws->push_flush(nvws, 0, NULL); - - screen->pipe.winsys = ws; - screen->pipe.destroy = nv40_screen_destroy; - - screen->pipe.get_name = nv40_screen_get_name; - screen->pipe.get_vendor = nv40_screen_get_vendor; - screen->pipe.get_param = nv40_screen_get_param; - screen->pipe.get_paramf = nv40_screen_get_paramf; - - screen->pipe.is_format_supported = nv40_screen_surface_format_supported; - - nv40_screen_init_miptree_functions(&screen->pipe); - nv40_screen_init_transfer_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); + nouveau_pushbuf_flush(chan, 0); - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h index 7b503bd207..57b4c8fc46 100644 --- a/src/gallium/drivers/nv40/nv40_screen.h +++ b/src/gallium/drivers/nv40/nv40_screen.h @@ -1,11 +1,11 @@ #ifndef __NV40_SCREEN_H__ #define __NV40_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" #include "nv04/nv04_surface_2d.h" struct nv40_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c index 10aae29832..198692965d 100644 --- a/src/gallium/drivers/nv40/nv40_state_emit.c +++ b/src/gallium/drivers/nv40/nv40_state_emit.c @@ -54,6 +54,7 @@ nv40_state_do_validate(struct nv40_context *nv40, void nv40_state_emit(struct nv40_context *nv40) { + struct nouveau_channel *chan = nv40->screen->base.channel; struct nv40_state *state = &nv40->state; struct nv40_screen *screen = nv40->screen; unsigned i, samplers; @@ -73,7 +74,7 @@ nv40_state_emit(struct nv40_context *nv40) continue; so_ref (state->hw[i], &nv40->screen->state[i]); if (state->hw[i]) - so_emit(nv40->nvws, nv40->screen->state[i]); + so_emit(chan, nv40->screen->state[i]); states &= ~(1ULL << i); } @@ -87,17 +88,17 @@ nv40_state_emit(struct nv40_context *nv40) state->dirty = 0; - so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FB]); + so_emit_reloc_markers(chan, state->hw[NV40_STATE_FB]); for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { if (!(samplers & (1 << i))) continue; - so_emit_reloc_markers(nv40->nvws, + so_emit_reloc_markers(chan, state->hw[NV40_STATE_FRAGTEX0+i]); samplers &= ~(1ULL << i); } - so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FRAGPROG]); + so_emit_reloc_markers(chan, state->hw[NV40_STATE_FRAGPROG]); if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW) - so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_VTXBUF]); + so_emit_reloc_markers(chan, state->hw[NV40_STATE_VTXBUF]); } boolean diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c index be618a306b..273142f9e0 100644 --- a/src/gallium/drivers/nv40/nv40_state_fb.c +++ b/src/gallium/drivers/nv40/nv40_state_fb.c @@ -2,15 +2,19 @@ #include "nouveau/nouveau_util.h" static struct pipe_buffer * -nv40_surface_buffer(struct pipe_surface *surface) +nv40_do_surface_buffer(struct pipe_surface *surface) { struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture; return mt->buffer; } +#define nv40_surface_buffer(ps) nouveau_bo(nv40_do_surface_buffer(ps)) + static boolean nv40_state_framebuffer_validate(struct nv40_context *nv40) { + struct nouveau_channel *chan = nv40->screen->base.channel; + struct nouveau_grobj *curie = nv40->screen->curie; struct pipe_framebuffer_state *fb = &nv40->framebuffer; struct nv04_surface *rt[4], *zeta; uint32_t rt_enable, rt_format; @@ -77,76 +81,80 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) } if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) { - so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1); - so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0, rt_flags | NOUVEAU_BO_OR, - nv40->nvws->channel->vram->handle, - nv40->nvws->channel->gart->handle); - so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2); + so_method(so, curie, NV40TCL_DMA_COLOR0, 1); + so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0, + rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, curie, NV40TCL_COLOR0_PITCH, 2); so_data (so, rt[0]->pitch); - so_reloc (so, nv40_surface_buffer(&rt[0]->base), rt[0]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, nv40_surface_buffer(&rt[0]->base), + rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW, + 0, 0); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) { - so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1); - so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0, rt_flags | NOUVEAU_BO_OR, - nv40->nvws->channel->vram->handle, - nv40->nvws->channel->gart->handle); - so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2); - so_reloc (so, nv40_surface_buffer(&rt[1]->base), rt[1]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); + so_method(so, curie, NV40TCL_DMA_COLOR1, 1); + so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0, + rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, curie, NV40TCL_COLOR1_OFFSET, 2); + so_reloc (so, nv40_surface_buffer(&rt[1]->base), + rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW, + 0, 0); so_data (so, rt[1]->pitch); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) { - so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1); - so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0, rt_flags | NOUVEAU_BO_OR, - nv40->nvws->channel->vram->handle, - nv40->nvws->channel->gart->handle); - so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(&rt[2]->base), rt[2]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); - so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1); + so_method(so, curie, NV40TCL_DMA_COLOR2, 1); + so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0, + rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, curie, NV40TCL_COLOR2_OFFSET, 1); + so_reloc (so, nv40_surface_buffer(&rt[2]->base), + rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW, + 0, 0); + so_method(so, curie, NV40TCL_COLOR2_PITCH, 1); so_data (so, rt[2]->pitch); } if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) { - so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1); - so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0, rt_flags | NOUVEAU_BO_OR, - nv40->nvws->channel->vram->handle, - nv40->nvws->channel->gart->handle); - so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(&rt[3]->base), rt[3]->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); - so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1); + so_method(so, curie, NV40TCL_DMA_COLOR3, 1); + so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0, + rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, curie, NV40TCL_COLOR3_OFFSET, 1); + so_reloc (so, nv40_surface_buffer(&rt[3]->base), + rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW, + 0, 0); + so_method(so, curie, NV40TCL_COLOR3_PITCH, 1); so_data (so, rt[3]->pitch); } if (zeta_format) { - so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1); - so_reloc (so, nv40_surface_buffer(&zeta->base), 0, rt_flags | NOUVEAU_BO_OR, - nv40->nvws->channel->vram->handle, - nv40->nvws->channel->gart->handle); - so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1); - so_reloc (so, nv40_surface_buffer(&zeta->base), zeta->base.offset, rt_flags | - NOUVEAU_BO_LOW, 0, 0); - so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1); + so_method(so, curie, NV40TCL_DMA_ZETA, 1); + so_reloc (so, nv40_surface_buffer(&zeta->base), 0, + rt_flags | NOUVEAU_BO_OR, + chan->vram->handle, chan->gart->handle); + so_method(so, curie, NV40TCL_ZETA_OFFSET, 1); + so_reloc (so, nv40_surface_buffer(&zeta->base), + zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0); + so_method(so, curie, NV40TCL_ZETA_PITCH, 1); so_data (so, zeta->pitch); } - so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1); + so_method(so, curie, NV40TCL_RT_ENABLE, 1); so_data (so, rt_enable); - so_method(so, nv40->screen->curie, NV40TCL_RT_HORIZ, 3); + so_method(so, curie, NV40TCL_RT_HORIZ, 3); so_data (so, (w << 16) | 0); so_data (so, (h << 16) | 0); so_data (so, rt_format); - so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_HORIZ, 2); + so_method(so, curie, NV40TCL_VIEWPORT_HORIZ, 2); so_data (so, (w << 16) | 0); so_data (so, (h << 16) | 0); - so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2); + so_method(so, curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2); so_data (so, ((w - 1) << 16) | 0); so_data (so, ((h - 1) << 16) | 0); - so_method(so, nv40->screen->curie, 0x1d88, 1); + so_method(so, curie, 0x1d88, 1); so_data (so, (1 << 12) | h); so_ref(so, &nv40->state.hw[NV40_STATE_FB]); diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c index 1a849da32d..a596547974 100644 --- a/src/gallium/drivers/nv40/nv40_surface.c +++ b/src/gallium/drivers/nv40/nv40_surface.c @@ -26,12 +26,13 @@ * **************************************************************************/ -#include "nv40_context.h" #include "pipe/p_defines.h" -#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" + #include "util/u_tile.h" +#include "nv40_context.h" + static void nv40_surface_copy(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c index 728e8b5674..ce45055fe8 100644 --- a/src/gallium/drivers/nv40/nv40_transfer.c +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -43,7 +43,6 @@ nv40_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; template->last_level = 0; - template->compressed = pt->compressed; template->nr_samples = pt->nr_samples; template->tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC | diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c index f3518b2e4f..b2753b8e2e 100644 --- a/src/gallium/drivers/nv40/nv40_vbo.c +++ b/src/gallium/drivers/nv40/nv40_vbo.c @@ -1,5 +1,6 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv40_context.h" #include "nv40_state.h" @@ -70,7 +71,7 @@ static boolean nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib, unsigned ib_size) { - struct pipe_screen *pscreen = &nv40->screen->pipe; + struct pipe_screen *pscreen = &nv40->screen->base.base; unsigned type; if (!ib) { @@ -108,7 +109,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so, int attrib, struct pipe_vertex_element *ve, struct pipe_vertex_buffer *vb) { - struct pipe_winsys *ws = nv40->pipe.winsys; + struct pipe_screen *pscreen = nv40->pipe.screen; struct nouveau_grobj *curie = nv40->screen->curie; unsigned type, ncomp; void *map; @@ -116,7 +117,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so, if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp)) return FALSE; - map = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ); map += vb->buffer_offset + ve->src_offset; switch (type) { @@ -148,17 +149,17 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so, so_data (so, fui(v[0])); break; default: - ws->buffer_unmap(ws, vb->buffer); + pipe_buffer_unmap(pscreen, vb->buffer); return FALSE; } } break; default: - ws->buffer_unmap(ws, vb->buffer); + pipe_buffer_unmap(pscreen, vb->buffer); return FALSE; } - ws->buffer_unmap(ws, vb->buffer); + pipe_buffer_unmap(pscreen, vb->buffer); return TRUE; } @@ -168,7 +169,7 @@ nv40_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv40_context *nv40 = nv40_context(pipe); - struct nouveau_channel *chan = nv40->nvws->channel; + struct nouveau_channel *chan = nv40->screen->base.channel; unsigned restart; nv40_vbo_set_idxbuf(nv40, NULL, 0); @@ -227,7 +228,7 @@ static INLINE void nv40_draw_elements_u08(struct nv40_context *nv40, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv40->nvws->channel; + struct nouveau_channel *chan = nv40->screen->base.channel; while (count) { uint8_t *elts = (uint8_t *)ib + start; @@ -276,7 +277,7 @@ static INLINE void nv40_draw_elements_u16(struct nv40_context *nv40, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv40->nvws->channel; + struct nouveau_channel *chan = nv40->screen->base.channel; while (count) { uint16_t *elts = (uint16_t *)ib + start; @@ -325,7 +326,7 @@ static INLINE void nv40_draw_elements_u32(struct nv40_context *nv40, void *ib, unsigned mode, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv40->nvws->channel; + struct nouveau_channel *chan = nv40->screen->base.channel; while (count) { uint32_t *elts = (uint32_t *)ib + start; @@ -367,10 +368,10 @@ nv40_draw_elements_inline(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv40_context *nv40 = nv40_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *pscreen = pipe->screen; void *map; - map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ); if (!ib) { NOUVEAU_ERR("failed mapping ib\n"); return FALSE; @@ -391,7 +392,7 @@ nv40_draw_elements_inline(struct pipe_context *pipe, break; } - ws->buffer_unmap(ws, ib); + pipe_buffer_unmap(pscreen, ib); return TRUE; } @@ -400,7 +401,7 @@ nv40_draw_elements_vbo(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv40_context *nv40 = nv40_context(pipe); - struct nouveau_channel *chan = nv40->nvws->channel; + struct nouveau_channel *chan = nv40->screen->base.channel; unsigned restart; while (count) { @@ -519,17 +520,20 @@ nv40_vbo_validate(struct nv40_context *nv40) return FALSE; } - so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset, - vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, - 0, NV40TCL_VTXBUF_ADDRESS_DMA1); + so_reloc(vtxbuf, nouveau_bo(vb->buffer), + vb->buffer_offset + ve->src_offset, + vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR, + 0, NV40TCL_VTXBUF_ADDRESS_DMA1); so_data (vtxfmt, ((vb->stride << NV40TCL_VTXFMT_STRIDE_SHIFT) | (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type)); } if (ib) { + struct nouveau_bo *bo = nouveau_bo(ib); + so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2); - so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); - so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR, + so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR, 0, NV40TCL_IDXBUF_FORMAT_DMA1); } diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c index 7df9a4d326..e75e8d3f42 100644 --- a/src/gallium/drivers/nv40/nv40_vertprog.c +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" @@ -855,8 +856,7 @@ out_err: static boolean nv40_vertprog_validate(struct nv40_context *nv40) { - struct nouveau_winsys *nvws = nv40->nvws; - struct pipe_winsys *ws = nv40->pipe.winsys; + struct pipe_screen *pscreen = nv40->pipe.screen; struct nouveau_grobj *curie = nv40->screen->curie; struct nv40_vertex_program *vp; struct pipe_buffer *constbuf; @@ -895,15 +895,15 @@ check_gpu_resources: struct nouveau_stateobj *so; uint vplen = vp->nr_insns; - if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) { + if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) { while (heap->next && heap->size < vplen) { struct nv40_vertex_program *evict; evict = heap->next->priv; - nvws->res_free(&evict->exec); + nouveau_resource_free(&evict->exec); } - if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) + if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) assert(0); } @@ -925,15 +925,15 @@ check_gpu_resources: if (vp->nr_consts && !vp->data) { struct nouveau_resource *heap = nv40->screen->vp_data_heap; - if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) { + if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) { while (heap->next && heap->size < vp->nr_consts) { struct nv40_vertex_program *evict; evict = heap->next->priv; - nvws->res_free(&evict->data); + nouveau_resource_free(&evict->data); } - if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) + if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) assert(0); } @@ -981,8 +981,8 @@ check_gpu_resources: float *map = NULL; if (constbuf) { - map = ws->buffer_map(ws, constbuf, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pscreen, constbuf, + PIPE_BUFFER_USAGE_CPU_READ); } for (i = 0; i < vp->nr_consts; i++) { @@ -1003,7 +1003,7 @@ check_gpu_resources: } if (constbuf) - ws->buffer_unmap(ws, constbuf); + pscreen->buffer_unmap(pscreen, constbuf); } /* Upload vtxprog */ @@ -1035,8 +1035,6 @@ check_gpu_resources: void nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp) { - struct nouveau_winsys *nvws = nv40->screen->nvws; - vp->translated = FALSE; if (vp->nr_insns) { @@ -1051,9 +1049,9 @@ nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp) vp->nr_consts = 0; } - nvws->res_free(&vp->exec); + nouveau_resource_free(&vp->exec); vp->exec_start = 0; - nvws->res_free(&vp->data); + nouveau_resource_free(&vp->data); vp->data_start = 0; vp->data_start_min = 0; diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c index db44a9da0e..e0b2d2880b 100644 --- a/src/gallium/drivers/nv50/nv50_clear.c +++ b/src/gallium/drivers/nv50/nv50_clear.c @@ -27,64 +27,42 @@ #include "nv50_context.h" void -nv50_clear(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) +nv50_clear(struct pipe_context *pipe, unsigned buffers, + const float *rgba, double depth, unsigned stencil) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->base.channel; struct nouveau_grobj *tesla = nv50->screen->tesla; - struct pipe_framebuffer_state fb, s_fb = nv50->framebuffer; - struct pipe_scissor_state sc, s_sc = nv50->scissor; - unsigned dirty = nv50->dirty; + struct pipe_framebuffer_state *fb = &nv50->framebuffer; + unsigned mode = 0, i; - nv50->dirty = 0; + if (!nv50_state_validate(nv50)) + return; - if (ps->format == PIPE_FORMAT_Z24S8_UNORM || - ps->format == PIPE_FORMAT_Z16_UNORM) { - fb.nr_cbufs = 0; - fb.zsbuf = ps; - } else { - fb.nr_cbufs = 1; - fb.cbufs[0] = ps; - fb.zsbuf = NULL; + if (buffers & PIPE_CLEAR_COLOR && fb->nr_cbufs) { + BEGIN_RING(chan, tesla, NV50TCL_CLEAR_COLOR(0), 4); + OUT_RING (chan, fui(rgba[0])); + OUT_RING (chan, fui(rgba[1])); + OUT_RING (chan, fui(rgba[2])); + OUT_RING (chan, fui(rgba[3])); + mode |= 0x3c; } - fb.width = ps->width; - fb.height = ps->height; - pipe->set_framebuffer_state(pipe, &fb); - sc.minx = sc.miny = 0; - sc.maxx = fb.width; - sc.maxy = fb.height; - pipe->set_scissor_state(pipe, &sc); + if (buffers & PIPE_CLEAR_DEPTHSTENCIL) { + BEGIN_RING(chan, tesla, NV50TCL_CLEAR_DEPTH, 1); + OUT_RING (chan, fui(depth)); + BEGIN_RING(chan, tesla, NV50TCL_CLEAR_STENCIL, 1); + OUT_RING (chan, stencil & 0xff); - nv50_state_validate(nv50); - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - BEGIN_RING(chan, tesla, 0x0d80, 4); - OUT_RINGf (chan, ubyte_to_float((clearValue >> 16) & 0xff)); - OUT_RINGf (chan, ubyte_to_float((clearValue >> 8) & 0xff)); - OUT_RINGf (chan, ubyte_to_float((clearValue >> 0) & 0xff)); - OUT_RINGf (chan, ubyte_to_float((clearValue >> 24) & 0xff)); - BEGIN_RING(chan, tesla, 0x19d0, 1); - OUT_RING (chan, 0x3c); - break; - case PIPE_FORMAT_Z24S8_UNORM: - BEGIN_RING(chan, tesla, 0x0d90, 1); - OUT_RINGf (chan, (float)(clearValue >> 8) * (1.0 / 16777215.0)); - BEGIN_RING(chan, tesla, 0x0da0, 1); - OUT_RING (chan, clearValue & 0xff); - BEGIN_RING(chan, tesla, 0x19d0, 1); - OUT_RING (chan, 0x03); - break; - default: - pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, - clearValue); - break; + mode |= 0x03; } - pipe->set_framebuffer_state(pipe, &s_fb); - pipe->set_scissor_state(pipe, &s_sc); - nv50->dirty |= dirty; + BEGIN_RING(chan, tesla, NV50TCL_CLEAR_BUFFERS, 1); + OUT_RING (chan, mode); + + for (i = 1; i < fb->nr_cbufs; i++) { + BEGIN_RING(chan, tesla, NV50TCL_CLEAR_BUFFERS, 1); + OUT_RING (chan, (i << 6) | 0x3c); + } } diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c index 565a5da668..e02afc4be9 100644 --- a/src/gallium/drivers/nv50/nv50_context.c +++ b/src/gallium/drivers/nv50/nv50_context.c @@ -33,7 +33,7 @@ nv50_flush(struct pipe_context *pipe, unsigned flags, { struct nv50_context *nv50 = (struct nv50_context *)pipe; - FIRE_RING(nv50->screen->nvws->channel); + FIRE_RING(nv50->screen->base.channel); } static void @@ -51,6 +51,29 @@ nv50_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield) { } +static unsigned int +nv50_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +nv50_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context * nv50_create(struct pipe_screen *pscreen, unsigned pctx_id) { @@ -76,6 +99,9 @@ nv50_create(struct pipe_screen *pscreen, unsigned pctx_id) nv50->pipe.flush = nv50_flush; + nv50->pipe.is_texture_referenced = nv50_is_texture_referenced; + nv50->pipe.is_buffer_referenced = nv50_is_buffer_referenced; + nv50_init_surface_functions(nv50); nv50_init_state_functions(nv50); nv50_init_query_functions(nv50); diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 7b67a75439..9b8cc4d37d 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -63,6 +63,11 @@ struct nv50_rasterizer_stateobj { struct nouveau_stateobj *so; }; +struct nv50_sampler_stateobj { + bool normalized; + unsigned tsc[8]; +}; + struct nv50_miptree_level { int *image_offset; unsigned pitch; @@ -70,7 +75,8 @@ struct nv50_miptree_level { struct nv50_miptree { struct pipe_texture base; - struct pipe_buffer *buffer; + + struct nouveau_bo *bo; struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS]; int image_nr; @@ -93,13 +99,6 @@ nv50_surface(struct pipe_surface *pt) return (struct nv50_surface *)pt; } -static INLINE struct pipe_buffer * -nv50_surface_buffer(struct pipe_surface *surface) -{ - struct nv50_miptree *mt = (struct nv50_miptree *)surface->texture; - return mt->buffer; -} - struct nv50_state { unsigned dirty; @@ -115,6 +114,7 @@ struct nv50_state { unsigned viewport_bypass; struct nouveau_stateobj *tsc_upload; struct nouveau_stateobj *tic_upload; + unsigned miptree_nr; struct nouveau_stateobj *vertprog; struct nouveau_stateobj *fragprog; struct nouveau_stateobj *vtxfmt; @@ -147,7 +147,7 @@ struct nv50_context { unsigned vtxbuf_nr; struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; unsigned vtxelt_nr; - unsigned *sampler[PIPE_MAX_SAMPLERS]; + struct nv50_sampler_stateobj *sampler[PIPE_MAX_SAMPLERS]; unsigned sampler_nr; struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS]; unsigned miptree_nr; diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index f79a7ca86c..6b605ba416 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -29,26 +29,35 @@ static struct pipe_texture * nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) { + struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); struct pipe_texture *pt = &mt->base; - unsigned usage, width = tmp->width[0], height = tmp->height[0]; + unsigned width = tmp->width[0], height = tmp->height[0]; unsigned depth = tmp->depth[0]; - int i, l; + uint32_t tile_mode, tile_flags, tile_h; + int ret, i, l; mt->base = *tmp; pipe_reference_init(&mt->base.reference, 1); mt->base.screen = pscreen; - usage = PIPE_BUFFER_USAGE_PIXEL; switch (pt->format) { case PIPE_FORMAT_Z24S8_UNORM: case PIPE_FORMAT_Z16_UNORM: - usage |= NOUVEAU_BUFFER_USAGE_ZETA; + tile_flags = 0x2800; break; default: + tile_flags = 0x7000; break; } + if (pt->height[0] > 32) tile_mode = 4; + else if (pt->height[0] > 16) tile_mode = 3; + else if (pt->height[0] > 8) tile_mode = 2; + else if (pt->height[0] > 4) tile_mode = 1; + else tile_mode = 0; + tile_h = 1 << (tile_mode + 2); + switch (pt->target) { case PIPE_TEXTURE_3D: mt->image_nr = pt->depth[0]; @@ -85,7 +94,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) size = align(pt->width[l], 8) * pt->block.size; size = align(size, 64); - size *= align(pt->height[l], 8) * pt->block.size; + size *= align(pt->height[l], tile_h) * pt->block.size; lvl->image_offset[i] = mt->total_size; @@ -93,12 +102,13 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) } } - mt->buffer = pscreen->buffer_create(pscreen, 256, usage, mt->total_size); - if (!mt->buffer) { + ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 256, mt->total_size, + tile_mode, tile_flags, &mt->bo); + if (ret) { FREE(mt); return NULL; } - + return &mt->base; } @@ -106,6 +116,7 @@ static struct pipe_texture * nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, const unsigned *stride, struct pipe_buffer *pb) { + struct nouveau_bo *bo = nouveau_bo(pb); struct nv50_miptree *mt; /* Only supports 2D, non-mipmapped textures for the moment */ @@ -124,7 +135,7 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, mt->level[0].pitch = *stride; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - pipe_buffer_reference(&mt->buffer, pb); + nouveau_bo_ref(bo, &mt->bo); return &mt->base; } @@ -133,7 +144,7 @@ nv50_miptree_destroy(struct pipe_texture *pt) { struct nv50_miptree *mt = nv50_miptree(pt); - pipe_buffer_reference(&mt->buffer, NULL); + nouveau_bo_ref(NULL, &mt->bo); FREE(mt); } diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 2d15868ae8..5f7d06dbec 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -85,6 +85,9 @@ struct nv50_reg { int hw; int neg; + + int rhw; /* result hw for FP outputs, or interpolant index */ + int acc; /* instruction where this reg is last read (first insn == 1) */ }; struct nv50_pc { @@ -108,12 +111,23 @@ struct nv50_pc { struct nv50_reg *temp_temp[16]; unsigned temp_temp_nr; + + unsigned interp_mode[32]; + /* perspective interpolation registers */ + struct nv50_reg *iv_p; + struct nv50_reg *iv_c; + + /* current instruction and total number of insns */ + unsigned insn_cur; + unsigned insn_nr; + + boolean allow32; }; static void alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg) { - int i; + int i = 0; if (reg->type == P_RESULT) { if (pc->p->cfg.high_result < (reg->hw + 1)) @@ -131,7 +145,22 @@ alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg) return; } - for (i = 0; i < NV50_SU_MAX_TEMP; i++) { + if (reg->rhw != -1) { + /* try to allocate temporary with index rhw first */ + if (!(pc->r_temp[reg->rhw])) { + pc->r_temp[reg->rhw] = reg; + reg->hw = reg->rhw; + if (pc->p->cfg.high_temp < (reg->rhw + 1)) + pc->p->cfg.high_temp = reg->rhw + 1; + return; + } + /* make sure we don't get things like $r0 needs to go + * in $r1 and $r1 in $r0 + */ + i = pc->result_nr * 4; + } + + for (; i < NV50_SU_MAX_TEMP; i++) { if (!(pc->r_temp[i])) { pc->r_temp[i] = reg; reg->hw = i; @@ -159,6 +188,7 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst) r->type = P_TEMP; r->index = -1; r->hw = i; + r->rhw = -1; pc->r_temp[i] = r; return r; } @@ -168,6 +198,38 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst) return NULL; } +/* Assign the hw of the discarded temporary register src + * to the tgsi register dst and free src. + */ +static void +assimilate_temp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) +{ + assert(src->index == -1 && src->hw != -1); + + if (dst->hw != -1) + pc->r_temp[dst->hw] = NULL; + pc->r_temp[src->hw] = dst; + dst->hw = src->hw; + + FREE(src); +} + +/* release the hardware resource held by r */ +static void +release_hw(struct nv50_pc *pc, struct nv50_reg *r) +{ + assert(r->type == P_TEMP); + if (r->hw == -1) + return; + + assert(pc->r_temp[r->hw] == r); + pc->r_temp[r->hw] = NULL; + + r->acc = 0; + if (r->index == -1) + FREE(r); +} + static void free_temp(struct nv50_pc *pc, struct nv50_reg *r) { @@ -250,7 +312,13 @@ alloc_immd(struct nv50_pc *pc, float f) struct nv50_reg *r = CALLOC_STRUCT(nv50_reg); unsigned hw; - hw = ctor_immd(pc, f, 0, 0, 0) * 4; + for (hw = 0; hw < pc->immd_nr * 4; hw++) + if (pc->immd_buf[hw] == f) + break; + + if (hw == pc->immd_nr * 4) + hw = ctor_immd(pc, f, -f, 0.5 * f, 0) * 4; + r->type = P_IMMD; r->hw = hw; r->index = -1; @@ -341,7 +409,8 @@ set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e) static INLINE void set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e) { - unsigned val = fui(pc->immd_buf[imm->hw]); /* XXX */ + float f = pc->immd_buf[imm->hw]; + unsigned val = fui(imm->neg ? -f : f); set_long(pc, e); /*XXX: can't be predicated - bits overlap.. catch cases where both @@ -354,20 +423,35 @@ set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e) e->inst[1] |= (val >> 6) << 2; } + +#define INTERP_LINEAR 0 +#define INTERP_FLAT 1 +#define INTERP_PERSPECTIVE 2 +#define INTERP_CENTROID 4 + +/* interpolant index has been stored in dst->rhw */ static void -emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, - struct nv50_reg *src, struct nv50_reg *iv) +emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *iv, + unsigned mode) { + assert(dst->rhw != -1); struct nv50_program_exec *e = exec(pc); e->inst[0] |= 0x80000000; set_dst(pc, dst, e); - alloc_reg(pc, src); - e->inst[0] |= (src->hw << 16); - if (iv) { - e->inst[0] |= (1 << 25); - alloc_reg(pc, iv); - e->inst[0] |= (iv->hw << 9); + e->inst[0] |= (dst->rhw << 16); + + if (mode & INTERP_FLAT) { + e->inst[0] |= (1 << 8); + } else { + if (mode & INTERP_PERSPECTIVE) { + e->inst[0] |= (1 << 25); + alloc_reg(pc, iv); + e->inst[0] |= (iv->hw << 9); + } + + if (mode & INTERP_CENTROID) + e->inst[0] |= (1 << 24); } emit(pc, e); @@ -378,22 +462,12 @@ set_data(struct nv50_pc *pc, struct nv50_reg *src, unsigned m, unsigned s, struct nv50_program_exec *e) { set_long(pc, e); -#if 1 - e->inst[1] |= (1 << 22); -#else - if (src->type == P_IMMD) { - e->inst[1] |= (NV50_CB_PMISC << 22); - } else { - if (pc->p->type == PIPE_SHADER_VERTEX) - e->inst[1] |= (NV50_CB_PVP << 22); - else - e->inst[1] |= (NV50_CB_PFP << 22); - } -#endif e->param.index = src->hw; e->param.shift = s; e->param.mask = m << (s % 32); + + e->inst[1] |= (((src->type == P_IMMD) ? 0 : 1) << 22); } static void @@ -405,12 +479,11 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) set_dst(pc, dst, e); - if (0 && dst->type != P_RESULT && src->type == P_IMMD) { + if (pc->allow32 && dst->type != P_RESULT && src->type == P_IMMD) { set_immd(pc, src, e); /*XXX: 32-bit, but steals part of "half" reg space - need to * catch and handle this case if/when we do half-regs */ - e->inst[0] |= 0x00008000; } else if (src->type == P_IMMD || src->type == P_CONST) { set_long(pc, e); @@ -426,18 +499,25 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) e->inst[0] |= (src->hw << 9); } - /* We really should support "half" instructions here at some point, - * but I don't feel confident enough about them yet. - */ - set_long(pc, e); if (is_long(e) && !is_immd(e)) { e->inst[1] |= 0x04000000; /* 32-bit */ - e->inst[1] |= 0x0003c000; /* "subsubop" 0xf == mov */ - } + e->inst[1] |= 0x0000c000; /* "subsubop" 0x3 */ + if (!(e->inst[1] & 0x20000000)) + e->inst[1] |= 0x00030000; /* "subsubop" 0xf */ + } else + e->inst[0] |= 0x00008000; emit(pc, e); } +static INLINE void +emit_mov_immdval(struct nv50_pc *pc, struct nv50_reg *dst, float f) +{ + struct nv50_reg *imm = alloc_immd(pc, f); + emit_mov(pc, dst, imm); + FREE(imm); +} + static boolean check_swap_src_0_1(struct nv50_pc *pc, struct nv50_reg **s0, struct nv50_reg **s1) @@ -541,12 +621,26 @@ emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_program_exec *e = exec(pc); e->inst[0] |= 0xc0000000; - set_long(pc, e); + + if (!pc->allow32) + set_long(pc, e); check_swap_src_0_1(pc, &src0, &src1); set_dst(pc, dst, e); set_src_0(pc, src0, e); - set_src_1(pc, src1, e); + if (src1->type == P_IMMD && !is_long(e)) { + if (src0->neg) + e->inst[0] |= 0x00008000; + set_immd(pc, src1, e); + } else { + set_src_1(pc, src1, e); + if (src0->neg ^ src1->neg) { + if (is_long(e)) + e->inst[1] |= 0x08000000; + else + e->inst[0] |= 0x00008000; + } + } emit(pc, e); } @@ -560,11 +654,20 @@ emit_add(struct nv50_pc *pc, struct nv50_reg *dst, e->inst[0] |= 0xb0000000; check_swap_src_0_1(pc, &src0, &src1); + + if (!pc->allow32 || src0->neg || src1->neg) { + set_long(pc, e); + e->inst[1] |= (src0->neg << 26) | (src1->neg << 27); + } + set_dst(pc, dst, e); set_src_0(pc, src0, e); - if (is_long(e)) + if (src1->type == P_CONST || src1->type == P_ATTR || is_long(e)) set_src_2(pc, src1, e); else + if (src1->type == P_IMMD) + set_immd(pc, src1, e); + else set_src_1(pc, src1, e); emit(pc, e); @@ -588,25 +691,13 @@ emit_minmax(struct nv50_pc *pc, unsigned sub, struct nv50_reg *dst, emit(pc, e); } -static void +static INLINE void emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1) { - struct nv50_program_exec *e = exec(pc); - - e->inst[0] |= 0xb0000000; - - set_long(pc, e); - if (check_swap_src_0_1(pc, &src0, &src1)) - e->inst[1] |= 0x04000000; - else - e->inst[1] |= 0x08000000; - - set_dst(pc, dst, e); - set_src_0(pc, src0, e); - set_src_2(pc, src1, e); - - emit(pc, e); + src1->neg ^= 1; + emit_add(pc, dst, src0, src1); + src1->neg ^= 1; } static void @@ -623,26 +714,21 @@ emit_mad(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, set_src_1(pc, src1, e); set_src_2(pc, src2, e); + if (src0->neg ^ src1->neg) + e->inst[1] |= 0x04000000; + if (src2->neg) + e->inst[1] |= 0x08000000; + emit(pc, e); } -static void +static INLINE void emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1, struct nv50_reg *src2) { - struct nv50_program_exec *e = exec(pc); - - e->inst[0] |= 0xe0000000; - set_long(pc, e); - e->inst[1] |= 0x08000000; /* src0 * src1 - src2 */ - - check_swap_src_0_1(pc, &src0, &src1); - set_dst(pc, dst, e); - set_src_0(pc, src0, e); - set_src_1(pc, src1, e); - set_src_2(pc, src2, e); - - emit(pc, e); + src2->neg ^= 1; + emit_mad(pc, dst, src0, src1, src2); + src2->neg ^= 1; } static void @@ -693,6 +779,48 @@ emit_precossin(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) emit(pc, e); } +#define CVTOP_RN 0x01 +#define CVTOP_FLOOR 0x03 +#define CVTOP_CEIL 0x05 +#define CVTOP_TRUNC 0x07 +#define CVTOP_SAT 0x08 +#define CVTOP_ABS 0x10 + +#define CVT_F32_F32 0xc4 +#define CVT_F32_S32 0x44 +#define CVT_F32_U32 0x64 +#define CVT_S32_F32 0x8c +#define CVT_S32_S32 0x0c +#define CVT_F32_F32_ROP 0xcc + +static void +emit_cvt(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src, + int wp, unsigned cop, unsigned fmt) +{ + struct nv50_program_exec *e; + + e = exec(pc); + set_long(pc, e); + + e->inst[0] |= 0xa0000000; + e->inst[1] |= 0x00004000; + e->inst[1] |= (cop << 16); + e->inst[1] |= (fmt << 24); + set_src_0(pc, src, e); + + if (wp >= 0) + set_pred_wr(pc, 1, wp, e); + + if (dst) + set_dst(pc, dst, e); + else { + e->inst[0] |= 0x000001fc; + e->inst[1] |= 0x00000008; + } + + emit(pc, e); +} + static void emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1) @@ -736,22 +864,10 @@ emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst, free_temp(pc, dst); } -static void +static INLINE void emit_flr(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) { - struct nv50_program_exec *e = exec(pc); - - e->inst[0] = 0xa0000000; /* cvt */ - set_long(pc, e); - e->inst[1] |= (6 << 29); /* cvt */ - e->inst[1] |= 0x08000000; /* integer mode */ - e->inst[1] |= 0x04000000; /* 32 bit */ - e->inst[1] |= ((0x1 << 3)) << 14; /* .rn */ - e->inst[1] |= (1 << 14); /* src .f32 */ - set_dst(pc, dst, e); - set_src_0(pc, src, e); - - emit(pc, e); + emit_cvt(pc, dst, src, -1, CVTOP_FLOOR, CVT_F32_F32_ROP); } static void @@ -768,21 +884,10 @@ emit_pow(struct nv50_pc *pc, struct nv50_reg *dst, free_temp(pc, temp); } -static void +static INLINE void emit_abs(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) { - struct nv50_program_exec *e = exec(pc); - - e->inst[0] = 0xa0000000; /* cvt */ - set_long(pc, e); - e->inst[1] |= (6 << 29); /* cvt */ - e->inst[1] |= 0x04000000; /* 32 bit */ - e->inst[1] |= (1 << 14); /* src .f32 */ - e->inst[1] |= ((1 << 6) << 14); /* .abs */ - set_dst(pc, dst, e); - set_src_0(pc, src, e); - - emit(pc, e); + emit_cvt(pc, dst, src, -1, CVTOP_ABS, CVT_F32_F32); } static void @@ -794,18 +899,12 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, struct nv50_reg *neg128 = alloc_immd(pc, -127.999999); struct nv50_reg *pos128 = alloc_immd(pc, 127.999999); struct nv50_reg *tmp[4]; + boolean allow32 = pc->allow32; - if (mask & (1 << 0)) - emit_mov(pc, dst[0], one); - - if (mask & (1 << 3)) - emit_mov(pc, dst[3], one); + pc->allow32 = FALSE; if (mask & (3 << 1)) { - if (mask & (1 << 1)) - tmp[0] = dst[1]; - else - tmp[0] = temp_temp(pc); + tmp[0] = alloc_temp(pc, NULL); emit_minmax(pc, 4, tmp[0], src[0], zero); } @@ -823,6 +922,26 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, emit_mov(pc, dst[2], zero); set_pred(pc, 3, 0, pc->p->exec_tail); } + + if (mask & (1 << 1)) + assimilate_temp(pc, dst[1], tmp[0]); + else + if (mask & (1 << 2)) + free_temp(pc, tmp[0]); + + pc->allow32 = allow32; + + /* do this last, in case src[i,j] == dst[0,3] */ + if (mask & (1 << 0)) + emit_mov(pc, dst[0], one); + + if (mask & (1 << 3)) + emit_mov(pc, dst[3], one); + + FREE(pos128); + FREE(neg128); + FREE(zero); + FREE(one); } static void @@ -853,6 +972,8 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src) e->inst[1] = 0xc4014788; set_src_0(pc, src, e); set_pred_wr(pc, 1, r_pred, e); + if (src->neg) + e->inst[1] |= 0x20000000; emit(pc, e); /* This is probably KILP */ @@ -863,6 +984,180 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src) emit(pc, e); } +static void +emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, + struct nv50_reg **src, unsigned unit, unsigned type, boolean proj) +{ + struct nv50_reg *temp, *t[4]; + struct nv50_program_exec *e; + + unsigned c, mode, dim; + + switch (type) { + case TGSI_TEXTURE_1D: + dim = 1; + break; + case TGSI_TEXTURE_UNKNOWN: + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_SHADOW1D: /* XXX: x, z */ + case TGSI_TEXTURE_RECT: + dim = 2; + break; + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: /* XXX */ + dim = 3; + break; + default: + assert(0); + break; + } + + alloc_temp4(pc, t, 0); + + if (proj) { + if (src[0]->type == P_TEMP && src[0]->rhw != -1) { + mode = pc->interp_mode[src[0]->index]; + + t[3]->rhw = src[3]->rhw; + emit_interp(pc, t[3], NULL, (mode & INTERP_CENTROID)); + emit_flop(pc, 0, t[3], t[3]); + + for (c = 0; c < dim; c++) { + t[c]->rhw = src[c]->rhw; + emit_interp(pc, t[c], t[3], + (mode | INTERP_PERSPECTIVE)); + } + } else { + emit_flop(pc, 0, t[3], src[3]); + for (c = 0; c < dim; c++) + emit_mul(pc, t[c], src[c], t[3]); + + /* XXX: for some reason the blob sometimes uses MAD: + * emit_mad(pc, t[c], src[0][c], t[3], t[3]) + * pc->p->exec_tail->inst[1] |= 0x080fc000; + */ + } + } else { + if (type == TGSI_TEXTURE_CUBE) { + temp = temp_temp(pc); + emit_minmax(pc, 4, temp, src[0], src[1]); + emit_minmax(pc, 4, temp, temp, src[2]); + emit_flop(pc, 0, temp, temp); + for (c = 0; c < 3; c++) + emit_mul(pc, t[c], src[c], temp); + } else { + for (c = 0; c < dim; c++) + emit_mov(pc, t[c], src[c]); + } + } + + e = exec(pc); + set_long(pc, e); + e->inst[0] |= 0xf0000000; + e->inst[1] |= 0x00000004; + set_dst(pc, t[0], e); + e->inst[0] |= (unit << 9); + + if (dim == 2) + e->inst[0] |= 0x00400000; + else + if (dim == 3) + e->inst[0] |= 0x00800000; + + e->inst[0] |= (mask & 0x3) << 25; + e->inst[1] |= (mask & 0xc) << 12; + + emit(pc, e); + +#if 1 + if (mask & 1) emit_mov(pc, dst[0], t[0]); + if (mask & 2) emit_mov(pc, dst[1], t[1]); + if (mask & 4) emit_mov(pc, dst[2], t[2]); + if (mask & 8) emit_mov(pc, dst[3], t[3]); + + free_temp4(pc, t); +#else + /* XXX: if p.e. MUL is used directly after TEX, it would still use + * the texture coordinates, not the fetched values: latency ? */ + + for (c = 0; c < 4; c++) { + if (mask & (1 << c)) + assimilate_temp(pc, dst[c], t[c]); + else + free_temp(pc, t[c]); + } +#endif +} + +static void +convert_to_long(struct nv50_pc *pc, struct nv50_program_exec *e) +{ + unsigned q = 0, m = ~0; + + assert(!is_long(e)); + + switch (e->inst[0] >> 28) { + case 0x1: + /* MOV */ + q = 0x0403c000; + m = 0xffff7fff; + break; + case 0x8: + /* INTERP */ + m = ~0x02000000; + if (e->inst[0] & 0x02000000) + q = 0x00020000; + break; + case 0x9: + /* RCP */ + break; + case 0xB: + /* ADD */ + m = ~(127 << 16); + q = ((e->inst[0] & (~m)) >> 2); + break; + case 0xC: + /* MUL */ + m = ~0x00008000; + q = ((e->inst[0] & (~m)) << 12); + break; + case 0xE: + /* MAD (if src2 == dst) */ + q = ((e->inst[0] & 0x1fc) << 12); + break; + default: + assert(0); + break; + } + + set_long(pc, e); + pc->p->exec_size++; + + e->inst[0] &= m; + e->inst[1] |= q; +} + +static boolean +negate_supported(const struct tgsi_full_instruction *insn, int i) +{ + switch (insn->Instruction.Opcode) { + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_MUL: + case TGSI_OPCODE_KIL: + case TGSI_OPCODE_ADD: + case TGSI_OPCODE_SUB: + case TGSI_OPCODE_MAD: + return TRUE; + case TGSI_OPCODE_POW: + return (i == 1) ? TRUE : FALSE; + default: + return FALSE; + } +} + static struct nv50_reg * tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst) { @@ -881,11 +1176,14 @@ tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst) } static struct nv50_reg * -tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src) +tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src, + boolean neg) { struct nv50_reg *r = NULL; struct nv50_reg *temp; - unsigned c; + unsigned sgn, c; + + sgn = tgsi_util_get_full_src_register_sign_mode(src, chan); c = tgsi_util_get_full_src_register_extswizzle(src, chan); switch (c) { @@ -915,16 +1213,17 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src) break; case TGSI_EXTSWIZZLE_ZERO: r = alloc_immd(pc, 0.0); - break; + return r; case TGSI_EXTSWIZZLE_ONE: - r = alloc_immd(pc, 1.0); - break; + if (sgn == TGSI_UTIL_SIGN_TOGGLE || sgn == TGSI_UTIL_SIGN_SET) + return alloc_immd(pc, -1.0); + return alloc_immd(pc, 1.0); default: assert(0); break; } - switch (tgsi_util_get_full_src_register_sign_mode(src, chan)) { + switch (sgn) { case TGSI_UTIL_SIGN_KEEP: break; case TGSI_UTIL_SIGN_CLEAR: @@ -933,14 +1232,21 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src) r = temp; break; case TGSI_UTIL_SIGN_TOGGLE: - temp = temp_temp(pc); - emit_neg(pc, temp, r); - r = temp; + if (neg) + r->neg = 1; + else { + temp = temp_temp(pc); + emit_neg(pc, temp, r); + r = temp; + } break; case TGSI_UTIL_SIGN_SET: temp = temp_temp(pc); emit_abs(pc, temp, r); - emit_neg(pc, temp, r); + if (neg) + temp->neg = 1; + else + emit_neg(pc, temp, temp); r = temp; break; default: @@ -951,12 +1257,40 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src) return r; } +/* returns TRUE if instruction can overwrite sources before they're read */ +static boolean +direct2dest_op(const struct tgsi_full_instruction *insn) +{ + if (insn->Instruction.Saturate) + return FALSE; + + switch (insn->Instruction.Opcode) { + case TGSI_OPCODE_COS: + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + case TGSI_OPCODE_KIL: + case TGSI_OPCODE_LIT: + case TGSI_OPCODE_POW: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SCS: + case TGSI_OPCODE_SIN: + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXP: + return FALSE; + default: + return TRUE; + } +} + static boolean nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) { const struct tgsi_full_instruction *inst = &tok->FullInstruction; struct nv50_reg *rdst[4], *dst[4], *src[3][4], *temp; unsigned mask, sat, unit; + boolean assimilate = FALSE; int i, c; mask = inst->FullDstRegisters[0].DstRegister.WriteMask; @@ -967,6 +1301,10 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) dst[c] = tgsi_dst(pc, c, &inst->FullDstRegisters[0]); else dst[c] = NULL; + rdst[c] = NULL; + src[0][c] = NULL; + src[1][c] = NULL; + src[2][c] = NULL; } for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { @@ -976,7 +1314,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) unit = fs->SrcRegister.Index; for (c = 0; c < 4; c++) - src[i][c] = tgsi_src(pc, c, fs); + src[i][c] = tgsi_src(pc, c, fs, + negate_supported(inst, i)); } if (sat) { @@ -984,6 +1323,25 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) rdst[c] = dst[c]; dst[c] = temp_temp(pc); } + } else + if (direct2dest_op(inst)) { + for (c = 0; c < 4; c++) { + if (!dst[c] || dst[c]->type != P_TEMP) + continue; + + for (i = c + 1; i < 4; i++) { + if (dst[c] == src[0][i] || + dst[c] == src[1][i] || + dst[c] == src[2][i]) + break; + } + if (i == 4) + continue; + + assimilate = TRUE; + rdst[c] = dst[c]; + dst[c] = alloc_temp(pc, NULL); + } } switch (inst->Instruction.Opcode) { @@ -1002,7 +1360,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_COS: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_precossin(pc, temp, src[0][0]); emit_flop(pc, 5, temp, temp); for (c = 0; c < 4; c++) { @@ -1012,7 +1370,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_DP3: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_mul(pc, temp, src[0][0], src[1][0]); emit_mad(pc, temp, src[0][1], src[1][1], temp); emit_mad(pc, temp, src[0][2], src[1][2], temp); @@ -1021,10 +1379,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) continue; emit_mov(pc, dst[c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_DP4: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_mul(pc, temp, src[0][0], src[1][0]); emit_mad(pc, temp, src[0][1], src[1][1], temp); emit_mad(pc, temp, src[0][2], src[1][2], temp); @@ -1034,10 +1391,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) continue; emit_mov(pc, dst[c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_DPH: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_mul(pc, temp, src[0][0], src[1][0]); emit_mad(pc, temp, src[0][1], src[1][1], temp); emit_mad(pc, temp, src[0][2], src[1][2], temp); @@ -1047,7 +1403,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) continue; emit_mov(pc, dst[c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_DST: { @@ -1064,7 +1419,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_EX2: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_preex2(pc, temp, src[0][0]); emit_flop(pc, 6, temp, temp); for (c = 0; c < 4; c++) { @@ -1072,7 +1427,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) continue; emit_mov(pc, dst[c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_FLR: for (c = 0; c < 4; c++) { @@ -1082,26 +1436,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_FRC: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); for (c = 0; c < 4; c++) { if (!(mask & (1 << c))) continue; emit_flr(pc, temp, src[0][c]); emit_sub(pc, dst[c], src[0][c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_KIL: emit_kil(pc, src[0][0]); emit_kil(pc, src[0][1]); emit_kil(pc, src[0][2]); emit_kil(pc, src[0][3]); + pc->p->cfg.fp.regs[2] |= 0x00100000; break; case TGSI_OPCODE_LIT: emit_lit(pc, &dst[0], mask, &src[0][0]); break; case TGSI_OPCODE_LG2: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_flop(pc, 3, temp, src[0][0]); for (c = 0; c < 4; c++) { if (!(mask & (1 << c))) @@ -1110,15 +1464,12 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_LRP: + temp = temp_temp(pc); for (c = 0; c < 4; c++) { if (!(mask & (1 << c))) continue; - /*XXX: we can do better than this */ - temp = alloc_temp(pc, NULL); - emit_neg(pc, temp, src[0][c]); - emit_mad(pc, temp, temp, src[2][c], src[2][c]); - emit_mad(pc, dst[c], src[0][c], src[1][c], temp); - free_temp(pc, temp); + emit_sub(pc, temp, src[1][c], src[2][c]); + emit_mad(pc, dst[c], temp, src[0][c], src[2][c]); } break; case TGSI_OPCODE_MAD: @@ -1157,36 +1508,39 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_POW: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_pow(pc, temp, src[0][0], src[1][0]); for (c = 0; c < 4; c++) { if (!(mask & (1 << c))) continue; emit_mov(pc, dst[c], temp); } - free_temp(pc, temp); break; case TGSI_OPCODE_RCP: - for (c = 0; c < 4; c++) { + for (c = 3; c >= 0; c--) { if (!(mask & (1 << c))) continue; emit_flop(pc, 0, dst[c], src[0][0]); } break; case TGSI_OPCODE_RSQ: - for (c = 0; c < 4; c++) { + for (c = 3; c >= 0; c--) { if (!(mask & (1 << c))) continue; emit_flop(pc, 2, dst[c], src[0][0]); } break; case TGSI_OPCODE_SCS: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_precossin(pc, temp, src[0][0]); if (mask & (1 << 0)) emit_flop(pc, 5, dst[0], temp); if (mask & (1 << 1)) emit_flop(pc, 4, dst[1], temp); + if (mask & (1 << 2)) + emit_mov_immdval(pc, dst[2], 0.0); + if (mask & (1 << 3)) + emit_mov_immdval(pc, dst[3], 1.0); break; case TGSI_OPCODE_SGE: for (c = 0; c < 4; c++) { @@ -1196,7 +1550,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_SIN: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); emit_precossin(pc, temp, src[0][0]); emit_flop(pc, 4, temp, temp); for (c = 0; c < 4; c++) { @@ -1220,33 +1574,15 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_TEX: + emit_tex(pc, dst, mask, src[0], unit, + inst->InstructionExtTexture.Texture, FALSE); + break; case TGSI_OPCODE_TXP: - { - struct nv50_reg *t[4]; - struct nv50_program_exec *e; - - alloc_temp4(pc, t, 0); - emit_mov(pc, t[0], src[0][0]); - emit_mov(pc, t[1], src[0][1]); - - e = exec(pc); - e->inst[0] = 0xf6400000; - e->inst[0] |= (unit << 9); - set_long(pc, e); - e->inst[1] |= 0x0000c004; - set_dst(pc, t[0], e); - emit(pc, e); - - if (mask & (1 << 0)) emit_mov(pc, dst[0], t[0]); - if (mask & (1 << 1)) emit_mov(pc, dst[1], t[1]); - if (mask & (1 << 2)) emit_mov(pc, dst[2], t[2]); - if (mask & (1 << 3)) emit_mov(pc, dst[3], t[3]); - - free_temp4(pc, t); - } + emit_tex(pc, dst, mask, src[0], unit, + inst->InstructionExtTexture.Texture, TRUE); break; case TGSI_OPCODE_XPD: - temp = alloc_temp(pc, NULL); + temp = temp_temp(pc); if (mask & (1 << 0)) { emit_mul(pc, temp, src[0][2], src[1][1]); emit_msb(pc, dst[0], src[0][1], src[1][2], temp); @@ -1259,7 +1595,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) emit_mul(pc, temp, src[0][1], src[1][0]); emit_msb(pc, dst[2], src[0][0], src[1][1], temp); } - free_temp(pc, temp); + if (mask & (1 << 3)) + emit_mov_immdval(pc, dst[3], 1.0); break; case TGSI_OPCODE_END: break; @@ -1270,21 +1607,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) if (sat) { for (c = 0; c < 4; c++) { - struct nv50_program_exec *e; - if (!(mask & (1 << c))) continue; - e = exec(pc); + emit_cvt(pc, rdst[c], dst[c], -1, CVTOP_SAT, + CVT_F32_F32); + } + } else if (assimilate) { + for (c = 0; c < 4; c++) + if (rdst[c]) + assimilate_temp(pc, rdst[c], dst[c]); + } - e->inst[0] = 0xa0000000; /* cvt */ - set_long(pc, e); - e->inst[1] |= (6 << 29); /* cvt */ - e->inst[1] |= 0x04000000; /* 32 bit */ - e->inst[1] |= (1 << 14); /* src .f32 */ - e->inst[1] |= ((1 << 5) << 14); /* .sat */ - set_dst(pc, rdst[c], e); - set_src_0(pc, dst[c], e); - emit(pc, e); + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + for (c = 0; c < 4; c++) { + if (!src[i][c]) + continue; + if (src[i][c]->index == -1 && src[i][c]->type == P_IMMD) + FREE(src[i][c]); + else + if (src[i][c]->acc == pc->insn_cur) + release_hw(pc, src[i][c]); } } @@ -1292,12 +1634,169 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) return TRUE; } +/* Adjust a bitmask that indicates what components of a source are used, + * we use this in tx_prep so we only load interpolants that are needed. + */ +static void +insn_adjust_mask(const struct tgsi_full_instruction *insn, unsigned *mask) +{ + const struct tgsi_instruction_ext_texture *tex; + + switch (insn->Instruction.Opcode) { + case TGSI_OPCODE_DP3: + *mask = 0x7; + break; + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + *mask = 0xF; + break; + case TGSI_OPCODE_LIT: + *mask = 0xB; + break; + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + *mask = 0x1; + break; + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXP: + assert(insn->Instruction.Extended); + tex = &insn->InstructionExtTexture; + + *mask = 0x7; + if (tex->Texture == TGSI_TEXTURE_1D) + *mask = 0x1; + else + if (tex->Texture == TGSI_TEXTURE_2D) + *mask = 0x3; + + if (insn->Instruction.Opcode == TGSI_OPCODE_TXP) + *mask |= 0x8; + break; + default: + break; + } +} + +static void +prep_inspect_insn(struct nv50_pc *pc, const union tgsi_full_token *tok, + unsigned *r_usage[2]) +{ + const struct tgsi_full_instruction *insn; + const struct tgsi_full_src_register *src; + const struct tgsi_dst_register *dst; + + unsigned i, c, k, n, mask, *acc_p; + + insn = &tok->FullInstruction; + dst = &insn->FullDstRegisters[0].DstRegister; + mask = dst->WriteMask; + + if (!r_usage[0]) + r_usage[0] = CALLOC(pc->temp_nr * 4, sizeof(unsigned)); + if (!r_usage[1]) + r_usage[1] = CALLOC(pc->attr_nr * 4, sizeof(unsigned)); + + if (dst->File == TGSI_FILE_TEMPORARY) { + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + r_usage[0][dst->Index * 4 + c] = pc->insn_nr; + } + } + + for (i = 0; i < insn->Instruction.NumSrcRegs; i++) { + src = &insn->FullSrcRegisters[i]; + + switch (src->SrcRegister.File) { + case TGSI_FILE_TEMPORARY: + acc_p = r_usage[0]; + break; + case TGSI_FILE_INPUT: + acc_p = r_usage[1]; + break; + default: + continue; + } + + insn_adjust_mask(insn, &mask); + + for (c = 0; c < 4; c++) { + if (!(mask & (1 << c))) + continue; + + k = tgsi_util_get_full_src_register_extswizzle(src, c); + switch (k) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + n = src->SrcRegister.Index * 4 + k; + acc_p[n] = pc->insn_nr; + break; + default: + break; + } + } + } +} + +static unsigned +load_fp_attrib(struct nv50_pc *pc, int i, unsigned *acc, int *mid, + int *aid, int *p_oid) +{ + struct nv50_reg *iv; + int oid, c, n; + unsigned mask = 0; + + iv = (pc->interp_mode[i] & INTERP_CENTROID) ? pc->iv_c : pc->iv_p; + + for (c = 0, n = i * 4; c < 4; c++, n++) { + oid = (*p_oid)++; + pc->attr[n].type = P_TEMP; + pc->attr[n].index = i; + + if (pc->attr[n].acc == acc[n]) + continue; + mask |= (1 << c); + + pc->attr[n].acc = acc[n]; + pc->attr[n].rhw = pc->attr[n].hw = -1; + alloc_reg(pc, &pc->attr[n]); + + pc->attr[n].rhw = (*aid)++; + emit_interp(pc, &pc->attr[n], iv, pc->interp_mode[i]); + + pc->p->cfg.fp.map[(*mid) / 4] |= oid << (8 * ((*mid) % 4)); + (*mid)++; + pc->p->cfg.fp.regs[1] += 0x00010001; + } + + return mask; +} + static boolean nv50_program_tx_prep(struct nv50_pc *pc) { struct tgsi_parse_context p; boolean ret = FALSE; unsigned i, c; + unsigned fcol, bcol, fcrd, depr; + + /* count (centroid) perspective interpolations */ + unsigned centroid_loads = 0; + unsigned perspect_loads = 0; + + /* track register access for temps and attrs */ + unsigned *r_usage[2]; + r_usage[0] = NULL; + r_usage[1] = NULL; + + depr = fcol = bcol = fcrd = 0xffff; + + if (pc->p->type == PIPE_SHADER_FRAGMENT) { + pc->p->cfg.fp.regs[0] = 0x01000404; + pc->p->cfg.fp.regs[1] = 0x00000400; + } tgsi_parse_init(&p, pc->p->pipe.tokens); while (!tgsi_parse_end_of_tokens(&p)) { @@ -1319,9 +1818,10 @@ nv50_program_tx_prep(struct nv50_pc *pc) case TGSI_TOKEN_TYPE_DECLARATION: { const struct tgsi_full_declaration *d; - unsigned last; + unsigned last, first, mode; d = &p.FullToken.FullDeclaration; + first = d->DeclarationRange.First; last = d->DeclarationRange.Last; switch (d->Declaration.File) { @@ -1332,10 +1832,69 @@ nv50_program_tx_prep(struct nv50_pc *pc) case TGSI_FILE_OUTPUT: if (pc->result_nr < (last + 1)) pc->result_nr = last + 1; + + if (!d->Declaration.Semantic) + break; + + switch (d->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + depr = first; + pc->p->cfg.fp.regs[2] |= 0x00000100; + pc->p->cfg.fp.regs[3] |= 0x00000011; + break; + default: + break; + } + break; case TGSI_FILE_INPUT: + { if (pc->attr_nr < (last + 1)) pc->attr_nr = last + 1; + + if (pc->p->type != PIPE_SHADER_FRAGMENT) + break; + + switch (d->Declaration.Interpolate) { + case TGSI_INTERPOLATE_CONSTANT: + mode = INTERP_FLAT; + break; + case TGSI_INTERPOLATE_PERSPECTIVE: + mode = INTERP_PERSPECTIVE; + break; + default: + mode = INTERP_LINEAR; + break; + } + + if (d->Declaration.Semantic) { + switch (d->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + fcrd = first; + break; + case TGSI_SEMANTIC_COLOR: + fcol = first; + mode = INTERP_PERSPECTIVE; + break; + case TGSI_SEMANTIC_BCOLOR: + bcol = first; + mode = INTERP_PERSPECTIVE; + break; + } + } + + if (d->Declaration.Centroid) { + mode |= INTERP_CENTROID; + if (mode & INTERP_PERSPECTIVE) + centroid_loads++; + } else + if (mode & INTERP_PERSPECTIVE) + perspect_loads++; + + assert(last < 32); + for (i = first; i <= last; i++) + pc->interp_mode[i] = mode; + } break; case TGSI_FILE_CONSTANT: if (pc->param_nr < (last + 1)) @@ -1351,6 +1910,8 @@ nv50_program_tx_prep(struct nv50_pc *pc) } break; case TGSI_TOKEN_TYPE_INSTRUCTION: + pc->insn_nr++; + prep_inspect_insn(pc, tok, r_usage); break; default: break; @@ -1366,56 +1927,95 @@ nv50_program_tx_prep(struct nv50_pc *pc) for (c = 0; c < 4; c++) { pc->temp[i*4+c].type = P_TEMP; pc->temp[i*4+c].hw = -1; + pc->temp[i*4+c].rhw = -1; pc->temp[i*4+c].index = i; + pc->temp[i*4+c].acc = r_usage[0][i*4+c]; } } } if (pc->attr_nr) { - struct nv50_reg *iv = NULL; - int aid = 0; + int oid = 4, mid = 4, aid = 0; + /* oid = VP output id + * aid = FP attribute/interpolant id + * mid = VP output mapping field ID + */ pc->attr = CALLOC(pc->attr_nr * 4, sizeof(struct nv50_reg)); if (!pc->attr) goto out_err; if (pc->p->type == PIPE_SHADER_FRAGMENT) { - iv = alloc_temp(pc, NULL); - emit_interp(pc, iv, iv, NULL); - emit_flop(pc, 0, iv, iv); - aid++; - } + /* position should be loaded first */ + if (fcrd != 0xffff) { + unsigned mask; + mid = 0; + mask = load_fp_attrib(pc, fcrd, r_usage[1], + &mid, &aid, &oid); + oid = 0; + pc->p->cfg.fp.regs[1] |= (mask << 24); + pc->p->cfg.fp.map[0] = 0x04040404 * fcrd; + } + pc->p->cfg.fp.map[0] += 0x03020100; - for (i = 0; i < pc->attr_nr; i++) { - struct nv50_reg *a = &pc->attr[i*4]; + /* should do MAD fcrd.xy, fcrd, SOME_CONST, fcrd */ - for (c = 0; c < 4; c++) { - if (pc->p->type == PIPE_SHADER_FRAGMENT) { - struct nv50_reg *at = - alloc_temp(pc, NULL); - pc->attr[i*4+c].type = at->type; - pc->attr[i*4+c].hw = at->hw; - pc->attr[i*4+c].index = at->index; + if (perspect_loads) { + pc->iv_p = alloc_temp(pc, NULL); + + if (!(pc->p->cfg.fp.regs[1] & 0x08000000)) { + pc->p->cfg.fp.regs[1] |= 0x08000000; + pc->iv_p->rhw = aid++; + emit_interp(pc, pc->iv_p, NULL, + INTERP_LINEAR); + emit_flop(pc, 0, pc->iv_p, pc->iv_p); } else { - pc->p->cfg.vp.attr[aid/32] |= - (1 << (aid % 32)); - pc->attr[i*4+c].type = P_ATTR; - pc->attr[i*4+c].hw = aid++; - pc->attr[i*4+c].index = i; + pc->iv_p->rhw = aid - 1; + emit_flop(pc, 0, pc->iv_p, + &pc->attr[fcrd * 4 + 3]); } } - if (pc->p->type != PIPE_SHADER_FRAGMENT) - continue; + if (centroid_loads) { + pc->iv_c = alloc_temp(pc, NULL); + pc->iv_c->rhw = pc->iv_p ? aid - 1 : aid++; + emit_interp(pc, pc->iv_c, NULL, + INTERP_CENTROID); + emit_flop(pc, 0, pc->iv_c, pc->iv_c); + pc->p->cfg.fp.regs[1] |= 0x08000000; + } - emit_interp(pc, &a[0], &a[0], iv); - emit_interp(pc, &a[1], &a[1], iv); - emit_interp(pc, &a[2], &a[2], iv); - emit_interp(pc, &a[3], &a[3], iv); - } + for (c = 0; c < 4; c++) { + /* I don't know what these values do, but + * let's set them like the blob does: + */ + if (fcol != 0xffff && r_usage[1][fcol * 4 + c]) + pc->p->cfg.fp.regs[0] += 0x00010000; + if (bcol != 0xffff && r_usage[1][bcol * 4 + c]) + pc->p->cfg.fp.regs[0] += 0x00010000; + } - if (iv) - free_temp(pc, iv); + for (i = 0; i < pc->attr_nr; i++) + load_fp_attrib(pc, i, r_usage[1], + &mid, &aid, &oid); + + if (pc->iv_p) + free_temp(pc, pc->iv_p); + if (pc->iv_c) + free_temp(pc, pc->iv_c); + + pc->p->cfg.fp.high_map = (mid / 4); + pc->p->cfg.fp.high_map += ((mid % 4) ? 1 : 0); + } else { + /* vertex program */ + for (i = 0; i < pc->attr_nr * 4; i++) { + pc->p->cfg.vp.attr[aid / 32] |= + (1 << (aid % 32)); + pc->attr[i].type = P_ATTR; + pc->attr[i].hw = aid++; + pc->attr[i].index = i / 4; + } + } } if (pc->result_nr) { @@ -1430,12 +2030,20 @@ nv50_program_tx_prep(struct nv50_pc *pc) if (pc->p->type == PIPE_SHADER_FRAGMENT) { pc->result[i*4+c].type = P_TEMP; pc->result[i*4+c].hw = -1; + pc->result[i*4+c].rhw = (i == depr) ? + -1 : rid++; } else { pc->result[i*4+c].type = P_RESULT; pc->result[i*4+c].hw = rid++; } pc->result[i*4+c].index = i; } + + if (pc->p->type == PIPE_SHADER_FRAGMENT && + depr != 0xffff) { + pc->result[depr * 4 + 2].rhw = + (pc->result_nr - 1) * 4; + } } } @@ -1456,7 +2064,7 @@ nv50_program_tx_prep(struct nv50_pc *pc) } if (pc->immd_nr) { - int rid = pc->param_nr * 4; + int rid = 0; pc->immd = CALLOC(pc->immd_nr * 4, sizeof(struct nv50_reg)); if (!pc->immd) @@ -1473,15 +2081,38 @@ nv50_program_tx_prep(struct nv50_pc *pc) ret = TRUE; out_err: + if (r_usage[0]) + FREE(r_usage[0]); + if (r_usage[1]) + FREE(r_usage[1]); + tgsi_parse_free(&p); return ret; } +static void +free_nv50_pc(struct nv50_pc *pc) +{ + if (pc->immd) + FREE(pc->immd); + if (pc->param) + FREE(pc->param); + if (pc->result) + FREE(pc->result); + if (pc->attr) + FREE(pc->attr); + if (pc->temp) + FREE(pc->temp); + + FREE(pc); +} + static boolean nv50_program_tx(struct nv50_program *p) { struct tgsi_parse_context parse; struct nv50_pc *pc; + unsigned k; boolean ret; pc = CALLOC_STRUCT(nv50_pc); @@ -1498,10 +2129,16 @@ nv50_program_tx(struct nv50_program *p) while (!tgsi_parse_end_of_tokens(&parse)) { const union tgsi_full_token *tok = &parse.FullToken; + /* don't allow half insn/immd on first and last instruction */ + pc->allow32 = TRUE; + if (pc->insn_cur == 0 || pc->insn_cur + 2 == pc->insn_nr) + pc->allow32 = FALSE; + tgsi_parse_token(&parse); switch (tok->Token.Type) { case TGSI_TOKEN_TYPE_INSTRUCTION: + ++pc->insn_cur; ret = nv50_program_tx_insn(pc, tok); if (ret == FALSE) goto out_err; @@ -1515,8 +2152,40 @@ nv50_program_tx(struct nv50_program *p) struct nv50_reg out; out.type = P_TEMP; - for (out.hw = 0; out.hw < pc->result_nr * 4; out.hw++) - emit_mov(pc, &out, &pc->result[out.hw]); + for (k = 0; k < pc->result_nr * 4; k++) { + if (pc->result[k].rhw == -1) + continue; + if (pc->result[k].hw != pc->result[k].rhw) { + out.hw = pc->result[k].rhw; + emit_mov(pc, &out, &pc->result[k]); + } + if (pc->p->cfg.high_result < (pc->result[k].rhw + 1)) + pc->p->cfg.high_result = pc->result[k].rhw + 1; + } + } + + /* look for single half instructions and make them long */ + struct nv50_program_exec *e, *e_prev; + + for (k = 0, e = pc->p->exec_head, e_prev = NULL; e; e = e->next) { + if (!is_long(e)) + k++; + + if (!e->next || is_long(e->next)) { + if (k & 1) + convert_to_long(pc, e); + k = 0; + } + + if (e->next) + e_prev = e; + } + + if (!is_long(pc->p->exec_tail)) { + /* this may occur if moving FP results */ + assert(e_prev && !is_long(e_prev)); + convert_to_long(pc, e_prev); + convert_to_long(pc, pc->p->exec_tail); } assert(is_long(pc->p->exec_tail) && !is_immd(pc->p->exec_head)); @@ -1530,6 +2199,7 @@ out_err: tgsi_parse_free(&parse); out_cleanup: + free_nv50_pc(pc); return ret; } @@ -1543,16 +2213,16 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p) static void nv50_program_upload_data(struct nv50_context *nv50, float *map, - unsigned start, unsigned count) + unsigned start, unsigned count, unsigned cbuf) { - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->base.channel; struct nouveau_grobj *tesla = nv50->screen->tesla; while (count) { unsigned nr = count > 2047 ? 