diff options
Diffstat (limited to 'src/mesa/drivers/dri/r200')
25 files changed, 2108 insertions, 5114 deletions
diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile index e9144ac75c..e593ed91c5 100644 --- a/src/mesa/drivers/dri/r200/Makefile +++ b/src/mesa/drivers/dri/r200/Makefile @@ -3,6 +3,8 @@ TOP = ../../../../.. include $(TOP)/configs/current +CFLAGS += $(RADEON_CFLAGS) + LIBNAME = r200_dri.so MINIGLX_SOURCES = server/radeon_dri.c @@ -11,25 +13,35 @@ ifeq ($(USING_EGL), 1) EGL_SOURCES = server/radeon_egl.c endif +RADEON_COMMON_SOURCES = \ + radeon_texture.c \ + radeon_common_context.c \ + radeon_common.c \ + radeon_dma.c \ + radeon_lock.c \ + radeon_bo_legacy.c \ + radeon_cs_legacy.c \ + radeon_mipmap_tree.c \ + radeon_span.c + + DRIVER_SOURCES = r200_context.c \ r200_ioctl.c \ - r200_lock.c \ r200_state.c \ r200_state_init.c \ r200_cmdbuf.c \ r200_pixel.c \ r200_tex.c \ - r200_texmem.c \ r200_texstate.c \ r200_tcl.c \ r200_swtcl.c \ - r200_span.c \ r200_maos.c \ r200_sanity.c \ r200_fragshader.c \ r200_vertprog.c \ radeon_screen.c \ - $(EGL_SOURCES) + $(EGL_SOURCES) \ + $(RADEON_COMMON_SOURCES) C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES) @@ -48,7 +60,29 @@ SYMLINKS = \ COMMON_SYMLINKS = \ radeon_chipset.h \ radeon_screen.c \ - radeon_screen.h + radeon_screen.h \ + radeon_bo_legacy.c \ + radeon_cs_legacy.c \ + radeon_bo_legacy.h \ + radeon_cs_legacy.h \ + radeon_bocs_wrapper.h \ + radeon_span.h \ + radeon_span.c \ + radeon_lock.c \ + radeon_lock.h \ + radeon_common.c \ + radeon_common_context.c \ + radeon_common_context.h \ + radeon_common.h \ + radeon_cmdbuf.h \ + radeon_mipmap_tree.c \ + radeon_mipmap_tree.h \ + radeon_texture.c \ + radeon_texture.h \ + radeon_dma.c \ + radeon_dma.h + +DRI_LIB_DEPS += $(RADEON_LDFLAGS) ##### TARGETS ##### diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c index e1633772a1..ae31bcb831 100644 --- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c +++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c @@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast/swrast.h" #include "main/simple_list.h" +#include "radeon_common.h" #include "r200_context.h" #include "r200_state.h" #include "r200_ioctl.h" @@ -45,18 +46,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_sanity.h" #include "radeon_reg.h" -static void print_state_atom( struct r200_state_atom *state ) -{ - int i; - - fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size); - - if (0 & R200_DEBUG & DEBUG_VERBOSE) - for (i = 0 ; i < state->cmd_size ; i++) - fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]); - -} - /* The state atoms will be emitted in the order they appear in the atom list, * so this step is important. */ @@ -64,141 +53,56 @@ void r200SetUpAtomList( r200ContextPtr rmesa ) { int i, mtu; - mtu = rmesa->glCtx->Const.MaxTextureUnits; - - make_empty_list(&rmesa->hw.atomlist); - rmesa->hw.atomlist.name = "atom-list"; - - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ctx ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.set ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lin ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msk ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpt ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vtx ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vap ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vte ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msc ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cst ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.zbs ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcl ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msl ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcg ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.grd ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.fog ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tam ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tf ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.atf ); + mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits; + + make_empty_list(&rmesa->radeon.hw.atomlist); + rmesa->radeon.hw.atomlist.name = "atom-list"; + + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ctx ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.set ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lin ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msk ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpt ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vtx ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vap ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vte ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msc ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cst ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.zbs ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcl ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msl ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcg ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.grd ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.fog ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tam ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tf ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.atf ); for (i = 0; i < mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tex[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i] ); for (i = 0; i < mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cube[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i] ); for (i = 0; i < 6; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pix[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[1] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pix[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[1] ); for (i = 0; i < 8; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lit[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i] ); for (i = 0; i < 3 + mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mat[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.eye ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.glt ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.eye ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.glt ); for (i = 0; i < 2; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mtl[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mtl[i] ); for (i = 0; i < 6; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ucp[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.spr ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ptp ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.prf ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pvs ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[1] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[1] ); -} - -static void r200SaveHwState( r200ContextPtr rmesa ) -{ - struct r200_state_atom *atom; - char * dest = rmesa->backup_store.cmd_buf; - - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->backup_store.cmd_used = 0; - - foreach( atom, &rmesa->hw.atomlist ) { - if ( atom->check( rmesa->glCtx, atom->idx ) ) { - int size = atom->cmd_size * 4; - memcpy( dest, atom->cmd, size); - dest += size; - rmesa->backup_store.cmd_used += size; - if (R200_DEBUG & DEBUG_STATE) - print_state_atom( atom ); - } - } - - assert( rmesa->backup_store.cmd_used <= R200_CMD_BUF_SZ ); - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "Returning to r200EmitState\n"); -} - -void r200EmitState( r200ContextPtr rmesa ) -{ - char *dest; - int mtu; - struct r200_state_atom *atom; - - if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->save_on_next_emit) { - r200SaveHwState(rmesa); - rmesa->save_on_next_emit = GL_FALSE; - } - - if (!rmesa->hw.is_dirty && !rmesa->hw.all_dirty) - return; - - mtu = rmesa->glCtx->Const.MaxTextureUnits; - - /* To avoid going across the entire set of states multiple times, just check - * for enough space for the case of emitting all state, and inline the - * r200AllocCmdBuf code here without all the checks. - */ - r200EnsureCmdBufSpace( rmesa, rmesa->hw.max_state_size ); - - /* we need to calculate dest after EnsureCmdBufSpace - as we may flush the buffer - airlied */ - dest = rmesa->store.cmd_buf + rmesa->store.cmd_used; - if (R200_DEBUG & DEBUG_STATE) { - foreach( atom, &rmesa->hw.atomlist ) { - if ( atom->dirty || rmesa->hw.all_dirty ) { - if ( atom->check( rmesa->glCtx, atom->idx ) ) - print_state_atom( atom ); - else - fprintf(stderr, "skip state %s\n", atom->name); - } - } - } - - foreach( atom, &rmesa->hw.atomlist ) { - if ( rmesa->hw.all_dirty ) - atom->dirty = GL_TRUE; - if ( atom->dirty ) { - if ( atom->check( rmesa->glCtx, atom->idx ) ) { - int size = atom->cmd_size * 4; - memcpy( dest, atom->cmd, size); - dest += size; - rmesa->store.cmd_used += size; - atom->dirty = GL_FALSE; - } - } - } - - assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ ); - - rmesa->hw.is_dirty = GL_FALSE; - rmesa->hw.all_dirty = GL_FALSE; + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.spr ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ptp ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.prf ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pvs ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[1] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[1] ); } /* Fire a section of the retained (indexed_verts) buffer as a regular @@ -209,50 +113,81 @@ void r200EmitVbufPrim( r200ContextPtr rmesa, GLuint vertex_nr ) { drm_radeon_cmd_header_t *cmd; + BATCH_LOCALS(&rmesa->radeon); assert(!(primitive & R200_VF_PRIM_WALK_IND)); - r200EmitState( rmesa ); + radeonEmitState(&rmesa->radeon); if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS)) fprintf(stderr, "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__, rmesa->store.cmd_used/4, primitive, vertex_nr); - - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VBUF_BUFSZ, - __FUNCTION__ ); - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = R200_CP_CMD_3D_DRAW_VBUF_2; - cmd[2].i = (primitive | - R200_VF_PRIM_WALK_LIST | - R200_VF_COLOR_ORDER_RGBA | - (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT)); + + BEGIN_BATCH(3); + OUT_BATCH_PACKET3_CLIP(R200_CP_CMD_3D_DRAW_VBUF_2, 0); + OUT_BATCH(primitive | R200_VF_PRIM_WALK_LIST | R200_VF_COLOR_ORDER_RGBA | + (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT)); + END_BATCH(); } +static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type) +{ + BATCH_LOCALS(&rmesa->radeon); + + if (vertex_count > 0) { + BEGIN_BATCH(8+2); + OUT_BATCH_PACKET3(R200_CP_CMD_3D_DRAW_INDX_2, 0); + OUT_BATCH(R200_VF_PRIM_WALK_IND | + ((vertex_count + 0) << 16) | + type); + + if (!rmesa->radeon.radeonScreen->kernel_mm) { + OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); + OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); + OUT_BATCH_RELOC(rmesa->tcl.elt_dma_offset, + rmesa->tcl.elt_dma_bo, + rmesa->tcl.elt_dma_offset, + RADEON_GEM_DOMAIN_GTT, 0, 0); + OUT_BATCH(vertex_count/2); + } else { + OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); + OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); + OUT_BATCH(rmesa->tcl.elt_dma_offset); + OUT_BATCH(vertex_count/2); + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.elt_dma_bo, + RADEON_GEM_DOMAIN_GTT, 0, 0); + } + END_BATCH(); + } +} -void r200FlushElts( r200ContextPtr rmesa ) +void r200FlushElts(GLcontext *ctx) { - int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start); + r200ContextPtr rmesa = R200_CONTEXT(ctx); int dwords; - int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 12)) / 2; + int nr, elt_used = rmesa->tcl.elt_used; if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); + fprintf(stderr, "%s %x %d\n", __FUNCTION__, rmesa->tcl.hw_primitive, elt_used); + + assert( rmesa->radeon.dma.flush == r200FlushElts ); + rmesa->radeon.dma.flush = NULL; + + elt_used = (elt_used + 2) & ~2; - assert( rmesa->dma.flush == r200FlushElts ); - rmesa->dma.flush = NULL; + nr = elt_used / 2; - /* Cope with odd number of elts: - */ - rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2; - dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4; + radeon_bo_unmap(rmesa->tcl.elt_dma_bo); - cmd[1] |= (dwords - 3) << 16; - cmd[2] |= nr << R200_VF_VERTEX_NUMBER_SHIFT; + r200FireEB(rmesa, nr, rmesa->tcl.hw_primitive); + + radeon_bo_unref(rmesa->tcl.elt_dma_bo); + rmesa->tcl.elt_dma_bo = NULL; if (R200_DEBUG & DEBUG_SYNC) { fprintf(stderr, "%s: Syncing\n", __FUNCTION__); - r200Finish( rmesa->glCtx ); + radeonFinish( rmesa->radeon.glCtx ); } } @@ -261,7 +196,6 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, GLuint primitive, GLuint min_nr ) { - drm_radeon_cmd_header_t *cmd; GLushort *retval; if (R200_DEBUG & DEBUG_IOCTL) @@ -269,30 +203,25 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, assert((primitive & R200_VF_PRIM_WALK_IND)); - r200EmitState( rmesa ); - - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, ELTS_BUFSZ(min_nr), - __FUNCTION__ ); - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = R200_CP_CMD_3D_DRAW_INDX_2; - cmd[2].i = (primitive | - R200_VF_PRIM_WALK_IND | - R200_VF_COLOR_ORDER_RGBA); + radeonEmitState(&rmesa->radeon); + rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, + 0, R200_ELT_BUF_SZ, 4, + RADEON_GEM_DOMAIN_GTT, 0); + rmesa->tcl.elt_dma_offset = 0; + rmesa->tcl.elt_used = min_nr * 2; + + radeon_bo_map(rmesa->tcl.elt_dma_bo, 1); + retval = rmesa->tcl.elt_dma_bo->ptr + rmesa->tcl.elt_dma_offset; - retval = (GLushort *)(cmd+3); if (R200_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: header 0x%x prim %x \n", - __FUNCTION__, - cmd[1].i, primitive); - - assert(!rmesa->dma.flush); - rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; - rmesa->dma.flush = r200FlushElts; + fprintf(stderr, "%s: header prim %x \n", + __FUNCTION__, primitive); - rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf; + assert(!rmesa->radeon.dma.flush); + rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; + rmesa->radeon.dma.flush = r200FlushElts; return retval; } @@ -300,129 +229,130 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, void r200EmitVertexAOS( r200ContextPtr rmesa, - GLuint vertex_size, - GLuint offset ) + GLuint vertex_size, + struct radeon_bo *bo, + GLuint offset ) { - drm_radeon_cmd_header_t *cmd; + BATCH_LOCALS(&rmesa->radeon); if (R200_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL)) fprintf(stderr, "%s: vertex_size 0x%x offset 0x%x \n", __FUNCTION__, vertex_size, offset); - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VERT_AOS_BUFSZ, - __FUNCTION__ ); - cmd[0].header.cmd_type = RADEON_CMD_PACKET3; - cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (2 << 16); - cmd[2].i = 1; - cmd[3].i = vertex_size | (vertex_size << 8); - cmd[4].i = offset; + BEGIN_BATCH(5); + OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, 2); + OUT_BATCH(1); + OUT_BATCH(vertex_size | (vertex_size << 8)); + OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); } - -void r200EmitAOS( r200ContextPtr rmesa, - struct r200_dma_region **component, - GLuint nr, - GLuint offset ) +void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset) { - drm_radeon_cmd_header_t *cmd; - int sz = AOS_BUFSZ(nr); + BATCH_LOCALS(&rmesa->radeon); + uint32_t voffset; + int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2; int i; - int *tmp; - - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s nr arrays: %d\n", __FUNCTION__, nr); - - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sz, __FUNCTION__ ); - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3; - cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (((sz / sizeof(int)) - 3) << 16); - cmd[2].i = nr; - tmp = &cmd[0].i; - cmd += 3; - - for (i = 0 ; i < nr ; i++) { - if (i & 1) { - cmd[0].i |= ((component[i]->aos_stride << 24) | - (component[i]->aos_size << 16)); - cmd[2].i = (component[i]->aos_start + - offset * component[i]->aos_stride * 4); - cmd += 3; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr, + offset); + + BEGIN_BATCH(sz+2+ (nr*2)); + OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, sz - 1); + OUT_BATCH(nr); + + + if (!rmesa->radeon.radeonScreen->kernel_mm) { + for (i = 0; i + 1 < nr; i += 2) { + OUT_BATCH((rmesa->tcl.aos[i].components << 0) | + (rmesa->tcl.aos[i].stride << 8) | + (rmesa->tcl.aos[i + 1].components << 16) | + (rmesa->tcl.aos[i + 1].stride << 24)); + + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[i].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[i+1].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); } - else { - cmd[0].i = ((component[i]->aos_stride << 8) | - (component[i]->aos_size << 0)); - cmd[1].i = (component[i]->aos_start + - offset * component[i]->aos_stride * 4); + + if (nr & 1) { + OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | + (rmesa->tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[nr - 1].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + } + } else { + for (i = 0; i + 1 < nr; i += 2) { + OUT_BATCH((rmesa->tcl.aos[i].components << 0) | + (rmesa->tcl.aos[i].stride << 8) | + (rmesa->tcl.aos[i + 1].components << 16) | + (rmesa->tcl.aos[i + 1].stride << 24)); + + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + OUT_BATCH(voffset); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + OUT_BATCH(voffset); + } + + if (nr & 1) { + OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | + (rmesa->tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH(voffset); + } + for (i = 0; i + 1 < nr; i += 2) { + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[i+0].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[i+1].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + } + if (nr & 1) { + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[nr-1].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); } } - - if (R200_DEBUG & DEBUG_VERTS) { - fprintf(stderr, "%s:\n", __FUNCTION__); - for (i = 0 ; i < sz ; i++) - fprintf(stderr, " %d: %x\n", i, tmp[i]); - } + END_BATCH(); } -void r200EmitBlit( r200ContextPtr rmesa, - GLuint color_fmt, - GLuint src_pitch, - GLuint src_offset, - GLuint dst_pitch, - GLuint dst_offset, - GLint srcx, GLint srcy, - GLint dstx, GLint dsty, - GLuint w, GLuint h ) +void r200FireAOS(r200ContextPtr rmesa, int vertex_count, int type) { - drm_radeon_cmd_header_t *cmd; + BATCH_LOCALS(&rmesa->radeon); - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n", - __FUNCTION__, - src_pitch, src_offset, srcx, srcy, - dst_pitch, dst_offset, dstx, dsty, - w, h); - - assert( (src_pitch & 63) == 0 ); - assert( (dst_pitch & 63) == 0 ); - assert( (src_offset & 1023) == 0 ); - assert( (dst_offset & 1023) == 0 ); - assert( w < (1<<16) ); - assert( h < (1<<16) ); - - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 8 * sizeof(int), - __FUNCTION__ ); - - - cmd[0].header.cmd_type = RADEON_CMD_PACKET3; - cmd[1].i = R200_CP_CMD_BITBLT_MULTI | (5 << 16); - cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL | - RADEON_GMC_DST_PITCH_OFFSET_CNTL | - RADEON_GMC_BRUSH_NONE | - (color_fmt << 8) | - RADEON_GMC_SRC_DATATYPE_COLOR | - RADEON_ROP3_S | - RADEON_DP_SRC_SOURCE_MEMORY | - RADEON_GMC_CLR_CMP_CNTL_DIS | - RADEON_GMC_WR_MSK_DIS ); - - cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10); - cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10); - cmd[5].i = (srcx << 16) | srcy; - cmd[6].i = (dstx << 16) | dsty; /* dst */ - cmd[7].i = (w << 16) | h; + BEGIN_BATCH(3); + OUT_BATCH_PACKET3(R200_CP_CMD_3D_DRAW_VBUF_2, 0); + OUT_BATCH(R200_VF_PRIM_WALK_LIST | (vertex_count << 16) | type); + END_BATCH(); } - -void r200EmitWait( r200ContextPtr rmesa, GLuint flags ) -{ - drm_radeon_cmd_header_t *cmd; - - assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) ); - - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 1 * sizeof(int), - __FUNCTION__ ); - cmd[0].i = 0; - cmd[0].wait.cmd_type = RADEON_CMD_WAIT; - cmd[0].wait.flags = flags; -} diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c index c06751516e..10a6362afd 100644 --- a/src/mesa/drivers/dri/r200/r200_context.c +++ b/src/mesa/drivers/dri/r200/r200_context.c @@ -54,7 +54,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_context.h" #include "r200_ioctl.h" #include "r200_state.h" -#include "r200_span.h" #include "r200_pixel.h" #include "r200_tex.h" #include "r200_swtcl.h" @@ -62,6 +61,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_maos.h" #include "r200_vertprog.h" +#include "radeon_span.h" + #define need_GL_ARB_vertex_program #define need_GL_ATI_fragment_shader #define need_GL_EXT_blend_minmax @@ -78,9 +79,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "vblank.h" #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -#ifndef R200_DEBUG -int R200_DEBUG = (0); -#endif /* Return various strings for glGetString(). */ @@ -89,8 +87,8 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name ) r200ContextPtr rmesa = R200_CONTEXT(ctx); static char buffer[128]; unsigned offset; - GLuint agp_mode = (rmesa->r200Screen->card_type == RADEON_CARD_PCI)? 0 : - rmesa->r200Screen->AGPMode; + GLuint agp_mode = (rmesa->radeon.radeonScreen->card_type == RADEON_CARD_PCI)? 0 : + rmesa->radeon.radeonScreen->AGPMode; switch ( name ) { case GL_VENDOR: @@ -101,7 +99,7 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name ) agp_mode ); sprintf( & buffer[ offset ], " %sTCL", - !(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE) + !(rmesa->radeon.TclFallback & R200_TCL_FALLBACK_TCL_DISABLE) ? "" : "NO-" ); return (GLubyte *)buffer; @@ -234,6 +232,40 @@ static const struct dri_debug_control debug_control[] = { NULL, 0 } }; +static void r200_get_lock(radeonContextPtr radeon) +{ + r200ContextPtr rmesa = (r200ContextPtr)radeon; + drm_radeon_sarea_t *sarea = radeon->sarea; + int i; + + R200_STATECHANGE( rmesa, ctx ); + if (rmesa->radeon.sarea->tiling_enabled) { + rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE; + } + else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE; + + if ( sarea->ctx_owner != rmesa->radeon.dri.hwContext ) { + sarea->ctx_owner = rmesa->radeon.dri.hwContext; + if (!radeon->radeonScreen->kernel_mm) + radeon_bo_legacy_texture_age(radeon->radeonScreen->bom); + } + +} + +static void r200_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa) +{ +} + + +static void r200_init_vtbl(radeonContextPtr radeon) +{ + radeon->vtbl.get_lock = r200_get_lock; + radeon->vtbl.update_viewport_offset = r200UpdateViewportOffset; + radeon->vtbl.emit_cs_header = r200_vtbl_emit_cs_header; + radeon->vtbl.swtcl_flush = r200_swtcl_flush; + radeon->vtbl.fallback = r200Fallback; +} + /* Create the device specific rendering context. */ @@ -245,9 +277,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private); struct dd_function_table functions; r200ContextPtr rmesa; - GLcontext *ctx, *shareCtx; + GLcontext *ctx; int i; - int tcl_mode, fthrottle_mode; + int tcl_mode; assert(glVisual); assert(driContextPriv); @@ -257,7 +289,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, rmesa = (r200ContextPtr) CALLOC( sizeof(*rmesa) ); if ( !rmesa ) return GL_FALSE; - + + r200_init_vtbl(&rmesa->radeon); /* init exp fog table data */ r200InitStaticFogData(); @@ -265,12 +298,12 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, * Do this here so that initialMaxAnisotropy is set before we create * the default textures. */ - driParseConfigFiles (&rmesa->optionCache, &screen->optionCache, + driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache, screen->driScreen->myNum, "r200"); - rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache, - "def_max_anisotropy"); + rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache, + "def_max_anisotropy"); - if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) { + if ( driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) { if ( sPriv->drm_version.minor < 13 ) fprintf( stderr, "DRM version 1.%d too old to support HyperZ, " "disabling.\n", sPriv->drm_version.minor ); @@ -291,59 +324,21 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, r200InitTextureFuncs(&functions); r200InitShaderFuncs(&functions); - /* Allocate and initialize the Mesa context */ - if (sharedContextPrivate) - shareCtx = ((r200ContextPtr) sharedContextPrivate)->glCtx; - else - shareCtx = NULL; - rmesa->glCtx = _mesa_create_context(glVisual, shareCtx, - &functions, (void *) rmesa); - if (!rmesa->glCtx) { - FREE(rmesa); - return GL_FALSE; - } - driContextPriv->driverPrivate = rmesa; - - /* Init r200 context data */ - rmesa->dri.context = driContextPriv; - rmesa->dri.screen = sPriv; - rmesa->dri.drawable = NULL; /* Set by XMesaMakeCurrent */ - rmesa->dri.hwContext = driContextPriv->hHWContext; - rmesa->dri.hwLock = &sPriv->pSAREA->lock; - rmesa->dri.fd = sPriv->fd; - rmesa->dri.drmMinor = sPriv->drm_version.minor; - - rmesa->r200Screen = screen; - rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA + - screen->sarea_priv_offset); - - - rmesa->dma.buf0_address = rmesa->r200Screen->buffers->list[0].address; - - (void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) ); - make_empty_list( & rmesa->swapped ); - - rmesa->nr_heaps = 1 /* screen->numTexHeaps */ ; - assert(rmesa->nr_heaps < RADEON_NR_TEX_HEAPS); - for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) { - rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa, - screen->texSize[i], - 12, - RADEON_NR_TEX_REGIONS, - (drmTextureRegionPtr)rmesa->sarea->tex_list[i], - & rmesa->sarea->tex_age[i], - & rmesa->swapped, - sizeof( r200TexObj ), - (destroy_texture_object_t *) r200DestroyTexObj ); + if (!radeonInitContext(&rmesa->radeon, &functions, + glVisual, driContextPriv, + sharedContextPrivate)) { + FREE(rmesa); + return GL_FALSE; } - rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache, + + rmesa->radeon.texture_depth = driQueryOptioni (&rmesa->radeon.optionCache, "texture_depth"); - if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB) - rmesa->texture_depth = ( screen->cpp == 4 ) ? + if (rmesa->radeon.texture_depth == DRI_CONF_TEXTURE_DEPTH_FB) + rmesa->radeon.texture_depth = ( screen->cpp == 4 ) ? DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16; - rmesa->swtcl.RenderIndex = ~0; - rmesa->hw.all_dirty = 1; + rmesa->radeon.swtcl.RenderIndex = ~0; + rmesa->radeon.hw.all_dirty = 1; /* Set the maximum texture size small enough that we can guarentee that * all texture units can bind a maximal texture and have all of them in @@ -351,29 +346,13 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, * setting allow larger textures. */ - ctx = rmesa->glCtx; - ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache, + ctx = rmesa->radeon.glCtx; + ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache, "texture_units"); ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits; ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits; - i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures"); - - driCalculateMaxTextureLevels( rmesa->texture_heaps, - rmesa->nr_heaps, - & ctx->Const, - 4, - 11, /* max 2D texture size is 2048x2048 */ -#if ENABLE_HW_3D_TEXTURE - 8, /* max 3D texture size is 256^3 */ -#else - 0, /* 3D textures unsupported */ -#endif - 11, /* max cube texture size is 2048x2048 */ - 11, /* max texture rectangle size is 2048x2048 */ - 12, - GL_FALSE, - i ); + i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures"); ctx->Const.MaxTextureMaxAnisotropy = 16.0; @@ -383,7 +362,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, ctx->Const.MinPointSizeAA = 1.0; ctx->Const.MaxPointSizeAA = 1.0; ctx->Const.PointSizeGranularity = 0.0625; - if (rmesa->r200Screen->drmSupportsPointSprites) + if (rmesa->radeon.radeonScreen->drmSupportsPointSprites) ctx->Const.MaxPointSize = 2047.0; else ctx->Const.MaxPointSize = 1.0; @@ -439,32 +418,32 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, _math_matrix_set_identity( &rmesa->tmpmat ); driInitExtensions( ctx, card_extensions, GL_TRUE ); - if (!(rmesa->r200Screen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) { + if (!(rmesa->radeon.radeonScreen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) { /* yuv textures don't work with some chips - R200 / rv280 okay so far others get the bit ordering right but don't actually do YUV-RGB conversion */ _mesa_enable_extension( ctx, "GL_MESA_ycbcr_texture" ); } - if (rmesa->glCtx->Mesa_DXTn) { + if (rmesa->radeon.glCtx->Mesa_DXTn) { _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" ); _mesa_enable_extension( ctx, "GL_S3_s3tc" ); } - else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) { + else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) { _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" ); } - if (rmesa->r200Screen->drmSupportsCubeMapsR200) + if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200) _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" ); - if (rmesa->r200Screen->drmSupportsBlendColor) { + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) { driInitExtensions( ctx, blend_extensions, GL_FALSE ); } - if(rmesa->r200Screen->drmSupportsVertexProgram) + if(rmesa->radeon.radeonScreen->drmSupportsVertexProgram) driInitSingleExtension( ctx, ARB_vp_extension ); - if(driQueryOptionb(&rmesa->optionCache, "nv_vertex_program")) + if(driQueryOptionb(&rmesa->radeon.optionCache, "nv_vertex_program")) driInitSingleExtension( ctx, NV_vp_extension ); - if ((ctx->Const.MaxTextureUnits == 6) && rmesa->r200Screen->drmSupportsFragShader) + if ((ctx->Const.MaxTextureUnits == 6) && rmesa->radeon.radeonScreen->drmSupportsFragShader) driInitSingleExtension( ctx, ATI_fs_extension ); - if (rmesa->r200Screen->drmSupportsPointSprites) + if (rmesa->radeon.radeonScreen->drmSupportsPointSprites) driInitExtensions( ctx, point_extensions, GL_FALSE ); #if 0 r200InitDriverFuncs( ctx ); @@ -474,33 +453,15 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, #endif /* plug in a few more device driver functions */ /* XXX these should really go right after _mesa_init_driver_functions() */ + radeonInitSpanFuncs( ctx ); r200InitPixelFuncs( ctx ); - r200InitSpanFuncs( ctx ); r200InitTnlFuncs( ctx ); r200InitState( rmesa ); r200InitSwtcl( ctx ); - fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode"); - rmesa->iw.irq_seq = -1; - rmesa->irqsEmitted = 0; - rmesa->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && - rmesa->r200Screen->irq); - - rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - - if (!rmesa->do_irqs) - fprintf(stderr, - "IRQ's not enabled, falling back to %s: %d %d\n", - rmesa->do_usleeps ? "usleeps" : "busy waits", - fthrottle_mode, - rmesa->r200Screen->irq); - rmesa->prefer_gart_client_texturing = (getenv("R200_GART_CLIENT_TEXTURES") != 0); - (*sPriv->systemTime->getUST)( & rmesa->swap_ust ); - - #if DO_DEBUG R200_DEBUG = driParseDebugString( getenv( "R200_DEBUG" ), debug_control ); @@ -508,18 +469,18 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, debug_control ); #endif - tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode"); - if (driQueryOptionb(&rmesa->optionCache, "no_rast")) { + tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode"); + if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) { fprintf(stderr, "disabling 3D acceleration\n"); FALLBACK(rmesa, R200_FALLBACK_DISABLE, 1); } else if (tcl_mode == DRI_CONF_TCL_SW || getenv("R200_NO_TCL") || - !(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) { - if (rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL) { - rmesa->r200Screen->chip_flags &= ~RADEON_CHIPSET_TCL; + !