2047 : count; BEGIN_RING(chan, tesla, 0x00000f00, 1); - OUT_RING (chan, (NV50_CB_PMISC << 0) | (start << 8)); + OUT_RING (chan, (cbuf << 0) | (start << 8)); BEGIN_RING(chan, tesla, 0x40000f04, nr); OUT_RINGp (chan, map, nr); @@ -1565,70 +2235,93 @@ nv50_program_upload_data(struct nv50_context *nv50, float *map, static void nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p) { - struct nouveau_winsys *nvws = nv50->screen->nvws; - struct pipe_winsys *ws = nv50->pipe.winsys; - unsigned nr = p->param_nr + p->immd_nr; + struct pipe_screen *pscreen = nv50->pipe.screen; - if (!p->data && nr) { - struct nouveau_resource *heap = nv50->screen->vp_data_heap; + if (!p->data[0] && p->immd_nr) { + struct nouveau_resource *heap = nv50->screen->immd_heap[0]; - if (nvws->res_alloc(heap, nr, p, &p->data)) { - while (heap->next && heap->size < nr) { + if (nouveau_resource_alloc(heap, p->immd_nr, p, &p->data[0])) { + while (heap->next && heap->size < p->immd_nr) { struct nv50_program *evict = heap->next->priv; - nvws->res_free(&evict->data); + nouveau_resource_free(&evict->data[0]); } - if (nvws->res_alloc(heap, nr, p, &p->data)) + if (nouveau_resource_alloc(heap, p->immd_nr, p, + &p->data[0])) assert(0); } + + /* immediates only need to be uploaded again when freed */ + nv50_program_upload_data(nv50, p->immd, p->data[0]->start, + p->immd_nr, NV50_CB_PMISC); } - if (p->param_nr) { - float *map = ws->buffer_map(ws, nv50->constbuf[p->type], - PIPE_BUFFER_USAGE_CPU_READ); - nv50_program_upload_data(nv50, map, p->data->start, - p->param_nr); - ws->buffer_unmap(ws, nv50->constbuf[p->type]); + if (!p->data[1] && p->param_nr) { + struct nouveau_resource *heap = + nv50->screen->parm_heap[p->type]; + + if (nouveau_resource_alloc(heap, p->param_nr, p, &p->data[1])) { + while (heap->next && heap->size < p->param_nr) { + struct nv50_program *evict = heap->next->priv; + nouveau_resource_free(&evict->data[1]); + } + + if (nouveau_resource_alloc(heap, p->param_nr, p, + &p->data[1])) + assert(0); + } } - if (p->immd_nr) { - nv50_program_upload_data(nv50, p->immd, - p->data->start + p->param_nr, - p->immd_nr); + if (p->param_nr) { + unsigned cbuf = NV50_CB_PVP; + float *map = pipe_buffer_map(pscreen, nv50->constbuf[p->type], + PIPE_BUFFER_USAGE_CPU_READ); + if (p->type == PIPE_SHADER_FRAGMENT) + cbuf = NV50_CB_PFP; + nv50_program_upload_data(nv50, map, p->data[1]->start, + p->param_nr, cbuf); + pipe_buffer_unmap(pscreen, nv50->constbuf[p->type]); } } static void nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) { - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->base.channel; struct nouveau_grobj *tesla = nv50->screen->tesla; - struct pipe_screen *screen = nv50->pipe.screen; struct nv50_program_exec *e; struct nouveau_stateobj *so; const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR; unsigned start, count, *up, *ptr; boolean upload = FALSE; - if (!p->buffer) { - p->buffer = screen->buffer_create(screen, 0x100, 0, p->exec_size * 4); + if (!p->bo) { + nouveau_bo_new(chan->device, NOUVEAU_BO_VRAM, 0x100, + p->exec_size * 4, &p->bo); upload = TRUE; } - if (p->data && p->data->start != p->data_start) { + if ((p->data[0] && p->data[0]->start != p->data_start[0]) || + (p->data[1] && p->data[1]->start != p->data_start[1])) { for (e = p->exec_head; e; e = e->next) { - unsigned ei, ci; + unsigned ei, ci, bs; if (e->param.index < 0) continue; + bs = (e->inst[1] >> 22) & 0x07; + assert(bs < 2); ei = e->param.shift >> 5; - ci = e->param.index + p->data->start; + ci = e->param.index + p->data[bs]->start; e->inst[ei] &= ~e->param.mask; e->inst[ei] |= (ci << e->param.shift); } - p->data_start = p->data->start; + if (p->data[0]) + p->data_start[0] = p->data[0]->start; + if (p->data[1]) + p->data_start[1] = p->data[1]->start; + upload = TRUE; } @@ -1637,13 +2330,11 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) #ifdef NV50_PROGRAM_DUMP NOUVEAU_ERR("-------\n"); - up = ptr = MALLOC(p->exec_size * 4); for (e = p->exec_head; e; e = e->next) { NOUVEAU_ERR("0x%08x\n", e->inst[0]); if (is_long(e)) NOUVEAU_ERR("0x%08x\n", e->inst[1]); } - #endif up = ptr = MALLOC(p->exec_size * 4); @@ -1655,20 +2346,20 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) so = so_new(4,2); so_method(so, nv50->screen->tesla, 0x1280, 3); - so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_HIGH, 0, 0); - so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4)); start = 0; count = p->exec_size; while (count) { - struct nouveau_winsys *nvws = nv50->screen->nvws; + struct nouveau_channel *chan = nv50->screen->base.channel; unsigned nr; - so_emit(nvws, so); + so_emit(chan, so); nr = MIN2(count, 2047); - nr = MIN2(nvws->channel->pushbuf->remaining, nr); - if (nvws->channel->pushbuf->remaining < (nr + 3)) { + nr = MIN2(chan->pushbuf->remaining, nr); + if (chan->pushbuf->remaining < (nr + 3)) { FIRE_RING(chan); continue; } @@ -1704,10 +2395,10 @@ nv50_vertprog_validate(struct nv50_context *nv50) so = so_new(13, 2); so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2); - so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | - NOUVEAU_BO_HIGH, 0, 0); - so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | - NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_LOW, 0, 0); so_method(so, tesla, 0x1650, 2); so_data (so, p->cfg.vp.attr[0]); so_data (so, p->cfg.vp.attr[1]); @@ -1728,6 +2419,7 @@ nv50_fragprog_validate(struct nv50_context *nv50) struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_program *p = nv50->fragprog; struct nouveau_stateobj *so; + unsigned i; if (!p->translated) { nv50_program_validate(nv50, p); @@ -1740,22 +2432,27 @@ nv50_fragprog_validate(struct nv50_context *nv50) so = so_new(64, 2); so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2); - so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | - NOUVEAU_BO_HIGH, 0, 0); - so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | - NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | + NOUVEAU_BO_LOW, 0, 0); so_method(so, tesla, 0x1904, 4); - so_data (so, 0x00040404); /* p: 0x01000404 */ + so_data (so, p->cfg.fp.regs[0]); /* 0x01000404 / 0x00040404 */ so_data (so, 0x00000004); so_data (so, 0x00000000); so_data (so, 0x00000000); - so_method(so, tesla, 0x16bc, 3); /*XXX: fixme */ - so_data (so, 0x03020100); - so_data (so, 0x07060504); - so_data (so, 0x0b0a0908); + so_method(so, tesla, 0x16bc, p->cfg.fp.high_map); + for (i = 0; i < p->cfg.fp.high_map; i++) + so_data(so, p->cfg.fp.map[i]); so_method(so, tesla, 0x1988, 2); - so_data (so, 0x08080408); //0x08040404); /* p: 0x0f000401 */ + so_data (so, p->cfg.fp.regs[1]); /* 0x08040404 / 0x0f000401 */ so_data (so, p->cfg.high_temp); + so_method(so, tesla, 0x1298, 1); + so_data (so, p->cfg.high_result); + so_method(so, tesla, 0x19a8, 1); + so_data (so, p->cfg.fp.regs[2]); + so_method(so, tesla, 0x196c, 1); + so_data (so, p->cfg.fp.regs[3]); so_method(so, tesla, 0x1414, 1); so_data (so, 0); /* program start offset */ so_ref(so, &nv50->state.fragprog); @@ -1765,8 +2462,6 @@ nv50_fragprog_validate(struct nv50_context *nv50) void nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) { - struct pipe_screen *pscreen = nv50->pipe.screen; - while (p->exec_head) { struct nv50_program_exec *e = p->exec_head; @@ -1776,10 +2471,10 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) p->exec_tail = NULL; p->exec_size = 0; - if (p->buffer) - pipe_buffer_reference(&p->buffer, NULL); + nouveau_bo_ref(NULL, &p->bo); - nv50->screen->nvws->res_free(&p->data); + nouveau_resource_free(&p->data[0]); + nouveau_resource_free(&p->data[1]); p->translated = 0; } diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h index 78deed6a38..096e0476aa 100644 --- a/src/gallium/drivers/nv50/nv50_program.h +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -24,10 +24,10 @@ struct nv50_program { struct nv50_program_exec *exec_head; struct nv50_program_exec *exec_tail; unsigned exec_size; - struct nouveau_resource *data; - unsigned data_start; + struct nouveau_resource *data[2]; + unsigned data_start[2]; - struct pipe_buffer *buffer; + struct nouveau_bo *bo; float *immd; unsigned immd_nr; @@ -39,6 +39,11 @@ struct nv50_program { struct { unsigned attr[2]; } vp; + struct { + unsigned regs[4]; + unsigned map[5]; + unsigned high_map; + } fp; } cfg; }; diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 35cebdbdc3..940e04365f 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -26,7 +26,7 @@ #include "nv50_context.h" struct nv50_query { - struct pipe_buffer *buffer; + struct nouveau_bo *bo; unsigned type; boolean ready; uint64_t result; @@ -41,14 +41,16 @@ nv50_query(struct pipe_query *pipe) static struct pipe_query * nv50_query_create(struct pipe_context *pipe, unsigned type) { - struct pipe_screen *screen = pipe->screen; + struct nouveau_device *dev = nouveau_screen(pipe->screen)->device; struct nv50_query *q = CALLOC_STRUCT(nv50_query); + int ret; assert (q->type == PIPE_QUERY_OCCLUSION_COUNTER); q->type = type; - q->buffer = screen->buffer_create(screen, 256, 0, 16); - if (!q->buffer) { + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM | NOUVEAU_BO_MAP, 256, + 16, &q->bo); + if (ret) { FREE(q); return NULL; } @@ -62,7 +64,7 @@ nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq) struct nv50_query *q = nv50_query(pq); if (q) { - pipe_buffer_reference(&q->buffer, NULL); + nouveau_bo_ref(NULL, &q->bo); FREE(q); } } @@ -71,7 +73,7 @@ static void nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->base.channel; struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_query *q = nv50_query(pq); @@ -87,15 +89,14 @@ static void nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->base.channel; struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_query *q = nv50_query(pq); - struct nouveau_bo *bo = nv50->screen->nvws->get_bo(q->buffer); WAIT_RING (chan, 5); BEGIN_RING(chan, tesla, 0x1b00, 4); - OUT_RELOCh(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RELOCl(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RING (chan, 0x00000000); OUT_RING (chan, 0x0100f002); FIRE_RING (chan); @@ -105,7 +106,6 @@ static boolean nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq, boolean wait, uint64_t *result) { - struct pipe_winsys *ws = pipe->winsys; struct nv50_query *q = nv50_query(pq); /*XXX: Want to be able to return FALSE here instead of blocking @@ -113,11 +113,10 @@ nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq, */ if (!q->ready) { - uint32_t *map = ws->buffer_map(ws, q->buffer, - PIPE_BUFFER_USAGE_CPU_READ); - q->result = map[1]; + nouveau_bo_map(q->bo, NOUVEAU_BO_RD); + q->result = ((uint32_t *)q->bo->map)[1]; q->ready = TRUE; - ws->buffer_unmap(ws, q->buffer); + nouveau_bo_unmap(q->bo); } *result = q->result; diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index 2980564594..fd39fa738b 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -22,8 +22,6 @@ #include "pipe/p_screen.h" -#include "util/u_simple_screen.h" - #include "nv50_context.h" #include "nv50_screen.h" @@ -68,23 +66,6 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen, return FALSE; } -static const char * -nv50_screen_get_name(struct pipe_screen *pscreen) -{ - struct nv50_screen *screen = nv50_screen(pscreen); - struct nouveau_device *dev = screen->nvws->channel->device; - static char buffer[128]; - - snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset); - return buffer; -} - -static const char * -nv50_screen_get_vendor(struct pipe_screen *pscreen) -{ - return "nouveau"; -} - static int nv50_screen_get_param(struct pipe_screen *pscreen, int param) { @@ -153,37 +134,64 @@ nv50_screen_get_paramf(struct pipe_screen *pscreen, int param) static void nv50_screen_destroy(struct pipe_screen *pscreen) { - FREE(pscreen); + struct nv50_screen *screen = nv50_screen(pscreen); + + nouveau_notifier_free(&screen->sync); + nouveau_grobj_free(&screen->tesla); + nouveau_grobj_free(&screen->eng2d); + nouveau_grobj_free(&screen->m2mf); + nouveau_screen_fini(&screen->base); + FREE(screen); } struct pipe_screen * -nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) +nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) { struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen); + struct nouveau_channel *chan; + struct pipe_screen *pscreen; struct nouveau_stateobj *so; - unsigned tesla_class = 0, ret; - unsigned chipset = nvws->channel->device->chipset; - int i; + unsigned chipset = dev->chipset; + unsigned tesla_class = 0; + int ret, i; if (!screen) return NULL; - screen->nvws = nvws; + pscreen = &screen->base.base; + + ret = nouveau_screen_init(&screen->base, dev); + if (ret) { + nv50_screen_destroy(pscreen); + return NULL; + } + chan = screen->base.channel; + + pscreen->winsys = ws; + pscreen->destroy = nv50_screen_destroy; + pscreen->get_param = nv50_screen_get_param; + pscreen->get_paramf = nv50_screen_get_paramf; + pscreen->is_format_supported = nv50_screen_is_format_supported; + + nv50_screen_init_miptree_functions(pscreen); + nv50_transfer_init_screen_functions(pscreen); /* DMA engine object */ - ret = nvws->grobj_alloc(nvws, 0x5039, &screen->m2mf); + ret = nouveau_grobj_alloc(chan, 0xbeef5039, 0x5039, &screen->m2mf); if (ret) { NOUVEAU_ERR("Error creating M2MF object: %d\n", ret); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } + BIND_RING(chan, screen->m2mf, 1); /* 2D object */ - ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d); + ret = nouveau_grobj_alloc(chan, 0xbeef502d, 0x502d, &screen->eng2d); if (ret) { NOUVEAU_ERR("Error creating 2D object: %d\n", ret); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } + BIND_RING(chan, screen->eng2d, 2); /* 3D object */ switch (chipset & 0xf0) { @@ -199,70 +207,55 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) break; default: NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", chipset); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } if (tesla_class == 0) { NOUVEAU_ERR("Unknown G8x chipset: NV%02x\n", chipset); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } - ret = nvws->grobj_alloc(nvws, tesla_class, &screen->tesla); + ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class, &screen->tesla); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } + BIND_RING(chan, screen->tesla, 3); /* Sync notifier */ - ret = nvws->notifier_alloc(nvws, 1, &screen->sync); + ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); if (ret) { NOUVEAU_ERR("Error creating notifier object: %d\n", ret); - nv50_screen_destroy(&screen->pipe); + nv50_screen_destroy(pscreen); return NULL; } - /* Setup the pipe */ - screen->pipe.winsys = ws; - - screen->pipe.destroy = nv50_screen_destroy; - - screen->pipe.get_name = nv50_screen_get_name; - screen->pipe.get_vendor = nv50_screen_get_vendor; - screen->pipe.get_param = nv50_screen_get_param; - screen->pipe.get_paramf = nv50_screen_get_paramf; - - screen->pipe.is_format_supported = nv50_screen_is_format_supported; - - nv50_screen_init_miptree_functions(&screen->pipe); - nv50_transfer_init_screen_functions(&screen->pipe); - u_simple_screen_init(&screen->pipe); - /* Static M2MF init */ so = so_new(32, 0); so_method(so, screen->m2mf, 0x0180, 3); so_data (so, screen->sync->handle); - so_data (so, screen->nvws->channel->vram->handle); - so_data (so, screen->nvws->channel->vram->handle); - so_emit(nvws, so); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); + so_emit(chan, so); so_ref (NULL, &so); /* Static 2D init */ so = so_new(64, 0); so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4); so_data (so, screen->sync->handle); - so_data (so, screen->nvws->channel->vram->handle); - so_data (so, screen->nvws->channel->vram->handle); - so_data (so, screen->nvws->channel->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); + so_data (so, chan->vram->handle); so_method(so, screen->eng2d, NV50_2D_OPERATION, 1); so_data (so, NV50_2D_OPERATION_SRCCOPY); so_method(so, screen->eng2d, 0x0290, 1); so_data (so, 0); so_method(so, screen->eng2d, 0x0888, 1); so_data (so, 1); - so_emit(nvws, so); + so_emit(chan, so); so_ref(NULL, &so); /* Static tesla init */ @@ -275,11 +268,11 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0), NV50TCL_DMA_UNK0__SIZE); for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++) - so_data(so, nvws->channel->vram->handle); + so_data(so, chan->vram->handle); so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0), NV50TCL_DMA_UNK1__SIZE); for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++) - so_data(so, nvws->channel->vram->handle); + so_data(so, chan->vram->handle); so_method(so, screen->tesla, 0x121c, 1); so_data (so, 1); @@ -290,27 +283,81 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->tesla, 0x16b8, 1); so_data (so, 8); - /* Shared constant buffer */ - screen->constbuf = screen->pipe.buffer_create(&screen->pipe, 0, 0, 128 * 4 * 4); - if (nvws->res_init(&screen->vp_data_heap, 0, 128)) { - NOUVEAU_ERR("Error initialising constant buffer\n"); - nv50_screen_destroy(&screen->pipe); + /* constant buffers for immediates and VP/FP parameters */ + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4, + &screen->constbuf_misc[0]); + if (ret) { + nv50_screen_destroy(pscreen); return NULL; } + for (i = 0; i < 2; i++) { + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4, + &screen->constbuf_parm[i]); + if (ret) { + nv50_screen_destroy(pscreen); + return NULL; + } + } + + if (nouveau_resource_init(&screen->immd_heap[0], 0, 128) || + nouveau_resource_init(&screen->parm_heap[0], 0, 128) || + nouveau_resource_init(&screen->parm_heap[1], 0, 128)) + { + NOUVEAU_ERR("Error initialising constant buffers.\n"); + nv50_screen_destroy(pscreen); + return NULL; + } + + /* + // map constant buffers: + // B = buffer ID (maybe more than 1 byte) + // N = CB index used in shader instruction + // P = program type (0 = VP, 2 = GP, 3 = FP) + so_method(so, screen->tesla, 0x1694, 1); + so_data (so, 0x000BBNP1); + */ + so_method(so, screen->tesla, 0x1280, 3); - so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM | + so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); - so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM | + so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); - so_data (so, (NV50_CB_PMISC << 16) | 0x00001000); + so_data (so, (NV50_CB_PMISC << 16) | 0x00000800); + so_method(so, screen->tesla, 0x1694, 1); + so_data (so, 0x00000001 | (NV50_CB_PMISC << 12)); + so_method(so, screen->tesla, 0x1694, 1); + so_data (so, 0x00000031 | (NV50_CB_PMISC << 12)); + + so_method(so, screen->tesla, 0x1280, 3); + so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_PVP << 16) | 0x00000800); + so_method(so, screen->tesla, 0x1694, 1); + so_data (so, 0x00000101 | (NV50_CB_PVP << 12)); + + so_method(so, screen->tesla, 0x1280, 3); + so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM | + NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + so_data (so, (NV50_CB_PFP << 16) | 0x00000800); + so_method(so, screen->tesla, 0x1694, 1); + so_data (so, 0x00000131 | (NV50_CB_PFP << 12)); /* Texture sampler/image unit setup - we abuse the constant buffer * upload mechanism for the moment to upload data to the tex config * blocks. At some point we *may* want to go the NVIDIA way of doing * things? */ - screen->tic = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4); + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tic); + if (ret) { + nv50_screen_destroy(pscreen); + return NULL; + } + so_method(so, screen->tesla, 0x1280, 3); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -324,7 +371,12 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, 0x00000800); - screen->tsc = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4); + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tsc); + if (ret) { + nv50_screen_destroy(pscreen); + return NULL; + } + so_method(so, screen->tesla, 0x1280, 3); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -352,14 +404,12 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws) so_method(so, screen->tesla, 0x1234, 1); so_data (so, 1); - so_method(so, screen->tesla, 0x1458, 1); - so_data (so, 1); - so_emit(nvws, so); + so_emit(chan, so); so_ref (so, &screen->static_init); so_ref (NULL, &so); - nvws->push_flush(nvws, 0, NULL); + nouveau_pushbuf_flush(chan, 0); - return &screen->pipe; + return pscreen; } diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h index db567aaac8..61e24a5b57 100644 --- a/src/gallium/drivers/nv50/nv50_screen.h +++ b/src/gallium/drivers/nv50/nv50_screen.h @@ -1,10 +1,10 @@ #ifndef __NV50_SCREEN_H__ #define __NV50_SCREEN_H__ -#include "pipe/p_screen.h" +#include "nouveau/nouveau_screen.h" struct nv50_screen { - struct pipe_screen pipe; + struct nouveau_screen base; struct nouveau_winsys *nvws; @@ -15,11 +15,14 @@ struct nv50_screen { struct nouveau_grobj *m2mf; struct nouveau_notifier *sync; - struct pipe_buffer *constbuf; - struct nouveau_resource *vp_data_heap; + struct nouveau_bo *constbuf_misc[1]; + struct nouveau_bo *constbuf_parm[2]; - struct pipe_buffer *tic; - struct pipe_buffer *tsc; + struct nouveau_resource *immd_heap[1]; + struct nouveau_resource *parm_heap[2]; + + struct nouveau_bo *tic; + struct nouveau_bo *tsc; struct nouveau_stateobj *static_init; }; diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index ba852194cd..116866a8e7 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -136,9 +136,11 @@ static void * nv50_sampler_state_create(struct pipe_context *pipe, const struct pipe_sampler_state *cso) { - unsigned *tsc = CALLOC(8, sizeof(unsigned)); + struct nv50_sampler_stateobj *sso = CALLOC(1, sizeof(*sso)); + unsigned *tsc = sso->tsc; + float limit; - tsc[0] = (0x00024000 | + tsc[0] = (0x00026000 | (wrap_mode(cso->wrap_s) << 0) | (wrap_mode(cso->wrap_t) << 3) | (wrap_mode(cso->wrap_r) << 6)); @@ -202,7 +204,14 @@ nv50_sampler_state_create(struct pipe_context *pipe, tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7); } - return (void *)tsc; + limit = CLAMP(cso->lod_bias, -16.0, 15.0); + tsc[1] |= ((int)(limit * 256.0) & 0x1fff) << 11; + + tsc[2] |= ((int)CLAMP(cso->max_lod, 0.0, 15.0) << 20) | + ((int)CLAMP(cso->min_lod, 0.0, 15.0) << 8); + + sso->normalized = cso->normalized_coords; + return (void *)sso; } static void diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index c13d3de1cb..0caf4b4e91 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -32,6 +32,9 @@ nv50_state_validate_fb(struct nv50_context *nv50) unsigned i, w, h, gw = 0; for (i = 0; i < fb->nr_cbufs; i++) { + struct pipe_texture *pt = fb->cbufs[i]->texture; + struct nouveau_bo *bo = nv50_miptree(pt)->bo; + if (!gw) { w = fb->cbufs[i]->width; h = fb->cbufs[i]->height; @@ -46,12 +49,10 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_data (so, fb->cbufs[i]->height); so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5); - so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset, - NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH | - NOUVEAU_BO_RDWR, 0, 0); - so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset, - NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | - NOUVEAU_BO_RDWR, 0, 0); + so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | + NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); + so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | + NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); switch (fb->cbufs[i]->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: so_data(so, 0xcf); @@ -65,7 +66,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_data(so, 0xe6); break; } - so_data(so, 0x00000000); + so_data(so, bo->tile_mode << 4); so_data(so, 0x00000000); so_method(so, tesla, 0x1224, 1); @@ -73,6 +74,9 @@ nv50_state_validate_fb(struct nv50_context *nv50) } if (fb->zsbuf) { + struct pipe_texture *pt = fb->zsbuf->texture; + struct nouveau_bo *bo = nv50_miptree(pt)->bo; + if (!gw) { w = fb->zsbuf->width; h = fb->zsbuf->height; @@ -83,12 +87,10 @@ nv50_state_validate_fb(struct nv50_context *nv50) } so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5); - so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset, - NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH | - NOUVEAU_BO_RDWR, 0, 0); - so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset, - NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | - NOUVEAU_BO_RDWR, 0, 0); + so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | + NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); + so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | + NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); switch (fb->zsbuf->format) { case PIPE_FORMAT_Z24S8_UNORM: so_data(so, 0x16); @@ -102,7 +104,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_data(so, 0x16); break; } - so_data(so, 0x00000000); + so_data(so, bo->tile_mode << 4); so_data(so, 0x00000000); so_method(so, tesla, 0x1538, 1); @@ -131,7 +133,7 @@ static void nv50_state_emit(struct nv50_context *nv50) { struct nv50_screen *screen = nv50->screen; - struct nouveau_winsys *nvws = screen->nvws; + struct nouveau_channel *chan = screen->base.channel; if (nv50->pctx_id != screen->cur_pctx) { nv50->state.dirty |= 0xffffffff; @@ -139,40 +141,40 @@ nv50_state_emit(struct nv50_context *nv50) } if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER) - so_emit(nvws, nv50->state.fb); + so_emit(chan, nv50->state.fb); if (nv50->state.dirty & NV50_NEW_BLEND) - so_emit(nvws, nv50->state.blend); + so_emit(chan, nv50->state.blend); if (nv50->state.dirty & NV50_NEW_ZSA) - so_emit(nvws, nv50->state.zsa); + so_emit(chan, nv50->state.zsa); if (nv50->state.dirty & NV50_NEW_VERTPROG) - so_emit(nvws, nv50->state.vertprog); + so_emit(chan, nv50->state.vertprog); if (nv50->state.dirty & NV50_NEW_FRAGPROG) - so_emit(nvws, nv50->state.fragprog); + so_emit(chan, nv50->state.fragprog); if (nv50->state.dirty & NV50_NEW_RASTERIZER) - so_emit(nvws, nv50->state.rast); + so_emit(chan, nv50->state.rast); if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR) - so_emit(nvws, nv50->state.blend_colour); + so_emit(chan, nv50->state.blend_colour); if (nv50->state.dirty & NV50_NEW_STIPPLE) - so_emit(nvws, nv50->state.stipple); + so_emit(chan, nv50->state.stipple); if (nv50->state.dirty & NV50_NEW_SCISSOR) - so_emit(nvws, nv50->state.scissor); + so_emit(chan, nv50->state.scissor); if (nv50->state.dirty & NV50_NEW_VIEWPORT) - so_emit(nvws, nv50->state.viewport); + so_emit(chan, nv50->state.viewport); if (nv50->state.dirty & NV50_NEW_SAMPLER) - so_emit(nvws, nv50->state.tsc_upload); + so_emit(chan, nv50->state.tsc_upload); if (nv50->state.dirty & NV50_NEW_TEXTURE) - so_emit(nvws, nv50->state.tic_upload); + so_emit(chan, nv50->state.tic_upload); if (nv50->state.dirty & NV50_NEW_ARRAYS) { - so_emit(nvws, nv50->state.vtxfmt); - so_emit(nvws, nv50->state.vtxbuf); + so_emit(chan, nv50->state.vtxfmt); + so_emit(chan, nv50->state.vtxbuf); } nv50->state.dirty = 0; - so_emit_reloc_markers(nvws, nv50->state.fb); - so_emit_reloc_markers(nvws, nv50->state.vertprog); - so_emit_reloc_markers(nvws, nv50->state.fragprog); - so_emit_reloc_markers(nvws, nv50->state.vtxbuf); - so_emit_reloc_markers(nvws, nv50->screen->static_init); + so_emit_reloc_markers(chan, nv50->state.fb); + so_emit_reloc_markers(chan, nv50->state.vertprog); + so_emit_reloc_markers(chan, nv50->state.fragprog); + so_emit_reloc_markers(chan, nv50->state.vtxbuf); + so_emit_reloc_markers(chan, nv50->screen->static_init); } boolean @@ -293,12 +295,12 @@ viewport_uptodate: so_data (so, NV50_CB_TSC); so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8); for (i = 0; i < nv50->sampler_nr; i++) - so_datap (so, nv50->sampler[i], 8); + so_datap (so, nv50->sampler[i]->tsc, 8); so_ref(so, &nv50->state.tsc_upload); so_ref(NULL, &so); } - if (nv50->dirty & NV50_NEW_TEXTURE) + if (nv50->dirty & (NV50_NEW_TEXTURE | NV50_NEW_SAMPLER)) nv50_tex_validate(nv50); if (nv50->dirty & NV50_NEW_ARRAYS) diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 0cc5168144..8db3b6d344 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -52,21 +52,17 @@ static int nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) { struct nv50_miptree *mt = nv50_miptree(ps->texture); - struct nouveau_channel *chan = screen->nvws->channel; + struct nouveau_channel *chan = screen->eng2d->channel; struct nouveau_grobj *eng2d = screen->eng2d; - struct nouveau_bo *bo; + struct nouveau_bo *bo = nv50_miptree(ps->texture)->bo; int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT; int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD); - - bo = screen->nvws->get_bo(nv50_miptree(ps->texture)->buffer); - if (!bo) - return 1; format = nv50_format(ps->format); if (format < 0) return 1; - if (!bo->tiled) { + if (!bo->tile_flags) { BEGIN_RING(chan, eng2d, mthd, 2); OUT_RING (chan, format); OUT_RING (chan, 1); @@ -80,7 +76,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) BEGIN_RING(chan, eng2d, mthd, 5); OUT_RING (chan, format); OUT_RING (chan, 0); - OUT_RING (chan, 0); + OUT_RING (chan, bo->tile_mode << 4); OUT_RING (chan, 1); OUT_RING (chan, 0); BEGIN_RING(chan, eng2d, mthd + 0x18, 4); @@ -108,7 +104,7 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst, int dx, int dy, struct pipe_surface *src, int sx, int sy, int w, int h) { - struct nouveau_channel *chan = screen->nvws->channel; + struct nouveau_channel *chan = screen->eng2d->channel; struct nouveau_grobj *eng2d = screen->eng2d; int ret; @@ -165,7 +161,7 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, { struct nv50_context *nv50 = (struct nv50_context *)pipe; struct nv50_screen *screen = nv50->screen; - struct nouveau_channel *chan = screen->nvws->channel; + struct nouveau_channel *chan = screen->eng2d->channel; struct nouveau_grobj *eng2d = screen->eng2d; int format, ret; diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 223c8a3a45..ff40c2ad81 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -26,7 +26,8 @@ #include "nouveau/nouveau_stateobj.h" static int -nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt) +nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, + struct nv50_miptree *mt, int unit) { switch (mt->base.format) { case PIPE_FORMAT_A8R8G8B8_UNORM: @@ -117,15 +118,18 @@ nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt) return 1; } - so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | + so_reloc(so, mt->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | NOUVEAU_BO_RD, 0, 0); - so_data (so, 0xd0005000); + if (nv50->sampler[unit]->normalized) + so_data (so, 0xd0005000 | mt->bo->tile_mode << 22); + else + so_data (so, 0x5001d000 | mt->bo->tile_mode << 22); so_data (so, 0x00300000); so_data (so, mt->base.width[0]); - so_data (so, (mt->base.depth[0] << 16) | mt->base.height[0]); + so_data (so, (mt->base.last_level << 28) | + (mt->base.depth[0] << 16) | mt->base.height[0]); so_data (so, 0x03000000); - so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH | - NOUVEAU_BO_RD, 0, 0); + so_data (so, mt->base.last_level << 4); return 0; } @@ -135,23 +139,35 @@ nv50_tex_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; struct nouveau_stateobj *so; - int unit; + int unit, push; + + push = nv50->miptree_nr * 9 + 2; + push += MAX2(nv50->miptree_nr, nv50->state.miptree_nr) * 2; - so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2); + so = so_new(push, nv50->miptree_nr * 2); so_method(so, tesla, 0x0f00, 1); so_data (so, NV50_CB_TIC); - so_method(so, tesla, 0x40000f04, nv50->miptree_nr * 8); for (unit = 0; unit < nv50->miptree_nr; unit++) { struct nv50_miptree *mt = nv50->miptree[unit]; - if (nv50_tex_construct(so, mt)) { + so_method(so, tesla, 0x40000f04, 8); + if (nv50_tex_construct(nv50, so, mt, unit)) { NOUVEAU_ERR("failed tex validate\n"); so_ref(NULL, &so); return; } + + so_method(so, tesla, 0x1458, 1); + so_data (so, (unit << 9) | (unit << 1) | 1); + } + + for (; unit < nv50->state.miptree_nr; unit++) { + so_method(so, tesla, 0x1458, 1); + so_data (so, (unit << 1) | 0); } so_ref(so, &nv50->state.tic_upload); so_ref(NULL, &so); + nv50->state.miptree_nr = nv50->miptree_nr; } diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index 747195b4f6..d0b7f0bef4 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -6,8 +6,8 @@ struct nv50_transfer { struct pipe_transfer base; - struct pipe_buffer *buffer; - struct nv50_miptree_level *level; + struct nouveau_bo *bo; + unsigned level_offset; int level_pitch; int level_width; int level_height; @@ -16,51 +16,48 @@ struct nv50_transfer { }; static void -nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src, - int src_pitch, int sx, int sy, int sw, int sh, - struct pipe_buffer *dst, int dst_pitch, int dx, int dy, +nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, + unsigned src_offset, int src_pitch, int sx, int sy, + int sw, int sh, struct nouveau_bo *dst_bo, + unsigned dst_offset, int dst_pitch, int dx, int dy, int dw, int dh, int cpp, int width, int height, unsigned src_reloc, unsigned dst_reloc) { struct nv50_screen *screen = nv50_screen(pscreen); - struct nouveau_winsys *nvws = screen->nvws; - struct nouveau_channel *chan = nvws->channel; + struct nouveau_channel *chan = screen->m2mf->channel; struct nouveau_grobj *m2mf = screen->m2mf; - struct nouveau_bo *src_bo = nvws->get_bo(src); - struct nouveau_bo *dst_bo = nvws->get_bo(dst); - unsigned src_offset = 0, dst_offset = 0; src_reloc |= NOUVEAU_BO_RD; dst_reloc |= NOUVEAU_BO_WR; WAIT_RING (chan, 14); - if (!src_bo->tiled) { + if (!src_bo->tile_flags) { BEGIN_RING(chan, m2mf, 0x0200, 1); OUT_RING (chan, 1); BEGIN_RING(chan, m2mf, 0x0314, 1); OUT_RING (chan, src_pitch); - src_offset = (sy * src_pitch) + (sx * cpp); + src_offset += (sy * src_pitch) + (sx * cpp); } else { BEGIN_RING(chan, m2mf, 0x0200, 6); OUT_RING (chan, 0); - OUT_RING (chan, 0); + OUT_RING (chan, src_bo->tile_mode << 4); OUT_RING (chan, sw * cpp); OUT_RING (chan, sh); OUT_RING (chan, 1); OUT_RING (chan, 0); } - if (!dst_bo->tiled) { + if (!dst_bo->tile_flags) { BEGIN_RING(chan, m2mf, 0x021c, 1); OUT_RING (chan, 1); BEGIN_RING(chan, m2mf, 0x0318, 1); OUT_RING (chan, dst_pitch); - dst_offset = (dy * dst_pitch) + (dx * cpp); + dst_offset += (dy * dst_pitch) + (dx * cpp); } else { BEGIN_RING(chan, m2mf, 0x021c, 6); OUT_RING (chan, 0); - OUT_RING (chan, 0); + OUT_RING (chan, dst_bo->tile_mode << 4); OUT_RING (chan, dw * cpp); OUT_RING (chan, dh); OUT_RING (chan, 1); @@ -77,13 +74,13 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src, BEGIN_RING(chan, m2mf, 0x030c, 2); OUT_RELOCl(chan, src_bo, src_offset, src_reloc); OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc); - if (src_bo->tiled) { + if (src_bo->tile_flags) { BEGIN_RING(chan, m2mf, 0x0218, 1); OUT_RING (chan, (dy << 16) | sx); } else { src_offset += (line_count * src_pitch); } - if (dst_bo->tiled) { + if (dst_bo->tile_flags) { BEGIN_RING(chan, m2mf, 0x0234, 1); OUT_RING (chan, (sy << 16) | dx); } else { @@ -108,10 +105,12 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, enum pipe_transfer_usage usage, unsigned x, unsigned y, unsigned w, unsigned h) { + struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct nv50_miptree *mt = nv50_miptree(pt); struct nv50_miptree_level *lvl = &mt->level[level]; struct nv50_transfer *tx; unsigned image = 0; + int ret; if (pt->target == PIPE_TEXTURE_CUBE) image = face; @@ -133,20 +132,24 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->base.stride = (w * pt->block.size); tx->base.usage = usage; - tx->level = lvl; tx->level_pitch = lvl->pitch; tx->level_width = mt->base.width[level]; tx->level_height = mt->base.height[level]; + tx->level_offset = lvl->image_offset[image]; tx->level_x = x; tx->level_y = y; - tx->buffer = - pipe_buffer_create(pscreen, 0, NOUVEAU_BUFFER_USAGE_TRANSFER, - w * tx->base.block.size * h); + ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, + w * pt->block.size * h, &tx->bo); + if (ret) { + FREE(tx); + return NULL; + } if (usage != PIPE_TRANSFER_WRITE) { - nv50_transfer_rect_m2mf(pscreen, mt->buffer, tx->level_pitch, - x, y, tx->level_width, tx->level_height, - tx->buffer, tx->base.stride, 0, 0, + nv50_transfer_rect_m2mf(pscreen, mt->bo, tx->level_offset, + tx->level_pitch, x, y, tx->level_width, + tx->level_height, tx->bo, 0, + tx->base.stride, 0, 0, tx->base.width, tx->base.height, tx->base.block.size, w, h, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, @@ -164,17 +167,18 @@ nv50_transfer_del(struct pipe_transfer *ptx) if (ptx->usage != PIPE_TRANSFER_READ) { struct pipe_screen *pscreen = ptx->texture->screen; - nv50_transfer_rect_m2mf(pscreen, tx->buffer, tx->base.stride, + nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride, 0, 0, tx->base.width, tx->base.height, - mt->buffer, tx->level_pitch, - tx->level_x, tx->level_y, - tx->level_width, tx->level_height, - tx->base.block.size, tx->base.width, - tx->base.height, NOUVEAU_BO_GART, - NOUVEAU_BO_VRAM | NOUVEAU_BO_GART); + mt->bo, tx->level_offset, + tx->level_pitch, tx->level_x, + tx->level_y, tx->level_width, + tx->level_height, tx->base.block.size, + tx->base.width, tx->base.height, + NOUVEAU_BO_GART, NOUVEAU_BO_VRAM | + NOUVEAU_BO_GART); } - pipe_buffer_reference(&tx->buffer, NULL); + nouveau_bo_ref(NULL, &tx->bo); pipe_texture_reference(&ptx->texture, NULL); FREE(ptx); } @@ -184,13 +188,17 @@ nv50_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) { struct nv50_transfer *tx = (struct nv50_transfer *)ptx; unsigned flags = 0; + int ret; if (ptx->usage & PIPE_TRANSFER_WRITE) - flags |= PIPE_BUFFER_USAGE_CPU_WRITE; + flags |= NOUVEAU_BO_WR; if (ptx->usage & PIPE_TRANSFER_READ) - flags |= PIPE_BUFFER_USAGE_CPU_READ; + flags |= NOUVEAU_BO_RD; - return pipe_buffer_map(pscreen, tx->buffer, flags); + ret = nouveau_bo_map(tx->bo, flags); + if (ret) + return NULL; + return tx->bo->map; } static void @@ -198,7 +206,7 @@ nv50_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx) { struct nv50_transfer *tx = (struct nv50_transfer *)ptx; - pipe_buffer_unmap(pscreen, tx->buffer); + nouveau_bo_unmap(tx->bo); } void diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 0749c90691..f81929f238 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -22,6 +22,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "nv50_context.h" @@ -53,7 +54,7 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; nv50_state_validate(nv50); @@ -83,7 +84,7 @@ static INLINE void nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; map += start; @@ -112,7 +113,7 @@ static INLINE void nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; map += start; @@ -141,7 +142,7 @@ static INLINE void nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map, unsigned start, unsigned count) { - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; map += start; @@ -163,10 +164,12 @@ nv50_draw_elements(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->nvws->channel; + struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; - struct pipe_winsys *ws = pipe->winsys; - void *map = ws->buffer_map(ws, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); + struct pipe_screen *pscreen = pipe->screen; + void *map; + + map = pipe_buffer_map(pscreen, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); nv50_state_validate(nv50); @@ -193,6 +196,7 @@ nv50_draw_elements(struct pipe_context *pipe, BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1); OUT_RING (chan, 0); + pipe_buffer_unmap(pscreen, indexBuffer); pipe->flush(pipe, 0, NULL); return TRUE; } @@ -212,6 +216,7 @@ nv50_vbo_validate(struct nv50_context *nv50) struct pipe_vertex_element *ve = &nv50->vtxelt[i]; struct pipe_vertex_buffer *vb = &nv50->vtxbuf[ve->vertex_buffer_index]; + struct nouveau_bo *bo = nouveau_bo(vb->buffer); switch (ve->src_format) { case PIPE_FORMAT_R32G32B32A32_FLOAT: @@ -240,10 +245,10 @@ nv50_vbo_validate(struct nv50_context *nv50) so_method(vtxbuf, tesla, 0x900 + (i * 16), 3); so_data (vtxbuf, 0x20000000 | vb->stride); - so_reloc (vtxbuf, vb->buffer, vb->buffer_offset + + so_reloc (vtxbuf, bo, vb->buffer_offset + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); - so_reloc (vtxbuf, vb->buffer, vb->buffer_offset + + so_reloc (vtxbuf, bo, vb->buffer_offset + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); } diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c index 9d95ad918c..00fae8d26f 100644 --- a/src/gallium/drivers/r300/r300_chipset.c +++ b/src/gallium/drivers/r300/r300_chipset.c @@ -34,7 +34,6 @@ void r300_parse_chipset(struct r300_capabilities* caps) caps->is_r500 = FALSE; caps->num_vert_fpus = 4; - /* Note: These are not ordered by PCI ID. I leave that task to GCC, * which will perform the ordering while collating jump tables. Instead, * I've tried to group them according to capabilities and age. */ @@ -150,6 +149,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) caps->num_vert_fpus = 6; break; + case 0x4B48: case 0x4B49: case 0x4B4A: case 0x4B4B: @@ -349,7 +349,4 @@ void r300_parse_chipset(struct r300_capabilities* caps) caps->pci_id); break; } - - /* XXX SW TCL is broken so no forcing it off right now - caps->has_tcl = FALSE; */ } diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h index 21eebeae60..5b2e1f0568 100644 --- a/src/gallium/drivers/r300/r300_chipset.h +++ b/src/gallium/drivers/r300/r300_chipset.h @@ -34,8 +34,6 @@ struct r300_capabilities { int family; /* The number of vertex floating-point units */ int num_vert_fpus; - /* The number of fragment pipes */ - int num_frag_pipes; /* Whether or not TCL is physically present */ boolean has_tcl; /* Whether or not this is an RV515 or newer; R500s have many differences diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 31efe91417..233a32b53c 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -34,10 +34,6 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); int i; - if (r300->dirty_state) { - r300_emit_dirty_state(r300); - } - for (i = 0; i < r300->vertex_buffer_count; i++) { void* buf = pipe_buffer_map(pipe->screen, r300->vertex_buffers[i].buffer, @@ -102,6 +98,29 @@ static void r300_destroy_context(struct pipe_context* context) { FREE(r300); } +static unsigned int +r300_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + +static unsigned int +r300_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + /** + * FIXME: Optimize. + */ + + return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +} + struct pipe_context* r300_create_context(struct pipe_screen* screen, struct r300_winsys* r300_winsys) { @@ -110,7 +129,6 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, if (!r300) return NULL; - /* XXX this could be refactored now? */ r300->winsys = r300_winsys; r300->context.winsys = (struct pipe_winsys*)r300_winsys; @@ -124,8 +142,18 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.draw_elements = r300_draw_elements; r300->context.draw_range_elements = r300_draw_range_elements; + r300->context.is_texture_referenced = r300_is_texture_referenced; + r300->context.is_buffer_referenced = r300_is_buffer_referenced; + + /* Create a Draw. This is used for vert collation and SW TCL. */ r300->draw = draw_create(); + /* Enable our renderer. */ draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300)); + /* Disable Draw's clipping if TCL is present. */ + draw_set_driver_clipping(r300->draw, r300_screen(screen)->caps->has_tcl); + /* Force Draw to never do viewport transform, since (again) we can do + * transform in hardware, always. */ + draw_set_viewport_state(r300->draw, &r300_viewport_identity); r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); r300->rs_block = CALLOC_STRUCT(r300_rs_block); diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index fec2bad546..27bc7fd1a9 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -60,9 +60,14 @@ struct r300_dsa_state { }; struct r300_rs_state { - /* XXX icky as fucking hell */ + /* Draw-specific rasterizer state */ struct pipe_rasterizer_state rs; + /* Whether or not to enable the VTE. This is referenced at the very + * last moment during emission of VTE state, to decide whether or not + * the VTE should be used for transformation. */ + boolean enable_vte; + uint32_t vap_control_status; /* R300_VAP_CNTL_STATUS: 0x2140 */ uint32_t point_size; /* R300_GA_POINT_SIZE: 0x421c */ uint32_t point_minmax; /* R300_GA_POINT_MINMAX: 0x4230 */ @@ -112,23 +117,24 @@ struct r300_viewport_state { uint32_t vte_control; /* R300_VAP_VTE_CNTL: 0x20b0 */ }; -#define R300_NEW_BLEND 0x0000001 -#define R300_NEW_BLEND_COLOR 0x0000002 -#define R300_NEW_CONSTANTS 0x0000004 -#define R300_NEW_DSA 0x0000008 -#define R300_NEW_FRAMEBUFFERS 0x0000010 -#define R300_NEW_FRAGMENT_SHADER 0x0000020 -#define R300_NEW_RASTERIZER 0x0000040 -#define R300_NEW_RS_BLOCK 0x0000080 -#define R300_NEW_SAMPLER 0x0000100 -#define R300_ANY_NEW_SAMPLERS 0x000ff00 -#define R300_NEW_SCISSOR 0x0010000 -#define R300_NEW_TEXTURE 0x0020000 -#define R300_ANY_NEW_TEXTURES 0x1fe0000 -#define R300_NEW_VERTEX_FORMAT 0x2000000 -#define R300_NEW_VERTEX_SHADER 0x4000000 -#define R300_NEW_VIEWPORT 0x8000000 -#define R300_NEW_KITCHEN_SINK 0xfffffff +#define R300_NEW_BLEND 0x00000001 +#define R300_NEW_BLEND_COLOR 0x00000002 +#define R300_NEW_CLIP 0x00000004 +#define R300_NEW_CONSTANTS 0x00000008 +#define R300_NEW_DSA 0x00000010 +#define R300_NEW_FRAMEBUFFERS 0x00000020 +#define R300_NEW_FRAGMENT_SHADER 0x00000040 +#define R300_NEW_RASTERIZER 0x00000080 +#define R300_NEW_RS_BLOCK 0x00000100 +#define R300_NEW_SAMPLER 0x00000200 +#define R300_ANY_NEW_SAMPLERS 0x0001fe00 +#define R300_NEW_SCISSOR 0x00020000 +#define R300_NEW_TEXTURE 0x00040000 +#define R300_ANY_NEW_TEXTURES 0x03fc0000 +#define R300_NEW_VERTEX_FORMAT 0x04000000 +#define R300_NEW_VERTEX_SHADER 0x08000000 +#define R300_NEW_VIEWPORT 0x10000000 +#define R300_NEW_KITCHEN_SINK 0x1fffffff /* The next several objects are not pure Radeon state; they inherit from * various Gallium classes. */ @@ -136,11 +142,11 @@ struct r300_viewport_state { struct r300_constant_buffer { /* Buffer of constants */ /* XXX first number should be raised */ - float constants[8][4]; + float constants[32][4]; /* Number of user-defined constants */ - int user_count; + unsigned user_count; /* Total number of constants */ - int count; + unsigned count; }; struct r3xx_fragment_shader { @@ -153,6 +159,10 @@ struct r3xx_fragment_shader { /* Pixel stack size */ int stack_size; + + /* Are there immediates in this shader? + * If not, we can heavily optimize recompilation. */ + boolean uses_imms; }; struct r300_fragment_shader { @@ -225,10 +235,11 @@ struct r300_vertex_format { uint32_t vap_prog_stream_cntl[8]; /* R300_VAP_PROG_STREAK_CNTL_EXT_[0-7] */ uint32_t vap_prog_stream_cntl_ext[8]; - /* This is a map of VAP/SW TCL outputs into the GA/RS. - * tab[i] is the location of input i in GA/RS input memory. - * Named tab for historical reasons. */ - int tab[16]; + /* Map of vertex attributes into PVS memory for HW TCL, + * or GA memory for SW TCL. */ + int vs_tab[16]; + /* Map of rasterizer attributes from GB through RS to US. */ + int fs_tab[16]; }; struct r300_vertex_shader { @@ -242,6 +253,10 @@ struct r300_vertex_shader { /* Has this shader been translated yet? */ boolean translated; + /* Are there immediates in this shader? + * If not, we can heavily optimize recompilation. */ + boolean uses_imms; + /* Number of used instructions */ int instruction_count; @@ -254,6 +269,11 @@ struct r300_vertex_shader { } instructions[128]; /*< XXX magic number */ }; +static struct pipe_viewport_state r300_viewport_identity = { + .scale = {1.0, 1.0, 1.0, 1.0}, + .translate = {0.0, 0.0, 0.0, 0.0}, +}; + struct r300_context { /* Parent class */ struct pipe_context context; @@ -263,11 +283,18 @@ struct r300_context { /* Draw module. Used mostly for SW TCL. */ struct draw_context* draw; + /* Vertex buffer for rendering. */ + struct pipe_buffer* vbo; + /* Offset into the VBO. */ + size_t vbo_offset; + /* Various CSO state objects. */ /* Blend state. */ struct r300_blend_state* blend_state; /* Blend color state. */ struct r300_blend_color_state* blend_color_state; + /* User clip planes. */ + struct pipe_clip_state clip_state; /* Shader constants. */ struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES]; /* Depth, stencil, and alpha state. */ @@ -288,7 +315,7 @@ struct r300_context { /* Texture states. */ struct r300_texture* textures[8]; int texture_count; - /* Vertex buffers. */ + /* Vertex buffers for Gallium. */ struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; int vertex_buffer_count; /* Vertex information. */ @@ -313,11 +340,4 @@ struct draw_stage* r300_draw_stage(struct r300_context* r300); void r300_init_state_functions(struct r300_context* r300); void r300_init_surface_functions(struct r300_context* r300); -/* Fun with includes: r300_winsys also declares this prototype. - * We'll just step out in that case... */ -#ifndef R300_WINSYS_H -struct pipe_context* r300_create_context(struct pipe_screen* screen, - struct r300_winsys* r300_winsys); -#endif - #endif /* R300_CONTEXT_H */ diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h index 5d9799dd72..71b142c0db 100644 --- a/src/gallium/drivers/r300/r300_cs.h +++ b/src/gallium/drivers/r300/r300_cs.h @@ -34,6 +34,7 @@ #define MAX_CS_SIZE 64 * 1024 / 4 +#define VERY_VERBOSE_CS 0 #define VERY_VERBOSE_REGISTERS 0 /* XXX stolen from radeon_drm.h */ @@ -49,27 +50,29 @@ #define CS_LOCALS(context) \ struct r300_winsys* cs_winsys = context->winsys; \ - struct radeon_cs* cs = cs_winsys->cs; \ int cs_count = 0; #define CHECK_CS(size) \ - cs_winsys->check_cs(cs, (size)) + cs_winsys->check_cs(cs_winsys, (size)) #define BEGIN_CS(size) do { \ CHECK_CS(size); \ - debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \ - size, __FUNCTION__, __FILE__, __LINE__); \ - cs_winsys->begin_cs(cs, (size), __FILE__, __FUNCTION__, __LINE__); \ + if (VERY_VERBOSE_CS) { \ + debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \ + size, __FUNCTION__, __FILE__, __LINE__); \ + } \ + cs_winsys->begin_cs(cs_winsys, (size), \ + __FILE__, __FUNCTION__, __LINE__); \ cs_count = size; \ } while (0) #define OUT_CS(value) do { \ - cs_winsys->write_cs_dword(cs, (value)); \ + cs_winsys->write_cs_dword(cs_winsys, (value)); \ cs_count--; \ } while (0) #define OUT_CS_32F(value) do { \ - cs_winsys->write_cs_dword(cs, fui(value)); \ + cs_winsys->write_cs_dword(cs_winsys, fui(value)); \ cs_count--; \ } while (0) @@ -93,26 +96,31 @@ } while (0) #define OUT_CS_RELOC(bo, offset, rd, wd, flags) do { \ - debug_printf("r300: writing relocation for buffer %p, offset %d\n", \ - bo, offset); \ + debug_printf("r300: writing relocation for buffer %p, offset %d, " \ + "domains (%d, %d, %d)\n", \ + bo, offset, rd, wd, flags); \ assert(bo); \ OUT_CS(offset); \ - cs_winsys->write_cs_reloc(cs, bo, rd, wd, flags); \ + cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \ cs_count -= 2; \ } while (0) #define END_CS do { \ - debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \ - __LINE__); \ + if (VERY_VERBOSE_CS) { \ + debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \ + __FILE__, __LINE__); \ + } \ if (cs_count != 0) \ debug_printf("r300: Warning: cs_count off by %d\n", cs_count); \ - cs_winsys->end_cs(cs, __FILE__, __FUNCTION__, __LINE__); \ + cs_winsys->end_cs(cs_winsys, __FILE__, __FUNCTION__, __LINE__); \ } while (0) #define FLUSH_CS do { \ - debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, __FILE__, \ - __LINE__); \ - cs_winsys->flush_cs(cs); \ + if (VERY_VERBOSE_CS) { \ + debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, \ + __FILE__, __LINE__); \ + } \ + cs_winsys->flush_cs(cs_winsys); \ } while (0) #define RADEON_ONE_REG_WR (1 << 15) @@ -138,7 +146,7 @@ assert(bo); \ OUT_CS(offset); \ OUT_CS(count); \ - cs_winsys->write_cs_reloc(cs, bo, rd, wd, flags); \ + cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \ cs_count -= 2; \ } while (0) diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c index dd63136c9d..678cd2b812 100644 --- a/src/gallium/drivers/r300/r300_debug.c +++ b/src/gallium/drivers/r300/r300_debug.c @@ -30,81 +30,6 @@ static void r300_dump_fs(struct r300_fragment_shader* fs) } } -static char* r500_fs_swiz[] = { - " R", - " G", - " B", - " A", - " 0", - ".5", - " 1", - " U", -}; - -static char* r500_fs_op_rgb[] = { - "MAD", - "DP3", - "DP4", - "D2A", - "MIN", - "MAX", - "---", - "CND", - "CMP", - "FRC", - "SOP", - "MDH", - "MDV", -}; - -static char* r500_fs_op_alpha[] = { - "MAD", - " DP", - "MIN", - "MAX", - "---", - "CND", - "CMP", - "FRC", - "EX2", - "LN2", - "RCP", - "RSQ", - "SIN", - "COS", - "MDH", - "MDV", -}; - -static char* r500_fs_mask[] = { - "NONE", - "R ", - " G ", - "RG ", - " B ", - "R B ", - " GB ", - "RGB ", - " A", - "R A", - " G A", - "RG A", - " BA", - "R BA", - " GBA", - "RGBA", -}; - -static char* r500_fs_tex[] = { - " NOP", - " LD", - "TEXKILL", - " PROJ", - "LODBIAS", - " LOD", - " DXDY", -}; - void r500_fs_dump(struct r500_fragment_shader* fs) { int i; @@ -225,14 +150,49 @@ void r500_fs_dump(struct r500_fragment_shader* fs) } } +static void r300_vs_op_dump(uint32_t op) +{ + debug_printf(" dst: %d%s op: ", + (op >> 13) & 0x7f, r300_vs_dst_debug[(op >> 8) & 0x7]); + if (op & 0x80) { + if (op & 0x1) { + debug_printf("PVS_MACRO_OP_2CLK_M2X_ADD\n"); + } else { + debug_printf(" PVS_MACRO_OP_2CLK_MADD\n"); + } + } else if (op & 0x40) { + debug_printf("%s\n", r300_vs_me_ops[op & 0x1f]); + } else { + debug_printf("%s\n", r300_vs_ve_ops[op & 0x1f]); + } +} + +void r300_vs_src_dump(uint32_t src) +{ + debug_printf(" reg: %d%s swiz: %s%s/%s%s/%s%s/%s%s\n", + (src >> 5) & 0x7f, r300_vs_src_debug[src & 0x3], + src & (1 << 25) ? "-" : " ", + r300_vs_swiz_debug[(src >> 13) & 0x7], + src & (1 << 26) ? "-" : " ", + r300_vs_swiz_debug[(src >> 16) & 0x7], + src & (1 << 27) ? "-" : " ", + r300_vs_swiz_debug[(src >> 19) & 0x7], + src & (1 << 28) ? "-" : " ", + r300_vs_swiz_debug[(src >> 22) & 0x7]); +} + void r300_vs_dump(struct r300_vertex_shader* vs) { int i; for (i = 0; i < vs->instruction_count; i++) { - debug_printf("inst0: 0x%x\n", vs->instructions[i].inst0); - debug_printf("inst1: 0x%x\n", vs->instructions[i].inst1); - debug_printf("inst2: 0x%x\n", vs->instructions[i].inst2); - debug_printf("inst3: 0x%x\n", vs->instructions[i].inst3); + debug_printf("%d: op: 0x%08x", i, vs->instructions[i].inst0); + r300_vs_op_dump(vs->instructions[i].inst0); + debug_printf(" src0: 0x%08x", vs->instructions[i].inst1); + r300_vs_src_dump(vs->instructions[i].inst1); + debug_printf(" src1: 0x%08x", vs->instructions[i].inst2); + r300_vs_src_dump(vs->instructions[i].inst2); + debug_printf(" src2: 0x%08x", vs->instructions[i].inst3); + r300_vs_src_dump(vs->instructions[i].inst3); } } diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r300_debug.h index a1f873656d..c86410ec0a 100644 --- a/src/gallium/drivers/r300/r300_debug.h +++ b/src/gallium/drivers/r300/r300_debug.h @@ -27,6 +27,182 @@ #include "r300_state_shader.h" #include "r300_state_tcl.h" +static char* r500_fs_swiz[] = { + " R", + " G", + " B", + " A", + " 0", + ".5", + " 1", + " U", +}; + +static char* r500_fs_op_rgb[] = { + "MAD", + "DP3", + "DP4", + "D2A", + "MIN", + "MAX", + "---", + "CND", + "CMP", + "FRC", + "SOP", + "MDH", + "MDV", +}; + +static char* r500_fs_op_alpha[] = { + "MAD", + " DP", + "MIN", + "MAX", + "---", + "CND", + "CMP", + "FRC", + "EX2", + "LN2", + "RCP", + "RSQ", + "SIN", + "COS", + "MDH", + "MDV", +}; + +static char* r500_fs_mask[] = { + "NONE", + "R ", + " G ", + "RG ", + " B ", + "R B ", + " GB ", + "RGB ", + " A", + "R A", + " G A", + "RG A", + " BA", + "R BA", + " GBA", + "RGBA", +}; + +static char* r500_fs_tex[] = { + " NOP", + " LD", + "TEXKILL", + " PROJ", + "LODBIAS", + " LOD", + " DXDY", +}; + +static char* r300_vs_ve_ops[] = { + /* R300 vector ops */ + " VE_NO_OP", + " VE_DOT_PRODUCT", + " VE_MULTIPLY", + " VE_ADD", + " VE_MULTIPLY_ADD", + " VE_DISTANCE_FACTOR", + " VE_FRACTION", + " VE_MAXIMUM", + " VE_MINIMUM", + "VE_SET_GREATER_THAN_EQUAL", + " VE_SET_LESS_THAN", + " VE_MULTIPLYX2_ADD", + " VE_MULTIPLY_CLAMP", + " VE_FLT2FIX_DX", + " VE_FLT2FIX_DX_RND", + /* R500 vector ops */ + " VE_PRED_SET_EQ_PUSH", + " VE_PRED_SET_GT_PUSH", + " VE_PRED_SET_GTE_PUSH", + " VE_PRED_SET_NEQ_PUSH", + " VE_COND_WRITE_EQ", + " VE_COND_WRITE_GT", + " VE_COND_WRITE_GTE", + " VE_COND_WRITE_NEQ", + " VE_SET_GREATER_THAN", + " VE_SET_EQUAL", + " VE_SET_NOT_EQUAL", + " (reserved)", + " (reserved)", + " (reserved)", +}; + +static char* r300_vs_me_ops[] = { + /* R300 math ops */ + " ME_NO_OP", + " ME_EXP_BASE2_DX", + " ME_LOG_BASE2_DX", + " ME_EXP_BASEE_FF", + " ME_LIGHT_COEFF_DX", + " ME_POWER_FUNC_FF", + " ME_RECIP_DX", + " ME_RECIP_FF", + " ME_RECIP_SQRT_DX", + " ME_RECIP_SQRT_FF", + " ME_MULTIPLY", + " ME_EXP_BASE2_FULL_DX", + " ME_LOG_BASE2_FULL_DX", + " ME_POWER_FUNC_FF_CLAMP_B", + "ME_POWER_FUNC_FF_CLAMP_B1", + "ME_POWER_FUNC_FF_CLAMP_01", + " ME_SIN", + " ME_COS", + /* R500 math ops */ + " ME_LOG_BASE2_IEEE", + " ME_RECIP_IEEE", + " ME_RECIP_SQRT_IEEE", + " ME_PRED_SET_EQ", + " ME_PRED_SET_GT", + " ME_PRED_SET_GTE", + " ME_PRED_SET_NEQ", + " ME_PRED_SET_CLR", + " ME_PRED_SET_INV", + " ME_PRED_SET_POP", + " ME_PRED_SET_RESTORE", + " (reserved)", + " (reserved)", + " (reserved)", +}; + +/* XXX refactor to avoid clashing symbols */ +static char* r300_vs_src_debug[] = { + "t", + "i", + "c", + "a", +}; + +static char* r300_vs_dst_debug[] = { + "t", + "a0", + "o", + "ox", + "a", + "i", + "u", + "u", +}; + +static char* r300_vs_swiz_debug[] = { + "X", + "Y", + "Z", + "W", + "0", + "1", + "U", + "U", +}; + void r500_fs_dump(struct r500_fragment_shader* fs); void r300_vs_dump(struct r300_vertex_shader* vs); diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index a3d83376b6..93cf6909a3 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -56,6 +56,36 @@ void r300_emit_blend_color_state(struct r300_context* r300, } } +void r300_emit_clip_state(struct r300_context* r300, + struct pipe_clip_state* clip) +{ + int i; + struct r300_screen* r300screen = r300_screen(r300->context.screen); + CS_LOCALS(r300); + + if (r300screen->caps->has_tcl) { + BEGIN_CS(5 + (6 * 4)); + OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, + (r300screen->caps->is_r500 ? + R500_PVS_UCP_START : R300_PVS_UCP_START)); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, 6 * 4); + for (i = 0; i < 6; i++) { + OUT_CS_32F(clip->ucp[i][0]); + OUT_CS_32F(clip->ucp[i][1]); + OUT_CS_32F(clip->ucp[i][2]); + OUT_CS_32F(clip->ucp[i][3]); + } + OUT_CS_REG(R300_VAP_CLIP_CNTL, ((1 << clip->nr) - 1) | + R300_PS_UCP_MODE_CLIP_AS_TRIFAN); + END_CS; + } else { + BEGIN_CS(2); + OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE); + END_CS; + } + +} + void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa) { @@ -152,34 +182,41 @@ void r500_emit_fragment_shader(struct r300_context* r300, END_CS; } -/* XXX add pitch, stride, clean up */ void r300_emit_fb_state(struct r300_context* r300, struct pipe_framebuffer_state* fb) { - int i; struct r300_texture* tex; + unsigned pixpitch; + int i; CS_LOCALS(r300); - BEGIN_CS((6 * fb->nr_cbufs) + (fb->zsbuf ? 6 : 0) + 4); + BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4); for (i = 0; i < fb->nr_cbufs; i++) { tex = (struct r300_texture*)fb->cbufs[i]->texture; + assert(tex && tex->buffer && "cbuf is marked, but NULL!"); + pixpitch = tex->stride / tex->tex.block.size; + OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), pixpitch | + r300_translate_colorformat(tex->tex.format)); + OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), r300_translate_out_fmt(fb->cbufs[i]->format)); } if (fb->zsbuf) { tex = (struct r300_texture*)fb->zsbuf->texture; + assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); + pixpitch = tex->stride / tex->tex.block.size; + OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - if (fb->zsbuf->format == PIPE_FORMAT_Z24S8_UNORM) { - OUT_CS_REG(R300_ZB_FORMAT, - R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL); - } else { - OUT_CS_REG(R300_ZB_FORMAT, 0x0); - } + + OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format)); + + OUT_CS_REG(R300_ZB_DEPTHPITCH, pixpitch); } OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, @@ -229,7 +266,7 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->ip[i]); - debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); + /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */ } OUT_CS_REG_SEQ(R300_RS_COUNT, 2); @@ -243,11 +280,11 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->inst[i]); - debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); + /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */ } - debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, - rs->inst_count); + /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, + * rs->inst_count); */ END_CS; } @@ -291,6 +328,30 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } +void r300_emit_vertex_buffer(struct r300_context* r300) +{ + CS_LOCALS(r300); + + debug_printf("r300: Preparing vertex buffer %p for render, " + "vertex size %d\n", r300->vbo, + r300->vertex_info.vinfo.size); + /* Set the pointer to our vertex buffer. The emitted values are this: + * PACKET3 [3D_LOAD_VBPNTR] + * COUNT [1] + * FORMAT [size | stride << 8] + * OFFSET [offset into BO] + * VBPNTR [relocated BO] + */ + BEGIN_CS(7); + OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3); + OUT_CS(1); + OUT_CS(r300->vertex_info.vinfo.size | + (r300->vertex_info.vinfo.size << 8)); + OUT_CS(r300->vbo_offset); + OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_CS; +} + void r300_emit_vertex_format_state(struct r300_context* r300) { int i; @@ -305,22 +366,22 @@ void r300_emit_vertex_format_state(struct r300_context* r300) OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); OUT_CS(r300->vertex_info.vinfo.hwfmt[2]); OUT_CS(r300->vertex_info.vinfo.hwfmt[3]); - for (i = 0; i < 4; i++) { - debug_printf("hwfmt%d: 0x%08x\n", i, - r300->vertex_info.vinfo.hwfmt[i]); - } + /* for (i = 0; i < 4; i++) { + * debug_printf("hwfmt%d: 0x%08x\n", i, + * r300->vertex_info.vinfo.hwfmt[i]); + * } */ OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]); - debug_printf("prog_stream_cntl%d: 0x%08x\n", i, - r300->vertex_info.vap_prog_stream_cntl[i]); + /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i, + * r300->vertex_info.vap_prog_stream_cntl[i]); */ } OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]); - debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, - r300->vertex_info.vap_prog_stream_cntl_ext[i]); + /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, + * r300->vertex_info.vap_prog_stream_cntl_ext[i]); */ } END_CS; } @@ -340,14 +401,23 @@ void r300_emit_vertex_shader(struct r300_context* r300, return; } - BEGIN_CS(13 + (vs->instruction_count * 4) + (constants->count * 4)); + if (constants->count) { + BEGIN_CS(14 + (vs->instruction_count * 4) + (constants->count * 4)); + } else { + BEGIN_CS(11 + (vs->instruction_count * 4)); + } - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) | + /* R300_VAP_PVS_CODE_CNTL_0 + * R300_VAP_PVS_CONST_CNTL + * R300_VAP_PVS_CODE_CNTL_1 + * See the r5xx docs for instructions on how to use these. + * XXX these could be optimized to select better values... */ + OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3); + OUT_CS(R300_PVS_FIRST_INST(0) | + R300_PVS_XYZW_VALID_INST(vs->instruction_count - 1) | R300_PVS_LAST_INST(vs->instruction_count - 1)); - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, vs->instruction_count - 1); - - /* XXX */ - OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x0); + OUT_CS(R300_PVS_MAX_CONST_ADDR(constants->count - 1)); + OUT_CS(vs->instruction_count - 1); OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0); OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4); @@ -377,7 +447,6 @@ void r300_emit_vertex_shader(struct r300_context* r300, R300_PVS_VF_MAX_VTX_NUM(12)); OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); END_CS; - } void r300_emit_viewport_state(struct r300_context* r300, @@ -394,7 +463,11 @@ void r300_emit_viewport_state(struct r300_context* r300, OUT_CS_32F(viewport->zscale); OUT_CS_32F(viewport->zoffset); - OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control); + if (r300->rs_state->enable_vte) { + OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control); + } else { + OUT_CS_REG(R300_VAP_VTE_CNTL, 0); + } END_CS; } @@ -412,16 +485,67 @@ void r300_flush_textures(struct r300_context* r300) void r300_emit_dirty_state(struct r300_context* r300) { struct r300_screen* r300screen = r300_screen(r300->context.screen); - int i; - int dirty_tex = 0; + struct r300_texture* tex; + int i, dirty_tex = 0; + boolean invalid = FALSE; - if (!(r300->dirty_state) && !(r300->dirty_hw)) { + if (!(r300->dirty_state)) { return; } r300_update_derived_state(r300); /* XXX check size */ +validate: + /* Color buffers... */ + for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) { + tex = (struct r300_texture*)r300->framebuffer_state.cbufs[i]->texture; + assert(tex && tex->buffer && "cbuf is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...depth buffer... */ + if (r300->framebuffer_state.zsbuf) { + tex = (struct r300_texture*)r300->framebuffer_state.zsbuf->texture; + assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...textures... */ + for (i = 0; i < r300->texture_count; i++) { + tex = r300->textures[i]; + assert(tex && tex->buffer && "texture is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...and vertex buffer. */ + if (r300->vbo) { + if (!r300->winsys->add_buffer(r300->winsys, r300->vbo, + RADEON_GEM_DOMAIN_GTT, 0)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } else { + debug_printf("No VBO while emitting dirty state!\n"); + } + if (r300->winsys->validate(r300->winsys)) { + r300->context.flush(&r300->context, 0, NULL); + if (invalid) { + /* Well, hell. */ + exit(1); + } + invalid = TRUE; + goto validate; + } if (r300->dirty_state & R300_NEW_BLEND) { r300_emit_blend_state(r300, r300->blend_state); @@ -433,6 +557,11 @@ void r300_emit_dirty_state(struct r300_context* r300) r300->dirty_state &= ~R300_NEW_BLEND_COLOR; } + if (r300->dirty_state & R300_NEW_CLIP) { + r300_emit_clip_state(r300, &r300->clip_state); + r300->dirty_state &= ~R300_NEW_CLIP; + } + if (r300->dirty_state & R300_NEW_DSA) { r300_emit_dsa_state(r300, r300->dsa_state); r300->dirty_state &= ~R300_NEW_DSA; @@ -502,4 +631,14 @@ void r300_emit_dirty_state(struct r300_context* r300) r300_emit_vertex_format_state(r300); r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT; } + + if (r300->dirty_state & R300_NEW_VERTEX_SHADER) { + r300_emit_vertex_shader(r300, r300->vs); + r300->dirty_state &= ~R300_NEW_VERTEX_SHADER; + } + + /* Finally, emit the VBO. */ + r300_emit_vertex_buffer(r300); + + r300->dirty_hw++; } diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 31dbc7ab85..946f625bd8 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -36,6 +36,9 @@ void r300_emit_blend_state(struct r300_context* r300, void r300_emit_blend_color_state(struct r300_context* r300, struct r300_blend_color_state* bc); +void r300_emit_clip_state(struct r300_context* r300, + struct pipe_clip_state* clip); + void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); @@ -62,6 +65,8 @@ void r300_emit_scissor_state(struct r300_context* r300, void r300_emit_texture(struct r300_context* r300, struct r300_texture* tex, unsigned offset); +void r300_emit_vertex_buffer(struct r300_context* r300); + void r300_emit_vertex_format_state(struct r300_context* r300); void r300_emit_vertex_shader(struct r300_context* r300, diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index 20ca6905ad..89a5f2b20c 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -29,6 +29,8 @@ static void r300_flush(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); CS_LOCALS(r300); + draw_flush(r300->draw); + if (r300->dirty_hw) { FLUSH_CS; r300_emit_invariant_state(r300); diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 660816e1da..3bb9bc47b5 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -511,11 +511,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_PVS_XYZW_VALID_INST_SHIFT 10 # define R300_PVS_LAST_INST_SHIFT 20 # define R300_PVS_FIRST_INST(x) ((x) << 0) +# define R300_PVS_XYZW_VALID_INST(x) ((x) << 10) # define R300_PVS_LAST_INST(x) ((x) << 20) /* Addresses are relative the the vertex program parameters area. */ #define R300_VAP_PVS_CONST_CNTL 0x22D4 # define R300_PVS_CONST_BASE_OFFSET_SHIFT 0 # define R300_PVS_MAX_CONST_ADDR_SHIFT 16 +# define R300_PVS_MAX_CONST_ADDR(x) ((x) << 16) #define R300_VAP_PVS_CODE_CNTL_1 0x22D8 # define R300_PVS_LAST_VTX_SRC_INST_SHIFT 0 #define R300_VAP_PVS_FLOW_CNTL_OPC 0x22DC @@ -3040,6 +3042,7 @@ enum { # define R500_INST_RGB_WMASK_R (1 << 11) # define R500_INST_RGB_WMASK_G (1 << 12) # define R500_INST_RGB_WMASK_B (1 << 13) +# define R500_INST_RGB_WMASK_RGB (7 << 11) # define R500_INST_ALPHA_WMASK (1 << 14) # define R500_INST_RGB_OMASK_R (1 << 15) # define R500_INST_RGB_OMASK_G (1 << 16) diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index b7ee8fb8a9..cd458d019a 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -45,11 +45,7 @@ struct r300_render { /* VBO */ struct pipe_buffer* vbo; - size_t vbo_size; - size_t vbo_offset; - void* vbo_map; size_t vbo_alloc_size; - size_t vbo_max_used; }; static INLINE struct r300_render* @@ -78,24 +74,21 @@ static boolean r300_render_allocate_vertices(struct vbuf_render* render, struct pipe_screen* screen = r300->context.screen; size_t size = (size_t)vertex_size * (size_t)count; - if (r300render->vbo) { + if (r300render->vbo && (size > r300render->vbo_alloc_size)) { pipe_buffer_reference(&r300render->vbo, NULL); } + + if (!r300render->vbo) { + r300render->vbo = pipe_buffer_create(screen, + 64, + PIPE_BUFFER_USAGE_VERTEX, + size); + } - r300render->vbo_size = MAX2(size, r300render->vbo_alloc_size); - r300render->vbo_offset = 0; - r300render->vbo = pipe_buffer_create(screen, - 64, - PIPE_BUFFER_USAGE_VERTEX, - r300render->vbo_size); - + r300render->vbo_alloc_size = MAX2(size, r300render->vbo_alloc_size); r300render->vertex_size = vertex_size; - if (r300render->vbo) { - return TRUE; - } else { - return FALSE; - } + return (r300render->vbo) ? TRUE : FALSE; } static void* r300_render_map_vertices(struct vbuf_render* render) @@ -103,10 +96,8 @@ static void* r300_render_map_vertices(struct vbuf_render* render) struct r300_render* r300render = r300_render(render); struct pipe_screen* screen = r300render->r300->context.screen; - r300render->vbo_map = pipe_buffer_map(screen, r300render->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); - - return (unsigned char*)r300render->vbo_map + r300render->vbo_offset; + return (unsigned char*)pipe_buffer_map(screen, r300render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); } static void r300_render_unmap_vertices(struct vbuf_render* render, @@ -116,9 +107,6 @@ static void r300_render_unmap_vertices(struct vbuf_render* render, struct r300_render* r300render = r300_render(render); struct pipe_screen* screen = r300render->r300->context.screen; - r300render->vbo_max_used = MAX2(r300render->vbo_max_used, - r300render->vertex_size * (max + 1)); - pipe_buffer_unmap(screen, r300render->vbo); } @@ -180,27 +168,9 @@ static void prepare_render(struct r300_render* render, unsigned count) CS_LOCALS(r300); - /* Make sure that all possible state is emitted. */ - r300_emit_dirty_state(r300); + r300->vbo = render->vbo; - debug_printf("r300: Preparing vertex buffer %p for render, " - "vertex size %d, vertex count %d\n", render->vbo, - r300->vertex_info.vinfo.size, count); - /* Set the pointer to our vertex buffer. The emitted values are this: - * PACKET3 [3D_LOAD_VBPNTR] - * COUNT [1] - * FORMAT [size | stride << 8] - * OFFSET [0] - * VBPNTR [relocated BO] - */ - BEGIN_CS(7); - OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3); - OUT_CS(1); - OUT_CS(r300->vertex_info.vinfo.size | - (r300->vertex_info.vinfo.size << 8)); - OUT_CS(render->vbo_offset); - OUT_CS_RELOC(render->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); - END_CS; + r300_emit_dirty_state(r300); } static void r300_render_draw_arrays(struct vbuf_render* render, @@ -212,8 +182,6 @@ static void r300_render_draw_arrays(struct vbuf_render* render, CS_LOCALS(r300); - r300render->vbo_offset = start; - prepare_render(r300render, count); debug_printf("r300: Doing vbuf render, count %d\n", count); @@ -234,6 +202,8 @@ static void r300_render_draw(struct vbuf_render* render, struct pipe_screen* screen = r300->context.screen; struct pipe_buffer* index_buffer; void* index_map; + int i; + uint32_t index; CS_LOCALS(r300); @@ -246,6 +216,7 @@ static void r300_render_draw(struct vbuf_render* render, return; } +/* index_map = pipe_buffer_map(screen, index_buffer, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(index_map, indices, count); @@ -253,13 +224,25 @@ static void r300_render_draw(struct vbuf_render* render, debug_printf("r300: Doing indexbuf render, count %d\n", count); - BEGIN_CS(6); + BEGIN_CS(8); OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0); OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | r300render->hwprim); OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2); OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2)); OUT_CS_INDEX_RELOC(index_buffer, 0, count, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_CS; */ + + BEGIN_CS(2 + (count+1)/2); + OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, (count+1)/2); + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | + r300render->hwprim); + for (i = 0; i < count-1; i += 2) { + OUT_CS(indices[i+1] << 16 | indices[i]); + } + if (count % 2) { + OUT_CS(indices[count-1]); + } END_CS; } diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index d2c5998c26..a6f1efe356 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -87,23 +87,25 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) } else { return 0; } - return 0; case PIPE_CAP_GLSL: - /* IN THEORY */ - return 0; + if (r300screen->caps->is_r500) { + return 1; + } else { + return 0; + } case PIPE_CAP_S3TC: /* IN THEORY */ return 0; case PIPE_CAP_ANISOTROPIC_FILTER: - /* IN THEORY */ - return 0; + return 1; case PIPE_CAP_POINT_SPRITE: /* IN THEORY */ return 0; case PIPE_CAP_MAX_RENDER_TARGETS: return 4; case PIPE_CAP_OCCLUSION_QUERY: - return 1; + /* IN THEORY */ + return 0; case PIPE_CAP_TEXTURE_SHADOW_MAP: /* IN THEORY */ return 0; @@ -152,17 +154,20 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) static float r300_get_paramf(struct pipe_screen* pscreen, int param) { + struct r300_screen* r300screen = r300_screen(pscreen); + switch (param) { case PIPE_CAP_MAX_LINE_WIDTH: case PIPE_CAP_MAX_LINE_WIDTH_AA: - /* XXX this is the biggest thing that will fit in that register. - * Perhaps the actual rendering limits are less? */ - return 10922.0f; case PIPE_CAP_MAX_POINT_WIDTH: case PIPE_CAP_MAX_POINT_WIDTH_AA: - /* XXX this is the biggest thing that will fit in that register. - * Perhaps the actual rendering limits are less? */ - return 10922.0f; + /* The maximum dimensions of the colorbuffer are our practical + * rendering limits. 