(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) { + if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) { + rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL; fprintf(stderr, "Disabling HW TCL support\n"); } - TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1); + TCL_FALLBACK(rmesa->radeon.glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1); } return GL_TRUE; @@ -538,55 +499,33 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv ) /* check if we're deleting the currently bound context */ if (rmesa == current) { - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); _mesa_make_current(NULL, NULL, NULL); } /* Free r200 context resources */ assert(rmesa); /* should never be null */ if ( rmesa ) { - GLboolean release_texture_heaps; - - release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1); - _swsetup_DestroyContext( rmesa->glCtx ); - _tnl_DestroyContext( rmesa->glCtx ); - _vbo_DestroyContext( rmesa->glCtx ); - _swrast_DestroyContext( rmesa->glCtx ); + _swsetup_DestroyContext( rmesa->radeon.glCtx ); + _tnl_DestroyContext( rmesa->radeon.glCtx ); + _vbo_DestroyContext( rmesa->radeon.glCtx ); + _swrast_DestroyContext( rmesa->radeon.glCtx ); - r200DestroySwtcl( rmesa->glCtx ); - r200ReleaseArrays( rmesa->glCtx, ~0 ); + r200DestroySwtcl( rmesa->radeon.glCtx ); + r200ReleaseArrays( rmesa->radeon.glCtx, ~0 ); - if (rmesa->dma.current.buf) { - r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ ); - r200FlushCmdBuf( rmesa, __FUNCTION__ ); + if (rmesa->radeon.dma.current) { + radeonReleaseDmaRegion( &rmesa->radeon ); + rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); } - if (rmesa->state.scissor.pClipRects) { - FREE(rmesa->state.scissor.pClipRects); - rmesa->state.scissor.pClipRects = NULL; - } - - if ( release_texture_heaps ) { - /* This share group is about to go away, free our private - * texture object data. - */ - int i; - - for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) { - driDestroyTextureHeap( rmesa->texture_heaps[ i ] ); - rmesa->texture_heaps[ i ] = NULL; - } - - assert( is_empty_list( & rmesa->swapped ) ); + if (rmesa->radeon.state.scissor.pClipRects) { + FREE(rmesa->radeon.state.scissor.pClipRects); + rmesa->radeon.state.scissor.pClipRects = NULL; } - /* free the Mesa context */ - rmesa->glCtx->DriverCtx = NULL; - _mesa_destroy_context( rmesa->glCtx ); - - /* free the option cache */ - driDestroyOptionCache (&rmesa->optionCache); + radeonCleanupContext(&rmesa->radeon); FREE( rmesa ); } @@ -594,107 +533,6 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv ) - -void -r200SwapBuffers( __DRIdrawablePrivate *dPriv ) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - r200ContextPtr rmesa; - GLcontext *ctx; - rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = rmesa->glCtx; - if (ctx->Visual.doubleBufferMode) { - _mesa_notifySwapBuffers( ctx ); /* flush pending rendering comands */ - if ( rmesa->doPageFlip ) { - r200PageFlip( dPriv ); - } - else { - r200CopyBuffer( dPriv, NULL ); - } - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__); - } -} - -void -r200CopySubBuffer( __DRIdrawablePrivate *dPriv, - int x, int y, int w, int h ) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - r200ContextPtr rmesa; - GLcontext *ctx; - rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = rmesa->glCtx; - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - rect.x1 = x + dPriv->x; - rect.y1 = (dPriv->h - y - h) + dPriv->y; - rect.x2 = rect.x1 + w; - rect.y2 = rect.y1 + h; - _mesa_notifySwapBuffers( ctx ); /* flush pending rendering comands */ - r200CopyBuffer( dPriv, &rect ); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__); - } -} - -/* Force the context `c' to be the current context and associate with it - * buffer `b'. - */ -GLboolean -r200MakeCurrent( __DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv ) -{ - if ( driContextPriv ) { - r200ContextPtr newCtx = - (r200ContextPtr) driContextPriv->driverPrivate; - - if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)newCtx->glCtx); - - newCtx->dri.readable = driReadPriv; - - if ( newCtx->dri.drawable != driDrawPriv || - newCtx->lastStamp != driDrawPriv->lastStamp ) { - if (driDrawPriv->swap_interval == (unsigned)-1) { - driDrawPriv->vblFlags = (newCtx->r200Screen->irq != 0) - ? driGetDefaultVBlankFlags(&newCtx->optionCache) - : VBLANK_FLAG_NO_IRQ; - - driDrawableInitVBlank( driDrawPriv ); - } - - newCtx->dri.drawable = driDrawPriv; - - r200SetCliprects(newCtx); - r200UpdateViewportOffset( newCtx->glCtx ); - } - - _mesa_make_current( newCtx->glCtx, - (GLframebuffer *) driDrawPriv->driverPrivate, - (GLframebuffer *) driReadPriv->driverPrivate ); - - _mesa_update_state( newCtx->glCtx ); - r200ValidateState( newCtx->glCtx ); - - } else { - if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx is null\n", __FUNCTION__); - _mesa_make_current( NULL, NULL, NULL ); - } - - if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "End %s\n", __FUNCTION__); - return GL_TRUE; -} - /* Force the context `c' to be unbound from its buffer. */ GLboolean @@ -703,7 +541,7 @@ r200UnbindContext( __DRIcontextPrivate *driContextPriv ) r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate; if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->glCtx); + fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->radeon.glCtx); return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h index 14a1dda46a..f7bad2a241 100644 --- a/src/mesa/drivers/dri/r200/r200_context.h +++ b/src/mesa/drivers/dri/r200/r200_context.h @@ -53,51 +53,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #error This driver requires a newer libdrm to compile #endif +#include "radeon_screen.h" +#include "radeon_common.h" + +#include "radeon_lock.h" + struct r200_context; typedef struct r200_context r200ContextRec; typedef struct r200_context *r200ContextPtr; -/* This union is used to avoid warnings/miscompilation - with float to uint32_t casts due to strict-aliasing */ -typedef union { GLfloat f; uint32_t ui32; } float_ui32_type; - -#include "r200_lock.h" -#include "radeon_screen.h" #include "main/mm.h" -/* Flags for software fallback cases */ -/* See correponding strings in r200_swtcl.c */ -#define R200_FALLBACK_TEXTURE 0x01 -#define R200_FALLBACK_DRAW_BUFFER 0x02 -#define R200_FALLBACK_STENCIL 0x04 -#define R200_FALLBACK_RENDER_MODE 0x08 -#define R200_FALLBACK_DISABLE 0x10 -#define R200_FALLBACK_BORDER_MODE 0x20 - -/* The blit width for texture uploads - */ -#define BLIT_WIDTH_BYTES 1024 - -/* Use the templated vertex format: - */ -#define COLOR_IS_RGBA -#define TAG(x) r200##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -typedef void (*r200_tri_func)( r200ContextPtr, - r200Vertex *, - r200Vertex *, - r200Vertex * ); - -typedef void (*r200_line_func)( r200ContextPtr, - r200Vertex *, - r200Vertex * ); - -typedef void (*r200_point_func)( r200ContextPtr, - r200Vertex * ); - - struct r200_vertex_program { struct gl_vertex_program mesa_program; /* Must be first */ int translated; @@ -112,93 +78,11 @@ struct r200_vertex_program { int fogmode; }; -struct r200_colorbuffer_state { - GLuint clear; -#if 000 - GLint drawOffset, drawPitch; -#endif - int roundEnable; -}; - - -struct r200_depthbuffer_state { - GLuint clear; - GLfloat scale; -}; - -#if 000 -struct r200_pixel_state { - GLint readOffset, readPitch; -}; -#endif - -struct r200_scissor_state { - drm_clip_rect_t rect; - GLboolean enabled; - - GLuint numClipRects; /* Cliprects active */ - GLuint numAllocedClipRects; /* Cliprects available */ - drm_clip_rect_t *pClipRects; -}; - -struct r200_stencilbuffer_state { - GLboolean hwBuffer; - GLuint clear; /* rb3d_stencilrefmask value */ -}; - -struct r200_stipple_state { - GLuint mask[32]; -}; - - - -#define TEX_0 0x1 -#define TEX_1 0x2 -#define TEX_2 0x4 -#define TEX_3 0x8 -#define TEX_4 0x10 -#define TEX_5 0x20 -#define TEX_ALL 0x3f - -typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr; - -/* Texture object in locally shared texture space. - */ -struct r200_tex_obj { - driTextureObject base; - - GLuint bufAddr; /* Offset to start of locally - shared texture block */ - - GLuint dirty_state; /* Flags (1 per texunit) for - whether or not this texobj - has dirty hardware state - (pp_*) that needs to be - brought into the - texunit. */ - - drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS]; - /* Six, for the cube faces */ - GLboolean image_override; /* Image overridden by GLX_EXT_tfp */ - - GLuint pp_txfilter; /* hardware register values */ - GLuint pp_txformat; - GLuint pp_txformat_x; - GLuint pp_txoffset; /* Image location in texmem. - All cube faces follow. */ - GLuint pp_txsize; /* npot only */ - GLuint pp_txpitch; /* npot only */ - GLuint pp_border_color; - GLuint pp_cubic_faces; /* cube face 1,2,3,4 log2 sizes */ - - GLboolean border_fallback; - - GLuint tile_bits; /* hw texture tile bits used on this texture */ -}; +#define R200_TEX_ALL 0x3f struct r200_texture_env_state { - r200TexObjPtr texobj; + radeonTexObjPtr texobj; GLuint outputreg; GLuint unitneeded; }; @@ -210,19 +94,6 @@ struct r200_texture_state { }; -struct r200_state_atom { - struct r200_state_atom *next, *prev; - const char *name; /* for debug */ - int cmd_size; /* size in bytes */ - GLuint idx; - int *cmd; /* one or more cmd's */ - int *lastcmd; /* one or more cmd's */ - GLboolean dirty; - GLboolean (*check)( GLcontext *, int ); /* is this state active? */ -}; - - - /* Trying to keep these relatively short as the variables are becoming * extravagently long. Drop the driver name prefix off the front of * everything - I think we know which driver we're in by now, and keep the @@ -597,181 +468,85 @@ struct r200_state_atom { struct r200_hw_state { - /* Head of the linked list of state atoms. */ - struct r200_state_atom atomlist; - /* Hardware state, stored as cmdbuf commands: * -- Need to doublebuffer for * - reviving state after loss of context * - eliding noop statechange loops? (except line stipple count) */ - struct r200_state_atom ctx; - struct r200_state_atom set; - struct r200_state_atom vte; - struct r200_state_atom lin; - struct r200_state_atom msk; - struct r200_state_atom vpt; - struct r200_state_atom vap; - struct r200_state_atom vtx; - struct r200_state_atom tcl; - struct r200_state_atom msl; - struct r200_state_atom tcg; - struct r200_state_atom msc; - struct r200_state_atom cst; - struct r200_state_atom tam; - struct r200_state_atom tf; - struct r200_state_atom tex[6]; - struct r200_state_atom cube[6]; - struct r200_state_atom zbs; - struct r200_state_atom mtl[2]; - struct r200_state_atom mat[9]; - struct r200_state_atom lit[8]; /* includes vec, scl commands */ - struct r200_state_atom ucp[6]; - struct r200_state_atom pix[6]; /* pixshader stages */ - struct r200_state_atom eye; /* eye pos */ - struct r200_state_atom grd; /* guard band clipping */ - struct r200_state_atom fog; - struct r200_state_atom glt; - struct r200_state_atom prf; - struct r200_state_atom afs[2]; - struct r200_state_atom pvs; - struct r200_state_atom vpi[2]; - struct r200_state_atom vpp[2]; - struct r200_state_atom atf; - struct r200_state_atom spr; - struct r200_state_atom ptp; - - int max_state_size; /* Number of bytes necessary for a full state emit. */ - GLboolean is_dirty, all_dirty; + struct radeon_state_atom ctx; + struct radeon_state_atom set; + struct radeon_state_atom vte; + struct radeon_state_atom lin; + struct radeon_state_atom msk; + struct radeon_state_atom vpt; + struct radeon_state_atom vap; + struct radeon_state_atom vtx; + struct radeon_state_atom tcl; + struct radeon_state_atom msl; + struct radeon_state_atom tcg; + struct radeon_state_atom msc; + struct radeon_state_atom cst; + struct radeon_state_atom tam; + struct radeon_state_atom tf; + struct radeon_state_atom tex[6]; + struct radeon_state_atom cube[6]; + struct radeon_state_atom zbs; + struct radeon_state_atom mtl[2]; + struct radeon_state_atom mat[9]; + struct radeon_state_atom lit[8]; /* includes vec, scl commands */ + struct radeon_state_atom ucp[6]; + struct radeon_state_atom pix[6]; /* pixshader stages */ + struct radeon_state_atom eye; /* eye pos */ + struct radeon_state_atom grd; /* guard band clipping */ + struct radeon_state_atom fog; + struct radeon_state_atom glt; + struct radeon_state_atom prf; + struct radeon_state_atom afs[2]; + struct radeon_state_atom pvs; + struct radeon_state_atom vpi[2]; + struct radeon_state_atom vpp[2]; + struct radeon_state_atom atf; + struct radeon_state_atom spr; + struct radeon_state_atom ptp; }; struct r200_state { /* Derived state for internal purposes: */ - struct r200_colorbuffer_state color; - struct r200_depthbuffer_state depth; -#if 00 - struct r200_pixel_state pixel; -#endif - struct r200_scissor_state scissor; - struct r200_stencilbuffer_state stencil; - struct r200_stipple_state stipple; + struct radeon_stipple_state stipple; struct r200_texture_state texture; GLuint envneeded; }; -/* Need refcounting on dma buffers: - */ -struct r200_dma_buffer { - int refcount; /* the number of retained regions in buf */ - drmBufPtr buf; -}; - -#define GET_START(rvb) (rmesa->r200Screen->gart_buffer_offset + \ - (rvb)->address - rmesa->dma.buf0_address + \ - (rvb)->start) - -/* A retained region, eg vertices for indexed vertices. - */ -struct r200_dma_region { - struct r200_dma_buffer *buf; - char *address; /* == buf->address */ - int start, end, ptr; /* offsets from start of buf */ - int aos_start; - int aos_stride; - int aos_size; -}; - - -struct r200_dma { - /* Active dma region. Allocations for vertices and retained - * regions come from here. Also used for emitting random vertices, - * these may be flushed by calling flush_current(); - */ - struct r200_dma_region current; - - void (*flush)( r200ContextPtr ); - - char *buf0_address; /* start of buf[0], for index calcs */ - GLuint nr_released_bufs; /* flush after so many buffers released */ -}; - -struct r200_dri_mirror { - __DRIcontextPrivate *context; /* DRI context */ - __DRIscreenPrivate *screen; /* DRI screen */ - __DRIdrawablePrivate *drawable; /* DRI drawable bound to this ctx */ - __DRIdrawablePrivate *readable; /* DRI readable bound to this ctx */ - - drm_context_t hwContext; - drm_hw_lock_t *hwLock; - int fd; - int drmMinor; -}; - - #define R200_CMD_BUF_SZ (16*1024) -struct r200_store { - GLuint statenr; - GLuint primnr; - char cmd_buf[R200_CMD_BUF_SZ]; - int cmd_used; - int elts_start; -}; - - +#define R200_ELT_BUF_SZ (16*1024) /* r200_tcl.c */ struct r200_tcl_info { GLuint hw_primitive; /* hw can handle 12 components max */ - struct r200_dma_region *aos_components[12]; + struct radeon_aos aos[12]; GLuint nr_aos_components; GLuint *Elts; - struct r200_dma_region indexed_verts; - struct r200_dma_region vertex_data[15]; + struct radeon_bo *elt_dma_bo; + int elt_dma_offset; /** Offset into this buffer object, in bytes */ + int elt_used; + }; /* r200_swtcl.c */ struct r200_swtcl_info { - GLuint RenderIndex; - - /** - * Size of a hardware vertex. This is calculated when \c ::vertex_attrs is - * installed in the Mesa state vector. - */ - GLuint vertex_size; - - /** - * Attributes instructing the Mesa TCL pipeline where / how to put vertex - * data in the hardware buffer. - */ - struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; - - /** - * Number of elements of \c ::vertex_attrs that are actually used. - */ - GLuint vertex_attr_count; - /** - * Cached pointer to the buffer where Mesa will store vertex data. - */ - GLubyte *verts; - /* Fallback rasterization functions - */ - r200_point_func draw_point; - r200_line_func draw_line; - r200_tri_func draw_tri; - - GLuint hw_primitive; - GLenum render_primitive; - GLuint numverts; + radeon_point_func draw_point; + radeon_line_func draw_line; + radeon_tri_func draw_tri; /** * Offset of the 4UB color data within a hardware (swtcl) vertex. @@ -787,27 +562,10 @@ struct r200_swtcl_info { * Should Mesa project vertex data or will the hardware do it? */ GLboolean needproj; - - struct r200_dma_region indexed_verts; -}; - - -struct r200_ioctl { - GLuint vertex_offset; - GLuint vertex_size; }; -#define R200_MAX_PRIMS 64 - - - -struct r200_prim { - GLuint start; - GLuint end; - GLuint prim; -}; /* A maximum total of 29 elements per vertex: 3 floats for position, 3 * floats for normal, 4 floats for color, 4 bytes for secondary color, @@ -822,9 +580,8 @@ struct r200_prim { #define R200_MAX_VERTEX_SIZE ((3*6)+11) - struct r200_context { - GLcontext *glCtx; /* Mesa context */ + struct radeon_context radeon; /* Driver and hardware state management */ @@ -832,56 +589,15 @@ struct r200_context { struct r200_state state; struct r200_vertex_program *curr_vp_hw; - /* Texture object bookkeeping - */ - unsigned nr_heaps; - driTexHeap * texture_heaps[ RADEON_NR_TEX_HEAPS ]; - driTextureObject swapped; - int texture_depth; - float initialMaxAnisotropy; - - /* Rasterization and vertex state: - */ - GLuint TclFallback; - GLuint Fallback; - GLuint NewGLState; - DECLARE_RENDERINPUTS(tnl_index_bitset); /* index of bits for last tnl_install_attrs */ - /* Vertex buffers */ - struct r200_ioctl ioctl; - struct r200_dma dma; - struct r200_store store; - /* A full state emit as of the first state emit in the main store, in case - * the context is lost. - */ - struct r200_store backup_store; - - /* Page flipping - */ - GLuint doPageFlip; - - /* Busy waiting - */ - GLuint do_usleeps; - GLuint do_irqs; - GLuint irqsEmitted; - drm_radeon_irq_wait_t iw; + struct radeon_ioctl ioctl; + struct radeon_store store; /* Clientdata textures; */ GLuint prefer_gart_client_texturing; - /* Drawable, cliprect and scissor information - */ - GLuint numClipRects; /* Cliprects for the draw buffer */ - drm_clip_rect_t *pClipRects; - unsigned int lastStamp; - GLboolean lost_context; - GLboolean save_on_next_emit; - radeonScreenPtr r200Screen; /* Screen private DRI data */ - drm_radeon_sarea_t *sarea; /* Private SAREA data */ - /* TCL stuff */ GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS]; @@ -893,15 +609,6 @@ struct r200_context { GLuint TexGenCompSel; GLmatrix tmpmat; - /* buffer swap - */ - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; - - /* r200_tcl.c */ struct r200_tcl_info tcl; @@ -910,45 +617,34 @@ struct r200_context { */ struct r200_swtcl_info swtcl; - /* Mirrors of some DRI state - */ - struct r200_dri_mirror dri; - - /* Configuration cache - */ - driOptionCache optionCache; - GLboolean using_hyperz; GLboolean texmicrotile; struct ati_fragment_shader *afs_loaded; -}; -#define R200_CONTEXT(ctx) ((r200ContextPtr)(ctx->DriverCtx)) + struct { + struct gl_fragment_program *bitmap_fp; + struct gl_vertex_program *passthrough_vp; + struct gl_fragment_program *saved_fp; + GLboolean saved_fp_enable; + struct gl_vertex_program *saved_vp; + GLboolean saved_vp_enable; -static INLINE GLuint r200PackColor( GLuint cpp, - GLubyte r, GLubyte g, - GLubyte b, GLubyte a ) -{ - switch ( cpp ) { - case 2: - return PACK_COLOR_565( r, g, b ); - case 4: - return PACK_COLOR_8888( a, r, g, b ); - default: - return 0; - } -} + GLint saved_vp_x, saved_vp_y; + GLsizei saved_vp_width, saved_vp_height; + GLenum saved_matrix_mode; + } meta; + +}; + +#define R200_CONTEXT(ctx) ((r200ContextPtr)(ctx->DriverCtx)) extern void r200DestroyContext( __DRIcontextPrivate *driContextPriv ); extern GLboolean r200CreateContext( const __GLcontextModes *glVisual, __DRIcontextPrivate *driContextPriv, void *sharedContextPrivate); -extern void r200SwapBuffers( __DRIdrawablePrivate *dPriv ); -extern void r200CopySubBuffer( __DRIdrawablePrivate * dPriv, - int x, int y, int w, int h ); extern GLboolean r200MakeCurrent( __DRIcontextPrivate *driContextPriv, __DRIdrawablePrivate *driDrawPriv, __DRIdrawablePrivate *driReadPriv ); @@ -957,28 +653,9 @@ extern GLboolean r200UnbindContext( __DRIcontextPrivate *driContextPriv ); /* ================================================================ * Debugging: */ -#define DO_DEBUG 1 -#if DO_DEBUG -extern int R200_DEBUG; -#else -#define R200_DEBUG 0 -#endif +#define R200_DEBUG RADEON_DEBUG + -#define DEBUG_TEXTURE 0x001 -#define DEBUG_STATE 0x002 -#define DEBUG_IOCTL 0x004 -#define DEBUG_PRIMS 0x008 -#define DEBUG_VERTS 0x010 -#define DEBUG_FALLBACKS 0x020 -#define DEBUG_VFMT 0x040 -#define DEBUG_CODEGEN 0x080 -#define DEBUG_VERBOSE 0x100 -#define DEBUG_DRI 0x200 -#define DEBUG_DMA 0x400 -#define DEBUG_SANITY 0x800 -#define DEBUG_SYNC 0x1000 -#define DEBUG_PIXEL 0x2000 -#define DEBUG_MEMORY 0x4000 #endif /* __R200_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r200/r200_fragshader.c b/src/mesa/drivers/dri/r200/r200_fragshader.c index d514b28219..85c1b7bdd1 100644 --- a/src/mesa/drivers/dri/r200/r200_fragshader.c +++ b/src/mesa/drivers/dri/r200/r200_fragshader.c @@ -522,7 +522,7 @@ static void r200UpdateFSConstants( GLcontext *ctx ) CLAMPED_FLOAT_TO_UBYTE(con_byte[2], ctx->ATIFragmentShader.GlobalConstants[i][2]); CLAMPED_FLOAT_TO_UBYTE(con_byte[3], ctx->ATIFragmentShader.GlobalConstants[i][3]); } - rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = r200PackColor ( + rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = radeonPackColor ( 4, con_byte[0], con_byte[1], con_byte[2], con_byte[3] ); } } diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c index 0741e57af7..96ed49665b 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.c +++ b/src/mesa/drivers/dri/r200/r200_ioctl.c @@ -41,6 +41,22 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/context.h" #include "swrast/swrast.h" +#include "main/blend.h" +#include "main/bufferobj.h" +#include "main/buffers.h" +#include "main/depth.h" +#include "main/shaders.h" +#include "main/texstate.h" +#include "main/varray.h" +#include "glapi/dispatch.h" +#include "swrast/swrast.h" +#include "main/stencil.h" +#include "main/matrix.h" +#include "main/attrib.h" +#include "main/enable.h" + +#include "radeon_common.h" +#include "radeon_lock.h" #include "r200_context.h" #include "r200_state.h" #include "r200_ioctl.h" @@ -54,635 +70,230 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R200_TIMEOUT 512 #define R200_IDLE_RETRY 16 - -static void r200WaitForIdle( r200ContextPtr rmesa ); - - -/* At this point we were in FlushCmdBufLocked but we had lost our context, so - * we need to unwire our current cmdbuf, hook the one with the saved state in - * it, flush it, and then put the current one back. This is so commands at the - * start of a cmdbuf can rely on the state being kept from the previous one. - */ -static void r200BackUpAndEmitLostStateLocked( r200ContextPtr rmesa ) +static void +r200_meta_set_passthrough_transform(r200ContextPtr r200) { - GLuint nr_released_bufs; - struct r200_store saved_store; + GLcontext *ctx = r200->radeon.glCtx; - if (rmesa->backup_store.cmd_used == 0) - return; + r200->meta.saved_vp_x = ctx->Viewport.X; + r200->meta.saved_vp_y = ctx->Viewport.Y; + r200->meta.saved_vp_width = ctx->Viewport.Width; + r200->meta.saved_vp_height = ctx->Viewport.Height; + r200->meta.saved_matrix_mode = ctx->Transform.MatrixMode; - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "Emitting backup state on lost context\n"); + _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height); - rmesa->lost_context = GL_FALSE; + _mesa_MatrixMode(GL_PROJECTION); + _mesa_PushMatrix(); + _mesa_LoadIdentity(); + _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1); - nr_released_bufs = rmesa->dma.nr_released_bufs; - saved_store = rmesa->store; - rmesa->dma.nr_released_bufs = 0; - rmesa->store = rmesa->backup_store; - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); - rmesa->dma.nr_released_bufs = nr_released_bufs; - rmesa->store = saved_store; + _mesa_MatrixMode(GL_MODELVIEW); + _mesa_PushMatrix(); + _mesa_LoadIdentity(); } -int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller ) +static void +r200_meta_restore_transform(r200ContextPtr r200) { - int ret, i; - drm_radeon_cmd_buffer_t cmd; - - if (rmesa->lost_context) - r200BackUpAndEmitLostStateLocked( rmesa ); - - if (R200_DEBUG & DEBUG_IOCTL) { - fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); - - if (0 & R200_DEBUG & DEBUG_VERBOSE) - for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 ) - fprintf(stderr, "%d: %x\n", i/4, - *(int *)(&rmesa->store.cmd_buf[i])); - } - - if (R200_DEBUG & DEBUG_DMA) - fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__, - rmesa->dma.nr_released_bufs); - - - if (R200_DEBUG & DEBUG_SANITY) { - if (rmesa->state.scissor.enabled) - ret = r200SanityCmdBuffer( rmesa, - rmesa->state.scissor.numClipRects, - rmesa->state.scissor.pClipRects); - else - ret = r200SanityCmdBuffer( rmesa, - rmesa->numClipRects, - rmesa->pClipRects); - if (ret) { - fprintf(stderr, "drmSanityCommandWrite: %d\n", ret); - goto out; - } - } - - - if (R200_DEBUG & DEBUG_MEMORY) { - if (! driValidateTextureHeaps( rmesa->texture_heaps, rmesa->nr_heaps, - & rmesa->swapped ) ) { - fprintf( stderr, "%s: texture memory is inconsistent - expect " - "mangled textures\n", __FUNCTION__ ); - } - } + _mesa_MatrixMode(GL_PROJECTION); + _mesa_PopMatrix(); + _mesa_MatrixMode(GL_MODELVIEW); + _mesa_PopMatrix(); + _mesa_MatrixMode(r200->meta.saved_matrix_mode); - cmd.bufsz = rmesa->store.cmd_used; - cmd.buf = rmesa->store.cmd_buf; - - if (rmesa->state.scissor.enabled) { - cmd.nbox = rmesa->state.scissor.numClipRects; - cmd.boxes = (drm_clip_rect_t *)rmesa->state.scissor.pClipRects; - } else { - cmd.nbox = rmesa->numClipRects; - cmd.boxes = (drm_clip_rect_t *)rmesa->pClipRects; - } - - ret = drmCommandWrite( rmesa->dri.fd, - DRM_RADEON_CMDBUF, - &cmd, sizeof(cmd) ); - - if (ret) - fprintf(stderr, "drmCommandWrite: %d\n", ret); - - if (R200_DEBUG & DEBUG_SYNC) { - fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__); - r200WaitForIdleLocked( rmesa ); - } - - - out: - rmesa->store.primnr = 0; - rmesa->store.statenr = 0; - rmesa->store.cmd_used = 0; - rmesa->dma.nr_released_bufs = 0; - rmesa->save_on_next_emit = 1; - - return ret; + _mesa_Viewport(r200->meta.saved_vp_x, r200->meta.saved_vp_y, + r200->meta.saved_vp_width, r200->meta.saved_vp_height); } - -/* Note: does not emit any commands to avoid recursion on - * r200AllocCmdBuf. - */ -void r200FlushCmdBuf( r200ContextPtr rmesa, const char *caller ) -{ - int ret; - - LOCK_HARDWARE( rmesa ); - - ret = r200FlushCmdBufLocked( rmesa, caller ); - - UNLOCK_HARDWARE( rmesa ); - - if (ret) { - fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret); - exit(ret); - } -} - - -/* ============================================================= - * Hardware vertex buffer handling +/** + * Perform glClear where mask contains only color, depth, and/or stencil. + * + * The implementation is based on calling into Mesa to set GL state and + * performing normal triangle rendering. The intent of this path is to + * have as generic a path as possible, so that any driver could make use of + * it. */ - - -void r200RefillCurrentDmaRegion( r200ContextPtr rmesa ) +static void radeon_clear_tris(GLcontext *ctx, GLbitfield mask) { - struct r200_dma_buffer *dmabuf; - int fd = rmesa->dri.fd; - int index = 0; - int size = 0; - drmDMAReq dma; - int ret; - - if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->dma.flush) { - rmesa->dma.flush( rmesa ); - } - - if (rmesa->dma.current.buf) - r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ ); - - if (rmesa->dma.nr_released_bufs > 4) - r200FlushCmdBuf( rmesa, __FUNCTION__ ); - - dma.context = rmesa->dri.hwContext; - dma.send_count = 0; - dma.send_list = NULL; - dma.send_sizes = NULL; - dma.flags = 0; - dma.request_count = 1; - dma.request_size = RADEON_BUFFER_SIZE; - dma.request_list = &index; - dma.request_sizes = &size; - dma.granted_count = 0; - - LOCK_HARDWARE(rmesa); /* no need to validate */ - - while (1) { - ret = drmDMA( fd, &dma ); - if (ret == 0) - break; + r200ContextPtr rmesa = R200_CONTEXT(ctx); + GLfloat vertices[4][3]; + GLfloat color[4][4]; + GLfloat dst_z; + struct gl_framebuffer *fb = ctx->DrawBuffer; + int i; + GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE; + GLboolean saved_shader_program = 0; + unsigned int saved_active_texture; - if (rmesa->dma.nr_released_bufs) { - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); - } - - if (rmesa->do_usleeps) { - UNLOCK_HARDWARE( rmesa ); - DO_USLEEP( 1 ); - LOCK_HARDWARE( rmesa ); - } + assert((mask & ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_FRONT_LEFT | + BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) == 0); + + _mesa_PushAttrib(GL_COLOR_BUFFER_BIT | + GL_CURRENT_BIT | + GL_DEPTH_BUFFER_BIT | + GL_ENABLE_BIT | + GL_STENCIL_BUFFER_BIT | + GL_TRANSFORM_BIT | + GL_CURRENT_BIT); + _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT); + saved_active_texture = ctx->Texture.CurrentUnit; + + /* Disable existing GL state we don't want to apply to a clear. */ + _mesa_Disable(GL_ALPHA_TEST); + _mesa_Disable(GL_BLEND); + _mesa_Disable(GL_CULL_FACE); + _mesa_Disable(GL_FOG); + _mesa_Disable(GL_POLYGON_SMOOTH); + _mesa_Disable(GL_POLYGON_STIPPLE); + _mesa_Disable(GL_POLYGON_OFFSET_FILL); + _mesa_Disable(GL_LIGHTING); + _mesa_Disable(GL_CLIP_PLANE0); + _mesa_Disable(GL_CLIP_PLANE1); + _mesa_Disable(GL_CLIP_PLANE2); + _mesa_Disable(GL_CLIP_PLANE3); + _mesa_Disable(GL_CLIP_PLANE4); + _mesa_Disable(GL_CLIP_PLANE5); + if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) { + saved_fp_enable = GL_TRUE; + _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB); + } + if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) { + saved_vp_enable = GL_TRUE; + _mesa_Disable(GL_VERTEX_PROGRAM_ARB); + } + if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) { + saved_shader_program = ctx->Shader.CurrentProgram->Name; + _mesa_UseProgramObjectARB(0); } - - UNLOCK_HARDWARE(rmesa); - - if (R200_DEBUG & DEBUG_DMA) - fprintf(stderr, "Allocated buffer %d\n", index); - - dmabuf = CALLOC_STRUCT( r200_dma_buffer ); - dmabuf->buf = &rmesa->r200Screen->buffers->list[index]; - dmabuf->refcount = 1; - - rmesa->dma.current.buf = dmabuf; - rmesa->dma.current.address = dmabuf->buf->address; - rmesa->dma.current.end = dmabuf->buf->total; - rmesa->dma.current.start = 0; - rmesa->dma.current.ptr = 0; -} - -void r200ReleaseDmaRegion( r200ContextPtr rmesa, - struct r200_dma_region *region, - const char *caller ) -{ - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); - if (!region->buf) - return; - - if (rmesa->dma.flush) - rmesa->dma.flush( rmesa ); - - if (--region->buf->refcount == 0) { - drm_radeon_cmd_header_t *cmd; - - if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA)) - fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__, - region->buf->buf->idx); + if (ctx->Texture._EnabledUnits != 0) { + int i; - cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sizeof(*cmd), - __FUNCTION__ ); - cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD; - cmd->dma.buf_idx = region->buf->buf->idx; - FREE(region->buf); - rmesa->dma.nr_released_bufs++; - } - - region->buf = NULL; - region->start = 0; -} - -/* Allocates a region from rmesa->dma.current. If there isn't enough - * space in current, grab a new buffer (and discard what was left of current) - */ -void r200AllocDmaRegion( r200ContextPtr rmesa, - struct r200_dma_region *region, - int bytes, - int alignment ) -{ - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %d\n", __FUNCTION__, bytes); - - if (rmesa->dma.flush) - rmesa->dma.flush( rmesa ); - - if (region->buf) - r200ReleaseDmaRegion( rmesa, region, __FUNCTION__ ); - - alignment--; - rmesa->dma.current.start = rmesa->dma.current.ptr = - (rmesa->dma.current.ptr + alignment) & ~alignment; - - if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) - r200RefillCurrentDmaRegion( rmesa ); - - region->start = rmesa->dma.current.start; - region->ptr = rmesa->dma.current.start; - region->end = rmesa->dma.current.start + bytes; - region->address = rmesa->dma.current.address; - region->buf = rmesa->dma.current.buf; - region->buf->refcount++; - - rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */ - rmesa->dma.current.start = - rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7; - - assert( rmesa->dma.current.ptr <= rmesa->dma.current.end ); -} - -/* ================================================================ - * SwapBuffers with client-side throttling - */ - -static uint32_t r200GetLastFrame(r200ContextPtr rmesa) -{ - drm_radeon_getparam_t gp; - int ret; - uint32_t frame; - - gp.param = RADEON_PARAM_LAST_FRAME; - gp.value = (int *)&frame; - ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM, - &gp, sizeof(gp) ); - if ( ret ) { - fprintf( stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, ret ); - exit(1); - } - - return frame; -} - -static void r200EmitIrqLocked( r200ContextPtr rmesa ) -{ - drm_radeon_irq_emit_t ie; - int ret; - - ie.irq_seq = &rmesa->iw.irq_seq; - ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, - &ie, sizeof(ie) ); - if ( ret ) { - fprintf( stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, ret ); - exit(1); - } -} - - -static void r200WaitIrq( r200ContextPtr rmesa ) -{ - int ret; - - do { - ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT, - &rmesa->iw, sizeof(rmesa->iw) ); - } while (ret && (errno == EINTR || errno == EBUSY)); - - if ( ret ) { - fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret ); - exit(1); - } -} - - -static void r200WaitForFrameCompletion( r200ContextPtr rmesa ) -{ - drm_radeon_sarea_t *sarea = rmesa->sarea; - - if (rmesa->do_irqs) { - if (r200GetLastFrame(rmesa) < sarea->last_frame) { - if (!rmesa->irqsEmitted) { - while (r200GetLastFrame (rmesa) < sarea->last_frame) - ; - } - else { - UNLOCK_HARDWARE( rmesa ); - r200WaitIrq( rmesa ); - LOCK_HARDWARE( rmesa ); + for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { + _mesa_ActiveTextureARB(GL_TEXTURE0 + i); + _mesa_Disable(GL_TEXTURE_1D); + _mesa_Disable(GL_TEXTURE_2D); + _mesa_Disable(GL_TEXTURE_3D); + if (ctx->Extensions.ARB_texture_cube_map) + _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB); + if (ctx->Extensions.NV_texture_rectangle) + _mesa_Disable(GL_TEXTURE_RECTANGLE_NV); + if (ctx->Extensions.MESA_texture_array) { + _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT); + _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT); } - rmesa->irqsEmitted = 10; - } - - if (rmesa->irqsEmitted) { - r200EmitIrqLocked( rmesa ); - rmesa->irqsEmitted--; - } - } - else { - while (r200GetLastFrame (rmesa) < sarea->last_frame) { - UNLOCK_HARDWARE( rmesa ); - if (rmesa->do_usleeps) - DO_USLEEP( 1 ); - LOCK_HARDWARE( rmesa ); } } -} - - - -/* Copy the back color buffer to the front color buffer. - */ -void r200CopyBuffer( __DRIdrawablePrivate *dPriv, - const drm_clip_rect_t *rect) -{ - r200ContextPtr rmesa; - GLint nbox, i, ret; - GLboolean missed_target; - int64_t ust; - __DRIscreenPrivate *psp = dPriv->driScreenPriv; - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate; - - if ( R200_DEBUG & DEBUG_IOCTL ) { - fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *)rmesa->glCtx ); + + r200_meta_set_passthrough_transform(rmesa); + + for (i = 0; i < 4; i++) { + color[i][0] = ctx->Color.ClearColor[0]; + color[i][1] = ctx->Color.ClearColor[1]; + color[i][2] = ctx->Color.ClearColor[2]; + color[i][3] = ctx->Color.ClearColor[3]; } - R200_FIREVERTICES( rmesa ); - - LOCK_HARDWARE( rmesa ); - - - /* Throttle the frame rate -- only allow one pending swap buffers - * request at a time. + /* convert clear Z from [0,1] to NDC coord in [-1,1] */ + dst_z = -1.0 + 2.0 * ctx->Depth.Clear; + + /* Prepare the vertices, which are the same regardless of which buffer we're + * drawing to. */ - r200WaitForFrameCompletion( rmesa ); - if (!rect) - { - UNLOCK_HARDWARE( rmesa ); - driWaitForVBlank( dPriv, & missed_target ); - LOCK_HARDWARE( rmesa ); - } - - nbox = dPriv->numClipRects; /* must be in locked region */ - - for ( i = 0 ; i < nbox ; ) { - GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox ); - drm_clip_rect_t *box = dPriv->pClipRects; - drm_clip_rect_t *b = rmesa->sarea->boxes; - GLint n = 0; - - for ( ; i < nr ; i++ ) { - - *b = box[i]; - - if (rect) - { - if (rect->x1 > b->x1) - b->x1 = rect->x1; - if (rect->y1 > b->y1) - b->y1 = rect->y1; - if (rect->x2 < b->x2) - b->x2 = rect->x2; - if (rect->y2 < b->y2) - b->y2 = rect->y2; - - if (b->x1 >= b->x2 || b->y1 >= b->y2) - continue; - } + vertices[0][0] = fb->_Xmin; + vertices[0][1] = fb->_Ymin; + vertices[0][2] = dst_z; + vertices[1][0] = fb->_Xmax; + vertices[1][1] = fb->_Ymin; + vertices[1][2] = dst_z; + vertices[2][0] = fb->_Xmax; + vertices[2][1] = fb->_Ymax; + vertices[2][2] = dst_z; + vertices[3][0] = fb->_Xmin; + vertices[3][1] = fb->_Ymax; + vertices[3][2] = dst_z; + + _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &color); + _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), &vertices); + _mesa_Enable(GL_COLOR_ARRAY); + _mesa_Enable(GL_VERTEX_ARRAY); + + while (mask != 0) { + GLuint this_mask = 0; + + if (mask & BUFFER_BIT_BACK_LEFT) + this_mask = BUFFER_BIT_BACK_LEFT; + else if (mask & BUFFER_BIT_FRONT_LEFT) + this_mask = BUFFER_BIT_FRONT_LEFT; + + /* Clear depth/stencil in the same pass as color. */ + this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)); + + /* Select the current color buffer and use the color write mask if + * we have one, otherwise don't write any color channels. + */ + if (this_mask & BUFFER_BIT_FRONT_LEFT) + _mesa_DrawBuffer(GL_FRONT_LEFT); + else if (this_mask & BUFFER_BIT_BACK_LEFT) + _mesa_DrawBuffer(GL_BACK_LEFT); + else + _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE); - b++; - n++; + /* Control writing of the depth clear value to depth. */ + if (this_mask & BUFFER_BIT_DEPTH) { + _mesa_DepthFunc(GL_ALWAYS); + _mesa_Enable(GL_DEPTH_TEST); + } else { + _mesa_Disable(GL_DEPTH_TEST); + _mesa_DepthMask(GL_FALSE); } - rmesa->sarea->nbox = n; - - if (!n) - continue; - ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP ); - - if ( ret ) { - fprintf( stderr, "DRM_R200_SWAP_BUFFERS: return = %d\n", ret ); - UNLOCK_HARDWARE( rmesa ); - exit( 1 ); + /* Control writing of the stencil clear value to stencil. */ + if (this_mask & BUFFER_BIT_STENCIL) { + _mesa_Enable(GL_STENCIL_TEST); + _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE); + _mesa_StencilFuncSeparate(GL_FRONT, GL_ALWAYS, ctx->Stencil.Clear, + ctx->Stencil.WriteMask[0]); + } else { + _mesa_Disable(GL_STENCIL_TEST); } - } - - UNLOCK_HARDWARE( rmesa ); - if (!rect) - { - rmesa->hw.all_dirty = GL_TRUE; - rmesa->swap_count++; - (*psp->systemTime->getUST)( & ust ); - if ( missed_target ) { - rmesa->swap_missed_count++; - rmesa->swap_missed_ust = ust - rmesa->swap_ust; - } + CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4)); - rmesa->swap_ust = ust; - - sched_yield(); + mask &= ~this_mask; } -} -void r200PageFlip( __DRIdrawablePrivate *dPriv ) -{ - r200ContextPtr rmesa; - GLint ret; - GLboolean missed_target; - __DRIscreenPrivate *psp = dPriv->driScreenPriv; + r200_meta_restore_transform(rmesa); - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); + _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture); + if (saved_fp_enable) + _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB); + if (saved_vp_enable) + _mesa_Enable(GL_VERTEX_PROGRAM_ARB); - rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate; + if (saved_shader_program) + _mesa_UseProgramObjectARB(saved_shader_program); - if ( R200_DEBUG & DEBUG_IOCTL ) { - fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__, - rmesa->sarea->pfCurrentPage); - } - - R200_FIREVERTICES( rmesa ); - LOCK_HARDWARE( rmesa ); - - if (!dPriv->numClipRects) { - UNLOCK_HARDWARE( rmesa ); - usleep( 10000 ); /* throttle invisible client 10ms */ - return; - } - - /* Need to do this for the perf box placement: - */ - { - drm_clip_rect_t *box = dPriv->pClipRects; - drm_clip_rect_t *b = rmesa->sarea->boxes; - b[0] = box[0]; - rmesa->sarea->nbox = 1; - } - - /* Throttle the frame rate -- only allow a few pending swap buffers - * request at a time. - */ - r200WaitForFrameCompletion( rmesa ); - UNLOCK_HARDWARE( rmesa ); - driWaitForVBlank( dPriv, & missed_target ); - if ( missed_target ) { - rmesa->swap_missed_count++; - (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust ); - } - LOCK_HARDWARE( rmesa ); - - ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP ); - - UNLOCK_HARDWARE( rmesa ); - - if ( ret ) { - fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret ); - exit( 1 ); - } - - rmesa->swap_count++; - (void) (*psp->systemTime->getUST)( & rmesa->swap_ust ); - -#if 000 - if ( rmesa->sarea->pfCurrentPage == 1 ) { - rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset; - rmesa->state.color.drawPitch = rmesa->r200Screen->frontPitch; - } else { - rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset; - rmesa->state.color.drawPitch = rmesa->r200Screen->backPitch; - } - - R200_STATECHANGE( rmesa, ctx ); - rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = rmesa->state.color.drawOffset - + rmesa->r200Screen->fbLocation; - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = rmesa->state.color.drawPitch; - if (rmesa->sarea->tiling_enabled) { - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE; - } -#else - /* Get ready for drawing next frame. Update the renderbuffers' - * flippedOffset/Pitch fields so we draw into the right place. - */ - driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer, - rmesa->sarea->pfCurrentPage); + _mesa_PopClientAttrib(); + _mesa_PopAttrib(); +} - r200UpdateDrawBuffer(rmesa->glCtx); -#endif +static void r200UserClear(GLcontext *ctx, GLuint mask) +{ + radeon_clear_tris(ctx, mask); } - -/* ================================================================ - * Buffer clear - */ -static void r200Clear( GLcontext *ctx, GLbitfield mask ) +static void r200KernelClear(GLcontext *ctx, GLuint flags) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; - GLuint flags = 0; - GLuint color_mask = 0; - GLint ret, i; - GLint cx, cy, cw, ch; - - if ( R200_DEBUG & DEBUG_IOCTL ) { - fprintf( stderr, "r200Clear\n"); - } - - { - LOCK_HARDWARE( rmesa ); - UNLOCK_HARDWARE( rmesa ); - if ( dPriv->numClipRects == 0 ) - return; - } - - r200Flush( ctx ); - - if ( mask & BUFFER_BIT_FRONT_LEFT ) { - flags |= RADEON_FRONT; - color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK]; - mask &= ~BUFFER_BIT_FRONT_LEFT; - } - - if ( mask & BUFFER_BIT_BACK_LEFT ) { - flags |= RADEON_BACK; - color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK]; - mask &= ~BUFFER_BIT_BACK_LEFT; - } - - if ( mask & BUFFER_BIT_DEPTH ) { - flags |= RADEON_DEPTH; - mask &= ~BUFFER_BIT_DEPTH; - } - - if ( (mask & BUFFER_BIT_STENCIL) && rmesa->state.stencil.hwBuffer ) { - flags |= RADEON_STENCIL; - mask &= ~BUFFER_BIT_STENCIL; - } - - if ( mask ) { - if (R200_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask); - _swrast_Clear( ctx, mask ); - } - - if ( !flags ) - return; - - if (rmesa->using_hyperz) { - flags |= RADEON_USE_COMP_ZBUF; -/* if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) - flags |= RADEON_USE_HIERZ; */ - if (!(rmesa->state.stencil.hwBuffer) || - ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) && - ((rmesa->state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) { - flags |= RADEON_CLEAR_FASTZ; - } - } + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; + GLint cx, cy, cw, ch, ret; + GLuint i; - LOCK_HARDWARE( rmesa ); - - /* compute region after locking: */ - cx = ctx->DrawBuffer->_Xmin; - cy = ctx->DrawBuffer->_Ymin; - cw = ctx->DrawBuffer->_Xmax - cx; - ch = ctx->DrawBuffer->_Ymax - cy; - - /* Flip top to bottom */ - cx += dPriv->x; - cy = dPriv->y + dPriv->h - cy - ch; + LOCK_HARDWARE( &rmesa->radeon ); /* Throttle the number of clear ioctls we do. */ @@ -693,7 +304,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) gp.param = RADEON_PARAM_LAST_CLEAR; gp.value = (int *)&clear; - ret = drmCommandWriteRead( rmesa->dri.fd, + ret = drmCommandWriteRead( rmesa->radeon.dri.fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp) ); if ( ret ) { @@ -703,24 +314,34 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) /* Clear throttling needs more thought. */ - if ( rmesa->sarea->last_clear - clear <= 25 ) { + if ( rmesa->radeon.sarea->last_clear - clear <= 25 ) { break; } - if (rmesa->do_usleeps) { - UNLOCK_HARDWARE( rmesa ); + if (rmesa->radeon.do_usleeps) { + UNLOCK_HARDWARE( &rmesa->radeon ); DO_USLEEP( 1 ); - LOCK_HARDWARE( rmesa ); + LOCK_HARDWARE( &rmesa->radeon ); } } /* Send current state to the hardware */ - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); + rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); + + /* compute region after locking: */ + cx = ctx->DrawBuffer->_Xmin; + cy = ctx->DrawBuffer->_Ymin; + cw = ctx->DrawBuffer->_Xmax - cx; + ch = ctx->DrawBuffer->_Ymax - cy; + + /* Flip top to bottom */ + cx += dPriv->x; + cy = dPriv->y + dPriv->h - cy - ch; for ( i = 0 ; i < dPriv->numClipRects ; ) { GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects ); drm_clip_rect_t *box = dPriv->pClipRects; - drm_clip_rect_t *b = rmesa->sarea->boxes; + drm_clip_rect_t *b = rmesa->radeon.sarea->boxes; drm_radeon_clear_t clear; drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS]; GLint n = 0; @@ -755,17 +376,17 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) } } - rmesa->sarea->nbox = n; + rmesa->radeon.sarea->nbox = n; clear.flags = flags; - clear.clear_color = rmesa->state.color.clear; - clear.clear_depth = rmesa->state.depth.clear; /* needed for hyperz */ + clear.clear_color = rmesa->radeon.state.color.clear; + clear.clear_depth = rmesa->radeon.state.depth.clear; /* needed for hyperz */ clear.color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK]; - clear.depth_mask = rmesa->state.stencil.clear; + clear.depth_mask = rmesa->radeon.state.stencil.clear; clear.depth_boxes = depth_boxes; n--; - b = rmesa->sarea->boxes; + b = rmesa->radeon.sarea->boxes; for ( ; n >= 0 ; n-- ) { depth_boxes[n].f[CLEAR_X1] = (float)b[n].x1; depth_boxes[n].f[CLEAR_Y1] = (float)b[n].y1; @@ -774,83 +395,92 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) depth_boxes[n].f[CLEAR_DEPTH] = ctx->Depth.Clear; } - ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR, + ret = drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_CLEAR, &clear, sizeof(clear)); if ( ret ) { - UNLOCK_HARDWARE( rmesa ); + UNLOCK_HARDWARE( &rmesa->radeon ); fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret ); exit( 1 ); } } - - UNLOCK_HARDWARE( rmesa ); - rmesa->hw.all_dirty = GL_TRUE; + UNLOCK_HARDWARE( &rmesa->radeon ); } - - -void r200WaitForIdleLocked( r200ContextPtr rmesa ) +/* ================================================================ + * Buffer clear + */ +static void r200Clear( GLcontext *ctx, GLbitfield mask ) { - int ret; - int i = 0; - - do { - ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_CP_IDLE); - if (ret) - DO_USLEEP( 1 ); - } while (ret && ++i < 100); - - if ( ret < 0 ) { - UNLOCK_HARDWARE( rmesa ); - fprintf( stderr, "Error: R200 timed out... exiting\n" ); - exit( -1 ); - } -} + r200ContextPtr rmesa = R200_CONTEXT(ctx); + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; + GLuint flags = 0; + GLuint color_mask = 0; + GLint ret; + GLuint orig_mask = mask; + if ( R200_DEBUG & DEBUG_IOCTL ) { + fprintf( stderr, "r200Clear %x %d\n", mask, rmesa->radeon.sarea->pfCurrentPage); + } -static void r200WaitForIdle( r200ContextPtr rmesa ) -{ - LOCK_HARDWARE(rmesa); - r200WaitForIdleLocked( rmesa ); - UNLOCK_HARDWARE(rmesa); -} + { + LOCK_HARDWARE( &rmesa->radeon ); + UNLOCK_HARDWARE( &rmesa->radeon ); + if ( dPriv->numClipRects == 0 ) + return; + } + radeonFlush( ctx ); -void r200Flush( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT( ctx ); + if ( mask & BUFFER_BIT_FRONT_LEFT ) { + flags |= RADEON_FRONT; + color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK]; + mask &= ~BUFFER_BIT_FRONT_LEFT; + } - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); + if ( mask & BUFFER_BIT_BACK_LEFT ) { + flags |= RADEON_BACK; + color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK]; + mask &= ~BUFFER_BIT_BACK_LEFT; + } - if (rmesa->dma.flush) - rmesa->dma.flush( rmesa ); + if ( mask & BUFFER_BIT_DEPTH ) { + flags |= RADEON_DEPTH; + mask &= ~BUFFER_BIT_DEPTH; + } - r200EmitState( rmesa ); - - if (rmesa->store.cmd_used) - r200FlushCmdBuf( rmesa, __FUNCTION__ ); -} + if ( (mask & BUFFER_BIT_STENCIL) && rmesa->radeon.state.stencil.hwBuffer ) { + flags |= RADEON_STENCIL; + mask &= ~BUFFER_BIT_STENCIL; + } -/* Make sure all commands have been sent to the hardware and have - * completed processing. - */ -void r200Finish( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - r200Flush( ctx ); + if ( mask ) { + if (R200_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask); + _swrast_Clear( ctx, mask ); + } + + if ( !flags ) + return; - if (rmesa->do_irqs) { - LOCK_HARDWARE( rmesa ); - r200EmitIrqLocked( rmesa ); - UNLOCK_HARDWARE( rmesa ); - r200WaitIrq( rmesa ); + if (rmesa->using_hyperz) { + flags |= RADEON_USE_COMP_ZBUF; +/* if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) + flags |= RADEON_USE_HIERZ; */ + if (!(rmesa->radeon.state.stencil.hwBuffer) || + ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) && + ((rmesa->radeon.state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) { + flags |= RADEON_CLEAR_FASTZ; + } } - else - r200WaitForIdle( rmesa ); -} + if (rmesa->radeon.radeonScreen->kernel_mm) + r200UserClear(ctx, orig_mask); + else + r200KernelClear(ctx, flags); + + rmesa->radeon.hw.all_dirty = GL_TRUE; +} /* This version of AllocateMemoryMESA allocates only GART memory, and * only does so after the point at which the driver has been @@ -875,7 +505,7 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size, fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, writefreq, priority); - if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map) + if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map) return NULL; if (getenv("R200_NO_ALLOC")) @@ -886,7 +516,7 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size, alloc.size = size; alloc.region_offset = ®ion_offset; - ret = drmCommandWriteRead( rmesa->r200Screen->driScreen->fd, + ret = drmCommandWriteRead( rmesa->radeon.radeonScreen->driScreen->fd, DRM_RADEON_ALLOC, &alloc, sizeof(alloc)); @@ -896,7 +526,7 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size, } { - char *region_start = (char *)rmesa->r200Screen->gartTextures.map; + char *region_start = (char *)rmesa->radeon.radeonScreen->gartTextures.map; return (void *)(region_start + region_offset); } } @@ -914,24 +544,24 @@ void r200FreeMemoryMESA(__DRIscreen *screen, GLvoid *pointer) if (R200_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s %p\n", __FUNCTION__, pointer); - if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map) { + if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map) { fprintf(stderr, "%s: no context\n", __FUNCTION__); return; } - region_offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map; + region_offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map; if (region_offset < 0 || - region_offset > rmesa->r200Screen->gartTextures.size) { + region_offset > rmesa->radeon.radeonScreen->gartTextures.size) { fprintf(stderr, "offset %d outside range 0..%d\n", region_offset, - rmesa->r200Screen->gartTextures.size); + rmesa->radeon.radeonScreen->gartTextures.size); return; } memfree.region = RADEON_MEM_REGION_GART; memfree.region_offset = region_offset; - ret = drmCommandWrite( rmesa->r200Screen->driScreen->fd, + ret = drmCommandWrite( rmesa->radeon.radeonScreen->driScreen->fd, DRM_RADEON_FREE, &memfree, sizeof(memfree)); @@ -956,16 +586,16 @@ GLuint r200GetMemoryOffsetMESA(__DRIscreen *screen, const GLvoid *pointer) card_offset = r200GartOffsetFromVirtual( rmesa, pointer ); - return card_offset - rmesa->r200Screen->gart_base; + return card_offset - rmesa->radeon.radeonScreen->gart_base; } GLboolean r200IsGartMemory( r200ContextPtr rmesa, const GLvoid *pointer, GLint size ) { - ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map; + ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map; int valid = (size >= 0 && offset >= 0 && - offset + size < rmesa->r200Screen->gartTextures.size); + offset + size < rmesa->radeon.radeonScreen->gartTextures.size); if (R200_DEBUG & DEBUG_IOCTL) fprintf(stderr, "r200IsGartMemory( %p ) : %d\n", pointer, valid ); @@ -976,12 +606,12 @@ GLboolean r200IsGartMemory( r200ContextPtr rmesa, const GLvoid *pointer, GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer ) { - ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map; + ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map; - if (offset < 0 || offset > rmesa->r200Screen->gartTextures.size) + if (offset < 0 || offset > rmesa->radeon.radeonScreen->gartTextures.size) return ~0; else - return rmesa->r200Screen->gart_texture_offset + offset; + return rmesa->radeon.radeonScreen->gart_texture_offset + offset; } @@ -989,7 +619,7 @@ GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer ) void r200InitIoctlFuncs( struct dd_function_table *functions ) { functions->Clear = r200Clear; - functions->Finish = r200Finish; - functions->Flush = r200Flush; + functions->Finish = radeonFinish; + functions->Flush = radeonFlush; } diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.h b/src/mesa/drivers/dri/r200/r200_ioctl.h index f7458e4a0e..2a4b8a11f4 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.h +++ b/src/mesa/drivers/dri/r200/r200_ioctl.h @@ -37,65 +37,30 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/simple_list.h" #include "radeon_dri.h" -#include "r200_lock.h" + +#include "radeon_bocs_wrapper.h" #include "xf86drm.h" #include "drm.h" #include "radeon_drm.h" -extern void r200EmitState( r200ContextPtr rmesa ); extern void r200EmitVertexAOS( r200ContextPtr rmesa, - GLuint vertex_size, - GLuint offset ); + GLuint vertex_size, + struct radeon_bo *bo, + GLuint offset ); extern void r200EmitVbufPrim( r200ContextPtr rmesa, GLuint primitive, GLuint vertex_nr ); -extern void r200FlushElts( r200ContextPtr rmesa ); +extern void r200FlushElts(GLcontext *ctx); extern GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, GLuint primitive, GLuint min_nr ); -extern void r200EmitAOS( r200ContextPtr rmesa, - struct r200_dma_region **regions, - GLuint n, - GLuint offset ); - -extern void r200EmitBlit( r200ContextPtr rmesa, - GLuint color_fmt, - GLuint src_pitch, - GLuint src_offset, - GLuint dst_pitch, - GLuint dst_offset, - GLint srcx, GLint srcy, - GLint dstx, GLint dsty, - GLuint w, GLuint h ); - -extern void r200EmitWait( r200ContextPtr rmesa, GLuint flags ); - -extern void r200FlushCmdBuf( r200ContextPtr rmesa, const char * ); -extern int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller ); - -extern void r200RefillCurrentDmaRegion( r200ContextPtr rmesa ); - -extern void r200AllocDmaRegion( r200ContextPtr rmesa, - struct r200_dma_region *region, - int bytes, - int alignment ); - -extern void r200ReleaseDmaRegion( r200ContextPtr rmesa, - struct r200_dma_region *region, - const char *caller ); - -extern void r200CopyBuffer( __DRIdrawablePrivate *drawable, - const drm_clip_rect_t *rect); -extern void r200PageFlip( __DRIdrawablePrivate *drawable ); -extern void r200Flush( GLcontext *ctx ); -extern void r200Finish( GLcontext *ctx ); -extern void r200WaitForIdleLocked( r200ContextPtr rmesa ); -extern void r200WaitForVBlank( r200ContextPtr rmesa ); +extern void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset); + extern void r200InitIoctlFuncs( struct dd_function_table *functions ); extern void *r200AllocateMemoryMESA( __DRIscreen *screen, GLsizei size, GLfloat readfreq, @@ -119,8 +84,8 @@ void r200SetUpAtomList( r200ContextPtr rmesa ); */ #define R200_NEWPRIM( rmesa ) \ do { \ - if ( rmesa->dma.flush ) \ - rmesa->dma.flush( rmesa ); \ + if ( rmesa->radeon.dma.flush ) \ + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); \ } while (0) /* Can accomodate several state changes and primitive changes without @@ -130,7 +95,7 @@ do { \ do { \ R200_NEWPRIM( rmesa ); \ rmesa->hw.ATOM.dirty = GL_TRUE; \ - rmesa->hw.is_dirty = GL_TRUE; \ + rmesa->radeon.hw.is_dirty = GL_TRUE; \ } while (0) #define R200_DB_STATE( ATOM ) \ @@ -139,13 +104,13 @@ do { \ static INLINE int R200_DB_STATECHANGE( r200ContextPtr rmesa, - struct r200_state_atom *atom ) + struct radeon_state_atom *atom ) { if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) { - int *tmp; + GLuint *tmp; R200_NEWPRIM( rmesa ); atom->dirty = GL_TRUE; - rmesa->hw.is_dirty = GL_TRUE; + rmesa->radeon.hw.is_dirty = GL_TRUE; tmp = atom->cmd; atom->cmd = atom->lastcmd; atom->lastcmd = tmp; @@ -156,15 +121,6 @@ static INLINE int R200_DB_STATECHANGE( } -/* Fire the buffered vertices no matter what. - */ -#define R200_FIREVERTICES( rmesa ) \ -do { \ - if ( rmesa->store.cmd_used || rmesa->dma.flush ) { \ - r200Flush( rmesa->glCtx ); \ - } \ -} while (0) - /* Command lengths. Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ * are available, you will also be adding an rmesa->state.max_state_size because * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts. @@ -174,36 +130,36 @@ do { \ #define ELTS_BUFSZ(nr) (12 + nr * 2) #define VBUF_BUFSZ (3 * sizeof(int)) -/* Ensure that a minimum amount of space is available in the command buffer. - * This is used to ensure atomicity of state updates with the rendering requests - * that rely on them. - * - * An alternative would be to implement a "soft lock" such that when the buffer - * wraps at an inopportune time, we grab the lock, flush the current buffer, - * and hang on to the lock until the critical section is finished and we flush - * the buffer again and unlock. - */ -static INLINE void r200EnsureCmdBufSpace( r200ContextPtr rmesa, int bytes ) +static inline uint32_t cmdpacket3(int cmd_type) { - if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ) - r200FlushCmdBuf( rmesa, __FUNCTION__ ); - assert( bytes <= R200_CMD_BUF_SZ ); -} + drm_radeon_cmd_header_t cmd; -/* Alloc space in the command buffer - */ -static INLINE char *r200AllocCmdBuf( r200ContextPtr rmesa, - int bytes, const char *where ) -{ - char * head; + cmd.i = 0; + cmd.header.cmd_type = cmd_type; - if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ) - r200FlushCmdBuf( rmesa, where ); + return (uint32_t)cmd.i; - head = rmesa->store.cmd_buf + rmesa->store.cmd_used; - rmesa->store.cmd_used += bytes; - assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ ); - return head; } +#define OUT_BATCH_PACKET3(packet, num_extra) do { \ + if (!b_l_rmesa->radeonScreen->kernel_mm) { \ + OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3)); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } else { \ + OUT_BATCH(CP_PACKET2); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } \ + } while(0) + +#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do { \ + if (!b_l_rmesa->radeonScreen->kernel_mm) { \ + OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP)); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } else { \ + OUT_BATCH(CP_PACKET2); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } \ + } while(0) + + #endif /* __R200_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r200/r200_lock.c b/src/mesa/drivers/dri/r200/r200_lock.c deleted file mode 100644 index 99661a4bfb..0000000000 --- a/src/mesa/drivers/dri/r200/r200_lock.c +++ /dev/null @@ -1,116 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -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, sublicense, 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 NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - */ - -#include "r200_context.h" -#include "r200_lock.h" -#include "r200_tex.h" -#include "r200_state.h" -#include "r200_ioctl.h" - -#include "drirenderbuffer.h" - - -#if DEBUG_LOCKING -char *prevLockFile = NULL; -int prevLockLine = 0; -#endif - -/* Turn on/off page flipping according to the flags in the sarea: - */ -static void -r200UpdatePageFlipping( r200ContextPtr rmesa ) -{ - rmesa->doPageFlip = rmesa->sarea->pfState; - if (rmesa->glCtx->WinSysDrawBuffer) { - driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer, - rmesa->sarea->pfCurrentPage); - } -} - - - -/* Update the hardware state. This is called if another main/context.has - * grabbed the hardware lock, which includes the X server. This - * function also updates the driver's window state after the X server - * moves, resizes or restacks a window -- the change will be reflected - * in the drawable position and clip rects. Since the X server grabs - * the hardware lock when it changes the window state, this routine will - * automatically be called after such a change. - */ -void r200GetLock( r200ContextPtr rmesa, GLuint flags ) -{ - __DRIdrawablePrivate *drawable = rmesa->dri.drawable; - __DRIdrawablePrivate *readable = rmesa->dri.readable; - __DRIscreenPrivate *sPriv = rmesa->dri.screen; - drm_radeon_sarea_t *sarea = rmesa->sarea; - int i; - - drmGetLock( rmesa->dri.fd, rmesa->dri.hwContext, flags ); - - /* The window might have moved, so we might need to get new clip - * rects. - * - * NOTE: This releases and regrabs the hw lock to allow the X server - * to respond to the DRI protocol request for new drawable info. - * Since the hardware state depends on having the latest drawable - * clip rects, all state checking must be done _after_ this call. - */ - DRI_VALIDATE_DRAWABLE_INFO( sPriv, drawable ); - if (drawable != readable) { - DRI_VALIDATE_DRAWABLE_INFO( sPriv, readable ); - } - - if ( rmesa->lastStamp != drawable->lastStamp ) { - r200UpdatePageFlipping( rmesa ); - r200SetCliprects( rmesa ); - r200UpdateViewportOffset( rmesa->glCtx ); - driUpdateFramebufferSize(rmesa->glCtx, drawable); - } - - R200_STATECHANGE( rmesa, ctx ); - if (rmesa->sarea->tiling_enabled) { - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE; - } - else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE; - - if ( sarea->ctx_owner != rmesa->dri.hwContext ) { - sarea->ctx_owner = rmesa->dri.hwContext; - } - - for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) { - DRI_AGE_TEXTURES( rmesa->texture_heaps[ i ] ); - } - - rmesa->lost_context = GL_TRUE; -} diff --git a/src/mesa/drivers/dri/r200/r200_lock.h b/src/mesa/drivers/dri/r200/r200_lock.h deleted file mode 100644 index 4ff98907fb..0000000000 --- a/src/mesa/drivers/dri/r200/r200_lock.h +++ /dev/null @@ -1,106 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -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, sublicense, 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 NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - */ - -#ifndef __R200_LOCK_H__ -#define __R200_LOCK_H__ - -extern void r200GetLock( r200ContextPtr rmesa, GLuint flags ); - -/* Turn DEBUG_LOCKING on to find locking conflicts. - */ -#define DEBUG_LOCKING 0 - -#if DEBUG_LOCKING -extern char *prevLockFile; -extern int prevLockLine; - -#define DEBUG_LOCK() \ - do { \ - prevLockFile = (__FILE__); \ - prevLockLine = (__LINE__); \ - } while (0) - -#define DEBUG_RESET() \ - do { \ - prevLockFile = 0; \ - prevLockLine = 0; \ - } while (0) - -#define DEBUG_CHECK_LOCK() \ - do { \ - if ( prevLockFile ) { \ - fprintf( stderr, \ - "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ - prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ - exit( 1 ); \ - } \ - } while (0) - -#else - -#define DEBUG_LOCK() -#define DEBUG_RESET() -#define DEBUG_CHECK_LOCK() - -#endif - -/* - * !!! We may want to separate locks from locks with validation. This - * could be used to improve performance for those things commands that - * do not do any drawing !!! - */ - - -/* Lock the hardware and validate our state. - */ -#define LOCK_HARDWARE( rmesa ) \ - do { \ - char __ret = 0; \ - DEBUG_CHECK_LOCK(); \ - DRM_CAS( rmesa->dri.hwLock, rmesa->dri.hwContext, \ - (DRM_LOCK_HELD | rmesa->dri.hwContext), __ret ); \ - if ( __ret ) \ - r200GetLock( rmesa, 0 ); \ - DEBUG_LOCK(); \ - } while (0) - -#define UNLOCK_HARDWARE( rmesa ) \ - do { \ - DRM_UNLOCK( rmesa->dri.fd, \ - rmesa->dri.hwLock, \ - rmesa->dri.hwContext ); \ - DEBUG_RESET(); \ - } while (0) - -#endif /* __R200_LOCK_H__ */ diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 8512b9af47..5dbc202330 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -50,110 +50,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_maos.h" #include "r200_tcl.h" - -#if 0 -/* Usage: - * - from r200_tcl_render - * - call r200EmitArrays to ensure uptodate arrays in dma - * - emit primitives (new type?) which reference the data - * -- need to use elts for lineloop, quads, quadstrip/flat - * -- other primitives are all well-formed (need tristrip-1,fake-poly) - * - */ -static void emit_ubyte_rgba3( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - r200_color_t *out = (r200_color_t *)(rvb->start + rvb->address); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p\n", - __FUNCTION__, count, stride, (void *)out); - - for (i = 0; i < count; i++) { - out->red = *data; - out->green = *(data+1); - out->blue = *(data+2); - out->alpha = 0xFF; - out++; - data += stride; - } -} - -static void emit_ubyte_rgba4( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - int *out = (int *)(rvb->address + rvb->start); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d\n", - __FUNCTION__, count, stride); - - if (stride == 4) { - for (i = 0; i < count; i++) - ((int *)out)[i] = LE32_TO_CPU(((int *)data)[i]); - } else { - for (i = 0; i < count; i++) { - *(int *)out++ = LE32_TO_CPU(*(int *)data); - data += stride; - } - } -} - - -static void emit_ubyte_rgba( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int size, - int stride, - int count ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size); - - assert (!rvb->buf); - - if (stride == 0) { - r200AllocDmaRegion( rmesa, rvb, 4, 4 ); - count = 1; - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = 0; - rvb->aos_size = 1; - } - else { - r200AllocDmaRegion( rmesa, rvb, 4 * count, 4 ); /* alignment? */ - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = 1; - rvb->aos_size = 1; - } - - /* Emit the data - */ - switch (size) { - case 3: - emit_ubyte_rgba3( ctx, rvb, data, stride, count ); - break; - case 4: - emit_ubyte_rgba4( ctx, rvb, data, stride, count ); - break; - default: - assert(0); - exit(1); - break; - } -} -#endif - - #if defined(USE_X86_ASM) #define COPY_DWORDS( dst, src, nr ) \ do { \ @@ -174,204 +70,34 @@ do { \ } while (0) #endif - -static void emit_vecfog( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) +static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, + GLvoid *data, int stride, int count) { - int i; - GLfloat *out; - - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d\n", - __FUNCTION__, count, stride); - - assert (!rvb->buf); - - if (stride == 0) { - r200AllocDmaRegion( rmesa, rvb, 4, 4 ); - count = 1; - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = 0; - rvb->aos_size = 1; - } - else { - r200AllocDmaRegion( rmesa, rvb, count * 4, 4 ); /* alignment? */ - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = 1; - rvb->aos_size = 1; - } - - /* Emit the data - */ - out = (GLfloat *)(rvb->address + rvb->start); - for (i = 0; i < count; i++) { - out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); - out++; - data += stride; - } - + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + uint32_t *out; + int i; + int size = 1; + + if (stride == 0) { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32); + count = 1; + aos->stride = 0; + } else { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32); + aos->stride = size; + } + + aos->components = size; + aos->count = count; + + out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); + for (i = 0; i < count; i++) { + out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); + out++; + data += stride; + } } - -static void emit_vec4( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - int *out = (int *)(rvb->address + rvb->start); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d\n", - __FUNCTION__, count, stride); - - if (stride == 4) - COPY_DWORDS( out, data, count ); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out++; - data += stride; - } -} - - -static void emit_vec8( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - int *out = (int *)(rvb->address + rvb->start); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d\n", - __FUNCTION__, count, stride); - - if (stride == 8) - COPY_DWORDS( out, data, count*2 ); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data+4); - out += 2; - data += stride; - } -} - -static void emit_vec12( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - int *out = (int *)(rvb->address + rvb->start); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 12) - COPY_DWORDS( out, data, count*3 ); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data+4); - out[2] = *(int *)(data+8); - out += 3; - data += stride; - } -} - -static void emit_vec16( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int stride, - int count ) -{ - int i; - int *out = (int *)(rvb->address + rvb->start); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d\n", - __FUNCTION__, count, stride); - - if (stride == 16) - COPY_DWORDS( out, data, count*4 ); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data+4); - out[2] = *(int *)(data+8); - out[3] = *(int *)(data+12); - out += 4; - data += stride; - } -} - - -static void emit_vector( GLcontext *ctx, - struct r200_dma_region *rvb, - char *data, - int size, - int stride, - int count ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (R200_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d size %d stride %d\n", - __FUNCTION__, count, size, stride); - - assert (!rvb->buf); - - if (stride == 0) { - r200AllocDmaRegion( rmesa, rvb, size * 4, 4 ); - count = 1; - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = 0; - rvb->aos_size = size; - } - else { - r200AllocDmaRegion( rmesa, rvb, size * count * 4, 4 ); /* alignment? */ - rvb->aos_start = GET_START(rvb); - rvb->aos_stride = size; - rvb->aos_size = size; - } - - /* Emit the data - */ - switch (size) { - case 1: - emit_vec4( ctx, rvb, data, stride, count ); - break; - case 2: - emit_vec8( ctx, rvb, data, stride, count ); - break; - case 3: - emit_vec12( ctx, rvb, data, stride, count ); - break; - case 4: - emit_vec16( ctx, rvb, data, stride, count ); - break; - default: - assert(0); - exit(1); - break; - } - -} - - - /* Emit any changed arrays to new GART memory, re-emit a packet to * update the arrays. */ @@ -379,12 +105,12 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ) { r200ContextPtr rmesa = R200_CONTEXT( ctx ); struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb; - struct r200_dma_region **component = rmesa->tcl.aos_components; GLuint nr = 0; GLuint vfmt0 = 0, vfmt1 = 0; GLuint count = VB->Count; GLuint i, emitsize; + // fprintf(stderr,"emit arrays\n"); for ( i = 0; i < 15; i++ ) { GLubyte attrib = vimap_rev[i]; if (attrib != 255) { @@ -416,20 +142,20 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ) case 3: /* special handling to fix up fog. Will get us into trouble with vbos...