2048 pixels should be enough for anybody. */ + if (r300screen->caps->is_r500) { + return 4096.0f; + } else { + return 2048.0f; + } case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: return 16.0f; case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: @@ -230,9 +235,16 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen, case PIPE_TEXTURE_2D: return check_tex_2d_format(format, r300_screen(pscreen)->caps->is_r500); + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_3D: + case PIPE_TEXTURE_CUBE: + debug_printf("r300: Implementation error: Unsupported format " + "target: %d\n", target); + break; default: - debug_printf("r300: Warning: Got unknown format target: %d\n", - format); + debug_printf("r300: Fatal: This is not a format target: %d\n", + target); + assert(0); break; } @@ -337,7 +349,6 @@ struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys) return NULL; caps->pci_id = r300_winsys->pci_id; - caps->num_frag_pipes = r300_winsys->gb_pipes; r300_parse_chipset(caps); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 2a77fd1739..01e2b51153 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -62,8 +62,6 @@ static void* r300_create_blend_state(struct pipe_context* pipe, } /* PIPE_LOGICOP_* don't need to be translated, fortunately. */ - /* XXX are logicops still allowed if blending's disabled? - * Does Gallium take care of it for us? */ if (state->logicop_enable) { blend->rop = R300_RB3D_ROPCNTL_ROP_ENABLE | (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT; @@ -121,9 +119,14 @@ static void r300_set_clip_state(struct pipe_context* pipe, const struct pipe_clip_state* state) { struct r300_context* r300 = r300_context(pipe); - /* XXX Draw */ - draw_flush(r300->draw); - draw_set_clip_state(r300->draw, state); + + if (r300_screen(pipe->screen)->caps->has_tcl) { + r300->clip_state = *state; + r300->dirty_state |= R300_NEW_CLIP; + } else { + draw_flush(r300->draw); + draw_set_clip_state(r300->draw, state); + } } static void @@ -153,10 +156,12 @@ static void /* If the number of constants have changed, invalidate the shader. */ if (r300->shader_constants[shader].user_count != i) { - if (shader == PIPE_SHADER_FRAGMENT && r300->fs) { + if (shader == PIPE_SHADER_FRAGMENT && r300->fs && + r300->fs->uses_imms) { r300->fs->translated = FALSE; r300_translate_fragment_shader(r300, r300->fs); - } else if (shader == PIPE_SHADER_VERTEX && r300->vs) { + } else if (shader == PIPE_SHADER_VERTEX && r300->vs && + r300->vs->uses_imms) { r300->vs->translated = FALSE; r300_translate_vertex_shader(r300, r300->vs); } @@ -257,6 +262,7 @@ static void r300_set_edgeflags(struct pipe_context* pipe, const unsigned* bitfield) { /* XXX you know it's bad when i915 has this blank too */ + /* XXX and even worse, I have no idea WTF the bitfield is */ } static void @@ -289,6 +295,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe, /* Copy state directly into shader. */ fs->state = *shader; + fs->state.tokens = tgsi_dup_tokens(shader->tokens); tgsi_scan_shader(shader->tokens, &fs->info); @@ -317,13 +324,15 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) /* Delete fragment shader state. */ static void r300_delete_fs_state(struct pipe_context* pipe, void* shader) { + struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader; + FREE(fs->state.tokens); FREE(shader); } static void r300_set_polygon_stipple(struct pipe_context* pipe, const struct pipe_poly_stipple* state) { - /* XXX */ + /* XXX no idea how to set this up, but not terribly important */ } /* Create a new rasterizer state based on the CSO rasterizer state. @@ -341,6 +350,8 @@ static void* r300_create_rs_state(struct pipe_context* pipe, /* Copy rasterizer state for Draw. */ rs->rs = *state; + rs->enable_vte = !state->bypass_vs_clip_and_viewport; + /* If bypassing TCL, or if no TCL engine is present, turn off the HW TCL. * Else, enable HW TCL and force Draw's TCL off. */ if (state->bypass_vs_clip_and_viewport || @@ -421,6 +432,7 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state) struct r300_context* r300 = r300_context(pipe); struct r300_rs_state* rs = (struct r300_rs_state*)state; + draw_flush(r300->draw); draw_set_rasterizer_state(r300->draw, &rs->rs); r300->rs_state = rs; @@ -528,7 +540,6 @@ static void r300_set_scissor_state(struct pipe_context* pipe, const struct pipe_scissor_state* state) { struct r300_context* r300 = r300_context(pipe); - draw_flush(r300->draw); if (r300_screen(r300->context.screen)->caps->is_r500) { r300->scissor_state->scissor_top_left = @@ -555,25 +566,38 @@ static void r300_set_viewport_state(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); - r300->viewport_state->xscale = state->scale[0]; - r300->viewport_state->yscale = state->scale[1]; - r300->viewport_state->zscale = state->scale[2]; - - r300->viewport_state->xoffset = state->translate[0]; - r300->viewport_state->yoffset = state->translate[1]; - r300->viewport_state->zoffset = state->translate[2]; - - r300->viewport_state->vte_control = 0; - if (r300_screen(r300->context.screen)->caps->has_tcl) { - /* Do the transform in HW. */ - r300->viewport_state->vte_control |= - R300_VPORT_X_SCALE_ENA | R300_VPORT_X_OFFSET_ENA | - R300_VPORT_Y_SCALE_ENA | R300_VPORT_Y_OFFSET_ENA | - R300_VPORT_Z_SCALE_ENA | R300_VPORT_Z_OFFSET_ENA; - } else { - /* Have Draw do the actual transform. */ - draw_set_viewport_state(r300->draw, state); + /* Do the transform in HW. */ + r300->viewport_state->vte_control = R300_VTX_W0_FMT; + + if (state->scale[0] != 1.0f) { + assert(state->scale[0] != 0.0f); + r300->viewport_state->xscale = state->scale[0]; + r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA; + } + if (state->scale[1] != 1.0f) { + assert(state->scale[1] != 0.0f); + r300->viewport_state->yscale = state->scale[1]; + r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA; + } + if (state->scale[2] != 1.0f) { + assert(state->scale[2] != 0.0f); + r300->viewport_state->zscale = state->scale[2]; + r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA; + } + if (state->translate[0] != 0.0f) { + r300->viewport_state->xoffset = state->translate[0]; + r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA; + } + if (state->translate[1] != 0.0f) { + r300->viewport_state->yoffset = state->translate[1]; + r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA; } + if (state->translate[2] != 0.0f) { + r300->viewport_state->zoffset = state->translate[2]; + r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA; + } + + r300->dirty_state |= R300_NEW_VIEWPORT; } static void r300_set_vertex_buffers(struct pipe_context* pipe, @@ -610,6 +634,7 @@ static void* r300_create_vs_state(struct pipe_context* pipe, struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader); /* Copy state directly into shader. */ vs->state = *shader; + vs->state.tokens = tgsi_dup_tokens(shader->tokens); tgsi_scan_shader(shader->tokens, &vs->info); @@ -626,6 +651,8 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader) { struct r300_context* r300 = r300_context(pipe); + draw_flush(r300->draw); + if (r300_screen(pipe->screen)->caps->has_tcl) { struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader; @@ -653,6 +680,7 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader) struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader; draw_delete_vertex_shader(r300->draw, vs->draw); + FREE(vs->state.tokens); FREE(shader); } else { draw_delete_vertex_shader(r300->draw, diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index f1feafbcf9..2477b30822 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -25,62 +25,88 @@ /* r300_state_derived: Various bits of state which are dependent upon * currently bound CSO data. */ -/* Update the vertex_info struct in our r300_context. - * - * The vertex_info struct describes the post-TCL format of vertices. It is - * required for Draw when doing SW TCL, and also for describing the - * dreaded RS block on R300 chipsets. */ -static void r300_update_vertex_layout(struct r300_context* r300) +/* Set up the vs_tab and routes. */ +static void r300_vs_tab_routes(struct r300_context* r300, + struct r300_vertex_format* vformat) { struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct r300_vertex_format vformat; - struct vertex_info vinfo; + struct vertex_info* vinfo = &vformat->vinfo; + int* tab = vformat->vs_tab; boolean pos = FALSE, psize = FALSE, fog = FALSE; int i, texs = 0, cols = 0; - int tab[16]; - - struct tgsi_shader_info* info = &r300->fs->info; + struct tgsi_shader_info* info; - memset(&vinfo, 0, sizeof(vinfo)); - for (i = 0; i < 16; i++) { - tab[i] = -1; + if (r300screen->caps->has_tcl) { + /* Use vertex shader to determine required routes. */ + info = &r300->vs->info; + } else { + /* Use fragment shader to determine required routes. */ + info = &r300->fs->info; } assert(info->num_inputs <= 16); - for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - pos = TRUE; - tab[i] = 0; - break; - case TGSI_SEMANTIC_COLOR: - tab[i] = 2 + cols++; - break; - case TGSI_SEMANTIC_PSIZE: - psize = TRUE; - tab[i] = 1; - break; - case TGSI_SEMANTIC_FOG: - fog = TRUE; - /* Fall through... */ - case TGSI_SEMANTIC_GENERIC: - tab[i] = 6 + texs++; - break; - default: - debug_printf("r300: Unknown vertex input %d\n", - info->input_semantic_name[i]); - break; - } - } - if (r300screen->caps->has_tcl) { + /* Just copy vert attribs over as-is. */ for (i = 0; i < info->num_inputs; i++) { - /* XXX should probably do real lookup with vert shader */ tab[i] = i; } + for (i = 0; i < info->num_outputs; i++) { + switch (info->output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + pos = TRUE; + break; + case TGSI_SEMANTIC_COLOR: + cols++; + break; + case TGSI_SEMANTIC_PSIZE: + psize = TRUE; + break; + case TGSI_SEMANTIC_FOG: + fog = TRUE; + /* Fall through */ + case TGSI_SEMANTIC_GENERIC: + texs++; + break; + default: + debug_printf("r300: Unknown vertex output %d\n", + info->output_semantic_name[i]); + break; + } + } + } else { + for (i = 0; i < info->num_inputs; i++) { + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + pos = TRUE; + tab[i] = 0; + break; + case TGSI_SEMANTIC_COLOR: + tab[i] = 2 + cols; + cols++; + break; + case TGSI_SEMANTIC_PSIZE: + psize = TRUE; + tab[i] = 15; + break; + case TGSI_SEMANTIC_FOG: + fog = TRUE; + /* Fall through */ + case TGSI_SEMANTIC_GENERIC: + tab[i] = 6 + texs; + texs++; + break; + default: + debug_printf("r300: Unknown vertex input %d\n", + info->input_semantic_name[i]); + break; + } + } } + /* XXX magic */ + assert(texs <= 8); + /* Do the actual vertex_info setup. * * vertex_info has four uints of hardware-specific data in it. @@ -89,7 +115,7 @@ static void r300_update_vertex_layout(struct r300_context* r300) * vinfo.hwfmt[2] is R300_VAP_OUTPUT_VTX_FMT_0 * vinfo.hwfmt[3] is R300_VAP_OUTPUT_VTX_FMT_1 */ - vinfo.hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */ + vinfo->hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */ if (!pos) { debug_printf("r300: Forcing vertex position attribute emit...\n"); @@ -99,109 +125,223 @@ static void r300_update_vertex_layout(struct r300_context* r300) tab[i] = tab[i-1]; } tab[0] = 0; - - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_POS, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); - } else { - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); } - vinfo.hwfmt[1] |= R300_INPUT_CNTL_POS; - vinfo.hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); + vinfo->hwfmt[1] |= R300_INPUT_CNTL_POS; + vinfo->hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; if (psize) { - draw_emit_vertex_attr(&vinfo, EMIT_1F_PSIZE, INTERP_POS, + draw_emit_vertex_attr(vinfo, EMIT_1F_PSIZE, INTERP_POS, draw_find_vs_output(r300->draw, TGSI_SEMANTIC_PSIZE, 0)); - vinfo.hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; + vinfo->hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; } for (i = 0; i < cols; i++) { - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, draw_find_vs_output(r300->draw, TGSI_SEMANTIC_COLOR, i)); - vinfo.hwfmt[1] |= R300_INPUT_CNTL_COLOR; - vinfo.hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i); + vinfo->hwfmt[1] |= R300_INPUT_CNTL_COLOR; + vinfo->hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i); } - for (i = 0; i < texs; i++) { - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i)); - vinfo.hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); - vinfo.hwfmt[3] |= (4 << (3 * i)); - } + /* Init i right here, increment it if fog is enabled. + * This gets around a double-increment problem. */ + i = 0; if (fog) { i++; - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0)); - vinfo.hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); - vinfo.hwfmt[3] |= (4 << (3 * i)); + vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); + vinfo->hwfmt[3] |= (4 << (3 * i)); + } + + for (i; i < texs; i++) { + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, + draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i)); + vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); + vinfo->hwfmt[3] |= (4 << (3 * i)); } - draw_compute_vertex_size(&vinfo); + /* Handle the case where the vertex shader will be generating some of + * the attribs based on its inputs. */ + if (r300screen->caps->has_tcl && + info->num_inputs < info->num_outputs) { + vinfo->num_attribs = info->num_inputs; + } - if (memcmp(&r300->vertex_info, &vinfo, sizeof(struct vertex_info))) { - uint32_t temp; - debug_printf("attrib count: %d, fp input count: %d\n", - vinfo.num_attribs, info->num_inputs); - for (i = 0; i < vinfo.num_attribs; i++) { - debug_printf("attrib: offset %d, interp %d, size %d," - " tab %d\n", vinfo.attrib[i].src_index, - vinfo.attrib[i].interp_mode, vinfo.attrib[i].emit, + draw_compute_vertex_size(vinfo); +} + +/* Update the PSC tables. */ +static void r300_vertex_psc(struct r300_context* r300, + struct r300_vertex_format* vformat) +{ + struct r300_screen* r300screen = r300_screen(r300->context.screen); + struct vertex_info* vinfo = &vformat->vinfo; + int* tab = vformat->vs_tab; + uint32_t temp; + int i, attrib_count; + + /* Vertex shaders have no semantics on their inputs, + * so PSC should just route stuff based on their info, + * and not on attrib information. */ + if (r300screen->caps->has_tcl) { + attrib_count = r300->vs->info.num_inputs; + debug_printf("r300: routing %d attribs in psc for vs\n", + attrib_count); + } else { + attrib_count = vinfo->num_attribs; + debug_printf("r300: attrib count: %d\n", attrib_count); + for (i = 0; i < attrib_count; i++) { + debug_printf("r300: attrib: offset %d, interp %d, size %d," + " tab %d\n", vinfo->attrib[i].src_index, + vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit, tab[i]); } + } - for (i = 0; i < vinfo.num_attribs; i++) { - /* Make sure we have a proper destination for our attribute */ - assert(tab[i] != -1); + for (i = 0; i < attrib_count; i++) { + /* Make sure we have a proper destination for our attribute */ + assert(tab[i] != -1); - temp = translate_vertex_data_type(vinfo.attrib[i].emit) | - (tab[i] << R300_DST_VEC_LOC_SHIFT); - if (i & 1) { - r300->vertex_info.vap_prog_stream_cntl[i >> 1] &= 0x0000ffff; - r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= temp << 16; - } else { - r300->vertex_info.vap_prog_stream_cntl[i >> 1] &= 0xffff0000; - r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= temp; - } + /* Add the attribute to the PSC table. */ + temp = r300screen->caps->has_tcl ? + R300_DATA_TYPE_FLOAT_4 : + translate_vertex_data_type(vinfo->attrib[i].emit); + temp |= tab[i] << R300_DST_VEC_LOC_SHIFT; + + if (i & 1) { + vformat->vap_prog_stream_cntl[i >> 1] &= 0x0000ffff; + vformat->vap_prog_stream_cntl[i >> 1] |= temp << 16; - r300->vertex_info.vap_prog_stream_cntl_ext[i >> 1] |= - (R300_VAP_SWIZZLE_XYZW << (i & 1 ? 16 : 0)); + vformat->vap_prog_stream_cntl_ext[i >> 1] |= + (R300_VAP_SWIZZLE_XYZW << 16); + } else { + vformat->vap_prog_stream_cntl[i >> 1] &= 0xffff0000; + vformat->vap_prog_stream_cntl[i >> 1] |= temp << 0; + + vformat->vap_prog_stream_cntl_ext[i >> 1] |= + (R300_VAP_SWIZZLE_XYZW << 0); } - /* Set the last vector. */ - i--; - r300->vertex_info.vap_prog_stream_cntl[i >> 1] |= (R300_LAST_VEC << - (i & 1 ? 16 : 0)); + } - memcpy(r300->vertex_info.tab, tab, sizeof(tab)); - memcpy(&r300->vertex_info, &vinfo, sizeof(struct vertex_info)); + /* Set the last vector in the PSC. */ + i--; + vformat->vap_prog_stream_cntl[i >> 1] |= + (R300_LAST_VEC << (i & 1 ? 16 : 0)); +} + +/* Update the vertex format. */ +static void r300_update_vertex_format(struct r300_context* r300) +{ + struct r300_vertex_format vformat; + int i; + + memset(&vformat, 0, sizeof(struct r300_vertex_format)); + for (i = 0; i < 16; i++) { + vformat.vs_tab[i] = -1; + vformat.fs_tab[i] = -1; + } + + r300_vs_tab_routes(r300, &vformat); + + r300_vertex_psc(r300, &vformat); + + if (memcmp(&r300->vertex_info, &vformat, + sizeof(struct r300_vertex_format))) { + memcpy(&r300->vertex_info, &vformat, + sizeof(struct r300_vertex_format)); r300->dirty_state |= R300_NEW_VERTEX_FORMAT; } } +/* Set up the mappings from GB to US, for RS block. */ +static void r300_update_fs_tab(struct r300_context* r300) +{ + struct r300_vertex_format* vformat = &r300->vertex_info; + struct tgsi_shader_info* info = &r300->fs->info; + int i, cols = 0, texs = 0, cols_emitted = 0; + int* tab = vformat->fs_tab; + + for (i = 0; i < 16; i++) { + tab[i] = -1; + } + + assert(info->num_inputs <= 16); + for (i = 0; i < info->num_inputs; i++) { + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: + tab[i] = INTERP_LINEAR; + cols++; + break; + case TGSI_SEMANTIC_POSITION: + case TGSI_SEMANTIC_PSIZE: + debug_printf("r300: Implementation error: Can't use " + "pos attribs in fragshader yet!\n"); + /* Pass through for now */ + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + tab[i] = INTERP_PERSPECTIVE; + break; + default: + debug_printf("r300: Unknown vertex input %d\n", + info->input_semantic_name[i]); + break; + } + } + + /* Now that we know where everything is... */ + debug_printf("r300: fp input count: %d\n", info->num_inputs); + for (i = 0; i < info->num_inputs; i++) { + switch (tab[i]) { + case INTERP_LINEAR: + debug_printf("r300: attrib: " + "stack offset %d, color, tab %d\n", + i, cols_emitted); + tab[i] = cols_emitted; + cols_emitted++; + break; + case INTERP_PERSPECTIVE: + debug_printf("r300: attrib: " + "stack offset %d, texcoord, tab %d\n", + i, cols + texs); + tab[i] = cols + texs; + texs++; + break; + case -1: + debug_printf("r300: Implementation error: Bad fp interp!\n"); + default: + break; + } + } + +} + /* Set up the RS block. This is the part of the chipset that actually does * the rasterization of vertices into fragments. This is also the part of the * chipset that locks up if any part of it is even slightly wrong. */ static void r300_update_rs_block(struct r300_context* r300) { struct r300_rs_block* rs = r300->rs_block; - struct vertex_info* vinfo = &r300->vertex_info.vinfo; - int* tab = r300->vertex_info.tab; + struct tgsi_shader_info* info = &r300->fs->info; + int* tab = r300->vertex_info.fs_tab; int col_count = 0, fp_offset = 0, i, memory_pos, tex_count = 0; memset(rs, 0, sizeof(struct r300_rs_block)); if (r300_screen(r300->context.screen)->caps->is_r500) { - for (i = 0; i < vinfo->num_attribs; i++) { - assert(tab[vinfo->attrib[i].src_index] != -1); - memory_pos = tab[vinfo->attrib[i].src_index] * 4; - switch (vinfo->attrib[i].interp_mode) { - case INTERP_LINEAR: + for (i = 0; i < info->num_inputs; i++) { + assert(tab[i] != -1); + memory_pos = tab[i] * 4; + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: rs->ip[col_count] |= R500_RS_COL_PTR(memory_pos) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA); col_count++; break; - case INTERP_PERSPECTIVE: + case TGSI_SEMANTIC_GENERIC: rs->ip[tex_count] |= R500_RS_SEL_S(memory_pos) | R500_RS_SEL_T(memory_pos + 1) | @@ -243,17 +383,17 @@ static void r300_update_rs_block(struct r300_context* r300) fp_offset++; } } else { - for (i = 0; i < vinfo->num_attribs; i++) { - memory_pos = tab[vinfo->attrib[i].src_index] * 4; - assert(tab[vinfo->attrib[i].src_index] != -1); - switch (vinfo->attrib[i].interp_mode) { - case INTERP_LINEAR: + for (i = 0; i < info->num_inputs; i++) { + assert(tab[i] != -1); + memory_pos = tab[i] * 4; + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: rs->ip[col_count] |= R300_RS_COL_PTR(memory_pos) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA); col_count++; break; - case INTERP_PERSPECTIVE: + case TGSI_SEMANTIC_GENERIC: rs->ip[tex_count] |= R300_RS_TEX_PTR(memory_pos) | R300_RS_SEL_S(R300_RS_SEL_C0) | @@ -307,10 +447,11 @@ void r300_update_derived_state(struct r300_context* r300) { if (r300->dirty_state & (R300_NEW_FRAGMENT_SHADER | R300_NEW_VERTEX_SHADER)) { - r300_update_vertex_layout(r300); + r300_update_vertex_format(r300); } if (r300->dirty_state & R300_NEW_VERTEX_FORMAT) { + r300_update_fs_tab(r300); r300_update_rs_block(r300); } } diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index 91b93fc367..22c8e199ae 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -353,25 +353,6 @@ static INLINE uint32_t r300_translate_out_fmt(enum pipe_format format) /* Non-CSO state. (For now.) */ -static INLINE uint32_t r300_translate_gb_pipes(int pipe_count) -{ - switch (pipe_count) { - case 1: - return R300_GB_TILE_PIPE_COUNT_RV300; - break; - case 2: - return R300_GB_TILE_PIPE_COUNT_R300; - break; - case 3: - return R300_GB_TILE_PIPE_COUNT_R420_3P; - break; - case 4: - return R300_GB_TILE_PIPE_COUNT_R420; - break; - } - return 0; -} - static INLINE uint32_t translate_vertex_data_type(int type) { switch (type) { case EMIT_1F: diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 8bd9b41bd7..e438114010 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -34,36 +34,19 @@ void r300_emit_invariant_state(struct r300_context* r300) struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); - BEGIN_CS(30 + (caps->has_tcl ? 2: 0)); + BEGIN_CS(26 + (caps->has_tcl ? 2: 0)); /*** Graphics Backend (GB) ***/ /* Various GB enables */ OUT_CS_REG(R300_GB_ENABLE, 0x0); /* Subpixel multisampling for AA */ OUT_CS_REG(R300_GB_MSPOS0, 0x66666666); - OUT_CS_REG(R300_GB_MSPOS1, 0x66666666); - /* GB tile config and pipe setup */ - OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_DISABLE | - r300_translate_gb_pipes(caps->num_frag_pipes)); + OUT_CS_REG(R300_GB_MSPOS1, 0x6666666); /* Source of fog depth */ OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W); /* AA enable */ OUT_CS_REG(R300_GB_AA_CONFIG, 0x0); - /*** Geometry Assembly (GA) ***/ - /* GA errata fixes. */ - if (caps->is_r500) { - OUT_CS_REG(R300_GA_ENHANCE, - R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL | - R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE | - R500_GA_ENHANCE_REG_READWRITE_ENABLE | - R500_GA_ENHANCE_REG_NOSTALL_ENABLE); - } else { - OUT_CS_REG(R300_GA_ENHANCE, - R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL | - R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE); - } - /*** Fog (FG) ***/ OUT_CS_REG(R300_FG_FOG_BLEND, 0x0); OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x0); @@ -86,7 +69,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(79 + (caps->has_tcl ? 7 : 0)); + BEGIN_CS(77 + (caps->has_tcl ? 5 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -97,8 +80,6 @@ void r300_emit_invariant_state(struct r300_context* r300) /* XXX endian */ if (caps->has_tcl) { OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP); - OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE | - R300_PS_UCP_MODE_CLIP_AS_TRIFAN); OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4); OUT_CS_32F(1.0); OUT_CS_32F(1.0); @@ -113,11 +94,14 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_32F(0.0); OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1); OUT_CS_32F(1.0); + /* XXX line tex stuffing */ + OUT_CS_REG_SEQ(R300_GA_LINE_S0, 1); + OUT_CS_32F(0.0); + OUT_CS_REG_SEQ(R300_GA_LINE_S1, 1); + OUT_CS_32F(1.0); OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 | (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT)); /* XXX this big chunk should be refactored into rs_state */ - OUT_CS_REG(R300_GA_LINE_S0, 0x00000000); - OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000); OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000); OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000); OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000); @@ -144,8 +128,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1); OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405); OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F); - /* Vertex size. */ - OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8); /* XXX */ OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa); diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r300_state_shader.c index 1b02239ee7..cc7f6a7c4b 100644 --- a/src/gallium/drivers/r300/r300_state_shader.c +++ b/src/gallium/drivers/r300/r300_state_shader.c @@ -22,24 +22,6 @@ #include "r300_state_shader.h" -static void r300_copy_passthrough_shader(struct r300_fragment_shader* fs) -{ - struct r300_fragment_shader* pt = &r300_passthrough_fragment_shader; - fs->shader.stack_size = pt->shader.stack_size; - fs->alu_instruction_count = pt->alu_instruction_count; - fs->tex_instruction_count = pt->tex_instruction_count; - fs->indirections = pt->indirections; - fs->instructions[0] = pt->instructions[0]; -} - -static void r500_copy_passthrough_shader(struct r500_fragment_shader* fs) -{ - struct r500_fragment_shader* pt = &r500_passthrough_fragment_shader; - fs->shader.stack_size = pt->shader.stack_size; - fs->instruction_count = pt->instruction_count; - fs->instructions[0] = pt->instructions[0]; -} - static void r300_fs_declare(struct r300_fs_asm* assembler, struct tgsi_full_declaration* decl) { @@ -49,6 +31,7 @@ static void r300_fs_declare(struct r300_fs_asm* assembler, case TGSI_SEMANTIC_COLOR: assembler->color_count++; break; + case TGSI_SEMANTIC_FOG: case TGSI_SEMANTIC_GENERIC: assembler->tex_count++; break; @@ -59,6 +42,12 @@ static void r300_fs_declare(struct r300_fs_asm* assembler, } break; case TGSI_FILE_OUTPUT: + /* Depth write. Mark the position of the output so we can + * identify it later. */ + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { + assembler->depth_output = decl->DeclarationRange.First; + } + break; case TGSI_FILE_CONSTANT: break; case TGSI_FILE_TEMPORARY: @@ -120,6 +109,14 @@ static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler, return 0; } +static INLINE boolean r300_fs_is_depr(struct r300_fs_asm* assembler, + struct tgsi_dst_register* dst) +{ + return (assembler->writes_depth && + (dst->File == TGSI_FILE_OUTPUT) && + (dst->Index == assembler->depth_output)); +} + static INLINE unsigned r500_fix_swiz(unsigned s) { /* For historical reasons, the swizzle values x, y, z, w, and 0 are @@ -194,11 +191,17 @@ static INLINE uint32_t r300_alpha_op(unsigned op) static INLINE uint32_t r500_rgba_op(unsigned op) { switch (op) { + case TGSI_OPCODE_COS: case TGSI_OPCODE_EX2: case TGSI_OPCODE_LG2: case TGSI_OPCODE_RCP: case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SIN: return R500_ALU_RGBA_OP_SOP; + case TGSI_OPCODE_DDX: + return R500_ALU_RGBA_OP_MDH; + case TGSI_OPCODE_DDY: + return R500_ALU_RGBA_OP_MDV; case TGSI_OPCODE_FRC: return R500_ALU_RGBA_OP_FRC; case TGSI_OPCODE_DP3: @@ -224,6 +227,8 @@ static INLINE uint32_t r500_rgba_op(unsigned op) static INLINE uint32_t r500_alpha_op(unsigned op) { switch (op) { + case TGSI_OPCODE_COS: + return R500_ALPHA_OP_COS; case TGSI_OPCODE_EX2: return R500_ALPHA_OP_EX2; case TGSI_OPCODE_LG2: @@ -234,6 +239,12 @@ static INLINE uint32_t r500_alpha_op(unsigned op) return R500_ALPHA_OP_RSQ; case TGSI_OPCODE_FRC: return R500_ALPHA_OP_FRC; + case TGSI_OPCODE_SIN: + return R500_ALPHA_OP_SIN; + case TGSI_OPCODE_DDX: + return R500_ALPHA_OP_MDH; + case TGSI_OPCODE_DDY: + return R500_ALPHA_OP_MDV; case TGSI_OPCODE_DP3: case TGSI_OPCODE_DP4: case TGSI_OPCODE_DPH: @@ -295,38 +306,34 @@ static INLINE void r300_emit_maths(struct r300_fragment_shader* fs, } /* Setup an ALU operation. */ -static INLINE void r500_emit_alu(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_dst_register* dst) +static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, + struct r300_fs_asm* assembler, + struct tgsi_full_src_register* src, + struct tgsi_full_dst_register* dst, + unsigned op, + unsigned count) { int i = fs->instruction_count; if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - fs->instructions[i].inst0 = R500_INST_TYPE_OUT | - R500_ALU_OMASK(dst->DstRegister.WriteMask); + fs->instructions[i].inst0 = R500_INST_TYPE_OUT; + if (r300_fs_is_depr(assembler, dst)) { + fs->instructions[i].inst4 = R500_W_OMASK; + } else { + fs->instructions[i].inst0 |= + R500_ALU_OMASK(dst->DstRegister.WriteMask); + } } else { fs->instructions[i].inst0 = R500_INST_TYPE_ALU | - R500_ALU_WMASK(dst->DstRegister.WriteMask); + R500_ALU_WMASK(dst->DstRegister.WriteMask); } fs->instructions[i].inst0 |= R500_INST_TEX_SEM_WAIT; - fs->instructions[i].inst4 = + fs->instructions[i].inst4 |= R500_ALPHA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); fs->instructions[i].inst5 = R500_ALU_RGBA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); -} - -static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_src_register* src, - struct tgsi_full_dst_register* dst, - unsigned op, - unsigned count) -{ - int i = fs->instruction_count; - - r500_emit_alu(fs, assembler, dst); switch (count) { case 3: @@ -348,8 +355,8 @@ static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, R500_ALU_RGB_SEL_B_SRC1 | R500_SWIZ_RGB_B(r500_rgb_swiz(&src[1])); fs->instructions[i].inst4 |= - R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])) | - R500_ALPHA_SEL_B_SRC1; + R500_ALPHA_SEL_B_SRC1 | + R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])); case 1: case 0: default: @@ -361,8 +368,8 @@ static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, R500_ALU_RGB_SEL_A_SRC0 | R500_SWIZ_RGB_A(r500_rgb_swiz(&src[0])); fs->instructions[i].inst4 |= - R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])) | - R500_ALPHA_SEL_A_SRC0; + R500_ALPHA_SEL_A_SRC0 | + R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])); break; } @@ -441,6 +448,9 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs, * AMD/ATI names for opcodes, please, as it facilitates using the * documentation. */ switch (inst->Instruction.Opcode) { + /* XXX trig needs extra prep */ + case TGSI_OPCODE_COS: + case TGSI_OPCODE_SIN: /* The simple scalar ops. */ case TGSI_OPCODE_EX2: case TGSI_OPCODE_LG2: @@ -452,6 +462,8 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs, inst->FullSrcRegisters[0].SrcRegister.SwizzleW = inst->FullSrcRegisters[0].SrcRegister.SwizzleX; /* Fall through */ + case TGSI_OPCODE_DDX: + case TGSI_OPCODE_DDY: case TGSI_OPCODE_FRC: r500_emit_maths(fs, assembler, inst->FullSrcRegisters, &inst->FullDstRegisters[0], inst->Instruction.Opcode, 1); @@ -527,6 +539,60 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs, &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); break; + /* The compound and hybrid insts. */ + case TGSI_OPCODE_LRP: + /* LRP DST A, B, C -> MAD TMP -A, C, C; MAD DST A, B, TMP */ + inst->FullSrcRegisters[3] = inst->FullSrcRegisters[1]; + inst->FullSrcRegisters[1] = inst->FullSrcRegisters[2]; + inst->FullSrcRegisters[0].SrcRegister.Negate = + !(inst->FullSrcRegisters[0].SrcRegister.Negate); + inst->FullDstRegisters[1] = inst->FullDstRegisters[0]; + inst->FullDstRegisters[0].DstRegister.Index = + assembler->temp_count; + inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3); + inst->FullSrcRegisters[2].SrcRegister.Index = + assembler->temp_count; + inst->FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY; + inst->FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst->FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst->FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; + inst->FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; + inst->FullSrcRegisters[1] = inst->FullSrcRegisters[3]; + inst->FullSrcRegisters[0].SrcRegister.Negate = + !(inst->FullSrcRegisters[0].SrcRegister.Negate); + inst->FullDstRegisters[0] = inst->FullDstRegisters[1]; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3); + break; + case TGSI_OPCODE_POW: + /* POW DST A, B -> LG2 TMP A; MUL TMP TMP, B; EX2 DST TMP */ + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX; + inst->FullSrcRegisters[0].SrcRegister.SwizzleW = + inst->FullSrcRegisters[0].SrcRegister.SwizzleX; + inst->FullDstRegisters[1] = inst->FullDstRegisters[0]; + inst->FullDstRegisters[0].DstRegister.Index = + assembler->temp_count; + inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], TGSI_OPCODE_LG2, 1); + inst->FullSrcRegisters[0].SrcRegister.Index = + assembler->temp_count; + inst->FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + inst->FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + inst->FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + inst->FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; + inst->FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; + inst->FullSrcRegisters[2] = r500_constant_zero; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], TGSI_OPCODE_MUL, 3); + inst->FullDstRegisters[0] = inst->FullDstRegisters[1]; + r500_emit_maths(fs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], TGSI_OPCODE_EX2, 1); + break; + /* The texture instruction set. */ case TGSI_OPCODE_KIL: case TGSI_OPCODE_TEX: @@ -555,7 +621,7 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs, static void r300_fs_finalize(struct r3xx_fragment_shader* fs, struct r300_fs_asm* assembler) { - fs->stack_size = assembler->temp_count + assembler->temp_offset; + fs->stack_size = assembler->temp_count + assembler->temp_offset + 1; } static void r500_fs_finalize(struct r500_fragment_shader* fs, @@ -581,6 +647,8 @@ void r300_translate_fragment_shader(struct r300_context* r300, } /* Setup starting offset for immediates. */ assembler->imm_offset = consts->user_count; + /* Enable depth writes, if needed. */ + assembler->writes_depth = fs->info.writes_z; /* Make sure we start at the beginning of the shader. */ if (is_r500) { @@ -630,6 +698,7 @@ void r300_translate_fragment_shader(struct r300_context* r300, assembler->tex_count + assembler->color_count); consts->count = consts->user_count + assembler->imm_count; + fs->uses_imms = assembler->imm_count; debug_printf("r300: fs: %d total constants, " "%d from user and %d from immediates\n", consts->count, consts->user_count, assembler->imm_count); diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r300_state_shader.h index 185fdd90f0..b6087404ce 100644 --- a/src/gallium/drivers/r300/r300_state_shader.h +++ b/src/gallium/drivers/r300/r300_state_shader.h @@ -57,6 +57,7 @@ #define R500_TEX_WMASK(x) ((x) << 11) #define R500_ALU_WMASK(x) ((x) << 11) #define R500_ALU_OMASK(x) ((x) << 15) +#define R500_W_OMASK (1 << 31) /* TGSI constants. TGSI is like XML: If it can't solve your problems, you're * not using enough of it. */ @@ -99,20 +100,17 @@ struct r300_fs_asm { unsigned imm_offset; /* Number of immediate constants. */ unsigned imm_count; + /* Are depth writes enabled? */ + boolean writes_depth; + /* Depth write offset. This is the TGSI output that corresponds to + * depth writes. */ + unsigned depth_output; }; void r300_translate_fragment_shader(struct r300_context* r300, struct r3xx_fragment_shader* fs); static struct r300_fragment_shader r300_passthrough_fragment_shader = { - /* XXX This is the emission code. TODO: decode - OUT_CS_REG(R300_US_CONFIG, 0); - OUT_CS_REG(R300_US_CODE_OFFSET, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000); -*/ .alu_instruction_count = 1, .tex_instruction_count = 0, .indirections = 0, @@ -159,14 +157,6 @@ static struct r500_fragment_shader r500_passthrough_fragment_shader = { }; static struct r300_fragment_shader r300_texture_fragment_shader = { - /* XXX This is the emission code. TODO: decode - OUT_CS_REG(R300_US_CONFIG, 0); - OUT_CS_REG(R300_US_CODE_OFFSET, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000); -*/ .alu_instruction_count = 1, .tex_instruction_count = 0, .indirections = 0, @@ -191,7 +181,7 @@ static struct r500_fragment_shader r500_texture_fragment_shader = { .instruction_count = 2, .instructions[0].inst0 = R500_INST_TYPE_TEX | R500_INST_TEX_SEM_WAIT | - R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | + R500_INST_RGB_WMASK_RGB | R500_INST_ALPHA_WMASK | R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, .instructions[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD | R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED, diff --git a/src/gallium/drivers/r300/r300_state_tcl.c b/src/gallium/drivers/r300/r300_state_tcl.c index 47d6c6dfcd..8cf8250425 100644 --- a/src/gallium/drivers/r300/r300_state_tcl.c +++ b/src/gallium/drivers/r300/r300_state_tcl.c @@ -34,11 +34,20 @@ static void r300_vs_declare(struct r300_vs_asm* assembler, assembler->tab[decl->DeclarationRange.First] = 0; break; case TGSI_SEMANTIC_COLOR: - assembler->tab[decl->DeclarationRange.First] = 2; + assembler->tab[decl->DeclarationRange.First] = + (assembler->point_size ? 1 : 0) + + assembler->out_colors; break; + case TGSI_SEMANTIC_FOG: case TGSI_SEMANTIC_GENERIC: /* XXX multiple? */ - assembler->tab[decl->DeclarationRange.First] = 6; + assembler->tab[decl->DeclarationRange.First] = + (assembler->point_size ? 1 : 0) + + assembler->out_colors + + assembler->out_texcoords; + break; + case TGSI_SEMANTIC_PSIZE: + assembler->tab[decl->DeclarationRange.First] = 1; break; default: debug_printf("r300: vs: Bad semantic declaration %d\n", @@ -62,16 +71,13 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler, { switch (src->File) { case TGSI_FILE_NULL: - /* Probably a zero or one swizzle */ - return R300_PVS_SRC_REG_INPUT; - break; case TGSI_FILE_INPUT: + /* Probably a zero or one swizzle */ return R300_PVS_SRC_REG_INPUT; - break; case TGSI_FILE_TEMPORARY: return R300_PVS_SRC_REG_TEMPORARY; - break; case TGSI_FILE_CONSTANT: + case TGSI_FILE_IMMEDIATE: return R300_PVS_SRC_REG_CONSTANT; default: debug_printf("r300: vs: Unimplemented src type %d\n", src->File); @@ -80,16 +86,32 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler, return 0; } +static INLINE unsigned r300_vs_src(struct r300_vs_asm* assembler, + struct tgsi_src_register* src) +{ + switch (src->File) { + case TGSI_FILE_NULL: + case TGSI_FILE_INPUT: + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_CONSTANT: + return src->Index; + case TGSI_FILE_IMMEDIATE: + return src->Index + assembler->imm_offset; + default: + debug_printf("r300: vs: Unimplemented src type %d\n", src->File); + break; + } + return 0; +} + static INLINE unsigned r300_vs_dst_type(struct r300_vs_asm* assembler, struct tgsi_dst_register* dst) { switch (dst->File) { case TGSI_FILE_TEMPORARY: return R300_PVS_DST_REG_TEMPORARY; - break; case TGSI_FILE_OUTPUT: return R300_PVS_DST_REG_OUT; - break; default: debug_printf("r300: vs: Unimplemented dst type %d\n", dst->File); break; @@ -103,10 +125,8 @@ static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler, switch (dst->File) { case TGSI_FILE_TEMPORARY: return dst->Index; - break; case TGSI_FILE_OUTPUT: return assembler->tab[dst->Index]; - break; default: debug_printf("r300: vs: Unimplemented dst %d\n", dst->File); break; @@ -117,12 +137,22 @@ static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler, static uint32_t r300_vs_op(unsigned op) { switch (op) { + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + return R300_VE_DOT_PRODUCT; case TGSI_OPCODE_MUL: return R300_VE_MULTIPLY; case TGSI_OPCODE_ADD: case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SUB: case TGSI_OPCODE_SWZ: return R300_VE_ADD; + case TGSI_OPCODE_MAX: + return R300_VE_MAXIMUM; + case TGSI_OPCODE_SLT: + return R300_VE_SET_LESS_THAN; + case TGSI_OPCODE_RSQ: + return R300_PVS_DST_MATH_INST | R300_ME_RECIP_DX; case TGSI_OPCODE_MAD: return R300_PVS_DST_MACRO_INST | R300_PVS_MACRO_OP_2CLK_MADD; default: @@ -134,51 +164,78 @@ static uint32_t r300_vs_op(unsigned op) static uint32_t r300_vs_swiz(struct tgsi_full_src_register* reg) { if (reg->SrcRegister.Extended) { - return reg->SrcRegisterExtSwz.ExtSwizzleX | + return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | + reg->SrcRegisterExtSwz.ExtSwizzleX | (reg->SrcRegisterExtSwz.ExtSwizzleY << 3) | (reg->SrcRegisterExtSwz.ExtSwizzleZ << 6) | (reg->SrcRegisterExtSwz.ExtSwizzleW << 9); } else { - return reg->SrcRegister.SwizzleX | + return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | + reg->SrcRegister.SwizzleX | (reg->SrcRegister.SwizzleY << 3) | (reg->SrcRegister.SwizzleZ << 6) | (reg->SrcRegister.SwizzleW << 9); } } +/* XXX icky icky icky icky */ +static uint32_t r300_vs_scalar_swiz(struct tgsi_full_src_register* reg) +{ + if (reg->SrcRegister.Extended) { + return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | + reg->SrcRegisterExtSwz.ExtSwizzleX | + (reg->SrcRegisterExtSwz.ExtSwizzleX << 3) | + (reg->SrcRegisterExtSwz.ExtSwizzleX << 6) | + (reg->SrcRegisterExtSwz.ExtSwizzleX << 9); + } else { + return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | + reg->SrcRegister.SwizzleX | + (reg->SrcRegister.SwizzleX << 3) | + (reg->SrcRegister.SwizzleX << 6) | + (reg->SrcRegister.SwizzleX << 9); + } +} + +/* XXX scalar stupidity */ static void r300_vs_emit_inst(struct r300_vertex_shader* vs, struct r300_vs_asm* assembler, struct tgsi_full_src_register* src, struct tgsi_full_dst_register* dst, unsigned op, - unsigned count) + unsigned count, + boolean is_scalar) { int i = vs->instruction_count; vs->instructions[i].inst0 = R300_PVS_DST_OPCODE(r300_vs_op(op)) | R300_PVS_DST_REG_TYPE(r300_vs_dst_type(assembler, &dst->DstRegister)) | R300_PVS_DST_OFFSET(r300_vs_dst(assembler, &dst->DstRegister)) | - R300_PVS_DST_WE_XYZW; + R300_PVS_DST_WE(dst->DstRegister.WriteMask); switch (count) { case 3: vs->instructions[i].inst3 = R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, &src[2].SrcRegister)) | - R300_PVS_SRC_OFFSET(src[2].SrcRegister.Index) | + R300_PVS_SRC_OFFSET(r300_vs_src(assembler, + &src[2].SrcRegister)) | R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[2])); /* Fall through */ case 2: vs->instructions[i].inst2 = R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, &src[1].SrcRegister)) | - R300_PVS_SRC_OFFSET(src[1].SrcRegister.Index) | + R300_PVS_SRC_OFFSET(r300_vs_src(assembler, + &src[1].SrcRegister)) | R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[1])); /* Fall through */ case 1: vs->instructions[i].inst1 = R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, &src[0].SrcRegister)) | - R300_PVS_SRC_OFFSET(src[0].SrcRegister.Index) | - R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[0])); + R300_PVS_SRC_OFFSET(r300_vs_src(assembler, + &src[0].SrcRegister)) | + /* XXX the icky, it burns */ + R300_PVS_SRC_SWIZZLE(is_scalar ? r300_vs_scalar_swiz(&src[0]) + : r300_vs_swiz(&src[0])); break; } vs->instruction_count++; @@ -189,23 +246,64 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs, struct tgsi_full_instruction* inst) { switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_RSQ: + r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, + 1, TRUE); + break; + case TGSI_OPCODE_SUB: + inst->FullSrcRegisters[1].SrcRegister.Negate = + !inst->FullSrcRegisters[1].SrcRegister.Negate; + /* Fall through */ case TGSI_OPCODE_ADD: case TGSI_OPCODE_MUL: + case TGSI_OPCODE_MAX: + case TGSI_OPCODE_SLT: + r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, + &inst->FullDstRegisters[0], inst->Instruction.Opcode, + 2, FALSE); + break; + case TGSI_OPCODE_DP3: + /* Set alpha swizzle to zero for src0 and src1 */ + if (!inst->FullSrcRegisters[0].SrcRegister.Extended) { + inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX = + inst->FullSrcRegisters[0].SrcRegister.SwizzleX; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY = + inst->FullSrcRegisters[0].SrcRegister.SwizzleY; + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ = + inst->FullSrcRegisters[0].SrcRegister.SwizzleZ; + } + inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = + TGSI_EXTSWIZZLE_ZERO; + if (!inst->FullSrcRegisters[1].SrcRegister.Extended) { + inst->FullSrcRegisters[1].SrcRegister.Extended = TRUE; + inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleX = + inst->FullSrcRegisters[1].SrcRegister.SwizzleX; + inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleY = + inst->FullSrcRegisters[1].SrcRegister.SwizzleY; + inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleZ = + inst->FullSrcRegisters[1].SrcRegister.SwizzleZ; + } + inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleW = + TGSI_EXTSWIZZLE_ZERO; + /* Fall through */ + case TGSI_OPCODE_DP4: r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 2); + 2, FALSE); break; case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: inst->FullSrcRegisters[1] = r300_constant_zero; r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 2); + 2, FALSE); break; case TGSI_OPCODE_MAD: r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 3); + 3, FALSE); break; case TGSI_OPCODE_END: break; @@ -216,6 +314,30 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs, } } +static void r300_vs_init(struct r300_vertex_shader* vs, + struct r300_vs_asm* assembler) +{ + struct tgsi_shader_info* info = &vs->info; + int i; + + for (i = 0; i < info->num_outputs; i++) { + switch (info->output_semantic_name[i]) { + case TGSI_SEMANTIC_PSIZE: + assembler->point_size = TRUE; + break; + case TGSI_SEMANTIC_COLOR: + assembler->out_colors++; + break; + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + assembler->out_texcoords++; + break; + } + } + + vs->instruction_count = 0; +} + void r300_translate_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs) { @@ -228,6 +350,10 @@ void r300_translate_vertex_shader(struct r300_context* r300, if (assembler == NULL) { return; } + + /* Init assembler. */ + r300_vs_init(vs, assembler); + /* Setup starting offset for immediates. */ assembler->imm_offset = consts->user_count; @@ -269,6 +395,7 @@ void r300_translate_vertex_shader(struct r300_context* r300, assembler->tex_count + assembler->color_count); consts->count = consts->user_count + assembler->imm_count; + vs->uses_imms = assembler->imm_count; debug_printf("r300: vs: %d total constants, " "%d from user and %d from immediates\n", consts->count, consts->user_count, assembler->imm_count); diff --git a/src/gallium/drivers/r300/r300_state_tcl.h b/src/gallium/drivers/r300/r300_state_tcl.h index 3d10e248e1..2c8b586c2f 100644 --- a/src/gallium/drivers/r300/r300_state_tcl.h +++ b/src/gallium/drivers/r300/r300_state_tcl.h @@ -32,8 +32,13 @@ /* XXX get these to r300_reg */ #define R300_PVS_DST_OPCODE(x) ((x) << 0) +# define R300_VE_DOT_PRODUCT 1 # define R300_VE_MULTIPLY 2 # define R300_VE_ADD 3 +# define R300_VE_MAXIMUM 7 +# define R300_VE_SET_LESS_THAN 10 +#define R300_PVS_DST_MATH_INST (1 << 6) +# define R300_ME_RECIP_DX 6 #define R300_PVS_DST_MACRO_INST (1 << 7) # define R300_PVS_MACRO_OP_2CLK_MADD 0 #define R300_PVS_DST_REG_TYPE(x) ((x) << 8) @@ -71,6 +76,13 @@ ((R300_PVS_SRC_SELECT_FORCE_1 | (R300_PVS_SRC_SELECT_FORCE_1 << 3) | \ (R300_PVS_SRC_SELECT_FORCE_1 << 6) | \ (R300_PVS_SRC_SELECT_FORCE_1 << 9)) << 13) +#define R300_PVS_MODIFIER_X (1 << 25) +#define R300_PVS_MODIFIER_Y (1 << 26) +#define R300_PVS_MODIFIER_Z (1 << 27) +#define R300_PVS_MODIFIER_W (1 << 28) +#define R300_PVS_NEGATE_XYZW \ + (R300_PVS_MODIFIER_X | R300_PVS_MODIFIER_Y | \ + R300_PVS_MODIFIER_Z | R300_PVS_MODIFIER_W) static const struct tgsi_full_src_register r300_constant_zero = { .SrcRegister.Extended = TRUE, @@ -98,7 +110,13 @@ struct r300_vs_asm { unsigned imm_offset; /* Number of immediate constants. */ unsigned imm_count; - /* Offsets into vertex output memory. */ + /* Number of colors to write. */ + unsigned out_colors; + /* Number of texcoords to write. */ + unsigned out_texcoords; + /* Whether to emit point size. */ + boolean point_size; + /* Tab of declared outputs to OVM outputs. */ unsigned tab[16]; }; @@ -114,7 +132,7 @@ static struct r300_vertex_shader r300_passthrough_vertex_shader = { .instructions[0].inst3 = 0x0, .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | - R300_PVS_DST_OFFSET(2) | R300_PVS_DST_WE_XYZW, + R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW, .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW, .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, @@ -133,7 +151,7 @@ static struct r300_vertex_shader r300_texture_vertex_shader = { .instructions[0].inst3 = 0x0, .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | - R300_PVS_DST_OFFSET(6) | R300_PVS_DST_WE_XYZW, + R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW, .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW, .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c index 79bed03253..c9e2dff14e 100644 --- a/src/gallium/drivers/r300/r300_surface.c +++ b/src/gallium/drivers/r300/r300_surface.c @@ -23,15 +23,13 @@ #include "r300_surface.h" -static void r300_surface_setup(struct pipe_context* pipe, - struct pipe_surface* dest, +static void r300_surface_setup(struct r300_context* r300, + struct r300_texture* dest, unsigned x, unsigned y, unsigned w, unsigned h) { - struct r300_context* r300 = r300_context(pipe); - struct r300_capabilities* caps = r300_screen(pipe->screen)->caps; - struct r300_texture* tex = (struct r300_texture*)dest->texture; - unsigned pixpitch = tex->stride / tex->tex.block.size; + struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; + unsigned pixpitch = dest->stride / dest->tex.block.size; CS_LOCALS(r300); r300_emit_blend_state(r300, &blend_clear_state); @@ -39,7 +37,22 @@ static void r300_surface_setup(struct pipe_context* pipe, r300_emit_dsa_state(r300, &dsa_clear_state); r300_emit_rs_state(r300, &rs_clear_state); - BEGIN_CS(15); + BEGIN_CS(24); + + /* Viewport setup */ + OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6); + OUT_CS_32F((float)w); + OUT_CS_32F((float)x); + OUT_CS_32F((float)h); + OUT_CS_32F((float)y); + OUT_CS_32F(1.0); + OUT_CS_32F(0.0); + + OUT_CS_REG(R300_VAP_VTE_CNTL, R300_VPORT_X_SCALE_ENA | + R300_VPORT_X_OFFSET_ENA | + R300_VPORT_Y_SCALE_ENA | + R300_VPORT_Y_OFFSET_ENA | + R300_VTX_XY_FMT | R300_VTX_Z_FMT); /* Pixel scissors. */ OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2); @@ -64,9 +77,9 @@ static void r300_surface_setup(struct pipe_context* pipe, /* Setup colorbuffer. */ OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_RELOC(dest->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG(R300_RB3D_COLORPITCH0, pixpitch | - r300_translate_colorformat(tex->tex.format)); + r300_translate_colorformat(dest->tex.format)); OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0xf); END_CS; @@ -86,6 +99,7 @@ static void r300_surface_fill(struct pipe_context* pipe, struct r300_capabilities* caps = r300_screen(pipe->screen)->caps; struct r300_texture* tex = (struct r300_texture*)dest->texture; unsigned pixpitch = tex->stride / tex->tex.block.size; + boolean invalid = FALSE; CS_LOCALS(r300); a = (float)((color >> 24) & 0xff) / 255.0f; @@ -98,12 +112,29 @@ static void r300_surface_fill(struct pipe_context* pipe, /* Fallback? */ if (FALSE) { +fallback: debug_printf("r300: Falling back on surface clear..."); util_surface_fill(pipe, dest, x, y, w, h, color); return; } - r300_surface_setup(r300, dest, x, y, w, h); + /* Make sure our target BO is okay. */ +validate: + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + if (r300->winsys->validate(r300->winsys)) { + r300->context.flush(&r300->context, 0, NULL); + if (invalid) { + goto fallback; + } + invalid = TRUE; + goto validate; + } + + r300_surface_setup(r300, tex, x, y, w, h); /* Vertex shader setup */ if (caps->has_tcl) { @@ -127,7 +158,7 @@ static void r300_surface_fill(struct pipe_context* pipe, r300_emit_rs_block_state(r300, &r300_rs_block_clear_state); } - BEGIN_CS(31); + BEGIN_CS(26); /* VAP stream control, mapping from input memory to PVS/RS memory */ if (caps->has_tcl) { @@ -154,27 +185,21 @@ static void r300_surface_fill(struct pipe_context* pipe, /* Disable textures */ OUT_CS_REG(R300_TX_ENABLE, 0x0); - /* Viewport setup */ - OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6); - OUT_CS_32F(1.0); - OUT_CS_32F((float)x); - OUT_CS_32F(1.0); - OUT_CS_32F((float)y); - OUT_CS_32F(1.0); - OUT_CS_32F(0.0); - /* The size of the point we're about to draw, in sixths of pixels */ OUT_CS_REG(R300_GA_POINT_SIZE, - ((h * 6) & R300_POINTSIZE_Y_MASK) | + ((h * 6) & R300_POINTSIZE_Y_MASK) | ((w * 6) << R300_POINTSIZE_X_SHIFT)); + /* Vertex size. */ + OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8); + /* Packet3 with our point vertex */ OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8); OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | (1 << R300_PRIM_NUM_VERTICES_SHIFT)); /* Position */ - OUT_CS_32F(w / 2.0); - OUT_CS_32F(h / 2.0); + OUT_CS_32F(0.5); + OUT_CS_32F(0.5); OUT_CS_32F(1.0); OUT_CS_32F(1.0); /* Color */ @@ -183,11 +208,7 @@ static void r300_surface_fill(struct pipe_context* pipe, OUT_CS_32F(b); OUT_CS_32F(a); - /* XXX figure out why this is 0xA and not 0x2 */ OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA); - /* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, - R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | - R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */ END_CS; @@ -206,6 +227,7 @@ static void r300_surface_copy(struct pipe_context* pipe, struct r300_texture* srctex = (struct r300_texture*)src->texture; struct r300_texture* desttex = (struct r300_texture*)dest->texture; unsigned pixpitch = srctex->stride / srctex->tex.block.size; + boolean invalid = FALSE; CS_LOCALS(r300); debug_printf("r300: Copying surface %p at (%d,%d) to %p at (%d, %d)," @@ -215,18 +237,48 @@ static void r300_surface_copy(struct pipe_context* pipe, if ((srctex == desttex) && ((destx < srcx + w) || (srcx < destx + w)) && ((desty < srcy + h) || (srcy < desty + h))) { +fallback: debug_printf("r300: Falling back on surface_copy\n"); util_surface_copy(pipe, FALSE, dest, destx, desty, src, srcx, srcy, w, h); } + /* Add our target BOs to the list. */ +validate: + if (!r300->winsys->add_buffer(r300->winsys, srctex->buffer, + RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + if (!r300->winsys->add_buffer(r300->winsys, desttex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + if (r300->winsys->validate(r300->winsys)) { + r300->context.flush(&r300->context, 0, NULL); + if (invalid) { + goto fallback; + } + invalid = TRUE; + goto validate; + } + + r300_surface_setup(r300, desttex, destx, desty, w, h); + + /* Setup the texture. */ r300_emit_sampler(r300, &r300_sampler_copy_state, 0); r300_emit_texture(r300, srctex, 0); - r300_flush_textures(r300); + + /* Flush and enable. */ + BEGIN_CS(4); + OUT_CS_REG(R300_TX_INVALTAGS, 0); + OUT_CS_REG(R300_TX_ENABLE, 0x1); + END_CS; /* Vertex shader setup */ if (caps->has_tcl) { - r300_emit_vertex_shader(r300, &r300_texture_vertex_shader); + r300_emit_vertex_shader(r300, &r300_passthrough_vertex_shader); } else { BEGIN_CS(4); OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS); @@ -246,6 +298,7 @@ static void r300_surface_copy(struct pipe_context* pipe, r300_emit_rs_block_state(r300, &r300_rs_block_copy_state); } + BEGIN_CS(30); /* VAP stream control, mapping from input memory to PVS/RS memory */ if (caps->has_tcl) { OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, @@ -268,33 +321,38 @@ static void r300_surface_copy(struct pipe_context* pipe, /* Two components of texture 0 */ OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x2); + /* Vertex size. */ + OUT_CS_REG(R300_VAP_VTX_SIZE, 0x4); + /* Packet3 with our texcoords */ - OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8); + OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 16); OUT_CS(R300_PRIM_TYPE_QUADS | R300_PRIM_WALK_RING | (4 << R300_PRIM_NUM_VERTICES_SHIFT)); /* (x , y ) */ - OUT_CS_32F((float)destx); - OUT_CS_32F((float)desty); - OUT_CS_32F((float)srcx); - OUT_CS_32F((float)srcy); + OUT_CS_32F((float)(destx / dest->width)); + OUT_CS_32F((float)(desty / dest->height)); + OUT_CS_32F((float)(srcx / dest->width)); + OUT_CS_32F((float)(srcy / dest->height)); /* (x , y + h) */ - OUT_CS_32F((float)destx); - OUT_CS_32F((float)(desty + h)); - OUT_CS_32F((float)srcx); - OUT_CS_32F((float)(srcy + h)); + OUT_CS_32F((float)(destx / dest->width)); + OUT_CS_32F((float)((desty + h) / dest->height)); + OUT_CS_32F((float)(srcx / dest->width)); + OUT_CS_32F((float)((srcy + h) / dest->height)); /* (x + w, y + h) */ - OUT_CS_32F((float)(destx + w)); - OUT_CS_32F((float)(desty + h)); - OUT_CS_32F((float)(srcx + w)); - OUT_CS_32F((float)(srcy + h)); + OUT_CS_32F((float)((destx + w) / dest->width)); + OUT_CS_32F((float)((desty + h) / dest->height)); + OUT_CS_32F((float)((srcx + w) / dest->width)); + OUT_CS_32F((float)((srcy + h) / dest->height)); /* (x + w, y ) */ - OUT_CS_32F((float)(destx + w)); - OUT_CS_32F((float)desty); - OUT_CS_32F((float)(srcx + w)); - OUT_CS_32F((float)srcy); + OUT_CS_32F((float)((destx + w) / dest->width)); + OUT_CS_32F((float)(desty / dest->height)); + OUT_CS_32F((float)((srcx + w) / dest->width)); + OUT_CS_32F((float)(srcy / dest->height)); OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA); + END_CS; + r300->dirty_hw++; } diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h index 894def07aa..9a4c39f58b 100644 --- a/src/gallium/drivers/r300/r300_surface.h +++ b/src/gallium/drivers/r300/r300_surface.h @@ -101,7 +101,7 @@ static struct r300_rs_block r300_rs_block_copy_state = { R500_RS_SEL_Q(R300_RS_SEL_K1), .inst[0] = R300_RS_INST_COL_CN_WRITE, .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN, - .inst_count = R300_RS_TX_OFFSET(6), + .inst_count = R300_RS_TX_OFFSET(0), }; static struct r300_rs_block r500_rs_block_copy_state = { @@ -111,7 +111,7 @@ static struct r300_rs_block r500_rs_block_copy_state = { R500_RS_SEL_Q(R500_RS_IP_PTR_K1), .inst[0] = R500_RS_INST_TEX_CN_WRITE, .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN, - .inst_count = R300_RS_TX_OFFSET(6), + .inst_count = R300_RS_TX_OFFSET(0), }; static struct r300_sampler_state r300_sampler_copy_state = { diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index fe91f4e184..5ea9f56247 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -86,8 +86,6 @@ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen, const struct pipe_texture* template) { - /* XXX struct r300_screen* r300screen = r300_screen(screen); */ - struct r300_texture* tex = CALLOC_STRUCT(r300_texture); if (!tex) { diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h index baa95282c3..d2893c3b9d 100644 --- a/src/gallium/drivers/r300/r300_winsys.h +++ b/src/gallium/drivers/r300/r300_winsys.h @@ -35,8 +35,6 @@ extern "C" { #include "pipe/p_state.h" #include "pipe/internal/p_winsys_screen.h" -struct radeon_cs; - struct r300_winsys { /* Parent class */ struct pipe_winsys base; @@ -47,42 +45,50 @@ struct r300_winsys { /* PCI ID */ uint32_t pci_id; - /* GB pipe count */ - uint32_t gb_pipes; + /* GART size. */ + uint32_t gart_size; + + /* VRAM size. */ + uint32_t vram_size; + + /* Add a pipe_buffer to the list of buffer objects to validate. */ + boolean (*add_buffer)(struct r300_winsys* winsys, + struct pipe_buffer* pbuffer, + uint32_t rd, + uint32_t wd); - /* CS object. This is very much like Intel's batchbuffer. - * Fill it full of dwords and relocs and then submit. - * Repeat as needed. */ - struct radeon_cs* cs; + /* Revalidate all currently setup pipe_buffers. + * Returns TRUE if a flush is required. */ + boolean (*validate)(struct r300_winsys* winsys); /* Check to see if there's room for commands. */ - boolean (*check_cs)(struct radeon_cs* cs, int size); + boolean (*check_cs)(struct r300_winsys* winsys, int size); /* Start a command emit. */ - void (*begin_cs)(struct radeon_cs* cs, - int size, - const char* file, - const char* function, - int line); + void (*begin_cs)(struct r300_winsys* winsys, + int size, + const char* file, + const char* function, + int line); /* Write a dword to the command buffer. */ - void (*write_cs_dword)(struct radeon_cs* cs, uint32_t dword); + void (*write_cs_dword)(struct r300_winsys* winsys, uint32_t dword); /* Write a relocated dword to the command buffer. */ - void (*write_cs_reloc)(struct radeon_cs* cs, - struct pipe_buffer* bo, - uint32_t rd, - uint32_t wd, - uint32_t flags); + void (*write_cs_reloc)(struct r300_winsys* winsys, + struct pipe_buffer* bo, + uint32_t rd, + uint32_t wd, + uint32_t flags); /* Finish a command emit. */ - void (*end_cs)(struct radeon_cs* cs, - const char* file, - const char* function, - int line); + void (*end_cs)(struct r300_winsys* winsys, + const char* file, + const char* function, + int line); /* Flush the CS. */ - void (*flush_cs)(struct radeon_cs* cs); + void (*flush_cs)(struct r300_winsys* winsys); }; struct pipe_context* r300_create_context(struct pipe_screen* screen, diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 06ace27d14..62e8d99cfd 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -121,11 +121,23 @@ static void softpipe_destroy( struct pipe_context *pipe ) FREE( softpipe ); } +static unsigned int +softpipe_is_texture_referenced( struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) +{ + return PIPE_UNREFERENCED; +} + +static unsigned int +softpipe_is_buffer_referenced( struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + return PIPE_UNREFERENCED; +} struct pipe_context * -softpipe_create( struct pipe_screen *screen, - struct pipe_winsys *pipe_winsys, - void *unused ) +softpipe_create( struct pipe_screen *screen ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; @@ -140,7 +152,7 @@ softpipe_create( struct pipe_screen *screen, softpipe->dump_fs = debug_get_bool_option( "GALLIUM_DUMP_FS", FALSE ); - softpipe->pipe.winsys = pipe_winsys; + softpipe->pipe.winsys = screen->winsys; softpipe->pipe.screen = screen; softpipe->pipe.destroy = softpipe_destroy; @@ -190,6 +202,9 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; + softpipe->pipe.is_texture_referenced = softpipe_is_texture_referenced; + softpipe->pipe.is_buffer_referenced = softpipe_is_buffer_referenced; + softpipe_init_query_funcs( softpipe ); softpipe_init_texture_funcs( softpipe ); diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 59d6df8f2d..b89a7292e5 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -92,6 +92,7 @@ struct softpipe_context { * queries. */ uint64_t occlusion_count; + unsigned active_query_count; /** Mapped vertex buffers */ ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS]; diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index eef6e5806c..06725fd09b 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -236,7 +236,6 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) } break; - case PIPE_PRIM_TRIANGLES: for (i = 2; i < nr; i += 3) { setup_tri( setup_ctx, @@ -256,7 +255,6 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) break; case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, indices[0], stride), @@ -264,6 +262,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) get_vert(vertex_buffer, indices[i-0], stride)); } break; + case PIPE_PRIM_QUADS: for (i = 3; i < nr; i += 4) { setup_tri( setup_ctx, @@ -277,6 +276,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) get_vert(vertex_buffer, indices[i-0], stride)); } break; + case PIPE_PRIM_QUAD_STRIP: for (i = 3; i < nr; i += 2) { setup_tri( setup_ctx, @@ -290,6 +290,16 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) get_vert(vertex_buffer, indices[i-0], stride)); } break; + + case PIPE_PRIM_POLYGON: + for (i = 2; i < nr; i += 1) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[0-1], stride), + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[0], stride)); + } + break; + default: assert(0); } @@ -378,7 +388,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) break; case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, 0, stride), @@ -386,6 +395,7 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) get_vert(vertex_buffer, i-0, stride)); } break; + case PIPE_PRIM_QUADS: for (i = 3; i < nr; i += 4) { setup_tri( setup_ctx, @@ -412,6 +422,20 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) get_vert(vertex_buffer, i-0, stride)); } break; + + case PIPE_PRIM_POLYGON: + /* Almost same as tri fan but the _first_ vertex specifies the flat + * shading color. Note that the first polygon vertex is passed as + * the last triangle vertex here. + */ + for (i = 2; i < nr; i += 1) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, 0, stride)); + } + break; + default: assert(0); } diff --git a/src/gallium/drivers/softpipe/sp_quad_pipe.c b/src/gallium/drivers/softpipe/sp_quad_pipe.c index 892ef87ee9..b5f69b7426 100644 --- a/src/gallium/drivers/softpipe/sp_quad_pipe.c +++ b/src/gallium/drivers/softpipe/sp_quad_pipe.c @@ -80,7 +80,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp_push_quad_first( sp, sp->quad[i].blend, i ); } - if (sp->depth_stencil->depth.occlusion_count) { + if (sp->active_query_count) { sp_push_quad_first( sp, sp->quad[i].occlusion, i ); } diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c index 93dab236d6..379cf4ad06 100644 --- a/src/gallium/drivers/softpipe/sp_query.c +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -34,6 +34,7 @@ #include "util/u_memory.h" #include "sp_context.h" #include "sp_query.h" +#include "sp_state.h" struct softpipe_query { uint64_t start; @@ -69,6 +70,8 @@ softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q) struct softpipe_query *sq = softpipe_query(q); sq->start = softpipe->occlusion_count; + softpipe->active_query_count++; + softpipe->dirty |= SP_NEW_QUERY; } @@ -78,7 +81,9 @@ softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q) struct softpipe_context *softpipe = softpipe_context( pipe ); struct softpipe_query *sq = softpipe_query(q); + softpipe->active_query_count--; sq->end = softpipe->occlusion_count; + softpipe->dirty |= SP_NEW_QUERY; } diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 7380a6ae2b..692deeb8fd 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -82,11 +82,11 @@ softpipe_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_TEXTURE_SHADOW_MAP: return 1; case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ + return 13; /* max 4Kx4K */ case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: return 8; /* max 128x128x128 */ case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ + return 13; /* max 4Kx4K */ default: return 0; } @@ -106,7 +106,7 @@ softpipe_get_paramf(struct pipe_screen *screen, int param) case PIPE_CAP_MAX_POINT_WIDTH_AA: return 255.0; /* arbitrary */ case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; + return 16.0; /* not actually signficant at this time */ case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: return 16.0; /* arbitrary */ default: @@ -127,6 +127,11 @@ softpipe_is_format_supported( struct pipe_screen *screen, unsigned tex_usage, unsigned geom_flags ) { + assert(target == PIPE_TEXTURE_1D || + target == PIPE_TEXTURE_2D || + target == PIPE_TEXTURE_3D || + target == PIPE_TEXTURE_CUBE); + switch(format) { case PIPE_FORMAT_DXT1_RGB: case PIPE_FORMAT_DXT1_RGBA: diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index accc692b66..c6844a2649 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -504,6 +504,8 @@ static void print_vertex(const struct setup_context *setup, #endif /** + * Sort the vertices from top to bottom order, setting up the triangle + * edge fields (ebot, emaj, etop). * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise */ static boolean setup_sort_vertices( struct setup_context *setup, diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index c0113c47ad..7a533dad9f 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -42,6 +42,7 @@ #include "sp_texture.h" #include "sp_tile_cache.h" #include "sp_screen.h" +#include "sp_winsys.h" /* Simple, maximally packed layout. @@ -399,3 +400,22 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen) screen->transfer_map = softpipe_transfer_map; screen->transfer_unmap = softpipe_transfer_unmap; } + + +boolean +softpipe_get_texture_buffer( struct pipe_texture *texture, + struct pipe_buffer **buf, + unsigned *stride ) +{ + struct softpipe_texture *tex = (struct softpipe_texture *)texture; + + if (!tex) + return FALSE; + + pipe_buffer_reference(buf, tex->buffer); + + if (stride) + *stride = tex->stride[0]; + + return TRUE; +} diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index 4ab666486c..9e571862b7 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -35,37 +35,29 @@ #define SP_WINSYS_H -#include "pipe/p_compiler.h" /* for boolean */ - - #ifdef __cplusplus extern "C" { #endif -enum pipe_format; - -struct softpipe_winsys { - /** test if the given format is supported for front/back color bufs */ - boolean (*is_format_supported)( struct softpipe_winsys *sws, - enum pipe_format format ); - -}; - struct pipe_screen; struct pipe_winsys; struct pipe_context; -struct pipe_context *softpipe_create( struct pipe_screen *, - struct pipe_winsys *, - void *unused ); +struct pipe_context *softpipe_create( struct pipe_screen * ); struct pipe_screen * softpipe_create_screen(struct pipe_winsys *); +boolean +softpipe_get_texture_buffer( struct pipe_texture *texture, + struct pipe_buffer **buf, + unsigned *stride ); + + #ifdef __cplusplus } #endif diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile index e087db169a..4aeb8e3d7e 100644 --- a/src/gallium/drivers/trace/Makefile +++ b/src/gallium/drivers/trace/Makefile @@ -7,8 +7,10 @@ C_SOURCES = \ tr_buffer.c \ tr_context.c \ tr_dump.c \ + tr_dump_state.c \ tr_screen.c \ tr_state.c \ + tr_rbug.c \ tr_texture.c include ../../Makefile.template diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 73dce20372..1000c31e49 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -3,7 +3,8 @@ = About = -This directory contains a Gallium3D pipe driver which traces all incoming calls. +This directory contains a Gallium3D debugger pipe driver. +It can traces all incoming calls and/or provide remote debugging functionality. = Build Instructions = @@ -23,7 +24,9 @@ ensure the right libGL.so is being picked by doing ldd progs/trivial/tri -and then try running +== Traceing == + +For traceing then do export XMESA_TRACE=y GALLIUM_TRACE=tri.trace progs/trivial/tri @@ -32,6 +35,16 @@ which should create a tri.trace file, which is an XML file. You can view copying trace.xsl to the same directory, and opening with a XSLT capable browser such as Firefox or Internet Explorer. +== Remote debugging == + +For remote debugging + + export XMESA_TRACE=y + GALLIUM_RBUG=true progs/trivial/tri + +which should open gallium remote debugging session. While the program is running +you can launch the small remote debugging application from progs/rbug. More +information is in that directory. = Integrating = @@ -62,3 +75,4 @@ trace_screen with real_screen when creating them. -- Jose Fonseca <jrfonseca@tungstengraphics.com> +Jakob Bornecrantz <jakob@vmware.com> diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index 4215215d1a..e635fed77d 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -8,8 +8,10 @@ trace = env.ConvenienceLibrary( 'tr_buffer.c', 'tr_context.c', 'tr_dump.c', + 'tr_dump_state.c', 'tr_screen.c', 'tr_state.c', + 'tr_rbug.c', 'tr_texture.c', ]) diff --git a/src/gallium/drivers/trace/tr_buffer.c b/src/gallium/drivers/trace/tr_buffer.c index 6ffce1660e..4f0eff6a5a 100644 --- a/src/gallium/drivers/trace/tr_buffer.c +++ b/src/gallium/drivers/trace/tr_buffer.c @@ -27,10 +27,10 @@ #include "util/u_memory.h" +#include "util/u_simple_list.h" #include "tr_buffer.h" - struct pipe_buffer * trace_buffer_create(struct trace_screen *tr_scr, struct pipe_buffer *buffer) @@ -52,6 +52,8 @@ trace_buffer_create(struct trace_screen *tr_scr, tr_buf->base.screen = &tr_scr->base; tr_buf->buffer = buffer; + trace_screen_add_to_list(tr_scr, buffers, tr_buf); + return &tr_buf->base; error: @@ -64,7 +66,10 @@ void trace_buffer_destroy(struct trace_screen *tr_scr, struct pipe_buffer *buffer) { - struct trace_buffer *tr_buf = trace_buffer(tr_scr, buffer); + struct trace_buffer *tr_buf = trace_buffer(buffer); + + trace_screen_remove_from_list(tr_scr, buffers, tr_buf); + pipe_buffer_reference(&tr_buf->buffer, NULL); FREE(tr_buf); } diff --git a/src/gallium/drivers/trace/tr_buffer.h b/src/gallium/drivers/trace/tr_buffer.h index e9e4d354da..1a2d0b9aea 100644 --- a/src/gallium/drivers/trace/tr_buffer.h +++ b/src/gallium/drivers/trace/tr_buffer.h @@ -41,19 +41,19 @@ struct trace_buffer struct pipe_buffer *buffer; + struct tr_list list; + void *map; boolean range_flushed; }; static INLINE struct trace_buffer * -trace_buffer(struct trace_screen *tr_scr, - struct pipe_buffer *buffer) +trace_buffer(struct pipe_buffer *buffer) { if(!buffer) return NULL; - assert(tr_scr); - assert(buffer->screen == &tr_scr->base); + (void)trace_screen(buffer->screen); return (struct trace_buffer *)buffer; } diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index d8d5821a1d..4ab718f233 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -26,14 +26,16 @@ **************************************************************************/ #include "util/u_memory.h" +#include "util/u_simple_list.h" + #include "pipe/p_screen.h" #include "tr_dump.h" +#include "tr_dump_state.h" #include "tr_state.h" #include "tr_buffer.h" #include "tr_screen.h" #include "tr_texture.h" -#include "tr_context.h" static INLINE struct pipe_buffer * @@ -46,7 +48,7 @@ trace_buffer_unwrap(struct trace_context *tr_ctx, if(!buffer) return NULL; - tr_buf = trace_buffer(tr_scr, buffer); + tr_buf = trace_buffer(buffer); assert(tr_buf->buffer); assert(tr_buf->buffer->screen == tr_scr->screen); @@ -111,6 +113,65 @@ trace_context_set_edgeflags(struct pipe_context *_pipe, } +static INLINE void +trace_context_draw_block(struct trace_context *tr_ctx, int flag) +{ + int k; + + pipe_mutex_lock(tr_ctx->draw_mutex); + + if (tr_ctx->draw_blocker & flag) { + tr_ctx->draw_blocked |= flag; + } else if ((tr_ctx->draw_rule.blocker & flag) && + (tr_ctx->draw_blocker & 4)) { + boolean block = FALSE; + debug_printf("%s (%lu %lu) (%lu %lu) (%lu %u) (%lu %u)\n", __FUNCTION__, + tr_ctx->draw_rule.fs, tr_ctx->curr.fs, + tr_ctx->draw_rule.vs, tr_ctx->curr.vs, + tr_ctx->draw_rule.surf, 0, + tr_ctx->draw_rule.tex, 0); + if (tr_ctx->draw_rule.fs && + tr_ctx->draw_rule.fs == tr_ctx->curr.fs) + block = TRUE; + if (tr_ctx->draw_rule.vs && + tr_ctx->draw_rule.vs == tr_ctx->curr.vs) + block = TRUE; + if (tr_ctx->draw_rule.surf && + tr_ctx->draw_rule.surf == tr_ctx->curr.zsbuf) + block = TRUE; + if (tr_ctx->draw_rule.surf) + for (k = 0; k < tr_ctx->curr.nr_cbufs; k++) + if (tr_ctx->draw_rule.surf == tr_ctx->curr.cbufs[k]) + block = TRUE; + if (tr_ctx->draw_rule.tex) + for (k = 0; k < tr_ctx->curr.num_texs; k++) + if (tr_ctx->draw_rule.tex == tr_ctx->curr.tex[k]) + block = TRUE; + + if (block) + tr_ctx->draw_blocked |= (flag | 4); + } + + if (tr_ctx->draw_blocked) + trace_rbug_notify_draw_blocked(tr_ctx); + + /* wait for rbug to clear the blocked flag */ + while (tr_ctx->draw_blocked & flag) { + tr_ctx->draw_blocked |= flag; +#ifdef PIPE_THREAD_HAVE_CONDVAR + pipe_condvar_wait(tr_ctx->draw_cond, tr_ctx->draw_mutex); +#else + pipe_mutex_unlock(tr_ctx->draw_mutex); +#ifdef PIPE_SUBSYSTEM_WINDOWS_USER + Sleep(1); +#endif + pipe_mutex_lock(tr_ctx->draw_mutex); +#endif + } + + pipe_mutex_unlock(tr_ctx->draw_mutex); +} + static INLINE boolean trace_context_draw_arrays(struct pipe_context *_pipe, unsigned mode, unsigned start, unsigned count) @@ -119,6 +180,11 @@ trace_context_draw_arrays(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; boolean result; + if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled) + return 0; + + trace_context_draw_block(tr_ctx, 1); + trace_dump_call_begin("pipe_context", "draw_arrays"); trace_dump_arg(ptr, pipe); @@ -132,6 +198,8 @@ trace_context_draw_arrays(struct pipe_context *_pipe, trace_dump_call_end(); + trace_context_draw_block(tr_ctx, 2); + return result; } @@ -142,13 +210,17 @@ trace_context_draw_elements(struct pipe_context *_pipe, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { - struct trace_screen *tr_scr = trace_screen(_pipe->screen); struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _indexBuffer); + struct trace_buffer *tr_buf = trace_buffer(_indexBuffer); struct pipe_context *pipe = tr_ctx->pipe; struct pipe_buffer *indexBuffer = tr_buf->buffer; boolean result; + if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled) + return 0; + + trace_context_draw_block(tr_ctx, 1); + trace_screen_user_buffer_update(_pipe->screen, indexBuffer); trace_dump_call_begin("pipe_context", "draw_elements"); @@ -166,6 +238,8 @@ trace_context_draw_elements(struct pipe_context *_pipe, trace_dump_call_end(); + trace_context_draw_block(tr_ctx, 2); + return result; } @@ -180,13 +254,17 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, unsigned start, unsigned count) { - struct trace_screen *tr_scr = trace_screen(_pipe->screen); struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _indexBuffer); + struct trace_buffer *tr_buf = trace_buffer(_indexBuffer); struct pipe_context *pipe = tr_ctx->pipe; struct pipe_buffer *indexBuffer = tr_buf->buffer; boolean result; + if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled) + return 0; + + trace_context_draw_block(tr_ctx, 1); + trace_screen_user_buffer_update(_pipe->screen, indexBuffer); trace_dump_call_begin("pipe_context", "draw_range_elements"); @@ -209,6 +287,8 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, trace_dump_call_end(); + trace_context_draw_block(tr_ctx, 2); + return result; } @@ -573,23 +653,32 @@ trace_context_create_fs_state(struct pipe_context *_pipe, trace_dump_call_end(); + result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_FRAGMENT); + return result; } static INLINE void trace_context_bind_fs_state(struct pipe_context *_pipe, - void *state) + void *_state) { struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_shader *tr_shdr = trace_shader(_state); struct pipe_context *pipe = tr_ctx->pipe; + void *state = tr_shdr ? tr_shdr->state : NULL; trace_dump_call_begin("pipe_context", "bind_fs_state"); trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->bind_fs_state(pipe, state);; + tr_ctx->curr.fs = tr_shdr; + + if (tr_shdr && tr_shdr->replaced) + state = tr_shdr->replaced; + + pipe->bind_fs_state(pipe, state); trace_dump_call_end(); } @@ -597,19 +686,23 @@ trace_context_bind_fs_state(struct pipe_context *_pipe, static INLINE void trace_context_delete_fs_state(struct pipe_context *_pipe, - void *state) + void *_state) { struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_shader *tr_shdr = trace_shader(_state); struct pipe_context *pipe = tr_ctx->pipe; + void *state = tr_shdr->state; trace_dump_call_begin("pipe_context", "delete_fs_state"); trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_fs_state(pipe, state);; + pipe->delete_fs_state(pipe, state); trace_dump_call_end(); + + trace_shader_destroy(tr_ctx, tr_shdr); } @@ -626,28 +719,37 @@ trace_context_create_vs_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(shader_state, state); - result = pipe->create_vs_state(pipe, state);; + result = pipe->create_vs_state(pipe, state); trace_dump_ret(ptr, result); trace_dump_call_end(); + result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_VERTEX); + return result; } static INLINE void trace_context_bind_vs_state(struct pipe_context *_pipe, - void *state) + void *_state) { struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_shader *tr_shdr = trace_shader(_state); struct pipe_context *pipe = tr_ctx->pipe; + void *state = tr_shdr ? tr_shdr->state : NULL; trace_dump_call_begin("pipe_context", "bind_vs_state"); trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); + tr_ctx->curr.vs = tr_shdr; + + if (tr_shdr && tr_shdr->replaced) + state = tr_shdr->replaced; + pipe->bind_vs_state(pipe, state);; trace_dump_call_end(); @@ -656,10 +758,12 @@ trace_context_bind_vs_state(struct pipe_context *_pipe, static INLINE void trace_context_delete_vs_state(struct pipe_context *_pipe, - void *state) + void *_state) { struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_shader *tr_shdr = trace_shader(_state); struct pipe_context *pipe = tr_ctx->pipe; + void *state = tr_shdr->state; trace_dump_call_begin("pipe_context", "delete_vs_state"); @@ -669,6 +773,8 @@ trace_context_delete_vs_state(struct pipe_context *_pipe, pipe->delete_vs_state(pipe, state);; trace_dump_call_end(); + + trace_shader_destroy(tr_ctx, tr_shdr); } @@ -747,6 +853,19 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, struct pipe_framebuffer_state unwrapped_state; unsigned i; + { + tr_ctx->curr.nr_cbufs = state->nr_cbufs; + for (i = 0; i < state->nr_cbufs; i++) + if (state->cbufs[i]) + tr_ctx->curr.cbufs[i] = trace_texture(state->cbufs[i]->texture); + else + tr_ctx->curr.cbufs[i] = NULL; + if (state->zsbuf) + tr_ctx->curr.zsbuf = trace_texture(state->zsbuf->texture); + else + tr_ctx->curr.zsbuf = NULL; + } + /* Unwrap the input state */ memcpy(&unwrapped_state, state, sizeof(unwrapped_state)); for(i = 0; i < state->nr_cbufs; ++i) @@ -827,12 +946,17 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, struct pipe_texture **textures) { struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_texture *tr_tex; struct pipe_context *pipe = tr_ctx->pipe; struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS]; unsigned i; - for(i = 0; i < num_textures; ++i) - unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]); + tr_ctx->curr.num_texs = num_textures; + for(i = 0; i < num_textures; ++i) { + tr_tex = trace_texture(textures[i]); + tr_ctx->curr.tex[i] = tr_tex; + unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL; + } textures = unwrapped_textures; trace_dump_call_begin("pipe_context", "set_sampler_textures"); @@ -1016,20 +1140,73 @@ trace_context_flush(struct pipe_context *_pipe, static INLINE void trace_context_destroy(struct pipe_context *_pipe) { + struct trace_screen *tr_scr = trace_screen(_pipe->screen); struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin("pipe_context", "destroy"); - trace_dump_arg(ptr, pipe); + trace_dump_call_end(); + + trace_screen_remove_from_list(tr_scr, contexts, tr_ctx); pipe->destroy(pipe); + FREE(tr_ctx); +} + +static unsigned int +trace_is_texture_referenced( struct pipe_context *_pipe, + struct pipe_texture *_texture, + unsigned face, unsigned level) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_texture *tr_tex = trace_texture(_texture); + struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_texture *texture = tr_tex->texture; + unsigned int referenced; + + trace_dump_call_begin("pipe_context", "is_texture_referenced"); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, texture); + trace_dump_arg(uint, face); + trace_dump_arg(uint, level); + + referenced = pipe->is_texture_referenced(pipe, texture, face, level); + + trace_dump_ret(uint, referenced); trace_dump_call_end(); - FREE(tr_ctx); + return referenced; } +static unsigned int +trace_is_buffer_referenced( struct pipe_context *_pipe, + struct pipe_buffer *_buf) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_buffer *tr_buf = trace_buffer(_buf); + struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_buffer *buf = tr_buf->buffer; + unsigned int referenced; + + trace_dump_call_begin("pipe_context", "is_buffer_referenced"); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, buf); + + referenced = pipe->is_buffer_referenced(pipe, buf); + + trace_dump_ret(uint, referenced); + trace_dump_call_end(); + + return referenced; +} + +static const struct debug_named_value rbug_blocker_flags[] = { + {"before", 1}, + {"after", 2}, + {NULL, 0}, +}; struct pipe_context * trace_context_create(struct pipe_screen *_screen, @@ -1042,7 +1219,7 @@ trace_context_create(struct pipe_screen *_screen, if(!pipe) goto error1; - if(!trace_dump_enabled()) + if(!trace_enabled()) goto error1; tr_scr = trace_screen(_screen); @@ -1052,6 +1229,14 @@ trace_context_create(struct pipe_screen *_screen, if(!tr_ctx) goto error1; + tr_ctx->draw_blocker = debug_get_flags_option("RBUG_BLOCK", + rbug_blocker_flags, + 0); + pipe_mutex_init(tr_ctx->draw_mutex); + pipe_condvar_init(tr_ctx->draw_cond); + pipe_mutex_init(tr_ctx->list_mutex); + make_empty_list(&tr_ctx->shaders); + tr_ctx->base.winsys = _screen->winsys; tr_ctx->base.screen = _screen; tr_ctx->base.destroy = trace_context_destroy; @@ -1096,6 +1281,8 @@ trace_context_create(struct pipe_screen *_screen, tr_ctx->base.surface_fill = trace_context_surface_fill; tr_ctx->base.clear = trace_context_clear; tr_ctx->base.flush = trace_context_flush; + tr_ctx->base.is_texture_referenced = trace_is_texture_referenced; + tr_ctx->base.is_buffer_referenced = trace_is_buffer_referenced; tr_ctx->pipe = pipe; @@ -1104,6 +1291,8 @@ trace_context_create(struct pipe_screen *_screen, trace_dump_ret(ptr, pipe); trace_dump_call_end(); + trace_screen_add_to_list(tr_scr, contexts, tr_ctx); + return &tr_ctx->base; error1: diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index d02b22a069..6febe4b411 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -33,6 +33,7 @@ #include "util/u_debug.h" #include "pipe/p_context.h" +#include "tr_screen.h" #ifdef __cplusplus extern "C" { @@ -44,6 +45,42 @@ struct trace_context struct pipe_context base; struct pipe_context *pipe; + + /* current state */ + struct { + struct trace_shader *fs; + struct trace_shader *vs; + + struct trace_texture *tex[PIPE_MAX_SAMPLERS]; + unsigned num_texs; + + unsigned nr_cbufs; + struct trace_texture *cbufs[PIPE_MAX_COLOR_BUFS]; + struct trace_texture *zsbuf; + } curr; + + struct { + struct trace_shader *fs; + struct trace_shader *vs; + + struct trace_texture *tex; + struct trace_texture *surf; + + int blocker; + } draw_rule; + unsigned draw_num_rules; + pipe_condvar draw_cond; + pipe_mutex draw_mutex; + int draw_blocker; + int draw_blocked; + + /* for list on screen */ + struct tr_list list; + + /* list of state objects */ + pipe_mutex list_mutex; + unsigned num_shaders; + struct tr_list shaders; }; @@ -60,6 +97,9 @@ struct pipe_context * trace_context_create(struct pipe_screen *screen, struct pipe_context *pipe); +void +trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx); + #ifdef __cplusplus } diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 2618883e70..3a1409e95a 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -61,6 +61,8 @@ static struct util_stream *stream = NULL; static unsigned refcount = 0; static pipe_mutex call_mutex; static long unsigned call_no = 0; +static boolean dumping = FALSE; +static boolean initialized = FALSE; static INLINE void @@ -226,10 +228,22 @@ trace_dump_trace_close(void) } } +void trace_dump_init() +{ + if (initialized) + return; + + pipe_mutex_init(call_mutex); + dumping = FALSE; + initialized = TRUE; +} + boolean trace_dump_trace_begin() { const char *filename; + assert(initialized); + filename = debug_get_option("GALLIUM_TRACE", NULL); if(!filename) return FALSE; @@ -240,8 +254,6 @@ boolean trace_dump_trace_begin() if(!stream) return FALSE; - pipe_mutex_init(call_mutex); - trace_dump_writes("<?xml version='1.0' encoding='UTF-8'?>\n"); trace_dump_writes("<?xml-stylesheet type='text/xsl' href='trace.xsl'?>\n"); trace_dump_writes("<trace version='0.1'>\n"); @@ -258,7 +270,7 @@ boolean trace_dump_trace_begin() return TRUE; } -boolean trace_dump_enabled(void) +boolean trace_dump_trace_enabled(void) { return stream ? TRUE : FALSE; } @@ -270,9 +282,71 @@ void trace_dump_trace_end(void) trace_dump_trace_close(); } -void trace_dump_call_begin(const char *klass, const char *method) +/* + * Call lock + */ + +void trace_dump_call_lock(void) { pipe_mutex_lock(call_mutex); +} + +void trace_dump_call_unlock(void) +{ + pipe_mutex_unlock(call_mutex); +} + +/* + * Dumping control + */ + +void trace_dumping_start_locked(void) +{ + dumping = TRUE; +} + +void trace_dumping_stop_locked(void) +{ + dumping = FALSE; +} + +boolean trace_dumping_enabled_locked(void) +{ + return dumping; +} + +void trace_dumping_start(void) +{ + pipe_mutex_lock(call_mutex); + trace_dumping_start_locked(); + pipe_mutex_unlock(call_mutex); +} + +void trace_dumping_stop(void) +{ + pipe_mutex_lock(call_mutex); + trace_dumping_stop_locked(); + pipe_mutex_unlock(call_mutex); +} + +boolean trace_dumping_enabled(void) +{ + boolean ret; + pipe_mutex_lock(call_mutex); + ret = trace_dumping_enabled_locked(); + pipe_mutex_unlock(call_mutex); + return ret; +} + +/* + * Dump functions + */ + +void trace_dump_call_begin_locked(const char *klass, const char *method) +{ + if (!dumping) + return; + ++call_no; trace_dump_indent(1); trace_dump_writes("<call no=\'"); @@ -285,56 +359,94 @@ void trace_dump_call_begin(const char *klass, const char *method) trace_dump_newline(); } -void trace_dump_call_end(void) +void trace_dump_call_end_locked(void) { + if (!dumping) + return; + trace_dump_indent(1); trace_dump_tag_end("call"); trace_dump_newline(); util_stream_flush(stream); +} + +void trace_dump_call_begin(const char *klass, const char *method) +{ + pipe_mutex_lock(call_mutex); + trace_dump_call_begin_locked(klass, method); +} + +void trace_dump_call_end(void) +{ + trace_dump_call_end_locked(); pipe_mutex_unlock(call_mutex); } void trace_dump_arg_begin(const char *name) { + if (!dumping) + return; + trace_dump_indent(2); trace_dump_tag_begin1("arg", "name", name); } void trace_dump_arg_end(void) { + if (!dumping) + return; + trace_dump_tag_end("arg"); trace_dump_newline(); } void trace_dump_ret_begin(void) { + if (!dumping) + return; + trace_dump_indent(2); trace_dump_tag_begin("ret"); } void trace_dump_ret_end(void) { + if (!dumping) + return; + trace_dump_tag_end("ret"); trace_dump_newline(); } void trace_dump_bool(int value) { + if (!dumping) + return; + trace_dump_writef("<bool>%c</bool>", value ? '1' : '0'); } void trace_dump_int(long long int value) { + if (!dumping) + return; + trace_dump_writef("<int>%lli</int>", value); } void trace_dump_uint(long long unsigned value) { + if (!dumping) + return; + trace_dump_writef("<uint>%llu</uint>", value); } void trace_dump_float(double value) { + if (!dumping) + return; + trace_dump_writef("<float>%g</float>", value); } @@ -344,6 +456,10 @@ void trace_dump_bytes(const void *data, static const char hex_table[16] = "0123456789ABCDEF"; const uint8_t *p = data; long unsigned i; + + if (!dumping) + return; + trace_dump_writes("<bytes>"); for(i = 0; i < size; ++i) { uint8_t byte = *p++; @@ -357,6 +473,9 @@ void trace_dump_bytes(const void *data, void trace_dump_string(const char *str) { + if (!dumping) + return; + trace_dump_writes("<string>"); trace_dump_escape(str); trace_dump_writes("</string>"); @@ -364,6 +483,9 @@ void trace_dump_string(const char *str) void trace_dump_enum(const char *value) { + if (!dumping) + return; + trace_dump_writes("<enum>"); trace_dump_escape(value); trace_dump_writes("</enum>"); @@ -371,51 +493,81 @@ void trace_dump_enum(const char *value) void trace_dump_array_begin(void) { + if (!dumping) + return; + trace_dump_writes("<array>"); } void trace_dump_array_end(void) { + if (!dumping) + return; + trace_dump_writes("</array>"); } void trace_dump_elem_begin(void) { + if (!dumping) + return; + trace_dump_writes("<elem>"); } void trace_dump_elem_end(void) { + if (!dumping) + return; + trace_dump_writes("</elem>"); } void trace_dump_struct_begin(const char *name) { + if (!dumping) + return; + trace_dump_writef("<struct name='%s'>", name); } void trace_dump_struct_end(void) { + if (!dumping) + return; + trace_dump_writes("</struct>"); } void trace_dump_member_begin(const char *name) { + if (!dumping) + return; + trace_dump_writef("<member name='%s'>", name); } void trace_dump_member_end(void) { + if (!dumping) + return; + trace_dump_writes("</member>"); } void trace_dump_null(void) { + if (!dumping) + return; + trace_dump_writes("<null/>"); } void trace_dump_ptr(const void *value) { + if (!dumping) + return; + if(value) trace_dump_writef("<ptr>0x%08lx</ptr>", (unsigned long)(uintptr_t)value); else @@ -424,9 +576,11 @@ void trace_dump_ptr(const void *value) void trace_dump_buffer_ptr(struct pipe_buffer *_buffer) { + if (!dumping) + return; + if (_buffer) { - struct trace_screen *tr_scr = trace_screen(_buffer->screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); trace_dump_ptr(tr_buf->buffer); } else { trace_dump_null(); @@ -435,6 +589,9 @@ void trace_dump_buffer_ptr(struct pipe_buffer *_buffer) void trace_dump_texture_ptr(struct pipe_texture *_texture) { + if (!dumping) + return; + if (_texture) { struct trace_texture *tr_tex = trace_texture(_texture); trace_dump_ptr(tr_tex->texture); @@ -445,6 +602,9 @@ void trace_dump_texture_ptr(struct pipe_texture *_texture) void trace_dump_surface_ptr(struct pipe_surface *_surface) { + if (!dumping) + return; + if (_surface) { struct trace_surface *tr_surf = trace_surface(_surface); trace_dump_ptr(tr_surf->surface); @@ -455,6 +615,9 @@ void trace_dump_surface_ptr(struct pipe_surface *_surface) void trace_dump_transfer_ptr(struct pipe_transfer *_transfer) { + if (!dumping) + return; + if (_transfer) { struct trace_transfer *tr_tran = trace_transfer(_transfer); trace_dump_ptr(tr_tran->transfer); diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 26409f26c6..31ac70802f 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -42,11 +42,47 @@ struct pipe_texture; struct pipe_surface; struct pipe_transfer; +/* + * Call before use. + */ +void trace_dump_init(void); + +/* + * Low level dumping controls. + * + * Opening the trace file and checking if that is opened. + */ boolean trace_dump_trace_begin(void); -boolean trace_dump_enabled(void); +boolean trace_dump_trace_enabled(void); void trace_dump_trace_end(void); + +/* + * Lock and unlock the call mutex. + * + * It used by the none locked version of dumping control + * and begin/end call dump functions. + * + * Begin takes the lock while end unlocks it. Use the _locked + * version to avoid locking/unlocking it. + */ +void trace_dump_call_lock(void); +void trace_dump_call_unlock(void); + +/* + * High level dumping control. + */ +void trace_dumping_start_locked(void); +void trace_dumping_stop_locked(void); +boolean trace_dumping_enabled_locked(void); +void trace_dumping_start(void); +void trace_dumping_stop(void); +boolean trace_dumping_enabled(void); + +void trace_dump_call_begin_locked(const char *klass, const char *method); +void trace_dump_call_end_locked(void); void trace_dump_call_begin(const char *klass, const char *method); void trace_dump_call_end(void); + void trace_dump_arg_begin(const char *name); void trace_dump_arg_end(void); void trace_dump_ret_begin(void); diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c new file mode 100644 index 0000000000..23a2473b57 --- /dev/null +++ b/src/gallium/drivers/trace/tr_dump_state.c @@ -0,0 +1,490 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "util/u_memory.h" +#include "tgsi/tgsi_dump.h" + +#include "tr_dump.h" +#include "tr_dump_state.h" + + +void trace_dump_format(enum pipe_format format) +{ + trace_dump_enum(pf_name(format) ); +} + + +void trace_dump_block(const struct pipe_format_block *block) +{ + trace_dump_struct_begin("pipe_format_block"); + trace_dump_member(uint, block, size); + trace_dump_member(uint, block, width); + trace_dump_member(uint, block, height); + trace_dump_struct_end(); +} + + +static void trace_dump_reference(const struct pipe_reference *reference) +{ + trace_dump_struct_begin("pipe_reference"); + trace_dump_member(int, &reference->count, count); + trace_dump_struct_end(); +} + + +void trace_dump_template(const struct pipe_texture *templat) +{ + if(!templat) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_texture"); + + trace_dump_member(int, templat, target); + trace_dump_member(format, templat, format); + + trace_dump_member_begin("width"); + trace_dump_array(uint, templat->width, 1); + trace_dump_member_end(); + + trace_dump_member_begin("height"); + trace_dump_array(uint, templat->height, 1); + trace_dump_member_end(); + + trace_dump_member_begin("depth"); + trace_dump_array(uint, templat->depth, 1); + trace_dump_member_end(); + + trace_dump_member_begin("block"); + trace_dump_block(&templat->block); + trace_dump_member_end(); + + trace_dump_member(uint, templat, last_level); + trace_dump_member(uint, templat, tex_usage); + + trace_dump_struct_end(); +} + + +void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_rasterizer_state"); + + trace_dump_member(bool, state, flatshade); + trace_dump_member(bool, state, light_twoside); + trace_dump_member(uint, state, front_winding); + trace_dump_member(uint, state, cull_mode); + trace_dump_member(uint, state, fill_cw); + trace_dump_member(uint, state, fill_ccw); + trace_dump_member(bool, state, offset_cw); + trace_dump_member(bool, state, offset_ccw); + trace_dump_member(bool, state, scissor); + trace_dump_member(bool, state, poly_smooth); + trace_dump_member(bool, state, poly_stipple_enable); + trace_dump_member(bool, state, point_smooth); + trace_dump_member(bool, state, point_sprite); + trace_dump_member(bool, state, point_size_per_vertex); + trace_dump_member(bool, state, multisample); + trace_dump_member(bool, state, line_smooth); + trace_dump_member(bool, state, line_stipple_enable); + trace_dump_member(uint, state, line_stipple_factor); + trace_dump_member(uint, state, line_stipple_pattern); + trace_dump_member(bool, state, line_last_pixel); + trace_dump_member(bool, state, bypass_vs_clip_and_viewport); + trace_dump_member(bool, state, flatshade_first); + trace_dump_member(bool, state, gl_rasterization_rules); + + trace_dump_member(float, state, line_width); + trace_dump_member(float, state, point_size); + trace_dump_member(float, state, point_size_min); + trace_dump_member(float, state, point_size_max); + trace_dump_member(float, state, offset_units); + trace_dump_member(float, state, offset_scale); + + trace_dump_member_array(uint, state, sprite_coord_mode); + + trace_dump_struct_end(); +} + + +void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_poly_stipple"); + + trace_dump_member_begin("stipple"); + trace_dump_array(uint, + state->stipple, + Elements(state->stipple)); + trace_dump_member_end(); + + trace_dump_struct_end(); +} + + +void trace_dump_viewport_state(const struct pipe_viewport_state *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_viewport_state"); + + trace_dump_member_array(float, state, scale); + trace_dump_member_array(float, state, translate); + + trace_dump_struct_end(); +} + + +void trace_dump_scissor_state(const struct pipe_scissor_state *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_scissor_state"); + + trace_dump_member(uint, state, minx); + trace_dump_member(uint, state, miny); + trace_dump_member(uint, state, maxx); + trace_dump_member(uint, state, maxy); + + trace_dump_struct_end(); +} + + +void trace_dump_clip_state(const struct pipe_clip_state *state) +{ + unsigned i; + + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_clip_state"); + + trace_dump_member_begin("ucp"); + trace_dump_array_begin(); + for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) { + trace_dump_elem_begin(); + trace_dump_array(float, state->ucp[i], 4); + trace_dump_elem_end(); + } + trace_dump_array_end(); + trace_dump_member_end(); + + trace_dump_member(uint, state, nr); + + trace_dump_struct_end(); +} + + +void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_constant_buffer"); + + trace_dump_member(buffer_ptr, state, buffer); + + trace_dump_struct_end(); +} + + +void trace_dump_shader_state(const struct pipe_shader_state *state) +{ + static char str[8192]; + + if(!state) { + trace_dump_null(); + return; + } + + tgsi_dump_str(state->tokens, 0, str, sizeof(str)); + + trace_dump_struct_begin("pipe_shader_state"); + + trace_dump_member_begin("tokens"); + trace_dump_string(str); + trace_dump_member_end(); + + trace_dump_struct_end(); +} + + +void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state) +{ + unsigned i; + + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_depth_stencil_alpha_state"); + + trace_dump_member_begin("depth"); + trace_dump_struct_begin("pipe_depth_state"); + trace_dump_member(bool, &state->depth, enabled); + trace_dump_member(bool, &state->depth, writemask); + trace_dump_member(uint, &state->depth, func); + trace_dump_struct_end(); + trace_dump_member_end(); + + trace_dump_member_begin("stencil"); + trace_dump_array_begin(); + for(i = 0; i < Elements(state->stencil); ++i) { + trace_dump_elem_begin(); + trace_dump_struct_begin("pipe_stencil_state"); + trace_dump_member(bool, &state->stencil[i], enabled); + trace_dump_member(uint, &state->stencil[i], func); + trace_dump_member(uint, &state->stencil[i], fail_op); + trace_dump_member(uint, &state->stencil[i], zpass_op); + trace_dump_member(uint, &state->stencil[i], zfail_op); + trace_dump_member(uint, &state->stencil[i], ref_value); + trace_dump_member(uint, &state->stencil[i], valuemask); + trace_dump_member(uint, &state->stencil[i], writemask); + trace_dump_struct_end(); + trace_dump_elem_end(); + } + trace_dump_array_end(); + trace_dump_member_end(); + + trace_dump_member_begin("alpha"); + trace_dump_struct_begin("pipe_alpha_state"); + trace_dump_member(bool, &state->alpha, enabled); + trace_dump_member(uint, &state->alpha, func); + trace_dump_member(float, &state->alpha, ref_value); + trace_dump_struct_end(); + trace_dump_member_end(); + + trace_dump_struct_end(); +} + + +void trace_dump_blend_state(const struct pipe_blend_state *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_blend_state"); + + trace_dump_member(bool, state, blend_enable); + + trace_dump_member(uint, state, rgb_func); + trace_dump_member(uint, state, rgb_src_factor); + trace_dump_member(uint, state, rgb_dst_factor); + + trace_dump_member(uint, state, alpha_func); + trace_dump_member(uint, state, alpha_src_factor); + trace_dump_member(uint, state, alpha_dst_factor); + + trace_dump_member(bool, state, logicop_enable); + trace_dump_member(uint, state, logicop_func); + + trace_dump_member(uint, state, colormask); + trace_dump_member(bool, state, dither); + + trace_dump_struct_end(); +} + + +void trace_dump_blend_color(const struct pipe_blend_color *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_blend_color"); + + trace_dump_member_array(float, state, color); + + trace_dump_struct_end(); +} + + +void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) +{ + trace_dump_struct_begin("pipe_framebuffer_state"); + + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); + trace_dump_member(uint, state, nr_cbufs); + trace_dump_member_array(ptr, state, cbufs); + trace_dump_member(ptr, state, zsbuf); + + trace_dump_struct_end(); +} + + +void trace_dump_sampler_state(const struct pipe_sampler_state *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_sampler_state"); + + trace_dump_member(uint, state, wrap_s); + trace_dump_member(uint, state, wrap_t); + trace_dump_member(uint, state, wrap_r); + trace_dump_member(uint, state, min_img_filter); + trace_dump_member(uint, state, min_mip_filter); + trace_dump_member(uint, state, mag_img_filter); + trace_dump_member(bool, state, compare_mode); + trace_dump_member(uint, state, compare_func); + trace_dump_member(bool, state, normalized_coords); + trace_dump_member(uint, state, prefilter); + trace_dump_member(float, state, shadow_ambient); + trace_dump_member(float, state, lod_bias); + trace_dump_member(float, state, min_lod); + trace_dump_member(float, state, max_lod); + trace_dump_member_array(float, state, border_color); + trace_dump_member(float, state, max_anisotropy); + + trace_dump_struct_end(); +} + + +void trace_dump_surface(const struct pipe_surface *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_surface"); + + trace_dump_reference(&state->reference); + + trace_dump_member(format, state, format); + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); + + trace_dump_member(uint, state, layout); + trace_dump_member(uint, state, offset); + trace_dump_member(uint, state, usage); + + trace_dump_member(ptr, state, texture); + trace_dump_member(uint, state, face); + trace_dump_member(uint, state, level); + trace_dump_member(uint, state, zslice); + + trace_dump_struct_end(); +} + + +void trace_dump_transfer(const struct pipe_transfer *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_transfer"); + + trace_dump_member(format, state, format); + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); + + trace_dump_member_begin("block"); + trace_dump_block(&state->block); + trace_dump_member_end(); + + trace_dump_member(uint, state, nblocksx); + trace_dump_member(uint, state, nblocksy); + trace_dump_member(uint, state, stride); + trace_dump_member(uint, state, usage); + + trace_dump_member(ptr, state, texture); + trace_dump_member(uint, state, face); + trace_dump_member(uint, state, level); + trace_dump_member(uint, state, zslice); + + trace_dump_struct_end(); +} + + +void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_vertex_buffer"); + + trace_dump_member(uint, state, stride); + trace_dump_member(uint, state, max_index); + trace_dump_member(uint, state, buffer_offset); + trace_dump_member(buffer_ptr, state, buffer); + + trace_dump_struct_end(); +} + + +void trace_dump_vertex_element(const struct pipe_vertex_element *state) +{ + if(!state) { + trace_dump_null(); + return; + } + + trace_dump_struct_begin("pipe_vertex_element"); + + trace_dump_member(uint, state, src_offset); + + trace_dump_member(uint, state, vertex_buffer_index); + trace_dump_member(uint, state, nr_components); + + trace_dump_member(format, state, src_format); + + trace_dump_struct_end(); +} diff --git a/src/gallium/drivers/trace/tr_dump_state.h b/src/gallium/drivers/trace/tr_dump_state.h new file mode 100644 index 0000000000..05b821adb6 --- /dev/null +++ b/src/gallium/drivers/trace/tr_dump_state.h @@ -0,0 +1,78 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_DUMP_STATE_H_ +#define TR_DUMP_STATE_H_ + +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" + + +void trace_dump_format(enum pipe_format format); + +void trace_dump_block(const struct pipe_format_block *block); + +void trace_dump_template(const struct pipe_texture *templat); + + +void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state); + +void trace_dump_poly_stipple(const struct pipe_poly_stipple *state); + +void trace_dump_viewport_state(const struct pipe_viewport_state *state); + +void trace_dump_scissor_state(const struct pipe_scissor_state *state); + +void trace_dump_clip_state(const struct pipe_clip_state *state); + +void trace_dump_constant_buffer(const struct pipe_constant_buffer *state); + +void trace_dump_token(const struct tgsi_token *token); + +void trace_dump_shader_state(const struct pipe_shader_state *state); + +void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state); + +void trace_dump_blend_state(const struct pipe_blend_state *state); + +void trace_dump_blend_color(const struct pipe_blend_color *state); + +void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state); + +void trace_dump_sampler_state(const struct pipe_sampler_state *state); + +void trace_dump_surface(const struct pipe_surface *state); + +void trace_dump_transfer(const struct pipe_transfer *state); + +void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state); + +void trace_dump_vertex_element(const struct pipe_vertex_element *state); + + +#endif /* TR_STATE_H */ diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c new file mode 100644 index 0000000000..e85ac15edc --- /dev/null +++ b/src/gallium/drivers/trace/tr_rbug.c @@ -0,0 +1,864 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_simple_list.h" +#include "util/u_network.h" + +#include "tgsi/tgsi_parse.h" + +#include "tr_dump.h" +#include "tr_state.h" +#include "tr_buffer.h" +#include "tr_texture.h" + +#include "rbug/rbug.h" + +#include <errno.h> + +#if defined(PIPE_SUBSYSTEM_WINDOWS_USER) +# define sleep Sleep +#elif defined(PIPE_OS_LINUX) +void usleep(int); +# define sleep usleep +#else +# warning "No socket implementation" +#endif + +#define U642VOID(x) ((void *)(unsigned long)(x)) +#define VOID2U64(x) ((uint64_t)(unsigned long)(x)) + +struct trace_rbug +{ + struct trace_screen *tr_scr; + struct rbug_connection *con; + pipe_thread thread; + boolean running; +}; + +PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug); + + +/********************************************************** + * Helper functions + */ + + +static struct trace_context * +trace_rbug_get_context_locked(struct trace_screen *tr_scr, rbug_context_t ctx) +{ + struct trace_context *tr_ctx = NULL; + struct tr_list *ptr; + + foreach(ptr, &tr_scr->contexts) { + tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list)); + if (ctx == VOID2U64(tr_ctx)) + break; + tr_ctx = NULL; + } + + return tr_ctx; +} + +static struct trace_shader * +trace_rbug_get_shader_locked(struct trace_context *tr_ctx, rbug_shader_t shdr) +{ + struct trace_shader *tr_shdr = NULL; + struct tr_list *ptr; + + foreach(ptr, &tr_ctx->shaders) { + tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list)); + if (shdr == VOID2U64(tr_shdr)) + break; + tr_shdr = NULL; + } + + return tr_shdr; +} + +static void * +trace_shader_create_locked(struct pipe_context *pipe, + struct trace_shader *tr_shdr, + struct tgsi_token *tokens) +{ + void *state = NULL; + struct pipe_shader_state pss = { 0 }; + pss.tokens = tokens; + + if (tr_shdr->type == TRACE_SHADER_FRAGMENT) { + state = pipe->create_fs_state(pipe, &pss); + } else if (tr_shdr->type == TRACE_SHADER_VERTEX) { + state = pipe->create_vs_state(pipe, &pss); + } else + assert(0); + + return state; +} + +static void +trace_shader_bind_locked(struct pipe_context *pipe, + struct trace_shader *tr_shdr, + void *state) +{ + if (tr_shdr->type == TRACE_SHADER_FRAGMENT) { + pipe->bind_fs_state(pipe, state); + } else if (tr_shdr->type == TRACE_SHADER_VERTEX) { + pipe->bind_vs_state(pipe, state); + } else + assert(0); +} + +static void +trace_shader_delete_locked(struct pipe_context *pipe, + struct trace_shader *tr_shdr, + void *state) +{ + if (tr_shdr->type == TRACE_SHADER_FRAGMENT) { + pipe->delete_fs_state(pipe, state); + } else if (tr_shdr->type == TRACE_SHADER_VERTEX) { + pipe->delete_vs_state(pipe, state); + } else + assert(0); +} + +/************************************************ + * Request handler functions + */ + + +static int +trace_rbug_texture_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_texture *tr_tex = NULL; + struct tr_list *ptr; + rbug_texture_t *texs; + int i = 0; + + pipe_mutex_lock(tr_scr->list_mutex); + texs = MALLOC(tr_scr->num_textures * sizeof(rbug_texture_t)); + foreach(ptr, &tr_scr->textures) { + tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list)); + texs[i++] = VOID2U64(tr_tex); + } + pipe_mutex_unlock(tr_scr->list_mutex); + + rbug_send_texture_list_reply(tr_rbug->con, serial, texs, i, NULL); + FREE(texs); + + return 0; +} + +static int +trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_texture *tr_tex; + struct rbug_proto_texture_info *gpti = (struct rbug_proto_texture_info *)header; + struct tr_list *ptr; + struct pipe_texture *t; + + pipe_mutex_lock(tr_scr->list_mutex); + foreach(ptr, &tr_scr->textures) { + tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list)); + if (gpti->texture == VOID2U64(tr_tex)) + break; + tr_tex = NULL; + } + + if (!tr_tex) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + t = tr_tex->texture; + rbug_send_texture_info_reply(tr_rbug->con, serial, + t->target, t->format, + t->width, t->last_level + 1, + t->height, t->last_level + 1, + t->depth, t->last_level + 1, + t->block.width, t->block.height, t->block.size, + t->last_level, + t->nr_samples, + t->tex_usage, + NULL); + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_texture_read *gptr = (struct rbug_proto_texture_read *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_texture *tr_tex; + struct tr_list *ptr; + + struct pipe_screen *screen = tr_scr->screen; + struct pipe_texture *tex; + struct pipe_transfer *t; + + void *map; + + pipe_mutex_lock(tr_scr->list_mutex); + foreach(ptr, &tr_scr->textures) { + tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list)); + if (gptr->texture == VOID2U64(tr_tex)) + break; + tr_tex = NULL; + } + + if (!tr_tex) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + tex = tr_tex->texture; + t = screen->get_tex_transfer(tr_scr->screen, tex, + gptr->face, gptr->level, gptr->zslice, + PIPE_TRANSFER_READ, + gptr->x, gptr->y, gptr->w, gptr->h); + + map = screen->transfer_map(screen, t); + + rbug_send_texture_read_reply(tr_rbug->con, serial, + t->format, + t->block.width, t->block.height, t->block.size, + (uint8_t*)map, t->stride * t->nblocksy, + t->stride, + NULL); + + screen->transfer_unmap(screen, t); + screen->tex_transfer_destroy(t); + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct tr_list *ptr; + struct trace_context *tr_ctx = NULL; + rbug_context_t *ctxs; + int i = 0; + + pipe_mutex_lock(tr_scr->list_mutex); + ctxs = MALLOC(tr_scr->num_contexts * sizeof(rbug_context_t)); + foreach(ptr, &tr_scr->contexts) { + tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list)); + ctxs[i++] = VOID2U64(tr_ctx); + } + pipe_mutex_unlock(tr_scr->list_mutex); + + rbug_send_context_list_reply(tr_rbug->con, serial, ctxs, i, NULL); + FREE(ctxs); + + return 0; +} + +static int +trace_rbug_context_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_info *info = (struct rbug_proto_context_info *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + rbug_texture_t cbufs[PIPE_MAX_COLOR_BUFS]; + rbug_texture_t texs[PIPE_MAX_SAMPLERS]; + int i; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + /* protect the pipe context */ + pipe_mutex_lock(tr_ctx->draw_mutex); + trace_dump_call_lock(); + + for (i = 0; i < tr_ctx->curr.nr_cbufs; i++) + cbufs[i] = VOID2U64(tr_ctx->curr.cbufs[i]); + + for (i = 0; i < tr_ctx->curr.num_texs; i++) + texs[i] = VOID2U64(tr_ctx->curr.tex[i]); + + rbug_send_context_info_reply(tr_rbug->con, serial, + VOID2U64(tr_ctx->curr.vs), VOID2U64(tr_ctx->curr.fs), + texs, tr_ctx->curr.num_texs, + cbufs, tr_ctx->curr.nr_cbufs, + VOID2U64(tr_ctx->curr.zsbuf), + tr_ctx->draw_blocker, tr_ctx->draw_blocked, NULL); + + trace_dump_call_unlock(); + pipe_mutex_unlock(tr_ctx->draw_mutex); + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_draw_block(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_draw_block *block = (struct rbug_proto_context_draw_block *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, block->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->draw_mutex); + tr_ctx->draw_blocker |= block->block; + pipe_mutex_unlock(tr_ctx->draw_mutex); + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_draw_step(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_draw_step *step = (struct rbug_proto_context_draw_step *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, step->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->draw_mutex); + if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) { + if (step->step & RBUG_BLOCK_RULE) + tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK; + } else { + tr_ctx->draw_blocked &= ~step->step; + } + pipe_mutex_unlock(tr_ctx->draw_mutex); + +#ifdef PIPE_THREAD_HAVE_CONDVAR + pipe_condvar_broadcast(tr_ctx->draw_cond); +#endif + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_draw_unblock(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_draw_unblock *unblock = (struct rbug_proto_context_draw_unblock *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, unblock->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->draw_mutex); + if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) { + if (unblock->unblock & RBUG_BLOCK_RULE) + tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK; + } else { + tr_ctx->draw_blocked &= ~unblock->unblock; + } + tr_ctx->draw_blocker &= ~unblock->unblock; + pipe_mutex_unlock(tr_ctx->draw_mutex); + +#ifdef PIPE_THREAD_HAVE_CONDVAR + pipe_condvar_broadcast(tr_ctx->draw_cond); +#endif + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_draw_rule(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_draw_rule *rule = (struct rbug_proto_context_draw_rule *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, rule->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->draw_mutex); + tr_ctx->draw_rule.vs = U642VOID(rule->vertex); + tr_ctx->draw_rule.fs = U642VOID(rule->fragment); + tr_ctx->draw_rule.tex = U642VOID(rule->texture); + tr_ctx->draw_rule.surf = U642VOID(rule->surface); + tr_ctx->draw_rule.blocker = rule->block; + tr_ctx->draw_blocker |= RBUG_BLOCK_RULE; + pipe_mutex_unlock(tr_ctx->draw_mutex); + +#ifdef PIPE_THREAD_HAVE_CONDVAR + pipe_condvar_broadcast(tr_ctx->draw_cond); +#endif + + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_context_flush(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_context_flush *flush = (struct rbug_proto_context_flush *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, flush->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + /* protect the pipe context */ + trace_dump_call_lock(); + + tr_ctx->pipe->flush(tr_ctx->pipe, flush->flags, NULL); + + trace_dump_call_unlock(); + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_shader_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_shader_list *list = (struct rbug_proto_shader_list *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + struct trace_shader *tr_shdr = NULL; + struct tr_list *ptr; + rbug_shader_t *shdrs; + int i = 0; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, list->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->list_mutex); + shdrs = MALLOC(tr_ctx->num_shaders * sizeof(rbug_shader_t)); + foreach(ptr, &tr_ctx->shaders) { + tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list)); + shdrs[i++] = VOID2U64(tr_shdr); + } + + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + + rbug_send_shader_list_reply(tr_rbug->con, serial, shdrs, i, NULL); + FREE(shdrs); + + return 0; +} + +static int +trace_rbug_shader_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + struct rbug_proto_shader_info *info = (struct rbug_proto_shader_info *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + struct trace_shader *tr_shdr = NULL; + unsigned original_len; + unsigned replaced_len; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->list_mutex); + + tr_shdr = trace_rbug_get_shader_locked(tr_ctx, info->shader); + + if (!tr_shdr) { + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + /* just in case */ + assert(sizeof(struct tgsi_token) == 4); + + original_len = tgsi_num_tokens(tr_shdr->tokens); + if (tr_shdr->replaced_tokens) + replaced_len = tgsi_num_tokens(tr_shdr->replaced_tokens); + else + replaced_len = 0; + + rbug_send_shader_info_reply(tr_rbug->con, serial, + (uint32_t*)tr_shdr->tokens, original_len, + (uint32_t*)tr_shdr->replaced_tokens, replaced_len, + tr_shdr->disabled, + NULL); + + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_shader_disable(struct trace_rbug *tr_rbug, struct rbug_header *header) +{ + struct rbug_proto_shader_disable *dis = (struct rbug_proto_shader_disable *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + struct trace_shader *tr_shdr = NULL; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, dis->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->list_mutex); + + tr_shdr = trace_rbug_get_shader_locked(tr_ctx, dis->shader); + + if (!tr_shdr) { + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + tr_shdr->disabled = dis->disable; + + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; +} + +static int +trace_rbug_shader_replace(struct trace_rbug *tr_rbug, struct rbug_header *header) +{ + struct rbug_proto_shader_replace *rep = (struct rbug_proto_shader_replace *)header; + + struct trace_screen *tr_scr = tr_rbug->tr_scr; + struct trace_context *tr_ctx = NULL; + struct trace_shader *tr_shdr = NULL; + struct pipe_context *pipe = NULL; + void *state; + + pipe_mutex_lock(tr_scr->list_mutex); + tr_ctx = trace_rbug_get_context_locked(tr_scr, rep->context); + + if (!tr_ctx) { + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + pipe_mutex_lock(tr_ctx->list_mutex); + + tr_shdr = trace_rbug_get_shader_locked(tr_ctx, rep->shader); + + if (!tr_shdr) { + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + return -ESRCH; + } + + /* protect the pipe context */ + trace_dump_call_lock(); + + pipe = tr_ctx->pipe; + + /* remove old replaced shader */ + if (tr_shdr->replaced) { + if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr) + trace_shader_bind_locked(pipe, tr_shdr, tr_shdr->state); + + FREE(tr_shdr->replaced_tokens); + trace_shader_delete_locked(pipe, tr_shdr, tr_shdr->replaced); + tr_shdr->replaced = NULL; + tr_shdr->replaced_tokens = NULL; + } + + /* empty inputs means restore old which we did above */ + if (rep->tokens_len == 0) + goto out; + + tr_shdr->replaced_tokens = tgsi_dup_tokens((struct tgsi_token *)rep->tokens); + if (!tr_shdr->replaced_tokens) + goto err; + + state = trace_shader_create_locked(pipe, tr_shdr, tr_shdr->replaced_tokens); + if (!state) + goto err; + + /* bind new shader if the shader is currently a bound */ + if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr) + trace_shader_bind_locked(pipe, tr_shdr, state); + + /* save state */ + tr_shdr->replaced = state; + +out: + trace_dump_call_unlock(); + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + + return 0; + +err: + FREE(tr_shdr->replaced_tokens); + tr_shdr->replaced = NULL; + tr_shdr->replaced_tokens = NULL; + + trace_dump_call_unlock(); + pipe_mutex_unlock(tr_ctx->list_mutex); + pipe_mutex_unlock(tr_scr->list_mutex); + return -EINVAL; +} + +static boolean +trace_rbug_header(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial) +{ + int ret = 0; + + switch(header->opcode) { + case RBUG_OP_PING: + rbug_send_ping_reply(tr_rbug->con, serial, NULL); + break; + case RBUG_OP_TEXTURE_LIST: + ret = trace_rbug_texture_list(tr_rbug, header, serial); + break; + case RBUG_OP_TEXTURE_INFO: + ret = trace_rbug_texture_info(tr_rbug, header, serial); + break; + case RBUG_OP_TEXTURE_READ: + ret = trace_rbug_texture_read(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_LIST: + ret = trace_rbug_context_list(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_INFO: + ret = trace_rbug_context_info(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_DRAW_BLOCK: + ret = trace_rbug_context_draw_block(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_DRAW_STEP: + ret = trace_rbug_context_draw_step(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_DRAW_UNBLOCK: + ret = trace_rbug_context_draw_unblock(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_DRAW_RULE: + ret = trace_rbug_context_draw_rule(tr_rbug, header, serial); + break; + case RBUG_OP_CONTEXT_FLUSH: + ret = trace_rbug_context_flush(tr_rbug, header, serial); + break; + case RBUG_OP_SHADER_LIST: + ret = trace_rbug_shader_list(tr_rbug, header, serial); + break; + case RBUG_OP_SHADER_INFO: + ret = trace_rbug_shader_info(tr_rbug, header, serial); + break; + case RBUG_OP_SHADER_DISABLE: + ret = trace_rbug_shader_disable(tr_rbug, header); + break; + case RBUG_OP_SHADER_REPLACE: + ret = trace_rbug_shader_replace(tr_rbug, header); + break; + default: + debug_printf("%s - unsupported opcode %u\n", __FUNCTION__, header->opcode); + ret = -ENOSYS; + break; + } + rbug_free_header(header); + + if (ret) + rbug_send_error_reply(tr_rbug->con, serial, ret, NULL); + + return TRUE; +} + +static void +trace_rbug_con(struct trace_rbug *tr_rbug) +{ + struct rbug_header *header; + uint32_t serial; + + debug_printf("%s - connection received\n", __FUNCTION__); + + while(tr_rbug->running) { + header = rbug_get_message(tr_rbug->con, &serial); + if (!header) + break; + + if (!trace_rbug_header(tr_rbug, header, serial)) + break; + } + + debug_printf("%s - connection closed\n", __FUNCTION__); + + rbug_disconnect(tr_rbug->con); + tr_rbug->con = NULL; +} + +PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug) +{ + struct trace_rbug *tr_rbug = void_tr_rbug; + uint16_t port = 13370; + int s = -1; + int c; + + u_socket_init(); + + for (;port <= 13379 && s < 0; port++) + s = u_socket_listen_on_port(port); + + if (s < 0) { + debug_printf("trace_rbug - failed to listen\n"); + return NULL; + } + + u_socket_block(s, false); + + debug_printf("trace_rbug - remote debugging listening on port %u\n", --port); + + while(tr_rbug->running) { + sleep(1); + + c = u_socket_accept(s); + if (c < 0) + continue; + + u_socket_block(c, true); + tr_rbug->con = rbug_from_socket(c); + + trace_rbug_con(tr_rbug); + + u_socket_close(c); + } + + u_socket_close(s); + + u_socket_stop(); + + return NULL; +} + +/********************************************************** + * + */ + +struct trace_rbug * +trace_rbug_start(struct trace_screen *tr_scr) +{ + struct trace_rbug *tr_rbug = CALLOC_STRUCT(trace_rbug); + if (!tr_rbug) + return NULL; + + tr_rbug->tr_scr = tr_scr; + tr_rbug->running = TRUE; + tr_rbug->thread = pipe_thread_create(trace_rbug_thread, tr_rbug); + + return tr_rbug; +} + +void +trace_rbug_stop(struct trace_rbug *tr_rbug) +{ + if (!tr_rbug) + return; + + tr_rbug->running = false; + pipe_thread_wait(tr_rbug->thread); + + FREE(tr_rbug); + + return; +} + +void +trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx) +{ + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_rbug *tr_rbug = tr_scr->rbug; + + if (tr_rbug && tr_rbug->con) + rbug_send_context_draw_blocked(tr_rbug->con, + VOID2U64(tr_ctx), tr_ctx->draw_blocked, NULL); +} diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 6792505383..920f418ebf 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -26,16 +26,19 @@ **************************************************************************/ #include "util/u_memory.h" +#include "util/u_simple_list.h" #include "tr_buffer.h" #include "tr_dump.h" -#include "tr_state.h" +#include "tr_dump_state.h" #include "tr_texture.h" #include "tr_screen.h" #include "pipe/p_inlines.h" +static boolean trace = FALSE; + static const char * trace_screen_get_name(struct pipe_screen *_screen) { @@ -211,10 +214,12 @@ static struct pipe_texture * trace_screen_texture_blanket(struct pipe_screen *_screen, const struct pipe_texture *templat, const unsigned *ppitch, - struct pipe_buffer *buffer) + struct pipe_buffer *_buffer) { struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; + struct pipe_buffer *buffer = tr_buf->buffer; unsigned pitch = *ppitch; struct pipe_texture *result; @@ -603,7 +608,7 @@ trace_screen_buffer_map(struct pipe_screen *_screen, unsigned usage) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; struct pipe_buffer *buffer = tr_buf->buffer; void *map; @@ -628,7 +633,7 @@ trace_screen_buffer_map_range(struct pipe_screen *_screen, unsigned usage) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; struct pipe_buffer *buffer = tr_buf->buffer; void *map; @@ -680,7 +685,7 @@ trace_screen_buffer_flush_mapped_range(struct pipe_screen *_screen, unsigned length) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; struct pipe_buffer *buffer = tr_buf->buffer; @@ -696,7 +701,7 @@ trace_screen_buffer_unmap(struct pipe_screen *_screen, struct pipe_buffer *_buffer) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; struct pipe_buffer *buffer = tr_buf->buffer; @@ -712,7 +717,7 @@ static void trace_screen_buffer_destroy(struct pipe_buffer *_buffer) { struct trace_screen *tr_scr = trace_screen(_buffer->screen); - struct trace_buffer *tr_buf = trace_buffer(tr_scr, _buffer); + struct trace_buffer *tr_buf = trace_buffer(_buffer); struct pipe_screen *screen = tr_scr->screen; struct pipe_buffer *buffer = tr_buf->buffer; @@ -734,19 +739,23 @@ trace_screen_buffer_destroy(struct pipe_buffer *_buffer) static void trace_screen_fence_reference(struct pipe_screen *_screen, - struct pipe_fence_handle **dst, + struct pipe_fence_handle **pdst, struct pipe_fence_handle *src) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; + struct pipe_fence_handle *dst; + assert(pdst); + dst = *pdst; + trace_dump_call_begin("pipe_screen", "fence_reference"); trace_dump_arg(ptr, screen); trace_dump_arg(ptr, dst); trace_dump_arg(ptr, src); - screen->fence_reference(screen, dst, src); + screen->fence_reference(screen, pdst, src); trace_dump_call_end(); } @@ -755,7 +764,7 @@ trace_screen_fence_reference(struct pipe_screen *_screen, static int trace_screen_fence_signalled(struct pipe_screen *_screen, struct pipe_fence_handle *fence, - unsigned flag) + unsigned flags) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; @@ -765,9 +774,9 @@ trace_screen_fence_signalled(struct pipe_screen *_screen, trace_dump_arg(ptr, screen); trace_dump_arg(ptr, fence); - trace_dump_arg(uint, flag); + trace_dump_arg(uint, flags); - result = screen->fence_signalled(screen, fence, flag); + result = screen->fence_signalled(screen, fence, flags); trace_dump_ret(int, result); @@ -780,7 +789,7 @@ trace_screen_fence_signalled(struct pipe_screen *_screen, static int trace_screen_fence_finish(struct pipe_screen *_screen, struct pipe_fence_handle *fence, - unsigned flag) + unsigned flags) { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; @@ -790,9 +799,9 @@ trace_screen_fence_finish(struct pipe_screen *_screen, trace_dump_arg(ptr, screen); trace_dump_arg(ptr, fence); - trace_dump_arg(uint, flag); + trace_dump_arg(uint, flags); - result = screen->fence_finish(screen, fence, flag); + result = screen->fence_finish(screen, fence, flags); trace_dump_ret(int, result); @@ -813,29 +822,47 @@ trace_screen_destroy(struct pipe_screen *_screen) struct pipe_screen *screen = tr_scr->screen; trace_dump_call_begin("pipe_screen", "destroy"); - trace_dump_arg(ptr, screen); - - screen->destroy(screen); - trace_dump_call_end(); - trace_dump_trace_end(); + if (tr_scr->rbug) + trace_rbug_stop(tr_scr->rbug); + + screen->destroy(screen); + FREE(tr_scr); } +boolean +trace_enabled(void) +{ + return trace; +} struct pipe_screen * trace_screen_create(struct pipe_screen *screen) { struct trace_screen *tr_scr; struct pipe_winsys *winsys; + boolean rbug = FALSE; if(!screen) goto error1; - if(!trace_dump_trace_begin()) + trace_dump_init(); + + if(trace_dump_trace_begin()) { + trace_dumping_start(); + trace = TRUE; + } + + if (debug_get_bool_option("GALLIUM_RBUG", FALSE)) { + trace = TRUE; + rbug = TRUE; + } + + if (!trace) goto error1; trace_dump_call_begin("", "pipe_screen_create"); @@ -851,6 +878,12 @@ trace_screen_create(struct pipe_screen *screen) #else winsys = screen->winsys; #endif + pipe_mutex_init(tr_scr->list_mutex); + make_empty_list(&tr_scr->buffers); + make_empty_list(&tr_scr->contexts); + make_empty_list(&tr_scr->textures); + make_empty_list(&tr_scr->surfaces); + make_empty_list(&tr_scr->transfers); tr_scr->base.winsys = winsys; tr_scr->base.destroy = trace_screen_destroy; @@ -889,6 +922,9 @@ trace_screen_create(struct pipe_screen *screen) trace_dump_ret(ptr, screen); trace_dump_call_end(); + if (rbug) + tr_scr->rbug = trace_rbug_start(tr_scr); + return &tr_scr->base; #if 0 diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 8c65516b50..dba8cd7c65 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -30,6 +30,7 @@ #include "pipe/p_screen.h" +#include "pipe/p_thread.h" #ifdef __cplusplus @@ -37,6 +38,11 @@ extern "C" { #endif +struct tr_list { + struct tr_list *next; + struct tr_list *prev; +}; + /** * It often happens that new data is written directly to the user buffers * without mapping/unmapping. This flag marks user buffers, so that their @@ -50,21 +56,72 @@ struct trace_screen struct pipe_screen base; struct pipe_screen *screen; + + /* remote debugger */ + struct trace_rbug *rbug; + + pipe_mutex list_mutex; + int num_buffers; + int num_contexts; + int num_textures; + int num_surfaces; + int num_transfers; + struct tr_list buffers; + struct tr_list contexts; + struct tr_list textures; + struct tr_list surfaces; + struct tr_list transfers; }; +/* + * tr_rbug.c + */ + + +struct trace_rbug; + +struct trace_rbug * +trace_rbug_start(struct trace_screen *tr_scr); + +void +trace_rbug_stop(struct trace_rbug *tr_rbug); + + +/* + * tr_screen.c + */ + + +boolean +trace_enabled(void); + struct trace_screen * trace_screen(struct pipe_screen *screen); - struct pipe_screen * trace_screen_create(struct pipe_screen *screen); - void trace_screen_user_buffer_update(struct pipe_screen *screen, struct pipe_buffer *buffer); +#define trace_screen_add_to_list(tr_scr, name, obj) \ + do { \ + pipe_mutex_lock(tr_scr->list_mutex); \ + insert_at_head(&tr_scr->name, &obj->list); \ + tr_scr->num_##name++; \ + pipe_mutex_unlock(tr_scr->list_mutex); \ + } while (0) + +#define trace_screen_remove_from_list(tr_scr, name, obj) \ + do { \ + pipe_mutex_lock(tr_scr->list_mutex); \ + remove_from_list(&obj->list); \ + tr_scr->num_##name--; \ + pipe_mutex_unlock(tr_scr->list_mutex); \ + } while (0) + #ifdef __cplusplus } diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index a9570c1aeb..d8c11640bf 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -1,491 +1,66 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +/* + * Copyright 2009 VMware, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. * - **************************************************************************/ - + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ -#include "pipe/p_compiler.h" -#include "util/u_memory.h" -#include "tgsi/tgsi_dump.h" - -#include "tr_dump.h" #include "tr_state.h" +#include "util/u_memory.h" +#include "util/u_simple_list.h" -void trace_dump_format(enum pipe_format format) -{ - trace_dump_enum(pf_name(format) ); -} - - -void trace_dump_block(const struct pipe_format_block *block) -{ - trace_dump_struct_begin("pipe_format_block"); - trace_dump_member(uint, block, size); - trace_dump_member(uint, block, width); - trace_dump_member(uint, block, height); - trace_dump_struct_end(); -} - - -static void trace_dump_reference(const struct pipe_reference *reference) -{ - trace_dump_struct_begin("pipe_reference"); - trace_dump_member(int, &reference->count, count); - trace_dump_struct_end(); -} - - -void trace_dump_template(const struct pipe_texture *templat) -{ - if(!templat) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_texture"); - - trace_dump_member(int, templat, target); - trace_dump_member(format, templat, format); - - trace_dump_member_begin("width"); - trace_dump_array(uint, templat->width, 1); - trace_dump_member_end(); - - trace_dump_member_begin("height"); - trace_dump_array(uint, templat->height, 1); - trace_dump_member_end(); - - trace_dump_member_begin("depth"); - trace_dump_array(uint, templat->depth, 1); - trace_dump_member_end(); - - trace_dump_member_begin("block"); - trace_dump_block(&templat->block); - trace_dump_member_end(); - - trace_dump_member(uint, templat, last_level); - trace_dump_member(uint, templat, tex_usage); - - trace_dump_struct_end(); -} - - -void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_rasterizer_state"); - - trace_dump_member(bool, state, flatshade); - trace_dump_member(bool, state, light_twoside); - trace_dump_member(uint, state, front_winding); - trace_dump_member(uint, state, cull_mode); - trace_dump_member(uint, state, fill_cw); - trace_dump_member(uint, state, fill_ccw); - trace_dump_member(bool, state, offset_cw); - trace_dump_member(bool, state, offset_ccw); - trace_dump_member(bool, state, scissor); - trace_dump_member(bool, state, poly_smooth); - trace_dump_member(bool, state, poly_stipple_enable); - trace_dump_member(bool, state, point_smooth); - trace_dump_member(bool, state, point_sprite); - trace_dump_member(bool, state, point_size_per_vertex); - trace_dump_member(bool, state, multisample); - trace_dump_member(bool, state, line_smooth); - trace_dump_member(bool, state, line_stipple_enable); - trace_dump_member(uint, state, line_stipple_factor); - trace_dump_member(uint, state, line_stipple_pattern); - trace_dump_member(bool, state, line_last_pixel); - trace_dump_member(bool, state, bypass_vs_clip_and_viewport); - trace_dump_member(bool, state, flatshade_first); - trace_dump_member(bool, state, gl_rasterization_rules); - - trace_dump_member(float, state, line_width); - trace_dump_member(float, state, point_size); - trace_dump_member(float, state, point_size_min); - trace_dump_member(float, state, point_size_max); - trace_dump_member(float, state, offset_units); - trace_dump_member(float, state, offset_scale); - - trace_dump_member_array(uint, state, sprite_coord_mode); - - trace_dump_struct_end(); -} - - -void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_poly_stipple"); - - trace_dump_member_begin("stipple"); - trace_dump_array(uint, - state->stipple, - Elements(state->stipple)); - trace_dump_member_end(); - - trace_dump_struct_end(); -} - - -void trace_dump_viewport_state(const struct pipe_viewport_state *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_viewport_state"); - - trace_dump_member_array(float, state, scale); - trace_dump_member_array(float, state, translate); - - trace_dump_struct_end(); -} - - -void trace_dump_scissor_state(const struct pipe_scissor_state *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_scissor_state"); - - trace_dump_member(uint, state, minx); - trace_dump_member(uint, state, miny); - trace_dump_member(uint, state, maxx); - trace_dump_member(uint, state, maxy); - - trace_dump_struct_end(); -} - - -void trace_dump_clip_state(const struct pipe_clip_state *state) -{ - unsigned i; - - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_clip_state"); - - trace_dump_member_begin("ucp"); - trace_dump_array_begin(); - for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) { - trace_dump_elem_begin(); - trace_dump_array(float, state->ucp[i], 4); - trace_dump_elem_end(); - } - trace_dump_array_end(); - trace_dump_member_end(); - - trace_dump_member(uint, state, nr); - - trace_dump_struct_end(); -} - - -void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_constant_buffer"); - - trace_dump_member(buffer_ptr, state, buffer); - - trace_dump_struct_end(); -} - - -void trace_dump_shader_state(const struct pipe_shader_state *state) -{ - static char str[8192]; - - if(!state) { - trace_dump_null(); - return; - } - - tgsi_dump_str(state->tokens, 0, str, sizeof(str)); - - trace_dump_struct_begin("pipe_shader_state"); - - trace_dump_member_begin("tokens"); - trace_dump_string(str); - trace_dump_member_end(); - - trace_dump_struct_end(); -} - - -void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state) -{ - unsigned i; - - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_depth_stencil_alpha_state"); - - trace_dump_member_begin("depth"); - trace_dump_struct_begin("pipe_depth_state"); - trace_dump_member(bool, &state->depth, enabled); - trace_dump_member(bool, &state->depth, writemask); - trace_dump_member(uint, &state->depth, func); - trace_dump_member(bool, &state->depth, occlusion_count); - trace_dump_struct_end(); - trace_dump_member_end(); - - trace_dump_member_begin("stencil"); - trace_dump_array_begin(); - for(i = 0; i < Elements(state->stencil); ++i) { - trace_dump_elem_begin(); - trace_dump_struct_begin("pipe_stencil_state"); - trace_dump_member(bool, &state->stencil[i], enabled); - trace_dump_member(uint, &state->stencil[i], func); - trace_dump_member(uint, &state->stencil[i], fail_op); - trace_dump_member(uint, &state->stencil[i], zpass_op); - trace_dump_member(uint, &state->stencil[i], zfail_op); - trace_dump_member(uint, &state->stencil[i], ref_value); - trace_dump_member(uint, &state->stencil[i], valuemask); - trace_dump_member(uint, &state->stencil[i], writemask); - trace_dump_struct_end(); - trace_dump_elem_end(); - } - trace_dump_array_end(); - trace_dump_member_end(); - - trace_dump_member_begin("alpha"); - trace_dump_struct_begin("pipe_alpha_state"); - trace_dump_member(bool, &state->alpha, enabled); - trace_dump_member(uint, &state->alpha, func); - trace_dump_member(float, &state->alpha, ref_value); - trace_dump_struct_end(); - trace_dump_member_end(); - - trace_dump_struct_end(); -} - - -void trace_dump_blend_state(const struct pipe_blend_state *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_blend_state"); - - trace_dump_member(bool, state, blend_enable); - - trace_dump_member(uint, state, rgb_func); - trace_dump_member(uint, state, rgb_src_factor); - trace_dump_member(uint, state, rgb_dst_factor); - - trace_dump_member(uint, state, alpha_func); - trace_dump_member(uint, state, alpha_src_factor); - trace_dump_member(uint, state, alpha_dst_factor); - - trace_dump_member(bool, state, logicop_enable); - trace_dump_member(uint, state, logicop_func); - - trace_dump_member(uint, state, colormask); - trace_dump_member(bool, state, dither); - - trace_dump_struct_end(); -} - - -void trace_dump_blend_color(const struct pipe_blend_color *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_blend_color"); - - trace_dump_member_array(float, state, color); - - trace_dump_struct_end(); -} - - -void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) -{ - trace_dump_struct_begin("pipe_framebuffer_state"); - - trace_dump_member(uint, state, width); - trace_dump_member(uint, state, height); - trace_dump_member(uint, state, nr_cbufs); - trace_dump_member_array(ptr, state, cbufs); - trace_dump_member(ptr, state, zsbuf); - - trace_dump_struct_end(); -} - - -void trace_dump_sampler_state(const struct pipe_sampler_state *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_sampler_state"); - - trace_dump_member(uint, state, wrap_s); - trace_dump_member(uint, state, wrap_t); - trace_dump_member(uint, state, wrap_r); - trace_dump_member(uint, state, min_img_filter); - trace_dump_member(uint, state, min_mip_filter); - trace_dump_member(uint, state, mag_img_filter); - trace_dump_member(bool, state, compare_mode); - trace_dump_member(uint, state, compare_func); - trace_dump_member(bool, state, normalized_coords); - trace_dump_member(uint, state, prefilter); - trace_dump_member(float, state, shadow_ambient); - trace_dump_member(float, state, lod_bias); - trace_dump_member(float, state, min_lod); - trace_dump_member(float, state, max_lod); - trace_dump_member_array(float, state, border_color); - trace_dump_member(float, state, max_anisotropy); - - trace_dump_struct_end(); -} - - -void trace_dump_surface(const struct pipe_surface *state) -{ - if(!state) { - trace_dump_null(); - return; - } - - trace_dump_struct_begin("pipe_surface"); - - trace_dump_reference(&state->reference); - - trace_dump_member(format, state, format); - trace_dump_member(uint, state, width); - trace_dump_member(uint, state, height); - - trace_dump_member(uint, state, layout); - trace_dump_member(uint, state, offset); - trace_dump_member(uint, state, usage); - - trace_dump_member(ptr, state, texture); - trace_dump_member(uint, state, face); - trace_dump_member(uint, state, level); - trace_dump_member(uint, state, zslice); - - trace_dump_struct_end(); -} - +#include "tgsi/tgsi_parse.h" -void trace_dump_transfer(const struct pipe_transfer *state) +struct trace_shader * trace_shader_create(struct trace_context *tr_ctx, + const struct pipe_shader_state *state, + void *result, + enum trace_shader_type type) { - if(!state) { - trace_dump_null(); - return; - } + struct trace_shader *tr_shdr = CALLOC_STRUCT(trace_shader); - trace_dump_struct_begin("pipe_transfer"); + tr_shdr->state = result; + tr_shdr->type = type; + tr_shdr->tokens = tgsi_dup_tokens(state->tokens); - trace_dump_member(format, state, format); - trace_dump_member(uint, state, width); - trace_dump_member(uint, state, height); + /* works on context as well */ + trace_screen_add_to_list(tr_ctx, shaders, tr_shdr); - trace_dump_member_begin("block"); - trace_dump_block(&state->block); - trace_dump_member_end(); - - trace_dump_member(uint, state, nblocksx); - trace_dump_member(uint, state, nblocksy); - trace_dump_member(uint, state, stride); - trace_dump_member(uint, state, usage); - - trace_dump_member(ptr, state, texture); - trace_dump_member(uint, state, face); - trace_dump_member(uint, state, level); - trace_dump_member(uint, state, zslice); - - trace_dump_struct_end(); + return tr_shdr; } - -void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) +void trace_shader_destroy(struct trace_context *tr_ctx, + struct trace_shader *tr_shdr) { - if(!state) { - trace_dump_null(); - return; - } + trace_screen_remove_from_list(tr_ctx, shaders, tr_shdr); - trace_dump_struct_begin("pipe_vertex_buffer"); - - trace_dump_member(uint, state, stride); - trace_dump_member(uint, state, max_index); - trace_dump_member(uint, state, buffer_offset); - trace_dump_member(buffer_ptr, state, buffer); - - trace_dump_struct_end(); -} - - -void trace_dump_vertex_element(const struct pipe_vertex_element *state) -{ - if(!state) { - trace_dump_null(); - return; + if (tr_shdr->replaced) { + if (tr_shdr->type == TRACE_SHADER_FRAGMENT) + tr_ctx->pipe->delete_fs_state(tr_ctx->pipe, tr_shdr->replaced); + else if (tr_shdr->type == TRACE_SHADER_VERTEX) + tr_ctx->pipe->delete_vs_state(tr_ctx->pipe, tr_shdr->replaced); + else + assert(0); } - trace_dump_struct_begin("pipe_vertex_element"); - - trace_dump_member(uint, state, src_offset); - - trace_dump_member(uint, state, vertex_buffer_index); - trace_dump_member(uint, state, nr_components); - - trace_dump_member(format, state, src_format); - - trace_dump_struct_end(); + FREE(tr_shdr->replaced_tokens); + FREE(tr_shdr->tokens); + FREE(tr_shdr); } diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h index 513ed0ac98..1c16042ee5 100644 --- a/src/gallium/drivers/trace/tr_state.h +++ b/src/gallium/drivers/trace/tr_state.h @@ -1,78 +1,68 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +/* + * Copyright 2009 VMware, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. * - **************************************************************************/ - -#ifndef TR_STATE_H -#define TR_STATE_H - -#include "pipe/p_format.h" -#include "pipe/p_state.h" -#include "pipe/p_shader_tokens.h" - - -void trace_dump_format(enum pipe_format format); - -void trace_dump_block(const struct pipe_format_block *block); - -void trace_dump_template(const struct pipe_texture *templat); - - -void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state); - -void trace_dump_poly_stipple(const struct pipe_poly_stipple *state); - -void trace_dump_viewport_state(const struct pipe_viewport_state *state); - -void trace_dump_scissor_state(const struct pipe_scissor_state *state); - -void trace_dump_clip_state(const struct pipe_clip_state *state); + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ -void trace_dump_constant_buffer(const struct pipe_constant_buffer *state); +#ifndef TR_STATE_H_ +#define TR_STATE_H_ -void trace_dump_token(const struct tgsi_token *token); +#include "tr_context.h" -void trace_dump_shader_state(const struct pipe_shader_state *state); +struct tgsi_token; -void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state); +enum trace_shader_type { + TRACE_SHADER_FRAGMENT = 0, + TRACE_SHADER_VERTEX = 1, + TRACE_SHADER_GEOMETRY = 2, +}; -void trace_dump_blend_state(const struct pipe_blend_state *state); +struct trace_shader +{ + struct tr_list list; -void trace_dump_blend_color(const struct pipe_blend_color *state); + enum trace_shader_type type; -void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state); + void *state; + void *replaced; -void trace_dump_sampler_state(const struct pipe_sampler_state *state); + struct tgsi_token *tokens; + struct tgsi_token *replaced_tokens; -void trace_dump_surface(const struct pipe_surface *state); + boolean disabled; +}; -void trace_dump_transfer(const struct pipe_transfer *state); -void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state); +static INLINE struct trace_shader * +trace_shader(void *state) +{ + return (struct trace_shader *)state; +} -void trace_dump_vertex_element(const struct pipe_vertex_element *state); +struct trace_shader * trace_shader_create(struct trace_context *tr_ctx, + const struct pipe_shader_state *state, + void *result, + enum trace_shader_type type); +void trace_shader_destroy(struct trace_context *tr_ctx, + struct trace_shader *tr_shdr); -#endif /* TR_STATE_H */ +#endif diff --git a/src/gallium/drivers/trace/tr_texture.c b/src/gallium/drivers/trace/tr_texture.c index f4e433792b..1f25fe38d4 100644 --- a/src/gallium/drivers/trace/tr_texture.c +++ b/src/gallium/drivers/trace/tr_texture.c @@ -27,6 +27,7 @@ #include "util/u_hash_table.h" #include "util/u_memory.h" +#include "util/u_simple_list.h" #include "tr_screen.h" #include "tr_texture.h" @@ -53,6 +54,8 @@ trace_texture_create(struct trace_screen *tr_scr, tr_tex->base.screen = &tr_scr->base; tr_tex->texture = texture; + trace_screen_add_to_list(tr_scr, textures, tr_tex); + return &tr_tex->base; error: @@ -64,6 +67,10 @@ error: void trace_texture_destroy(struct trace_texture *tr_tex) { + struct trace_screen *tr_scr = trace_screen(tr_tex->base.screen); + + trace_screen_remove_from_list(tr_scr, textures, tr_tex); + pipe_texture_reference(&tr_tex->texture, NULL); FREE(tr_tex); } @@ -73,6 +80,7 @@ struct pipe_surface * trace_surface_create(struct trace_texture *tr_tex, struct pipe_surface *surface) { + struct trace_screen *tr_scr = trace_screen(tr_tex->base.screen); struct trace_surface *tr_surf; if(!surface) @@ -91,6 +99,8 @@ trace_surface_create(struct trace_texture *tr_tex, pipe_texture_reference(&tr_surf->base.texture, &tr_tex->base); tr_surf->surface = surface; + trace_screen_add_to_list(tr_scr, surfaces, tr_surf); + return &tr_surf->base; error: @@ -102,6 +112,10 @@ error: void trace_surface_destroy(struct trace_surface *tr_surf) { + struct trace_screen *tr_scr = trace_screen(tr_surf->base.texture->screen); + + trace_screen_remove_from_list(tr_scr, surfaces, tr_surf); + pipe_texture_reference(&tr_surf->base.texture, NULL); pipe_surface_reference(&tr_surf->surface, NULL); FREE(tr_surf); @@ -112,6 +126,7 @@ struct pipe_transfer * trace_transfer_create(struct trace_texture *tr_tex, struct pipe_transfer *transfer) { + struct trace_screen *tr_scr = trace_screen(tr_tex->base.screen); struct trace_transfer *tr_trans; if(!transfer) @@ -130,6 +145,8 @@ trace_transfer_create(struct trace_texture *tr_tex, tr_trans->transfer = transfer; assert(tr_trans->base.texture == &tr_tex->base); + trace_screen_add_to_list(tr_scr, transfers, tr_trans); + return &tr_trans->base; error: @@ -141,7 +158,11 @@ error: void trace_transfer_destroy(struct trace_transfer *tr_trans) { + struct trace_screen *tr_scr = trace_screen(tr_trans->base.texture->screen); struct pipe_screen *screen = tr_trans->transfer->texture->screen; + + trace_screen_remove_from_list(tr_scr, transfers, tr_trans); + pipe_texture_reference(&tr_trans->base.texture, NULL); screen->tex_transfer_destroy(tr_trans->transfer); FREE(tr_trans); diff --git a/src/gallium/drivers/trace/tr_texture.h b/src/gallium/drivers/trace/tr_texture.h index 14dafd8b2c..395e523e73 100644 --- a/src/gallium/drivers/trace/tr_texture.h +++ b/src/gallium/drivers/trace/tr_texture.h @@ -40,6 +40,8 @@ struct trace_texture struct pipe_texture base; struct pipe_texture *texture; + + struct tr_list list; }; @@ -48,6 +50,8 @@ struct trace_surface struct pipe_surface base; struct pipe_surface *surface; + + struct tr_list list; }; @@ -57,6 +61,8 @@ struct trace_transfer struct pipe_transfer *transfer; + struct tr_list list; + void *map; }; diff --git a/src/gallium/include/pipe/p_atomic.h b/src/gallium/include/pipe/p_atomic.h index f2fe083efa..0c3fbae428 100644 --- a/src/gallium/include/pipe/p_atomic.h +++ b/src/gallium/include/pipe/p_atomic.h @@ -18,58 +18,29 @@ extern "C" { /* Favor OS-provided implementations. + * + * Where no OS-provided implementation is available, fall back to + * locally coded assembly, compiler intrinsic or ultimately a + * mutex-based implementation. */ -#define PIPE_ATOMIC_OS_UNLOCKED \ - (defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || \ - defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)) - -#define PIPE_ATOMIC_OS_MS_INTERLOCK \ - (!defined(PIPE_CC_GCC) && \ - !PIPE_ATOMIC_OS_UNLOCKED && \ - defined(PIPE_SUBSYSTEM_WINDOWS_USER)) - -#define PIPE_ATOMIC_OS_PROVIDED \ - (PIPE_ATOMIC_OS_UNLOCKED || \ - PIPE_ATOMIC_OS_MS_INTERLOCK) - -/* Where no OS-provided implementation is available, fall back to - * either locally coded assembly or ultimately a mutex-based - * implementation: - */ -#define PIPE_ATOMIC_ASM_GCC_X86 \ - (!PIPE_ATOMIC_OS_PROVIDED && \ - defined(PIPE_CC_GCC) && \ - defined(PIPE_ARCH_X86)) - -/* KW: this was originally used when x86 asm wasn't available. - * Maintain that logic here. - */ -#define PIPE_ATOMIC_GCC_INTRINISIC \ - (!PIPE_ATOMIC_OS_PROVIDED && \ - !PIPE_ATOMIC_ASM_GCC_X86 && \ - defined(PIPE_CC_GCC)) - -#define PIPE_ATOMIC_ASM_MSVC_X86 \ - (!PIPE_ATOMIC_OS_PROVIDED && \ - defined(PIPE_CC_MSVC) && \ - defined(PIPE_ARCH_X86)) - -#define PIPE_ATOMIC_ASM \ - (PIPE_ATOMIC_ASM_GCC_X86 || \ - PIPE_ATOMIC_ASM_GCC_INTRINSIC || \ - PIPE_ATOMIC_ASM_MSVC_X86) - - -/* Where no OS-provided or locally-coded assembly implemenation is - * available, use pipe_mutex: - */ -#define PIPE_ATOMIC_MUTEX \ - (!PIPE_ATOMIC_OS_PROVIDED && \ - !PIPE_ATOMIC_ASM) +#if (defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || \ + defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)) +#define PIPE_ATOMIC_OS_UNLOCKED +#elif (defined(PIPE_CC_MSVC) && defined(PIPE_SUBSYSTEM_WINDOWS_USER)) +#define PIPE_ATOMIC_OS_MS_INTERLOCK +#elif (defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)) +#define PIPE_ATOMIC_ASM_MSVC_X86 +#elif (defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)) +#define PIPE_ATOMIC_ASM_GCC_X86 +#elif defined(PIPE_CC_GCC) +#define PIPE_ATOMIC_GCC_INTRINSIC +#else +#define PIPE_ATOMIC_MUTEX +#endif -#if (PIPE_ATOMIC_ASM_GCC_X86) +#if defined(PIPE_ATOMIC_ASM_GCC_X86) #define PIPE_ATOMIC "GCC x86 assembly" @@ -115,7 +86,7 @@ p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) /* Implementation using GCC-provided synchronization intrinsics */ -#if (PIPE_ATOMIC_ASM_GCC_INTRINSIC) +#if defined(PIPE_ATOMIC_GCC_INTRINSIC) #define PIPE_ATOMIC "GCC Sync Intrinsics" @@ -157,7 +128,7 @@ p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) /* Unlocked version for single threaded environments, such as some * windows kernel modules. */ -#if (PIPE_ATOMIC_OS_UNLOCKED) +#if defined(PIPE_ATOMIC_OS_UNLOCKED) #define PIPE_ATOMIC "Unlocked" @@ -178,7 +149,7 @@ struct pipe_atomic /* Locally coded assembly for MSVC on x86: */ -#if (PIPE_ATOMIC_ASM_MSVC_X86) +#if defined(PIPE_ATOMIC_ASM_MSVC_X86) #define PIPE_ATOMIC "MSVC x86 assembly" @@ -246,7 +217,7 @@ p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) #endif -#if (PIPE_ATOMIC_OS_MS_INTERLOCK) +#if defined(PIPE_ATOMIC_OS_MS_INTERLOCK) #define PIPE_ATOMIC "MS userspace interlocks" @@ -254,7 +225,7 @@ p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) struct pipe_atomic { - long count; + volatile long count; }; #define p_atomic_set(_v, _i) ((_v)->count = (_i)) @@ -263,7 +234,7 @@ struct pipe_atomic static INLINE boolean p_atomic_dec_zero(struct pipe_atomic *v) { - return InterlockedDecrement(&v->count); + return InterlockedDecrement(&v->count) == 0; } static INLINE void @@ -288,7 +259,7 @@ p_atomic_cmpxchg(struct pipe_atomic *v, int32_t old, int32_t _new) -#if (PIPE_ATOMIC_MUTEX) +#if defined(PIPE_ATOMIC_MUTEX) #define PIPE_ATOMIC "mutex-based fallback" diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 29095dcdc3..57e966ac3b 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -42,7 +42,6 @@ struct pipe_state_cache; struct pipe_query; struct pipe_winsys; - /** * Gallium rendering context. Basically: * - state setting functions @@ -192,14 +191,21 @@ struct pipe_context { * Surface functions */ /*@{*/ + + /** + * Copy a block of pixels from one surface to another. + * The surfaces must be of the same format. + */ void (*surface_copy)(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, - struct pipe_surface *src, /* don't make this const - - need to map/unmap */ + struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height); + /** + * Fill a region of a surface with a constant value. + */ void (*surface_fill)(struct pipe_context *pipe, struct pipe_surface *dst, unsigned dstx, unsigned dsty, @@ -224,6 +230,34 @@ struct pipe_context { void (*flush)( struct pipe_context *pipe, unsigned flags, struct pipe_fence_handle **fence ); + + /** + * Check whether a texture is referenced by an unflushed hw command. + * The state-tracker uses this function to optimize away unnecessary + * flushes. It is safe (but wasteful) to always return. + * PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE. + * \param pipe The pipe context whose unflushed hw commands will be + * checked. + * \param level mipmap level. + * \param texture texture to check. + * \param face cubemap face. Use 0 for non-cubemap texture. + */ + + unsigned int (*is_texture_referenced) (struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level); + /** + * Check whether a buffer is referenced by an unflushed hw command. + * The state-tracker uses this function to optimize away unnecessary + * flushes. It is safe (but wasteful) to always return + * PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE. + * \param pipe The pipe context whose unflushed hw commands will be + * checked. + * \param buf Buffer to check. + */ + + unsigned int (*is_buffer_referenced) (struct pipe_context *pipe, + struct pipe_buffer *buf); }; diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 81defa445b..9924046cb2 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -132,6 +132,7 @@ enum pipe_texture_target { #define PIPE_TEX_FACE_NEG_Y 3 #define PIPE_TEX_FACE_POS_Z 4 #define PIPE_TEX_FACE_NEG_Z 5 +#define PIPE_TEX_FACE_MAX 6 #define PIPE_TEX_WRAP_REPEAT 0 #define PIPE_TEX_WRAP_CLAMP 1 @@ -158,14 +159,6 @@ enum pipe_texture_target { #define PIPE_TEX_COMPARE_NONE 0 #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 -#define PIPE_TEX_FACE_POS_X 0 -#define PIPE_TEX_FACE_NEG_X 1 -#define PIPE_TEX_FACE_POS_Y 2 -#define PIPE_TEX_FACE_NEG_Y 3 -#define PIPE_TEX_FACE_POS_Z 4 -#define PIPE_TEX_FACE_NEG_Z 5 -#define PIPE_TEX_FACE_MAX 6 - #define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ #define PIPE_TEXTURE_USAGE_PRIMARY 0x4 /* ie a frontbuffer */ @@ -200,7 +193,7 @@ enum pipe_texture_target { enum pipe_transfer_usage { PIPE_TRANSFER_READ, PIPE_TRANSFER_WRITE, - PIPE_TRANSFER_READ_WRITE //< Read/modify/write + PIPE_TRANSFER_READ_WRITE /**< Read/modify/write */ }; @@ -280,9 +273,8 @@ enum pipe_transfer_usage { /** - * Implementation capabilities/limits - * Passed to pipe->get_param() - * XXX this will need some fine tuning... + * Implementation capabilities/limits which are queried through + * pipe_screen::get_param() and pipe_screen::get_paramf(). */ #define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1 #define PIPE_CAP_NPOT_TEXTURES 2 @@ -312,6 +304,13 @@ enum pipe_transfer_usage { #define PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS 26 +/** + * Referenced query flags. + */ + +#define PIPE_UNREFERENCED 0 +#define PIPE_REFERENCED_FOR_READ (1 << 0) +#define PIPE_REFERENCED_FOR_WRITE (1 << 1) #ifdef __cplusplus } diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index a279eefef9..e31538c95f 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -536,14 +536,36 @@ pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned h return pf_get_nblocksx(block, width)*pf_get_nblocksy(block, height); } +static INLINE size_t +pf_get_stride(const struct pipe_format_block *block, unsigned width) +{ + return pf_get_nblocksx(block, width)*block->size; +} + +static INLINE size_t +pf_get_2d_size(const struct pipe_format_block *block, size_t stride, unsigned height) +{ + return pf_get_nblocksy(block, height)*stride; +} + static INLINE boolean -pf_is_depth_stencil( enum pipe_format format ) +pf_is_depth_or_stencil( enum pipe_format format ) { return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; } static INLINE boolean +pf_is_depth_and_stencil( enum pipe_format format ) +{ + return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) != 0 && + pf_get_component_bits( format, PIPE_FORMAT_COMP_S ) != 0); +} + +/** DEPRECATED: For backwards compatibility */ +#define pf_is_depth_stencil pf_is_depth_or_stencil + +static INLINE boolean pf_is_compressed( enum pipe_format format ) { return pf_layout(format) == PIPE_FORMAT_LAYOUT_DXT ? TRUE : FALSE; diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index ceac755e71..b449522fac 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -87,7 +87,7 @@ struct pipe_screen { * Check if the given pipe_format is supported as a texture or * drawing surface. * \param tex_usage bitmask of PIPE_TEXTURE_USAGE_* - * \param flags bitmask of PIPE_TEXTURE_GEOM_* + * \param geom_flags bitmask of PIPE_TEXTURE_GEOM_* */ boolean (*is_format_supported)( struct pipe_screen *, enum pipe_format format, @@ -102,7 +102,7 @@ struct pipe_screen { const struct pipe_texture *templat); /** - * Create a new texture object, using the given template info, but on top of + * Create a new texture object, using the given template info, but on top of * existing memory. * * It is assumed that the buffer data is layed out according to the expected @@ -144,8 +144,10 @@ struct pipe_screen { /** - * Buffer management. Buffer attributes are mostly fixed over its lifetime. - * + * Create a new buffer. + * \param alignment buffer start address alignment in bytes + * \param usage bitmask of PIPE_BUFFER_USAGE_x + * \param size size in bytes */ struct pipe_buffer *(*buffer_create)( struct pipe_screen *screen, unsigned alignment, @@ -264,7 +266,7 @@ struct pipe_screen { */ int (*fence_signalled)( struct pipe_screen *screen, struct pipe_fence_handle *fence, - unsigned flag ); + unsigned flags ); /** * Wait for the fence to finish. @@ -273,7 +275,7 @@ struct pipe_screen { */ int (*fence_finish)( struct pipe_screen *screen, struct pipe_fence_handle *fence, - unsigned flag ); + unsigned flags ); }; diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 705ae68ec6..4b590bdc90 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -336,7 +336,6 @@ struct pipe_texture unsigned nblocksy[PIPE_MAX_TEXTURE_LEVELS]; /**< allocated height in blocks */ unsigned last_level:8; /**< Index of last mipmap level present/defined */ - unsigned compressed:1; unsigned nr_samples:8; /**< for multisampled surfaces, nr of samples */ diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index de55e99ed4..df6d38904a 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -43,6 +43,8 @@ #include <pthread.h> /* POSIX threads headers */ #include <stdio.h> /* for perror() */ +#define PIPE_THREAD_HAVE_CONDVAR + typedef pthread_t pipe_thread; #define PIPE_THREAD_ROUTINE( name, param ) \ diff --git a/src/gallium/include/state_tracker/dri1_api.h b/src/gallium/include/state_tracker/dri1_api.h new file mode 100644 index 0000000000..b173ba3683 --- /dev/null +++ b/src/gallium/include/state_tracker/dri1_api.h @@ -0,0 +1,82 @@ +#ifndef _DRI1_API_H_ +#define _DRI1_API_H_ + +#include "pipe/p_compiler.h" +#include "pipe/p_screen.h" +#include "pipe/p_format.h" + +#include "state_tracker/drm_api.h" + +#include <drm.h> + +struct pipe_screen; +struct pipe_winsys; +struct pipe_buffer; +struct pipe_context; +struct pipe_texture; + +struct dri1_api_version +{ + int major; + int minor; + int patch_level; +}; + +/** + * This callback struct is intended for drivers that need to take + * the hardware lock on command submission. + */ + +struct dri1_api_lock_funcs +{ + void (*lock) (struct pipe_context * pipe); + void (*unlock) (struct pipe_context * locked_pipe); + boolean(*is_locked) (struct pipe_context * locked_pipe); + boolean(*is_lock_lost) (struct pipe_context * locked_pipe); + void (*clear_lost_lock) (struct pipe_context * locked_pipe); +}; + +struct dri1_api +{ + /** + * For flushing to the front buffer. A driver should implement one and only + * one of the functions below. The present_locked functions allows a dri1 + * driver to pageflip. + */ + + /*@{ */ + + struct pipe_surface *(*front_srf_locked) (struct pipe_context * + locked_pipe); + + void (*present_locked) (struct pipe_context * locked_pipe, + struct pipe_surface * surf, + const struct drm_clip_rect * rect, + unsigned int num_clip, + int x_draw, int y_draw, + const struct drm_clip_rect * src_bbox, + struct pipe_fence_handle ** fence); + /*@} */ +}; + +struct dri1_create_screen_arg +{ + struct drm_create_screen_arg base; + + struct dri1_api_lock_funcs *lf; + void *ddx_info; + int ddx_info_size; + void *sarea; + + struct dri1_api_version ddx_version; + struct dri1_api_version dri_version; + struct dri1_api_version drm_version; + + /* + * out parameters; + */ + + struct dri1_api *api; +}; + +#endif diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h index 435435da29..5790b2f6c7 100644 --- a/src/gallium/include/state_tracker/drm_api.h +++ b/src/gallium/include/state_tracker/drm_api.h @@ -10,13 +10,30 @@ struct pipe_buffer; struct pipe_context; struct pipe_texture; +enum drm_create_screen_mode { + DRM_CREATE_NORMAL = 0, + DRM_CREATE_DRI1, + DRM_CREATE_DRIVER = 1024, + DRM_CREATE_MAX +}; + +/** + * Modes other than DRM_CREATE_NORMAL derive from this struct. + */ +/*@{*/ +struct drm_create_screen_arg { + enum drm_create_screen_mode mode; +}; +/*@}*/ + struct drm_api { /** * Special buffer functions */ /*@{*/ - struct pipe_screen* (*create_screen)(int drmFB, int pciID); + struct pipe_screen* (*create_screen)(int drm_fd, + struct drm_create_screen_arg *arg); struct pipe_context* (*create_context)(struct pipe_screen *screen); /*@}*/ diff --git a/src/gallium/state_trackers/dri2/Makefile b/src/gallium/state_trackers/dri/Makefile index 47750e997e..ef8f19709a 100644 --- a/src/gallium/state_trackers/dri2/Makefile +++ b/src/gallium/state_trackers/dri/Makefile @@ -1,7 +1,7 @@ TOP = ../../../.. include $(TOP)/configs/current -LIBNAME = dri2drm +LIBNAME = dridrm LIBRARY_INCLUDES = \ -I$(TOP)/include \ diff --git a/src/gallium/state_trackers/dri/SConscript b/src/gallium/state_trackers/dri/SConscript new file mode 100644 index 0000000000..ce2c273597 --- /dev/null +++ b/src/gallium/state_trackers/dri/SConscript @@ -0,0 +1,23 @@ +####################################################################### +# SConscript for dri state_tracker + +Import('*') + +if env['dri']: + + env = env.Clone() + + env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/drivers/dri/common', + ]) + + st_dri = env.ConvenienceLibrary( + target = 'st_dri', + source = [ 'dri_context.c', + 'dri_drawable.c', + 'dri_extensions.c', + 'dri_screen.c', + ] + ) + Export('st_dri') diff --git a/src/gallium/state_trackers/dri2/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c index 92c26ac70f..45eaec4ed3 100644 --- a/src/gallium/state_trackers/dri2/dri_context.c +++ b/src/gallium/state_trackers/dri/dri_context.c @@ -32,9 +32,8 @@ #include "dri_screen.h" #include "dri_drawable.h" - - #include "state_tracker/drm_api.h" +#include "state_tracker/dri1_api.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "pipe/p_context.h" @@ -43,11 +42,9 @@ #include "util/u_memory.h" - GLboolean -dri_create_context(const __GLcontextModes *visual, - __DRIcontextPrivate *cPriv, - void *sharedContextPrivate) +dri_create_context(const __GLcontextModes * visual, + __DRIcontextPrivate * cPriv, void *sharedContextPrivate) { __DRIscreenPrivate *sPriv = cPriv->driScreenPriv; struct dri_screen *screen = dri_screen(sPriv); @@ -55,7 +52,7 @@ dri_create_context(const __GLcontextModes *visual, struct st_context *st_share = NULL; if (sharedContextPrivate) { - st_share = ((struct dri_context *) sharedContextPrivate)->st; + st_share = ((struct dri_context *)sharedContextPrivate)->st; } ctx = CALLOC_STRUCT(dri_context); @@ -65,11 +62,12 @@ dri_create_context(const __GLcontextModes *visual, cPriv->driverPrivate = ctx; ctx->cPriv = cPriv; ctx->sPriv = sPriv; + ctx->lock = screen->drmLock; + ctx->d_stamp = -1; + ctx->r_stamp = -1; driParseConfigFiles(&ctx->optionCache, - &screen->optionCache, - sPriv->myNum, - "dri"); + &screen->optionCache, sPriv->myNum, "dri"); ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen); @@ -87,7 +85,7 @@ dri_create_context(const __GLcontextModes *visual, return GL_TRUE; -fail: + fail: if (ctx && ctx->st) st_destroy_context(ctx->st); @@ -98,9 +96,8 @@ fail: return FALSE; } - void -dri_destroy_context(__DRIcontextPrivate *cPriv) +dri_destroy_context(__DRIcontextPrivate * cPriv) { struct dri_context *ctx = dri_context(cPriv); struct dri_screen *screen = dri_screen(cPriv->driScreenPriv); @@ -122,44 +119,64 @@ dri_destroy_context(__DRIcontextPrivate *cPriv) FREE(ctx); } - GLboolean -dri_unbind_context(__DRIcontextPrivate *cPriv) +dri_unbind_context(__DRIcontextPrivate * cPriv) { - struct dri_context *ctx = dri_context(cPriv); - st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* XXX make_current(NULL)? */ + if (cPriv) { + struct dri_context *ctx = dri_context(cPriv); + + if (--ctx->bind_count == 0) { + if (ctx->st && ctx->st == st_get_current()) { + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + st_make_current(NULL, NULL, NULL); + } + } + } + return GL_TRUE; } - GLboolean -dri_make_current(__DRIcontextPrivate *cPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv) +dri_make_current(__DRIcontextPrivate * cPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) { if (cPriv) { struct dri_context *ctx = dri_context(cPriv); struct dri_screen *screen = dri_screen(cPriv->driScreenPriv); struct dri_drawable *draw = dri_drawable(driDrawPriv); struct dri_drawable *read = dri_drawable(driReadPriv); + struct st_context *old_st = st_get_current(); + + if (old_st && old_st != ctx->st) + st_flush(old_st, PIPE_FLUSH_RENDER_CACHE, NULL); + + ++ctx->bind_count; /* This is for situations in which we need a rendering context but * there may not be any currently bound. */ screen->dummyContext = ctx; - st_make_current(ctx->st, - draw->stfb, - read->stfb); - - /* used in dri_flush_frontbuffer */ - ctx->dPriv = driDrawPriv; - - if (driDrawPriv) - dri_get_buffers(driDrawPriv); - if (driDrawPriv != driReadPriv && driReadPriv) - dri_get_buffers(driReadPriv); + if (ctx->dPriv != driDrawPriv) { + ctx->dPriv = driDrawPriv; + ctx->d_stamp = driDrawPriv->lastStamp - 1; + } + if (ctx->rPriv != driReadPriv) { + ctx->rPriv = driReadPriv; + ctx->r_stamp = driReadPriv->lastStamp - 1; + } + + st_make_current(ctx->st, draw->stfb, read->stfb); + + if (__dri1_api_hooks) { + dri1_update_drawables(ctx, draw, read); + } else { + if (driDrawPriv) + dri_get_buffers(driDrawPriv); + if (driDrawPriv != driReadPriv && driReadPriv) + dri_get_buffers(driReadPriv); + } } else { st_make_current(NULL, NULL, NULL); } @@ -167,4 +184,42 @@ dri_make_current(__DRIcontextPrivate *cPriv, return GL_TRUE; } +static void +st_dri_lock(struct pipe_context *pipe) +{ + dri_lock((struct dri_context *)pipe->priv); +} + +static void +st_dri_unlock(struct pipe_context *pipe) +{ + dri_unlock((struct dri_context *)pipe->priv); +} + +static boolean +st_dri_is_locked(struct pipe_context *pipe) +{ + return ((struct dri_context *)pipe->priv)->isLocked; +} + +static boolean +st_dri_lost_lock(struct pipe_context *pipe) +{ + return ((struct dri_context *)pipe->priv)->wsLostLock; +} + +static void +st_dri_clear_lost_lock(struct pipe_context *pipe) +{ + ((struct dri_context *)pipe->priv)->wsLostLock = FALSE; +} + +struct dri1_api_lock_funcs dri1_lf = { + .lock = st_dri_lock, + .unlock = st_dri_unlock, + .is_locked = st_dri_is_locked, + .is_lock_lost = st_dri_lost_lock, + .clear_lost_lock = st_dri_clear_lost_lock +}; + /* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/dri2/dri_context.h b/src/gallium/state_trackers/dri/dri_context.h index e910472700..4650178734 100644 --- a/src/gallium/state_trackers/dri2/dri_context.h +++ b/src/gallium/state_trackers/dri/dri_context.h @@ -36,60 +36,87 @@ #include "drm.h" #include "dri_util.h" - struct pipe_context; struct pipe_fence; struct st_context; struct dri_drawable; - struct dri_context { /* dri */ __DRIscreenPrivate *sPriv; __DRIcontextPrivate *cPriv; __DRIdrawablePrivate *dPriv; + __DRIdrawablePrivate *rPriv; driOptionCache optionCache; + unsigned int d_stamp; + unsigned int r_stamp; + + drmLock *lock; + boolean isLocked; + boolean stLostLock; + boolean wsLostLock; + + unsigned int bind_count; + /* gallium */ struct st_context *st; struct pipe_context *pipe; }; - static INLINE struct dri_context * -dri_context(__DRIcontextPrivate *driContextPriv) +dri_context(__DRIcontextPrivate * driContextPriv) +{ + return (struct dri_context *)driContextPriv->driverPrivate; +} + +static INLINE void +dri_lock(struct dri_context *ctx) { - return (struct dri_context *) driContextPriv->driverPrivate; + drm_context_t hw_context = ctx->cPriv->hHWContext; + char ret = 0; + + DRM_CAS(ctx->lock, hw_context, DRM_LOCK_HELD | hw_context, ret); + if (ret) { + drmGetLock(ctx->sPriv->fd, hw_context, 0); + ctx->stLostLock = TRUE; + ctx->wsLostLock = TRUE; + } + ctx->isLocked = TRUE; } +static INLINE void +dri_unlock(struct dri_context *ctx) +{ + ctx->isLocked = FALSE; + DRM_UNLOCK(ctx->sPriv->fd, ctx->lock, ctx->cPriv->hHWContext); +} /*********************************************************************** * dri_context.c */ -void -dri_destroy_context(__DRIcontextPrivate * driContextPriv); +extern struct dri1_api_lock_funcs dri1_lf; -boolean -dri_unbind_context(__DRIcontextPrivate * driContextPriv); +void dri_destroy_context(__DRIcontextPrivate * driContextPriv); + +boolean dri_unbind_context(__DRIcontextPrivate * driContextPriv); boolean dri_make_current(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); boolean dri_create_context(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); /*********************************************************************** * dri_extensions.c */ -void -dri_init_extensions(struct dri_context *ctx); +void dri_init_extensions(struct dri_context *ctx); #endif diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c new file mode 100644 index 0000000000..09cd3091d2 --- /dev/null +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -0,0 +1,605 @@ +/************************************************************************** + * + * Copyright 2009, VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +/* + * Author: Keith Whitwell <keithw@vmware.com> + * Author: Jakob Bornecrantz <wallbraker@gmail.com> + */ + +#include "dri_screen.h" +#include "dri_context.h" +#include "dri_drawable.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "state_tracker/drm_api.h" +#include "state_tracker/dri1_api.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" + +#include "util/u_memory.h" + +static void +dri_copy_to_front(__DRIdrawablePrivate * dPriv, + struct pipe_surface *from, + int x, int y, unsigned w, unsigned h) +{ + /* TODO send a message to the Xserver to copy to the real front buffer */ +} + +static struct pipe_surface * +dri_surface_from_handle(struct pipe_screen *screen, + unsigned handle, + enum pipe_format format, + unsigned width, unsigned height, unsigned pitch) +{ + struct pipe_surface *surface = NULL; + struct pipe_texture *texture = NULL; + struct pipe_texture templat; + struct pipe_buffer *buf = NULL; + + buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle); + if (!buf) + return NULL; + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = format; + templat.width[0] = width; + templat.height[0] = height; + pf_get_block(templat.format, &templat.block); + + texture = screen->texture_blanket(screen, &templat, &pitch, buf); + + /* we don't need the buffer from this point on */ + pipe_buffer_reference(&buf, NULL); + + if (!texture) + return NULL; + + surface = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* we don't need the texture from this point on */ + pipe_texture_reference(&texture, NULL); + return surface; +} + +/** + * This will be called a drawable is known to have been resized. + */ +void +dri_get_buffers(__DRIdrawablePrivate * dPriv) +{ + struct dri_drawable *drawable = dri_drawable(dPriv); + struct pipe_surface *surface = NULL; + struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen; + __DRIbuffer *buffers = NULL; + __DRIscreen *dri_screen = drawable->sPriv; + __DRIdrawable *dri_drawable = drawable->dPriv; + boolean have_depth = FALSE; + int i, count; + + buffers = (*dri_screen->dri2.loader->getBuffers) (dri_drawable, + &dri_drawable->w, + &dri_drawable->h, + drawable->attachments, + drawable-> + num_attachments, &count, + dri_drawable-> + loaderPrivate); + + if (buffers == NULL) { + return; + } + + /* set one cliprect to cover the whole dri_drawable */ + dri_drawable->x = 0; + dri_drawable->y = 0; + dri_drawable->backX = 0; + dri_drawable->backY = 0; + dri_drawable->numClipRects = 1; + dri_drawable->pClipRects[0].x1 = 0; + dri_drawable->pClipRects[0].y1 = 0; + dri_drawable->pClipRects[0].x2 = dri_drawable->w; + dri_drawable->pClipRects[0].y2 = dri_drawable->h; + dri_drawable->numBackClipRects = 1; + dri_drawable->pBackClipRects[0].x1 = 0; + dri_drawable->pBackClipRects[0].y1 = 0; + dri_drawable->pBackClipRects[0].x2 = dri_drawable->w; + dri_drawable->pBackClipRects[0].y2 = dri_drawable->h; + + if (drawable->old_num == count && + drawable->old_w == dri_drawable->w && + drawable->old_h == dri_drawable->h && + memcmp(drawable->old, buffers, sizeof(__DRIbuffer) * count) == 0) { + return; + } else { + drawable->old_num = count; + drawable->old_w = dri_drawable->w; + drawable->old_h = dri_drawable->h; + memcpy(drawable->old, buffers, sizeof(__DRIbuffer) * count); + } + + for (i = 0; i < count; i++) { + enum pipe_format format = 0; + int index = 0; + + switch (buffers[i].attachment) { + case __DRI_BUFFER_FRONT_LEFT: + index = ST_SURFACE_FRONT_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_FAKE_FRONT_LEFT: + index = ST_SURFACE_FRONT_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_BACK_LEFT: + index = ST_SURFACE_BACK_LEFT; + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + case __DRI_BUFFER_DEPTH: + index = ST_SURFACE_DEPTH; + format = PIPE_FORMAT_Z24S8_UNORM; + break; + case __DRI_BUFFER_STENCIL: + index = ST_SURFACE_DEPTH; + format = PIPE_FORMAT_Z24S8_UNORM; + break; + case __DRI_BUFFER_ACCUM: + default: + assert(0); + } + assert(buffers[i].cpp == 4); + + if (index == ST_SURFACE_DEPTH) { + if (have_depth) + continue; + else + have_depth = TRUE; + } + + surface = dri_surface_from_handle(screen, + buffers[i].name, + format, + dri_drawable->w, + dri_drawable->h, buffers[i].pitch); + + st_set_framebuffer_surface(drawable->stfb, index, surface); + pipe_surface_reference(&surface, NULL); + } + /* this needed, or else the state tracker fails to pick the new buffers */ + st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h); +} + +void +dri_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, void *context_private) +{ + struct dri_context *ctx = (struct dri_context *)context_private; + + dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height); +} + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +boolean +dri_create_buffer(__DRIscreenPrivate * sPriv, + __DRIdrawablePrivate * dPriv, + const __GLcontextModes * visual, boolean isPixmap) +{ + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct dri_screen *screen = sPriv->private; + struct dri_drawable *drawable = NULL; + struct pipe_screen *pscreen = screen->pipe_screen; + int i; + + if (isPixmap) + goto fail; /* not implemented */ + + drawable = CALLOC_STRUCT(dri_drawable); + if (drawable == NULL) + goto fail; + + /* XXX: todo: use the pipe_screen queries to figure out which + * render targets are supportable. + */ + assert(visual->redBits == 8); + assert(visual->depthBits == 24 || visual->depthBits == 0); + assert(visual->stencilBits == 8 || visual->stencilBits == 0); + + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits) { + if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) + depthFormat = PIPE_FORMAT_Z24S8_UNORM; + else + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + } else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits) { + if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) + stencilFormat = PIPE_FORMAT_Z24S8_UNORM; + else + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + } else + stencilFormat = PIPE_FORMAT_NONE; + + drawable->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + dPriv->w, + dPriv->h, (void *)drawable); + if (drawable->stfb == NULL) + goto fail; + + drawable->sPriv = sPriv; + drawable->dPriv = dPriv; + dPriv->driverPrivate = (void *)drawable; + + /* setup dri2 buffers information */ + i = 0; + drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT; +#if 0 + /* TODO incase of double buffer visual, delay fake creation */ + drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; +#endif + if (visual->doubleBufferMode) + drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; + if (visual->depthBits) + drawable->attachments[i++] = __DRI_BUFFER_DEPTH; + if (visual->stencilBits) + drawable->attachments[i++] = __DRI_BUFFER_STENCIL; + drawable->num_attachments = i; + + drawable->desired_fences = 2; + + return GL_TRUE; + fail: + FREE(drawable); + return GL_FALSE; +} + +static struct pipe_fence_handle * +dri_swap_fences_pop_front(struct dri_drawable *draw) +{ + struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen; + struct pipe_fence_handle *fence = NULL; + + if (draw->cur_fences >= draw->desired_fences) { + screen->fence_reference(screen, &fence, draw->swap_fences[draw->tail]); + screen->fence_reference(screen, &draw->swap_fences[draw->tail++], NULL); + --draw->cur_fences; + draw->tail &= DRI_SWAP_FENCES_MASK; + } + return fence; +} + +static void +dri_swap_fences_push_back(struct dri_drawable *draw, + struct pipe_fence_handle *fence) +{ + struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen; + + if (!fence) + return; + + if (draw->cur_fences < DRI_SWAP_FENCES_MAX) { + draw->cur_fences++; + screen->fence_reference(screen, &draw->swap_fences[draw->head++], + fence); + draw->head &= DRI_SWAP_FENCES_MASK; + } +} + +void +dri_destroy_buffer(__DRIdrawablePrivate * dPriv) +{ + struct dri_drawable *drawable = dri_drawable(dPriv); + struct pipe_fence_handle *fence; + struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen; + + st_unreference_framebuffer(drawable->stfb); + drawable->desired_fences = 0; + while (drawable->cur_fences) { + fence = dri_swap_fences_pop_front(drawable); + screen->fence_reference(screen, &fence, NULL); + } + + FREE(drawable); +} + +static void +dri1_update_drawables_locked(struct dri_context *ctx, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + if (ctx->stLostLock) { + ctx->stLostLock = FALSE; + if (driDrawPriv == driReadPriv) + DRI_VALIDATE_DRAWABLE_INFO(ctx->sPriv, driDrawPriv); + else + DRI_VALIDATE_TWO_DRAWABLES_INFO(ctx->sPriv, driDrawPriv, + driReadPriv); + } +} + +/** + * This ensures all contexts which bind to a drawable pick up the + * drawable change and signal new buffer state. + * Calling st_resize_framebuffer for each context may seem like overkill, + * but no new buffers will actually be allocated if the dimensions don't + * change. + */ + +static void +dri1_propagate_drawable_change(struct dri_context *ctx) +{ + __DRIdrawablePrivate *dPriv = ctx->dPriv; + __DRIdrawablePrivate *rPriv = ctx->rPriv; + boolean flushed = FALSE; + + if (dPriv && ctx->d_stamp != dPriv->lastStamp) { + + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + flushed = TRUE; + ctx->d_stamp = dPriv->lastStamp; + st_resize_framebuffer(dri_drawable(dPriv)->stfb, dPriv->w, dPriv->h); + + } + + if (rPriv && dPriv != rPriv && ctx->r_stamp != rPriv->lastStamp) { + + if (!flushed) + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + ctx->r_stamp = rPriv->lastStamp; + st_resize_framebuffer(dri_drawable(rPriv)->stfb, rPriv->w, rPriv->h); + + } else if (rPriv && dPriv == rPriv) { + + ctx->r_stamp = ctx->d_stamp; + + } +} + +void +dri1_update_drawables(struct dri_context *ctx, + struct dri_drawable *draw, struct dri_drawable *read) +{ + dri_lock(ctx); + dri1_update_drawables_locked(ctx, draw->dPriv, read->dPriv); + dri_unlock(ctx); + + dri1_propagate_drawable_change(ctx); +} + +static INLINE boolean +dri1_intersect_src_bbox(struct drm_clip_rect *dst, + int dst_x, + int dst_y, + const struct drm_clip_rect *src, + const struct drm_clip_rect *bbox) +{ + int xy1; + int xy2; + + xy1 = ((int)src->x1 > (int)bbox->x1 + dst_x) ? src->x1 : + (int)bbox->x1 + dst_x; + xy2 = ((int)src->x2 < (int)bbox->x2 + dst_x) ? src->x2 : + (int)bbox->x2 + dst_x; + if (xy1 >= xy2 || xy1 < 0) + return FALSE; + + dst->x1 = xy1; + dst->x2 = xy2; + + xy1 = ((int)src->y1 > (int)bbox->y1 + dst_y) ? src->y1 : + (int)bbox->y1 + dst_y; + xy2 = ((int)src->y2 < (int)bbox->y2 + dst_y) ? src->y2 : + (int)bbox->y2 + dst_y; + if (xy1 >= xy2 || xy1 < 0) + return FALSE; + + dst->y1 = xy1; + dst->y2 = xy2; + return TRUE; +} + +static void +dri1_swap_copy(struct dri_context *ctx, + struct pipe_surface *dst, + struct pipe_surface *src, + __DRIdrawablePrivate * dPriv, const struct drm_clip_rect *bbox) +{ + struct pipe_context *pipe = ctx->pipe; + struct drm_clip_rect clip; + struct drm_clip_rect *cur; + int i; + + cur = dPriv->pClipRects; + + for (i = 0; i < dPriv->numClipRects; ++i) { + if (dri1_intersect_src_bbox(&clip, dPriv->x, dPriv->y, cur++, bbox)) + pipe->surface_copy(pipe, dst, clip.x1, clip.y1, + src, + (int)clip.x1 - dPriv->x, + (int)clip.y1 - dPriv->y, + clip.x2 - clip.x1, clip.y2 - clip.y1); + } +} + +static void +dri1_copy_to_front(struct dri_context *ctx, + struct pipe_surface *surf, + __DRIdrawablePrivate * dPriv, + const struct drm_clip_rect *sub_box, + struct pipe_fence_handle **fence) +{ + struct pipe_context *pipe = ctx->pipe; + boolean save_lost_lock; + uint cur_w; + uint cur_h; + struct drm_clip_rect bbox; + boolean visible = TRUE; + + *fence = NULL; + + dri_lock(ctx); + save_lost_lock = ctx->stLostLock; + dri1_update_drawables_locked(ctx, dPriv, dPriv); + st_get_framebuffer_dimensions(dri_drawable(dPriv)->stfb, &cur_w, &cur_h); + + bbox.x1 = 0; + bbox.x2 = cur_w; + bbox.y1 = 0; + bbox.y2 = cur_h; + + if (sub_box) + visible = dri1_intersect_src_bbox(&bbox, 0, 0, &bbox, sub_box); + + if (visible && __dri1_api_hooks->present_locked) { + + __dri1_api_hooks->present_locked(pipe, + surf, + dPriv->pClipRects, + dPriv->numClipRects, + dPriv->x, dPriv->y, &bbox, fence); + + } else if (visible && __dri1_api_hooks->front_srf_locked) { + + struct pipe_surface *front = __dri1_api_hooks->front_srf_locked(pipe); + + if (front) + dri1_swap_copy(ctx, front, surf, dPriv, &bbox); + + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, fence); + } + + ctx->stLostLock = save_lost_lock; + + /** + * FIXME: Revisit this: Update drawables on copy_sub_buffer ? + */ + + if (!sub_box) + dri1_update_drawables_locked(ctx, ctx->dPriv, ctx->rPriv); + + dri_unlock(ctx); + dri1_propagate_drawable_change(ctx); +} + +void +dri1_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, void *context_private) +{ + struct dri_context *ctx = (struct dri_context *)context_private; + struct pipe_fence_handle *dummy_fence; + + dri1_copy_to_front(ctx, surf, ctx->dPriv, NULL, &dummy_fence); + screen->fence_reference(screen, &dummy_fence, NULL); + + /** + * FIXME: Do we need swap throttling here? + */ +} + +void +dri_swap_buffers(__DRIdrawablePrivate * dPriv) +{ + struct dri_context *ctx; + struct pipe_surface *back_surf; + struct dri_drawable *draw = dri_drawable(dPriv); + struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen; + struct pipe_fence_handle *fence; + struct st_context *st = st_get_current(); + + assert(__dri1_api_hooks != NULL); + + if (!st) + return; /* For now */ + + ctx = (struct dri_context *)st->pipe->priv; + + st_get_framebuffer_surface(draw->stfb, ST_SURFACE_BACK_LEFT, &back_surf); + if (back_surf) { + st_notify_swapbuffers(draw->stfb); + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + fence = dri_swap_fences_pop_front(draw); + if (fence) { + (void)screen->fence_finish(screen, fence, 0); + screen->fence_reference(screen, &fence, NULL); + } + dri1_copy_to_front(ctx, back_surf, dPriv, NULL, &fence); + dri_swap_fences_push_back(draw, fence); + screen->fence_reference(screen, &fence, NULL); + } +} + +void +dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + struct pipe_screen *screen = dri_screen(dPriv->driScreenPriv)->pipe_screen; + struct drm_clip_rect sub_bbox; + struct dri_context *ctx; + struct pipe_surface *back_surf; + struct dri_drawable *draw = dri_drawable(dPriv); + struct pipe_fence_handle *dummy_fence; + struct st_context *st = st_get_current(); + + assert(__dri1_api_hooks != NULL); + + if (!st) + return; + + ctx = (struct dri_context *)st->pipe->priv; + + sub_bbox.x1 = x; + sub_bbox.x2 = x + w; + sub_bbox.y1 = y; + sub_bbox.y2 = y + h; + + st_get_framebuffer_surface(draw->stfb, ST_SURFACE_BACK_LEFT, &back_surf); + if (back_surf) { + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + dri1_copy_to_front(ctx, back_surf, dPriv, &sub_bbox, &dummy_fence); + screen->fence_reference(screen, &dummy_fence, NULL); + } +} + +/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/dri2/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h index 185c657b35..0f654d804a 100644 --- a/src/gallium/state_trackers/dri2/dri_drawable.h +++ b/src/gallium/state_trackers/dri/dri_drawable.h @@ -31,9 +31,11 @@ #include "pipe/p_compiler.h" struct pipe_surface; -struct pipe_fence; +struct pipe_fence_handle; struct st_framebuffer; +#define DRI_SWAP_FENCES_MAX 8 +#define DRI_SWAP_FENCES_MASK 7 struct dri_drawable { @@ -44,46 +46,54 @@ struct dri_drawable unsigned attachments[8]; unsigned num_attachments; + __DRIbuffer old[8]; + unsigned old_num; + unsigned old_w; + unsigned old_h; + /* gallium */ struct st_framebuffer *stfb; + struct pipe_fence_handle *swap_fences[DRI_SWAP_FENCES_MAX]; + unsigned int head; + unsigned int tail; + unsigned int desired_fences; + unsigned int cur_fences; }; - static INLINE struct dri_drawable * dri_drawable(__DRIdrawablePrivate * driDrawPriv) { - return (struct dri_drawable *) driDrawPriv->driverPrivate; + return (struct dri_drawable *)driDrawPriv->driverPrivate; } - /*********************************************************************** * dri_drawable.c */ boolean -dri_create_buffer(__DRIscreenPrivate *sPriv, - __DRIdrawablePrivate *dPriv, - const __GLcontextModes *visual, - boolean isPixmap); +dri_create_buffer(__DRIscreenPrivate * sPriv, + __DRIdrawablePrivate * dPriv, + const __GLcontextModes * visual, boolean isPixmap); void dri_flush_frontbuffer(struct pipe_screen *screen, - struct pipe_surface *surf, - void *context_private); + struct pipe_surface *surf, void *context_private); -void -dri_swap_buffers(__DRIdrawablePrivate * dPriv); +void dri_swap_buffers(__DRIdrawablePrivate * dPriv); void -dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, - int x, int y, - int w, int h); +dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); -void -dri_get_buffers(__DRIdrawablePrivate * dPriv); +void dri_get_buffers(__DRIdrawablePrivate * dPriv); + +void dri_destroy_buffer(__DRIdrawablePrivate * dPriv); void -dri_destroy_buffer(__DRIdrawablePrivate *dPriv); +dri1_update_drawables(struct dri_context *ctx, + struct dri_drawable *draw, struct dri_drawable *read); +void +dri1_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, void *context_private); #endif /* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/dri2/dri_extensions.c b/src/gallium/state_trackers/dri/dri_extensions.c index 732d1e89b0..2f48162526 100644 --- a/src/gallium/state_trackers/dri2/dri_extensions.c +++ b/src/gallium/state_trackers/dri/dri_extensions.c @@ -36,9 +36,11 @@ #define need_GL_ARB_multisample #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters +#define need_GL_ARB_shader_objects #define need_GL_ARB_texture_compression #define need_GL_ARB_vertex_buffer_object #define need_GL_ARB_vertex_program +#define need_GL_ARB_vertex_shader #define need_GL_ARB_window_pos #define need_GL_EXT_blend_color #define need_GL_EXT_blend_equation_separate @@ -50,17 +52,23 @@ #define need_GL_EXT_multi_draw_arrays #define need_GL_EXT_secondary_color #define need_GL_NV_vertex_program +#define need_GL_VERSION_2_0 +#define need_GL_VERSION_2_1 #include "extension_helper.h" - /** * Extension strings exported by the driver. */ const struct dri_extension card_extensions[] = { + {"GL_ARB_fragment_shader", NULL}, {"GL_ARB_multisample", GL_ARB_multisample_functions}, {"GL_ARB_multitexture", NULL}, {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, + {"GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, + {"GL_ARB_shader_objects", GL_ARB_shader_objects_functions}, {"GL_ARB_texture_border_clamp", NULL}, {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, {"GL_ARB_texture_cube_map", NULL}, @@ -70,11 +78,12 @@ const struct dri_extension card_extensions[] = { {"GL_ARB_texture_mirrored_repeat", NULL}, {"GL_ARB_texture_rectangle", NULL}, {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_equation_separate", + GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, {"GL_EXT_blend_subtract", NULL}, @@ -98,11 +107,10 @@ const struct dri_extension card_extensions[] = { {"GL_NV_blend_square", NULL}, {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, + {"GL_SGIS_generate_mipmap", NULL}, {NULL, NULL} }; - void dri_init_extensions(struct dri_context *ctx) { diff --git a/src/gallium/state_trackers/dri2/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c index ab5878a4bc..d3392ee690 100644 --- a/src/gallium/state_trackers/dri2/dri_screen.c +++ b/src/gallium/state_trackers/dri/dri_screen.c @@ -40,68 +40,38 @@ #include "pipe/p_context.h" #include "pipe/p_screen.h" #include "pipe/p_inlines.h" +#include "pipe/p_format.h" #include "state_tracker/drm_api.h" +#include "state_tracker/dri1_api.h" #include "state_tracker/st_public.h" #include "state_tracker/st_cb_fbo.h" - PUBLIC const char __driConfigOptions[] = DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - /*DRI_CONF_FORCE_S3TC_ENABLE(false)*/ - DRI_CONF_ALLOW_LARGE_TEXTURES(1) + /*DRI_CONF_FORCE_S3TC_ENABLE(false) */ + DRI_CONF_ALLOW_LARGE_TEXTURES(1) DRI_CONF_SECTION_END DRI_CONF_END; + const uint __driNConfigOptions = 3; -const uint __driNConfigOptions = 3; - - -static const __DRIextension *dri_screen_extensions[] = { - &driReadDrawableExtension, - &driCopySubBufferExtension.base, - &driSwapControlExtension.base, - &driFrameTrackingExtension.base, - &driMediaStreamCounterExtension.base, - NULL -}; - - -static void -dri_get_drm_minor(struct dri_screen *screen) -{ - /* TODO get the real minor */ - screen->minor = 0; -} - - -static void -dri_get_device_id(struct dri_screen *screen) -{ - char path[512]; - FILE *file; - - /* - * There must be a better way to get the deviceID. - * XXX this only works on Linux. - */ - snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", screen->minor); - file = fopen(path, "r"); - if (!file) { - return; - } - - fgets(path, sizeof(path), file); - sscanf(path, "%x", &screen->deviceID); - fclose(file); -} + static const __DRIextension *dri_screen_extensions[] = { + &driReadDrawableExtension, + &driCopySubBufferExtension.base, + &driSwapControlExtension.base, + &driFrameTrackingExtension.base, + &driMediaStreamCounterExtension.base, + NULL + }; +struct dri1_api *__dri1_api_hooks = NULL; static const __DRIconfig ** -dri_fill_in_modes(__DRIscreenPrivate *psp, - unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) +dri_fill_in_modes(__DRIscreenPrivate * psp, + unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) { __DRIconfig **configs; __GLcontextModes *m; @@ -125,9 +95,9 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, depth_bits_array[1] = 24; depth_bits_array[2] = 24; - stencil_bits_array[0] = 0; /* no depth or stencil */ - stencil_bits_array[1] = 0; /* z24x8 */ - stencil_bits_array[2] = 8; /* z24s8 */ + stencil_bits_array[0] = 0; /* no depth or stencil */ + stencil_bits_array[1] = 0; /* z24x8 */ + stencil_bits_array[2] = 8; /* z24s8 */ msaa_samples_array[0] = 0; @@ -135,22 +105,22 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, back_buffer_factor = 3; msaa_samples_factor = 1; - num_modes = depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4; + num_modes = + depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4; if (pixel_bits == 16) { fb_format = GL_RGB; fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { + } else { fb_format = GL_BGRA; fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; } configs = driCreateConfigs(fb_format, fb_type, depth_bits_array, - stencil_bits_array, depth_buffer_factor, - back_buffer_modes, back_buffer_factor, - msaa_samples_array, msaa_samples_factor); + stencil_bits_array, depth_buffer_factor, + back_buffer_modes, back_buffer_factor, + msaa_samples_array, msaa_samples_factor); if (configs == NULL) { debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__); return NULL; @@ -159,49 +129,99 @@ dri_fill_in_modes(__DRIscreenPrivate *psp, for (i = 0; configs[i]; i++) { m = &configs[i]->modes; if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; + m->visualRating = GLX_SLOW_CONFIG; } } - return (const const __DRIconfig **) configs; + return (const const __DRIconfig **)configs; } - /** * Get information about previous buffer swaps. */ -int -dri_get_swap_info(__DRIdrawablePrivate * dPriv, - __DRIswapInfo * sInfo) +static int +dri_get_swap_info(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) { - if (dPriv == NULL || - dPriv->driverPrivate == NULL || - sInfo == NULL) + if (dPriv == NULL || dPriv->driverPrivate == NULL || sInfo == NULL) return -1; else return 0; } +static INLINE void +dri_copy_version(struct dri1_api_version *dst, + const struct __DRIversionRec *src) +{ + dst->major = src->major; + dst->minor = src->minor; + dst->patch_level = src->patch; +} -/** - * NULL stub for old dri loaders - */ -const __DRIconfig ** -dri_init_screen(__DRIscreenPrivate *sPriv) +static const __DRIconfig ** +dri_init_screen(__DRIscreenPrivate * sPriv) { + struct dri_screen *screen; + const __DRIconfig **configs; + struct dri1_create_screen_arg arg; + + dri_init_extensions(NULL); + + screen = CALLOC_STRUCT(dri_screen); + if (!screen) + return NULL; + + screen->sPriv = sPriv; + screen->fd = sPriv->fd; + screen->drmLock = (drmLock *) & sPriv->pSAREA->lock; + + sPriv->private = (void *)screen; + sPriv->extensions = dri_screen_extensions; + + arg.base.mode = DRM_CREATE_DRI1; + arg.lf = &dri1_lf; + arg.ddx_info = sPriv->pDevPriv; + arg.ddx_info_size = sPriv->devPrivSize; + arg.sarea = sPriv->pSAREA; + dri_copy_version(&arg.ddx_version, &sPriv->ddx_version); + dri_copy_version(&arg.dri_version, &sPriv->dri_version); + dri_copy_version(&arg.drm_version, &sPriv->drm_version); + arg.api = NULL; + + screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg.base); + + if (!screen->pipe_screen || !arg.api) { + debug_printf("%s: failed to create dri1 screen\n", __FUNCTION__); + goto out_no_screen; + } + + __dri1_api_hooks = arg.api; + + screen->pipe_screen->flush_frontbuffer = dri1_flush_frontbuffer; + driParseOptionInfo(&screen->optionCache, + __driConfigOptions, __driNConfigOptions); + + configs = dri_fill_in_modes(sPriv, sPriv->fbBPP, 24, 8, 1); + if (!configs) + goto out_no_configs; + + return configs; + out_no_configs: + screen->pipe_screen->destroy(screen->pipe_screen); + out_no_screen: + FREE(screen); return NULL; } - /** * This is the driver specific part of the createNewScreen entry point. * * Returns the __GLcontextModes supported by this driver. */ -const __DRIconfig ** -dri_init_screen2(__DRIscreenPrivate *sPriv) +static const __DRIconfig ** +dri_init_screen2(__DRIscreenPrivate * sPriv) { struct dri_screen *screen; + struct drm_create_screen_arg arg; /* Set up dispatch table to cope with all known extensions */ dri_init_extensions(NULL); @@ -212,13 +232,11 @@ dri_init_screen2(__DRIscreenPrivate *sPriv) screen->sPriv = sPriv; screen->fd = sPriv->fd; - dri_get_drm_minor(screen); - dri_get_device_id(screen); - sPriv->private = (void *) screen; + sPriv->private = (void *)screen; sPriv->extensions = dri_screen_extensions; + arg.mode = DRM_CREATE_NORMAL; - - screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, screen->deviceID); + screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg); if (!screen->pipe_screen) { debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__); goto fail; @@ -228,20 +246,14 @@ dri_init_screen2(__DRIscreenPrivate *sPriv) screen->pipe_screen->flush_frontbuffer = dri_flush_frontbuffer; driParseOptionInfo(&screen->optionCache, - __driConfigOptions, - __driNConfigOptions); - - return dri_fill_in_modes(sPriv, - 4 * 8, - 24, - 8, - 1); -fail: + __driConfigOptions, __driNConfigOptions); + + return dri_fill_in_modes(sPriv, 4 * 8, 24, 8, 1); + fail: return NULL; } - -void +static void dri_destroy_screen(__DRIscreenPrivate * sPriv) { struct dri_screen *screen = dri_screen(sPriv); @@ -251,22 +263,22 @@ dri_destroy_screen(__DRIscreenPrivate * sPriv) sPriv->private = NULL; } - PUBLIC const struct __DriverAPIRec driDriverAPI = { - .InitScreen = dri_init_screen, /* not supported but exported */ - .DestroyScreen = dri_destroy_screen, - .CreateContext = dri_create_context, - .DestroyContext = dri_destroy_context, - .CreateBuffer = dri_create_buffer, - .DestroyBuffer = dri_destroy_buffer, - .SwapBuffers = dri_swap_buffers, /* not supported but exported */ - .MakeCurrent = dri_make_current, - .UnbindContext = dri_unbind_context, - .GetSwapInfo = dri_get_swap_info, - .GetDrawableMSC = driDrawableGetMSC32, - .WaitForMSC = driWaitForMSC32, - .CopySubBuffer = dri_copy_sub_buffer, /* not supported but exported */ - .InitScreen2 = dri_init_screen2, + .InitScreen = dri_init_screen, + .DestroyScreen = dri_destroy_screen, + .CreateContext = dri_create_context, + .DestroyContext = dri_destroy_context, + .CreateBuffer = dri_create_buffer, + .DestroyBuffer = dri_destroy_buffer, + .SwapBuffers = dri_swap_buffers, + .MakeCurrent = dri_make_current, + .UnbindContext = dri_unbind_context, + .GetSwapInfo = dri_get_swap_info, + .GetDrawableMSC = driDrawableGetMSC32, + .WaitForMSC = driWaitForMSC32, + .CopySubBuffer = dri_copy_sub_buffer, + .InitScreen = dri_init_screen, + .InitScreen2 = dri_init_screen2, }; /* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/dri2/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h index fe2676d0be..100d9e50e0 100644 --- a/src/gallium/state_trackers/dri2/dri_screen.h +++ b/src/gallium/state_trackers/dri/dri_screen.h @@ -37,6 +37,8 @@ #include "pipe/p_compiler.h" +#include "state_tracker/dri1_api.h" + struct dri_screen { /* dri */ @@ -54,36 +56,26 @@ struct dri_screen struct dri_context *dummyContext; /* drm */ - int deviceID; int fd; - int minor; + drmLock *drmLock; /* gallium */ struct pipe_winsys *pipe_winsys; struct pipe_screen *pipe_screen; }; - /** cast wrapper */ static INLINE struct dri_screen * -dri_screen(__DRIscreenPrivate *sPriv) +dri_screen(__DRIscreenPrivate * sPriv) { - return (struct dri_screen *) sPriv->private; + return (struct dri_screen *)sPriv->private; } - /*********************************************************************** * dri_screen.c */ -const __DRIconfig ** -dri_init_screen2(__DRIscreenPrivate *sPriv); - -void -dri_destroy_screen(__DRIscreenPrivate * sPriv); -int -dri_get_swap_info(__DRIdrawablePrivate * dPriv, - __DRIswapInfo * sInfo); +extern struct dri1_api *__dri1_api_hooks; #endif diff --git a/src/gallium/state_trackers/dri2/dri_drawable.c b/src/gallium/state_trackers/dri2/dri_drawable.c deleted file mode 100644 index 2e3f4099e2..0000000000 --- a/src/gallium/state_trackers/dri2/dri_drawable.c +++ /dev/null @@ -1,325 +0,0 @@ -/************************************************************************** - * - * Copyright 2009, VMware, Inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ -/* - * Author: Keith Whitwell <keithw@vmware.com> - * Author: Jakob Bornecrantz <wallbraker@gmail.com> - */ - -#include "dri_screen.h" -#include "dri_context.h" -#include "dri_drawable.h" - -#include "pipe/p_context.h" -#include "pipe/p_screen.h" -#include "pipe/p_inlines.h" -#include "state_tracker/drm_api.