*/ assert(attrib == VERT_ATTRIB_FOG); - if (!rmesa->tcl.vertex_data[i].buf) { + if (!rmesa->tcl.aos[i].bo) { if (ctx->VertexProgram._Enabled) - emit_vector( ctx, - &(rmesa->tcl.vertex_data[i]), - (char *)VB->AttribPtr[attrib]->data, - 1, - VB->AttribPtr[attrib]->stride, - count); + rcommon_emit_vector( ctx, + &(rmesa->tcl.aos[nr]), + (char *)VB->AttribPtr[attrib]->data, + 1, + VB->AttribPtr[attrib]->stride, + count); else - emit_vecfog( ctx, - &(rmesa->tcl.vertex_data[i]), - (char *)VB->AttribPtr[attrib]->data, - VB->AttribPtr[attrib]->stride, - count); + r200_emit_vecfog( ctx, + &(rmesa->tcl.aos[nr]), + (char *)VB->AttribPtr[attrib]->data, + VB->AttribPtr[attrib]->stride, + count); } vfmt0 |= R200_VTX_DISCRETE_FOG; goto after_emit; @@ -473,17 +199,17 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ) default: assert(0); } - if (!rmesa->tcl.vertex_data[i].buf) { - emit_vector( ctx, - &(rmesa->tcl.vertex_data[i]), - (char *)VB->AttribPtr[attrib]->data, - emitsize, - VB->AttribPtr[attrib]->stride, - count ); + if (!rmesa->tcl.aos[nr].bo) { + rcommon_emit_vector( ctx, + &(rmesa->tcl.aos[nr]), + (char *)VB->AttribPtr[attrib]->data, + emitsize, + VB->AttribPtr[attrib]->stride, + count ); } after_emit: assert(nr < 12); - component[nr++] = &rmesa->tcl.vertex_data[i]; + nr++; } } @@ -501,12 +227,11 @@ after_emit: void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs ) { r200ContextPtr rmesa = R200_CONTEXT( ctx ); - - /* only do it for changed inputs ? */ int i; - for (i = 0; i < 15; i++) { - if (newinputs & (1 << i)) - r200ReleaseDmaRegion( rmesa, - &rmesa->tcl.vertex_data[i], __FUNCTION__ ); + for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { + if (rmesa->tcl.aos[i].bo) { + radeon_bo_unref(rmesa->tcl.aos[i].bo); + rmesa->tcl.aos[i].bo = NULL; + } } } diff --git a/src/mesa/drivers/dri/r200/r200_pixel.c b/src/mesa/drivers/dri/r200/r200_pixel.c index 2797cbb3dc..354daef07f 100644 --- a/src/mesa/drivers/dri/r200/r200_pixel.c +++ b/src/mesa/drivers/dri/r200/r200_pixel.c @@ -51,7 +51,7 @@ check_color( const GLcontext *ctx, GLenum type, GLenum format, const void *pixels, GLint sz, GLint pitch ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - GLuint cpp = rmesa->r200Screen->cpp; + GLuint cpp = rmesa->radeon.radeonScreen->cpp; if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); @@ -137,8 +137,8 @@ clip_pixelrect( const GLcontext *ctx, if (*height <= 0) return GL_FALSE; - *size = ((*y + *height - 1) * rmesa->r200Screen->frontPitch + - (*x + *width - 1) * rmesa->r200Screen->cpp); + *size = ((*y + *height - 1) * rmesa->radeon.radeonScreen->frontPitch + + (*x + *width - 1) * rmesa->radeon.radeonScreen->cpp); return GL_TRUE; } @@ -153,19 +153,20 @@ r200TryReadPixels( GLcontext *ctx, r200ContextPtr rmesa = R200_CONTEXT(ctx); GLint pitch = pack->RowLength ? pack->RowLength : width; GLint blit_format; - GLuint cpp = rmesa->r200Screen->cpp; + GLuint cpp = rmesa->radeon.radeonScreen->cpp; GLint size = width * height * cpp; + return GL_FALSE; +#if 0 if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); /* Only accelerate reading to GART buffers. */ if ( !r200IsGartMemory(rmesa, pixels, - pitch * height * rmesa->r200Screen->cpp ) ) { + pitch * height * rmesa->radeon.radeonScreen->cpp ) ) { if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s: dest not GART\n", __FUNCTION__); - return GL_FALSE; } /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from @@ -180,7 +181,7 @@ r200TryReadPixels( GLcontext *ctx, if (!check_color(ctx, type, format, pack, pixels, size, pitch)) return GL_FALSE; - switch ( rmesa->r200Screen->cpp ) { + switch ( rmesa->radeon.radeonScreen->cpp ) { case 4: blit_format = R200_CP_COLOR_FORMAT_ARGB8888; break; @@ -197,14 +198,14 @@ r200TryReadPixels( GLcontext *ctx, * a full command buffer expects to be called unlocked. As a * workaround, immediately flush the buffer on aquiring the lock. */ - LOCK_HARDWARE( rmesa ); + LOCK_HARDWARE( &rmesa->radeon ); if (rmesa->store.cmd_used) - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); + rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); if (!clip_pixelrect(ctx, ctx->ReadBuffer, &x, &y, &width, &height, &size)) { - UNLOCK_HARDWARE( rmesa ); + UNLOCK_HARDWARE( &rmesa->radeon ); if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s totally clipped -- nothing to do\n", __FUNCTION__); @@ -212,14 +213,14 @@ r200TryReadPixels( GLcontext *ctx, } { - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; driRenderbuffer *drb = (driRenderbuffer *) ctx->ReadBuffer->_ColorReadBuffer; int nbox = dPriv->numClipRects; int src_offset = drb->offset - + rmesa->r200Screen->fbLocation; + + rmesa->radeon.radeonScreen->fbLocation; int src_pitch = drb->pitch * drb->cpp; int dst_offset = r200GartOffsetFromVirtual( rmesa, pixels ); - int dst_pitch = pitch * rmesa->r200Screen->cpp; + int dst_pitch = pitch * rmesa->radeon.radeonScreen->cpp; drm_clip_rect_t *box = dPriv->pClipRects; int i; @@ -257,12 +258,12 @@ r200TryReadPixels( GLcontext *ctx, bw, bh ); } - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); + rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); } - UNLOCK_HARDWARE( rmesa ); - - r200Finish( ctx ); /* required by GL */ + UNLOCK_HARDWARE( &rmesa->radeon ); + radeonFinish( ctx ); /* required by GL */ +#endif return GL_TRUE; } @@ -292,7 +293,7 @@ static void do_draw_pix( GLcontext *ctx, GLuint planemask) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; drm_clip_rect_t *box = dPriv->pClipRects; struct gl_renderbuffer *rb = ctx->ReadBuffer->_ColorDrawBuffers[0]; driRenderbuffer *drb = (driRenderbuffer *) rb; @@ -301,12 +302,12 @@ static void do_draw_pix( GLcontext *ctx, int blit_format; int size; int src_offset = r200GartOffsetFromVirtual( rmesa, pixels ); - int src_pitch = pitch * rmesa->r200Screen->cpp; + int src_pitch = pitch * rmesa->radeon.radeonScreen->cpp; if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); - - switch ( rmesa->r200Screen->cpp ) { +#if 0 + switch ( rmesa->radeon.radeonScreen->cpp ) { case 2: blit_format = R200_CP_COLOR_FORMAT_RGB565; break; @@ -318,17 +319,17 @@ static void do_draw_pix( GLcontext *ctx, } - LOCK_HARDWARE( rmesa ); + LOCK_HARDWARE( &rmesa->radeon ); if (rmesa->store.cmd_used) - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); + rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); y -= height; /* cope with pixel zoom */ if (!clip_pixelrect(ctx, ctx->DrawBuffer, &x, &y, &width, &height, &size)) { - UNLOCK_HARDWARE( rmesa ); + UNLOCK_HARDWARE( &rmesa->radeon ); return; } @@ -357,15 +358,16 @@ static void do_draw_pix( GLcontext *ctx, blit_format, src_pitch, src_offset, drb->pitch * drb->cpp, - drb->offset + rmesa->r200Screen->fbLocation, + drb->offset + rmesa->radeon.radeonScreen->fbLocation, bx - x, by - y, bx, by, bw, bh ); } - r200FlushCmdBufLocked( rmesa, __FUNCTION__ ); - r200WaitForIdleLocked( rmesa ); /* required by GL */ - UNLOCK_HARDWARE( rmesa ); + rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); + radeonWaitForIdleLocked( &rmesa->radeon ); /* required by GL */ + UNLOCK_HARDWARE( &rmesa->radeon ); +#endif } @@ -381,7 +383,7 @@ r200TryDrawPixels( GLcontext *ctx, r200ContextPtr rmesa = R200_CONTEXT(ctx); GLint pitch = unpack->RowLength ? unpack->RowLength : width; GLuint planemask; - GLuint cpp = rmesa->r200Screen->cpp; + GLuint cpp = rmesa->radeon.radeonScreen->cpp; GLint size = height * pitch * cpp; if (R200_DEBUG & DEBUG_PIXEL) @@ -395,7 +397,7 @@ r200TryDrawPixels( GLcontext *ctx, case GL_RGB: case GL_RGBA: case GL_BGRA: - planemask = r200PackColor(cpp, + planemask = radeonPackColor(cpp, ctx->Color.ColorMask[RCOMP], ctx->Color.ColorMask[GCOMP], ctx->Color.ColorMask[BCOMP], @@ -431,7 +433,7 @@ r200TryDrawPixels( GLcontext *ctx, return GL_FALSE; } - if ( r200IsGartMemory(rmesa, pixels, size) ) + if (0)// r200IsGartMemory(rmesa, pixels, size) ) { do_draw_pix( ctx, x, y, width, height, pitch, pixels, planemask ); return GL_TRUE; @@ -471,7 +473,7 @@ r200Bitmap( GLcontext *ctx, GLint px, GLint py, { r200ContextPtr rmesa = R200_CONTEXT(ctx); - if (rmesa->Fallback) + if (rmesa->radeon.Fallback) _swrast_Bitmap( ctx, px, py, width, height, unpack, bitmap ); else r200PointsBitmap( ctx, px, py, width, height, unpack, bitmap ); diff --git a/src/mesa/drivers/dri/r200/r200_reg.h b/src/mesa/drivers/dri/r200/r200_reg.h index 5ce287f7a5..526a624b69 100644 --- a/src/mesa/drivers/dri/r200/r200_reg.h +++ b/src/mesa/drivers/dri/r200/r200_reg.h @@ -463,8 +463,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R200_VSC_UPDATE_USER_COLOR_1_ENABLE 0x00020000 /* gap */ #define R200_SE_TCL_VECTOR_INDX_REG 0x2200 +# define RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT 16 +# define RADEON_VEC_INDX_DWORD_COUNT_SHIFT 28 #define R200_SE_TCL_VECTOR_DATA_REG 0x2204 #define R200_SE_TCL_SCALAR_INDX_REG 0x2208 +# define RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT 16 #define R200_SE_TCL_SCALAR_DATA_REG 0x220c /* gap */ #define R200_SE_TCL_MATRIX_SEL_0 0x2230 @@ -949,6 +952,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R200_LOD_BIAS_MASK (0xfff80000) #define R200_LOD_BIAS_SHIFT 19 #define R200_PP_TXSIZE_0 0x2c0c /* NPOT only */ +#define R200_PP_TX_WIDTHMASK_SHIFT 0 +#define R200_PP_TX_HEIGHTMASK_SHIFT 16 + #define R200_PP_TXPITCH_0 0x2c10 /* NPOT only */ #define R200_PP_BORDER_COLOR_0 0x2c14 #define R200_PP_CUBIC_FACES_0 0x2c18 diff --git a/src/mesa/drivers/dri/r200/r200_span.c b/src/mesa/drivers/dri/r200/r200_span.c deleted file mode 100644 index 9783678028..0000000000 --- a/src/mesa/drivers/dri/r200/r200_span.c +++ /dev/null @@ -1,307 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -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, sublicense, 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 NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - */ - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/colormac.h" -#include "swrast/swrast.h" - -#include "r200_context.h" -#include "r200_ioctl.h" -#include "r200_state.h" -#include "r200_span.h" -#include "r200_tex.h" - -#define DBG 0 - -/* - * Note that all information needed to access pixels in a renderbuffer - * should be obtained through the gl_renderbuffer parameter, not per-context - * information. - */ -#define LOCAL_VARS \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - const __DRIdrawablePrivate *dPriv = drb->dPriv; \ - const GLuint bottom = dPriv->h - 1; \ - GLubyte *buf = (GLubyte *) drb->flippedData \ - + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp; \ - GLuint p; \ - (void) p; - -#define LOCAL_DEPTH_VARS \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - const __DRIdrawablePrivate *dPriv = drb->dPriv; \ - const GLuint bottom = dPriv->h - 1; \ - GLuint xo = dPriv->x; \ - GLuint yo = dPriv->y; \ - GLubyte *buf = (GLubyte *) drb->Base.Data; - -#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS - -#define Y_FLIP(Y) (bottom - (Y)) - -#define HW_LOCK() - -#define HW_UNLOCK() - - - -/* ================================================================ - * Color buffer - */ - -/* 16 bit, RGB565 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_RGB -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5 - -#define TAG(x) r200##x##_RGB565 -#define TAG2(x,y) r200##x##_RGB565##y -#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2) -#include "spantmp2.h" - -/* 32 bit, ARGB8888 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_BGRA -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV - -#define TAG(x) r200##x##_ARGB8888 -#define TAG2(x,y) r200##x##_ARGB8888##y -#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4) -#include "spantmp2.h" - - -/* ================================================================ - * Depth buffer - */ - -/* The Radeon family has depth tiling on all the time, so we have to convert - * the x,y coordinates into the memory bus address (mba) in the same - * manner as the engine. In each case, the linear block address (ba) - * is calculated, and then wired with x and y to produce the final - * memory address. - * The chip will do address translation on its own if the surface registers - * are set up correctly. It is not quite enough to get it working with hyperz too... - */ - -/* extract bit 'b' of x, result is zero or one */ -#define BIT(x,b) ((x & (1<<b))>>b) - -static GLuint -r200_mba_z32( driRenderbuffer *drb, GLint x, GLint y ) -{ - GLuint pitch = drb->pitch; - if (drb->depthHasSurface) { - return 4 * (x + y * pitch); - } - else { - GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x7FF) >> 5); - GLuint a = - (BIT(x,0) << 2) | - (BIT(y,0) << 3) | - (BIT(x,1) << 4) | - (BIT(y,1) << 5) | - (BIT(x,3) << 6) | - (BIT(x,4) << 7) | - (BIT(x,2) << 8) | - (BIT(y,2) << 9) | - (BIT(y,3) << 10) | - (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) | - ((b >> 1) << 12); - return a; - } -} - -static GLuint -r200_mba_z16( driRenderbuffer *drb, GLint x, GLint y ) -{ - GLuint pitch = drb->pitch; - if (drb->depthHasSurface) { - return 2 * (x + y * pitch); - } - else { - GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x7FF) >> 6); - GLuint a = - (BIT(x,0) << 1) | - (BIT(y,0) << 2) | - (BIT(x,1) << 3) | - (BIT(y,1) << 4) | - (BIT(x,2) << 5) | - (BIT(x,4) << 6) | - (BIT(x,5) << 7) | - (BIT(x,3) << 8) | - (BIT(y,2) << 9) | - (BIT(y,3) << 10) | - (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) | - ((b >> 1) << 12); - return a; - } -} - - -/* 16-bit depth buffer functions - */ -#define VALUE_TYPE GLushort - -#define WRITE_DEPTH( _x, _y, d ) \ - *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo )) = d; - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo )); - -#define TAG(x) r200##x##_z16 -#include "depthtmp.h" - - -/* 24 bit depth, 8 bit stencil depthbuffer functions - */ -#define VALUE_TYPE GLuint - -#define WRITE_DEPTH( _x, _y, d ) \ -do { \ - GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ - tmp &= 0xff000000; \ - tmp |= ((d) & 0x00ffffff); \ - *(GLuint *)(buf + offset) = tmp; \ -} while (0) - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + r200_mba_z32( drb, _x + xo, \ - _y + yo )) & 0x00ffffff; - -#define TAG(x) r200##x##_z24_s8 -#include "depthtmp.h" - - -/* ================================================================ - * Stencil buffer - */ - -/* 24 bit depth, 8 bit stencil depthbuffer functions - */ -#define WRITE_STENCIL( _x, _y, d ) \ -do { \ - GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ - tmp &= 0x00ffffff; \ - tmp |= (((d) & 0xff) << 24); \ - *(GLuint *)(buf + offset) = tmp; \ -} while (0) - -#define READ_STENCIL( d, _x, _y ) \ -do { \ - GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ - tmp &= 0xff000000; \ - d = tmp >> 24; \ -} while (0) - -#define TAG(x) r200##x##_z24_s8 -#include "stenciltmp.h" - - -/* Move locking out to get reasonable span performance (10x better - * than doing this in HW_LOCK above). WaitForIdle() is the main - * culprit. - */ - -static void r200SpanRenderStart( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - - R200_FIREVERTICES( rmesa ); - LOCK_HARDWARE( rmesa ); - r200WaitForIdleLocked( rmesa ); - - /* Read & rewrite the first pixel in the frame buffer. This should - * be a noop, right? In fact without this conform fails as reading - * from the framebuffer sometimes produces old results -- the - * on-card read cache gets mixed up and doesn't notice that the - * framebuffer has been updated. - * - * In the worst case this is buggy too as p might get the wrong - * value first time, so really need a hidden pixel somewhere for this. - */ - { - int p; - driRenderbuffer *drb = - (driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0]; - volatile int *buf = - (volatile int *)(rmesa->dri.screen->pFB + drb->offset); - p = *buf; - *buf = p; - } -} - -static void r200SpanRenderFinish( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - _swrast_flush( ctx ); - UNLOCK_HARDWARE( rmesa ); -} - -void r200InitSpanFuncs( GLcontext *ctx ) -{ - struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); - swdd->SpanRenderStart = r200SpanRenderStart; - swdd->SpanRenderFinish = r200SpanRenderFinish; -} - - - -/** - * Plug in the Get/Put routines for the given driRenderbuffer. - */ -void -radeonSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) -{ - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - r200InitPointers_RGB565(&drb->Base); - } - else { - r200InitPointers_ARGB8888(&drb->Base); - } - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { - r200InitDepthPointers_z16(&drb->Base); - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { - r200InitDepthPointers_z24_s8(&drb->Base); - } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { - r200InitStencilPointers_z24_s8(&drb->Base); - } -} diff --git a/src/mesa/drivers/dri/r200/r200_span.h b/src/mesa/drivers/dri/r200/r200_span.h deleted file mode 100644 index bae5644309..0000000000 --- a/src/mesa/drivers/dri/r200/r200_span.h +++ /dev/null @@ -1,45 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -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, sublicense, 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 NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - */ - -#ifndef __R200_SPAN_H__ -#define __R200_SPAN_H__ - -#include "drirenderbuffer.h" - -extern void r200InitSpanFuncs( GLcontext *ctx ); - -extern void -radeonSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis); - -#endif diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c index 0eaaaf69ac..5eb61c8804 100644 --- a/src/mesa/drivers/dri/r200/r200_state.c +++ b/src/mesa/drivers/dri/r200/r200_state.c @@ -47,6 +47,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_pipeline.h" #include "swrast_setup/swrast_setup.h" +#include "radeon_common.h" +#include "radeon_mipmap_tree.h" #include "r200_context.h" #include "r200_ioctl.h" #include "r200_state.h" @@ -114,8 +116,8 @@ static void r200BlendColor( GLcontext *ctx, const GLfloat cf[4] ) CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]); CLAMPED_FLOAT_TO_UBYTE(color[2], cf[2]); CLAMPED_FLOAT_TO_UBYTE(color[3], cf[3]); - if (rmesa->r200Screen->drmSupportsBlendColor) - rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = r200PackColor( 4, color[0], color[1], color[2], color[3] ); + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) + rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = radeonPackColor( 4, color[0], color[1], color[2], color[3] ); } /** @@ -213,7 +215,7 @@ static void r200_set_blend_state( GLcontext * ctx ) R200_STATECHANGE( rmesa, ctx ); - if (rmesa->r200Screen->drmSupportsBlendColor) { + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) { if (ctx->Color.ColorLogicOpEnabled) { rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl | R200_ROP_ENABLE; rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = eqn | func; @@ -278,7 +280,7 @@ static void r200_set_blend_state( GLcontext * ctx ) return; } - if (!rmesa->r200Screen->drmSupportsBlendColor) { + if (!rmesa->radeon.radeonScreen->drmSupportsBlendColor) { rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func; return; } @@ -383,10 +385,10 @@ static void r200ClearDepth( GLcontext *ctx, GLclampd d ) switch ( format ) { case R200_DEPTH_FORMAT_16BIT_INT_Z: - rmesa->state.depth.clear = d * 0x0000ffff; + rmesa->radeon.state.depth.clear = d * 0x0000ffff; break; case R200_DEPTH_FORMAT_24BIT_INT_Z: - rmesa->state.depth.clear = d * 0x00ffffff; + rmesa->radeon.state.depth.clear = d * 0x00ffffff; break; } } @@ -480,7 +482,7 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param ) case GL_FOG_COLOR: R200_STATECHANGE( rmesa, ctx ); UNCLAMPED_FLOAT_TO_RGB_CHAN( col, ctx->Fog.Color ); - i = r200PackColor( 4, col[0], col[1], col[2], 0 ); + i = radeonPackColor( 4, col[0], col[1], col[2], 0 ); rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_COLOR_MASK; rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= i; break; @@ -521,102 +523,6 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param ) } } - -/* ============================================================= - * Scissoring - */ - - -static GLboolean intersect_rect( drm_clip_rect_t *out, - drm_clip_rect_t *a, - drm_clip_rect_t *b ) -{ - *out = *a; - if ( b->x1 > out->x1 ) out->x1 = b->x1; - if ( b->y1 > out->y1 ) out->y1 = b->y1; - if ( b->x2 < out->x2 ) out->x2 = b->x2; - if ( b->y2 < out->y2 ) out->y2 = b->y2; - if ( out->x1 >= out->x2 ) return GL_FALSE; - if ( out->y1 >= out->y2 ) return GL_FALSE; - return GL_TRUE; -} - - -void r200RecalcScissorRects( r200ContextPtr rmesa ) -{ - drm_clip_rect_t *out; - int i; - - /* Grow cliprect store? - */ - if (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) { - while (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) { - rmesa->state.scissor.numAllocedClipRects += 1; /* zero case */ - rmesa->state.scissor.numAllocedClipRects *= 2; - } - - if (rmesa->state.scissor.pClipRects) - FREE(rmesa->state.scissor.pClipRects); - - rmesa->state.scissor.pClipRects = - MALLOC( rmesa->state.scissor.numAllocedClipRects * - sizeof(drm_clip_rect_t) ); - - if ( rmesa->state.scissor.pClipRects == NULL ) { - rmesa->state.scissor.numAllocedClipRects = 0; - return; - } - } - - out = rmesa->state.scissor.pClipRects; - rmesa->state.scissor.numClipRects = 0; - - for ( i = 0 ; i < rmesa->numClipRects ; i++ ) { - if ( intersect_rect( out, - &rmesa->pClipRects[i], - &rmesa->state.scissor.rect ) ) { - rmesa->state.scissor.numClipRects++; - out++; - } - } -} - - -static void r200UpdateScissor( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if ( rmesa->dri.drawable ) { - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; - - int x = ctx->Scissor.X; - int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height; - int w = ctx->Scissor.X + ctx->Scissor.Width - 1; - int h = dPriv->h - ctx->Scissor.Y - 1; - - rmesa->state.scissor.rect.x1 = x + dPriv->x; - rmesa->state.scissor.rect.y1 = y + dPriv->y; - rmesa->state.scissor.rect.x2 = w + dPriv->x + 1; - rmesa->state.scissor.rect.y2 = h + dPriv->y + 1; - - r200RecalcScissorRects( rmesa ); - } -} - - -static void r200Scissor( GLcontext *ctx, - GLint x, GLint y, GLsizei w, GLsizei h ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if ( ctx->Scissor.Enabled ) { - R200_FIREVERTICES( rmesa ); /* don't pipeline cliprect changes */ - r200UpdateScissor( ctx ); - } - -} - - /* ============================================================= * Culling */ @@ -803,7 +709,7 @@ static void r200ColorMask( GLcontext *ctx, GLboolean b, GLboolean a ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - GLuint mask = r200PackColor( rmesa->r200Screen->cpp, + GLuint mask = radeonPackColor( rmesa->radeon.radeonScreen->cpp, ctx->Color.ColorMask[RCOMP], ctx->Color.ColorMask[GCOMP], ctx->Color.ColorMask[BCOMP], @@ -834,7 +740,7 @@ static void r200PolygonOffset( GLcontext *ctx, GLfloat factor, GLfloat units ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - float_ui32_type constant = { units * rmesa->state.depth.scale }; + float_ui32_type constant = { units * rmesa->radeon.state.depth.scale }; float_ui32_type factoru = { factor }; /* factor *= 2; */ @@ -861,15 +767,15 @@ static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask ) /* TODO: push this into cmd mechanism */ - R200_FIREVERTICES( rmesa ); - LOCK_HARDWARE( rmesa ); + radeon_firevertices(&rmesa->radeon); + LOCK_HARDWARE( &rmesa->radeon ); /* FIXME: Use window x,y offsets into stipple RAM. */ stipple.mask = rmesa->state.stipple.mask; - drmCommandWrite( rmesa->dri.fd, DRM_RADEON_STIPPLE, + drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_STIPPLE, &stipple, sizeof(stipple) ); - UNLOCK_HARDWARE( rmesa ); + UNLOCK_HARDWARE( &rmesa->radeon ); } static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode ) @@ -881,7 +787,7 @@ static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode ) * cases work. */ TCL_FALLBACK( ctx, R200_TCL_FALLBACK_UNFILLED, flag); - if (rmesa->TclFallback) { + if (rmesa->radeon.TclFallback) { r200ChooseRenderState( ctx ); r200ChooseVertexState( ctx ); } @@ -958,7 +864,7 @@ static void r200UpdateSpecular( GLcontext *ctx ) /* Update vertex/render formats */ - if (rmesa->TclFallback) { + if (rmesa->radeon.TclFallback) { r200ChooseRenderState( ctx ); r200ChooseVertexState( ctx ); } @@ -1430,7 +1336,7 @@ static void r200LightModelfv( GLcontext *ctx, GLenum pname, rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHT_TWOSIDE; else rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= ~(R200_LIGHT_TWOSIDE); - if (rmesa->TclFallback) { + if (rmesa->radeon.TclFallback) { r200ChooseRenderState( ctx ); r200ChooseVertexState( ctx ); } @@ -1675,7 +1581,7 @@ static void r200ClearStencil( GLcontext *ctx, GLint s ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - rmesa->state.stencil.clear = + rmesa->radeon.state.stencil.clear = ((GLuint) (ctx->Stencil.Clear & 0xff) | (0xff << R200_STENCIL_MASK_SHIFT) | ((ctx->Stencil.WriteMask[0] & 0xff) << R200_STENCIL_WRITEMASK_SHIFT)); @@ -1700,19 +1606,18 @@ static void r200ClearStencil( GLcontext *ctx, GLint s ) void r200UpdateWindow( GLcontext *ctx ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; - GLfloat xoffset = (GLfloat)dPriv->x; - GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h; + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; + GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0; + GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0; const GLfloat *v = ctx->Viewport._WindowMap.m; float_ui32_type sx = { v[MAT_SX] }; float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X }; float_ui32_type sy = { - v[MAT_SY] }; float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y }; - float_ui32_type sz = { v[MAT_SZ] * rmesa->state.depth.scale }; - float_ui32_type tz = { v[MAT_TZ] * rmesa->state.depth.scale }; + float_ui32_type sz = { v[MAT_SZ] * rmesa->radeon.state.depth.scale }; + float_ui32_type tz = { v[MAT_TZ] * rmesa->radeon.state.depth.scale }; - R200_FIREVERTICES( rmesa ); R200_STATECHANGE( rmesa, vpt ); rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE] = sx.ui32; @@ -1733,6 +1638,8 @@ static void r200Viewport( GLcontext *ctx, GLint x, GLint y, * values, or keep the originals hanging around. */ r200UpdateWindow( ctx ); + + radeon_viewport(ctx, x, y, width, height); } static void r200DepthRange( GLcontext *ctx, GLclampd nearval, @@ -1744,7 +1651,7 @@ static void r200DepthRange( GLcontext *ctx, GLclampd nearval, void r200UpdateViewportOffset( GLcontext *ctx ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; + __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; GLfloat xoffset = (GLfloat)dPriv->x; GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h; const GLfloat *v = ctx->Viewport._WindowMap.m; @@ -1774,8 +1681,8 @@ void r200UpdateViewportOffset( GLcontext *ctx ) R200_STIPPLE_Y_OFFSET_MASK); /* add magic offsets, then invert */ - stx = 31 - ((rmesa->dri.drawable->x - 1) & R200_STIPPLE_COORD_MASK); - sty = 31 - ((rmesa->dri.drawable->y + rmesa->dri.drawable->h - 1) + stx = 31 - ((rmesa->radeon.dri.drawable->x - 1) & R200_STIPPLE_COORD_MASK); + sty = 31 - ((rmesa->radeon.dri.drawable->y + rmesa->radeon.dri.drawable->h - 1) & R200_STIPPLE_COORD_MASK); m |= ((stx << R200_STIPPLE_X_OFFSET_SHIFT) | @@ -1788,7 +1695,7 @@ void r200UpdateViewportOffset( GLcontext *ctx ) } } - r200UpdateScissor( ctx ); + radeonUpdateScissor( ctx ); } @@ -1805,7 +1712,7 @@ static void r200ClearColor( GLcontext *ctx, const GLfloat c[4] ) CLAMPED_FLOAT_TO_UBYTE(color[1], c[1]); CLAMPED_FLOAT_TO_UBYTE(color[2], c[2]); CLAMPED_FLOAT_TO_UBYTE(color[3], c[3]); - rmesa->state.color.clear = r200PackColor( rmesa->r200Screen->cpp, + rmesa->radeon.state.color.clear = radeonPackColor( rmesa->radeon.radeonScreen->cpp, color[0], color[1], color[2], color[3] ); } @@ -1848,96 +1755,6 @@ static void r200LogicOpCode( GLcontext *ctx, GLenum opcode ) rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = r200_rop_tab[rop]; } - -/* - * Set up the cliprects for either front or back-buffer drawing. - */ -void r200SetCliprects( r200ContextPtr rmesa ) -{ - __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; - __DRIdrawablePrivate *const readable = rmesa->dri.readable; - GLframebuffer *const draw_fb = (GLframebuffer*) drawable->driverPrivate; - GLframebuffer *const read_fb = (GLframebuffer*) readable->driverPrivate; - - if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BIT_BACK_LEFT) { - /* Can't ignore 2d windows if we are page flipping. - */ - if ( drawable->numBackClipRects == 0 || rmesa->doPageFlip ) { - rmesa->numClipRects = drawable->numClipRects; - rmesa->pClipRects = drawable->pClipRects; - } - else { - rmesa->numClipRects = drawable->numBackClipRects; - rmesa->pClipRects = drawable->pBackClipRects; - } - } - else { - /* front buffer (or none, or multiple buffers) */ - rmesa->numClipRects = drawable->numClipRects; - rmesa->pClipRects = drawable->pClipRects; - } - - if ((draw_fb->Width != drawable->w) || (draw_fb->Height != drawable->h)) { - _mesa_resize_framebuffer(rmesa->glCtx, draw_fb, - drawable->w, drawable->h); - draw_fb->Initialized = GL_TRUE; - } - - if (drawable != readable) { - if ((read_fb->Width != readable->w) || - (read_fb->Height != readable->h)) { - _mesa_resize_framebuffer(rmesa->glCtx, read_fb, - readable->w, readable->h); - read_fb->Initialized = GL_TRUE; - } - } - - if (rmesa->state.scissor.enabled) - r200RecalcScissorRects( rmesa ); - - rmesa->lastStamp = drawable->lastStamp; -} - - -static void r200DrawBuffer( GLcontext *ctx, GLenum mode ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr( mode )); - - R200_FIREVERTICES(rmesa); /* don't pipeline cliprect changes */ - - if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) { - /* 0 (GL_NONE) buffers or multiple color drawing buffers */ - FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE ); - return; - } - - switch ( ctx->DrawBuffer->_ColorDrawBufferIndexes[0] ) { - case BUFFER_FRONT_LEFT: - case BUFFER_BACK_LEFT: - FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_FALSE ); - break; - default: - FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE ); - return; - } - - r200SetCliprects( rmesa ); - - /* We'll set the drawing engine's offset/pitch parameters later - * when we update other state. - */ -} - - -static void r200ReadBuffer( GLcontext *ctx, GLenum mode ) -{ - /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */ -} - /* ============================================================= * State enable/disable */ @@ -2013,10 +1830,10 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state ) R200_STATECHANGE(rmesa, ctx ); if ( state ) { rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_ENABLE; - rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->state.color.roundEnable; + rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->radeon.state.color.roundEnable; } else { rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_DITHER_ENABLE; - rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable; + rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable; } break; @@ -2031,7 +1848,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state ) rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_TCL_FOG_MASK; } r200UpdateSpecular( ctx ); /* for PK_SPEC */ - if (rmesa->TclFallback) + if (rmesa->radeon.TclFallback) r200ChooseVertexState( ctx ); _mesa_allow_light_in_model( ctx, !state ); break; @@ -2068,7 +1885,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state ) case GL_LIGHTING: r200UpdateSpecular(ctx); /* for reflection map fixup - might set recheck_texgen for all units too */ - rmesa->NewGLState |= _NEW_TEXTURE; + rmesa->radeon.NewGLState |= _NEW_TEXTURE; break; case GL_LINE_SMOOTH: @@ -2181,13 +1998,13 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state ) } case GL_SCISSOR_TEST: - R200_FIREVERTICES( rmesa ); - rmesa->state.scissor.enabled = state; - r200UpdateScissor( ctx ); + radeon_firevertices(&rmesa->radeon); + rmesa->radeon.state.scissor.enabled = state; + radeonUpdateScissor( ctx ); break; case GL_STENCIL_TEST: - if ( rmesa->state.stencil.hwBuffer ) { + if ( rmesa->radeon.state.stencil.hwBuffer ) { R200_STATECHANGE( rmesa, ctx ); if ( state ) { rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_STENCIL_ENABLE; @@ -2432,64 +2249,88 @@ static void update_texturematrix( GLcontext *ctx ) } } - - -/** - * Tell the card where to render (offset, pitch). - * Effected by glDrawBuffer, etc - */ -void -r200UpdateDrawBuffer(GLcontext *ctx) +static GLboolean r200ValidateBuffers(GLcontext *ctx) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - driRenderbuffer *drb; - - if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { - /* draw to front */ - drb = (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - } - else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) { - /* draw to back */ - drb = (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - } - else { - /* drawing to multiple buffers, or none */ - return; + struct radeon_cs_space_check bos[8]; + struct radeon_renderbuffer *rrb; + int num_bo = 0; + int i; + int flushed = 0, ret; +again: + num_bo = 0; + + rrb = radeon_get_colorbuffer(&rmesa->radeon); + /* color buffer */ + if (rrb && rrb->bo) { + bos[num_bo].bo = rrb->bo; + bos[num_bo].read_domains = 0; + bos[num_bo].write_domain = RADEON_GEM_DOMAIN_VRAM; + bos[num_bo].new_accounted = 0; + num_bo++; + } + + /* depth buffer */ + rrb = radeon_get_depthbuffer(&rmesa->radeon); + /* color buffer */ + if (rrb && rrb->bo) { + bos[num_bo].bo = rrb->bo; + bos[num_bo].read_domains = 0; + bos[num_bo].write_domain = RADEON_GEM_DOMAIN_VRAM; + bos[num_bo].new_accounted = 0; + num_bo++; + } + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) { + radeonTexObj *t; + + if (!ctx->Texture.Unit[i]._ReallyEnabled) + continue; + + t = radeon_tex_obj(ctx->Texture.Unit[i]._Current); + bos[num_bo].bo = t->mt->bo; + bos[num_bo].read_domains = RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM; + bos[num_bo].write_domain = 0; + bos[num_bo].new_accounted = 0; + num_bo++; } - - assert(drb); - assert(drb->flippedPitch); - - R200_STATECHANGE( rmesa, ctx ); - - /* Note: we used the (possibly) page-flipped values */ - rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] - = ((drb->flippedOffset + rmesa->r200Screen->fbLocation) - & R200_COLOROFFSET_MASK); - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch; - if (rmesa->sarea->tiling_enabled) { - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE; + + ret = radeon_cs_space_check(rmesa->radeon.cmdbuf.cs, bos, num_bo); + if (ret == RADEON_CS_SPACE_OP_TO_BIG) + return GL_FALSE; + if (ret == RADEON_CS_SPACE_FLUSH) { + radeonFlush(ctx); + if (flushed) + return GL_FALSE; + flushed = 1; + goto again; } + return GL_TRUE; } - - -void r200ValidateState( GLcontext *ctx ) +GLboolean r200ValidateState( GLcontext *ctx ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - GLuint new_state = rmesa->NewGLState; + GLuint new_state = rmesa->radeon.NewGLState; if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { - r200UpdateDrawBuffer(ctx); + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + + R200_STATECHANGE(rmesa, ctx); } if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM)) { r200UpdateTextureState( ctx ); - new_state |= rmesa->NewGLState; /* may add TEXTURE_MATRIX */ + new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */ r200UpdateLocalViewer( ctx ); } + /* we need to do a space check here */ + if (!r200ValidateBuffers(ctx)) + return GL_FALSE; + /* FIXME: don't really need most of these when vertex progs are enabled */ /* Need an event driven matrix update? @@ -2533,7 +2374,8 @@ void r200ValidateState( GLcontext *ctx ) else TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, 0); } - rmesa->NewGLState = 0; + rmesa->radeon.NewGLState = 0; + return GL_TRUE; } @@ -2544,7 +2386,7 @@ static void r200InvalidateState( GLcontext *ctx, GLuint new_state ) _vbo_InvalidateState( ctx, new_state ); _tnl_InvalidateState( ctx, new_state ); _ae_invalidate_state( ctx, new_state ); - R200_CONTEXT(ctx)->NewGLState |= new_state; + R200_CONTEXT(ctx)->radeon.NewGLState |= new_state; } /* A hack. The r200 can actually cope just fine with materials @@ -2573,12 +2415,13 @@ static void r200WrapRunPipeline( GLcontext *ctx ) GLboolean has_material; if (0) - fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->NewGLState); + fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->radeon.NewGLState); /* Validate state: */ - if (rmesa->NewGLState) - r200ValidateState( ctx ); + if (rmesa->radeon.NewGLState) + if (!r200ValidateState( ctx )) + FALLBACK(rmesa, RADEON_FALLBACK_TEXTURE, GL_TRUE); has_material = !ctx->VertexProgram._Enabled && ctx->Light.Enabled && check_material( ctx ); @@ -2603,8 +2446,8 @@ void r200InitStateFuncs( struct dd_function_table *functions ) functions->UpdateState = r200InvalidateState; functions->LightingSpaceChange = r200LightingSpaceChange; - functions->DrawBuffer = r200DrawBuffer; - functions->ReadBuffer = r200ReadBuffer; + functions->DrawBuffer = radeonDrawBuffer; + functions->ReadBuffer = radeonReadBuffer; functions->AlphaFunc = r200AlphaFunc; functions->BlendColor = r200BlendColor; @@ -2636,7 +2479,7 @@ void r200InitStateFuncs( struct dd_function_table *functions ) functions->PointParameterfv = r200PointParameter; functions->PointSize = r200PointSize; functions->RenderMode = r200RenderMode; - functions->Scissor = r200Scissor; + functions->Scissor = radeonScissor; functions->ShadeModel = r200ShadeModel; functions->StencilFuncSeparate = r200StencilFuncSeparate; functions->StencilMaskSeparate = r200StencilMaskSeparate; diff --git a/src/mesa/drivers/dri/r200/r200_state.h b/src/mesa/drivers/dri/r200/r200_state.h index a917163a00..1dddbfdbfe 100644 --- a/src/mesa/drivers/dri/r200/r200_state.h +++ b/src/mesa/drivers/dri/r200/r200_state.h @@ -43,13 +43,11 @@ extern void r200InitTnlFuncs( GLcontext *ctx ); extern void r200UpdateMaterial( GLcontext *ctx ); -extern void r200SetCliprects( r200ContextPtr rmesa ); -extern void r200RecalcScissorRects( r200ContextPtr rmesa ); extern void r200UpdateViewportOffset( GLcontext *ctx ); extern void r200UpdateWindow( GLcontext *ctx ); extern void r200UpdateDrawBuffer(GLcontext *ctx); -extern void r200ValidateState( GLcontext *ctx ); +extern GLboolean r200ValidateState( GLcontext *ctx ); extern void r200PrintDirty( r200ContextPtr rmesa, const char *msg ); @@ -59,7 +57,7 @@ extern void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ); #define FALLBACK( rmesa, bit, mode ) do { \ if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n", \ __FUNCTION__, bit, mode ); \ - r200Fallback( rmesa->glCtx, bit, mode ); \ + r200Fallback( rmesa->radeon.glCtx, bit, mode ); \ } while (0) extern void r200LightingSpaceChange( GLcontext *ctx ); diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c index 9e4677eda4..a71f33ca3b 100644 --- a/src/mesa/drivers/dri/r200/r200_state_init.c +++ b/src/mesa/drivers/dri/r200/r200_state_init.c @@ -43,6 +43,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_pipeline.h" #include "swrast_setup/swrast_setup.h" +#include "radeon_common.h" +#include "radeon_mipmap_tree.h" #include "r200_context.h" #include "r200_ioctl.h" #include "r200_state.h" @@ -52,31 +54,145 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "xmlpool.h" +/* New (1.3) state mechanism. 3 commands (packet, scalar, vector) in + * 1.3 cmdbuffers allow all previous state to be updated as well as + * the tcl scalar and vector areas. + */ +static struct { + int start; + int len; + const char *name; +} packet[RADEON_MAX_STATE_PACKETS] = { + {RADEON_PP_MISC, 7, "RADEON_PP_MISC"}, + {RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"}, + {RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"}, + {RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"}, + {RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"}, + {RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"}, + {RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"}, + {RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"}, + {RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"}, + {RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"}, + {RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"}, + {RADEON_RE_MISC, 1, "RADEON_RE_MISC"}, + {RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"}, + {RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"}, + {RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"}, + {RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"}, + {RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"}, + {RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"}, + {RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"}, + {RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"}, + {RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17, + "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"}, + {R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"}, + {R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"}, + {R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"}, + {R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"}, + {R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"}, + {R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"}, + {R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"}, + {R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"}, + {R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"}, + {R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"}, + {R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"}, + {R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"}, + {R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"}, + {R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"}, + {R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"}, + {R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"}, + {R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"}, + {R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"}, + {R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"}, + {R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"}, + {R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"}, + {R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"}, + {R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"}, + {R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"}, + {R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"}, + {R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"}, + {R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"}, + {R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"}, + {R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1, + "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"}, + {R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"}, + {R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"}, + {R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"}, + {R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"}, + {R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"}, + {R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"}, + {R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"}, + {R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"}, + {R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"}, + {R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"}, + {R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4, + "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"}, + {R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"}, /* 61 */ + {R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */ + {R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"}, + {R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"}, + {R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"}, + {R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"}, + {R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"}, + {R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"}, + {R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"}, + {R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"}, + {R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"}, + {R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"}, + {RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"}, + {RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"}, + {RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"}, + {R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"}, + {R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"}, + {RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"}, + {RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"}, + {RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"}, + {RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"}, + {RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"}, + {RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"}, + {R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"}, + {R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"}, /* 85 */ + {R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"}, + {R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"}, + {R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"}, + {R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"}, + {R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"}, + {R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"}, + {R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"}, + {R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"}, + {R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"}, +}; + /* ============================================================= * State initialization */ void r200PrintDirty( r200ContextPtr rmesa, const char *msg ) { - struct r200_state_atom *l; + struct radeon_state_atom *l; fprintf(stderr, msg); fprintf(stderr, ": "); - foreach(l, &rmesa->hw.atomlist) { - if (l->dirty || rmesa->hw.all_dirty) + foreach(l, &rmesa->radeon.hw.atomlist) { + if (l->dirty || rmesa->radeon.hw.all_dirty) fprintf(stderr, "%s, ", l->name); } fprintf(stderr, "\n"); } -static int cmdpkt( int id ) +static int cmdpkt( r200ContextPtr rmesa, int id ) { drm_radeon_cmd_header_t h; - h.i = 0; - h.packet.cmd_type = RADEON_CMD_PACKET; - h.packet.packet_id = id; + + if (rmesa->radeon.radeonScreen->kernel_mm) { + return CP_PACKET0(packet[id].start, packet[id].len - 1); + } else { + h.i = 0; + h.packet.cmd_type = RADEON_CMD_PACKET; + h.packet.packet_id = id; + } return h.i; } @@ -127,150 +243,480 @@ static int cmdscl2( int offset, int stride, int count ) } #define CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx, int idx ) \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \ { \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - (void) idx; \ (void) rmesa; \ - return FLAG; \ + return (FLAG) ? atom->cmd_size : 0; \ } #define TCL_CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx, int idx ) \ -{ \ - r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - (void) idx; \ - return !rmesa->TclFallback && !ctx->VertexProgram._Enabled && (FLAG); \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \ +{ \ + r200ContextPtr rmesa = R200_CONTEXT(ctx); \ + return (!rmesa->radeon.TclFallback && !ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size : 0; \ } #define TCL_OR_VP_CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx, int idx ) \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ { \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - (void) idx; \ - return !rmesa->TclFallback && (FLAG); \ + return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size : 0; \ } #define VP_CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx, int idx ) \ -{ \ - r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - (void) idx; \ - return !rmesa->TclFallback && ctx->VertexProgram._Enabled && (FLAG); \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ +{ \ + r200ContextPtr rmesa = R200_CONTEXT(ctx); \ + (void) atom; \ + return (!rmesa->radeon.TclFallback && ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size : 0; \ } - CHECK( always, GL_TRUE ) CHECK( never, GL_FALSE ) CHECK( tex_any, ctx->Texture._EnabledUnits ) CHECK( tf, (ctx->Texture._EnabledUnits && !ctx->ATIFragmentShader._Enabled) ); -CHECK( tex_pair, (rmesa->state.texture.unit[idx].unitneeded | rmesa->state.texture.unit[idx & ~1].unitneeded) ) -CHECK( tex, rmesa->state.texture.unit[idx].unitneeded ) +CHECK( tex_pair, (rmesa->state.texture.unit[atom->idx].unitneeded | rmesa->state.texture.unit[atom->idx & ~1].unitneeded) ) +CHECK( tex, rmesa->state.texture.unit[atom->idx].unitneeded ) CHECK( pix_zero, !ctx->ATIFragmentShader._Enabled ) -CHECK( texenv, (rmesa->state.envneeded & (1 << idx) && !ctx->ATIFragmentShader._Enabled) ) + CHECK( texenv, (rmesa->state.envneeded & (1 << (atom->idx)) && !ctx->ATIFragmentShader._Enabled) ) CHECK( afs_pass1, (ctx->ATIFragmentShader._Enabled && (ctx->ATIFragmentShader.Current->NumPasses > 1)) ) CHECK( afs, ctx->ATIFragmentShader._Enabled ) -CHECK( tex_cube, rmesa->state.texture.unit[idx].unitneeded & TEXTURE_CUBE_BIT ) +CHECK( tex_cube, rmesa->state.texture.unit[atom->idx].unitneeded & TEXTURE_CUBE_BIT ) TCL_CHECK( tcl_fog, ctx->Fog.Enabled ) TCL_CHECK( tcl, GL_TRUE ) -TCL_CHECK( tcl_tex, rmesa->state.texture.unit[idx].unitneeded ) +TCL_CHECK( tcl_tex, rmesa->state.texture.unit[atom->idx].unitneeded ) TCL_CHECK( tcl_lighting, ctx->Light.Enabled ) -TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[idx].Enabled ) -TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << idx)) ) +TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[atom->idx].Enabled ) +TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << (atom->idx))) ) TCL_OR_VP_CHECK( tcl_or_vp, GL_TRUE ) VP_CHECK( tcl_vp, GL_TRUE ) VP_CHECK( tcl_vp_size, ctx->VertexProgram.Current->Base.NumNativeInstructions > 64 ) VP_CHECK( tcl_vpp_size, ctx->VertexProgram.Current->Base.NumNativeParameters > 96 ) +#define OUT_VEC(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0)); \ + OUT_BATCH(0); \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0)); \ + OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1)); \ + OUT_BATCH_TABLE((data), h.vectors.count); \ + } while(0) + +#define OUT_VECLINEAR(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + uint32_t _start = h.veclinear.addr_lo | (h.veclinear.addr_hi << 8); \ + uint32_t _sz = h.veclinear.count * 4; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0)); \ + OUT_BATCH(0); \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0)); \ + OUT_BATCH(_start | (1 << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, _sz - 1)); \ + OUT_BATCH_TABLE((data), _sz); \ + } while(0) + +#define OUT_SCL(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0)); \ + OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1)); \ + OUT_BATCH_TABLE((data), h.scalars.count); \ + } while(0) + +#define OUT_SCL2(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0)); \ + OUT_BATCH((h.scalars.offset + 0x100) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1)); \ + OUT_BATCH_TABLE((data), h.scalars.count); \ + } while(0) + +static void mtl_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 6; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VEC(atom->cmd[MTL_CMD_0], (atom->cmd+1)); + OUT_SCL2(atom->cmd[MTL_CMD_1], (atom->cmd + 18)); + END_BATCH(); +} + +static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 8; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1); + OUT_VEC(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1); + END_BATCH(); +} + +static void ptp_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 8; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VEC(atom->cmd[PTP_CMD_0], atom->cmd+1); + OUT_VEC(atom->cmd[PTP_CMD_1], atom->cmd+PTP_CMD_1+1); + END_BATCH(); +} + +static void veclinear_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 4; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VECLINEAR(atom->cmd[0], atom->cmd+1); + END_BATCH(); +} + +static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 2; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_SCL(atom->cmd[0], atom->cmd+1); + END_BATCH(); +} + + +static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + + dwords += 4; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VEC(atom->cmd[0], atom->cmd+1); + END_BATCH(); +} + +static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + struct radeon_renderbuffer *rrb; + uint32_t cbpitch; + uint32_t zbpitch, depth_fmt; + uint32_t dwords = atom->cmd_size; + + /* output the first 7 bytes of context */ + BEGIN_BATCH_NO_AUTOSTATE(dwords+2+2); + OUT_BATCH_TABLE(atom->cmd, 5); + + rrb = radeon_get_depthbuffer(&r200->radeon); + if (!rrb) { + OUT_BATCH(0); + OUT_BATCH(0); + } else { + zbpitch = (rrb->pitch / rrb->cpp); + if (r200->using_hyperz) + zbpitch |= RADEON_DEPTH_HYPERZ; + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_BATCH(zbpitch); + if (rrb->cpp == 4) + depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; + else + depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; + atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK; + atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt; + } + + OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]); + OUT_BATCH(atom->cmd[CTX_CMD_1]); + OUT_BATCH(atom->cmd[CTX_PP_CNTL]); + + rrb = radeon_get_colorbuffer(&r200->radeon); + if (!rrb || !rrb->bo) { + OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); + OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]); + } else { + atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); + if (rrb->cpp == 4) + atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; + else + atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; + + OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + } + + OUT_BATCH(atom->cmd[CTX_CMD_2]); + + if (!rrb || !rrb->bo) { + OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]); + } else { + cbpitch = (rrb->pitch / rrb->cpp); + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + cbpitch |= R200_COLOR_TILE_ENABLE; + OUT_BATCH(cbpitch); + } + + if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) + OUT_BATCH_TABLE((atom->cmd + 14), 4); + + END_BATCH(); +} + +static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + struct radeon_renderbuffer *rrb, *drb; + uint32_t cbpitch = 0; + uint32_t zbpitch = 0; + uint32_t dwords = atom->cmd_size; + uint32_t depth_fmt; + + rrb = radeon_get_colorbuffer(&r200->radeon); + if (!rrb || !rrb->bo) { + return; + } + + atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); + if (rrb->cpp == 4) + atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; + else + atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; + + cbpitch = (rrb->pitch / rrb->cpp); + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + cbpitch |= R200_COLOR_TILE_ENABLE; + + drb = radeon_get_depthbuffer(&r200->radeon); + if (drb) { + zbpitch = (drb->pitch / drb->cpp); + if (drb->cpp == 4) + depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; + else + depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; + atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK; + atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt; + } + + dwords = 10; + if (drb) + dwords += 6; + if (rrb) + dwords += 6; + + /* output the first 7 bytes of context */ + BEGIN_BATCH_NO_AUTOSTATE(dwords); + + /* In the CS case we need to split this up */ + OUT_BATCH(CP_PACKET0(packet[0].start, 3)); + OUT_BATCH_TABLE((atom->cmd + 1), 4); + + if (drb) { + OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0)); + OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + + OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0)); + OUT_BATCH(zbpitch); + } + + OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0)); + OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]); + OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1)); + OUT_BATCH(atom->cmd[CTX_PP_CNTL]); + OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); + + + if (rrb) { + OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0)); + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + + OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0)); + OUT_BATCH(cbpitch); + } + + if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) { + OUT_BATCH_TABLE((atom->cmd + 14), 4); + } + + END_BATCH(); +} + +static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + int i = atom->idx; + radeonTexObj *t = r200->state.texture.unit[i].texobj; + radeon_mipmap_level *lvl; + + if (t && t->mt && !t->image_override) + dwords += 2; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_BATCH_TABLE(atom->cmd, 10); + + if (t && t->mt && !t->image_override) { + if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) { + lvl = &t->mt->levels[0]; + OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + } else { + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + } + } else if (!t) { + /* workaround for old CS mechanism */ + OUT_BATCH(r200->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]); + } else { + OUT_BATCH(t->override_offset); + } + + END_BATCH(); +} + +static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + int i = atom->idx; + radeonTexObj *t = r200->state.texture.unit[i].texobj; + radeon_mipmap_level *lvl; + int hastexture = 1; + + if (!t) + hastexture = 0; + else { + if (!t->mt && !t->bo) + hastexture = 0; + } + + dwords += 2; + if (hastexture) + dwords += 2; + else + dwords -= 2; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + + OUT_BATCH(CP_PACKET0(R200_PP_TXFILTER_0 + (24 * i), 7)); + OUT_BATCH_TABLE((atom->cmd + 1), 8); + + if (hastexture) { + OUT_BATCH(CP_PACKET0(R200_PP_TXOFFSET_0 + (24 * i), 0)); + if (t->mt && !t->image_override) { + if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) { + lvl = &t->mt->levels[0]; + OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + } else { + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + } + } else { + if (t->bo) + OUT_BATCH_RELOC(t->tile_bits, t->bo, 0, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + } + } + END_BATCH(); +} + + +static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r200ContextPtr r200 = R200_CONTEXT(ctx); + BATCH_LOCALS(&r200->radeon); + uint32_t dwords = atom->cmd_size; + int i = atom->idx; + radeonTexObj *t = r200->state.texture.unit[i].texobj; + GLuint size; + + BEGIN_BATCH_NO_AUTOSTATE(dwords + (2 * 5)); + OUT_BATCH_TABLE(atom->cmd, 3); + + if (t && !t->image_override) { + size = t->mt->totalsize / 6; + OUT_BATCH_RELOC(0, t->mt->bo, size, RADEON_GEM_DOMAIN_VRAM, 0, 0); + OUT_BATCH_RELOC(0, t->mt->bo, size * 2, RADEON_GEM_DOMAIN_VRAM, 0, 0); + OUT_BATCH_RELOC(0, t->mt->bo, size * 3, RADEON_GEM_DOMAIN_VRAM, 0, 0); + OUT_BATCH_RELOC(0, t->mt->bo, size * 4, RADEON_GEM_DOMAIN_VRAM, 0, 0); + OUT_BATCH_RELOC(0, t->mt->bo, size * 5, RADEON_GEM_DOMAIN_VRAM, 0, 0); + } + END_BATCH(); +} /* Initialize the context's hardware state. */ void r200InitState( r200ContextPtr rmesa ) { - GLcontext *ctx = rmesa->glCtx; - GLuint color_fmt, depth_fmt, i; - GLint drawPitch, drawOffset; + GLcontext *ctx = rmesa->radeon.glCtx; + GLuint i; - switch ( rmesa->r200Screen->cpp ) { - case 2: - color_fmt = R200_COLOR_FORMAT_RGB565; - break; - case 4: - color_fmt = R200_COLOR_FORMAT_ARGB8888; - break; - default: - fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" ); - exit( -1 ); - } - - rmesa->state.color.clear = 0x00000000; + rmesa->radeon.state.color.clear = 0x00000000; switch ( ctx->Visual.depthBits ) { case 16: - rmesa->state.depth.clear = 0x0000ffff; - rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff; - depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z; - rmesa->state.stencil.clear = 0x00000000; + rmesa->radeon.state.depth.clear = 0x0000ffff; + rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffff; + rmesa->radeon.state.stencil.clear = 0x00000000; break; case 24: - rmesa->state.depth.clear = 0x00ffffff; - rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff; - depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z; - rmesa->state.stencil.clear = 0xffff0000; - break; default: - fprintf( stderr, "Error: Unsupported depth %d... exiting\n", - ctx->Visual.depthBits ); - exit( -1 ); + rmesa->radeon.state.depth.clear = 0x00ffffff; + rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffffff; + rmesa->radeon.state.stencil.clear = 0xffff0000; + break; } /* Only have hw stencil when depth buffer is 24 bits deep */ - rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 && + rmesa->radeon.state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 && ctx->Visual.depthBits == 24 ); - rmesa->Fallback = 0; - - if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) { - drawOffset = rmesa->r200Screen->backOffset; - drawPitch = rmesa->r200Screen->backPitch; - } else { - drawOffset = rmesa->r200Screen->frontOffset; - drawPitch = rmesa->r200Screen->frontPitch; - } -#if 000 - if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) { - rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset; - rmesa->state.color.drawPitch = rmesa->r200Screen->backPitch; - } else { - rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset; - rmesa->state.color.drawPitch = rmesa->r200Screen->frontPitch; - } - - rmesa->state.pixel.readOffset = rmesa->state.color.drawOffset; - rmesa->state.pixel.readPitch = rmesa->state.color.drawPitch; -#endif + rmesa->radeon.Fallback = 0; - rmesa->hw.max_state_size = 0; + rmesa->radeon.hw.max_state_size = 0; #define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX ) \ do { \ rmesa->hw.ATOM.cmd_size = SZ; \ - rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int)); \ - rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int)); \ + rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int)); \ + rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int)); \ rmesa->hw.ATOM.name = NM; \ rmesa->hw.ATOM.idx = IDX; \ rmesa->hw.ATOM.check = check_##CHK; \ rmesa->hw.ATOM.dirty = GL_FALSE; \ - rmesa->hw.max_state_size += SZ * sizeof(int); \ + rmesa->radeon.hw.max_state_size += SZ * sizeof(int); \ } while (0) /* Allocate state buffers: */ - if (rmesa->r200Screen->drmSupportsBlendColor) + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) ALLOC_STATE( ctx, always, CTX_STATE_SIZE_NEWDRM, "CTX/context", 0 ); else ALLOC_STATE( ctx, always, CTX_STATE_SIZE_OLDDRM, "CTX/context", 0 ); + + if (rmesa->radeon.radeonScreen->kernel_mm) + rmesa->hw.ctx.emit = ctx_emit_cs; + else + rmesa->hw.ctx.emit = ctx_emit; ALLOC_STATE( set, always, SET_STATE_SIZE, "SET/setup", 0 ); ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 ); ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 ); @@ -282,8 +728,8 @@ void r200InitState( r200ContextPtr rmesa ) ALLOC_STATE( cst, always, CST_STATE_SIZE, "CST/constant", 0 ); ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 ); ALLOC_STATE( tf, tf, TF_STATE_SIZE, "TF/tfactor", 0 ); - if (rmesa->r200Screen->drmSupportsFragShader) { - if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) { + if (rmesa->radeon.radeonScreen->drmSupportsFragShader) { + if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) { /* make sure texture units 0/1 are emitted pair-wise for r200 t0 hang workaround */ ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-0", 0 ); ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-1", 1 ); @@ -303,7 +749,7 @@ void r200InitState( r200ContextPtr rmesa ) ALLOC_STATE( afs[1], afs, AFS_STATE_SIZE, "AFS/afsinst-1", 1 ); } else { - if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) { + if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) { ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-0", 0 ); ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-1", 1 ); ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 ); @@ -321,13 +767,21 @@ void r200InitState( r200ContextPtr rmesa ) ALLOC_STATE( afs[0], never, AFS_STATE_SIZE, "AFS/afsinst-0", 0 ); ALLOC_STATE( afs[1], never, AFS_STATE_SIZE, "AFS/afsinst-1", 1 ); } - if (rmesa->r200Screen->drmSupportsCubeMapsR200) { + + for (i = 0; i < 5; i++) + if (rmesa->radeon.radeonScreen->kernel_mm) + rmesa->hw.tex[i].emit = tex_emit_cs; + else + rmesa->hw.tex[i].emit = tex_emit; + if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200) { ALLOC_STATE( cube[0], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-0", 0 ); ALLOC_STATE( cube[1], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-1", 1 ); ALLOC_STATE( cube[2], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-2", 2 ); ALLOC_STATE( cube[3], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-3", 3 ); ALLOC_STATE( cube[4], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-4", 4 ); ALLOC_STATE( cube[5], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-5", 5 ); + for (i = 0; i < 5; i++) + rmesa->hw.cube[i].emit = cube_emit; } else { ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/tex-0", 0 ); @@ -337,7 +791,8 @@ void r200InitState( r200ContextPtr rmesa ) ALLOC_STATE( cube[4], never, CUBE_STATE_SIZE, "CUBE/tex-4", 4 ); ALLOC_STATE( cube[5], never, CUBE_STATE_SIZE, "CUBE/tex-5", 5 ); } - if (rmesa->r200Screen->drmSupportsVertexProgram) { + + if (rmesa->radeon.radeonScreen->drmSupportsVertexProgram) { ALLOC_STATE( pvs, tcl_vp, PVS_STATE_SIZE, "PVS/pvscntl", 0 ); ALLOC_STATE( vpi[0], tcl_vp, VPI_STATE_SIZE, "VP/vertexprog-0", 0 ); ALLOC_STATE( vpi[1], tcl_vp_size, VPI_STATE_SIZE, "VP/vertexprog-1", 1 ); @@ -390,13 +845,13 @@ void r200InitState( r200ContextPtr rmesa ) ALLOC_STATE( pix[3], texenv, PIX_STATE_SIZE, "PIX/pixstage-3", 3 ); ALLOC_STATE( pix[4], texenv, PIX_STATE_SIZE, "PIX/pixstage-4", 4 ); ALLOC_STATE( pix[5], texenv, PIX_STATE_SIZE, "PIX/pixstage-5", 5 ); - if (rmesa->r200Screen->drmSupportsTriPerf) { + if (rmesa->radeon.radeonScreen->drmSupportsTriPerf) { ALLOC_STATE( prf, always, PRF_STATE_SIZE, "PRF/performance-tri", 0 ); } else { ALLOC_STATE( prf, never, PRF_STATE_SIZE, "PRF/performance-tri", 0 ); } - if (rmesa->r200Screen->drmSupportsPointSprites) { + if (rmesa->radeon.radeonScreen->drmSupportsPointSprites) { ALLOC_STATE( spr, always, SPR_STATE_SIZE, "SPR/pointsprite", 0 ); ALLOC_STATE( ptp, tcl, PTP_STATE_SIZE, "PTP/pointparams", 0 ); } @@ -409,87 +864,115 @@ void r200InitState( r200ContextPtr rmesa ) /* Fill in the packet headers: */ - rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC); - rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL); - rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH); - if (rmesa->r200Screen->drmSupportsBlendColor) - rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(R200_EMIT_RB3D_BLENDCOLOR); - rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN); - rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH); - rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK); - rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE); - rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL); - rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC); - rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(R200_EMIT_PP_CNTL_X); - rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(R200_EMIT_RB3D_DEPTHXY_OFFSET); - rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(R200_EMIT_RE_AUX_SCISSOR_CNTL); - rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(R200_EMIT_RE_SCISSOR_TL_0); - rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(R200_EMIT_SE_VAP_CNTL_STATUS); - rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(R200_EMIT_RE_POINTSIZE); - rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0); - rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(R200_EMIT_PP_TAM_DEBUG3); - rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(R200_EMIT_TFACTOR_0); - if (rmesa->r200Screen->drmSupportsFragShader) { - rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(R200_EMIT_ATF_TFACTOR); - rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_0); - rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0); - rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_1); - rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1); - rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_2); - rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2); - rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_3); - rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3); - rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_4); - rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4); - rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_5); - rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5); + rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC); + rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL); + rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH); + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) + rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(rmesa, R200_EMIT_RB3D_BLENDCOLOR); + rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN); + rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH); + rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK); + rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE); + rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL); + rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC); + rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CNTL_X); + rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(rmesa, R200_EMIT_RB3D_DEPTHXY_OFFSET); + rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(rmesa, R200_EMIT_RE_AUX_SCISSOR_CNTL); + rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(rmesa, R200_EMIT_RE_SCISSOR_TL_0); + rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(rmesa, R200_EMIT_SE_VAP_CNTL_STATUS); + rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(rmesa, R200_EMIT_RE_POINTSIZE); + rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(rmesa, R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0); + rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TAM_DEBUG3); + rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(rmesa, R200_EMIT_TFACTOR_0); + if (rmesa->radeon.radeonScreen->drmSupportsFragShader) { + rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(rmesa, R200_EMIT_ATF_TFACTOR); + rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_0); + rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0); + rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_1); + rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1); + rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_2); + rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2); + rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_3); + rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3); + rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_4); + rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4); + rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_5); + rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5); } else { - rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_0); - rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0); - rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_1); - rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1); - rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_2); - rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2); - rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_3); - rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3); - rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_4); - rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4); - rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_5); - rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5); - } - rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_0); - rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_1); - rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(R200_EMIT_VAP_PVS_CNTL); - rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_0); - rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_0); - rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_1); - rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_1); - rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_2); - rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_2); - rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_3); - rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_3); - rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_4); - rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_4); - rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_5); - rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_5); - rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_0); - rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_1); - rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_2); - rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_3); - rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_4); - rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_5); - rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR); - rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(R200_EMIT_TCL_LIGHT_MODEL_CTL_0); - rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(R200_EMIT_TCL_UCP_VERT_BLEND_CTL); - rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(R200_EMIT_TEX_PROC_CTL_2); - rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(R200_EMIT_MATRIX_SELECT_0); - rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(R200_EMIT_VAP_CTL); - rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(R200_EMIT_VTX_FMT_0); - rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(R200_EMIT_OUTPUT_VTX_COMP_SEL); - rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(R200_EMIT_SE_VTX_STATE_CNTL); - rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(R200_EMIT_VTE_CNTL); - rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(R200_EMIT_PP_TRI_PERF_CNTL); - rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(R200_EMIT_TCL_POINT_SPRITE_CNTL); + rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_0); + rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0); + rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_1); + rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1); + rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_2); + rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2); + rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_3); + rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3); + rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_4); + rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4); + rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_5); + rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5); + } + rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_0); + rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_1); + rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_PVS_CNTL); + rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_0); + rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_0); + rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_1); + rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_1); + rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_2); + rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_2); + rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_3); + rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_3); + rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_4); + rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_4); + rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_5); + rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_5); + rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_0); + rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_1); + rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_2); + rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_3); + rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_4); + rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_5); + rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR); + rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_LIGHT_MODEL_CTL_0); + rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(rmesa, R200_EMIT_TCL_UCP_VERT_BLEND_CTL); + rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(rmesa, R200_EMIT_TEX_PROC_CTL_2); + rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(rmesa, R200_EMIT_MATRIX_SELECT_0); + rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_CTL); + rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTX_FMT_0); + rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(rmesa, R200_EMIT_OUTPUT_VTX_COMP_SEL); + rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(rmesa, R200_EMIT_SE_VTX_STATE_CNTL); + rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTE_CNTL); + rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TRI_PERF_CNTL); + rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_POINT_SPRITE_CNTL); + if (rmesa->radeon.radeonScreen->kernel_mm) { + rmesa->hw.mtl[0].emit = mtl_emit; + rmesa->hw.mtl[1].emit = mtl_emit; + + rmesa->hw.vpi[0].emit = veclinear_emit; + rmesa->hw.vpi[1].emit = veclinear_emit; + rmesa->hw.vpp[0].emit = veclinear_emit; + rmesa->hw.vpp[1].emit = veclinear_emit; + + rmesa->hw.grd.emit = scl_emit; + rmesa->hw.fog.emit = vec_emit; + rmesa->hw.glt.emit = vec_emit; + rmesa->hw.eye.emit = vec_emit; + + for (i = R200_MTX_MV; i <= R200_MTX_TEX5; i++) + rmesa->hw.mat[i].emit = vec_emit; + + for (i = 0; i < 8; i++) + rmesa->hw.lit[i].emit = lit_emit; + + for (i = 0; i < 6; i++) + rmesa->hw.ucp[i].emit = vec_emit; + + rmesa->hw.ptp.emit = ptp_emit; + } + + + rmesa->hw.mtl[0].cmd[MTL_CMD_0] = cmdvec( R200_VS_MAT_0_EMISS, 1, 16 ); rmesa->hw.mtl[0].cmd[MTL_CMD_1] = @@ -567,7 +1050,7 @@ void r200InitState( r200ContextPtr rmesa ) (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) | (R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT)); - if (rmesa->r200Screen->drmSupportsBlendColor) { + if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) { rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = 0x00000000; rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = (R200_COMB_FCN_ADD_CLAMP | (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) | @@ -578,18 +1061,17 @@ void r200InitState( r200ContextPtr rmesa ) } rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] = - rmesa->r200Screen->depthOffset + rmesa->r200Screen->fbLocation; + rmesa->radeon.radeonScreen->depthOffset + rmesa->radeon.radeonScreen->fbLocation; rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = - ((rmesa->r200Screen->depthPitch & + ((rmesa->radeon.radeonScreen->depthPitch & R200_DEPTHPITCH_MASK) | R200_DEPTH_ENDIAN_NO_SWAP); if (rmesa->using_hyperz) rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] |= R200_DEPTH_HYPERZ; - rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt | - R200_Z_TEST_LESS | + rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (R200_Z_TEST_LESS | R200_STENCIL_TEST_ALWAYS | R200_STENCIL_FAIL_KEEP | R200_STENCIL_ZPASS_KEEP | @@ -599,15 +1081,14 @@ void r200InitState( r200ContextPtr rmesa ) if (rmesa->using_hyperz) { rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= R200_Z_COMPRESSION_ENABLE | R200_Z_DECOMPRESSION_ENABLE; -/* if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) +/* if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/ } rmesa->hw.ctx.cmd[CTX_PP_CNTL] = (R200_ANTI_ALIAS_NONE | R200_TEX_BLEND_0_ENABLE); - rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = color_fmt; - switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) { + switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) { case DRI_CONF_DITHER_XERRORDIFFRESET: rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_INIT; break; @@ -615,41 +1096,19 @@ void r200InitState( r200ContextPtr rmesa ) rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_SCALE_DITHER_ENABLE; break; } - if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) == + if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) == DRI_CONF_ROUND_ROUND ) - rmesa->state.color.roundEnable = R200_ROUND_ENABLE; + rmesa->radeon.state.color.roundEnable = R200_ROUND_ENABLE; else - rmesa->state.color.roundEnable = 0; - if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) == + rmesa->radeon.state.color.roundEnable = 0; + if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) == DRI_CONF_COLOR_REDUCTION_DITHER ) rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_ENABLE; else - rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable; - -#if 000 - rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((rmesa->state.color.drawOffset + - rmesa->r200Screen->fbLocation) - & R200_COLOROFFSET_MASK); - - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((rmesa->state.color.drawPitch & - R200_COLORPITCH_MASK) | - R200_COLOR_ENDIAN_NO_SWAP); -#else - rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset + - rmesa->r200Screen->fbLocation) - & R200_COLOROFFSET_MASK); - - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch & - R200_COLORPITCH_MASK) | - R200_COLOR_ENDIAN_NO_SWAP); -#endif - /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */ - if (rmesa->sarea->tiling_enabled) { - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE; - } + rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable; rmesa->hw.prf.cmd[PRF_PP_TRI_PERF] = R200_TRI_CUTOFF_MASK - R200_TRI_CUTOFF_MASK * - driQueryOptionf (&rmesa->optionCache,"texture_blend_quality"); + driQueryOptionf (&rmesa->radeon.optionCache,"texture_blend_quality"); rmesa->hw.prf.cmd[PRF_PP_PERF_CNTL] = 0; rmesa->hw.set.cmd[SET_SE_CNTL] = (R200_FFACE_CULL_CCW | @@ -704,7 +1163,7 @@ void r200InitState( r200ContextPtr rmesa ) R200_VC_NO_SWAP; #endif - if (!(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) { + if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) { /* Bypass TCL */ rmesa->hw.cst.cmd[CST_SE_VAP_CNTL_STATUS] |= (1<<8); } @@ -743,28 +1202,28 @@ void r200InitState( r200ContextPtr rmesa ) rmesa->hw.tex[i].cmd[TEX_PP_TXFORMAT_X] = (/* R200_TEXCOORD_PROJ | */ 0x100000); /* Small default bias */ - if (rmesa->r200Screen->drmSupportsFragShader) { + if (rmesa->radeon.radeonScreen->drmSupportsFragShader) { rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_NEWDRM] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.tex[i].cmd[TEX_PP_CUBIC_FACES] = 0; rmesa->hw.tex[i].cmd[TEX_PP_TXMULTI_CTL] = 0; } else { rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_OLDDRM] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; } rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F1] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F2] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F3] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F4] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F5] = - rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.pix[i].cmd[PIX_PP_TXCBLEND] = (R200_TXC_ARG_A_ZERO | @@ -967,5 +1426,7 @@ void r200InitState( r200ContextPtr rmesa ) r200LightingSpaceChange( ctx ); - rmesa->hw.all_dirty = GL_TRUE; + rmesa->radeon.hw.all_dirty = GL_TRUE; + + rcommonInitCmdBuf(&rmesa->radeon); } diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index b25f028244..b006409987 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -55,27 +55,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_tcl.h" -static void flush_last_swtcl_prim( r200ContextPtr rmesa ); - - /*********************************************************************** * Initialization ***********************************************************************/ #define EMIT_ATTR( ATTR, STYLE, F0 ) \ do { \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR); \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE); \ - rmesa->swtcl.vertex_attr_count++; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \ + rmesa->radeon.swtcl.vertex_attr_count++; \ fmt_0 |= F0; \ } while (0) #define EMIT_PAD( N ) \ do { \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0; \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD; \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N); \ - rmesa->swtcl.vertex_attr_count++; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N); \ + rmesa->radeon.swtcl.vertex_attr_count++; \ } while (0) static void r200SetVertexFormat( GLcontext *ctx ) @@ -100,7 +97,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) } assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL ); - rmesa->swtcl.vertex_attr_count = 0; + rmesa->radeon.swtcl.vertex_attr_count = 0; /* EMIT_ATTR's must be in order as they tell t_vertex.c how to * build up a hardware vertex. @@ -185,7 +182,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_SPEC_ALPHA; } - if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) || + if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) || (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0) || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) { R200_NEWPRIM(rmesa); @@ -193,26 +190,20 @@ static void r200SetVertexFormat( GLcontext *ctx ) rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0; rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1; - rmesa->swtcl.vertex_size = + rmesa->radeon.swtcl.vertex_size = _tnl_install_attrs( ctx, - rmesa->swtcl.vertex_attrs, - rmesa->swtcl.vertex_attr_count, + rmesa->radeon.swtcl.vertex_attrs, + rmesa->radeon.swtcl.vertex_attr_count, NULL, 0 ); - rmesa->swtcl.vertex_size /= 4; - RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset ); + rmesa->radeon.swtcl.vertex_size /= 4; + RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset ); } } static void r200RenderStart( GLcontext *ctx ) { - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - r200SetVertexFormat( ctx ); - - if (rmesa->dma.flush != 0 && - rmesa->dma.flush != flush_last_swtcl_prim) - rmesa->dma.flush( rmesa ); } @@ -232,7 +223,7 @@ void r200ChooseVertexState( GLcontext *ctx ) * rasterization fallback. As this function will be called again when we * leave a rasterization fallback, we can just skip it for now. */ - if (rmesa->Fallback != 0) + if (rmesa->radeon.Fallback != 0) return; vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL]; @@ -273,78 +264,27 @@ void r200ChooseVertexState( GLcontext *ctx ) } } - -/* Flush vertices in the current dma region. - */ -static void flush_last_swtcl_prim( r200ContextPtr rmesa ) -{ - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->dma.flush = NULL; - - if (rmesa->dma.current.buf) { - struct r200_dma_region *current = &rmesa->dma.current; - GLuint current_offset = (rmesa->r200Screen->gart_buffer_offset + - current->buf->buf->idx * RADEON_BUFFER_SIZE + - current->start); - - assert (!(rmesa->swtcl.hw_primitive & R200_VF_PRIM_WALK_IND)); - - assert (current->start + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - current->ptr); - - if (rmesa->dma.current.start != rmesa->dma.current.ptr) { - r200EnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ + - rmesa->hw.max_state_size + VBUF_BUFSZ ); - r200EmitVertexAOS( rmesa, - rmesa->swtcl.vertex_size, - current_offset); - - r200EmitVbufPrim( rmesa, - rmesa->swtcl.hw_primitive, - rmesa->swtcl.numverts); - } - - rmesa->swtcl.numverts = 0; - current->start = current->ptr; - } -} - - -/* Alloc space in the current dma region. - */ -static INLINE void * -r200AllocDmaLowVerts( r200ContextPtr rmesa, int nverts, int vsize ) +void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { - GLuint bytes = vsize * nverts; - - if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) - r200RefillCurrentDmaRegion( rmesa ); - - if (!rmesa->dma.flush) { - rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; - rmesa->dma.flush = flush_last_swtcl_prim; - } + r200ContextPtr rmesa = R200_CONTEXT(ctx); + rcommonEnsureCmdBufSpace(&rmesa->radeon, + rmesa->radeon.hw.max_state_size + (12*sizeof(int)), + __FUNCTION__); - ASSERT( vsize == rmesa->swtcl.vertex_size * 4 ); - ASSERT( rmesa->dma.flush == flush_last_swtcl_prim ); - ASSERT( rmesa->dma.current.start + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - rmesa->dma.current.ptr ); + radeonEmitState(&rmesa->radeon); + r200EmitVertexAOS( rmesa, + rmesa->radeon.swtcl.vertex_size, + rmesa->radeon.dma.current, + current_offset); - { - GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr); - rmesa->dma.current.ptr += bytes; - rmesa->swtcl.numverts += nverts; - return head; - } + + r200EmitVbufPrim( rmesa, + rmesa->radeon.swtcl.hw_primitive, + rmesa->radeon.swtcl.numverts); } - /**************************************************************************/ @@ -392,13 +332,13 @@ static void r200ResetLineStipple( GLcontext *ctx ); #undef LOCAL_VARS #undef ALLOC_VERTS #define CTX_ARG r200ContextPtr rmesa -#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size -#define ALLOC_VERTS( n, size ) r200AllocDmaLowVerts( rmesa, n, size * 4 ) +#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size +#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 ) #define LOCAL_VARS \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - const char *r200verts = (char *)rmesa->swtcl.verts; -#define VERT(x) (r200Vertex *)(r200verts + ((x) * vertsize * sizeof(int))) -#define VERTEX r200Vertex + const char *r200verts = (char *)rmesa->radeon.swtcl.verts; +#define VERT(x) (radeonVertex *)(r200verts + ((x) * vertsize * sizeof(int))) +#define VERTEX radeonVertex #define DO_DEBUG_VERTS (1 && (R200_DEBUG & DEBUG_VERTS)) #undef TAG @@ -456,11 +396,11 @@ static struct { #define VERT_Y(_v) _v->v.y #define VERT_Z(_v) _v->v.z #define AREA_IS_CCW( a ) (a < 0) -#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int))) +#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int))) #define VERT_SET_RGBA( v, c ) \ do { \ - r200_color_t *color = (r200_color_t *)&((v)->ui[coloroffset]); \ + radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]); \ UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \ UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \ UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \ @@ -472,7 +412,7 @@ do { \ #define VERT_SET_SPEC( v, c ) \ do { \ if (specoffset) { \ - r200_color_t *spec = (r200_color_t *)&((v)->ui[specoffset]); \ + radeon_color_t *spec = (radeon_color_t *)&((v)->ui[specoffset]); \ UNCLAMPED_FLOAT_TO_UBYTE(spec->red, (c)[0]); \ UNCLAMPED_FLOAT_TO_UBYTE(spec->green, (c)[1]); \ UNCLAMPED_FLOAT_TO_UBYTE(spec->blue, (c)[2]); \ @@ -481,8 +421,8 @@ do { \ #define VERT_COPY_SPEC( v0, v1 ) \ do { \ if (specoffset) { \ - r200_color_t *spec0 = (r200_color_t *)&((v0)->ui[specoffset]); \ - r200_color_t *spec1 = (r200_color_t *)&((v1)->ui[specoffset]); \ + radeon_color_t *spec0 = (radeon_color_t *)&((v0)->ui[specoffset]); \ + radeon_color_t *spec1 = (radeon_color_t *)&((v1)->ui[specoffset]); \ spec0->red = spec1->red; \ spec0->green = spec1->green; \ spec0->blue = spec1->blue; \ @@ -513,7 +453,7 @@ do { \ ***********************************************************************/ #define RASTERIZE(x) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, x) ) -#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive +#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive #undef TAG #define TAG(x) x #include "tnl_dd/t_dd_unfilled.h" @@ -569,8 +509,8 @@ static void init_rast_tab( void ) #undef LOCAL_VARS #define LOCAL_VARS \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - const GLuint vertsize = rmesa->swtcl.vertex_size; \ - const char *r200verts = (char *)rmesa->swtcl.verts; \ + const GLuint vertsize = rmesa->radeon.swtcl.vertex_size; \ + const char *r200verts = (char *)rmesa->radeon.swtcl.verts; \ const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ const GLboolean stipple = ctx->Line.StippleFlag; \ (void) elt; (void) stipple; @@ -599,13 +539,13 @@ void r200ChooseRenderState( GLcontext *ctx ) GLuint index = 0; GLuint flags = ctx->_TriangleCaps; - if (!rmesa->TclFallback || rmesa->Fallback) + if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback) return; if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R200_TWOSIDE_BIT; if (flags & DD_TRI_UNFILLED) index |= R200_UNFILLED_BIT; - if (index != rmesa->swtcl.RenderIndex) { + if (index != rmesa->radeon.swtcl.RenderIndex) { tnl->Driver.Render.Points = rast_tab[index].points; tnl->Driver.Render.Line = rast_tab[index].line; tnl->Driver.Render.ClippedLine = rast_tab[index].line; @@ -622,7 +562,7 @@ void r200ChooseRenderState( GLcontext *ctx ) tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon; } - rmesa->swtcl.RenderIndex = index; + rmesa->radeon.swtcl.RenderIndex = index; } } @@ -636,7 +576,7 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - if (rmesa->swtcl.hw_primitive != hwprim) { + if (rmesa->radeon.swtcl.hw_primitive != hwprim) { /* need to disable perspective-correct texturing for point sprites */ if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) { if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) { @@ -649,14 +589,14 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim ) rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE; } R200_NEWPRIM( rmesa ); - rmesa->swtcl.hw_primitive = hwprim; + rmesa->radeon.swtcl.hw_primitive = hwprim; } } static void r200RenderPrimitive( GLcontext *ctx, GLenum prim ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - rmesa->swtcl.render_primitive = prim; + rmesa->radeon.swtcl.render_primitive = prim; if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, prim) ); } @@ -701,15 +641,15 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = rmesa->Fallback; + GLuint oldfallback = rmesa->radeon.Fallback; if (mode) { - rmesa->Fallback |= bit; + rmesa->radeon.Fallback |= bit; if (oldfallback == 0) { - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE ); _swsetup_Wakeup( ctx ); - rmesa->swtcl.RenderIndex = ~0; + rmesa->radeon.swtcl.RenderIndex = ~0; if (R200_DEBUG & DEBUG_FALLBACKS) { fprintf(stderr, "R200 begin rasterization fallback: 0x%x %s\n", bit, getFallbackString(bit)); @@ -717,7 +657,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ) } } else { - rmesa->Fallback &= ~bit; + rmesa->radeon.Fallback &= ~bit; if (oldfallback == bit) { _swrast_flush( ctx ); @@ -731,14 +671,14 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ) tnl->Driver.Render.ResetLineStipple = r200ResetLineStipple; TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_FALSE ); - if (rmesa->TclFallback) { - /* These are already done if rmesa->TclFallback goes to + if (rmesa->radeon.TclFallback) { + /* These are already done if rmesa->radeon.TclFallback goes to * zero above. But not if it doesn't (R200_NO_TCL for * example?) */ _tnl_invalidate_vertex_state( ctx, ~0 ); _tnl_invalidate_vertices( ctx, ~0 ); - RENDERINPUTS_ZERO( rmesa->tnl_index_bitset ); + RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset ); r200ChooseVertexState( ctx ); r200ChooseRenderState( ctx ); } @@ -772,7 +712,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py, r200ContextPtr rmesa = R200_CONTEXT(ctx); const GLfloat *rc = ctx->Current.RasterColor; GLint row, col; - r200Vertex vert; + radeonVertex vert; GLuint orig_vte; GLuint h; @@ -794,7 +734,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py, vte |= R200_VTX_W0_FMT; vap &= ~R200_VAP_FORCE_W_TO_ONE; - rmesa->swtcl.vertex_size = 5; + rmesa->radeon.swtcl.vertex_size = 5; if ( (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0) || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) { @@ -871,10 +811,10 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py, /* Update window height */ - LOCK_HARDWARE( rmesa ); - UNLOCK_HARDWARE( rmesa ); - h = rmesa->dri.drawable->h + rmesa->dri.drawable->y; - px += rmesa->dri.drawable->x; + LOCK_HARDWARE( &rmesa->radeon ); + UNLOCK_HARDWARE( &rmesa->radeon ); + h = rmesa->radeon.dri.drawable->h + rmesa->radeon.dri.drawable->y; + px += rmesa->radeon.dri.drawable->x; /* Clipping handled by existing mechansims in r200_ioctl.c? */ @@ -929,7 +869,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py, /* Need to restore vertexformat? */ - if (rmesa->TclFallback) + if (rmesa->radeon.TclFallback) r200ChooseVertexState( ctx ); } @@ -962,17 +902,13 @@ void r200InitSwtcl( GLcontext *ctx ) _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 36 * sizeof(GLfloat) ); - rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf; - rmesa->swtcl.RenderIndex = ~0; - rmesa->swtcl.render_primitive = GL_TRIANGLES; - rmesa->swtcl.hw_primitive = 0; + rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf; + rmesa->radeon.swtcl.RenderIndex = ~0; + rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES; + rmesa->radeon.swtcl.hw_primitive = 0; } void r200DestroySwtcl( GLcontext *ctx ) { - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (rmesa->swtcl.indexed_verts.buf) - r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, __FUNCTION__ ); } diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h index 8c29fd0c99..a4051a4b76 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.h +++ b/src/mesa/drivers/dri/r200/r200_swtcl.h @@ -52,15 +52,11 @@ extern void r200BuildVertices( GLcontext *ctx, GLuint start, GLuint count, extern void r200PrintSetupFlags(char *msg, GLuint flags ); -extern void r200_emit_indexed_verts( GLcontext *ctx, - GLuint start, - GLuint count ); - extern void r200_translate_vertex( GLcontext *ctx, - const r200Vertex *src, + const radeonVertex *src, SWvertex *dst ); -extern void r200_print_vertex( GLcontext *ctx, const r200Vertex *v ); +extern void r200_print_vertex( GLcontext *ctx, const radeonVertex *v ); extern