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - -#include "util/u_memory.h" - - -static void -dri_copy_to_front(__DRIdrawablePrivate *dPriv, - struct pipe_surface *from, - int x, int y, unsigned w, unsigned h) -{ - /* TODO send a message to the Xserver to copy to the real front buffer */ -} - - -static struct pipe_surface * -dri_surface_from_handle(struct pipe_screen *screen, - unsigned handle, - enum pipe_format format, - unsigned width, - unsigned height, - unsigned pitch) -{ - struct pipe_surface *surface = NULL; - struct pipe_texture *texture = NULL; - struct pipe_texture templat; - struct pipe_buffer *buf = NULL; - - buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle); - if (!buf) - return NULL; - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = format; - templat.width[0] = width; - templat.height[0] = height; - pf_get_block(templat.format, &templat.block); - - texture = screen->texture_blanket(screen, - &templat, - &pitch, - buf); - - /* we don't need the buffer from this point on */ - pipe_buffer_reference(&buf, NULL); - - if (!texture) - return NULL; - - surface = screen->get_tex_surface(screen, texture, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - - /* we don't need the texture from this point on */ - pipe_texture_reference(&texture, NULL); - return surface; -} - - -/** - * This will be called a drawable is known to have been resized. - */ -void -dri_get_buffers(__DRIdrawablePrivate *dPriv) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - struct pipe_surface *surface = NULL; - struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen; - __DRIbuffer *buffers = NULL; - __DRIscreen *dri_screen = drawable->sPriv; - __DRIdrawable *dri_drawable = drawable->dPriv; - boolean have_depth = FALSE; - int i, count; - - buffers = (*dri_screen->dri2.loader->getBuffers)(dri_drawable, - &dri_drawable->w, - &dri_drawable->h, - drawable->attachments, - drawable->num_attachments, - &count, - dri_drawable->loaderPrivate); - - if (buffers == NULL) { - return; - } - - /* set one cliprect to cover the whole dri_drawable */ - dri_drawable->x = 0; - dri_drawable->y = 0; - dri_drawable->backX = 0; - dri_drawable->backY = 0; - dri_drawable->numClipRects = 1; - dri_drawable->pClipRects[0].x1 = 0; - dri_drawable->pClipRects[0].y1 = 0; - dri_drawable->pClipRects[0].x2 = dri_drawable->w; - dri_drawable->pClipRects[0].y2 = dri_drawable->h; - dri_drawable->numBackClipRects = 1; - dri_drawable->pBackClipRects[0].x1 = 0; - dri_drawable->pBackClipRects[0].y1 = 0; - dri_drawable->pBackClipRects[0].x2 = dri_drawable->w; - dri_drawable->pBackClipRects[0].y2 = dri_drawable->h; - - for (i = 0; i < count; i++) { - enum pipe_format format = 0; - int index = 0; - - switch (buffers[i].attachment) { - case __DRI_BUFFER_FRONT_LEFT: - index = ST_SURFACE_FRONT_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - case __DRI_BUFFER_FAKE_FRONT_LEFT: - index = ST_SURFACE_FRONT_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - case __DRI_BUFFER_BACK_LEFT: - index = ST_SURFACE_BACK_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; - case __DRI_BUFFER_DEPTH: - index = ST_SURFACE_DEPTH; - format = PIPE_FORMAT_Z24S8_UNORM; - break; - case __DRI_BUFFER_STENCIL: - index = ST_SURFACE_DEPTH; - format = PIPE_FORMAT_Z24S8_UNORM; - break; - case __DRI_BUFFER_ACCUM: - default: - assert(0); - } - assert(buffers[i].cpp == 4); - - if (index == ST_SURFACE_DEPTH) { - if (have_depth) - continue; - else - have_depth = TRUE; - } - - surface = dri_surface_from_handle(screen, - buffers[i].name, - format, - dri_drawable->w, - dri_drawable->h, - buffers[i].pitch); - - st_set_framebuffer_surface(drawable->stfb, index, surface); - pipe_surface_reference(&surface, NULL); - } - /* this needed, or else the state tracker fails to pick the new buffers */ - st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h); -} - - -void -dri_flush_frontbuffer(struct pipe_screen *screen, - struct pipe_surface *surf, - void *context_private) -{ - struct dri_context *ctx = (struct dri_context *)context_private; - dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height); -} - - -void -dri_swap_buffers(__DRIdrawablePrivate * dPriv) -{ - /* not needed for dri2 */ - assert(0); -} - - -void -dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - /* not needed for dri2 */ - assert(0); -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -boolean -dri_create_buffer(__DRIscreenPrivate *sPriv, - __DRIdrawablePrivate *dPriv, - const __GLcontextModes *visual, - boolean isPixmap) -{ - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct dri_screen *screen = sPriv->private; - struct dri_drawable *drawable = NULL; - struct pipe_screen *pscreen = screen->pipe_screen; - int i; - - if (isPixmap) - goto fail; /* not implemented */ - - drawable = CALLOC_STRUCT(dri_drawable); - if (drawable == NULL) - goto fail; - - /* XXX: todo: use the pipe_screen queries to figure out which - * render targets are supportable. - */ - assert(visual->redBits == 8); - assert(visual->depthBits == 24 || visual->depthBits == 0); - assert(visual->stencilBits == 8 || visual->stencilBits == 0); - - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits) { - if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, - PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET | - PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) - depthFormat = PIPE_FORMAT_Z24S8_UNORM; - else - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - } else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits) { - if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, - PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET | - PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) - stencilFormat = PIPE_FORMAT_Z24S8_UNORM; - else - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - } else - stencilFormat = PIPE_FORMAT_NONE; - - drawable->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - dPriv->w, - dPriv->h, - (void*) drawable); - if (drawable->stfb == NULL) - goto fail; - - drawable->sPriv = sPriv; - drawable->dPriv = dPriv; - dPriv->driverPrivate = (void *) drawable; - - /* setup dri2 buffers information */ - i = 0; - drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT; -#if 0 - /* TODO incase of double buffer visual, delay fake creation */ - drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; -#endif - if (visual->doubleBufferMode) - drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; - if (visual->depthBits) - drawable->attachments[i++] = __DRI_BUFFER_DEPTH; - if (visual->stencilBits) - drawable->attachments[i++] = __DRI_BUFFER_STENCIL; - drawable->num_attachments = i; - - return GL_TRUE; -fail: - FREE(drawable); - return GL_FALSE; -} - - -void -dri_destroy_buffer(__DRIdrawablePrivate *dPriv) -{ - struct dri_drawable *drawable = dri_drawable(dPriv); - - st_unreference_framebuffer(drawable->stfb); - - FREE(drawable); -} - -/* vim: set sw=3 ts=8 sts=3 expandtab: */ diff --git a/src/gallium/state_trackers/egl/Makefile b/src/gallium/state_trackers/egl/Makefile index 692a3c8b76..e825aa718b 100644 --- a/src/gallium/state_trackers/egl/Makefile +++ b/src/gallium/state_trackers/egl/Makefile @@ -1,29 +1,19 @@ -TARGET = libegldrm.a -CFILES = $(wildcard ./*.c) -OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) -GALLIUMDIR = ../.. -TOP = ../../../.. +TOP = ../../../.. +include $(TOP)/configs/current -include ${TOP}/configs/current +LIBNAME = egldrm -CFLAGS := \ - -I${GALLIUMDIR}/include \ - -I${GALLIUMDIR}/auxiliary \ - -I${TOP}/src/mesa/drivers/dri/common \ - -I${TOP}/src/mesa \ - -I$(TOP)/include \ - -I$(TOP)/src/egl/main \ - ${LIBDRM_CFLAGS} \ - ${CFLAGS} +LIBRARY_INCLUDES = \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main \ + $(shell pkg-config --cflags-only-I libdrm) -############################################# -.PHONY = all clean +C_SOURCES = $(wildcard ./*.c) -all: ${TARGET} -${TARGET}: ${OBJECTS} - ar rcs $@ $^ - -clean: - rm -rf ${OBJECTS} ${TARGET} +include ../../Makefile.template diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c index abdf84544f..8e62008461 100644 --- a/src/gallium/state_trackers/egl/egl_tracker.c +++ b/src/gallium/state_trackers/egl/egl_tracker.c @@ -146,7 +146,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) dev->drmFD = fd; drm_get_device_id(dev); - dev->screen = drm_api_hooks.create_screen(dev->drmFD, dev->deviceID); + dev->screen = drm_api_hooks.create_screen(dev->drmFD, NULL); if (!dev->screen) goto err_screen; dev->winsys = dev->screen->winsys; diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/fakeglx.c index 65e7048188..85e7ecfb9e 100644 --- a/src/gallium/state_trackers/glx/xlib/fakeglx.c +++ b/src/gallium/state_trackers/glx/xlib/fakeglx.c @@ -97,6 +97,9 @@ struct fake_glx_context { +#define DEFAULT_DIRECT GL_TRUE + + /**********************************************************************/ /*** GLX Visual Code ***/ /**********************************************************************/ @@ -1059,7 +1062,7 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, return NULL; } - glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.isDirect = DEFAULT_DIRECT; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ @@ -1296,9 +1299,9 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) static Bool Fake_glXIsDirect( Display *dpy, GLXContext ctx ) { - (void) dpy; + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; (void) ctx; - return False; + return glxCtx->glxContext.isDirect; } @@ -2055,7 +2058,7 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, return NULL; } - glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.isDirect = DEFAULT_DIRECT; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ @@ -2277,7 +2280,7 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re return NULL; } - glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.isDirect = DEFAULT_DIRECT; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c index a3d1651653..79c2230588 100644 --- a/src/gallium/state_trackers/glx/xlib/xm_api.c +++ b/src/gallium/state_trackers/glx/xlib/xm_api.c @@ -1100,26 +1100,19 @@ XMesaContext XMesaGetCurrentContext( void ) - - - -/* - * Copy the back buffer to the front buffer. If there's no back buffer - * this is a no-op. +/** + * Swap front and back color buffers and have winsys display front buffer. + * If there's no front color buffer no swap actually occurs. */ PUBLIC void XMesaSwapBuffers( XMesaBuffer b ) { - struct pipe_surface *surf; + struct pipe_surface *frontLeftSurf; - /* If we're swapping the buffer associated with the current context - * we have to flush any pending rendering commands first. - */ - st_notify_swapbuffers(b->stfb); + st_swapbuffers(b->stfb, &frontLeftSurf, NULL); - st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf); - if (surf) { - driver.display_surface(b, surf); + if (frontLeftSurf) { + driver.display_surface(b, frontLeftSurf); } xmesa_check_and_update_buffer_size(NULL, b); diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i index f41a95e6eb..1d513abf3c 100644 --- a/src/gallium/state_trackers/python/p_texture.i +++ b/src/gallium/state_trackers/python/p_texture.i @@ -426,7 +426,7 @@ struct st_surface if(!*STRING) return; - pipe_buffer_read(screen, $self, 0, $self->size, STRING); + pipe_buffer_read(screen, $self, 0, $self->size, *STRING); } %cstring_input_binary(const char *STRING, unsigned LENGTH); diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index 5ea07724a5..5885e162c2 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -32,7 +32,7 @@ import struct import gallium import model -import parser +import parse as parser try: @@ -43,19 +43,27 @@ except ImportError: return struct.unpack(fmt, buf[offset:offset + size]) -def make_image(surface): +def make_image(surface, x=None, y=None, w=None, h=None): + if x is None: + x = 0 + if y is None: + y = 0 + if w is None: + w = surface.width - x + if h is None: + h = surface.height - y data = surface.get_tile_rgba8(0, 0, surface.width, surface.height) import Image outimage = Image.fromstring('RGBA', (surface.width, surface.height), data, "raw", 'RGBA', 0, 1) return outimage -def save_image(filename, surface): - outimage = make_image(surface) +def save_image(filename, surface, x=None, y=None, w=None, h=None): + outimage = make_image(surface, x, y, w, h) outimage.save(filename, "PNG") -def show_image(surface, title): - outimage = make_image(surface) +def show_image(surface, title, x=None, y=None, w=None, h=None): + outimage = make_image(surface, x, y, w, h) import Tkinter as tk from PIL import Image, ImageTk @@ -305,7 +313,11 @@ class Screen(Object): def get_tex_transfer(self, texture, face, level, zslice, usage, x, y, w, h): if texture is None: return None - return Transfer(texture.get_surface(face, level, zslice), x, y, w, h) + transfer = Transfer(texture.get_surface(face, level, zslice), x, y, w, h) + if transfer and usage != gallium.PIPE_TRANSFER_WRITE: + if self.interpreter.options.all: + self.interpreter.present(transfer.surface, 'transf_read', x, y, w, h) + return transfer def tex_transfer_destroy(self, transfer): self.interpreter.unregister_object(transfer) @@ -314,6 +326,8 @@ class Screen(Object): if transfer is None: return transfer.surface.put_tile_raw(transfer.x, transfer.y, transfer.w, transfer.h, data, stride) + if self.interpreter.options.all: + self.interpreter.present(transfer.surface, 'transf_write', transfer.x, transfer.y, transfer.w, transfer.h) def user_buffer_create(self, data, size): # We don't really care to distinguish between user and regular buffers @@ -577,6 +591,14 @@ class Context(Object): self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count) self._set_dirty() + def is_texture_referenced(self, texture, face, level): + #return self.real.is_texture_referenced(format, texture, face, level) + pass + + def is_buffer_referenced(self, buf): + #return self.real.is_buffer_referenced(format, buf) + pass + def _set_dirty(self): if self.interpreter.options.step: self._present() @@ -602,6 +624,9 @@ class Context(Object): if self.cbufs and self.cbufs[0]: self.interpreter.present(self.cbufs[0], "cbuf") + if self.zsbuf: + if self.interpreter.options.all: + self.interpreter.present(self.zsbuf, "zsbuf") class Interpreter(parser.TraceDumper): @@ -671,16 +696,16 @@ class Interpreter(parser.TraceDumper): def verbosity(self, level): return self.options.verbosity >= level - def present(self, surface, description): + def present(self, surface, description, x=None, y=None, w=None, h=None): if self.call_no < self.options.start: return if self.options.images: - filename = '%s_%04u.png' % (description, self.call_no) - save_image(filename, surface) + filename = '%04u_%s.png' % (self.call_no, description) + save_image(filename, surface, x, y, w, h) else: title = '%u. %s' % (self.call_no, description) - show_image(surface, title) + show_image(surface, title, x, y, w, h) class Main(parser.Main): @@ -690,6 +715,7 @@ class Main(parser.Main): optparser.add_option("-q", "--quiet", action="store_const", const=0, dest="verbosity", help="no messages") optparser.add_option("-v", "--verbose", action="count", dest="verbosity", default=1, help="increase verbosity level") optparser.add_option("-i", "--images", action="store_true", dest="images", default=False, help="save images instead of showing them") + optparser.add_option("-a", "--all", action="store_true", dest="all", default=False, help="show depth, stencil, and transfers") optparser.add_option("-s", "--step", action="store_true", dest="step", default=False, help="step trhough every draw") optparser.add_option("-f", "--from", action="store", type="int", dest="start", default=0, help="from call no") optparser.add_option("-t", "--to", action="store", type="int", dest="stop", default=0, help="until call no") diff --git a/src/gallium/state_trackers/python/retrace/parse.py b/src/gallium/state_trackers/python/retrace/parse.py new file mode 100755 index 0000000000..b0f3e8a432 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/parse.py @@ -0,0 +1,392 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +import sys +import xml.parsers.expat +import binascii +import optparse + +from model import * + + +ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF = range(4) + + +class XmlToken: + + def __init__(self, type, name_or_data, attrs = None, line = None, column = None): + assert type in (ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF) + self.type = type + self.name_or_data = name_or_data + self.attrs = attrs + self.line = line + self.column = column + + def __str__(self): + if self.type == ELEMENT_START: + return '<' + self.name_or_data + ' ...>' + if self.type == ELEMENT_END: + return '</' + self.name_or_data + '>' + if self.type == CHARACTER_DATA: + return self.name_or_data + if self.type == EOF: + return 'end of file' + assert 0 + + +class XmlTokenizer: + """Expat based XML tokenizer.""" + + def __init__(self, fp, skip_ws = True): + self.fp = fp + self.tokens = [] + self.index = 0 + self.final = False + self.skip_ws = skip_ws + + self.character_pos = 0, 0 + self.character_data = '' + + self.parser = xml.parsers.expat.ParserCreate() + self.parser.StartElementHandler = self.handle_element_start + self.parser.EndElementHandler = self.handle_element_end + self.parser.CharacterDataHandler = self.handle_character_data + + def handle_element_start(self, name, attributes): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_START, name, attributes, line, column) + self.tokens.append(token) + + def handle_element_end(self, name): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_END, name, None, line, column) + self.tokens.append(token) + + def handle_character_data(self, data): + if not self.character_data: + self.character_pos = self.pos() + self.character_data += data + + def finish_character_data(self): + if self.character_data: + if not self.skip_ws or not self.character_data.isspace(): + line, column = self.character_pos + token = XmlToken(CHARACTER_DATA, self.character_data, None, line, column) + self.tokens.append(token) + self.character_data = '' + + def next(self): + size = 16*1024 + while self.index >= len(self.tokens) and not self.final: + self.tokens = [] + self.index = 0 + data = self.fp.read(size) + self.final = len(data) < size + data = data.rstrip('\0') + try: + self.parser.Parse(data, self.final) + except xml.parsers.expat.ExpatError, e: + #if e.code == xml.parsers.expat.errors.XML_ERROR_NO_ELEMENTS: + if e.code == 3: + pass + else: + raise e + if self.index >= len(self.tokens): + line, column = self.pos() + token = XmlToken(EOF, None, None, line, column) + else: + token = self.tokens[self.index] + self.index += 1 + return token + + def pos(self): + return self.parser.CurrentLineNumber, self.parser.CurrentColumnNumber + + +class TokenMismatch(Exception): + + def __init__(self, expected, found): + self.expected = expected + self.found = found + + def __str__(self): + return '%u:%u: %s expected, %s found' % (self.found.line, self.found.column, str(self.expected), str(self.found)) + + + +class XmlParser: + """Base XML document parser.""" + + def __init__(self, fp): + self.tokenizer = XmlTokenizer(fp) + self.consume() + + def consume(self): + self.token = self.tokenizer.next() + + def match_element_start(self, name): + return self.token.type == ELEMENT_START and self.token.name_or_data == name + + def match_element_end(self, name): + return self.token.type == ELEMENT_END and self.token.name_or_data == name + + def element_start(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_START: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + attrs = self.token.attrs + self.consume() + return attrs + + def element_end(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_END: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + self.consume() + + def character_data(self, strip = True): + data = '' + while self.token.type == CHARACTER_DATA: + data += self.token.name_or_data + self.consume() + if strip: + data = data.strip() + return data + + +class TraceParser(XmlParser): + + def __init__(self, fp): + XmlParser.__init__(self, fp) + self.last_call_no = 0 + + def parse(self): + self.element_start('trace') + while self.token.type not in (ELEMENT_END, EOF): + call = self.parse_call() + self.handle_call(call) + if self.token.type != EOF: + self.element_end('trace') + + def parse_call(self): + attrs = self.element_start('call') + try: + no = int(attrs['no']) + except KeyError: + self.last_call_no += 1 + no = self.last_call_no + else: + self.last_call_no = no + klass = attrs['class'] + method = attrs['method'] + args = [] + ret = None + while self.token.type == ELEMENT_START: + if self.token.name_or_data == 'arg': + arg = self.parse_arg() + args.append(arg) + elif self.token.name_or_data == 'ret': + ret = self.parse_ret() + elif self.token.name_or_data == 'call': + # ignore nested function calls + self.parse_call() + else: + raise TokenMismatch("<arg ...> or <ret ...>", self.token) + self.element_end('call') + + return Call(no, klass, method, args, ret) + + def parse_arg(self): + attrs = self.element_start('arg') + name = attrs['name'] + value = self.parse_value() + self.element_end('arg') + + return name, value + + def parse_ret(self): + attrs = self.element_start('ret') + value = self.parse_value() + self.element_end('ret') + + return value + + def parse_value(self): + expected_tokens = ('null', 'bool', 'int', 'uint', 'float', 'string', 'enum', 'array', 'struct', 'ptr', 'bytes') + if self.token.type == ELEMENT_START: + if self.token.name_or_data in expected_tokens: + method = getattr(self, 'parse_' + self.token.name_or_data) + return method() + raise TokenMismatch(" or " .join(expected_tokens), self.token) + + def parse_null(self): + self.element_start('null') + self.element_end('null') + return Literal(None) + + def parse_bool(self): + self.element_start('bool') + value = int(self.character_data()) + self.element_end('bool') + return Literal(value) + + def parse_int(self): + self.element_start('int') + value = int(self.character_data()) + self.element_end('int') + return Literal(value) + + def parse_uint(self): + self.element_start('uint') + value = int(self.character_data()) + self.element_end('uint') + return Literal(value) + + def parse_float(self): + self.element_start('float') + value = float(self.character_data()) + self.element_end('float') + return Literal(value) + + def parse_enum(self): + self.element_start('enum') + name = self.character_data() + self.element_end('enum') + return NamedConstant(name) + + def parse_string(self): + self.element_start('string') + value = self.character_data() + self.element_end('string') + return Literal(value) + + def parse_bytes(self): + self.element_start('bytes') + value = binascii.a2b_hex(self.character_data()) + self.element_end('bytes') + return Literal(value) + + def parse_array(self): + self.element_start('array') + elems = [] + while self.token.type != ELEMENT_END: + elems.append(self.parse_elem()) + self.element_end('array') + return Array(elems) + + def parse_elem(self): + self.element_start('elem') + value = self.parse_value() + self.element_end('elem') + return value + + def parse_struct(self): + attrs = self.element_start('struct') + name = attrs['name'] + members = [] + while self.token.type != ELEMENT_END: + members.append(self.parse_member()) + self.element_end('struct') + return Struct(name, members) + + def parse_member(self): + attrs = self.element_start('member') + name = attrs['name'] + value = self.parse_value() + self.element_end('member') + + return name, value + + def parse_ptr(self): + self.element_start('ptr') + address = self.character_data() + self.element_end('ptr') + + return Pointer(address) + + def handle_call(self, call): + pass + + +class TraceDumper(TraceParser): + + def __init__(self, fp): + TraceParser.__init__(self, fp) + self.formatter = format.DefaultFormatter(sys.stdout) + self.pretty_printer = PrettyPrinter(self.formatter) + + def handle_call(self, call): + call.visit(self.pretty_printer) + self.formatter.newline() + + +class Main: + '''Common main class for all retrace command line utilities.''' + + def __init__(self): + pass + + def main(self): + optparser = self.get_optparser() + (options, args) = optparser.parse_args(sys.argv[1:]) + + if args: + for arg in args: + if arg.endswith('.gz'): + from gzip import GzipFile + stream = GzipFile(arg, 'rt') + elif arg.endswith('.bz2'): + from bz2 import BZ2File + stream = BZ2File(arg, 'rt') + else: + stream = open(arg, 'rt') + self.process_arg(stream, options) + else: + self.process_arg(stream, options) + + def get_optparser(self): + optparser = optparse.OptionParser( + usage="\n\t%prog [options] [traces] ...") + return optparser + + def process_arg(self, stream, options): + parser = TraceDumper(stream) + parser.parse() + + +if __name__ == '__main__': + Main().main() diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py index b0f3e8a432..bd47c9a6b0 100755 --- a/src/gallium/state_trackers/python/retrace/parser.py +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -27,365 +27,7 @@ ########################################################################## -import sys -import xml.parsers.expat -import binascii -import optparse - -from model import * - - -ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF = range(4) - - -class XmlToken: - - def __init__(self, type, name_or_data, attrs = None, line = None, column = None): - assert type in (ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF) - self.type = type - self.name_or_data = name_or_data - self.attrs = attrs - self.line = line - self.column = column - - def __str__(self): - if self.type == ELEMENT_START: - return '<' + self.name_or_data + ' ...>' - if self.type == ELEMENT_END: - return '</' + self.name_or_data + '>' - if self.type == CHARACTER_DATA: - return self.name_or_data - if self.type == EOF: - return 'end of file' - assert 0 - - -class XmlTokenizer: - """Expat based XML tokenizer.""" - - def __init__(self, fp, skip_ws = True): - self.fp = fp - self.tokens = [] - self.index = 0 - self.final = False - self.skip_ws = skip_ws - - self.character_pos = 0, 0 - self.character_data = '' - - self.parser = xml.parsers.expat.ParserCreate() - self.parser.StartElementHandler = self.handle_element_start - self.parser.EndElementHandler = self.handle_element_end - self.parser.CharacterDataHandler = self.handle_character_data - - def handle_element_start(self, name, attributes): - self.finish_character_data() - line, column = self.pos() - token = XmlToken(ELEMENT_START, name, attributes, line, column) - self.tokens.append(token) - - def handle_element_end(self, name): - self.finish_character_data() - line, column = self.pos() - token = XmlToken(ELEMENT_END, name, None, line, column) - self.tokens.append(token) - - def handle_character_data(self, data): - if not self.character_data: - self.character_pos = self.pos() - self.character_data += data - - def finish_character_data(self): - if self.character_data: - if not self.skip_ws or not self.character_data.isspace(): - line, column = self.character_pos - token = XmlToken(CHARACTER_DATA, self.character_data, None, line, column) - self.tokens.append(token) - self.character_data = '' - - def next(self): - size = 16*1024 - while self.index >= len(self.tokens) and not self.final: - self.tokens = [] - self.index = 0 - data = self.fp.read(size) - self.final = len(data) < size - data = data.rstrip('\0') - try: - self.parser.Parse(data, self.final) - except xml.parsers.expat.ExpatError, e: - #if e.code == xml.parsers.expat.errors.XML_ERROR_NO_ELEMENTS: - if e.code == 3: - pass - else: - raise e - if self.index >= len(self.tokens): - line, column = self.pos() - token = XmlToken(EOF, None, None, line, column) - else: - token = self.tokens[self.index] - self.index += 1 - return token - - def pos(self): - return self.parser.CurrentLineNumber, self.parser.CurrentColumnNumber - - -class TokenMismatch(Exception): - - def __init__(self, expected, found): - self.expected = expected - self.found = found - - def __str__(self): - return '%u:%u: %s expected, %s found' % (self.found.line, self.found.column, str(self.expected), str(self.found)) - - - -class XmlParser: - """Base XML document parser.""" - - def __init__(self, fp): - self.tokenizer = XmlTokenizer(fp) - self.consume() - - def consume(self): - self.token = self.tokenizer.next() - - def match_element_start(self, name): - return self.token.type == ELEMENT_START and self.token.name_or_data == name - - def match_element_end(self, name): - return self.token.type == ELEMENT_END and self.token.name_or_data == name - - def element_start(self, name): - while self.token.type == CHARACTER_DATA: - self.consume() - if self.token.type != ELEMENT_START: - raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) - if self.token.name_or_data != name: - raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) - attrs = self.token.attrs - self.consume() - return attrs - - def element_end(self, name): - while self.token.type == CHARACTER_DATA: - self.consume() - if self.token.type != ELEMENT_END: - raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) - if self.token.name_or_data != name: - raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) - self.consume() - - def character_data(self, strip = True): - data = '' - while self.token.type == CHARACTER_DATA: - data += self.token.name_or_data - self.consume() - if strip: - data = data.strip() - return data - - -class TraceParser(XmlParser): - - def __init__(self, fp): - XmlParser.__init__(self, fp) - self.last_call_no = 0 - - def parse(self): - self.element_start('trace') - while self.token.type not in (ELEMENT_END, EOF): - call = self.parse_call() - self.handle_call(call) - if self.token.type != EOF: - self.element_end('trace') - - def parse_call(self): - attrs = self.element_start('call') - try: - no = int(attrs['no']) - except KeyError: - self.last_call_no += 1 - no = self.last_call_no - else: - self.last_call_no = no - klass = attrs['class'] - method = attrs['method'] - args = [] - ret = None - while self.token.type == ELEMENT_START: - if self.token.name_or_data == 'arg': - arg = self.parse_arg() - args.append(arg) - elif self.token.name_or_data == 'ret': - ret = self.parse_ret() - elif self.token.name_or_data == 'call': - # ignore nested function calls - self.parse_call() - else: - raise TokenMismatch("<arg ...> or <ret ...>", self.token) - self.element_end('call') - - return Call(no, klass, method, args, ret) - - def parse_arg(self): - attrs = self.element_start('arg') - name = attrs['name'] - value = self.parse_value() - self.element_end('arg') - - return name, value - - def parse_ret(self): - attrs = self.element_start('ret') - value = self.parse_value() - self.element_end('ret') - - return value - - def parse_value(self): - expected_tokens = ('null', 'bool', 'int', 'uint', 'float', 'string', 'enum', 'array', 'struct', 'ptr', 'bytes') - if self.token.type == ELEMENT_START: - if self.token.name_or_data in expected_tokens: - method = getattr(self, 'parse_' + self.token.name_or_data) - return method() - raise TokenMismatch(" or " .join(expected_tokens), self.token) - - def parse_null(self): - self.element_start('null') - self.element_end('null') - return Literal(None) - - def parse_bool(self): - self.element_start('bool') - value = int(self.character_data()) - self.element_end('bool') - return Literal(value) - - def parse_int(self): - self.element_start('int') - value = int(self.character_data()) - self.element_end('int') - return Literal(value) - - def parse_uint(self): - self.element_start('uint') - value = int(self.character_data()) - self.element_end('uint') - return Literal(value) - - def parse_float(self): - self.element_start('float') - value = float(self.character_data()) - self.element_end('float') - return Literal(value) - - def parse_enum(self): - self.element_start('enum') - name = self.character_data() - self.element_end('enum') - return NamedConstant(name) - - def parse_string(self): - self.element_start('string') - value = self.character_data() - self.element_end('string') - return Literal(value) - - def parse_bytes(self): - self.element_start('bytes') - value = binascii.a2b_hex(self.character_data()) - self.element_end('bytes') - return Literal(value) - - def parse_array(self): - self.element_start('array') - elems = [] - while self.token.type != ELEMENT_END: - elems.append(self.parse_elem()) - self.element_end('array') - return Array(elems) - - def parse_elem(self): - self.element_start('elem') - value = self.parse_value() - self.element_end('elem') - return value - - def parse_struct(self): - attrs = self.element_start('struct') - name = attrs['name'] - members = [] - while self.token.type != ELEMENT_END: - members.append(self.parse_member()) - self.element_end('struct') - return Struct(name, members) - - def parse_member(self): - attrs = self.element_start('member') - name = attrs['name'] - value = self.parse_value() - self.element_end('member') - - return name, value - - def parse_ptr(self): - self.element_start('ptr') - address = self.character_data() - self.element_end('ptr') - - return Pointer(address) - - def handle_call(self, call): - pass - - -class TraceDumper(TraceParser): - - def __init__(self, fp): - TraceParser.__init__(self, fp) - self.formatter = format.DefaultFormatter(sys.stdout) - self.pretty_printer = PrettyPrinter(self.formatter) - - def handle_call(self, call): - call.visit(self.pretty_printer) - self.formatter.newline() - - -class Main: - '''Common main class for all retrace command line utilities.''' - - def __init__(self): - pass - - def main(self): - optparser = self.get_optparser() - (options, args) = optparser.parse_args(sys.argv[1:]) - - if args: - for arg in args: - if arg.endswith('.gz'): - from gzip import GzipFile - stream = GzipFile(arg, 'rt') - elif arg.endswith('.bz2'): - from bz2 import BZ2File - stream = BZ2File(arg, 'rt') - else: - stream = open(arg, 'rt') - self.process_arg(stream, options) - else: - self.process_arg(stream, options) - - def get_optparser(self): - optparser = optparse.OptionParser( - usage="\n\t%prog [options] [traces] ...") - return optparser - - def process_arg(self, stream, options): - parser = TraceDumper(stream) - parser.parse() +from parse import * if __name__ == '__main__': diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 41cdeaa6fd..f0a4826a00 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -260,7 +260,7 @@ st_softpipe_screen_create(void) static struct pipe_context * st_softpipe_context_create(struct pipe_screen *screen) { - return softpipe_create(screen, screen->winsys, NULL); + return softpipe_create(screen); } diff --git a/src/gallium/state_trackers/python/tests/texture_sample.py b/src/gallium/state_trackers/python/tests/texture_sample.py index a382424667..06130ea1c9 100755 --- a/src/gallium/state_trackers/python/tests/texture_sample.py +++ b/src/gallium/state_trackers/python/tests/texture_sample.py @@ -99,7 +99,7 @@ def is_pot(n): return n & (n - 1) == 0 -class TextureTest(TestCase): +class TextureColorSampleTest(TestCase): tags = ( 'target', @@ -286,6 +286,206 @@ class TextureTest(TestCase): self.assert_rgba(cbuf, x, y, w, h, expected_rgba, 4.0/256, 0.85) +class TextureDepthSampleTest(TestCase): + + tags = ( + 'target', + 'format', + 'width', + 'height', + 'depth', + 'last_level', + 'face', + 'level', + 'zslice', + ) + + def test(self): + dev = self.dev + + target = self.target + format = self.format + width = self.width + height = self.height + depth = self.depth + last_level = self.last_level + face = self.face + level = self.level + zslice = self.zslice + + tex_usage = PIPE_TEXTURE_USAGE_SAMPLER + geom_flags = 0 + if width != height: + geom_flags |= PIPE_TEXTURE_GEOM_NON_SQUARE + if not is_pot(width) or not is_pot(height) or not is_pot(depth): + geom_flags |= PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO + + if not dev.is_format_supported(format, target, tex_usage, geom_flags): + raise TestSkip + + ctx = self.dev.context_create() + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + depth_stencil_alpha.depth.enabled = 1 + depth_stencil_alpha.depth.writemask = 1 + depth_stencil_alpha.depth.func = PIPE_FUNC_LESS + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_vs_clip_and_viewport = 1 + ctx.set_rasterizer(rasterizer) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + sampler.min_lod = 0 + sampler.max_lod = PIPE_MAX_TEXTURE_LEVELS - 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create( + target = target, + format = format, + width = width, + height = height, + depth = depth, + last_level = last_level, + tex_usage = tex_usage, + ) + + expected_rgba = FloatArray(height*width*4) + texture.get_surface( + face = face, + level = level, + zslice = zslice, + ).sample_rgba(expected_rgba) + + ctx.set_sampler_texture(0, texture) + + # framebuffer + cbuf_tex = dev.texture_create( + PIPE_FORMAT_A8R8G8B8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) + + zsbuf_tex = dev.texture_create( + PIPE_FORMAT_Z24X8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) + + cbuf = cbuf_tex.get_surface() + zsbuf = zsbuf_tex.get_surface() + fb = Framebuffer() + fb.width = width + fb.height = height + fb.nr_cbufs = 1 + fb.set_cbuf(0, cbuf) + fb.set_zsbuf(zsbuf) + ctx.set_framebuffer(fb) + rgba = FloatArray(4); + rgba[0] = 0.5 + rgba[1] = 0.5 + rgba[2] = 0.5 + rgba[3] = 0.5 + ctx.clear(PIPE_CLEAR_DEPTHSTENCIL, rgba, 1.0, 0) + del fb + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], GENERIC, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], GENERIC, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) + + # fragment shader + op = { + PIPE_TEXTURE_1D: "1D", + PIPE_TEXTURE_2D: "2D", + PIPE_TEXTURE_3D: "3D", + PIPE_TEXTURE_CUBE: "CUBE", + }[target] + fs = Shader(''' + FRAG1.1 + DCL IN[0], GENERIC[0], LINEAR + DCL SAMP[0], CONSTANT + DCL OUT[0].z, POSITION + 0:TEX OUT[0].z, IN[0], SAMP[0], %s + 1:END + ''' % op) + #fs.dump() + ctx.set_fragment_shader(fs) + + nverts = 4 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + x = 0 + y = 0 + w, h = minify((width, height), level) + + pos = [ + [x, y], + [x+w, y], + [x+w, y+h], + [x, y+h], + ] + + tex = tex_coords(texture, face, level, zslice) + + for i in range(0, 4): + j = 8*i + verts[j + 0] = pos[i][0] # x + verts[j + 1] = pos[i][1] # y + verts[j + 2] = 0.0 # z + verts[j + 3] = 1.0 # w + verts[j + 4] = tex[i][0] # s + verts[j + 5] = tex[i][1] # r + verts[j + 6] = tex[i][2] # q + verts[j + 7] = 1.0 + + ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, + nverts, + nattrs, + verts) + + ctx.flush() + + zsbuf = zsbuf_tex.get_surface() + + self.assert_rgba(zsbuf, x, y, w, h, expected_rgba, 4.0/256, 0.85) + + + def main(): dev = Device() @@ -304,11 +504,10 @@ def main(): PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_A1R5G5B5_UNORM, PIPE_FORMAT_A4R4G4B4_UNORM, - #PIPE_FORMAT_Z32_UNORM, - #PIPE_FORMAT_Z24S8_UNORM, - #PIPE_FORMAT_Z24X8_UNORM, - #PIPE_FORMAT_Z16_UNORM, - #PIPE_FORMAT_S8_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_Z16_UNORM, PIPE_FORMAT_A8_UNORM, PIPE_FORMAT_L8_UNORM, PIPE_FORMAT_YCBCR, @@ -347,7 +546,17 @@ def main(): for level in range(0, last_level + 1): zslice = 0 while zslice < depth >> level: - test = TextureTest( + if format in ( + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_Z16_UNORM, + ): + klass = TextureDepthSampleTest + else: + klass = TextureColorSampleTest + + test = klass( dev = dev, target = target, format = format, diff --git a/src/gallium/state_trackers/wgl/SConscript b/src/gallium/state_trackers/wgl/SConscript index 038a7a31b3..5bbcc7175f 100644 --- a/src/gallium/state_trackers/wgl/SConscript +++ b/src/gallium/state_trackers/wgl/SConscript @@ -26,8 +26,7 @@ if env['platform'] in ['windows']: 'shared/stw_device.c', 'shared/stw_framebuffer.c', 'shared/stw_pixelformat.c', - 'shared/stw_quirks.c', - 'shared/stw_arbextensionsstring.c', + 'shared/stw_extensionsstring.c', 'shared/stw_getprocaddress.c', 'shared/stw_arbpixelformat.c', 'shared/stw_tls.c', diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.c b/src/gallium/state_trackers/wgl/icd/stw_icd.c index faf7f2f410..347f40aa06 100644 --- a/src/gallium/state_trackers/wgl/icd/stw_icd.c +++ b/src/gallium/state_trackers/wgl/icd/stw_icd.c @@ -38,9 +38,6 @@ #define DBG 0 -static GLCLTPROCTABLE cpt; -static boolean cpt_initialized = FALSE; - BOOL APIENTRY DrvCopyContext( @@ -57,7 +54,15 @@ DrvCreateLayerContext( HDC hdc, INT iLayerPlane ) { - return stw_create_layer_context( hdc, iLayerPlane ); + DHGLRC r; + + r = stw_create_layer_context( hdc, iLayerPlane ); + + if (DBG) + debug_printf( "%s( %p, %i ) = %u\n", + __FUNCTION__, hdc, iLayerPlane, r ); + + return r; } DHGLRC APIENTRY @@ -71,7 +76,15 @@ BOOL APIENTRY DrvDeleteContext( DHGLRC dhglrc ) { - return stw_delete_context( dhglrc ); + BOOL r; + + r = stw_delete_context( dhglrc ); + + if (DBG) + debug_printf( "%s( %u ) = %u\n", + __FUNCTION__, dhglrc, r ); + + return r; } BOOL APIENTRY @@ -129,7 +142,7 @@ DrvGetProcAddress( r = stw_get_proc_address( lpszProc ); if (DBG) - debug_printf( "%s( \", __FUNCTION__%s\" ) = %p\n", lpszProc, r ); + debug_printf( "%s( \"%s\" ) = %p\n", __FUNCTION__, lpszProc, r ); return r; } @@ -165,351 +178,352 @@ DrvSetCallbackProcs( } -static void init_proc_table( GLCLTPROCTABLE *cpt ) -{ - GLDISPATCHTABLE *disp = &cpt->glDispatchTable; - - memset( cpt, 0, sizeof *cpt ); - cpt->cEntries = OPENGL_VERSION_110_ENTRIES; - -#define GPA_GL( NAME ) disp->NAME = gl##NAME - GPA_GL( NewList ); - GPA_GL( EndList ); - GPA_GL( CallList ); - GPA_GL( CallLists ); - GPA_GL( DeleteLists ); - GPA_GL( GenLists ); - GPA_GL( ListBase ); - GPA_GL( Begin ); - GPA_GL( Bitmap ); - GPA_GL( Color3b ); - GPA_GL( Color3bv ); - GPA_GL( Color3d ); - GPA_GL( Color3dv ); - GPA_GL( Color3f ); - GPA_GL( Color3fv ); - GPA_GL( Color3i ); - GPA_GL( Color3iv ); - GPA_GL( Color3s ); - GPA_GL( Color3sv ); - GPA_GL( Color3ub ); - GPA_GL( Color3ubv ); - GPA_GL( Color3ui ); - GPA_GL( Color3uiv ); - GPA_GL( Color3us ); - GPA_GL( Color3usv ); - GPA_GL( Color4b ); - GPA_GL( Color4bv ); - GPA_GL( Color4d ); - GPA_GL( Color4dv ); - GPA_GL( Color4f ); - GPA_GL( Color4fv ); - GPA_GL( Color4i ); - GPA_GL( Color4iv ); - GPA_GL( Color4s ); - GPA_GL( Color4sv ); - GPA_GL( Color4ub ); - GPA_GL( Color4ubv ); - GPA_GL( Color4ui ); - GPA_GL( Color4uiv ); - GPA_GL( Color4us ); - GPA_GL( Color4usv ); - GPA_GL( EdgeFlag ); - GPA_GL( EdgeFlagv ); - GPA_GL( End ); - GPA_GL( Indexd ); - GPA_GL( Indexdv ); - GPA_GL( Indexf ); - GPA_GL( Indexfv ); - GPA_GL( Indexi ); - GPA_GL( Indexiv ); - GPA_GL( Indexs ); - GPA_GL( Indexsv ); - GPA_GL( Normal3b ); - GPA_GL( Normal3bv ); - GPA_GL( Normal3d ); - GPA_GL( Normal3dv ); - GPA_GL( Normal3f ); - GPA_GL( Normal3fv ); - GPA_GL( Normal3i ); - GPA_GL( Normal3iv ); - GPA_GL( Normal3s ); - GPA_GL( Normal3sv ); - GPA_GL( RasterPos2d ); - GPA_GL( RasterPos2dv ); - GPA_GL( RasterPos2f ); - GPA_GL( RasterPos2fv ); - GPA_GL( RasterPos2i ); - GPA_GL( RasterPos2iv ); - GPA_GL( RasterPos2s ); - GPA_GL( RasterPos2sv ); - GPA_GL( RasterPos3d ); - GPA_GL( RasterPos3dv ); - GPA_GL( RasterPos3f ); - GPA_GL( RasterPos3fv ); - GPA_GL( RasterPos3i ); - GPA_GL( RasterPos3iv ); - GPA_GL( RasterPos3s ); - GPA_GL( RasterPos3sv ); - GPA_GL( RasterPos4d ); - GPA_GL( RasterPos4dv ); - GPA_GL( RasterPos4f ); - GPA_GL( RasterPos4fv ); - GPA_GL( RasterPos4i ); - GPA_GL( RasterPos4iv ); - GPA_GL( RasterPos4s ); - GPA_GL( RasterPos4sv ); - GPA_GL( Rectd ); - GPA_GL( Rectdv ); - GPA_GL( Rectf ); - GPA_GL( Rectfv ); - GPA_GL( Recti ); - GPA_GL( Rectiv ); - GPA_GL( Rects ); - GPA_GL( Rectsv ); - GPA_GL( TexCoord1d ); - GPA_GL( TexCoord1dv ); - GPA_GL( TexCoord1f ); - GPA_GL( TexCoord1fv ); - GPA_GL( TexCoord1i ); - GPA_GL( TexCoord1iv ); - GPA_GL( TexCoord1s ); - GPA_GL( TexCoord1sv ); - GPA_GL( TexCoord2d ); - GPA_GL( TexCoord2dv ); - GPA_GL( TexCoord2f ); - GPA_GL( TexCoord2fv ); - GPA_GL( TexCoord2i ); - GPA_GL( TexCoord2iv ); - GPA_GL( TexCoord2s ); - GPA_GL( TexCoord2sv ); - GPA_GL( TexCoord3d ); - GPA_GL( TexCoord3dv ); - GPA_GL( TexCoord3f ); - GPA_GL( TexCoord3fv ); - GPA_GL( TexCoord3i ); - GPA_GL( TexCoord3iv ); - GPA_GL( TexCoord3s ); - GPA_GL( TexCoord3sv ); - GPA_GL( TexCoord4d ); - GPA_GL( TexCoord4dv ); - GPA_GL( TexCoord4f ); - GPA_GL( TexCoord4fv ); - GPA_GL( TexCoord4i ); - GPA_GL( TexCoord4iv ); - GPA_GL( TexCoord4s ); - GPA_GL( TexCoord4sv ); - GPA_GL( Vertex2d ); - GPA_GL( Vertex2dv ); - GPA_GL( Vertex2f ); - GPA_GL( Vertex2fv ); - GPA_GL( Vertex2i ); - GPA_GL( Vertex2iv ); - GPA_GL( Vertex2s ); - GPA_GL( Vertex2sv ); - GPA_GL( Vertex3d ); - GPA_GL( Vertex3dv ); - GPA_GL( Vertex3f ); - GPA_GL( Vertex3fv ); - GPA_GL( Vertex3i ); - GPA_GL( Vertex3iv ); - GPA_GL( Vertex3s ); - GPA_GL( Vertex3sv ); - GPA_GL( Vertex4d ); - GPA_GL( Vertex4dv ); - GPA_GL( Vertex4f ); - GPA_GL( Vertex4fv ); - GPA_GL( Vertex4i ); - GPA_GL( Vertex4iv ); - GPA_GL( Vertex4s ); - GPA_GL( Vertex4sv ); - GPA_GL( ClipPlane ); - GPA_GL( ColorMaterial ); - GPA_GL( CullFace ); - GPA_GL( Fogf ); - GPA_GL( Fogfv ); - GPA_GL( Fogi ); - GPA_GL( Fogiv ); - GPA_GL( FrontFace ); - GPA_GL( Hint ); - GPA_GL( Lightf ); - GPA_GL( Lightfv ); - GPA_GL( Lighti ); - GPA_GL( Lightiv ); - GPA_GL( LightModelf ); - GPA_GL( LightModelfv ); - GPA_GL( LightModeli ); - GPA_GL( LightModeliv ); - GPA_GL( LineStipple ); - GPA_GL( LineWidth ); - GPA_GL( Materialf ); - GPA_GL( Materialfv ); - GPA_GL( Materiali ); - GPA_GL( Materialiv ); - GPA_GL( PointSize ); - GPA_GL( PolygonMode ); - GPA_GL( PolygonStipple ); - GPA_GL( Scissor ); - GPA_GL( ShadeModel ); - GPA_GL( TexParameterf ); - GPA_GL( TexParameterfv ); - GPA_GL( TexParameteri ); - GPA_GL( TexParameteriv ); - GPA_GL( TexImage1D ); - GPA_GL( TexImage2D ); - GPA_GL( TexEnvf ); - GPA_GL( TexEnvfv ); - GPA_GL( TexEnvi ); - GPA_GL( TexEnviv ); - GPA_GL( TexGend ); - GPA_GL( TexGendv ); - GPA_GL( TexGenf ); - GPA_GL( TexGenfv ); - GPA_GL( TexGeni ); - GPA_GL( TexGeniv ); - GPA_GL( FeedbackBuffer ); - GPA_GL( SelectBuffer ); - GPA_GL( RenderMode ); - GPA_GL( InitNames ); - GPA_GL( LoadName ); - GPA_GL( PassThrough ); - GPA_GL( PopName ); - GPA_GL( PushName ); - GPA_GL( DrawBuffer ); - GPA_GL( Clear ); - GPA_GL( ClearAccum ); - GPA_GL( ClearIndex ); - GPA_GL( ClearColor ); - GPA_GL( ClearStencil ); - GPA_GL( ClearDepth ); - GPA_GL( StencilMask ); - GPA_GL( ColorMask ); - GPA_GL( DepthMask ); - GPA_GL( IndexMask ); - GPA_GL( Accum ); - GPA_GL( Disable ); - GPA_GL( Enable ); - GPA_GL( Finish ); - GPA_GL( Flush ); - GPA_GL( PopAttrib ); - GPA_GL( PushAttrib ); - GPA_GL( Map1d ); - GPA_GL( Map1f ); - GPA_GL( Map2d ); - GPA_GL( Map2f ); - GPA_GL( MapGrid1d ); - GPA_GL( MapGrid1f ); - GPA_GL( MapGrid2d ); - GPA_GL( MapGrid2f ); - GPA_GL( EvalCoord1d ); - GPA_GL( EvalCoord1dv ); - GPA_GL( EvalCoord1f ); - GPA_GL( EvalCoord1fv ); - GPA_GL( EvalCoord2d ); - GPA_GL( EvalCoord2dv ); - GPA_GL( EvalCoord2f ); - GPA_GL( EvalCoord2fv ); - GPA_GL( EvalMesh1 ); - GPA_GL( EvalPoint1 ); - GPA_GL( EvalMesh2 ); - GPA_GL( EvalPoint2 ); - GPA_GL( AlphaFunc ); - GPA_GL( BlendFunc ); - GPA_GL( LogicOp ); - GPA_GL( StencilFunc ); - GPA_GL( StencilOp ); - GPA_GL( DepthFunc ); - GPA_GL( PixelZoom ); - GPA_GL( PixelTransferf ); - GPA_GL( PixelTransferi ); - GPA_GL( PixelStoref ); - GPA_GL( PixelStorei ); - GPA_GL( PixelMapfv ); - GPA_GL( PixelMapuiv ); - GPA_GL( PixelMapusv ); - GPA_GL( ReadBuffer ); - GPA_GL( CopyPixels ); - GPA_GL( ReadPixels ); - GPA_GL( DrawPixels ); - GPA_GL( GetBooleanv ); - GPA_GL( GetClipPlane ); - GPA_GL( GetDoublev ); - GPA_GL( GetError ); - GPA_GL( GetFloatv ); - GPA_GL( GetIntegerv ); - GPA_GL( GetLightfv ); - GPA_GL( GetLightiv ); - GPA_GL( GetMapdv ); - GPA_GL( GetMapfv ); - GPA_GL( GetMapiv ); - GPA_GL( GetMaterialfv ); - GPA_GL( GetMaterialiv ); - GPA_GL( GetPixelMapfv ); - GPA_GL( GetPixelMapuiv ); - GPA_GL( GetPixelMapusv ); - GPA_GL( GetPolygonStipple ); - GPA_GL( GetString ); - GPA_GL( GetTexEnvfv ); - GPA_GL( GetTexEnviv ); - GPA_GL( GetTexGendv ); - GPA_GL( GetTexGenfv ); - GPA_GL( GetTexGeniv ); - GPA_GL( GetTexImage ); - GPA_GL( GetTexParameterfv ); - GPA_GL( GetTexParameteriv ); - GPA_GL( GetTexLevelParameterfv ); - GPA_GL( GetTexLevelParameteriv ); - GPA_GL( IsEnabled ); - GPA_GL( IsList ); - GPA_GL( DepthRange ); - GPA_GL( Frustum ); - GPA_GL( LoadIdentity ); - GPA_GL( LoadMatrixf ); - GPA_GL( LoadMatrixd ); - GPA_GL( MatrixMode ); - GPA_GL( MultMatrixf ); - GPA_GL( MultMatrixd ); - GPA_GL( Ortho ); - GPA_GL( PopMatrix ); - GPA_GL( PushMatrix ); - GPA_GL( Rotated ); - GPA_GL( Rotatef ); - GPA_GL( Scaled ); - GPA_GL( Scalef ); - GPA_GL( Translated ); - GPA_GL( Translatef ); - GPA_GL( Viewport ); - GPA_GL( ArrayElement ); - GPA_GL( BindTexture ); - GPA_GL( ColorPointer ); - GPA_GL( DisableClientState ); - GPA_GL( DrawArrays ); - GPA_GL( DrawElements ); - GPA_GL( EdgeFlagPointer ); - GPA_GL( EnableClientState ); - GPA_GL( IndexPointer ); - GPA_GL( Indexub ); - GPA_GL( Indexubv ); - GPA_GL( InterleavedArrays ); - GPA_GL( NormalPointer ); - GPA_GL( PolygonOffset ); - GPA_GL( TexCoordPointer ); - GPA_GL( VertexPointer ); - GPA_GL( AreTexturesResident ); - GPA_GL( CopyTexImage1D ); - GPA_GL( CopyTexImage2D ); - GPA_GL( CopyTexSubImage1D ); - GPA_GL( CopyTexSubImage2D ); - GPA_GL( DeleteTextures ); - GPA_GL( GenTextures ); - GPA_GL( GetPointerv ); - GPA_GL( IsTexture ); - GPA_GL( PrioritizeTextures ); - GPA_GL( TexSubImage1D ); - GPA_GL( TexSubImage2D ); - GPA_GL( PopClientAttrib ); - GPA_GL( PushClientAttrib ); -} +/** + * Although WGL allows different dispatch entrypoints per context + */ +static const GLCLTPROCTABLE cpt = +{ + OPENGL_VERSION_110_ENTRIES, + { + &glNewList, + &glEndList, + &glCallList, + &glCallLists, + &glDeleteLists, + &glGenLists, + &glListBase, + &glBegin, + &glBitmap, + &glColor3b, + &glColor3bv, + &glColor3d, + &glColor3dv, + &glColor3f, + &glColor3fv, + &glColor3i, + &glColor3iv, + &glColor3s, + &glColor3sv, + &glColor3ub, + &glColor3ubv, + &glColor3ui, + &glColor3uiv, + &glColor3us, + &glColor3usv, + &glColor4b, + &glColor4bv, + &glColor4d, + &glColor4dv, + &glColor4f, + &glColor4fv, + &glColor4i, + &glColor4iv, + &glColor4s, + &glColor4sv, + &glColor4ub, + &glColor4ubv, + &glColor4ui, + &glColor4uiv, + &glColor4us, + &glColor4usv, + &glEdgeFlag, + &glEdgeFlagv, + &glEnd, + &glIndexd, + &glIndexdv, + &glIndexf, + &glIndexfv, + &glIndexi, + &glIndexiv, + &glIndexs, + &glIndexsv, + &glNormal3b, + &glNormal3bv, + &glNormal3d, + &glNormal3dv, + &glNormal3f, + &glNormal3fv, + &glNormal3i, + &glNormal3iv, + &glNormal3s, + &glNormal3sv, + &glRasterPos2d, + &glRasterPos2dv, + &glRasterPos2f, + &glRasterPos2fv, + &glRasterPos2i, + &glRasterPos2iv, + &glRasterPos2s, + &glRasterPos2sv, + &glRasterPos3d, + &glRasterPos3dv, + &glRasterPos3f, + &glRasterPos3fv, + &glRasterPos3i, + &glRasterPos3iv, + &glRasterPos3s, + &glRasterPos3sv, + &glRasterPos4d, + &glRasterPos4dv, + &glRasterPos4f, + &glRasterPos4fv, + &glRasterPos4i, + &glRasterPos4iv, + &glRasterPos4s, + &glRasterPos4sv, + &glRectd, + &glRectdv, + &glRectf, + &glRectfv, + &glRecti, + &glRectiv, + &glRects, + &glRectsv, + &glTexCoord1d, + &glTexCoord1dv, + &glTexCoord1f, + &glTexCoord1fv, + &glTexCoord1i, + &glTexCoord1iv, + &glTexCoord1s, + &glTexCoord1sv, + &glTexCoord2d, + &glTexCoord2dv, + &glTexCoord2f, + &glTexCoord2fv, + &glTexCoord2i, + &glTexCoord2iv, + &glTexCoord2s, + &glTexCoord2sv, + &glTexCoord3d, + &glTexCoord3dv, + &glTexCoord3f, + &glTexCoord3fv, + &glTexCoord3i, + &glTexCoord3iv, + &glTexCoord3s, + &glTexCoord3sv, + &glTexCoord4d, + &glTexCoord4dv, + &glTexCoord4f, + &glTexCoord4fv, + &glTexCoord4i, + &glTexCoord4iv, + &glTexCoord4s, + &glTexCoord4sv, + &glVertex2d, + &glVertex2dv, + &glVertex2f, + &glVertex2fv, + &glVertex2i, + &glVertex2iv, + &glVertex2s, + &glVertex2sv, + &glVertex3d, + &glVertex3dv, + &glVertex3f, + &glVertex3fv, + &glVertex3i, + &glVertex3iv, + &glVertex3s, + &glVertex3sv, + &glVertex4d, + &glVertex4dv, + &glVertex4f, + &glVertex4fv, + &glVertex4i, + &glVertex4iv, + &glVertex4s, + &glVertex4sv, + &glClipPlane, + &glColorMaterial, + &glCullFace, + &glFogf, + &glFogfv, + &glFogi, + &glFogiv, + &glFrontFace, + &glHint, + &glLightf, + &glLightfv, + &glLighti, + &glLightiv, + &glLightModelf, + &glLightModelfv, + &glLightModeli, + &glLightModeliv, + &glLineStipple, + &glLineWidth, + &glMaterialf, + &glMaterialfv, + &glMateriali, + &glMaterialiv, + &glPointSize, + &glPolygonMode, + &glPolygonStipple, + &glScissor, + &glShadeModel, + &glTexParameterf, + &glTexParameterfv, + &glTexParameteri, + &glTexParameteriv, + &glTexImage1D, + &glTexImage2D, + &glTexEnvf, + &glTexEnvfv, + &glTexEnvi, + &glTexEnviv, + &glTexGend, + &glTexGendv, + &glTexGenf, + &glTexGenfv, + &glTexGeni, + &glTexGeniv, + &glFeedbackBuffer, + &glSelectBuffer, + &glRenderMode, + &glInitNames, + &glLoadName, + &glPassThrough, + &glPopName, + &glPushName, + &glDrawBuffer, + &glClear, + &glClearAccum, + &glClearIndex, + &glClearColor, + &glClearStencil, + &glClearDepth, + &glStencilMask, + &glColorMask, + &glDepthMask, + &glIndexMask, + &glAccum, + &glDisable, + &glEnable, + &glFinish, + &glFlush, + &glPopAttrib, + &glPushAttrib, + &glMap1d, + &glMap1f, + &glMap2d, + &glMap2f, + &glMapGrid1d, + &glMapGrid1f, + &glMapGrid2d, + &glMapGrid2f, + &glEvalCoord1d, + &glEvalCoord1dv, + &glEvalCoord1f, + &glEvalCoord1fv, + &glEvalCoord2d, + &glEvalCoord2dv, + &glEvalCoord2f, + &glEvalCoord2fv, + &glEvalMesh1, + &glEvalPoint1, + &glEvalMesh2, + &glEvalPoint2, + &glAlphaFunc, + &glBlendFunc, + &glLogicOp, + &glStencilFunc, + &glStencilOp, + &glDepthFunc, + &glPixelZoom, + &glPixelTransferf, + &glPixelTransferi, + &glPixelStoref, + &glPixelStorei, + &glPixelMapfv, + &glPixelMapuiv, + &glPixelMapusv, + &glReadBuffer, + &glCopyPixels, + &glReadPixels, + &glDrawPixels, + &glGetBooleanv, + &glGetClipPlane, + &glGetDoublev, + &glGetError, + &glGetFloatv, + &glGetIntegerv, + &glGetLightfv, + &glGetLightiv, + &glGetMapdv, + &glGetMapfv, + &glGetMapiv, + &glGetMaterialfv, + &glGetMaterialiv, + &glGetPixelMapfv, + &glGetPixelMapuiv, + &glGetPixelMapusv, + &glGetPolygonStipple, + &glGetString, + &glGetTexEnvfv, + &glGetTexEnviv, + &glGetTexGendv, + &glGetTexGenfv, + &glGetTexGeniv, + &glGetTexImage, + &glGetTexParameterfv, + &glGetTexParameteriv, + &glGetTexLevelParameterfv, + &glGetTexLevelParameteriv, + &glIsEnabled, + &glIsList, + &glDepthRange, + &glFrustum, + &glLoadIdentity, + &glLoadMatrixf, + &glLoadMatrixd, + &glMatrixMode, + &glMultMatrixf, + &glMultMatrixd, + &glOrtho, + &glPopMatrix, + &glPushMatrix, + &glRotated, + &glRotatef, + &glScaled, + &glScalef, + &glTranslated, + &glTranslatef, + &glViewport, + &glArrayElement, + &glBindTexture, + &glColorPointer, + &glDisableClientState, + &glDrawArrays, + &glDrawElements, + &glEdgeFlagPointer, + &glEnableClientState, + &glIndexPointer, + &glIndexub, + &glIndexubv, + &glInterleavedArrays, + &glNormalPointer, + &glPolygonOffset, + &glTexCoordPointer, + &glVertexPointer, + &glAreTexturesResident, + &glCopyTexImage1D, + &glCopyTexImage2D, + &glCopyTexSubImage1D, + &glCopyTexSubImage2D, + &glDeleteTextures, + &glGenTextures, + &glGetPointerv, + &glIsTexture, + &glPrioritizeTextures, + &glTexSubImage1D, + &glTexSubImage2D, + &glPopClientAttrib, + &glPushClientAttrib + } +}; + PGLCLTPROCTABLE APIENTRY DrvSetContext( @@ -517,21 +531,16 @@ DrvSetContext( DHGLRC dhglrc, PFN_SETPROCTABLE pfnSetProcTable ) { - if (DBG) - debug_printf( "%s( 0x%p, %u, 0x%p )\n", - __FUNCTION__, hdc, dhglrc, pfnSetProcTable ); - - /* Although WGL allows different dispatch entrypoints per - */ - if (!cpt_initialized) { - init_proc_table( &cpt ); - cpt_initialized = TRUE; - } - + PGLCLTPROCTABLE r = (PGLCLTPROCTABLE)&cpt; + if (!stw_make_current( hdc, dhglrc )) - return NULL; + r = NULL; - return &cpt; + if (DBG) + debug_printf( "%s( 0x%p, %u, 0x%p ) = %p\n", + __FUNCTION__, hdc, dhglrc, pfnSetProcTable, r ); + + return r; } int APIENTRY @@ -571,7 +580,7 @@ DrvShareLists( if (DBG) debug_printf( "%s\n", __FUNCTION__ ); - return FALSE; + return stw_share_lists(dhglrc1, dhglrc2); } BOOL APIENTRY @@ -592,7 +601,7 @@ DrvSwapLayerBuffers( if (DBG) debug_printf( "%s\n", __FUNCTION__ ); - return FALSE; + return stw_swap_layer_buffers( hdc, fuPlanes ); } BOOL APIENTRY diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.h b/src/gallium/state_trackers/wgl/icd/stw_icd.h index 8e676fb5b7..cbc1a66548 100644 --- a/src/gallium/state_trackers/wgl/icd/stw_icd.h +++ b/src/gallium/state_trackers/wgl/icd/stw_icd.h @@ -25,8 +25,8 @@ * **************************************************************************/ -#ifndef DRV_H -#define DRV_H +#ifndef STW_ICD_H +#define STW_ICD_H #include <windows.h> @@ -486,4 +486,4 @@ BOOL APIENTRY DrvValidateVersion( ULONG ulVersion ); -#endif /* DRV_H */ +#endif /* STW_ICD_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c index f563635420..0e2d407699 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c +++ b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c @@ -25,75 +25,30 @@ * **************************************************************************/ +/** + * @file + * + * WGL_ARB_pixel_format extension implementation. + * + * @sa http://www.opengl.org/registry/specs/ARB/wgl_pixel_format.txt + */ + + #include <windows.h> +#define WGL_WGLEXT_PROTOTYPES + +#include <GL/gl.h> +#include <GL/wglext.h> + #include "pipe/p_compiler.h" #include "util/u_memory.h" #include "stw_public.h" #include "stw_pixelformat.h" -#include "stw_arbpixelformat.h" - -#define WGL_NUMBER_PIXEL_FORMATS_ARB 0x2000 -#define WGL_DRAW_TO_WINDOW_ARB 0x2001 -#define WGL_DRAW_TO_BITMAP_ARB 0x2002 -#define WGL_ACCELERATION_ARB 