void r200_import_float_colors( GLcontext *ctx ); extern void r200_import_float_spec_colors( GLcontext *ctx ); @@ -70,5 +66,5 @@ extern void r200PointsBitmap( GLcontext *ctx, GLint px, GLint py, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ); - +void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset); #endif diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index 99aecfe1e9..8e0fb14e70 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -123,7 +123,7 @@ static GLboolean discrete_prim[0x10] = { #define RESET_STIPPLE() do { \ R200_STATECHANGE( rmesa, lin ); \ - r200EmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) #define AUTO_STIPPLE( mode ) do { \ @@ -134,7 +134,7 @@ static GLboolean discrete_prim[0x10] = { else \ rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &= \ ~R200_LINE_PATTERN_AUTO_RESET; \ - r200EmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) @@ -142,25 +142,23 @@ static GLboolean discrete_prim[0x10] = { static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) { - if (rmesa->dma.flush == r200FlushElts && - rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) { + if (rmesa->radeon.dma.flush == r200FlushElts && + rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) { - GLushort *dest = (GLushort *)(rmesa->store.cmd_buf + - rmesa->store.cmd_used); + GLushort *dest = (GLushort *)(rmesa->tcl.elt_dma_bo->ptr + + rmesa->tcl.elt_used); - rmesa->store.cmd_used += nr*2; + rmesa->tcl.elt_used += nr*2; return dest; } else { - if (rmesa->dma.flush) - rmesa->dma.flush( rmesa ); + if (rmesa->radeon.dma.flush) + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + ELTS_BUFSZ(nr) ); + rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__); r200EmitAOS( rmesa, - rmesa->tcl.aos_components, rmesa->tcl.nr_aos_components, 0 ); return r200AllocEltsOpenEnded( rmesa, rmesa->tcl.hw_primitive, nr ); @@ -188,13 +186,14 @@ static void r200EmitPrim( GLcontext *ctx, r200ContextPtr rmesa = R200_CONTEXT( ctx ); r200TclPrimitive( ctx, prim, hwprim ); - r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + VBUF_BUFSZ ); + // fprintf(stderr,"Emit prim %d\n", rmesa->tcl.nr_aos_components); + rcommonEnsureCmdBufSpace( &rmesa->radeon, + AOS_BUFSZ(rmesa->tcl.nr_aos_components) + + rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); r200EmitAOS( rmesa, - rmesa->tcl.aos_components, - rmesa->tcl.nr_aos_components, - start ); + rmesa->tcl.nr_aos_components, + start ); /* Why couldn't this packet have taken an offset param? */ @@ -394,7 +393,7 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, /* TODO: separate this from the swtnl pipeline */ - if (rmesa->TclFallback) + if (rmesa->radeon.TclFallback) return GL_TRUE; /* fallback to software t&l */ if (R200_DEBUG & DEBUG_PRIMS) @@ -405,8 +404,9 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, /* Validate state: */ - if (rmesa->NewGLState) - r200ValidateState( ctx ); + if (rmesa->radeon.NewGLState) + if (!r200ValidateState( ctx )) + return GL_TRUE; /* fallback to sw t&l */ if (!ctx->VertexProgram._Enabled) { /* NOTE: inputs != tnl->render_inputs - these are the untransformed @@ -565,15 +565,11 @@ static void transition_to_hwtnl( GLcontext *ctx ) tnl->Driver.NotifyMaterialChange = r200UpdateMaterial; - if ( rmesa->dma.flush ) - rmesa->dma.flush( rmesa ); + if ( rmesa->radeon.dma.flush ) + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - rmesa->dma.flush = NULL; + rmesa->radeon.dma.flush = NULL; - if (rmesa->swtcl.indexed_verts.buf) - r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, - __FUNCTION__ ); - R200_STATECHANGE( rmesa, vap ); rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE; rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_FORCE_W_TO_ONE; @@ -631,10 +627,10 @@ static char *getFallbackString(GLuint bit) void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - GLuint oldfallback = rmesa->TclFallback; + GLuint oldfallback = rmesa->radeon.TclFallback; if (mode) { - rmesa->TclFallback |= bit; + rmesa->radeon.TclFallback |= bit; if (oldfallback == 0) { if (R200_DEBUG & DEBUG_FALLBACKS) fprintf(stderr, "R200 begin tcl fallback %s\n", @@ -643,7 +639,7 @@ void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) } } else { - rmesa->TclFallback &= ~bit; + rmesa->radeon.TclFallback &= ~bit; if (oldfallback == bit) { if (R200_DEBUG & DEBUG_FALLBACKS) fprintf(stderr, "R200 end tcl fallback %s\n", diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c index 5a4db33f44..19a6cad80b 100644 --- a/src/mesa/drivers/dri/r200/r200_tex.c +++ b/src/mesa/drivers/dri/r200/r200_tex.c @@ -43,8 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/teximage.h" #include "main/texobj.h" -#include "texmem.h" - +#include "radeon_mipmap_tree.h" #include "r200_context.h" #include "r200_state.h" #include "r200_ioctl.h" @@ -63,10 +62,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * \param twrap Wrap mode for the \a t texture coordinate */ -static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap ) +static void r200SetTexWrap( radeonTexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap ) { GLboolean is_clamp = GL_FALSE; GLboolean is_clamp_to_border = GL_FALSE; + struct gl_texture_object *tObj = &t->base; t->pp_txfilter &= ~(R200_CLAMP_S_MASK | R200_CLAMP_T_MASK | R200_BORDER_MODE_D3D); @@ -103,7 +103,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum _mesa_problem(NULL, "bad S wrap mode in %s", __FUNCTION__); } - if (t->base.tObj->Target != GL_TEXTURE_1D) { + if (tObj->Target != GL_TEXTURE_1D) { switch ( twrap ) { case GL_REPEAT: t->pp_txfilter |= R200_CLAMP_T_WRAP; @@ -180,7 +180,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum t->border_fallback = (is_clamp && is_clamp_to_border); } -static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max ) +static void r200SetTexMaxAnisotropy( radeonTexObjPtr t, GLfloat max ) { t->pp_txfilter &= ~R200_MAX_ANISO_MASK; @@ -205,10 +205,13 @@ static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max ) * \param magf Texture magnification mode */ -static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf ) +static void r200SetTexFilter( radeonTexObjPtr t, GLenum minf, GLenum magf ) { GLuint anisotropy = (t->pp_txfilter & R200_MAX_ANISO_MASK); + /* Force revalidation to account for switches from/to mipmapping. */ + t->validated = GL_FALSE; + t->pp_txfilter &= ~(R200_MIN_FILTER_MASK | R200_MAG_FILTER_MASK); t->pp_txformat_x &= ~R200_VOLUME_FILTER_MASK; @@ -267,693 +270,12 @@ static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf ) } } -static void r200SetTexBorderColor( r200TexObjPtr t, GLubyte c[4] ) -{ - t->pp_border_color = r200PackColor( 4, c[0], c[1], c[2], c[3] ); -} - - -/** - * Allocate space for and load the mesa images into the texture memory block. - * This will happen before drawing with a new texture, or drawing with a - * texture after it was swapped out or teximaged again. - */ - -static r200TexObjPtr r200AllocTexObj( struct gl_texture_object *texObj ) -{ - r200TexObjPtr t; - - t = CALLOC_STRUCT( r200_tex_obj ); - texObj->DriverData = t; - if ( t != NULL ) { - if ( R200_DEBUG & DEBUG_TEXTURE ) { - fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)texObj, - (void *)t ); - } - - /* Initialize non-image-dependent parts of the state: - */ - t->base.tObj = texObj; - t->border_fallback = GL_FALSE; - - make_empty_list( & t->base ); - - r200SetTexWrap( t, texObj->WrapS, texObj->WrapT, texObj->WrapR ); - r200SetTexMaxAnisotropy( t, texObj->MaxAnisotropy ); - r200SetTexFilter( t, texObj->MinFilter, texObj->MagFilter ); - r200SetTexBorderColor( t, texObj->_BorderChan ); - } - - return t; -} - -/* try to find a format which will only need a memcopy */ -static const struct gl_texture_format * -r200Choose8888TexFormat( GLenum srcFormat, GLenum srcType ) -{ - const GLuint ui = 1; - const GLubyte littleEndian = *((const GLubyte *) &ui); - - if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) || - (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) { - return &_mesa_texformat_rgba8888; - } - else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || - (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) { - return &_mesa_texformat_rgba8888_rev; - } - else return _dri_texformat_argb8888; -} - -static const struct gl_texture_format * -r200ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, - GLenum format, GLenum type ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - const GLboolean do32bpt = - ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32 ); - const GLboolean force16bpt = - ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16 ); - (void) format; - - switch ( internalFormat ) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - switch ( type ) { - case GL_UNSIGNED_INT_10_10_10_2: - case GL_UNSIGNED_INT_2_10_10_10_REV: - return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb1555; - case GL_UNSIGNED_SHORT_4_4_4_4: - case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return _dri_texformat_argb4444; - case GL_UNSIGNED_SHORT_5_5_5_1: - case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return _dri_texformat_argb1555; - default: - return do32bpt ? - r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444; - } - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - switch ( type ) { - case GL_UNSIGNED_SHORT_4_4_4_4: - case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return _dri_texformat_argb4444; - case GL_UNSIGNED_SHORT_5_5_5_1: - case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return _dri_texformat_argb1555; - case GL_UNSIGNED_SHORT_5_6_5: - case GL_UNSIGNED_SHORT_5_6_5_REV: - return _dri_texformat_rgb565; - default: - return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565; - } - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return !force16bpt ? - r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return _dri_texformat_argb4444; - - case GL_RGB5_A1: - return _dri_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return !force16bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return _dri_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - /* can't use a8 format since interpreting hw I8 as a8 would result - in wrong rgb values (same as alpha value instead of 0). */ - return _dri_texformat_al88; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return _dri_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return _dri_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return _dri_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_APPLE || - type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - default: - _mesa_problem(ctx, - "unexpected internalFormat 0x%x in r200ChooseTextureFormat", - (int) internalFormat); - return NULL; - } - - return NULL; /* never get here */ -} - - -static GLboolean -r200ValidateClientStorage( GLcontext *ctx, GLenum target, - GLint internalFormat, - GLint srcWidth, GLint srcHeight, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) - -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf(stderr, "intformat %s format %s type %s\n", - _mesa_lookup_enum_by_nr( internalFormat ), - _mesa_lookup_enum_by_nr( format ), - _mesa_lookup_enum_by_nr( type )); - - if (!ctx->Unpack.ClientStorage) - return 0; - - if (ctx->_ImageTransferState || - texImage->IsCompressed || - texObj->GenerateMipmap) - return 0; - - - /* This list is incomplete, may be different on ppc??? - */ - switch ( internalFormat ) { - case GL_RGBA: - if ( format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV ) { - texImage->TexFormat = _dri_texformat_argb8888; - } - else - return 0; - break; - - case GL_RGB: - if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) { - texImage->TexFormat = _dri_texformat_rgb565; - } - else - return 0; - break; - - case GL_YCBCR_MESA: - if ( format == GL_YCBCR_MESA && - type == GL_UNSIGNED_SHORT_8_8_REV_APPLE ) { - texImage->TexFormat = &_mesa_texformat_ycbcr_rev; - } - else if ( format == GL_YCBCR_MESA && - (type == GL_UNSIGNED_SHORT_8_8_APPLE || - type == GL_UNSIGNED_BYTE)) { - texImage->TexFormat = &_mesa_texformat_ycbcr; - } - else - return 0; - break; - - default: - return 0; - } - - /* Could deal with these packing issues, but currently don't: - */ - if (packing->SkipPixels || - packing->SkipRows || - packing->SwapBytes || - packing->LsbFirst) { - return 0; - } - - { - GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth, - format, type); - - - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf(stderr, "%s: srcRowStride %d/%x\n", - __FUNCTION__, srcRowStride, srcRowStride); - - /* Could check this later in upload, pitch restrictions could be - * relaxed, but would need to store the image pitch somewhere, - * as packing details might change before image is uploaded: - */ - if (!r200IsGartMemory( rmesa, pixels, srcHeight * srcRowStride ) || - (srcRowStride & 63)) - return 0; - - - /* Have validated that _mesa_transfer_teximage would be a straight - * memcpy at this point. NOTE: future calls to TexSubImage will - * overwrite the client data. This is explicitly mentioned in the - * extension spec. - */ - texImage->Data = (void *)pixels; - texImage->IsClientData = GL_TRUE; - texImage->RowStride = srcRowStride / texImage->TexFormat->TexelBytes; - - return 1; - } -} - - -static void r200TexImage1D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D"); - return; - } - } - - /* Note, this will call ChooseTextureFormat */ - _mesa_store_teximage1d(ctx, target, level, internalFormat, - width, border, format, type, pixels, - &ctx->Unpack, texObj, texImage); - - t->dirty_images[0] |= (1 << level); -} - - -static void r200TexSubImage1D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - assert( t ); /* this _should_ be true */ - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D"); - return; - } - } - - _mesa_store_texsubimage1d(ctx, target, level, xoffset, width, - format, type, pixels, packing, texObj, - texImage); - - t->dirty_images[0] |= (1 << level); -} - - -static void r200TexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } - - if ( t != NULL ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D"); - return; - } - } - - texImage->IsClientData = GL_FALSE; - - if (r200ValidateClientStorage( ctx, target, - internalFormat, - width, height, - format, type, pixels, - packing, texObj, texImage)) { - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); - } - else { - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); - - /* Normal path: copy (to cached memory) and eventually upload - * via another copy to GART memory and then a blit... Could - * eliminate one copy by going straight to (permanent) GART. - * - * Note, this will call r200ChooseTextureFormat. - */ - _mesa_store_teximage2d(ctx, target, level, internalFormat, - width, height, border, format, type, pixels, - &ctx->Unpack, texObj, texImage); - - t->dirty_images[face] |= (1 << level); - } -} - - -static void r200TexSubImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } - - assert( t ); /* this _should_ be true */ - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D"); - return; - } - } - - _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, type, pixels, packing, texObj, - texImage); - - t->dirty_images[face] |= (1 << level); -} - - -static void r200CompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) +static void r200SetTexBorderColor( radeonTexObjPtr t, GLubyte c[4] ) { - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } - - if ( t != NULL ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D"); - return; - } - } - - texImage->IsClientData = GL_FALSE; -/* can't call this, different parameters. Would never evaluate to true anyway currently - if (r200ValidateClientStorage( ctx, target, - internalFormat, - width, height, - format, type, pixels, - packing, texObj, texImage)) { - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); - } - else */{ - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); - - /* Normal path: copy (to cached memory) and eventually upload - * via another copy to GART memory and then a blit... Could - * eliminate one copy by going straight to (permanent) GART. - * - * Note, this will call r200ChooseTextureFormat. - */ - _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width, - height, border, imageSize, data, texObj, texImage); - - t->dirty_images[face] |= (1 << level); - } -} - - -static void r200CompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } - - assert( t ); /* this _should_ be true */ - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexSubImage2D"); - return; - } - } - - _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, imageSize, data, texObj, texImage); - - t->dirty_images[face] |= (1 << level); -} - - -#if ENABLE_HW_3D_TEXTURE -static void r200TexImage3D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D"); - return; - } - } - - texImage->IsClientData = GL_FALSE; - -#if 0 - if (r200ValidateClientStorage( ctx, target, - internalFormat, - width, height, - format, type, pixels, - packing, texObj, texImage)) { - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); - } - else -#endif - { - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); - - /* Normal path: copy (to cached memory) and eventually upload - * via another copy to GART memory and then a blit... Could - * eliminate one copy by going straight to (permanent) GART. - * - * Note, this will call r200ChooseTextureFormat. - */ - _mesa_store_teximage3d(ctx, target, level, internalFormat, - width, height, depth, border, - format, type, pixels, - &ctx->Unpack, texObj, texImage); - - t->dirty_images[0] |= (1 << level); - } + t->pp_border_color = radeonPackColor( 4, c[0], c[1], c[2], c[3] ); } -#endif - -#if ENABLE_HW_3D_TEXTURE -static void -r200TexSubImage3D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - -/* fprintf(stderr, "%s\n", __FUNCTION__); */ - - assert( t ); /* this _should_ be true */ - if ( t ) { - driSwapOutTextureObject( t ); - } - else { - t = (driTextureObject *) r200AllocTexObj( texObj ); - if (!t) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D"); - return; - } - texObj->DriverData = t; - } - _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - - t->dirty_images[0] |= (1 << level); -} -#endif @@ -978,7 +300,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target, GLubyte c[4]; GLuint envColor; UNCLAMPED_FLOAT_TO_RGBA_CHAN( c, texUnit->EnvColor ); - envColor = r200PackColor( 4, c[0], c[1], c[2], c[3] ); + envColor = radeonPackColor( 4, c[0], c[1], c[2], c[3] ); if ( rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] != envColor ) { R200_STATECHANGE( rmesa, tf ); rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] = envColor; @@ -997,7 +319,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target, * NOTE: Add a small bias to the bias for conform mipsel.c test. */ bias = *param + .01; - min = driQueryOptionb (&rmesa->optionCache, "no_neg_lod_bias") ? + min = driQueryOptionb (&rmesa->radeon.optionCache, "no_neg_lod_bias") ? 0.0 : -16.0; bias = CLAMP( bias, min, 16.0 ); b = (int)(bias * fixed_one) & R200_LOD_BIAS_MASK; @@ -1034,7 +356,7 @@ static void r200TexParameter( GLcontext *ctx, GLenum target, struct gl_texture_object *texObj, GLenum pname, const GLfloat *params ) { - r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData; + radeonTexObj* t = radeon_tex_obj(texObj); if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { fprintf( stderr, "%s( %s )\n", __FUNCTION__, @@ -1068,59 +390,46 @@ static void r200TexParameter( GLcontext *ctx, GLenum target, * we just have to rely on loading the right subset of mipmap levels * to simulate a clamped LOD. */ - driSwapOutTextureObject( (driTextureObject *) t ); + if (t->mt) { + radeon_miptree_unreference(t->mt); + t->mt = 0; + t->validated = GL_FALSE; + } break; default: return; } - - /* Mark this texobj as dirty (one bit per tex unit) - */ - t->dirty_state = TEX_ALL; } - -static void r200BindTexture( GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj ) -{ - if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { - fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, (void *)texObj, - ctx->Texture.CurrentUnit ); - } - - if ( (target == GL_TEXTURE_1D) - || (target == GL_TEXTURE_2D) -#if ENABLE_HW_3D_TEXTURE - || (target == GL_TEXTURE_3D) -#endif - || (target == GL_TEXTURE_CUBE_MAP) - || (target == GL_TEXTURE_RECTANGLE_NV) ) { - assert( texObj->DriverData != NULL ); - } -} - - -static void r200DeleteTexture( GLcontext *ctx, - struct gl_texture_object *texObj ) +static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { - fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj, - _mesa_lookup_enum_by_nr( texObj->Target ) ); + radeonTexObj* t = radeon_tex_obj(texObj); + + if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) { + fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__, + (void *)texObj, + _mesa_lookup_enum_by_nr(texObj->Target)); + } + + if (rmesa) { + int i; + radeon_firevertices(&rmesa->radeon); + for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) { + if ( t == rmesa->state.texture.unit[i].texobj ) { + rmesa->state.texture.unit[i].texobj = NULL; + rmesa->hw.tex[i].dirty = GL_FALSE; + rmesa->hw.cube[i].dirty = GL_FALSE; + } + } } - - if ( t != NULL ) { - if ( rmesa ) { - R200_FIREVERTICES( rmesa ); - } - - driDestroyTextureObject( t ); + + if (t->mt) { + radeon_miptree_unreference(t->mt); + t->mt = 0; } - /* Free mipmap images and the texture object itself */ _mesa_delete_texture_object(ctx, texObj); } @@ -1150,46 +459,59 @@ static void r200TexGen( GLcontext *ctx, * Called via ctx->Driver.NewTextureObject. * Note: this function will be called during context creation to * allocate the default texture objects. - * Note: we could use containment here to 'derive' the driver-specific - * texture object from the core mesa gl_texture_object. Not done at this time. * Fixup MaxAnisotropy according to user preference. */ -static struct gl_texture_object * -r200NewTextureObject( GLcontext *ctx, GLuint name, GLenum target ) +static struct gl_texture_object *r200NewTextureObject(GLcontext * ctx, + GLuint name, + GLenum target) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_object *obj; - obj = _mesa_new_texture_object(ctx, name, target); - if (!obj) - return NULL; - obj->MaxAnisotropy = rmesa->initialMaxAnisotropy; - r200AllocTexObj( obj ); - return obj; + radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj); + + + if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) { + fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__, + t, _mesa_lookup_enum_by_nr(target)); + } + + _mesa_initialize_texture_object(&t->base, name, target); + t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy; + + /* Initialize hardware state */ + r200SetTexWrap( t, t->base.WrapS, t->base.WrapT, t->base.WrapR ); + r200SetTexMaxAnisotropy( t, t->base.MaxAnisotropy ); + r200SetTexFilter(t, t->base.MinFilter, t->base.MagFilter); + r200SetTexBorderColor(t, t->base._BorderChan); + + return &t->base; } + void r200InitTextureFuncs( struct dd_function_table *functions ) { /* Note: we only plug in the functions we implement in the driver * since _mesa_init_driver_functions() was already called. */ - functions->ChooseTextureFormat = r200ChooseTextureFormat; - functions->TexImage1D = r200TexImage1D; - functions->TexImage2D = r200TexImage2D; + functions->ChooseTextureFormat = radeonChooseTextureFormat; + functions->TexImage1D = radeonTexImage1D; + functions->TexImage2D = radeonTexImage2D; #if ENABLE_HW_3D_TEXTURE - functions->TexImage3D = r200TexImage3D; + functions->TexImage3D = radeonTexImage3D; #else functions->TexImage3D = _mesa_store_teximage3d; #endif - functions->TexSubImage1D = r200TexSubImage1D; - functions->TexSubImage2D = r200TexSubImage2D; + functions->TexSubImage1D = radeonTexSubImage1D; + functions->TexSubImage2D = radeonTexSubImage2D; #if ENABLE_HW_3D_TEXTURE - functions->TexSubImage3D = r200TexSubImage3D; + functions->TexSubImage3D = radeonTexSubImage3D; #else functions->TexSubImage3D = _mesa_store_texsubimage3d; #endif + functions->GetTexImage = radeonGetTexImage; + functions->GetCompressedTexImage = radeonGetCompressedTexImage; functions->NewTextureObject = r200NewTextureObject; - functions->BindTexture = r200BindTexture; + // functions->BindTexture = r200BindTexture; functions->DeleteTexture = r200DeleteTexture; functions->IsTextureResident = driIsTextureResident; @@ -1197,22 +519,16 @@ void r200InitTextureFuncs( struct dd_function_table *functions ) functions->TexParameter = r200TexParameter; functions->TexGen = r200TexGen; - functions->CompressedTexImage2D = r200CompressedTexImage2D; - functions->CompressedTexSubImage2D = r200CompressedTexSubImage2D; + functions->CompressedTexImage2D = radeonCompressedTexImage2D; + functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D; + + functions->GenerateMipmap = radeonGenerateMipmap; + + functions->NewTextureImage = radeonNewTextureImage; + functions->FreeTexImageData = radeonFreeTexImageData; + functions->MapTexture = radeonMapTexture; + functions->UnmapTexture = radeonUnmapTexture; driInitTextureFormats(); -#if 000 - /* moved or obsolete code */ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - driInitTextureObjects( ctx, & rmesa->swapped, - DRI_TEXMGR_DO_TEXTURE_1D - | DRI_TEXMGR_DO_TEXTURE_2D ); - - /* Hack: r200NewTextureObject is not yet installed when the - * default textures are created. Therefore set MaxAnisotropy of the - * default 2D texture now. */ - ctx->Shared->Default2D->MaxAnisotropy = driQueryOptionf (&rmesa->optionCache, - "def_max_anisotropy"); -#endif } diff --git a/src/mesa/drivers/dri/r200/r200_tex.h b/src/mesa/drivers/dri/r200/r200_tex.h index 10ff8e8a66..886682b36b 100644 --- a/src/mesa/drivers/dri/r200/r200_tex.h +++ b/src/mesa/drivers/dri/r200/r200_tex.h @@ -35,15 +35,16 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifndef __R200_TEX_H__ #define __R200_TEX_H__ +extern void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv); extern void r200SetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); extern void r200UpdateTextureState( GLcontext *ctx ); -extern int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face ); +extern int r200UploadTexImages( r200ContextPtr rmesa, radeonTexObjPtr t, GLuint face ); -extern void r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t ); +extern void r200DestroyTexObj( r200ContextPtr rmesa, radeonTexObjPtr t ); extern void r200InitTextureFuncs( struct dd_function_table *functions ); diff --git a/src/mesa/drivers/dri/r200/r200_texmem.c b/src/mesa/drivers/dri/r200/r200_texmem.c deleted file mode 100644 index 3b81ac0c80..0000000000 --- a/src/mesa/drivers/dri/r200/r200_texmem.c +++ /dev/null @@ -1,530 +0,0 @@ -/************************************************************************** - -Copyright (C) Tungsten Graphics 2002. All Rights Reserved. -The Weather Channel, Inc. funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 -license. This notice must be preserved. - -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 ATI, VA LINUX SYSTEMS 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. - -**************************************************************************/ - -/* - * Authors: - * Kevin E. Martin <martin@valinux.com> - * Gareth Hughes <gareth@valinux.com> - * - */ - -#include <errno.h> - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/context.h" -#include "main/colormac.h" -#include "main/macros.h" -#include "r200_context.h" -#include "r200_ioctl.h" -#include "r200_tex.h" -#include "radeon_reg.h" - -#include <unistd.h> /* for usleep() */ - - -/** - * Destroy any device-dependent state associated with the texture. This may - * include NULLing out hardware state that points to the texture. - */ -void -r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t ) -{ - if ( R200_DEBUG & DEBUG_TEXTURE ) { - fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, - (void *)t, (void *)t->base.tObj ); - } - - if ( rmesa != NULL ) { - unsigned i; - - - for ( i = 0 ; i < rmesa->glCtx->Const.MaxTextureUnits ; i++ ) { - if ( t == rmesa->state.texture.unit[i].texobj ) { - rmesa->state.texture.unit[i].texobj = NULL; - rmesa->hw.tex[i].dirty = GL_FALSE; - rmesa->hw.cube[i].dirty = GL_FALSE; - } - } - } -} - - -/* ------------------------------------------------------------ - * Texture image conversions - */ - - -static void r200UploadGARTClientSubImage( r200ContextPtr rmesa, - r200TexObjPtr t, - struct gl_texture_image *texImage, - GLint hwlevel, - GLint x, GLint y, - GLint width, GLint height ) -{ - const struct gl_texture_format *texFormat = texImage->TexFormat; - GLuint srcPitch, dstPitch; - int blit_format; - int srcOffset; - - /* - * XXX it appears that we always upload the full image, not a subimage. - * I.e. x==0, y==0, width=texWidth, height=texWidth. If this is ever - * changed, the src pitch will have to change. - */ - switch ( texFormat->TexelBytes ) { - case 1: - blit_format = R200_CP_COLOR_FORMAT_CI8; - srcPitch = t->image[0][0].width * texFormat->TexelBytes; - dstPitch = t->image[0][0].width * texFormat->TexelBytes; - break; - case 2: - blit_format = R200_CP_COLOR_FORMAT_RGB565; - srcPitch = t->image[0][0].width * texFormat->TexelBytes; - dstPitch = t->image[0][0].width * texFormat->TexelBytes; - break; - case 4: - blit_format = R200_CP_COLOR_FORMAT_ARGB8888; - srcPitch = t->image[0][0].width * texFormat->TexelBytes; - dstPitch = t->image[0][0].width * texFormat->TexelBytes; - break; - default: - return; - } - - t->image[0][hwlevel].data = texImage->Data; - srcOffset = r200GartOffsetFromVirtual( rmesa, texImage->Data ); - - assert( srcOffset != ~0 ); - - /* Don't currently need to cope with small pitches? - */ - width = texImage->Width; - height = texImage->Height; - - r200EmitWait( rmesa, RADEON_WAIT_3D ); - - r200EmitBlit( rmesa, blit_format, - srcPitch, - srcOffset, - dstPitch, - t->bufAddr, - x, - y, - t->image[0][hwlevel].x + x, - t->image[0][hwlevel].y + y, - width, - height ); - - r200EmitWait( rmesa, RADEON_WAIT_2D ); -} - -static void r200UploadRectSubImage( r200ContextPtr rmesa, - r200TexObjPtr t, - struct gl_texture_image *texImage, - GLint x, GLint y, - GLint width, GLint height ) -{ - const struct gl_texture_format *texFormat = texImage->TexFormat; - int blit_format, dstPitch, done; - - switch ( texFormat->TexelBytes ) { - case 1: - blit_format = R200_CP_COLOR_FORMAT_CI8; - break; - case 2: - blit_format = R200_CP_COLOR_FORMAT_RGB565; - break; - case 4: - blit_format = R200_CP_COLOR_FORMAT_ARGB8888; - break; - default: - return; - } - - t->image[0][0].data = texImage->Data; - - /* Currently don't need to cope with small pitches. - */ - width = texImage->Width; - height = texImage->Height; - dstPitch = t->pp_txpitch + 32; - - if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) { - /* In this case, could also use GART texturing. This is - * currently disabled, but has been tested & works. - */ - if ( !t->image_override ) - t->pp_txoffset = r200GartOffsetFromVirtual( rmesa, texImage->Data ); - t->pp_txpitch = texImage->RowStride * texFormat->TexelBytes - 32; - - if (R200_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, - "Using GART texturing for rectangular client texture\n"); - - /* Release FB memory allocated for this image: - */ - /* FIXME This may not be correct as driSwapOutTextureObject sets - * FIXME dirty_images. It may be fine, though. - */ - if ( t->base.memBlock ) { - driSwapOutTextureObject( (driTextureObject *) t ); - } - } - else if (texImage->IsClientData) { - /* Data already in GART memory, with usable pitch. - */ - GLuint srcPitch; - srcPitch = texImage->RowStride * texFormat->TexelBytes; - r200EmitBlit( rmesa, - blit_format, - srcPitch, - r200GartOffsetFromVirtual( rmesa, texImage->Data ), - dstPitch, t->bufAddr, - 0, 0, - 0, 0, - width, height ); - } - else { - /* Data not in GART memory, or bad pitch. - */ - for (done = 0; done < height ; ) { - struct r200_dma_region region; - int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch ); - int src_pitch; - char *tex; - - src_pitch = texImage->RowStride * texFormat->TexelBytes; - - tex = (char *)texImage->Data + done * src_pitch; - - memset(®ion, 0, sizeof(region)); - r200AllocDmaRegion( rmesa, ®ion, lines * dstPitch, 1024 ); - - /* Copy texdata to dma: - */ - if (0) - fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n", - __FUNCTION__, src_pitch, dstPitch); - - if (src_pitch == dstPitch) { - memcpy( region.address + region.start, tex, lines * src_pitch ); - } - else { - char *buf = region.address + region.start; - int i; - for (i = 0 ; i < lines ; i++) { - memcpy( buf, tex, src_pitch ); - buf += dstPitch; - tex += src_pitch; - } - } - - r200EmitWait( rmesa, RADEON_WAIT_3D ); - - /* Blit to framebuffer - */ - r200EmitBlit( rmesa, - blit_format, - dstPitch, GET_START( ®ion ), - dstPitch | (t->tile_bits >> 16), - t->bufAddr, - 0, 0, - 0, done, - width, lines ); - - r200EmitWait( rmesa, RADEON_WAIT_2D ); - - r200ReleaseDmaRegion( rmesa, ®ion, __FUNCTION__ ); - done += lines; - } - } -} - - -/** - * Upload the texture image associated with texture \a t at the specified - * level at the address relative to \a start. - */ -static void uploadSubImage( r200ContextPtr rmesa, r200TexObjPtr t, - GLint hwlevel, - GLint x, GLint y, GLint width, GLint height, - GLuint face ) -{ - struct gl_texture_image *texImage = NULL; - GLuint offset; - GLint imageWidth, imageHeight; - GLint ret; - drm_radeon_texture_t tex; - drm_radeon_tex_image_t tmp; - const int level = hwlevel + t->base.firstLevel; - - if ( R200_DEBUG & DEBUG_TEXTURE ) { - fprintf( stderr, "%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n", - __FUNCTION__, (void *)t, (void *)t->base.tObj, - level, width, height, face ); - } - - ASSERT(face < 6); - - /* Ensure we have a valid texture to upload */ - if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) { - _mesa_problem(NULL, "bad texture level in %s", __FUNCTION__); - return; - } - - texImage = t->base.tObj->Image[face][level]; - - if ( !texImage ) { - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level ); - return; - } - if ( !texImage->Data ) { - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ ); - return; - } - - - if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) { - assert(level == 0); - assert(hwlevel == 0); - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__); - r200UploadRectSubImage( rmesa, t, texImage, x, y, width, height ); - return; - } - else if (texImage->IsClientData) { - if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf( stderr, "%s: image data is in GART client storage\n", - __FUNCTION__); - r200UploadGARTClientSubImage( rmesa, t, texImage, hwlevel, - x, y, width, height ); - return; - } - else if ( R200_DEBUG & DEBUG_TEXTURE ) - fprintf( stderr, "%s: image data is in normal memory\n", - __FUNCTION__); - - - imageWidth = texImage->Width; - imageHeight = texImage->Height; - - offset = t->bufAddr + t->base.totalSize / 6 * face; - - if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) { - GLint imageX = 0; - GLint imageY = 0; - GLint blitX = t->image[face][hwlevel].x; - GLint blitY = t->image[face][hwlevel].y; - GLint blitWidth = t->image[face][hwlevel].width; - GLint blitHeight = t->image[face][hwlevel].height; - fprintf( stderr, " upload image: %d,%d at %d,%d\n", - imageWidth, imageHeight, imageX, imageY ); - fprintf( stderr, " upload blit: %d,%d at %d,%d\n", - blitWidth, blitHeight, blitX, blitY ); - fprintf( stderr, " blit ofs: 0x%07x level: %d/%d\n", - (GLuint)offset, hwlevel, level ); - } - - t->image[face][hwlevel].data = texImage->Data; - - /* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct. - * NOTE: we're always use a 1KB-wide blit and I8 texture format. - * We used to use 1, 2 and 4-byte texels and used to use the texture - * width to dictate the blit width - but that won't work for compressed - * textures. (Brian) - * NOTE: can't do that with texture tiling. (sroland) - */ - tex.offset = offset; - tex.image = &tmp; - /* copy (x,y,width,height,data) */ - memcpy( &tmp, &t->image[face][hwlevel], sizeof(tmp) ); - - if (texImage->TexFormat->TexelBytes) { - /* use multi-byte upload scheme */ - tex.height = imageHeight; - tex.width = imageWidth; - tex.format = t->pp_txformat & R200_TXFORMAT_FORMAT_MASK; - if (tex.format == R200_TXFORMAT_ABGR8888) { - /* drm will refuse abgr8888 textures. */ - tex.format = R200_TXFORMAT_ARGB8888; - } - tex.pitch = MAX2((texImage->Width * texImage->TexFormat->TexelBytes) / 64, 1); - tex.offset += tmp.x & ~1023; - tmp.x = tmp.x % 1024; - if (t->tile_bits & R200_TXO_MICRO_TILE) { - /* need something like "tiled coordinates" ? */ - tmp.y = tmp.x / (tex.pitch * 128) * 2; - tmp.x = tmp.x % (tex.pitch * 128) / 2 / texImage->TexFormat->TexelBytes; - tex.pitch |= RADEON_DST_TILE_MICRO >> 22; - } - else { - tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1); - } - if ((t->tile_bits & R200_TXO_MACRO_TILE) && - (texImage->Width * texImage->TexFormat->TexelBytes >= 256) && - ((!