0x2003 -#define WGL_NEED_PALETTE_ARB 0x2004 -#define WGL_NEED_SYSTEM_PALETTE_ARB 0x2005 -#define WGL_SWAP_LAYER_BUFFERS_ARB 0x2006 -#define WGL_SWAP_METHOD_ARB 0x2007 -#define WGL_NUMBER_OVERLAYS_ARB 0x2008 -#define WGL_NUMBER_UNDERLAYS_ARB 0x2009 -#define WGL_TRANSPARENT_ARB 0x200A -#define WGL_TRANSPARENT_RED_VALUE_ARB 0x2037 -#define WGL_TRANSPARENT_GREEN_VALUE_ARB 0x2038 -#define WGL_TRANSPARENT_BLUE_VALUE_ARB 0x2039 -#define WGL_TRANSPARENT_ALPHA_VALUE_ARB 0x203A -#define WGL_TRANSPARENT_INDEX_VALUE_ARB 0x203B -#define WGL_SHARE_DEPTH_ARB 0x200C -#define WGL_SHARE_STENCIL_ARB 0x200D -#define WGL_SHARE_ACCUM_ARB 0x200E -#define WGL_SUPPORT_GDI_ARB 0x200F -#define WGL_SUPPORT_OPENGL_ARB 0x2010 -#define WGL_DOUBLE_BUFFER_ARB 0x2011 -#define WGL_STEREO_ARB 0x2012 -#define WGL_PIXEL_TYPE_ARB 0x2013 -#define WGL_COLOR_BITS_ARB 0x2014 -#define WGL_RED_BITS_ARB 0x2015 -#define WGL_RED_SHIFT_ARB 0x2016 -#define WGL_GREEN_BITS_ARB 0x2017 -#define WGL_GREEN_SHIFT_ARB 0x2018 -#define WGL_BLUE_BITS_ARB 0x2019 -#define WGL_BLUE_SHIFT_ARB 0x201A -#define WGL_ALPHA_BITS_ARB 0x201B -#define WGL_ALPHA_SHIFT_ARB 0x201C -#define WGL_ACCUM_BITS_ARB 0x201D -#define WGL_ACCUM_RED_BITS_ARB 0x201E -#define WGL_ACCUM_GREEN_BITS_ARB 0x201F -#define WGL_ACCUM_BLUE_BITS_ARB 0x2020 -#define WGL_ACCUM_ALPHA_BITS_ARB 0x2021 -#define WGL_DEPTH_BITS_ARB 0x2022 -#define WGL_STENCIL_BITS_ARB 0x2023 -#define WGL_AUX_BUFFERS_ARB 0x2024 - -#define WGL_NO_ACCELERATION_ARB 0x2025 -#define WGL_GENERIC_ACCELERATION_ARB 0x2026 -#define WGL_FULL_ACCELERATION_ARB 0x2027 - -#define WGL_SWAP_EXCHANGE_ARB 0x2028 -#define WGL_SWAP_COPY_ARB 0x2029 -#define WGL_SWAP_UNDEFINED_ARB 0x202A - -#define WGL_TYPE_RGBA_ARB 0x202B -#define WGL_TYPE_COLORINDEX_ARB 0x202C - -/* From arb_multisample: - */ -#define WGL_SAMPLE_BUFFERS_ARB 0x2041 -#define WGL_SAMPLES_ARB 0x2042 static boolean -query_attrib( +stw_query_attrib( int iPixelFormat, int iLayerPlane, int attrib, @@ -101,9 +56,9 @@ query_attrib( { uint count; uint index; - const struct pixelformat_info *pf; + const struct stw_pixelformat_info *pfi; - count = pixelformat_get_extended_count(); + count = stw_pixelformat_get_extended_count(); if (attrib == WGL_NUMBER_PIXEL_FORMATS_ARB) { *pvalue = (int) count; @@ -114,30 +69,27 @@ query_attrib( if (index >= count) return FALSE; - pf = pixelformat_get_info( index ); + pfi = stw_pixelformat_get_info( index ); switch (attrib) { case WGL_DRAW_TO_WINDOW_ARB: - *pvalue = TRUE; + *pvalue = pfi->pfd.dwFlags & PFD_DRAW_TO_WINDOW ? TRUE : FALSE; return TRUE; case WGL_DRAW_TO_BITMAP_ARB: - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_DRAW_TO_BITMAP ? TRUE : FALSE; return TRUE; case WGL_NEED_PALETTE_ARB: - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_NEED_PALETTE ? TRUE : FALSE; return TRUE; case WGL_NEED_SYSTEM_PALETTE_ARB: - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_NEED_SYSTEM_PALETTE ? TRUE : FALSE; return TRUE; case WGL_SWAP_METHOD_ARB: - if (pf->flags & PF_FLAG_DOUBLEBUFFER) - *pvalue = WGL_SWAP_COPY_ARB; - else - *pvalue = WGL_SWAP_UNDEFINED_ARB; + *pvalue = pfi->pfd.dwFlags & PFD_SWAP_COPY ? WGL_SWAP_COPY_ARB : WGL_SWAP_UNDEFINED_ARB; return TRUE; case WGL_SWAP_LAYER_BUFFERS_ARB: @@ -179,96 +131,108 @@ query_attrib( break; case WGL_SUPPORT_GDI_ARB: - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_SUPPORT_GDI ? TRUE : FALSE; break; case WGL_SUPPORT_OPENGL_ARB: - *pvalue = TRUE; + *pvalue = pfi->pfd.dwFlags & PFD_SUPPORT_OPENGL ? TRUE : FALSE; break; case WGL_DOUBLE_BUFFER_ARB: - if (pf->flags & PF_FLAG_DOUBLEBUFFER) - *pvalue = TRUE; - else - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_DOUBLEBUFFER ? TRUE : FALSE; break; case WGL_STEREO_ARB: - *pvalue = FALSE; + *pvalue = pfi->pfd.dwFlags & PFD_STEREO ? TRUE : FALSE; break; case WGL_PIXEL_TYPE_ARB: - *pvalue = WGL_TYPE_RGBA_ARB; + switch (pfi->pfd.iPixelType) { + case PFD_TYPE_RGBA: + *pvalue = WGL_TYPE_RGBA_ARB; + break; + case PFD_TYPE_COLORINDEX: + *pvalue = WGL_TYPE_COLORINDEX_ARB; + break; + default: + return FALSE; + } break; case WGL_COLOR_BITS_ARB: - *pvalue = (int) (pf->color.redbits + pf->color.greenbits + pf->color.bluebits); + *pvalue = pfi->pfd.cColorBits; break; case WGL_RED_BITS_ARB: - *pvalue = (int) pf->color.redbits; + *pvalue = pfi->pfd.cRedBits; break; case WGL_RED_SHIFT_ARB: - *pvalue = (int) pf->color.redshift; + *pvalue = pfi->pfd.cRedShift; break; case WGL_GREEN_BITS_ARB: - *pvalue = (int) pf->color.greenbits; + *pvalue = pfi->pfd.cGreenBits; break; case WGL_GREEN_SHIFT_ARB: - *pvalue = (int) pf->color.greenshift; + *pvalue = pfi->pfd.cGreenShift; break; case WGL_BLUE_BITS_ARB: - *pvalue = (int) pf->color.bluebits; + *pvalue = pfi->pfd.cBlueBits; break; case WGL_BLUE_SHIFT_ARB: - *pvalue = (int) pf->color.blueshift; + *pvalue = pfi->pfd.cBlueShift; break; case WGL_ALPHA_BITS_ARB: - *pvalue = (int) pf->alpha.alphabits; + *pvalue = pfi->pfd.cAlphaBits; break; case WGL_ALPHA_SHIFT_ARB: - *pvalue = (int) pf->alpha.alphashift; + *pvalue = pfi->pfd.cAlphaShift; break; case WGL_ACCUM_BITS_ARB: + *pvalue = pfi->pfd.cAccumBits; + break; + case WGL_ACCUM_RED_BITS_ARB: + *pvalue = pfi->pfd.cAccumRedBits; + break; + case WGL_ACCUM_GREEN_BITS_ARB: + *pvalue = pfi->pfd.cAccumGreenBits; + break; + case WGL_ACCUM_BLUE_BITS_ARB: + *pvalue = pfi->pfd.cAccumBlueBits; + break; + case WGL_ACCUM_ALPHA_BITS_ARB: - *pvalue = 0; + *pvalue = pfi->pfd.cAccumAlphaBits; break; case WGL_DEPTH_BITS_ARB: - *pvalue = (int) pf->depth.depthbits; + *pvalue = pfi->pfd.cDepthBits; break; case WGL_STENCIL_BITS_ARB: - *pvalue = (int) pf->depth.stencilbits; + *pvalue = pfi->pfd.cStencilBits; break; case WGL_AUX_BUFFERS_ARB: - *pvalue = 0; + *pvalue = pfi->pfd.cAuxBuffers; break; case WGL_SAMPLE_BUFFERS_ARB: - if (pf->flags & PF_FLAG_MULTISAMPLED) - *pvalue = stw_query_sample_buffers(); - else - *pvalue = 0; + *pvalue = pfi->numSampleBuffers; break; case WGL_SAMPLES_ARB: - if (pf->flags & PF_FLAG_MULTISAMPLED) - *pvalue = stw_query_samples(); - else - *pvalue = 0; + *pvalue = pfi->numSamples; break; default: @@ -285,7 +249,7 @@ struct attrib_match_info BOOL exact; }; -static struct attrib_match_info attrib_match[] = { +static const struct attrib_match_info attrib_match[] = { /* WGL_ARB_pixel_format */ { WGL_DRAW_TO_WINDOW_ARB, 0, TRUE }, @@ -324,7 +288,7 @@ static struct attrib_match_info attrib_match[] = { { WGL_SAMPLES_ARB, 2, FALSE } }; -struct pixelformat_score +struct stw_pixelformat_score { int points; uint index; @@ -332,13 +296,13 @@ struct pixelformat_score static BOOL score_pixelformats( - struct pixelformat_score *scores, + struct stw_pixelformat_score *scores, uint count, int attribute, int expected_value ) { uint i; - struct attrib_match_info *ami = NULL; + const struct attrib_match_info *ami = NULL; uint index; /* Find out if a given attribute should be considered for score calculation. @@ -358,7 +322,7 @@ score_pixelformats( for (index = 0; index < count; index++) { int actual_value; - if (!query_attrib( index + 1, 0, attribute, &actual_value )) + if (!stw_query_attrib( index + 1, 0, attribute, &actual_value )) return FALSE; if (ami->exact) { @@ -395,7 +359,7 @@ wglChoosePixelFormatARB( UINT *nNumFormats ) { uint count; - struct pixelformat_score *scores; + struct stw_pixelformat_score *scores; uint i; *nNumFormats = 0; @@ -405,8 +369,8 @@ wglChoosePixelFormatARB( * points for a mismatch when the match does not have to be exact. * Set a score to 0 if there is a mismatch for an exact match criteria. */ - count = pixelformat_get_extended_count(); - scores = (struct pixelformat_score *) MALLOC( count * sizeof( struct pixelformat_score ) ); + count = stw_pixelformat_get_extended_count(); + scores = (struct stw_pixelformat_score *) MALLOC( count * sizeof( struct stw_pixelformat_score ) ); if (scores == NULL) return FALSE; for (i = 0; i < count; i++) { @@ -446,7 +410,7 @@ wglChoosePixelFormatARB( swapped = FALSE; for (i = 1; i < n; i++) { if (scores[i - 1].points < scores[i].points) { - struct pixelformat_score score = scores[i - 1]; + struct stw_pixelformat_score score = scores[i - 1]; scores[i - 1] = scores[i]; scores[i] = score; @@ -489,7 +453,7 @@ wglGetPixelFormatAttribfvARB( for (i = 0; i < nAttributes; i++) { int value; - if (!query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &value )) + if (!stw_query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &value )) return FALSE; pfValues[i] = (FLOAT) value; } @@ -511,7 +475,7 @@ wglGetPixelFormatAttribivARB( (void) hdc; for (i = 0; i < nAttributes; i++) { - if (!query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &piValues[i] )) + if (!stw_query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &piValues[i] )) return FALSE; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h deleted file mode 100644 index a6c4259942..0000000000 --- a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h +++ /dev/null @@ -1,61 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef WGL_ARBPIXELFORMAT_H -#define WGL_ARBPIXELFORMAT_H - - -/* Extension functions for get_proc_address: - */ -WINGDIAPI BOOL APIENTRY -wglChoosePixelFormatARB( - HDC hdc, - const int *piAttribIList, - const FLOAT *pfAttribFList, - UINT nMaxFormats, - int *piFormats, - UINT *nNumFormats ); - -WINGDIAPI BOOL APIENTRY -wglGetPixelFormatAttribfvARB( - HDC hdc, - int iPixelFormat, - int iLayerPlane, - UINT nAttributes, - const int *piAttributes, - FLOAT *pfValues ); - -WINGDIAPI BOOL APIENTRY -wglGetPixelFormatAttribivARB( - HDC hdc, - int iPixelFormat, - int iLayerPlane, - UINT nAttributes, - const int *piAttributes, - int *piValues ); - -#endif /* WGL_ARBPIXELFORMAT_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c index f890225242..662b5fbcd2 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_context.c +++ b/src/gallium/state_trackers/wgl/shared/stw_context.c @@ -59,11 +59,12 @@ stw_copy_context( pipe_mutex_lock( stw_dev->mutex ); - src = stw_lookup_context( hglrcSrc ); - dst = stw_lookup_context( hglrcDst ); + src = stw_lookup_context_locked( hglrcSrc ); + dst = stw_lookup_context_locked( hglrcDst ); if (src && dst) { /* FIXME */ + assert(0); (void) src; (void) dst; (void) mask; @@ -74,58 +75,61 @@ stw_copy_context( return ret; } +BOOL +stw_share_lists( + UINT_PTR hglrc1, + UINT_PTR hglrc2 ) +{ + struct stw_context *ctx1; + struct stw_context *ctx2; + BOOL ret = FALSE; + + pipe_mutex_lock( stw_dev->mutex ); + + ctx1 = stw_lookup_context_locked( hglrc1 ); + ctx2 = stw_lookup_context_locked( hglrc2 ); + + if (ctx1 && ctx2 && + ctx1->iPixelFormat == ctx2->iPixelFormat) { + ret = _mesa_share_state(ctx2->st->ctx, ctx1->st->ctx); + } + + pipe_mutex_unlock( stw_dev->mutex ); + + return ret; +} + UINT_PTR stw_create_layer_context( HDC hdc, int iLayerPlane ) { - uint pfi; - const struct pixelformat_info *pf = NULL; + int iPixelFormat; + const struct stw_pixelformat_info *pfi; + GLvisual visual; struct stw_context *ctx = NULL; - GLvisual *visual = NULL; struct pipe_screen *screen = NULL; struct pipe_context *pipe = NULL; - UINT_PTR hglrc = 0; - + if(!stw_dev) return 0; if (iLayerPlane != 0) return 0; - pfi = stw_pixelformat_get( hdc ); - if (pfi == 0) + iPixelFormat = GetPixelFormat(hdc); + if(!iPixelFormat) return 0; - - pf = pixelformat_get_info( pfi - 1 ); - + + pfi = stw_pixelformat_get_info( iPixelFormat - 1 ); + stw_pixelformat_visual(&visual, pfi); + ctx = CALLOC_STRUCT( stw_context ); if (ctx == NULL) - return 0; + goto no_ctx; ctx->hdc = hdc; - ctx->color_bits = GetDeviceCaps( ctx->hdc, BITSPIXEL ); - - /* Create visual based on flags - */ - visual = _mesa_create_visual( - GL_TRUE, - (pf->flags & PF_FLAG_DOUBLEBUFFER) ? GL_TRUE : GL_FALSE, - GL_FALSE, - pf->color.redbits, - pf->color.greenbits, - pf->color.bluebits, - pf->alpha.alphabits, - 0, - pf->depth.depthbits, - pf->depth.stencilbits, - 0, - 0, - 0, - 0, - (pf->flags & PF_FLAG_MULTISAMPLED) ? stw_query_samples() : 0 ); - if (visual == NULL) - goto fail; + ctx->iPixelFormat = iPixelFormat; screen = stw_dev->screen; @@ -137,7 +141,7 @@ stw_create_layer_context( pipe = stw_dev->stw_winsys->create_context( screen ); if (pipe == NULL) - goto fail; + goto no_pipe; #ifdef DEBUG /* Wrap context */ @@ -145,34 +149,32 @@ stw_create_layer_context( pipe = trace_context_create(stw_dev->screen, pipe); #endif + /* pass to stw_flush_frontbuffer as context_private */ assert(!pipe->priv); pipe->priv = hdc; - ctx->st = st_create_context( pipe, visual, NULL ); + ctx->st = st_create_context( pipe, &visual, NULL ); if (ctx->st == NULL) - goto fail; + goto no_st_ctx; ctx->st->ctx->DriverCtx = ctx; pipe_mutex_lock( stw_dev->mutex ); - { - hglrc = handle_table_add(stw_dev->ctx_table, ctx); - } + ctx->hglrc = handle_table_add(stw_dev->ctx_table, ctx); pipe_mutex_unlock( stw_dev->mutex ); - - /* Success? - */ - if (hglrc != 0) - return hglrc; - -fail: - if (visual) - _mesa_destroy_visual( visual ); - - if (pipe) - pipe->destroy( pipe ); - - FREE( ctx ); + if (!ctx->hglrc) + goto no_hglrc; + + return ctx->hglrc; + +no_hglrc: + st_destroy_context(ctx->st); + goto no_pipe; /* st_context_destroy already destroys pipe */ +no_st_ctx: + pipe->destroy( pipe ); +no_pipe: + FREE(ctx); +no_ctx: return 0; } @@ -187,38 +189,24 @@ stw_delete_context( return FALSE; pipe_mutex_lock( stw_dev->mutex ); + ctx = stw_lookup_context_locked(hglrc); + handle_table_remove(stw_dev->ctx_table, hglrc); + pipe_mutex_unlock( stw_dev->mutex ); - ctx = stw_lookup_context(hglrc); if (ctx) { GLcontext *glctx = ctx->st->ctx; GET_CURRENT_CONTEXT( glcurctx ); - struct stw_framebuffer *fb; - /* Unbind current if deleting current context. - */ + /* Unbind current if deleting current context. */ if (glcurctx == glctx) st_make_current( NULL, NULL, NULL ); - fb = framebuffer_from_hdc( ctx->hdc ); - if (fb) - framebuffer_destroy( fb ); - - if (WindowFromDC( ctx->hdc ) != NULL) - ReleaseDC( WindowFromDC( ctx->hdc ), ctx->hdc ); - - pipe_mutex_lock(stw_dev->mutex); - { - st_destroy_context(ctx->st); - FREE(ctx); - handle_table_remove(stw_dev->ctx_table, hglrc); - } - pipe_mutex_unlock(stw_dev->mutex); + st_destroy_context(ctx->st); + FREE(ctx); ret = TRUE; } - pipe_mutex_unlock( stw_dev->mutex ); - return ret; } @@ -226,62 +214,69 @@ BOOL stw_release_context( UINT_PTR hglrc ) { - BOOL ret = FALSE; + struct stw_context *ctx; if (!stw_dev) - return ret; + return FALSE; pipe_mutex_lock( stw_dev->mutex ); - { - struct stw_context *ctx; - - /* XXX: The expectation is that ctx is the same context which is - * current for this thread. We should check that and return False - * if not the case. - */ - ctx = stw_lookup_context( hglrc ); - if (ctx == NULL) - goto done; + ctx = stw_lookup_context_locked( hglrc ); + pipe_mutex_unlock( stw_dev->mutex ); - if (stw_make_current( NULL, 0 ) == FALSE) - goto done; + if (!ctx) + return FALSE; + + /* The expectation is that ctx is the same context which is + * current for this thread. We should check that and return False + * if not the case. + */ + { + GLcontext *glctx = ctx->st->ctx; + GET_CURRENT_CONTEXT( glcurctx ); - ret = TRUE; + if (glcurctx != glctx) + return FALSE; } -done: - pipe_mutex_unlock( stw_dev->mutex ); - - return ret; -} -/* Find the width and height of the window named by hdc. - */ -static void -get_window_size( HDC hdc, GLuint *width, GLuint *height ) -{ - if (WindowFromDC( hdc )) { - RECT rect; + if (stw_make_current( NULL, 0 ) == FALSE) + return FALSE; - GetClientRect( WindowFromDC( hdc ), &rect ); - *width = rect.right - rect.left; - *height = rect.bottom - rect.top; - } - else { - *width = GetDeviceCaps( hdc, HORZRES ); - *height = GetDeviceCaps( hdc, VERTRES ); - } + return TRUE; } + UINT_PTR stw_get_current_context( void ) { - return stw_tls_get_data()->currentGLRC; + GET_CURRENT_CONTEXT( glcurctx ); + struct stw_context *ctx; + + if(!glcurctx) + return 0; + + ctx = (struct stw_context *)glcurctx->DriverCtx; + assert(ctx); + if(!ctx) + return 0; + + return ctx->hglrc; } HDC stw_get_current_dc( void ) { - return stw_tls_get_data()->currentDC; + GET_CURRENT_CONTEXT( glcurctx ); + struct stw_context *ctx; + + if(!glcurctx) + return NULL; + + ctx = (struct stw_context *)glcurctx->DriverCtx; + assert(ctx); + if(!ctx) + return NULL; + + return ctx->hdc; } BOOL @@ -292,64 +287,68 @@ stw_make_current( struct stw_context *ctx; GET_CURRENT_CONTEXT( glcurctx ); struct stw_framebuffer *fb; - GLuint width = 0; - GLuint height = 0; - struct stw_context *curctx; if (!stw_dev) - return FALSE; - - pipe_mutex_lock( stw_dev->mutex ); - ctx = stw_lookup_context( hglrc ); - pipe_mutex_unlock( stw_dev->mutex ); - - stw_tls_get_data()->currentDC = hdc; - stw_tls_get_data()->currentGLRC = hglrc; + goto fail; if (glcurctx != NULL) { + struct stw_context *curctx; curctx = (struct stw_context *) glcurctx->DriverCtx; - if (curctx != ctx) + if (curctx->hglrc != hglrc) st_flush(glcurctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + + /* Return if already current. */ + if (curctx->hglrc == hglrc && curctx->hdc == hdc) + return TRUE; } if (hdc == NULL || hglrc == 0) { - st_make_current( NULL, NULL, NULL ); - return TRUE; + return st_make_current( NULL, NULL, NULL ); } - /* Return if already current. - */ - if (glcurctx != NULL) { - if (curctx != NULL && curctx == ctx && ctx->hdc == hdc) - return TRUE; - } + pipe_mutex_lock( stw_dev->mutex ); - fb = framebuffer_from_hdc( hdc ); + ctx = stw_lookup_context_locked( hglrc ); + if(!ctx) + goto fail; - if (hdc != NULL) - get_window_size( hdc, &width, &height ); + fb = stw_framebuffer_from_hdc_locked( hdc ); + if(!fb) { + /* Applications should call SetPixelFormat before creating a context, + * but not all do, and the opengl32 runtime seems to use a default pixel + * format in some cases, so we must create a framebuffer for those here + */ + int iPixelFormat = GetPixelFormat(hdc); + if(iPixelFormat) + fb = stw_framebuffer_create_locked( hdc, iPixelFormat ); + if(!fb) + goto fail; + } + + pipe_mutex_unlock( stw_dev->mutex ); - /* Lazy creation of framebuffers. - */ - if (fb == NULL && ctx != NULL && hdc != NULL) { - GLvisual *visual = &ctx->st->ctx->Visual; + if(fb->iPixelFormat != ctx->iPixelFormat) + goto fail; - fb = framebuffer_create( hdc, visual, width, height ); - if (fb == NULL) - return FALSE; - } + /* Lazy allocation of the frame buffer */ + if(!stw_framebuffer_allocate(fb)) + goto fail; - if (ctx && fb) { - st_make_current( ctx->st, fb->stfb, fb->stfb ); - framebuffer_resize( fb, width, height ); - ctx->hdc = hdc; - ctx->st->pipe->priv = hdc; - } - else { - /* Detach */ - st_make_current( NULL, NULL, NULL ); - } + /* Bind the new framebuffer */ + ctx->hdc = hdc; + + /* pass to stw_flush_frontbuffer as context_private */ + ctx->st->pipe->priv = hdc; + + if(!st_make_current( ctx->st, fb->stfb, fb->stfb )) + goto fail; + stw_framebuffer_resize(fb); + return TRUE; + +fail: + st_make_current( NULL, NULL, NULL ); + return FALSE; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.h b/src/gallium/state_trackers/wgl/shared/stw_context.h index b289615272..166471de5e 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_context.h +++ b/src/gallium/state_trackers/wgl/shared/stw_context.h @@ -35,8 +35,9 @@ struct st_context; struct stw_context { struct st_context *st; + UINT_PTR hglrc; + int iPixelFormat; HDC hdc; - DWORD color_bits; }; #endif /* STW_CONTEXT_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.c b/src/gallium/state_trackers/wgl/shared/stw_device.c index 51936c2bdd..1a6b29807d 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.c +++ b/src/gallium/state_trackers/wgl/shared/stw_device.c @@ -41,6 +41,7 @@ #include "shared/stw_pixelformat.h" #include "shared/stw_public.h" #include "shared/stw_tls.h" +#include "shared/stw_framebuffer.h" #ifdef WIN32_THREADS extern _glthread_Mutex OneTimeLock; @@ -56,7 +57,7 @@ struct stw_device *stw_dev = NULL; * stw_winsys::flush_front_buffer. */ static void -st_flush_frontbuffer(struct pipe_screen *screen, +stw_flush_frontbuffer(struct pipe_screen *screen, struct pipe_surface *surface, void *context_private ) { @@ -75,7 +76,7 @@ st_flush_frontbuffer(struct pipe_screen *screen, boolean -st_init(const struct stw_winsys *stw_winsys) +stw_init(const struct stw_winsys *stw_winsys) { static struct stw_device stw_dev_storage; struct pipe_screen *screen; @@ -110,7 +111,7 @@ st_init(const struct stw_winsys *stw_winsys) stw_dev->screen = screen; #endif - stw_dev->screen->flush_frontbuffer = st_flush_frontbuffer; + stw_dev->screen->flush_frontbuffer = &stw_flush_frontbuffer; pipe_mutex_init( stw_dev->mutex ); @@ -119,7 +120,7 @@ st_init(const struct stw_winsys *stw_winsys) goto error1; } - pixelformat_init(); + stw_pixelformat_init(); return TRUE; @@ -130,25 +131,28 @@ error1: boolean -st_init_thread(void) +stw_init_thread(void) { - if (!stw_tls_init_thread()) { + if (!stw_tls_init_thread()) + return FALSE; + + if (!stw_framebuffer_init_thread()) return FALSE; - } return TRUE; } void -st_cleanup_thread(void) +stw_cleanup_thread(void) { + stw_framebuffer_cleanup_thread(); stw_tls_cleanup_thread(); } void -st_cleanup(void) +stw_cleanup(void) { unsigned i; @@ -169,6 +173,8 @@ st_cleanup(void) } pipe_mutex_unlock( stw_dev->mutex ); + stw_framebuffer_cleanup(); + pipe_mutex_destroy( stw_dev->mutex ); stw_dev->screen->destroy(stw_dev->screen); @@ -189,7 +195,7 @@ st_cleanup(void) struct stw_context * -stw_lookup_context( UINT_PTR dhglrc ) +stw_lookup_context_locked( UINT_PTR dhglrc ) { if (dhglrc == 0) return NULL; diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.h b/src/gallium/state_trackers/wgl/shared/stw_device.h index 703cb67081..e097f1f71e 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.h +++ b/src/gallium/state_trackers/wgl/shared/stw_device.h @@ -29,12 +29,19 @@ #define STW_DEVICE_H_ +#include <windows.h> + #include "pipe/p_compiler.h" #include "pipe/p_thread.h" #include "util/u_handle_table.h" +#include "stw_pixelformat.h" + + +#define STW_MAX_PIXELFORMATS 256 struct pipe_screen; +struct stw_framebuffer; struct stw_device { @@ -45,18 +52,24 @@ struct stw_device #ifdef DEBUG boolean trace_running; #endif - + + struct stw_pixelformat_info pixelformats[STW_MAX_PIXELFORMATS]; + unsigned pixelformat_count; + unsigned pixelformat_extended_count; + pipe_mutex mutex; struct handle_table *ctx_table; + struct stw_framebuffer *fb_head; + #ifdef DEBUG unsigned long memdbg_no; #endif }; struct stw_context * -stw_lookup_context( UINT_PTR hglrc ); +stw_lookup_context_locked( UINT_PTR hglrc ); extern struct stw_device *stw_dev; diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c index b3934cb464..2660c591f9 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c +++ b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c @@ -1,5 +1,6 @@ /************************************************************************** * + * Copyright 2009 VMware, Inc. * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * @@ -27,7 +28,18 @@ #include <windows.h> -#include "stw_arbextensionsstring.h" +#define WGL_WGLEXT_PROTOTYPES + +#include <GL/gl.h> +#include <GL/wglext.h> + + +static const char *stw_extension_string = + "WGL_ARB_extensions_string " + "WGL_ARB_multisample " + "WGL_ARB_pixel_format " + "WGL_EXT_extensions_string"; + WINGDIAPI const char * APIENTRY wglGetExtensionsStringARB( @@ -35,8 +47,12 @@ wglGetExtensionsStringARB( { (void) hdc; - return - "WGL_ARB_extensions_string " - "WGL_ARB_multisample " - "WGL_ARB_pixel_format"; + return stw_extension_string; +} + + +WINGDIAPI const char * APIENTRY +wglGetExtensionsStringEXT( void ) +{ + return stw_extension_string; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c index c96c4b8dfa..58f1830319 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c +++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c @@ -42,208 +42,329 @@ #include "stw_device.h" #include "stw_public.h" #include "stw_winsys.h" +#include "stw_tls.h" -void -framebuffer_resize( - struct stw_framebuffer *fb, - GLuint width, - GLuint height ) +static INLINE struct stw_framebuffer * +stw_framebuffer_from_hwnd_locked( + HWND hwnd ) +{ + struct stw_framebuffer *fb; + + for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next) + if (fb->hWnd == hwnd) + break; + + return fb; +} + + +static INLINE void +stw_framebuffer_destroy_locked( + struct stw_framebuffer *fb ) { - st_resize_framebuffer( fb->stfb, width, height ); + struct stw_framebuffer **link; + + link = &stw_dev->fb_head; + while (*link != fb) + link = &(*link)->next; + assert(*link); + *link = fb->next; + fb->next = NULL; + + st_unreference_framebuffer(fb->stfb); + + pipe_mutex_destroy( fb->mutex ); + + FREE( fb ); } -static struct stw_framebuffer *fb_head = NULL; +/** + * @sa http://msdn.microsoft.com/en-us/library/ms644975(VS.85).aspx + * @sa http://msdn.microsoft.com/en-us/library/ms644960(VS.85).aspx + */ static LRESULT CALLBACK -window_proc( - HWND hWnd, - UINT uMsg, +stw_call_window_proc( + int nCode, WPARAM wParam, LPARAM lParam ) { - struct stw_framebuffer *fb; - - for (fb = fb_head; fb != NULL; fb = fb->next) - if (fb->hWnd == hWnd) - break; - assert( fb != NULL ); + struct stw_tls_data *tls_data; + PCWPSTRUCT pParams = (PCWPSTRUCT)lParam; + + tls_data = stw_tls_get_data(); + if(!tls_data) + return 0; + + if (nCode < 0) + return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam); + + if (pParams->message == WM_SIZE && pParams->wParam != SIZE_MINIMIZED) { + struct stw_framebuffer *fb; + + pipe_mutex_lock( stw_dev->mutex ); + fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd ); + pipe_mutex_unlock( stw_dev->mutex ); + + if(fb) { + unsigned width = LOWORD( pParams->lParam ); + unsigned height = HIWORD( pParams->lParam ); + + /* FIXME: The mesa statetracker makes the assumptions that only + * one context is using the framebuffer, and that that context is the + * current one. However neither holds true, as WGL allows more than + * one context to be bound to the same drawable, and this function can + * be called from any thread. + */ + pipe_mutex_lock( fb->mutex ); + if (fb->stfb) + st_resize_framebuffer( fb->stfb, width, height ); + pipe_mutex_unlock( fb->mutex ); + } + } - if (uMsg == WM_SIZE && wParam != SIZE_MINIMIZED) - framebuffer_resize( fb, LOWORD( lParam ), HIWORD( lParam ) ); + if (pParams->message == WM_DESTROY) { + struct stw_framebuffer *fb; - return CallWindowProc( fb->WndProc, hWnd, uMsg, wParam, lParam ); -} + pipe_mutex_lock( stw_dev->mutex ); + + fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd ); + if(fb) + stw_framebuffer_destroy_locked(fb); + + pipe_mutex_unlock( stw_dev->mutex ); + } -static INLINE boolean -stw_is_supported_color(enum pipe_format format) -{ - struct pipe_screen *screen = stw_dev->screen; - return screen->is_format_supported(screen, format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET, 0); + return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam); } -static INLINE boolean -stw_is_supported_depth_stencil(enum pipe_format format) -{ - struct pipe_screen *screen = stw_dev->screen; - return screen->is_format_supported(screen, format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); -} -/* Create a new framebuffer object which will correspond to the given HDC. +/** + * Create a new framebuffer object which will correspond to the given HDC. */ struct stw_framebuffer * -framebuffer_create( +stw_framebuffer_create_locked( HDC hdc, - GLvisual *visual, - GLuint width, - GLuint height ) + int iPixelFormat ) { + HWND hWnd; struct stw_framebuffer *fb; - enum pipe_format colorFormat, depthFormat, stencilFormat; - - /* Determine PIPE_FORMATs for buffers. - */ + const struct stw_pixelformat_info *pfi; - if(visual->alphaBits <= 0 && visual->redBits <= 5 && visual->blueBits <= 6 && visual->greenBits <= 5 && - stw_is_supported_color(PIPE_FORMAT_R5G6B5_UNORM)) { - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - } - else if(visual->alphaBits <= 0 && visual->redBits <= 8 && visual->blueBits <= 8 && visual->greenBits <= 8 && - stw_is_supported_color(PIPE_FORMAT_X8R8G8B8_UNORM)) { - colorFormat = PIPE_FORMAT_X8R8G8B8_UNORM; - } - else if(visual->alphaBits <= 1 && visual->redBits <= 5 && visual->blueBits <= 5 && visual->greenBits <= 5 && - stw_is_supported_color(PIPE_FORMAT_A1R5G5B5_UNORM)) { - colorFormat = PIPE_FORMAT_A1R5G5B5_UNORM; - } - else if(visual->alphaBits <= 4 && visual->redBits <= 4 && visual->blueBits <= 4 && visual->greenBits <= 4 && - stw_is_supported_color(PIPE_FORMAT_A4R4G4B4_UNORM)) { - colorFormat = PIPE_FORMAT_A4R4G4B4_UNORM; - } - else if(visual->alphaBits <= 8 && visual->redBits <= 8 && visual->blueBits <= 8 && visual->greenBits <= 8 && - stw_is_supported_color(PIPE_FORMAT_A8R8G8B8_UNORM)) { - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - } - else { - assert(0); + /* We only support drawing to a window. */ + hWnd = WindowFromDC( hdc ); + if(!hWnd) + return NULL; + + fb = CALLOC_STRUCT( stw_framebuffer ); + if (fb == NULL) return NULL; - } - if (visual->depthBits == 0) - depthFormat = PIPE_FORMAT_NONE; - else if (visual->depthBits <= 16 && - stw_is_supported_depth_stencil(PIPE_FORMAT_Z16_UNORM)) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits <= 24 && visual->stencilBits != 8 && - stw_is_supported_depth_stencil(PIPE_FORMAT_X8Z24_UNORM)) { - depthFormat = PIPE_FORMAT_X8Z24_UNORM; - } - else if (visual->depthBits <= 24 && visual->stencilBits != 8 && - stw_is_supported_depth_stencil(PIPE_FORMAT_Z24X8_UNORM)) { - depthFormat = PIPE_FORMAT_Z24X8_UNORM; - } - else if (visual->depthBits <= 24 && visual->stencilBits == 8 && - stw_is_supported_depth_stencil(PIPE_FORMAT_S8Z24_UNORM)) { - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - } - else if (visual->depthBits <= 24 && visual->stencilBits == 8 && - stw_is_supported_depth_stencil(PIPE_FORMAT_Z24S8_UNORM)) { - depthFormat = PIPE_FORMAT_Z24S8_UNORM; - } - else if(stw_is_supported_depth_stencil(PIPE_FORMAT_Z32_UNORM)) { - depthFormat = PIPE_FORMAT_Z32_UNORM; - } - else if(stw_is_supported_depth_stencil(PIPE_FORMAT_Z32_FLOAT)) { - depthFormat = PIPE_FORMAT_Z32_FLOAT; - } - else { - assert(0); - depthFormat = PIPE_FORMAT_NONE; - } + fb->hDC = hdc; + fb->hWnd = hWnd; + fb->iPixelFormat = iPixelFormat; - if (depthFormat == PIPE_FORMAT_S8Z24_UNORM || - depthFormat == PIPE_FORMAT_Z24S8_UNORM) { - stencilFormat = depthFormat; - } - else if (visual->stencilBits == 8 && - stw_is_supported_depth_stencil(PIPE_FORMAT_S8_UNORM)) { - stencilFormat = PIPE_FORMAT_S8_UNORM; + fb->pfi = pfi = stw_pixelformat_get_info( iPixelFormat - 1 ); + + stw_pixelformat_visual(&fb->visual, pfi); + + pipe_mutex_init( fb->mutex ); + + fb->next = stw_dev->fb_head; + stw_dev->fb_head = fb; + + return fb; +} + + +static void +stw_framebuffer_get_size( struct stw_framebuffer *fb, GLuint *pwidth, GLuint *pheight ) +{ + GLuint width, height; + + if (fb->hWnd) { + RECT rect; + GetClientRect( fb->hWnd, &rect ); + width = rect.right - rect.left; + height = rect.bottom - rect.top; } else { - stencilFormat = PIPE_FORMAT_NONE; + width = GetDeviceCaps( fb->hDC, HORZRES ); + height = GetDeviceCaps( fb->hDC, VERTRES ); } - fb = CALLOC_STRUCT( stw_framebuffer ); - if (fb == NULL) - return NULL; + if(width < 1) + width = 1; + if(height < 1) + height = 1; - fb->stfb = st_create_framebuffer( - visual, - colorFormat, - depthFormat, - stencilFormat, - width, - height, - (void *) fb ); + *pwidth = width; + *pheight = height; +} - fb->cColorBits = GetDeviceCaps( hdc, BITSPIXEL ); - fb->hDC = hdc; - /* Subclass a window associated with the device context. - */ - fb->hWnd = WindowFromDC( hdc ); - if (fb->hWnd != NULL) { - fb->WndProc = (WNDPROC) SetWindowLongPtr( - fb->hWnd, - GWLP_WNDPROC, - (LONG_PTR) window_proc ); +BOOL +stw_framebuffer_allocate( + struct stw_framebuffer *fb) +{ + pipe_mutex_lock( fb->mutex ); + + if(!fb->stfb) { + const struct stw_pixelformat_info *pfi = fb->pfi; + enum pipe_format colorFormat, depthFormat, stencilFormat; + GLuint width, height; + + colorFormat = pfi->color_format; + + assert(pf_layout( pfi->depth_stencil_format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); + + if(pf_get_component_bits( pfi->depth_stencil_format, PIPE_FORMAT_COMP_Z )) + depthFormat = pfi->depth_stencil_format; + else + depthFormat = PIPE_FORMAT_NONE; + + if(pf_get_component_bits( pfi->depth_stencil_format, PIPE_FORMAT_COMP_S )) + stencilFormat = pfi->depth_stencil_format; + else + stencilFormat = PIPE_FORMAT_NONE; + + stw_framebuffer_get_size(fb, &width, &height); + + fb->stfb = st_create_framebuffer( + &fb->visual, + colorFormat, + depthFormat, + stencilFormat, + width, + height, + (void *) fb ); } + + pipe_mutex_unlock( fb->mutex ); - fb->next = fb_head; - fb_head = fb; - return fb; + return fb->stfb ? TRUE : FALSE; } + void -framebuffer_destroy( - struct stw_framebuffer *fb ) +stw_framebuffer_resize( + struct stw_framebuffer *fb) { - struct stw_framebuffer **link = &fb_head; - struct stw_framebuffer *pfb = fb_head; - - while (pfb != NULL) { - if (pfb == fb) { - if (fb->hWnd != NULL) { - SetWindowLongPtr( - fb->hWnd, - GWLP_WNDPROC, - (LONG_PTR) fb->WndProc ); - } - - *link = fb->next; - FREE( fb ); - return; - } + GLuint width, height; + assert(fb->stfb); + stw_framebuffer_get_size(fb, &width, &height); + st_resize_framebuffer(fb->stfb, width, height); +} + + +void +stw_framebuffer_cleanup( void ) +{ + struct stw_framebuffer *fb; + struct stw_framebuffer *next; + + pipe_mutex_lock( stw_dev->mutex ); - link = &pfb->next; - pfb = pfb->next; + fb = stw_dev->fb_head; + while (fb) { + next = fb->next; + stw_framebuffer_destroy_locked(fb); + fb = next; } + stw_dev->fb_head = NULL; + + pipe_mutex_unlock( stw_dev->mutex ); } -/* Given an hdc, return the corresponding stw_framebuffer. + +/** + * Given an hdc, return the corresponding stw_framebuffer. */ struct stw_framebuffer * -framebuffer_from_hdc( +stw_framebuffer_from_hdc_locked( HDC hdc ) { struct stw_framebuffer *fb; - for (fb = fb_head; fb != NULL; fb = fb->next) + for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next) if (fb->hDC == hdc) - return fb; - return NULL; + break; + + return fb; +} + + +/** + * Given an hdc, return the corresponding stw_framebuffer. + */ +struct stw_framebuffer * +stw_framebuffer_from_hdc( + HDC hdc ) +{ + struct stw_framebuffer *fb; + + pipe_mutex_lock( stw_dev->mutex ); + fb = stw_framebuffer_from_hdc_locked(hdc); + pipe_mutex_unlock( stw_dev->mutex ); + + return fb; +} + + +BOOL +stw_pixelformat_set( + HDC hdc, + int iPixelFormat ) +{ + uint count; + uint index; + struct stw_framebuffer *fb; + + index = (uint) iPixelFormat - 1; + count = stw_pixelformat_get_extended_count(); + if (index >= count) + return FALSE; + + pipe_mutex_lock( stw_dev->mutex ); + + fb = stw_framebuffer_from_hdc_locked(hdc); + if(fb) { + /* SetPixelFormat must be called only once */ + pipe_mutex_unlock( stw_dev->mutex ); + return FALSE; + } + + fb = stw_framebuffer_create_locked(hdc, iPixelFormat); + if(!fb) { + pipe_mutex_unlock( stw_dev->mutex ); + return FALSE; + } + + pipe_mutex_unlock( stw_dev->mutex ); + + /* Some applications mistakenly use the undocumented wglSetPixelFormat + * function instead of SetPixelFormat, so we call SetPixelFormat here to + * avoid opengl32.dll's wglCreateContext to fail */ + if (GetPixelFormat(hdc) == 0) { + SetPixelFormat(hdc, iPixelFormat, NULL); + } + + return TRUE; +} + + +int +stw_pixelformat_get( + HDC hdc ) +{ + struct stw_framebuffer *fb; + + fb = stw_framebuffer_from_hdc(hdc); + if(!fb) + return 0; + + return fb->iPixelFormat; } @@ -255,10 +376,15 @@ stw_swap_buffers( struct pipe_screen *screen; struct pipe_surface *surface; - fb = framebuffer_from_hdc( hdc ); + fb = stw_framebuffer_from_hdc( hdc ); if (fb == NULL) return FALSE; + if (!(fb->pfi->pfd.dwFlags & PFD_DOUBLEBUFFER)) + return TRUE; + + pipe_mutex_lock( fb->mutex ); + /* If we're swapping the buffer associated with the current context * we have to flush any pending rendering commands first. */ @@ -266,9 +392,11 @@ stw_swap_buffers( screen = stw_dev->screen; - if(!st_get_framebuffer_surface( fb->stfb, ST_SURFACE_BACK_LEFT, &surface )) + if(!st_get_framebuffer_surface( fb->stfb, ST_SURFACE_BACK_LEFT, &surface )) { /* FIXME: this shouldn't happen, but does on glean */ + pipe_mutex_unlock( fb->mutex ); return FALSE; + } #ifdef DEBUG if(stw_dev->trace_running) { @@ -279,5 +407,54 @@ stw_swap_buffers( stw_dev->stw_winsys->flush_frontbuffer( screen, surface, hdc ); + pipe_mutex_unlock( fb->mutex ); + return TRUE; } + + +BOOL +stw_swap_layer_buffers( + HDC hdc, + UINT fuPlanes ) +{ + if(fuPlanes & WGL_SWAP_MAIN_PLANE) + return stw_swap_buffers(hdc); + + return FALSE; +} + + +boolean +stw_framebuffer_init_thread(void) +{ + struct stw_tls_data *tls_data; + + tls_data = stw_tls_get_data(); + if(!tls_data) + return FALSE; + + tls_data->hCallWndProcHook = SetWindowsHookEx(WH_CALLWNDPROC, + stw_call_window_proc, + NULL, + GetCurrentThreadId()); + if(tls_data->hCallWndProcHook == NULL) + return FALSE; + + return TRUE; +} + +void +stw_framebuffer_cleanup_thread(void) +{ + struct stw_tls_data *tls_data; + + tls_data = stw_tls_get_data(); + if(!tls_data) + return; + + if(tls_data->hCallWndProcHook) { + UnhookWindowsHookEx(tls_data->hCallWndProcHook); + tls_data->hCallWndProcHook = NULL; + } +} diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h index 5abdf18997..e7fa51c3a8 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h +++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h @@ -28,39 +28,61 @@ #ifndef STW_FRAMEBUFFER_H #define STW_FRAMEBUFFER_H +#include <windows.h> + #include "main/mtypes.h" -/* Windows framebuffer, derived from gl_framebuffer. +#include "pipe/p_thread.h" + +struct stw_pixelformat_info; + +/** + * Windows framebuffer, derived from gl_framebuffer. */ struct stw_framebuffer { - struct st_framebuffer *stfb; HDC hDC; - BYTE cColorBits; HWND hWnd; - WNDPROC WndProc; + + int iPixelFormat; + const struct stw_pixelformat_info *pfi; + GLvisual visual; + + pipe_mutex mutex; + struct st_framebuffer *stfb; + + /** This is protected by stw_device::mutex, not the mutex above */ struct stw_framebuffer *next; }; struct stw_framebuffer * -framebuffer_create( +stw_framebuffer_create_locked( HDC hdc, - GLvisual *visual, - GLuint width, - GLuint height ); + int iPixelFormat ); -void -framebuffer_destroy( +BOOL +stw_framebuffer_allocate( struct stw_framebuffer *fb ); void -framebuffer_resize( - struct stw_framebuffer *fb, - GLuint width, - GLuint height ); +stw_framebuffer_resize( + struct stw_framebuffer *fb); + +void +stw_framebuffer_cleanup(void); + +struct stw_framebuffer * +stw_framebuffer_from_hdc_locked( + HDC hdc ); struct stw_framebuffer * -framebuffer_from_hdc( +stw_framebuffer_from_hdc( HDC hdc ); +boolean +stw_framebuffer_init_thread(void); + +void +stw_framebuffer_cleanup_thread(void); + #endif /* STW_FRAMEBUFFER_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c index ac2d6fc260..4070cbd5c0 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c +++ b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c @@ -27,28 +27,34 @@ #include <windows.h> +#define WGL_WGLEXT_PROTOTYPES + +#include <GL/gl.h> +#include <GL/wglext.h> + #include "glapi/glapi.h" -#include "stw_arbextensionsstring.h" -#include "stw_arbpixelformat.h" #include "stw_public.h" -struct extension_entry +struct stw_extension_entry { const char *name; PROC proc; }; -#define EXTENTRY(P) { #P, (PROC) P } +#define STW_EXTENSION_ENTRY(P) { #P, (PROC) P } -static struct extension_entry extension_entries[] = { +static const struct stw_extension_entry stw_extension_entries[] = { /* WGL_ARB_extensions_string */ - EXTENTRY( wglGetExtensionsStringARB ), + STW_EXTENSION_ENTRY( wglGetExtensionsStringARB ), /* WGL_ARB_pixel_format */ - EXTENTRY( wglChoosePixelFormatARB ), - EXTENTRY( wglGetPixelFormatAttribfvARB ), - EXTENTRY( wglGetPixelFormatAttribivARB ), + STW_EXTENSION_ENTRY( wglChoosePixelFormatARB ), + STW_EXTENSION_ENTRY( wglGetPixelFormatAttribfvARB ), + STW_EXTENSION_ENTRY( wglGetPixelFormatAttribivARB ), + + /* WGL_EXT_extensions_string */ + STW_EXTENSION_ENTRY( wglGetExtensionsStringEXT ), { NULL, NULL } }; @@ -57,13 +63,13 @@ PROC stw_get_proc_address( LPCSTR lpszProc ) { - struct extension_entry *entry; + const struct stw_extension_entry *entry; - PROC p = (PROC) _glapi_get_proc_address( (const char *) lpszProc ); + PROC p = (PROC) _glapi_get_proc_address( lpszProc ); if (p) return p; - for (entry = extension_entries; entry->name; entry++) + for (entry = stw_extension_entries; entry->name; entry++) if (strcmp( lpszProc, entry->name ) == 0) return entry->proc; diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c index b216ca5c82..c296744838 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c +++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c @@ -25,113 +25,266 @@ * **************************************************************************/ +#include "main/mtypes.h" +#include "main/context.h" + +#include "pipe/p_format.h" +#include "pipe/p_defines.h" +#include "pipe/p_screen.h" + #include "util/u_debug.h" + +#include "stw_device.h" #include "stw_pixelformat.h" #include "stw_public.h" #include "stw_tls.h" -#define MAX_PIXELFORMATS 16 -static struct pixelformat_info pixelformats[MAX_PIXELFORMATS]; -static uint pixelformat_count = 0; -static uint pixelformat_extended_count = 0; +struct stw_pf_color_info +{ + enum pipe_format format; + struct { + unsigned char red; + unsigned char green; + unsigned char blue; + unsigned char alpha; + } bits; + struct { + unsigned char red; + unsigned char green; + unsigned char blue; + unsigned char alpha; + } shift; +}; + +struct stw_pf_depth_info +{ + enum pipe_format format; + struct { + unsigned char depth; + unsigned char stencil; + } bits; +}; + + +/* NOTE: order matters, since in otherwise equal circumstances the first + * format listed will get chosen */ + +static const struct stw_pf_color_info +stw_pf_color[] = { + /* no-alpha */ + { PIPE_FORMAT_X8R8G8B8_UNORM, { 8, 8, 8, 0}, {16, 8, 0, 0} }, + { PIPE_FORMAT_B8G8R8X8_UNORM, { 8, 8, 8, 0}, { 8, 16, 24, 0} }, + { PIPE_FORMAT_R5G6B5_UNORM, { 5, 6, 5, 0}, {11, 5, 0, 0} }, + /* alpha */ + { PIPE_FORMAT_A8R8G8B8_UNORM, { 8, 8, 8, 8}, {16, 8, 0, 24} }, + { PIPE_FORMAT_B8G8R8A8_UNORM, { 8, 8, 8, 8}, { 8, 16, 24, 0} }, +#if 0 + { PIPE_FORMAT_A2B10G10R10_UNORM, {10, 10, 10, 2}, { 0, 10, 20, 30} }, +#endif + { PIPE_FORMAT_A1R5G5B5_UNORM, { 5, 5, 5, 1}, {10, 5, 0, 15} }, + { PIPE_FORMAT_A4R4G4B4_UNORM, { 4, 4, 4, 4}, {16, 4, 0, 12} } +}; + + +static const struct stw_pf_depth_info +stw_pf_depth_stencil[] = { + /* pure depth */ + { PIPE_FORMAT_Z32_UNORM, {32, 0} }, + { PIPE_FORMAT_Z24X8_UNORM, {24, 0} }, + { PIPE_FORMAT_X8Z24_UNORM, {24, 0} }, + { PIPE_FORMAT_Z16_UNORM, {16, 0} }, + /* pure stencil */ + { PIPE_FORMAT_S8_UNORM, { 0, 8} }, + /* combined depth-stencil */ + { PIPE_FORMAT_S8Z24_UNORM, {24, 8} }, + { PIPE_FORMAT_Z24S8_UNORM, {24, 8} } +}; + + +static const boolean +stw_pf_doublebuffer[] = { + FALSE, + TRUE, +}; + + +const unsigned +stw_pf_multisample[] = { + 0, + 4 +}; static void -add_standard_pixelformats( - struct pixelformat_info **ppf, - uint flags ) +stw_pixelformat_add( + struct stw_device *stw_dev, + const struct stw_pf_color_info *color, + const struct stw_pf_depth_info *depth, + unsigned accum, + boolean doublebuffer, + unsigned samples ) { - struct pixelformat_info *pf = *ppf; - struct pixelformat_color_info color24 = { 8, 0, 8, 8, 8, 16 }; - struct pixelformat_alpha_info alpha8 = { 8, 24 }; - struct pixelformat_alpha_info noalpha = { 0, 0 }; - struct pixelformat_depth_info depth24s8 = { 24, 8 }; - struct pixelformat_depth_info depth16 = { 16, 0 }; - - pf->flags = PF_FLAG_DOUBLEBUFFER | flags; - pf->color = color24; - pf->alpha = alpha8; - pf->depth = depth16; - pf++; - - pf->flags = PF_FLAG_DOUBLEBUFFER | flags; - pf->color = color24; - pf->alpha = alpha8; - pf->depth = depth24s8; - pf++; - - pf->flags = PF_FLAG_DOUBLEBUFFER | flags; - pf->color = color24; - pf->alpha = noalpha; - pf->depth = depth16; - pf++; - - pf->flags = PF_FLAG_DOUBLEBUFFER | flags; - pf->color = color24; - pf->alpha = noalpha; - pf->depth = depth24s8; - pf++; - - pf->flags = flags; - pf->color = color24; - pf->alpha = alpha8; - pf->depth = depth16; - pf++; - - pf->flags = flags; - pf->color = color24; - pf->alpha = alpha8; - pf->depth = depth24s8; - pf++; - - pf->flags = flags; - pf->color = color24; - pf->alpha = noalpha; - pf->depth = depth16; - pf++; - - pf->flags = flags; - pf->color = color24; - pf->alpha = noalpha; - pf->depth = depth24s8; - pf++; - - *ppf = pf; + boolean extended = FALSE; + struct stw_pixelformat_info *pfi; + + assert(stw_dev->pixelformat_extended_count < STW_MAX_PIXELFORMATS); + if(stw_dev->pixelformat_extended_count >= STW_MAX_PIXELFORMATS) + return; + + assert(pf_layout( color->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); + assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_R ) == color->bits.red ); + assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_G ) == color->bits.green ); + assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_B ) == color->bits.blue ); + assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_A ) == color->bits.alpha ); + assert(pf_layout( depth->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); + assert(pf_get_component_bits( depth->format, PIPE_FORMAT_COMP_Z ) == depth->bits.depth ); + assert(pf_get_component_bits( depth->format, PIPE_FORMAT_COMP_S ) == depth->bits.stencil ); + + pfi = &stw_dev->pixelformats[stw_dev->pixelformat_extended_count]; + + memset(pfi, 0, sizeof *pfi); + + pfi->color_format = color->format; + pfi->depth_stencil_format = depth->format; + + pfi->pfd.nSize = sizeof pfi->pfd; + pfi->pfd.nVersion = 1; + + pfi->pfd.dwFlags = PFD_SUPPORT_OPENGL; + + /* TODO: also support non-native pixel formats */ + pfi->pfd.dwFlags |= PFD_DRAW_TO_WINDOW ; + + if (doublebuffer) + pfi->pfd.dwFlags |= PFD_DOUBLEBUFFER | PFD_SWAP_COPY; + + pfi->pfd.iPixelType = PFD_TYPE_RGBA; + + pfi->pfd.cColorBits = color->bits.red + color->bits.green + color->bits.blue + color->bits.alpha; + pfi->pfd.cRedBits = color->bits.red; + pfi->pfd.cRedShift = color->shift.red; + pfi->pfd.cGreenBits = color->bits.green; + pfi->pfd.cGreenShift = color->shift.green; + pfi->pfd.cBlueBits = color->bits.blue; + pfi->pfd.cBlueShift = color->shift.blue; + pfi->pfd.cAlphaBits = color->bits.alpha; + pfi->pfd.cAlphaShift = color->shift.alpha; + pfi->pfd.cAccumBits = 4*accum; + pfi->pfd.cAccumRedBits = accum; + pfi->pfd.cAccumGreenBits = accum; + pfi->pfd.cAccumBlueBits = accum; + pfi->pfd.cAccumAlphaBits = accum; + pfi->pfd.cDepthBits = depth->bits.depth; + pfi->pfd.cStencilBits = depth->bits.stencil; + pfi->pfd.cAuxBuffers = 0; + pfi->pfd.iLayerType = 0; + pfi->pfd.bReserved = 0; + pfi->pfd.dwLayerMask = 0; + pfi->pfd.dwVisibleMask = 0; + pfi->pfd.dwDamageMask = 0; + + if(samples) { + pfi->numSampleBuffers = 1; + pfi->numSamples = samples; + extended = TRUE; + } + + ++stw_dev->pixelformat_extended_count; + + if(!extended) { + ++stw_dev->pixelformat_count; + assert(stw_dev->pixelformat_count == stw_dev->pixelformat_extended_count); + } } void -pixelformat_init( void ) +stw_pixelformat_init( void ) { - struct pixelformat_info *pf = pixelformats; - - add_standard_pixelformats( &pf, 0 ); - pixelformat_count = pf - pixelformats; - - add_standard_pixelformats( &pf, PF_FLAG_MULTISAMPLED ); - pixelformat_extended_count = pf - pixelformats; + struct pipe_screen *screen = stw_dev->screen; + unsigned i, j, k, l; + + assert( !stw_dev->pixelformat_count ); + assert( !stw_dev->pixelformat_extended_count ); + + for(i = 0; i < Elements(stw_pf_multisample); ++i) { + unsigned samples = stw_pf_multisample[i]; + + /* FIXME: re-enabled MSAA when we can query it */ + if(samples) + continue; - assert( pixelformat_extended_count <= MAX_PIXELFORMATS ); + for(j = 0; j < Elements(stw_pf_color); ++j) { + const struct stw_pf_color_info *color = &stw_pf_color[j]; + + if(!screen->is_format_supported(screen, color->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) + continue; + + for(k = 0; k < Elements(stw_pf_doublebuffer); ++k) { + unsigned doublebuffer = stw_pf_doublebuffer[k]; + + for(l = 0; l < Elements(stw_pf_depth_stencil); ++l) { + const struct stw_pf_depth_info *depth = &stw_pf_depth_stencil[l]; + + if(!screen->is_format_supported(screen, depth->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) + continue; + + stw_pixelformat_add( stw_dev, color, depth, 0, doublebuffer, samples ); + stw_pixelformat_add( stw_dev, color, depth, 16, doublebuffer, samples ); + } + } + } + } + + assert( stw_dev->pixelformat_count <= stw_dev->pixelformat_extended_count ); + assert( stw_dev->pixelformat_extended_count <= STW_MAX_PIXELFORMATS ); } uint -pixelformat_get_count( void ) +stw_pixelformat_get_count( void ) { - return pixelformat_count; + return stw_dev->pixelformat_count; } uint -pixelformat_get_extended_count( void ) +stw_pixelformat_get_extended_count( void ) { - return pixelformat_extended_count; + return stw_dev->pixelformat_extended_count; } -const struct pixelformat_info * -pixelformat_get_info( uint index ) +const struct stw_pixelformat_info * +stw_pixelformat_get_info( uint index ) { - assert( index < pixelformat_extended_count ); + assert( index < stw_dev->pixelformat_extended_count ); + + return &stw_dev->pixelformats[index]; +} + - return &pixelformats[index]; +void +stw_pixelformat_visual(GLvisual *visual, + const struct stw_pixelformat_info *pfi ) +{ + memset(visual, 0, sizeof *visual); + _mesa_initialize_visual( + visual, + (pfi->pfd.iPixelType == PFD_TYPE_RGBA) ? GL_TRUE : GL_FALSE, + (pfi->pfd.dwFlags & PFD_DOUBLEBUFFER) ? GL_TRUE : GL_FALSE, + (pfi->pfd.dwFlags & PFD_STEREO) ? GL_TRUE : GL_FALSE, + pfi->pfd.cRedBits, + pfi->pfd.cGreenBits, + pfi->pfd.cBlueBits, + pfi->pfd.cAlphaBits, + (pfi->pfd.iPixelType == PFD_TYPE_COLORINDEX) ? pfi->pfd.cColorBits : 0, + pfi->pfd.cDepthBits, + pfi->pfd.cStencilBits, + pfi->pfd.cAccumRedBits, + pfi->pfd.cAccumGreenBits, + pfi->pfd.cAccumBlueBits, + pfi->pfd.cAccumAlphaBits, + pfi->numSamples ); } @@ -144,11 +297,11 @@ stw_pixelformat_describe( { uint count; uint index; - const struct pixelformat_info *pf; + const struct stw_pixelformat_info *pfi; (void) hdc; - count = pixelformat_get_extended_count(); + count = stw_pixelformat_get_extended_count(); index = (uint) iPixelFormat - 1; if (ppfd == NULL) @@ -156,36 +309,9 @@ stw_pixelformat_describe( if (index >= count || nBytes != sizeof( PIXELFORMATDESCRIPTOR )) return 0; - pf = pixelformat_get_info( index ); - - ppfd->nSize = sizeof( PIXELFORMATDESCRIPTOR ); - ppfd->nVersion = 1; - ppfd->dwFlags = PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL; - if (pf->flags & PF_FLAG_DOUBLEBUFFER) - ppfd->dwFlags |= PFD_DOUBLEBUFFER | PFD_SWAP_COPY; - ppfd->iPixelType = PFD_TYPE_RGBA; - ppfd->cColorBits = pf->color.redbits + pf->color.greenbits + pf->color.bluebits; - ppfd->cRedBits = pf->color.redbits; - ppfd->cRedShift = pf->color.redshift; - ppfd->cGreenBits = pf->color.greenbits; - ppfd->cGreenShift = pf->color.greenshift; - ppfd->cBlueBits = pf->color.bluebits; - ppfd->cBlueShift = pf->color.blueshift; - ppfd->cAlphaBits = pf->alpha.alphabits; - ppfd->cAlphaShift = pf->alpha.alphashift; - ppfd->cAccumBits = 0; - ppfd->cAccumRedBits = 0; - ppfd->cAccumGreenBits = 0; - ppfd->cAccumBlueBits = 0; - ppfd->cAccumAlphaBits = 0; - ppfd->cDepthBits = pf->depth.depthbits; - ppfd->cStencilBits = pf->depth.stencilbits; - ppfd->cAuxBuffers = 0; - ppfd->iLayerType = 0; - ppfd->bReserved = 0; - ppfd->dwLayerMask = 0; - ppfd->dwVisibleMask = 0; - ppfd->dwDamageMask = 0; + pfi = stw_pixelformat_get_info( index ); + + memcpy(ppfd, &pfi->pfd, sizeof( PIXELFORMATDESCRIPTOR )); return count; } @@ -203,29 +329,30 @@ int stw_pixelformat_choose( HDC hdc, (void) hdc; - count = pixelformat_get_count(); + count = stw_pixelformat_get_count(); bestindex = count; - bestdelta = 0xffffffff; + bestdelta = ~0U; for (index = 0; index < count; index++) { uint delta = 0; - const struct pixelformat_info *pf = pixelformat_get_info( index ); + const struct stw_pixelformat_info *pfi = stw_pixelformat_get_info( index ); if (!(ppfd->dwFlags & PFD_DOUBLEBUFFER_DONTCARE) && !!(ppfd->dwFlags & PFD_DOUBLEBUFFER) != - !!(pf->flags & PF_FLAG_DOUBLEBUFFER)) + !!(pfi->pfd.dwFlags & PFD_DOUBLEBUFFER)) continue; - if (ppfd->cColorBits != pf->color.redbits + pf->color.greenbits + pf->color.bluebits) + /* FIXME: Take in account individual channel bits */ + if (ppfd->cColorBits != pfi->pfd.cColorBits) delta += 8; - if (ppfd->cDepthBits != pf->depth.depthbits) + if (ppfd->cDepthBits != pfi->pfd.cDepthBits) delta += 4; - if (ppfd->cStencilBits != pf->depth.stencilbits) + if (ppfd->cStencilBits != pfi->pfd.cStencilBits) delta += 2; - if (ppfd->cAlphaBits != pf->alpha.alphabits) + if (ppfd->cAlphaBits != pfi->pfd.cAlphaBits) delta++; if (delta < bestdelta) { @@ -241,57 +368,3 @@ int stw_pixelformat_choose( HDC hdc, return bestindex + 1; } - - -int -stw_pixelformat_get( - HDC hdc ) -{ - return stw_tls_get_data()->currentPixelFormat; -} - - -BOOL -stw_pixelformat_set( - HDC hdc, - int iPixelFormat ) -{ - uint count; - uint index; - - (void) hdc; - - index = (uint) iPixelFormat - 1; - count = pixelformat_get_extended_count(); - if (index >= count) - return FALSE; - - stw_tls_get_data()->currentPixelFormat = iPixelFormat; - - /* Some applications mistakenly use the undocumented wglSetPixelFormat - * function instead of SetPixelFormat, so we call SetPixelFormat here to - * avoid opengl32.dll's wglCreateContext to fail */ - if (GetPixelFormat(hdc) == 0) { - SetPixelFormat(hdc, iPixelFormat, NULL); - } - - return TRUE; -} - - - -/* XXX: this needs to be turned into queries on pipe_screen or - * stw_winsys. - */ -int -stw_query_sample_buffers( void ) -{ - return 1; -} - -int -stw_query_samples( void ) -{ - return 4; -} - diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h index 7ca4194a2a..bec429231b 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h +++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h @@ -25,59 +25,41 @@ * **************************************************************************/ -#ifndef PIXELFORMAT_H -#define PIXELFORMAT_H +#ifndef STW_PIXELFORMAT_H +#define STW_PIXELFORMAT_H #include <windows.h> -#include "pipe/p_compiler.h" - -#define PF_FLAG_DOUBLEBUFFER 0x00000001 -#define PF_FLAG_MULTISAMPLED 0x00000002 - -struct pixelformat_color_info -{ - uint redbits; - uint redshift; - uint greenbits; - uint greenshift; - uint bluebits; - uint blueshift; -}; -struct pixelformat_alpha_info -{ - uint alphabits; - uint alphashift; -}; +#include "main/mtypes.h" -struct pixelformat_depth_info -{ - uint depthbits; - uint stencilbits; -}; +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" -struct pixelformat_info +struct stw_pixelformat_info { - uint flags; - struct pixelformat_color_info color; - struct pixelformat_alpha_info alpha; - struct pixelformat_depth_info depth; + enum pipe_format color_format; + enum pipe_format depth_stencil_format; + + PIXELFORMATDESCRIPTOR pfd; + + unsigned numSampleBuffers; + unsigned numSamples; }; void -pixelformat_init( void ); +stw_pixelformat_init( void ); uint -pixelformat_get_count( void ); +stw_pixelformat_get_count( void ); uint -pixelformat_get_extended_count( void ); - -const struct pixelformat_info * -pixelformat_get_info( uint index ); +stw_pixelformat_get_extended_count( void ); -int stw_query_sample_buffers( void ); -int stw_query_samples( void ); +const struct stw_pixelformat_info * +stw_pixelformat_get_info( uint index ); +void +stw_pixelformat_visual(GLvisual *visual, + const struct stw_pixelformat_info *pfi ); -#endif /* PIXELFORMAT_H */ +#endif /* STW_PIXELFORMAT_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_public.h b/src/gallium/state_trackers/wgl/shared/stw_public.h index 39d377c16b..7fe9cfb356 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_public.h +++ b/src/gallium/state_trackers/wgl/shared/stw_public.h @@ -37,6 +37,8 @@ BOOL stw_copy_context( UINT_PTR hglrcSrc, UINT_PTR stw_create_layer_context( HDC hdc, int iLayerPlane ); +BOOL stw_share_lists( UINT_PTR hglrc1, UINT_PTR hglrc2 ); + BOOL stw_delete_context( UINT_PTR hglrc ); BOOL @@ -50,6 +52,9 @@ BOOL stw_make_current( HDC hdc, UINT_PTR hglrc ); BOOL stw_swap_buffers( HDC hdc ); +BOOL +stw_swap_layer_buffers( HDC hdc, UINT fuPlanes ); + PROC stw_get_proc_address( LPCSTR lpszProc ); int stw_pixelformat_describe( HDC hdc, diff --git a/src/gallium/state_trackers/wgl/shared/stw_quirks.c b/src/gallium/state_trackers/wgl/shared/stw_quirks.c deleted file mode 100644 index 2f7091a52c..0000000000 --- a/src/gallium/state_trackers/wgl/shared/stw_quirks.c +++ /dev/null @@ -1,116 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * @file - * - * This is hopefully a temporary hack to define some needed dispatch - * table entries. Hopefully, I'll find a better solution. The - * dispatch table generation scripts ought to be making these dummy - * stubs as well. - */ - -void gl_dispatch_stub_543(void){} -void gl_dispatch_stub_544(void){} -void gl_dispatch_stub_545(void){} -void gl_dispatch_stub_546(void){} -void gl_dispatch_stub_547(void){} -void gl_dispatch_stub_548(void){} -void gl_dispatch_stub_549(void){} -void gl_dispatch_stub_550(void){} -void gl_dispatch_stub_551(void){} -void gl_dispatch_stub_552(void){} -void gl_dispatch_stub_553(void){} -void gl_dispatch_stub_554(void){} -void gl_dispatch_stub_555(void){} -void gl_dispatch_stub_556(void){} -void gl_dispatch_stub_557(void){} -void gl_dispatch_stub_558(void){} -void gl_dispatch_stub_559(void){} -void gl_dispatch_stub_560(void){} -void gl_dispatch_stub_561(void){} -void gl_dispatch_stub_565(void){} -void gl_dispatch_stub_566(void){} -void gl_dispatch_stub_570(void){} -void gl_dispatch_stub_577(void){} -void gl_dispatch_stub_578(void){} -void gl_dispatch_stub_582(void){} -void gl_dispatch_stub_603(void){} -void gl_dispatch_stub_607(void){} -void gl_dispatch_stub_645(void){} -void gl_dispatch_stub_646(void){} -void gl_dispatch_stub_647(void){} -void gl_dispatch_stub_648(void){} -void gl_dispatch_stub_649(void){} -void gl_dispatch_stub_650(void){} -void gl_dispatch_stub_651(void){} -void gl_dispatch_stub_652(void){} -void gl_dispatch_stub_653(void){} -void gl_dispatch_stub_657(void){} -void gl_dispatch_stub_733(void){} -void gl_dispatch_stub_734(void){} -void gl_dispatch_stub_735(void){} -void gl_dispatch_stub_736(void){} -void gl_dispatch_stub_737(void){} -void gl_dispatch_stub_738(void){} -void gl_dispatch_stub_744(void){} -void gl_dispatch_stub_745(void){} -void gl_dispatch_stub_746(void){} -void gl_dispatch_stub_760(void){} -void gl_dispatch_stub_761(void){} -void gl_dispatch_stub_763(void){} -void gl_dispatch_stub_764(void){} -void gl_dispatch_stub_765(void){} -void gl_dispatch_stub_766(void){} -void gl_dispatch_stub_767(void){} -void gl_dispatch_stub_768(void){} - -void gl_dispatch_stub_562(void){} -void gl_dispatch_stub_563(void){} -void gl_dispatch_stub_564(void){} -void gl_dispatch_stub_567(void){} -void gl_dispatch_stub_568(void){} -void gl_dispatch_stub_569(void){} -void gl_dispatch_stub_580(void){} -void gl_dispatch_stub_581(void){} -void gl_dispatch_stub_606(void){} -void gl_dispatch_stub_654(void){} -void gl_dispatch_stub_655(void){} -void gl_dispatch_stub_656(void){} -void gl_dispatch_stub_739(void){} -void gl_dispatch_stub_740(void){} -void gl_dispatch_stub_741(void){} -void gl_dispatch_stub_748(void){} -void gl_dispatch_stub_749(void){} -void gl_dispatch_stub_769(void){} -void gl_dispatch_stub_770(void){} -void gl_dispatch_stub_771(void){} -void gl_dispatch_stub_772(void){} -void gl_dispatch_stub_773(void){} -void gl_dispatch_stub_774(void){} -void gl_dispatch_stub_750(void){} -void gl_dispatch_stub_742(void){} diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.c b/src/gallium/state_trackers/wgl/shared/stw_tls.c index e72bafb880..0c18a52352 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_tls.c +++ b/src/gallium/state_trackers/wgl/shared/stw_tls.c @@ -44,6 +44,18 @@ stw_tls_init(void) return TRUE; } +static INLINE struct stw_tls_data * +stw_tls_data_create() +{ + struct stw_tls_data *data; + + data = CALLOC_STRUCT(stw_tls_data); + if (!data) + return NULL; + + return data; +} + boolean stw_tls_init_thread(void) { @@ -53,14 +65,9 @@ stw_tls_init_thread(void) return FALSE; } - data = MALLOC(sizeof(*data)); - if (!data) { + data = stw_tls_data_create(); + if(!data) return FALSE; - } - - data->currentPixelFormat = 0; - data->currentDC = NULL; - data->currentGLRC = 0; TlsSetValue(tlsIndex, data); @@ -93,9 +100,23 @@ stw_tls_cleanup(void) struct stw_tls_data * stw_tls_get_data(void) { + struct stw_tls_data *data; + if (tlsIndex == TLS_OUT_OF_INDEXES) { return NULL; } + + data = (struct stw_tls_data *) TlsGetValue(tlsIndex); + if(!data) { + /* DllMain is called with DLL_THREAD_ATTACH only by threads created after + * the DLL is loaded by the process */ + + data = stw_tls_data_create(); + if(!data) + return NULL; + + TlsSetValue(tlsIndex, data); + } - return (struct stw_tls_data *) TlsGetValue(tlsIndex); + return data; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.h b/src/gallium/state_trackers/wgl/shared/stw_tls.h index 23b61e68ff..6af8be70c9 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_tls.h +++ b/src/gallium/state_trackers/wgl/shared/stw_tls.h @@ -28,11 +28,11 @@ #ifndef STW_TLS_H #define STW_TLS_H +#include <windows.h> + struct stw_tls_data { - uint currentPixelFormat; - HDC currentDC; - UINT_PTR currentGLRC; + HHOOK hCallWndProcHook; }; boolean diff --git a/src/gallium/state_trackers/wgl/shared/stw_winsys.h b/src/gallium/state_trackers/wgl/shared/stw_winsys.h index e4a1d4f979..c0bf82c9ed 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_winsys.h +++ b/src/gallium/state_trackers/wgl/shared/stw_winsys.h @@ -51,15 +51,15 @@ struct stw_winsys }; boolean -st_init(const struct stw_winsys *stw_winsys); +stw_init(const struct stw_winsys *stw_winsys); boolean -st_init_thread(void); +stw_init_thread(void); void -st_cleanup_thread(void); +stw_cleanup_thread(void); void -st_cleanup(void); +stw_cleanup(void); #endif /* STW_WINSYS_H */ diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c index e06d2640b4..a131292f7a 100644 --- a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c +++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c @@ -100,10 +100,7 @@ wglSwapLayerBuffers( HDC hdc, UINT fuPlanes ) { - (void) hdc; - (void) fuPlanes; - - return FALSE; + return stw_swap_layer_buffers( hdc, fuPlanes ); } WINGDIAPI PROC APIENTRY @@ -189,12 +186,7 @@ wglShareLists( HGLRC hglrc1, HGLRC hglrc2 ) { - (void) hglrc1; - (void) hglrc2; - - assert( 0 ); - - return FALSE; + return stw_share_lists( (UINT_PTR)hglrc1, (UINT_PTR)hglrc2);; } WINGDIAPI BOOL APIENTRY diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index d04204e1bf..401bd39dac 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -28,6 +28,7 @@ * */ +#include "xorg-server.h" #include "xf86.h" #include "xf86_OSproc.h" @@ -85,7 +86,6 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) memset(&template, 0, sizeof(template)); template.target = PIPE_TEXTURE_2D; - template.compressed = 0; template.format = PIPE_FORMAT_S8Z24_UNORM; pf_get_block(template.format, &template.block); template.width[0] = pDraw->width; @@ -98,7 +98,6 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) struct pipe_texture template; memset(&template, 0, sizeof(template)); template.target = PIPE_TEXTURE_2D; - template.compressed = 0; template.format = PIPE_FORMAT_A8R8G8B8_UNORM; pf_get_block(template.format, &template.block); template.width[0] = pDraw->width; diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c index 8a2711e70c..45e831f0c2 100644 --- a/src/gallium/state_trackers/xorg/xorg_driver.c +++ b/src/gallium/state_trackers/xorg/xorg_driver.c @@ -476,7 +476,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv) } if (!ms->screen) { - ms->screen = drm_api_hooks.create_screen(ms->fd, ms->PciInfo->device_id); + ms->screen = drm_api_hooks.create_screen(ms->fd, NULL); if (!ms->screen) { FatalError("Could not init pipe_screen\n"); diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c index 56c8fdccb2..7913174354 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.c +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -28,6 +28,7 @@ * */ +#include "xorg-server.h" #include "xf86.h" #include "xorg_tracker.h" @@ -36,6 +37,8 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_rect.h" + struct exa_entity { ExaDriverPtr pExa; @@ -425,7 +428,6 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, memset(&template, 0, sizeof(template)); template.target = PIPE_TEXTURE_2D; - template.compressed = 0; exa_get_pipe_format(depth, &template.format, &bitsPerPixel); pf_get_block(template.format, &template.block); template.width[0] = width; @@ -436,6 +438,18 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, priv->tex = exa->scrn->texture_create(exa->scrn, &template); } + if (pPixData) { + struct pipe_transfer *transfer = + exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, + PIPE_TRANSFER_WRITE, + 0, 0, width, height); + pipe_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), + &priv->tex->block, transfer->stride, 0, 0, + width, height, pPixData, pPixmap->devKind, 0, 0); + exa->scrn->transfer_unmap(exa->scrn, transfer); + exa->scrn->tex_transfer_destroy(transfer); + } + return TRUE; } diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template index 9f92cb4207..985e5a861f 100644 --- a/src/gallium/winsys/drm/Makefile.template +++ b/src/gallium/winsys/drm/Makefile.template @@ -118,7 +118,7 @@ clean: install: $(LIBNAME) $(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR) + $(MINSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR) include depend diff --git a/src/gallium/winsys/drm/intel/SConscript b/src/gallium/winsys/drm/intel/SConscript new file mode 100644 index 0000000000..50d7b75ed6 --- /dev/null +++ b/src/gallium/winsys/drm/intel/SConscript @@ -0,0 +1,7 @@ +Import('*') + +SConscript(['gem/SConscript',]) + +if 'mesa' in env['statetrackers']: + + SConscript(['dri/SConscript']) diff --git a/src/gallium/winsys/drm/intel/dri2/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile index 1a02109274..de39e759d8 100644 --- a/src/gallium/winsys/drm/intel/dri2/Makefile +++ b/src/gallium/winsys/drm/intel/dri/Makefile @@ -4,8 +4,10 @@ include $(TOP)/configs/current LIBNAME = i915_dri.so PIPE_DRIVERS = \ - $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ + $(TOP)/src/gallium/state_trackers/dri/libdridrm.a \ $(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ $(TOP)/src/gallium/drivers/i915simple/libi915simple.a @@ -19,4 +21,6 @@ include ../../Makefile.template DRI_LIB_DEPS += -ldrm_intel -symlinks: +symlinks: $(TOP)/$(LIB_DIR)/gallium + @rm -f $(TOP)/$(LIB_DIR)/gallium/i965_dri.so + ln -s i915_dri.so $(TOP)/$(LIB_DIR)/gallium/i965_dri.so diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript new file mode 100644 index 0000000000..e14e96e32f --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/SConscript @@ -0,0 +1,16 @@ +Import('*') + +env = drienv.Clone() + +drivers = [ + softpipe, + i915simple, + trace, + inteldrm +] + +env.SharedLibrary( + target ='i915_dri.so', + source = COMMON_GALLIUM_SOURCES, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], +) diff --git a/src/gallium/winsys/drm/intel/gem/SConscript b/src/gallium/winsys/drm/intel/gem/SConscript new file mode 100644 index 0000000000..ea8a2e55f6 --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/SConscript @@ -0,0 +1,17 @@ +Import('*') + +env = drienv.Clone() + +inteldrm_sources = [ + 'intel_be_api.c', + 'intel_be_batchbuffer.c', + 'intel_be_context.c', + 'intel_be_device.c', +] + +inteldrm = env.ConvenienceLibrary( + target ='inteldrm', + source = inteldrm_sources, +) + +Export('inteldrm') diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c index f4ef7c2d88..a74be13bf7 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.c @@ -2,7 +2,13 @@ #include "intel_be_api.h" #include "i915simple/i915_winsys.h" +#ifdef DEBUG +#include "trace/trace_drm.h" + +struct drm_api hooks = +#else struct drm_api drm_api_hooks = +#endif { /* intel_be_context.c */ .create_context = intel_be_create_context, diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h index 73e458d4ba..1c622f3b97 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.h @@ -8,7 +8,8 @@ #include "intel_be_device.h" -struct pipe_screen *intel_be_create_screen(int drmFD, int pciID); +struct pipe_screen *intel_be_create_screen(int drmFD, + struct drm_create_screen_arg *arg); struct pipe_context *intel_be_create_context(struct pipe_screen *screen); #endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c index bb6f1b916c..fe0b138fbe 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c @@ -1,6 +1,8 @@ #include "pipe/p_screen.h" +#include "softpipe/sp_winsys.h" + #include "intel_be_device.h" #include "intel_be_context.h" #include "intel_be_batchbuffer.h" @@ -106,13 +108,13 @@ intel_be_create_context(struct pipe_screen *screen) intel_be_init_context(intel, device); -#if 0 - pipe = intel_create_softpipe(intel, screen->winsys); -#else - pipe = i915_create_context(screen, &device->base, &intel->base); -#endif + if (device->softpipe) + pipe = softpipe_create(screen); + else + pipe = i915_create_context(screen, &device->base, &intel->base); - pipe->priv = intel; + if (pipe) + pipe->priv = intel; return pipe; } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c index 0f6300323b..907ac86637 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c @@ -6,12 +6,15 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" #include "util/u_memory.h" +#include "util/u_debug.h" #include "intel_be_fence.h" #include "i915simple/i915_winsys.h" +#include "softpipe/sp_winsys.h" #include "intel_be_api.h" +#include <stdio.h> /* * Buffer @@ -175,8 +178,6 @@ intel_be_handle_from_buffer(struct pipe_screen *screen, struct pipe_buffer *buffer, unsigned *handle) { - drm_intel_bo *bo; - if (!buffer) return FALSE; @@ -285,14 +286,47 @@ intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size); + dev->softpipe = debug_get_bool_option("INTEL_SOFTPIPE", FALSE); + return true; } +static void +intel_be_get_device_id(unsigned int *device_id) +{ + char path[512]; + FILE *file; + + /* + * FIXME: Fix this up to use a drm ioctl or whatever. + */ + + snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device"); + file = fopen(path, "r"); + if (!file) { + return; + } + + fgets(path, sizeof(path), file); + sscanf(path, "%x", device_id); + fclose(file); +} + struct pipe_screen * -intel_be_create_screen(int drmFD, int deviceID) +intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg) { struct intel_be_device *dev; struct pipe_screen *screen; + unsigned int deviceID; + + if (arg != NULL) { + switch(arg->mode) { + case DRM_CREATE_NORMAL: + break; + default: + return NULL; + } + } /* Allocate the private area */ dev = malloc(sizeof(*dev)); @@ -300,9 +334,14 @@ intel_be_create_screen(int drmFD, int deviceID) return NULL; memset(dev, 0, sizeof(*dev)); + intel_be_get_device_id(&deviceID); intel_be_init_device(dev, drmFD, deviceID); - screen = i915_create_screen(&dev->base, deviceID); + if (dev->softpipe) { + screen = softpipe_create_screen(&dev->base); + drm_api_hooks.buffer_from_texture = softpipe_get_texture_buffer; + } else + screen = i915_create_screen(&dev->base, deviceID); return screen; } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h index 47d2176cb4..b32637ece2 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h @@ -16,6 +16,8 @@ struct intel_be_device { struct pipe_winsys base; + boolean softpipe; + int fd; /**< Drm file discriptor */ unsigned id; diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile index a45ca570db..d9aa5d54e1 100644 --- a/src/gallium/winsys/drm/intel/xorg/Makefile +++ b/src/gallium/winsys/drm/intel/xorg/Makefile @@ -21,6 +21,7 @@ LIBS = \ $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \ $(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \ $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ $(GALLIUM_AUXILIARIES) ############################################# @@ -29,7 +30,7 @@ LIBS = \ all default: $(TARGET) -$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a +$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a $(LIBS) $(TOP)/bin/mklib -noprefix -o $@ \ $(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel @@ -38,6 +39,6 @@ clean: install: $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + $(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) .PHONY = all clean install diff --git a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c index aea39247e5..46a7971f60 100644 --- a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c +++ b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c @@ -37,20 +37,17 @@ static Bool intel_xorg_pci_probe(DriverPtr driver, intptr_t match_data); static const struct pci_id_match intel_xorg_device_match[] = { - {0x8086, 0x2592, 0xffff, 0xffff, 0, 0, 0}, - {0x8086, 0x27A2, 0xffff, 0xffff, 0, 0, 0}, + {0x8086, PCI_MATCH_ANY, PCI_MATCH_ANY, PCI_MATCH_ANY, 0, 0, 0}, {0, 0, 0}, }; static SymTabRec intel_xorg_chipsets[] = { - {0x2592, "Intel Graphics Device"}, - {0x27A2, "Intel Graphics Device"}, + {PCI_MATCH_ANY, "Intel Graphics Device"}, {-1, NULL} }; static PciChipsets intel_xorg_pci_devices[] = { - {0x2592, 0x2592, RES_SHARED_VGA}, - {0x27A2, 0x27A2, RES_SHARED_VGA}, + {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA}, {-1, -1, RES_UNDEFINED} }; diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile index f8c8135854..6c9cbef26d 100644 --- a/src/gallium/winsys/drm/nouveau/Makefile +++ b/src/gallium/winsys/drm/nouveau/Makefile @@ -2,7 +2,7 @@ TOP = ../../../../.. include $(TOP)/configs/current -SUBDIRS = drm dri dri2 +SUBDIRS = drm $(GALLIUM_STATE_TRACKERS_DIRS) default install clean: @for dir in $(SUBDIRS) ; do \ diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile index f7db6201fe..0937f68c34 100644 --- a/src/gallium/winsys/drm/nouveau/dri/Makefile +++ b/src/gallium/winsys/drm/nouveau/dri/Makefile @@ -3,32 +3,25 @@ include $(TOP)/configs/current LIBNAME = nouveau_dri.so -MINIGLX_SOURCES = - PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/dri/libdridrm.a \ $(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \ $(TOP)/src/gallium/drivers/nv04/libnv04.a \ $(TOP)/src/gallium/drivers/nv10/libnv10.a \ $(TOP)/src/gallium/drivers/nv20/libnv20.a \ $(TOP)/src/gallium/drivers/nv30/libnv30.a \ $(TOP)/src/gallium/drivers/nv40/libnv40.a \ - $(TOP)/src/gallium/drivers/nv50/libnv50.a - -DRIVER_SOURCES = \ - nouveau_context.c \ - nouveau_screen.c \ - nouveau_swapbuffers.c \ - nouveau_lock.c + $(TOP)/src/gallium/drivers/nv50/libnv50.a \ + $(TOP)/src/gallium/drivers/nouveau/libnouveau.a + +DRIVER_SOURCES = C_SOURCES = \ $(COMMON_GALLIUM_SOURCES) \ $(DRIVER_SOURCES) -ASM_SOURCES = +include ../../Makefile.template -DRIVER_DEFINES = $(shell pkg-config libdrm_nouveau --cflags) DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs) -include ../../Makefile.template - symlinks: diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c deleted file mode 100644 index deb6ffcff1..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c +++ /dev/null @@ -1,118 +0,0 @@ -#include <main/glheader.h> -#include <glapi/glthread.h> -#include <GL/internal/glcore.h> -#include <utils.h> - -#include <state_tracker/st_public.h> -#include <state_tracker/st_context.h> -#include <state_tracker/drm_api.h> -#include <pipe/p_defines.h> -#include <pipe/p_context.h> -#include <pipe/p_screen.h> - -#include "nouveau_context.h" -#include "nouveau_screen.h" - -#include "nouveau_drmif.h" - -GLboolean -nouveau_context_create(const __GLcontextModes *glVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) -{ - __DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv; - struct nouveau_screen *nv_screen = driScrnPriv->private; - struct nouveau_context *nv; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - if (sharedContextPrivate) - st_share = ((struct nouveau_context *)sharedContextPrivate)->st; - - nv = CALLOC_STRUCT(nouveau_context); - if (!nv) - return GL_FALSE; - - { - struct nouveau_device_priv *nvdev = - nouveau_device(nv_screen->device); - - nvdev->ctx = driContextPriv->hHWContext; - nvdev->lock = (drmLock *)&driScrnPriv->pSAREA->lock; - } - - pipe = drm_api_hooks.create_context(nv_screen->pscreen); - if (!pipe) { - FREE(nv); - return GL_FALSE; - } - pipe->priv = nv; - - driContextPriv->driverPrivate = nv; - nv->dri_screen = driScrnPriv; - - driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache, - nv->dri_screen->myNum, "nouveau"); - - nv->st = st_create_context(pipe, glVis, st_share); - return GL_TRUE; -} - -void -nouveau_context_destroy(__DRIcontextPrivate *driContextPriv) -{ - struct nouveau_context *nv = driContextPriv->driverPrivate; - - assert(nv); - - st_finish(nv->st); - st_destroy_context(nv->st); - - FREE(nv); -} - -GLboolean -nouveau_context_bind(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv) -{ - struct nouveau_context *nv; - struct nouveau_framebuffer *draw, *read; - - if (!driContextPriv) { - st_make_current(NULL, NULL, NULL); - return GL_TRUE; - } - - nv = driContextPriv->driverPrivate; - draw = driDrawPriv->driverPrivate; - read = driReadPriv->driverPrivate; - - st_make_current(nv->st, draw->stfb, read->stfb); - - if ((nv->dri_drawable != driDrawPriv) || - (nv->last_stamp != driDrawPriv->lastStamp)) { - nv->dri_drawable = driDrawPriv; - st_resize_framebuffer(draw->stfb, driDrawPriv->w, - driDrawPriv->h); - nv->last_stamp = driDrawPriv->lastStamp; - } - - if (driDrawPriv != driReadPriv) { - st_resize_framebuffer(read->stfb, driReadPriv->w, - driReadPriv->h); - } - - return GL_TRUE; -} - -GLboolean -nouveau_context_unbind(__DRIcontextPrivate *driContextPriv) -{ - struct nouveau_context *nv = driContextPriv->driverPrivate; - (void)nv; - - st_flush(nv->st, 0, NULL); - return GL_TRUE; -} - diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h deleted file mode 100644 index 2779b092e6..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h +++ /dev/null @@ -1,53 +0,0 @@ -#ifndef __NOUVEAU_CONTEXT_DRI_H__ -#define __NOUVEAU_CONTEXT_DRI_H__ - -#include <dri_util.h> -#include <xmlconfig.h> - -#include "nouveau/nouveau_winsys.h" - -#define NOUVEAU_ERR(fmt, args...) debug_printf("%s: "fmt, __func__, ##args) - -struct nouveau_framebuffer { - struct st_framebuffer *stfb; -}; - -struct nouveau_context { - struct st_context *st; - - /* DRI stuff */ - __DRIscreenPrivate *dri_screen; - __DRIdrawablePrivate *dri_drawable; - unsigned int last_stamp; - driOptionCache dri_option_cache; - drm_context_t drm_context; - drmLock drm_lock; - int locked; -}; - -extern GLboolean nouveau_context_create(const __GLcontextModes *, - __DRIcontextPrivate *, void *); -extern void nouveau_context_destroy(__DRIcontextPrivate *); -extern GLboolean nouveau_context_bind(__DRIcontextPrivate *, - __DRIdrawablePrivate *draw, - __DRIdrawablePrivate *read); -extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *); - -extern void nouveau_contended_lock(struct nouveau_context *nv); -extern void LOCK_HARDWARE(struct nouveau_context *nv); -extern void UNLOCK_HARDWARE(struct nouveau_context *nv); - -#ifdef DEBUG -extern int __nouveau_debug; - -#define DEBUG_BO (1 << 0) - -#define DBG(flag, ...) do { \ - if (__nouveau_debug & (DEBUG_##flag)) \ - NOUVEAU_ERR(__VA_ARGS__); \ -} while(0) -#else -#define DBG(flag, ...) -#endif - -#endif diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c deleted file mode 100644 index 92f5bd09c9..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include <pipe/p_thread.h> -#include "nouveau_context.h" -#include "nouveau_screen.h" -#include "nouveau_drmif.h" - -pipe_static_mutex(lockMutex); - -/* Lock the hardware and validate our state. - */ -void -LOCK_HARDWARE(struct nouveau_context *nv) -{ - struct nouveau_screen *nv_screen = nv->dri_screen->private; - struct nouveau_device *dev = nv_screen->device; - struct nouveau_device_priv *nvdev = nouveau_device(dev); - char __ret=0; - - assert(!nv->locked); - pipe_mutex_lock(lockMutex); - - DRM_CAS(nvdev->lock, nvdev->ctx, - (DRM_LOCK_HELD | nvdev->ctx), __ret); - - if (__ret) { - drmGetLock(nvdev->fd, nvdev->ctx, 0); - nouveau_contended_lock(nv); - } - nv->locked = 1; -} - -/* Unlock the hardware using the global current context - */ -void -UNLOCK_HARDWARE(struct nouveau_context *nv) -{ - struct nouveau_screen *nv_screen = nv->dri_screen->private; - struct nouveau_device *dev = nv_screen->device; - struct nouveau_device_priv *nvdev = nouveau_device(dev); - - assert(nv->locked); - nv->locked = 0; - - DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx); - - pipe_mutex_unlock(lockMutex); -} diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c deleted file mode 100644 index 0b45b1ff1f..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c +++ /dev/null @@ -1,330 +0,0 @@ -#include <utils.h> -#include <vblank.h> -#include <xmlpool.h> - -#include <pipe/p_context.h> -#include <state_tracker/st_public.h> -#include <state_tracker/st_cb_fbo.h> -#include <state_tracker/drm_api.h> - -#include "nouveau_context.h" -#include "nouveau_screen.h" -#include "nouveau_swapbuffers.h" -#include "nouveau_dri.h" - -#include "nouveau_drm.h" -#include "nouveau_drmif.h" - -#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12 -#error nouveau_drm.h version does not match expected version -#endif - -/* Extension stuff, enabling of extensions handled by Gallium's GL state - * tracker. But, we still need to define the entry points we want. - */ -#define need_GL_ARB_fragment_program -#define need_GL_ARB_multisample -#define need_GL_ARB_occlusion_query -#define need_GL_ARB_point_parameters -#define need_GL_ARB_shader_objects -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_program -#define need_GL_ARB_vertex_shader -#define need_GL_ARB_vertex_buffer_object -#define need_GL_EXT_compiled_vertex_array -#define need_GL_EXT_fog_coord -#define need_GL_EXT_secondary_color -#define need_GL_EXT_framebuffer_object -#define need_GL_VERSION_2_0 -#define need_GL_VERSION_2_1 -#include "extension_helper.h" - -const struct dri_extension card_extensions[] = -{ - { "GL_ARB_multisample", GL_ARB_multisample_functions }, - { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, - { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, - { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, - { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, - { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, - { "GL_ARB_texture_compression", GL_ARB_texture_compression_functions }, - { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, - { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions }, - { "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions }, - { "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions }, - { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, - { "GL_EXT_secondary_color", GL_EXT_secondary_color_functions }, - { NULL, 0 } -}; - -PUBLIC const char __driConfigOptions[] = -DRI_CONF_BEGIN -DRI_CONF_END; -static const GLuint __driNConfigOptions = 0; - -extern const struct dri_extension common_extensions[]; -extern const struct dri_extension nv40_extensions[]; - -static GLboolean -nouveau_create_buffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes *glVis, GLboolean pixmapBuffer) -{ - struct nouveau_framebuffer *nvfb; - enum pipe_format colour, depth, stencil; - - if (pixmapBuffer) - return GL_FALSE; - - nvfb = CALLOC_STRUCT(nouveau_framebuffer); - if (!nvfb) - return GL_FALSE; - - if (glVis->redBits == 5) - colour = PIPE_FORMAT_R5G6B5_UNORM; - else - colour = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (glVis->depthBits == 16) - depth = PIPE_FORMAT_Z16_UNORM; - else if (glVis->depthBits == 24) - depth = PIPE_FORMAT_Z24S8_UNORM; - else - depth = PIPE_FORMAT_NONE; - - if (glVis->stencilBits == 8) - stencil = PIPE_FORMAT_Z24S8_UNORM; - else - stencil = PIPE_FORMAT_NONE; - - nvfb->stfb = st_create_framebuffer(glVis, colour, depth, stencil, - driDrawPriv->w, driDrawPriv->h, - (void*)nvfb); - if (!nvfb->stfb) { - free(nvfb); - return GL_FALSE; - } - - driDrawPriv->driverPrivate = (void *)nvfb; - return GL_TRUE; -} - -static void -nouveau_destroy_buffer(__DRIdrawablePrivate * driDrawPriv) -{ - struct nouveau_framebuffer *nvfb; - - nvfb = (struct nouveau_framebuffer *)driDrawPriv->driverPrivate; - st_unreference_framebuffer(nvfb->stfb); - free(nvfb); -} - -static __DRIconfig ** -nouveau_fill_in_modes(__DRIscreenPrivate *psp, - unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) -{ - __DRIconfig **configs; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, - }; - - uint8_t depth_bits_array[3]; - uint8_t stencil_bits_array[3]; - uint8_t msaa_samples_array[1]; - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - msaa_samples_array[0] = 0; - - depth_buffer_factor = - ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - configs = driCreateConfigs(fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, msaa_samples_array, 1); - if (configs == NULL) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", - __func__, __LINE__); - return NULL; - } - - return configs; -} - -static struct pipe_surface * -dri_surface_from_handle(struct pipe_screen *screen, - unsigned handle, - enum pipe_format format, - unsigned width, - unsigned height, - unsigned pitch) -{ - struct pipe_surface *surface = NULL; - struct pipe_texture *texture = NULL; - struct pipe_texture templat; - struct pipe_buffer *buf = NULL; - - buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle); - if (!buf) - return NULL; - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY | - NOUVEAU_TEXTURE_USAGE_LINEAR; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = format; - templat.width[0] = width; - templat.height[0] = height; - pf_get_block(templat.format, &templat.block); - - texture = screen->texture_blanket(screen, - &templat, - &pitch, - buf); - - /* we don't need the buffer from this point on */ - pipe_buffer_reference(&buf, NULL); - - if (!texture) - return NULL; - - surface = screen->get_tex_surface(screen, texture, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - - /* we don't need the texture from this point on */ - pipe_texture_reference(&texture, NULL); - return surface; -} - -static const __DRIconfig ** -nouveau_screen_create(__DRIscreenPrivate *psp) -{ - struct nouveau_dri *nv_dri = psp->pDevPriv; - struct nouveau_screen *nv_screen; - static const __DRIversion ddx_expected = - { 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = - { 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL }; - - if (!driCheckDriDdxDrmVersions2("nouveau", - &psp->dri_version, &dri_expected, - &psp->ddx_version, &ddx_expected, - &psp->drm_version, &drm_expected)) { - return NULL; - } - - if (drm_expected.patch != psp->drm_version.patch) { - fprintf(stderr, "Incompatible DRM patch level.\n" - "Expected: %d\n" "Current : %d\n", - drm_expected.patch, psp->drm_version.patch); - return NULL; - } - - driInitExtensions(NULL, card_extensions, GL_FALSE); - - if (psp->devPrivSize != sizeof(struct nouveau_dri)) { - NOUVEAU_ERR("DRI struct mismatch between DDX/DRI\n"); - return NULL; - } - - nv_screen = CALLOC_STRUCT(nouveau_screen); - if (!nv_screen) - return NULL; - - nouveau_device_open_existing(&nv_screen->device, 0, psp->fd, 0); - - nv_screen->pscreen = drm_api_hooks.create_screen(psp->fd, 0); - if (!nv_screen->pscreen) { - FREE(nv_screen); - return NULL; - } - nv_screen->pscreen->flush_frontbuffer = nouveau_flush_frontbuffer; - - { - enum pipe_format format; - - if (nv_dri->bpp == 16) - format = PIPE_FORMAT_R5G6B5_UNORM; - else - format = PIPE_FORMAT_A8R8G8B8_UNORM; - - nv_screen->fb = dri_surface_from_handle(nv_screen->pscreen, - nv_dri->front_offset, - format, - nv_dri->width, - nv_dri->height, - nv_dri->front_pitch * - nv_dri->bpp / 8); - } - - driParseOptionInfo(&nv_screen->option_cache, - __driConfigOptions, __driNConfigOptions); - - nv_screen->driScrnPriv = psp; - psp->private = (void *)nv_screen; - - return (const __DRIconfig **) - nouveau_fill_in_modes(psp, nv_dri->bpp, - (nv_dri->bpp == 16) ? 16 : 24, - (nv_dri->bpp == 16) ? 0 : 8, 1); -} - -static void -nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv) -{ - struct nouveau_screen *nv_screen = driScrnPriv->private; - - driScrnPriv->private = NULL; - FREE(nv_screen); -} - -const struct __DriverAPIRec -driDriverAPI = { - .InitScreen = nouveau_screen_create, - .DestroyScreen = nouveau_screen_destroy, - .CreateContext = nouveau_context_create, - .DestroyContext = nouveau_context_destroy, - .CreateBuffer = nouveau_create_buffer, - .DestroyBuffer = nouveau_destroy_buffer, - .SwapBuffers = nouveau_swap_buffers, - .MakeCurrent = nouveau_context_bind, - .UnbindContext = nouveau_context_unbind, - .CopySubBuffer = nouveau_copy_sub_buffer, - - .InitScreen2 = NULL, /* one day, I promise! */ -}; - diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h deleted file mode 100644 index ac078f3c63..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h +++ /dev/null @@ -1,16 +0,0 @@ -#ifndef __NOUVEAU_SCREEN_DRI_H__ -#define __NOUVEAU_SCREEN_DRI_H__ - -#include "xmlconfig.h" - -struct nouveau_screen { - __DRIscreenPrivate *driScrnPriv; - driOptionCache option_cache; - - struct nouveau_device *device; - - struct pipe_screen *pscreen; - struct pipe_surface *fb; -}; - -#endif diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c deleted file mode 100644 index 9c841a0b2d..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c +++ /dev/null @@ -1,115 +0,0 @@ -#include <main/glheader.h> -#include <glapi/glthread.h> -#include <GL/internal/glcore.h> - -#include <pipe/p_context.h> -#include <state_tracker/st_public.h> -#include <state_tracker/st_context.h> -#include <state_tracker/st_cb_fbo.h> - -#include "nouveau_context.h" -#include "nouveau_screen.h" -#include "nouveau_swapbuffers.h" - -#include "nouveau_pushbuf.h" - -void -nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf, - const drm_clip_rect_t *rect) -{ - struct nouveau_context *nv = dPriv->driContextPriv->driverPrivate; - struct nouveau_screen *nv_screen = nv->dri_screen->private; - struct pipe_context *pipe = nv->st->pipe; - drm_clip_rect_t *pbox; - int nbox, i; - - LOCK_HARDWARE(nv); - if (!dPriv->numClipRects) { - UNLOCK_HARDWARE(nv); - return; - } - pbox = dPriv->pClipRects; - nbox = dPriv->numClipRects; - - for (i = 0; i < nbox; i++, pbox++) { - int sx, sy, dx, dy, w, h; - - sx = pbox->x1 - dPriv->x; - sy = pbox->y1 - dPriv->y; - dx = pbox->x1; - dy = pbox->y1; - w = pbox->x2 - pbox->x1; - h = pbox->y2 - pbox->y1; - - pipe->surface_copy(pipe, nv_screen->fb, dx, dy, surf, - sx, sy, w, h); - } - - pipe->flush(pipe, 0, NULL); - UNLOCK_HARDWARE(nv); - - if (nv->last_stamp != dPriv->lastStamp) { - struct nouveau_framebuffer *nvfb = dPriv->driverPrivate; - st_resize_framebuffer(nvfb->stfb, dPriv->w, dPriv->h); - nv->last_stamp = dPriv->lastStamp; - } -} - -void -nouveau_copy_sub_buffer(__DRIdrawablePrivate *dPriv, int x, int y, int w, int h) -{ - struct nouveau_framebuffer *nvfb = dPriv->driverPrivate; - struct pipe_surface *surf; - - st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf); - if (surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = x + w; - rect.y2 = y + h; - - st_notify_swapbuffers(nvfb->stfb); - nouveau_copy_buffer(dPriv, surf, &rect); - } -} - -void -nouveau_swap_buffers(__DRIdrawablePrivate *dPriv) -{ - struct nouveau_framebuffer *nvfb = dPriv->driverPrivate; - struct pipe_surface *surf; - - st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf); - if (surf) { - st_notify_swapbuffers(nvfb->stfb); - nouveau_copy_buffer(dPriv, surf, NULL); - } -} - -void -nouveau_flush_frontbuffer(struct pipe_screen *pscreen, struct pipe_surface *ps, - void *context_private) -{ - struct nouveau_context *nv = context_private; - __DRIdrawablePrivate *dPriv = nv->dri_drawable; - - nouveau_copy_buffer(dPriv, ps, NULL); -} - -void -nouveau_contended_lock(struct nouveau_context *nv) -{ - struct nouveau_context *nv_sub = (struct nouveau_context*)nv; - __DRIdrawablePrivate *dPriv = nv_sub->dri_drawable; - __DRIscreenPrivate *sPriv = nv_sub->dri_screen; - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); -} - diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h deleted file mode 100644 index 4ca9cc2283..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h +++ /dev/null @@ -1,11 +0,0 @@ -#ifndef __NOUVEAU_SWAPBUFFERS_H__ -#define __NOUVEAU_SWAPBUFFERS_H__ - -void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *, - const drm_clip_rect_t *); -void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, int x, int y, int w, int h); -void nouveau_swap_buffers(__DRIdrawablePrivate *); -void nouveau_flush_frontbuffer(struct pipe_screen *, struct pipe_surface *, - void *context_private); - -#endif diff --git a/src/gallium/winsys/drm/nouveau/dri2/Makefile b/src/gallium/winsys/drm/nouveau/dri2/Makefile deleted file mode 100644 index 728870d2e1..0000000000 --- a/src/gallium/winsys/drm/nouveau/dri2/Makefile +++ /dev/null @@ -1,26 +0,0 @@ -TOP = ../../../../../.. -include $(TOP)/configs/current - -LIBNAME = nouveau_dri2.so - -PIPE_DRIVERS = \ - $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ - $(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \ - $(TOP)/src/gallium/drivers/nv04/libnv04.a \ - $(TOP)/src/gallium/drivers/nv10/libnv10.a \ - $(TOP)/src/gallium/drivers/nv20/libnv20.a \ - $(TOP)/src/gallium/drivers/nv30/libnv30.a \ - $(TOP)/src/gallium/drivers/nv40/libnv40.a \ - $(TOP)/src/gallium/drivers/nv50/libnv50.a - -DRIVER_SOURCES = - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(DRIVER_SOURCES) - -include ../../Makefile.template - -DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs) - -symlinks: diff --git a/src/gallium/winsys/drm/nouveau/drm/Makefile b/src/gallium/winsys/drm/nouveau/drm/Makefile index 2da78d8690..54c3b26c75 100644 --- a/src/gallium/winsys/drm/nouveau/drm/Makefile +++ b/src/gallium/winsys/drm/nouveau/drm/Makefile @@ -3,9 +3,7 @@ include $(TOP)/configs/current LIBNAME = nouveaudrm -C_SOURCES = nouveau_drm_api.c \ - nouveau_winsys_pipe.c \ - nouveau_winsys.c +C_SOURCES = nouveau_drm_api.c LIBRARY_INCLUDES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-I) LIBRARY_DEFINES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-other) diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h index 1207c2d609..1207c2d609 100644 --- a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index c0127e803f..395b21ec7a 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -1,20 +1,83 @@ +#include "pipe/p_context.h" +#include "pipe/p_state.h" #include "util/u_memory.h" #include "nouveau_drm_api.h" -#include "nouveau_winsys_pipe.h" #include "nouveau_drmif.h" #include "nouveau_channel.h" #include "nouveau_bo.h" +#include "nouveau/nouveau_winsys.h" +#include "nouveau/nouveau_screen.h" + +static struct pipe_surface * +dri_surface_from_handle(struct pipe_screen *screen, + unsigned handle, + enum pipe_format format, + unsigned width, + unsigned height, + unsigned pitch) +{ + struct pipe_surface *surface = NULL; + struct pipe_texture *texture = NULL; + struct pipe_texture templat; + struct pipe_buffer *buf = NULL; + + buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle); + if (!buf) + return NULL; + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY | + NOUVEAU_TEXTURE_USAGE_LINEAR; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = format; + templat.width[0] = width; + templat.height[0] = height; + pf_get_block(templat.format, &templat.block); + + texture = screen->texture_blanket(screen, + &templat, + &pitch, + buf); + + /* we don't need the buffer from this point on */ + pipe_buffer_reference(&buf, NULL); + + if (!texture) + return NULL; + + surface = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* we don't need the texture from this point on */ + pipe_texture_reference(&texture, NULL); + return surface; +} + +static struct pipe_surface * +nouveau_dri1_front_surface(struct pipe_context *pipe) +{ + return nouveau_winsys_screen(pipe->screen)->front; +} + +static struct dri1_api nouveau_dri1_api = { + nouveau_dri1_front_surface, +}; + static struct pipe_screen * -nouveau_drm_create_screen(int fd, int pciid) +nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg) { - struct pipe_winsys *ws; + struct dri1_create_screen_arg *dri1 = (void *)arg; struct nouveau_winsys *nvws; + struct pipe_winsys *ws; struct nouveau_device *dev = NULL; struct pipe_screen *(*init)(struct pipe_winsys *, - struct nouveau_winsys *); + struct nouveau_device *); int ret; ret = nouveau_device_open_existing(&dev, 0, fd, 0); @@ -49,33 +112,54 @@ nouveau_drm_create_screen(int fd, int pciid) return NULL; } - ws = nouveau_pipe_winsys_new(dev); - if (!ws) { + nvws = CALLOC_STRUCT(nouveau_winsys); + if (!nvws) { nouveau_device_close(&dev); return NULL; } + ws = &nvws->base; - nvws = nouveau_winsys_new(ws); - if (!nvws) { + nvws->pscreen = init(ws, dev); + if (!nvws->pscreen) { ws->destroy(ws); return NULL; } - nouveau_pipe_winsys(ws)->pscreen = init(ws, nvws); - if (!nouveau_pipe_winsys(ws)->pscreen) { - ws->destroy(ws); - return NULL; + if (arg->mode == DRM_CREATE_DRI1) { + struct nouveau_dri *nvdri = dri1->ddx_info; + enum pipe_format format; + + if (nvdri->bpp == 16) + format = PIPE_FORMAT_R5G6B5_UNORM; + else + format = PIPE_FORMAT_A8R8G8B8_UNORM; + + nvws->front = dri_surface_from_handle(nvws->pscreen, + nvdri->front_offset, + format, + nvdri->width, + nvdri->height, + nvdri->front_pitch * + (nvdri->bpp / 8)); + if (!nvws->front) { + debug_printf("%s: error referencing front buffer\n", + __func__); + ws->destroy(ws); + return NULL; + } + + dri1->api = &nouveau_dri1_api; } - return nouveau_pipe_winsys(ws)->pscreen; + return nvws->pscreen; } static struct pipe_context * nouveau_drm_create_context(struct pipe_screen *pscreen) { - struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen); + struct nouveau_winsys *nvws = nouveau_winsys_screen(pscreen); struct pipe_context *(*init)(struct pipe_screen *, unsigned); - unsigned chipset = nvpws->channel->device->chipset; + unsigned chipset = nouveau_screen(pscreen)->device->chipset; int i; switch (chipset & 0xf0) { @@ -106,19 +190,19 @@ nouveau_drm_create_context(struct pipe_screen *pscreen) } /* Find a free slot for a pipe context, allocate a new one if needed */ - for (i = 0; i < nvpws->nr_pctx; i++) { - if (nvpws->pctx[i] == NULL) + for (i = 0; i < nvws->nr_pctx; i++) { + if (nvws->pctx[i] == NULL) break; } - if (i == nvpws->nr_pctx) { - nvpws->nr_pctx++; - nvpws->pctx = realloc(nvpws->pctx, - sizeof(*nvpws->pctx) * nvpws->nr_pctx); + if (i == nvws->nr_pctx) { + nvws->nr_pctx++; + nvws->pctx = realloc(nvws->pctx, + sizeof(*nvws->pctx) * nvws->nr_pctx); } - nvpws->pctx[i] = init(pscreen, i); - return nvpws->pctx[i]; + nvws->pctx[i] = init(pscreen, i); + return nvws->pctx[i]; } static boolean @@ -132,42 +216,41 @@ static struct pipe_buffer * nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name, unsigned handle) { - struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen); - struct nouveau_device *dev = nvpws->channel->device; - struct nouveau_pipe_buffer *nvpb; + struct nouveau_device *dev = nouveau_screen(pscreen)->device; + struct pipe_buffer *pb; int ret; - nvpb = CALLOC_STRUCT(nouveau_pipe_buffer); - if (!nvpb) + pb = CALLOC(1, sizeof(struct pipe_buffer) + sizeof(struct nouveau_bo*)); + if (!pb) return NULL; - ret = nouveau_bo_handle_ref(dev, handle, &nvpb->bo); + ret = nouveau_bo_handle_ref(dev, handle, (struct nouveau_bo**)(pb+1)); if (ret) { debug_printf("%s: ref name 0x%08x failed with %d\n", __func__, handle, ret); - FREE(nvpb); + FREE(pb); return NULL; } - pipe_reference_init(&nvpb->base.reference, 1); - nvpb->base.screen = pscreen; - nvpb->base.alignment = 0; - nvpb->base.usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE | - PIPE_BUFFER_USAGE_CPU_READ_WRITE; - nvpb->base.size = nvpb->bo->size; - return &nvpb->base; + pipe_reference_init(&pb->reference, 1); + pb->screen = pscreen; + pb->alignment = 0; + pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE | + PIPE_BUFFER_USAGE_CPU_READ_WRITE; + pb->size = nouveau_bo(pb)->size; + return pb; } static boolean nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, unsigned *handle) { - struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb); + struct nouveau_bo *bo = nouveau_bo(pb); - if (!nvpb) + if (!bo) return FALSE; - *handle = nvpb->bo->handle; + *handle = bo->handle; return TRUE; } @@ -175,12 +258,12 @@ static boolean nouveau_drm_name_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, unsigned *handle) { - struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb); + struct nouveau_bo *bo = nouveau_bo(pb); - if (!nvpb) + if (!bo) return FALSE; - return nouveau_bo_handle_get(nvpb->bo, handle) == 0; + return nouveau_bo_handle_get(bo, handle) == 0; } struct drm_api drm_api_hooks = { diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h index 2782c83c0e..e61e0e0957 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h @@ -1,5 +1,34 @@ #ifndef __NOUVEAU_DRM_API_H__ #define __NOUVEAU_DRM_API_H__ + #include "state_tracker/drm_api.h" +#include "state_tracker/dri1_api.h" + +#include "pipe/internal/p_winsys_screen.h" + +#include "nouveau_dri.h" + +struct nouveau_winsys { + struct pipe_winsys base; + + struct pipe_screen *pscreen; + + unsigned nr_pctx; + struct pipe_context **pctx; + + struct pipe_surface *front; +}; + +static INLINE struct nouveau_winsys * +nouveau_winsys(struct pipe_winsys *ws) +{ + return (struct nouveau_winsys *)ws; +} + +static INLINE struct nouveau_winsys * +nouveau_winsys_screen(struct pipe_screen *pscreen) +{ + return nouveau_winsys(pscreen->winsys); +} #endif diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c deleted file mode 100644 index e3175fd775..0000000000 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c +++ /dev/null @@ -1,94 +0,0 @@ -#include "util/u_memory.h" - -#include "nouveau_winsys_pipe.h" - -static int -nouveau_pipe_notifier_alloc(struct nouveau_winsys *nvws, int count, - struct nouveau_notifier **notify) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws); - - return nouveau_notifier_alloc(nvpws->channel, nvpws->next_handle++, - count, notify); -} - -static int -nouveau_pipe_grobj_alloc(struct nouveau_winsys *nvws, int grclass, - struct nouveau_grobj **grobj) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws); - struct nouveau_channel *chan = nvpws->channel; - int ret; - - ret = nouveau_grobj_alloc(chan, nvpws->next_handle++, grclass, grobj); - if (ret) - return ret; - - BEGIN_RING(chan, *grobj, 0x0000, 1); - OUT_RING (chan, (*grobj)->handle); - (*grobj)->bound = NOUVEAU_GROBJ_BOUND_EXPLICIT; - return 0; -} - -static int -nouveau_pipe_push_reloc(struct nouveau_winsys *nvws, void *ptr, - struct pipe_buffer *buf, uint32_t data, - uint32_t flags, uint32_t vor, uint32_t tor) -{ - struct nouveau_bo *bo = nouveau_pipe_buffer(buf)->bo; - - return nouveau_pushbuf_emit_reloc(nvws->channel, ptr, bo, - data, flags, vor, tor); -} - -static int -nouveau_pipe_push_flush(struct nouveau_winsys *nvws, unsigned size, - struct pipe_fence_handle **fence) -{ - if (fence) - *fence = NULL; - - return nouveau_pushbuf_flush(nvws->channel, size); -} - -static struct nouveau_bo * -nouveau_pipe_get_bo(struct pipe_buffer *pb) -{ - return nouveau_pipe_buffer(pb)->bo; -} - -struct nouveau_winsys * -nouveau_winsys_new(struct pipe_winsys *ws) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); - struct nouveau_winsys *nvws; - - nvws = CALLOC_STRUCT(nouveau_winsys); - if (!nvws) - return NULL; - - nvws->ws = ws; - nvws->channel = nvpws->channel; - - nvws->res_init = nouveau_resource_init; - nvws->res_alloc = nouveau_resource_alloc; - nvws->res_free = nouveau_resource_free; - - nvws->push_reloc = nouveau_pipe_push_reloc; - nvws->push_flush = nouveau_pipe_push_flush; - - nvws->grobj_alloc = nouveau_pipe_grobj_alloc; - nvws->grobj_free = nouveau_grobj_free; - - nvws->notifier_alloc = nouveau_pipe_notifier_alloc; - nvws->notifier_free = nouveau_notifier_free; - nvws->notifier_reset = nouveau_notifier_reset; - nvws->notifier_status = nouveau_notifier_status; - nvws->notifier_retval = nouveau_notifier_return_val; - nvws->notifier_wait = nouveau_notifier_wait_status; - - nvws->get_bo = nouveau_pipe_get_bo; - - return nvws; -} - diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c deleted file mode 100644 index 9e03a9f5db..0000000000 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c +++ /dev/null @@ -1,204 +0,0 @@ -#include "pipe/internal/p_winsys_screen.h" -#include <pipe/p_defines.h> -#include <pipe/p_inlines.h> -#include <util/u_memory.h> - -#include "nouveau_winsys_pipe.h" - -#include "nouveau_drmif.h" -#include "nouveau_bo.h" - -static const char * -nouveau_get_name(struct pipe_winsys *pws) -{ - return "Nouveau/DRI"; -} - -static uint32_t -nouveau_flags_from_usage(struct pipe_winsys *ws, unsigned usage) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); - struct pipe_screen *pscreen = nvpws->pscreen; - uint32_t flags = NOUVEAU_BO_LOCAL; - - if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER) - flags |= NOUVEAU_BO_GART; - - if (usage & PIPE_BUFFER_USAGE_PIXEL) { - if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE) - flags |= NOUVEAU_BO_GART; - if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE)) - flags |= NOUVEAU_BO_VRAM; - - switch (nvpws->channel->device->chipset & 0xf0) { - case 0x50: - case 0x80: - case 0x90: - flags |= NOUVEAU_BO_TILED; - if (usage & NOUVEAU_BUFFER_USAGE_ZETA) - flags |= NOUVEAU_BO_ZTILE; - break; - default: - break; - } - } - - if (usage & PIPE_BUFFER_USAGE_VERTEX) { - if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF)) - flags |= NOUVEAU_BO_GART; - } - - if (usage & PIPE_BUFFER_USAGE_INDEX) { - if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF)) - flags |= NOUVEAU_BO_GART; - } - - return flags; -} - -static struct pipe_buffer * -nouveau_pipe_bo_create(struct pipe_winsys *ws, unsigned alignment, - unsigned usage, unsigned size) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); - struct nouveau_device *dev = nvpws->channel->device; - struct nouveau_pipe_buffer *nvbuf; - uint32_t flags; - - nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); - if (!nvbuf) - return NULL; - pipe_reference_init(&nvbuf->base.reference, 1); - nvbuf->base.alignment = alignment; - nvbuf->base.usage = usage; - nvbuf->base.size = size; - - flags = nouveau_flags_from_usage(ws, usage); - if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) { - FREE(nvbuf); - return NULL; - } - - return &nvbuf->base; -} - -static struct pipe_buffer * -nouveau_pipe_bo_user_create(struct pipe_winsys *ws, void *ptr, unsigned bytes) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); - struct nouveau_device *dev = nvpws->channel->device; - struct nouveau_pipe_buffer *nvbuf; - - nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer); - if (!nvbuf) - return NULL; - pipe_reference_init(&nvbuf->base.reference, 1); - nvbuf->base.size = bytes; - - if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) { - FREE(nvbuf); - return NULL; - } - - return &nvbuf->base; -} - -static void -nouveau_pipe_bo_del(struct pipe_buffer *buf) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - - nouveau_bo_ref(NULL, &nvbuf->bo); - FREE(nvbuf); -} - -static void * -nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf, - unsigned flags) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - uint32_t map_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - map_flags |= NOUVEAU_BO_RD; - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - map_flags |= NOUVEAU_BO_WR; - - if (nouveau_bo_map(nvbuf->bo, map_flags)) - return NULL; - return nvbuf->bo->map; -} - -static void -nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) -{ - struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf); - - nouveau_bo_unmap(nvbuf->bo); -} - -static void -nouveau_pipe_fence_reference(struct pipe_winsys *ws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *pfence) -{ - *ptr = pfence; -} - -static int -nouveau_pipe_fence_signalled(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, unsigned flag) -{ - return 0; -} - -static int -nouveau_pipe_fence_finish(struct pipe_winsys *ws, - struct pipe_fence_handle *pfence, unsigned flag) -{ - return 0; -} - -static void -nouveau_destroy(struct pipe_winsys *ws) -{ - struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws); - - nouveau_device_close(&nvpws->channel->device); - FREE(nvpws); -} - -struct pipe_winsys * -nouveau_pipe_winsys_new(struct nouveau_device *dev) -{ - struct nouveau_pipe_winsys *nvpws; - int ret; - - nvpws = CALLOC_STRUCT(nouveau_pipe_winsys); - if (!nvpws) - return NULL; - - ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202, - &nvpws->channel); - if (ret) { - debug_printf("%s: error opening GPU channel: %d\n", - __func__, ret); - FREE(nvpws); - return NULL; - } - nvpws->next_handle = 0x77000000; - - nvpws->base.buffer_create = nouveau_pipe_bo_create; - nvpws->base.buffer_destroy = nouveau_pipe_bo_del; - nvpws->base.user_buffer_create = nouveau_pipe_bo_user_create; - nvpws->base.buffer_map = nouveau_pipe_bo_map; - nvpws->base.buffer_unmap = nouveau_pipe_bo_unmap; - - nvpws->base.fence_reference = nouveau_pipe_fence_reference; - nvpws->base.fence_signalled = nouveau_pipe_fence_signalled; - nvpws->base.fence_finish = nouveau_pipe_fence_finish; - - nvpws->base.get_name = nouveau_get_name; - nvpws->base.destroy = nouveau_destroy; - return &nvpws->base; -} diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h deleted file mode 100644 index 10e1e269e8..0000000000 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h +++ /dev/null @@ -1,52 +0,0 @@ -#ifndef NOUVEAU_PIPE_WINSYS_H -#define NOUVEAU_PIPE_WINSYS_H - -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_context.h" - -#include "nouveau/nouveau_winsys.h" - -#include "nouveau_device.h" - -struct nouveau_pipe_buffer { - struct pipe_buffer base; - struct nouveau_bo *bo; -}; - -static INLINE struct nouveau_pipe_buffer * -nouveau_pipe_buffer(struct pipe_buffer *buf) -{ - return (struct nouveau_pipe_buffer *)buf; -} - -struct nouveau_pipe_winsys { - struct pipe_winsys base; - - struct pipe_screen *pscreen; - - struct nouveau_channel *channel; - uint32_t next_handle; - - unsigned nr_pctx; - struct pipe_context **pctx; -}; - -static INLINE struct nouveau_pipe_winsys * -nouveau_pipe_winsys(struct pipe_winsys *ws) -{ - return (struct nouveau_pipe_winsys *)ws; -} - -static INLINE struct nouveau_pipe_winsys * -nouveau_screen(struct pipe_screen *pscreen) -{ - return nouveau_pipe_winsys(pscreen->winsys); -} - -struct pipe_winsys * -nouveau_pipe_winsys_new(struct nouveau_device *); - -struct nouveau_winsys * -nouveau_winsys_new(struct pipe_winsys *ws); - -#endif diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 611ee68da6..0d0fdc5bd8 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -68,8 +68,8 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws, domain |= RADEON_GEM_DOMAIN_GTT; } - radeon_buffer->bo = radeon_bo_open(radeon_ws->bom, 0, size, alignment, - domain, 0); + radeon_buffer->bo = radeon_bo_open(radeon_ws->priv->bom, 0, size, + alignment, domain, 0); if (radeon_buffer->bo == NULL) { FREE(radeon_buffer); } @@ -93,6 +93,29 @@ static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws, return &radeon_buffer->base; } +static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws, + unsigned width, + unsigned height, + enum pipe_format format, + unsigned usage, + unsigned *stride) +{ + struct pipe_format_block block; + unsigned nblocksx, nblocksy, size; + + pf_get_block(format, &block); + + nblocksx = pf_get_nblocksx(&block, width); + nblocksy = pf_get_nblocksy(&block, height); + + /* Radeons enjoy things in multiples of 32. */ + /* XXX this can be 32 when POT */ + *stride = (nblocksx * block.size + 63) & ~63; + size = *stride * nblocksy; + + return radeon_buffer_create(ws, 64, usage, size); +} + static void radeon_buffer_del(struct pipe_buffer *buffer) { struct radeon_pipe_buffer *radeon_buffer = @@ -169,16 +192,22 @@ struct radeon_winsys* radeon_pipe_winsys(int fd) return NULL; } - bom = radeon_bo_manager_gem_ctor(fd); - radeon_ws->bom = bom; + radeon_ws->priv = CALLOC_STRUCT(radeon_winsys_priv); + if (radeon_ws->priv == NULL) { + FREE(radeon_ws); + return NULL; + } + + radeon_ws->priv->bom = radeon_bo_manager_gem_ctor(fd); radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer; radeon_ws->base.buffer_create = radeon_buffer_create; - radeon_ws->base.buffer_destroy = radeon_buffer_del; radeon_ws->base.user_buffer_create = radeon_buffer_user_create; + radeon_ws->base.surface_buffer_create = radeon_surface_buffer_create; radeon_ws->base.buffer_map = radeon_buffer_map; radeon_ws->base.buffer_unmap = radeon_buffer_unmap; + radeon_ws->base.buffer_destroy = radeon_buffer_del; radeon_ws->base.fence_reference = radeon_fence_reference; radeon_ws->base.fence_signalled = radeon_fence_signalled; diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h index 163422f296..ca8bbb3c11 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h @@ -41,6 +41,7 @@ #include "util/u_memory.h" #include "radeon_bo.h" +#include "radeon_cs.h" #include "radeon_drm.h" @@ -49,13 +50,30 @@ struct radeon_pipe_buffer { struct radeon_bo *bo; }; +#define RADEON_MAX_BOS 24 + +struct radeon_winsys_priv { + /* Radeon BO manager. */ + struct radeon_bo_manager* bom; + + /* Radeon BO space checker. */ + struct radeon_cs_space_check sc[RADEON_MAX_BOS]; + /* Current BO count. */ + unsigned bo_count; + + /* Radeon CS manager. */ + struct radeon_cs_manager* csm; + + /* Current CS. */ + struct radeon_cs* cs; +}; + struct radeon_winsys { /* Parent class. */ struct pipe_winsys base; - /* Radeon BO manager. - * This corresponds to void* radeon_winsys in r300_winsys. */ - struct radeon_bo_manager* bom; + /* This corresponds to void* radeon_winsys in r300_winsys. */ + struct radeon_winsys_priv* priv; }; struct radeon_winsys* radeon_pipe_winsys(int fb); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c index 3446654e28..5406d2bbea 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c @@ -30,8 +30,13 @@ #include "radeon_drm.h" +#ifdef DEBUG +#include "trace/trace_drm.h" +#endif + /* Create a pipe_screen. */ -struct pipe_screen* radeon_create_screen(int drmFB, int pciID) +struct pipe_screen* radeon_create_screen(int drmFB, + struct drm_create_screen_arg *arg) { struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB); @@ -68,7 +73,7 @@ struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, unsigned handle) { struct radeon_bo_manager* bom = - ((struct radeon_winsys*)screen->winsys)->bom; + ((struct radeon_winsys*)screen->winsys)->priv->bom; struct radeon_pipe_buffer* radeon_buffer; struct radeon_bo* bo = NULL; @@ -111,7 +116,11 @@ boolean radeon_global_handle_from_buffer(struct pipe_screen* screen, return TRUE; } +#ifdef DEBUG +struct drm_api hooks = { +#else struct drm_api drm_api_hooks = { +#endif .create_screen = radeon_create_screen, .create_context = radeon_create_context, /* XXX fix this */ diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h index ca2d98ed1a..049f9984db 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h @@ -40,7 +40,8 @@ #include "radeon_r300.h" #include "radeon_winsys_softpipe.h" -struct pipe_screen* radeon_create_screen(int drmFB, int pciID); +struct pipe_screen* radeon_create_screen(int drmFB, + struct drm_create_screen_arg *arg); struct pipe_context* radeon_create_context(struct pipe_screen* screen); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index 293b6c2d38..63aa3179ac 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -22,86 +22,224 @@ #include "radeon_r300.h" -static boolean radeon_r300_check_cs(struct radeon_cs* cs, int size) +static boolean radeon_r300_add_buffer(struct r300_winsys* winsys, + struct pipe_buffer* pbuffer, + uint32_t rd, + uint32_t wd) +{ + int i; + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + struct radeon_cs_space_check* sc = priv->sc; + struct radeon_bo* bo = ((struct radeon_pipe_buffer*)pbuffer)->bo; + + /* Check to see if this BO is already in line for validation; + * find a slot for it otherwise. */ + for (i = 0; i < priv->bo_count; i++) { + if (sc[i].bo == bo) { + sc[i].read_domains |= rd; + sc[i].write_domain |= wd; + return; + } + } + + if (priv->bo_count >= RADEON_MAX_BOS) { + /* Dohoho. Not falling for that one again. Request a flush. */ + return FALSE; + } + + sc[priv->bo_count].bo = bo; + sc[priv->bo_count].read_domains = rd; + sc[priv->bo_count].write_domain = wd; + priv->bo_count++; + + return TRUE; +} + +static boolean radeon_r300_validate(struct r300_winsys* winsys) +{ + int retval, i; + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + struct radeon_cs_space_check* sc = priv->sc; + + retval = radeon_cs_space_check(priv->cs, sc, priv->bo_count); + + if (retval == RADEON_CS_SPACE_OP_TO_BIG) { + /* We might as well HCF, since this is not going to fit in the card, + * period. */ + /* XXX just drop it on the floor instead */ + exit(1); + } else if (retval == RADEON_CS_SPACE_FLUSH) { + /* We must flush before more rendering can commence. */ + return TRUE; + } + + /* XXX should probably be its own function */ + for (i = 0; i < priv->bo_count; i++) { + if (sc[i].read_domains && sc[i].write_domain) { + /* Cute, cute. We need to flush first. */ + debug_printf("radeon: BO %p can't be read and written; " + "requesting flush.\n", sc[i].bo); + return TRUE; + } + } + + /* Things are fine, we can proceed as normal. */ + return FALSE; +} + +static boolean radeon_r300_check_cs(struct r300_winsys* winsys, int size) { /* XXX check size here, lazy ass! */ + /* XXX also validate buffers */ return TRUE; } -static void radeon_r300_write_cs_reloc(struct radeon_cs* cs, - struct pipe_buffer* pbuffer, - uint32_t rd, - uint32_t wd, - uint32_t flags) +static void radeon_r300_begin_cs(struct r300_winsys* winsys, + int size, + const char* file, + const char* function, + int line) { - radeon_cs_write_reloc(cs, ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags); + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + + radeon_cs_begin(priv->cs, size, file, function, line); } -static void radeon_r300_flush_cs(struct radeon_cs* cs) +static void radeon_r300_write_cs_dword(struct r300_winsys* winsys, + uint32_t dword) { + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + + radeon_cs_write_dword(priv->cs, dword); +} + +static void radeon_r300_write_cs_reloc(struct r300_winsys* winsys, + struct pipe_buffer* pbuffer, + uint32_t rd, + uint32_t wd, + uint32_t flags) +{ + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; int retval = 0; - retval = radeon_cs_emit(cs); + retval = radeon_cs_write_reloc(priv->cs, + ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags); + + if (retval) { + debug_printf("radeon: Relocation of %p (%d, %d, %d) failed!\n", + pbuffer, rd, wd, flags); + } +} + +static void radeon_r300_end_cs(struct r300_winsys* winsys, + const char* file, + const char* function, + int line) +{ + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + + radeon_cs_end(priv->cs, file, function, line); +} + +static void radeon_r300_flush_cs(struct r300_winsys* winsys) +{ + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + struct radeon_cs_space_check* sc = priv->sc; + int retval = 1; + + /* Emit the CS. */ + retval = radeon_cs_emit(priv->cs); if (retval) { debug_printf("radeon: Bad CS, dumping...\n"); - radeon_cs_print(cs, stderr); + radeon_cs_print(priv->cs, stderr); } - radeon_cs_erase(cs); + radeon_cs_erase(priv->cs); + + /* Clean out BOs. */ + memset(sc, 0, sizeof(struct radeon_cs_space_check) * RADEON_MAX_BOS); + priv->bo_count = 0; } /* Helper function to do the ioctls needed for setup and init. */ static void do_ioctls(struct r300_winsys* winsys, int fd) { - drm_radeon_getparam_t gp; - uint32_t target; + struct drm_radeon_gem_info gem_info = {0}; + drm_radeon_getparam_t gp = {0}; + struct drm_radeon_info info = {0}; + int target = 0; int retval; - /* XXX is this cast safe? */ - gp.value = (int*)⌖ + info.value = ⌖ + gp.value = ⌖ /* First, get PCI ID */ - gp.param = RADEON_PARAM_DEVICE_ID; - retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp)); + info.request = RADEON_INFO_DEVICE_ID; + retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)); if (retval) { - fprintf(stderr, "%s: Failed to get PCI ID, error number %d", + fprintf(stderr, "%s: New ioctl for PCI ID failed " + "(error number %d), trying classic ioctl...\n", __FUNCTION__, retval); - exit(1); + gp.param = RADEON_PARAM_DEVICE_ID; + retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, + sizeof(gp)); + if (retval) { + fprintf(stderr, "%s: Failed to get PCI ID, " + "error number %d\n", __FUNCTION__, retval); + exit(1); + } } winsys->pci_id = target; - /* Then, get the number of pixel pipes */ - gp.param = RADEON_PARAM_NUM_GB_PIPES; - retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp)); + /* Then, retrieve MM info */ + retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO, + &gem_info, sizeof(gem_info)); if (retval) { - fprintf(stderr, "%s: Failed to get GB pipe count, error number %d", + fprintf(stderr, "%s: Failed to get MM info, error number %d\n", __FUNCTION__, retval); exit(1); } - winsys->gb_pipes = target; - + winsys->gart_size = gem_info.gart_size; + /* XXX */ + winsys->vram_size = gem_info.vram_visible; } struct r300_winsys* radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys) { struct r300_winsys* winsys = CALLOC_STRUCT(r300_winsys); - struct radeon_cs_manager* csm; + struct radeon_winsys_priv* priv; if (winsys == NULL) { return NULL; } + priv = old_winsys->priv; + do_ioctls(winsys, fd); - csm = radeon_cs_manager_gem_ctor(fd); + priv->csm = radeon_cs_manager_gem_ctor(fd); + + priv->cs = radeon_cs_create(priv->csm, 1024 * 64 / 4); + radeon_cs_set_limit(priv->cs, + RADEON_GEM_DOMAIN_GTT, winsys->gart_size); + radeon_cs_set_limit(priv->cs, + RADEON_GEM_DOMAIN_VRAM, winsys->vram_size); - winsys->cs = radeon_cs_create(csm, 1024 * 64 / 4); + winsys->add_buffer = radeon_r300_add_buffer; + winsys->validate = radeon_r300_validate; winsys->check_cs = radeon_r300_check_cs; - winsys->begin_cs = radeon_cs_begin; - winsys->write_cs_dword = radeon_cs_write_dword; + winsys->begin_cs = radeon_r300_begin_cs; + winsys->write_cs_dword = radeon_r300_write_cs_dword; winsys->write_cs_reloc = radeon_r300_write_cs_reloc; - winsys->end_cs = radeon_cs_end; + winsys->end_cs = radeon_r300_end_cs; winsys->flush_cs = radeon_r300_flush_cs; memcpy(winsys, old_winsys, sizeof(struct radeon_winsys)); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h index 5c373cd084..a2e0e58248 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h @@ -31,5 +31,18 @@ #include "radeon_buffer.h" +/* protect us from bonghits */ +#ifndef RADEON_INFO_DEVICE_ID +#define RADEON_INFO_DEVICE_ID 0 +#endif +#ifndef DRM_RADEON_INFO +#define DRM_RADEON_INFO 0x1 +struct drm_radeon_info { + uint32_t request; + uint32_t pad; + uint64_t value; +}; +#endif + struct r300_winsys* radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c index 33f9ac15ab..f038bfa40e 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_winsys_softpipe.c @@ -31,42 +31,11 @@ #include "radeon_winsys_softpipe.h" -struct radeon_softpipe_winsys { - struct softpipe_winsys sp_winsys; - struct radeon_context *radeon_context; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean radeon_is_format_supported(struct softpipe_winsys *sws, - uint format) -{ - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_Z24S8_UNORM: - return TRUE; - default: - break; - } - return FALSE; -} - struct pipe_context *radeon_create_softpipe(struct pipe_winsys* winsys) { - struct softpipe_winsys *sp_winsys; struct pipe_screen *pipe_screen; pipe_screen = softpipe_create_screen(winsys); - sp_winsys = CALLOC_STRUCT(softpipe_winsys); - if (sp_winsys == NULL) { - return NULL; - } - - sp_winsys->is_format_supported = radeon_is_format_supported; - return softpipe_create(pipe_screen, - winsys, - sp_winsys); + return softpipe_create(pipe_screen); } diff --git a/src/gallium/winsys/drm/radeon/dri2/Makefile b/src/gallium/winsys/drm/radeon/dri/Makefile index f471c44349..a9889444de 100644 --- a/src/gallium/winsys/drm/radeon/dri2/Makefile +++ b/src/gallium/winsys/drm/radeon/dri/Makefile @@ -7,9 +7,10 @@ LIBNAME = radeon_dri.so MINIGLX_SOURCES = PIPE_DRIVERS = \ - $(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \ + $(TOP)/src/gallium/state_trackers/dri/libdridrm.a \ $(TOP)/src/gallium/winsys/drm/radeon/core/libradeonwinsys.a \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ $(TOP)/src/gallium/drivers/r300/libr300.a C_SOURCES = \ diff --git a/src/gallium/winsys/drm/radeon/dri2/SConscript b/src/gallium/winsys/drm/radeon/dri/SConscript index f2cdee97d9..f2cdee97d9 100644 --- a/src/gallium/winsys/drm/radeon/dri2/SConscript +++ b/src/gallium/winsys/drm/radeon/dri/SConscript diff --git a/src/gallium/winsys/drm/radeon/egl/Makefile b/src/gallium/winsys/drm/radeon/egl/Makefile index d989b3aa93..6a1448d1b9 100644 --- a/src/gallium/winsys/drm/radeon/egl/Makefile +++ b/src/gallium/winsys/drm/radeon/egl/Makefile @@ -8,6 +8,7 @@ PIPE_DRIVERS = \ $(TOP)/src/gallium/state_trackers/egl/libegldrm.a \ $(GALLIUMDIR)/winsys/drm/radeon/core/libradeonwinsys.a \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ $(TOP)/src/gallium/drivers/r300/libr300.a DRIVER_SOURCES = diff --git a/src/gallium/winsys/drm/radeon/xorg/Makefile b/src/gallium/winsys/drm/radeon/xorg/Makefile index 6ffd4a3a54..0241625f69 100644 --- a/src/gallium/winsys/drm/radeon/xorg/Makefile +++ b/src/gallium/winsys/drm/radeon/xorg/Makefile @@ -37,6 +37,6 @@ clean: install: $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + $(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) .PHONY = all clean install diff --git a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c index 6f77fbe5de..36824251f0 100644 --- a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c +++ b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c @@ -38,20 +38,17 @@ static Bool radeon_xorg_pci_probe(DriverPtr driver, intptr_t match_data); static const struct pci_id_match radeon_xorg_device_match[] = { - {0x1002, 0x5E4D, 0xffff, 0xffff, 0, 0, 0}, - {0x1002, 0x7249, 0xffff, 0xffff, 0, 0, 0}, + {0x1002, PCI_MATCH_ANY, PCI_MATCH_ANY, PCI_MATCH_ANY, 0, 0, 0}, {0, 0, 0}, }; static SymTabRec radeon_xorg_chipsets[] = { - {0x5E4D, "Radeon RV410 PCIE (X700)"}, - {0x7249, "Radeon R580 PCIE (X1900 XT)"}, + {PCI_MATCH_ANY, "ATI/AMD Radeon Graphics Chipset"}, {-1, NULL} }; static PciChipsets radeon_xorg_pci_devices[] = { - {0x5E4D, 0x5E4D, RES_SHARED_VGA}, - {0x7249, 0x7249, RES_SHARED_VGA}, + {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA}, {-1, -1, RES_UNDEFINED} }; diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile index 02ac47caa4..a33a50ec22 100644 --- a/src/gallium/winsys/egl_xlib/Makefile +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -60,7 +60,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS) -noprefix \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \ - --whole-archive $(LIBS) --no-whole-archive + -Wl,--whole-archive $(LIBS) -Wl,--no-whole-archive depend: $(ALL_SOURCES) @@ -74,7 +74,7 @@ depend: $(ALL_SOURCES) install: default $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) @if [ -e $(TOP)/$(LIB_DIR) ]; then \ - $(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \ + $(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \ fi diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 9ceb67d2ac..c10e3c00ff 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -359,7 +359,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, /* fall-through */ case EGL_OPENGL_API: /* create a softpipe context */ - ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); + ctx->pipe = softpipe_create(xdrv->screen); /* Now do xlib / state tracker inits here */ _eglConfigToContextModesRec(conf, &visual); ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); @@ -388,6 +388,7 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) /* API-dependent clean-up */ switch (context->Base.ClientAPI) { case EGL_OPENGL_ES_API: + case EGL_OPENVG_API: /* fall-through */ case EGL_OPENGL_API: st_destroy_context(context->Context); @@ -587,7 +588,9 @@ find_supported_apis(void) EGLint mask = 0; void *handle; - handle = dlopen(NULL, 0); + handle = dlopen(NULL, RTLD_LAZY | RTLD_LOCAL); + if(!handle) + return mask; if (dlsym(handle, "st_api_OpenGL_ES1")) mask |= EGL_OPENGL_ES_BIT; diff --git a/src/gallium/winsys/g3dvl/xsp_winsys.c b/src/gallium/winsys/g3dvl/xsp_winsys.c index 5b9fdb5c1f..698c2856a4 100644 --- a/src/gallium/winsys/g3dvl/xsp_winsys.c +++ b/src/gallium/winsys/g3dvl/xsp_winsys.c @@ -261,7 +261,7 @@ struct pipe_context* create_pipe_context(Display *display, int screen) } sp_screen = softpipe_create_screen((struct pipe_winsys*)xsp_winsys); - sp_pipe = softpipe_create(sp_screen, (struct pipe_winsys*)xsp_winsys, NULL); + sp_pipe = softpipe_create(sp_screen); xsp_context = calloc(1, sizeof(struct xsp_context)); xsp_context->display = display; diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript index aabab95f3a..86eb9ef55e 100644 --- a/src/gallium/winsys/gdi/SConscript +++ b/src/gallium/winsys/gdi/SConscript @@ -15,6 +15,7 @@ if env['platform'] == 'windows': 'gdi32', 'user32', 'kernel32', + 'ws2_32', ]) sources = [ diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c index d5d9431865..33826524d7 100644 --- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c @@ -263,7 +263,7 @@ gdi_softpipe_screen_create(void) static struct pipe_context * gdi_softpipe_context_create(struct pipe_screen *screen) { - return softpipe_create(screen, screen->winsys, NULL); + return softpipe_create(screen); } @@ -312,21 +312,21 @@ DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpReserved) { switch (fdwReason) { case DLL_PROCESS_ATTACH: - if (!st_init(&stw_winsys)) { + if (!stw_init(&stw_winsys)) { return FALSE; } - return st_init_thread(); + return stw_init_thread(); case DLL_THREAD_ATTACH: - return st_init_thread(); + return stw_init_thread(); case DLL_THREAD_DETACH: - st_cleanup_thread(); + stw_cleanup_thread(); break; case DLL_PROCESS_DETACH: - st_cleanup_thread(); - st_cleanup(); + stw_cleanup_thread(); + stw_cleanup(); break; } return TRUE; diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index bb187cc14a..522f6dc5ae 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -74,7 +74,7 @@ $(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME): $(XLIB_WINSYS_OBJECTS) $(LIBS) Makefil -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR)/gallium \ $(MKLIB_OPTIONS) $(XLIB_WINSYS_OBJECTS) \ - --start-group $(LIBS) --end-group $(GL_LIB_DEPS) + -Wl,--start-group $(LIBS) -Wl,--end-group $(GL_LIB_DEPS) depend: $(XLIB_WINSYS_SOURCES) @@ -90,7 +90,7 @@ install: default $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL @if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \ - $(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ + $(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ fi diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c index 762ebd9847..44b8464518 100644 --- a/src/gallium/winsys/xlib/xlib_softpipe.c +++ b/src/gallium/winsys/xlib/xlib_softpipe.c @@ -482,7 +482,7 @@ xlib_create_softpipe_context( struct pipe_screen *screen, { struct pipe_context *pipe; - pipe = softpipe_create(screen, screen->winsys, NULL); + pipe = softpipe_create(screen); if (pipe == NULL) goto fail; diff --git a/src/gallium/winsys/xlib/xlib_trace.c b/src/gallium/winsys/xlib/xlib_trace.c index 37095c5d8e..dbea655ab4 100644 --- a/src/gallium/winsys/xlib/xlib_trace.c +++ b/src/gallium/winsys/xlib/xlib_trace.c @@ -37,6 +37,7 @@ #include "trace/tr_screen.h" #include "trace/tr_context.h" +#include "trace/tr_texture.h" #include "pipe/p_screen.h" @@ -66,9 +67,11 @@ fail: } static struct pipe_context * -xlib_create_trace_context( struct pipe_screen *screen, +xlib_create_trace_context( struct pipe_screen *_screen, void *priv ) { + struct trace_screen *tr_scr = trace_screen( _screen ); + struct pipe_screen *screen = tr_scr->screen; struct pipe_context *pipe, *trace_pipe; pipe = xlib_softpipe_driver.create_pipe_context( screen, priv ); @@ -77,7 +80,7 @@ xlib_create_trace_context( struct pipe_screen *screen, /* Wrap it: */ - trace_pipe = trace_context_create(screen, pipe); + trace_pipe = trace_context_create(_screen, pipe); if (trace_pipe == NULL) goto fail; @@ -86,15 +89,18 @@ xlib_create_trace_context( struct pipe_screen *screen, return trace_pipe; fail: + if (pipe) + pipe->destroy( pipe ); return NULL; } static void xlib_trace_display_surface( struct xmesa_buffer *buffer, - struct pipe_surface *surf ) + struct pipe_surface *_surf ) { - /* ?? - */ + struct trace_surface *tr_surf = trace_surface( _surf ); + struct pipe_surface *surf = tr_surf->surface; + xlib_softpipe_driver.display_surface( buffer, surf ); } |