(t->tile_bits & R200_TXO_MICRO_TILE) && (texImage->Height >= 8)) || - (texImage->Height >= 16))) { - /* weird: R200 disables macro tiling if mip width is smaller than 256 bytes, - OR if height is smaller than 8 automatically, but if micro tiling is active - the limit is height 16 instead ? */ - tex.pitch |= RADEON_DST_TILE_MACRO >> 22; - } - } - else { - /* In case of for instance 8x8 texture (2x2 dxt blocks), padding after the first two blocks is - needed (only with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */ - /* set tex.height to 1/4 since 1 "macropixel" (dxt-block) has 4 real pixels. Needed - so the kernel module reads the right amount of data. */ - tex.format = R200_TXFORMAT_I8; /* any 1-byte texel format */ - tex.pitch = (BLIT_WIDTH_BYTES / 64); - tex.height = (imageHeight + 3) / 4; - tex.width = (imageWidth + 3) / 4; - switch (t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) { - case R200_TXFORMAT_DXT1: - tex.width *= 8; - break; - case R200_TXFORMAT_DXT23: - case R200_TXFORMAT_DXT45: - tex.width *= 16; - break; - default: - fprintf(stderr, "unknown compressed tex format in uploadSubImage\n"); - } - } - - LOCK_HARDWARE( rmesa ); - do { - ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE, - &tex, sizeof(drm_radeon_texture_t) ); - if (ret) { - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "DRM_RADEON_TEXTURE: again!\n"); - usleep(1); - } - } while ( ret == -EAGAIN ); - - UNLOCK_HARDWARE( rmesa ); - - if ( ret ) { - fprintf( stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret ); - fprintf( stderr, " offset=0x%08x\n", - offset ); - fprintf( stderr, " image width=%d height=%d\n", - imageWidth, imageHeight ); - fprintf( stderr, " blit width=%d height=%d data=%p\n", - t->image[face][hwlevel].width, t->image[face][hwlevel].height, - t->image[face][hwlevel].data ); - exit( 1 ); - } -} - - -/** - * Upload the texture images associated with texture \a t. This might - * require the allocation of texture memory. - * - * \param rmesa Context pointer - * \param t Texture to be uploaded - * \param face Cube map face to be uploaded. Zero for non-cube maps. - */ - -int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face ) -{ - const int numLevels = t->base.lastLevel - t->base.firstLevel + 1; - - if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) { - fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__, - (void *)rmesa->glCtx, (void *)t->base.tObj, t->base.totalSize, - t->base.firstLevel, t->base.lastLevel ); - } - - if ( !t || t->base.totalSize == 0 || t->image_override ) - return 0; - - if (R200_DEBUG & DEBUG_SYNC) { - fprintf(stderr, "%s: Syncing\n", __FUNCTION__ ); - r200Finish( rmesa->glCtx ); - } - - LOCK_HARDWARE( rmesa ); - - if ( t->base.memBlock == NULL ) { - int heap; - - heap = driAllocateTexture( rmesa->texture_heaps, rmesa->nr_heaps, - (driTextureObject *) t ); - if ( heap == -1 ) { - UNLOCK_HARDWARE( rmesa ); - return -1; - } - - /* Set the base offset of the texture image */ - t->bufAddr = rmesa->r200Screen->texOffset[heap] - + t->base.memBlock->ofs; - t->pp_txoffset = t->bufAddr; - - if (!(t->base.tObj->Image[0][0]->IsClientData)) { - /* hope it's safe to add that here... */ - t->pp_txoffset |= t->tile_bits; - } - - /* Mark this texobj as dirty on all units: - */ - t->dirty_state = TEX_ALL; - } - - /* Let the world know we've used this memory recently. - */ - driUpdateTextureLRU( (driTextureObject *) t ); - UNLOCK_HARDWARE( rmesa ); - - /* Upload any images that are new */ - if (t->base.dirty_images[face]) { - int i; - for ( i = 0 ; i < numLevels ; i++ ) { - if ( (t->base.dirty_images[face] & (1 << (i+t->base.firstLevel))) != 0 ) { - uploadSubImage( rmesa, t, i, 0, 0, t->image[face][i].width, - t->image[face][i].height, face ); - } - } - t->base.dirty_images[face] = 0; - } - - - if (R200_DEBUG & DEBUG_SYNC) { - fprintf(stderr, "%s: Syncing\n", __FUNCTION__ ); - r200Finish( rmesa->glCtx ); - } - - return 0; -} diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c index 0ad5651cd4..e9cee1a637 100644 --- a/src/mesa/drivers/dri/r200/r200_texstate.c +++ b/src/mesa/drivers/dri/r200/r200_texstate.c @@ -37,9 +37,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/context.h" #include "main/macros.h" #include "main/texformat.h" +#include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" +#include "radeon_common.h" +#include "radeon_mipmap_tree.h" #include "r200_context.h" #include "r200_state.h" #include "r200_ioctl.h" @@ -139,257 +142,6 @@ static const struct tx_table tx_table_le[] = #undef _ALPHA #undef _INVALID -/** - * This function computes the number of bytes of storage needed for - * the given texture object (all mipmap levels, all cube faces). - * The \c image[face][level].x/y/width/height parameters for upload/blitting - * are computed here. \c pp_txfilter, \c pp_txformat, etc. will be set here - * too. - * - * \param rmesa Context pointer - * \param tObj GL texture object whose images are to be posted to - * hardware state. - */ -static void r200SetTexImages( r200ContextPtr rmesa, - struct gl_texture_object *tObj ) -{ - r200TexObjPtr t = (r200TexObjPtr)tObj->DriverData; - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - GLint curOffset, blitWidth; - GLint i, texelBytes; - GLint numLevels; - GLint log2Width, log2Height, log2Depth; - - /* Set the hardware texture format - */ - if ( !t->image_override ) { - if ( VALID_FORMAT( baseImage->TexFormat->MesaFormat ) ) { - const struct tx_table *table = _mesa_little_endian() ? tx_table_le : - tx_table_be; - - t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK | - R200_TXFORMAT_ALPHA_IN_MAP); - t->pp_txfilter &= ~R200_YUV_TO_RGB; - - t->pp_txformat |= table[ baseImage->TexFormat->MesaFormat ].format; - t->pp_txfilter |= table[ baseImage->TexFormat->MesaFormat ].filter; - } - else { - _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__); - return; - } - } - - texelBytes = baseImage->TexFormat->TexelBytes; - - /* Compute which mipmap levels we really want to send to the hardware. - */ - - driCalculateTextureFirstLastLevel( (driTextureObject *) t ); - log2Width = tObj->Image[0][t->base.firstLevel]->WidthLog2; - log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2; - log2Depth = tObj->Image[0][t->base.firstLevel]->DepthLog2; - - numLevels = t->base.lastLevel - t->base.firstLevel + 1; - - assert(numLevels <= RADEON_MAX_TEXTURE_LEVELS); - - /* Calculate mipmap offsets and dimensions for blitting (uploading) - * The idea is that we lay out the mipmap levels within a block of - * memory organized as a rectangle of width BLIT_WIDTH_BYTES. - */ - curOffset = 0; - blitWidth = BLIT_WIDTH_BYTES; - t->tile_bits = 0; - - /* figure out if this texture is suitable for tiling. */ - if (texelBytes) { - if (rmesa->texmicrotile && (tObj->Target != GL_TEXTURE_RECTANGLE_NV) && - /* texrect might be able to use micro tiling too in theory? */ - (baseImage->Height > 1)) { - /* allow 32 (bytes) x 1 mip (which will use two times the space - the non-tiled version would use) max if base texture is large enough */ - if ((numLevels == 1) || - (((baseImage->Width * texelBytes / baseImage->Height) <= 32) && - (baseImage->Width * texelBytes > 64)) || - ((baseImage->Width * texelBytes / baseImage->Height) <= 16)) { - t->tile_bits |= R200_TXO_MICRO_TILE; - } - } - if (tObj->Target != GL_TEXTURE_RECTANGLE_NV) { - /* we can set macro tiling even for small textures, they will be untiled anyway */ - t->tile_bits |= R200_TXO_MACRO_TILE; - } - } - - for (i = 0; i < numLevels; i++) { - const struct gl_texture_image *texImage; - GLuint size; - - texImage = tObj->Image[0][i + t->base.firstLevel]; - if ( !texImage ) - break; - - /* find image size in bytes */ - if (texImage->IsCompressed) { - /* need to calculate the size AFTER padding even though the texture is - submitted without padding. - Only handle pot textures currently - don't know if npot is even possible, - size calculation would certainly need (trivial) adjustments. - Align (and later pad) to 32byte, not sure what that 64byte blit width is - good for? */ - if ((t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) == R200_TXFORMAT_DXT1) { - /* RGB_DXT1/RGBA_DXT1, 8 bytes per block */ - if ((texImage->Width + 3) < 8) /* width one block */ - size = texImage->CompressedSize * 4; - else if ((texImage->Width + 3) < 16) - size = texImage->CompressedSize * 2; - else size = texImage->CompressedSize; - } - else /* DXT3/5, 16 bytes per block */ - if ((texImage->Width + 3) < 8) - size = texImage->CompressedSize * 2; - else size = texImage->CompressedSize; - } - else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) { - size = ((texImage->Width * texelBytes + 63) & ~63) * texImage->Height; - } - else if (t->tile_bits & R200_TXO_MICRO_TILE) { - /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned, - though the actual offset may be different (if texture is less than - 32 bytes width) to the untiled case */ - int w = (texImage->Width * texelBytes * 2 + 31) & ~31; - size = (w * ((texImage->Height + 1) / 2)) * texImage->Depth; - blitWidth = MAX2(texImage->Width, 64 / texelBytes); - } - else { - int w = (texImage->Width * texelBytes + 31) & ~31; - size = w * texImage->Height * texImage->Depth; - blitWidth = MAX2(texImage->Width, 64 / texelBytes); - } - assert(size > 0); - - /* Align to 32-byte offset. It is faster to do this unconditionally - * (no branch penalty). - */ - - curOffset = (curOffset + 0x1f) & ~0x1f; - - if (texelBytes) { - t->image[0][i].x = curOffset; /* fix x and y coords up later together with offset */ - t->image[0][i].y = 0; - t->image[0][i].width = MIN2(size / texelBytes, blitWidth); - t->image[0][i].height = (size / texelBytes) / t->image[0][i].width; - } - else { - t->image[0][i].x = curOffset % BLIT_WIDTH_BYTES; - t->image[0][i].y = curOffset / BLIT_WIDTH_BYTES; - t->image[0][i].width = MIN2(size, BLIT_WIDTH_BYTES); - t->image[0][i].height = size / t->image[0][i].width; - } - -#if 0 - /* for debugging only and only applicable to non-rectangle targets */ - assert(size % t->image[0][i].width == 0); - assert(t->image[0][i].x == 0 - || (size < BLIT_WIDTH_BYTES && t->image[0][i].height == 1)); -#endif - - if (0) - fprintf(stderr, - "level %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n", - i, texImage->Width, texImage->Height, - t->image[0][i].x, t->image[0][i].y, - t->image[0][i].width, t->image[0][i].height, size, curOffset); - - curOffset += size; - - } - - /* Align the total size of texture memory block. - */ - t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK; - - /* Setup remaining cube face blits, if needed */ - if (tObj->Target == GL_TEXTURE_CUBE_MAP) { - const GLuint faceSize = t->base.totalSize; - GLuint face; - /* reuse face 0 x/y/width/height - just update the offset when uploading */ - for (face = 1; face < 6; face++) { - for (i = 0; i < numLevels; i++) { - t->image[face][i].x = t->image[0][i].x; - t->image[face][i].y = t->image[0][i].y; - t->image[face][i].width = t->image[0][i].width; - t->image[face][i].height = t->image[0][i].height; - } - } - t->base.totalSize = 6 * faceSize; /* total texmem needed */ - } - - - /* Hardware state: - */ - t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK; - t->pp_txfilter |= (numLevels - 1) << R200_MAX_MIP_LEVEL_SHIFT; - - t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK | - R200_TXFORMAT_HEIGHT_MASK | - R200_TXFORMAT_CUBIC_MAP_ENABLE | - R200_TXFORMAT_F5_WIDTH_MASK | - R200_TXFORMAT_F5_HEIGHT_MASK); - t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) | - (log2Height << R200_TXFORMAT_HEIGHT_SHIFT)); - - t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK); - if (tObj->Target == GL_TEXTURE_3D) { - t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT); - t->pp_txformat_x |= R200_TEXCOORD_VOLUME; - } - else if (tObj->Target == GL_TEXTURE_CUBE_MAP) { - ASSERT(log2Width == log2Height); - t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) | - (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) | -/* don't think we need this bit, if it exists at all - fglrx does not set it */ - (R200_TXFORMAT_CUBIC_MAP_ENABLE)); - t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV; - t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) | - (log2Height << R200_FACE_HEIGHT_1_SHIFT) | - (log2Width << R200_FACE_WIDTH_2_SHIFT) | - (log2Height << R200_FACE_HEIGHT_2_SHIFT) | - (log2Width << R200_FACE_WIDTH_3_SHIFT) | - (log2Height << R200_FACE_HEIGHT_3_SHIFT) | - (log2Width << R200_FACE_WIDTH_4_SHIFT) | - (log2Height << R200_FACE_HEIGHT_4_SHIFT)); - } - else { - /* If we don't in fact send enough texture coordinates, q will be 1, - * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?) - */ - t->pp_txformat_x |= R200_TEXCOORD_PROJ; - } - - t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) | - ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16)); - - /* Only need to round to nearest 32 for textures, but the blitter - * requires 64-byte aligned pitches, and we may/may not need the - * blitter. NPOT only! - */ - if ( !t->image_override ) { - if (baseImage->IsCompressed) - t->pp_txpitch = (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63); - else - t->pp_txpitch = ((tObj->Image[0][t->base.firstLevel]->Width * texelBytes) + 63) & ~(63); - t->pp_txpitch -= 32; - } - - t->dirty_state = TEX_ALL; - - /* FYI: r200UploadTexImages( rmesa, t ) used to be called here */ -} - - - /* ================================================================ * Texture combine functions */ @@ -981,20 +733,19 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname, { r200ContextPtr rmesa = pDRICtx->driverPrivate; struct gl_texture_object *tObj = - _mesa_lookup_texture(rmesa->glCtx, texname); - r200TexObjPtr t; + _mesa_lookup_texture(rmesa->radeon.glCtx, texname); + radeonTexObjPtr t = radeon_tex_obj(tObj); if (!tObj) return; - t = (r200TexObjPtr) tObj->DriverData; - t->image_override = GL_TRUE; if (!offset) return; - t->pp_txoffset = offset; + t->bo = NULL; + t->override_offset = offset; t->pp_txpitch = pitch - 32; switch (depth) { @@ -1014,6 +765,109 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname, } } +void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv) +{ + struct gl_texture_unit *texUnit; + struct gl_texture_object *texObj; + struct gl_texture_image *texImage; + struct radeon_renderbuffer *rb; + radeon_texture_image *rImage; + radeonContextPtr radeon; + r200ContextPtr rmesa; + struct radeon_framebuffer *rfb; + radeonTexObjPtr t; + uint32_t pitch_val; + + target = GL_TEXTURE_RECTANGLE_ARB; + + radeon = pDRICtx->driverPrivate; + rmesa = pDRICtx->driverPrivate; + + rfb = dPriv->driverPrivate; + texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit]; + texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target); + texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0); + + rImage = get_radeon_texture_image(texImage); + t = radeon_tex_obj(texObj); + if (t == NULL) { + return; + } + + radeon_update_renderbuffers(pDRICtx, dPriv); + /* back & depth buffer are useless free them right away */ + rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = rfb->color_rb[0]; + if (rb->bo == NULL) { + /* Failed to BO for the buffer */ + return; + } + + _mesa_lock_texture(radeon->glCtx, texObj); + if (t->bo) { + radeon_bo_unref(t->bo); + t->bo = NULL; + } + if (rImage->bo) { + radeon_bo_unref(rImage->bo); + rImage->bo = NULL; + } + if (t->mt) { + radeon_miptree_unreference(t->mt); + t->mt = NULL; + } + if (rImage->mt) { + radeon_miptree_unreference(rImage->mt); + rImage->mt = NULL; + } + fprintf(stderr,"settexbuf %d %dx%d@%d\n", rb->pitch, rb->width, rb->height, rb->cpp); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, + rb->width, rb->height, 1, 0, rb->cpp); + texImage->TexFormat = &_mesa_texformat_rgba8888_rev; + rImage->bo = rb->bo; + radeon_bo_ref(rImage->bo); + t->bo = rb->bo; + radeon_bo_ref(t->bo); + t->tile_bits = 0; + t->image_override = GL_TRUE; + t->override_offset = 0; + t->pp_txpitch &= (1 << 13) -1; + pitch_val = rb->pitch; + switch (rb->cpp) { + case 4: + t->pp_txformat = tx_table_le[MESA_FORMAT_ARGB8888].format; + t->pp_txfilter |= tx_table_le[MESA_FORMAT_ARGB8888].filter; + break; + case 3: + default: + t->pp_txformat = tx_table_le[MESA_FORMAT_RGB888].format; + t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB888].filter; + break; + case 2: + t->pp_txformat = tx_table_le[MESA_FORMAT_RGB565].format; + t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB565].filter; + break; + } + t->pp_txsize = ((rb->width - 1) << RADEON_TEX_USIZE_SHIFT) + | ((rb->height - 1) << RADEON_TEX_VSIZE_SHIFT); + t->pp_txformat |= R200_TXFORMAT_NON_POWER2; + t->pp_txpitch = pitch_val; + t->pp_txpitch -= 32; + + t->validated = GL_TRUE; + _mesa_unlock_texture(radeon->glCtx, texObj); + return; +} + #define REF_COLOR 1 #define REF_ALPHA 2 @@ -1207,12 +1061,41 @@ static GLboolean r200UpdateAllTexEnv( GLcontext *ctx ) R200_VOLUME_FILTER_MASK) +static void disable_tex_obj_state( r200ContextPtr rmesa, + int unit ) +{ + + R200_STATECHANGE( rmesa, vtx ); + rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3)); + + if (rmesa->radeon.TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) { + TCL_FALLBACK( rmesa->radeon.glCtx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE); + } + + /* Actually want to keep all units less than max active texture + * enabled, right? Fix this for >2 texunits. + */ + + { + GLuint tmp = rmesa->TexGenEnabled; + + rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit); + rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit); + rmesa->TexGenNeedNormals[unit] = GL_FALSE; + rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit); + + if (tmp != rmesa->TexGenEnabled) { + rmesa->recheck_texgen[unit] = GL_TRUE; + rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX; + } + } +} static void import_tex_obj_state( r200ContextPtr rmesa, int unit, - r200TexObjPtr texobj ) + radeonTexObjPtr texobj ) { /* do not use RADEON_DB_STATE to avoid stale texture caches */ - int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0]; + GLuint *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0]; R200_STATECHANGE( rmesa, tex[unit] ); @@ -1225,36 +1108,21 @@ static void import_tex_obj_state( r200ContextPtr rmesa, cmd[TEX_PP_TXSIZE] = texobj->pp_txsize; /* NPOT only! */ cmd[TEX_PP_TXPITCH] = texobj->pp_txpitch; /* NPOT only! */ cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color; - if (rmesa->r200Screen->drmSupportsFragShader) { - cmd[TEX_PP_TXOFFSET_NEWDRM] = texobj->pp_txoffset; - } - else { - cmd[TEX_PP_TXOFFSET_OLDDRM] = texobj->pp_txoffset; - } - if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) { - int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0]; - GLuint bytesPerFace = texobj->base.totalSize / 6; - ASSERT(texobj->base.totalSize % 6 == 0); + if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) { + GLuint *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0]; R200_STATECHANGE( rmesa, cube[unit] ); cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces; - if (rmesa->r200Screen->drmSupportsFragShader) { + if (rmesa->radeon.radeonScreen->drmSupportsFragShader) { /* that value is submitted twice. could change cube atom to not include that command when new drm is used */ cmd[TEX_PP_CUBIC_FACES] = texobj->pp_cubic_faces; } - cube_cmd[CUBE_PP_CUBIC_OFFSET_F1] = texobj->pp_txoffset + 1 * bytesPerFace; - cube_cmd[CUBE_PP_CUBIC_OFFSET_F2] = texobj->pp_txoffset + 2 * bytesPerFace; - cube_cmd[CUBE_PP_CUBIC_OFFSET_F3] = texobj->pp_txoffset + 3 * bytesPerFace; - cube_cmd[CUBE_PP_CUBIC_OFFSET_F4] = texobj->pp_txoffset + 4 * bytesPerFace; - cube_cmd[CUBE_PP_CUBIC_OFFSET_F5] = texobj->pp_txoffset + 5 * bytesPerFace; } - texobj->dirty_state &= ~(1<<unit); } - static void set_texgen_matrix( r200ContextPtr rmesa, GLuint unit, const GLfloat *s_plane, @@ -1377,7 +1245,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit ) } else { tgcm |= R200_TEXGEN_COMP_T << (unit * 4); } - if (texUnit->TexGenEnabled & R_BIT) { if (texUnit->GenR.Mode != mode) mixed_fallback = GL_TRUE; @@ -1517,52 +1384,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit ) return GL_TRUE; } - -static void disable_tex( GLcontext *ctx, int unit ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if (rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit)) { - /* Texture unit disabled */ - if ( rmesa->state.texture.unit[unit].texobj != NULL ) { - /* The old texture is no longer bound to this texture unit. - * Mark it as such. - */ - - rmesa->state.texture.unit[unit].texobj->base.bound &= ~(1UL << unit); - rmesa->state.texture.unit[unit].texobj = NULL; - } - - R200_STATECHANGE( rmesa, ctx ); - rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~(R200_TEX_0_ENABLE << unit); - - R200_STATECHANGE( rmesa, vtx ); - rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3)); - - if (rmesa->TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) { - TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE); - } - - /* Actually want to keep all units less than max active texture - * enabled, right? Fix this for >2 texunits. - */ - - { - GLuint tmp = rmesa->TexGenEnabled; - - rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit); - rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit); - rmesa->TexGenNeedNormals[unit] = GL_FALSE; - rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit); - - if (tmp != rmesa->TexGenEnabled) { - rmesa->recheck_texgen[unit] = GL_TRUE; - rmesa->NewGLState |= _NEW_TEXTURE_MATRIX; - } - } - } -} - void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); @@ -1579,237 +1400,169 @@ void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d ) } } -static GLboolean enable_tex_2d( GLcontext *ctx, int unit ) +/** + * Compute the cached hardware register values for the given texture object. + * + * \param rmesa Context pointer + * \param t the r300 texture object + */ +static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) { - r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData; - - /* Need to load the 2d images associated with this unit. - */ - if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) { - t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2; - t->base.dirty_images[0] = ~0; + const struct gl_texture_image *firstImage = + t->base.Image[0][t->mt->firstLevel]; + GLint log2Width, log2Height, log2Depth, texelBytes; + + if ( t->bo ) { + return; } - ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D); + log2Width = firstImage->WidthLog2; + log2Height = firstImage->HeightLog2; + log2Depth = firstImage->DepthLog2; + texelBytes = firstImage->TexFormat->TexelBytes; - if ( t->base.dirty_images[0] ) { - R200_FIREVERTICES( rmesa ); - r200SetTexImages( rmesa, tObj ); - r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 ); - if ( !t->base.memBlock && !t->image_override ) - return GL_FALSE; - } - set_re_cntl_d3d( ctx, unit, GL_FALSE ); - - return GL_TRUE; -} - -#if ENABLE_HW_3D_TEXTURE -static GLboolean enable_tex_3d( GLcontext *ctx, int unit ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData; - - /* Need to load the 3d images associated with this unit. - */ - if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) { - t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2; - t->base.dirty_images[0] = ~0; + if (!t->image_override) { + if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) { + const struct tx_table *table = _mesa_little_endian() ? tx_table_le : + tx_table_be; + + t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK | + R200_TXFORMAT_ALPHA_IN_MAP); + t->pp_txfilter &= ~R200_YUV_TO_RGB; + + t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format; + t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter; + } else { + _mesa_problem(NULL, "unexpected texture format in %s", + __FUNCTION__); + return; + } } + + t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK; + t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << R200_MAX_MIP_LEVEL_SHIFT; + + t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK | + R200_TXFORMAT_HEIGHT_MASK | + R200_TXFORMAT_CUBIC_MAP_ENABLE | + R200_TXFORMAT_F5_WIDTH_MASK | + R200_TXFORMAT_F5_HEIGHT_MASK); + t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) | + (log2Height << R200_TXFORMAT_HEIGHT_SHIFT)); + + t->tile_bits = 0; + + t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK); + if (t->base.Target == GL_TEXTURE_3D) { + t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT); + t->pp_txformat_x |= R200_TEXCOORD_VOLUME; - ASSERT(tObj->Target == GL_TEXTURE_3D); - - /* R100 & R200 do not support mipmaps for 3D textures. - */ - if ( (tObj->MinFilter != GL_NEAREST) && (tObj->MinFilter != GL_LINEAR) ) { - return GL_FALSE; } - - if ( t->base.dirty_images[0] ) { - R200_FIREVERTICES( rmesa ); - r200SetTexImages( rmesa, tObj ); - r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 ); - if ( !t->base.memBlock ) - return GL_FALSE; + else if (t->base.Target == GL_TEXTURE_CUBE_MAP) { + ASSERT(log2Width == log2Height); + t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) | + (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) | + /* don't think we need this bit, if it exists at all - fglrx does not set it */ + (R200_TXFORMAT_CUBIC_MAP_ENABLE)); + t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV; + t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) | + (log2Height << R200_FACE_HEIGHT_1_SHIFT) | + (log2Width << R200_FACE_WIDTH_2_SHIFT) | + (log2Height << R200_FACE_HEIGHT_2_SHIFT) | + (log2Width << R200_FACE_WIDTH_3_SHIFT) | + (log2Height << R200_FACE_HEIGHT_3_SHIFT) | + (log2Width << R200_FACE_WIDTH_4_SHIFT) | + (log2Height << R200_FACE_HEIGHT_4_SHIFT)); } - - set_re_cntl_d3d( ctx, unit, GL_TRUE ); - - return GL_TRUE; -} -#endif - -static GLboolean enable_tex_cube( GLcontext *ctx, int unit ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData; - GLuint face; - - /* Need to load the 2d images associated with this unit. - */ - if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) { - t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2; - for (face = 0; face < 6; face++) - t->base.dirty_images[face] = ~0; + else { + /* If we don't in fact send enough texture coordinates, q will be 1, + * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?) + */ + t->pp_txformat_x |= R200_TEXCOORD_PROJ; } - ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP); - - if ( t->base.dirty_images[0] || t->base.dirty_images[1] || - t->base.dirty_images[2] || t->base.dirty_images[3] || - t->base.dirty_images[4] || t->base.dirty_images[5] ) { - /* flush */ - R200_FIREVERTICES( rmesa ); - /* layout memory space, once for all faces */ - r200SetTexImages( rmesa, tObj ); - } + t->pp_txsize = (((firstImage->Width - 1) << R200_PP_TX_WIDTHMASK_SHIFT) + | ((firstImage->Height - 1) << R200_PP_TX_HEIGHTMASK_SHIFT)); - /* upload (per face) */ - for (face = 0; face < 6; face++) { - if (t->base.dirty_images[face]) { - r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, face ); - } - } - - if ( !t->base.memBlock ) { - /* texmem alloc failed, use s/w fallback */ - return GL_FALSE; + if ( !t->image_override ) { + if (firstImage->IsCompressed) + t->pp_txpitch = (firstImage->Width + 63) & ~(63); + else + t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63); + t->pp_txpitch -= 32; } - set_re_cntl_d3d( ctx, unit, GL_TRUE ); - - return GL_TRUE; -} - -static GLboolean enable_tex_rect( GLcontext *ctx, int unit ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData; - - if (!(t->pp_txformat & R200_TXFORMAT_NON_POWER2)) { + if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) { t->pp_txformat |= R200_TXFORMAT_NON_POWER2; - t->base.dirty_images[0] = ~0; } - ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV); - - if ( t->base.dirty_images[0] ) { - R200_FIREVERTICES( rmesa ); - r200SetTexImages( rmesa, tObj ); - r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 ); - if ( !t->base.memBlock && - !t->image_override && - !rmesa->prefer_gart_client_texturing ) - return GL_FALSE; - } - - set_re_cntl_d3d( ctx, unit, GL_FALSE ); - - return GL_TRUE; } - -static GLboolean update_tex_common( GLcontext *ctx, int unit ) +static GLboolean r200_validate_texture(GLcontext *ctx, struct gl_texture_object *texObj, int unit) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData; - - /* Fallback if there's a texture border */ - if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 ) - return GL_FALSE; + radeonTexObj *t = radeon_tex_obj(texObj); - /* Update state if this is a different texture object to last - * time. - */ - if ( rmesa->state.texture.unit[unit].texobj != t ) { - if ( rmesa->state.texture.unit[unit].texobj != NULL ) { - /* The old texture is no longer bound to this texture unit. - * Mark it as such. - */ - - rmesa->state.texture.unit[unit].texobj->base.bound &= - ~(1UL << unit); - } - - rmesa->state.texture.unit[unit].texobj = t; - t->base.bound |= (1UL << unit); - t->dirty_state |= 1<<unit; - driUpdateTextureLRU( (driTextureObject *) t ); /* XXX: should be locked! */ - } - - - /* Newly enabled? - */ - if ( 1|| !(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit))) { - R200_STATECHANGE( rmesa, ctx ); - rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit; - - R200_STATECHANGE( rmesa, vtx ); - rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3)); - rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3); + if (!radeon_validate_texture_miptree(ctx, texObj)) + return GL_FALSE; - rmesa->recheck_texgen[unit] = GL_TRUE; - } + r200_validate_texgen(ctx, unit); + /* Configure the hardware registers (more precisely, the cached version + * of the hardware registers). */ + setup_hardware_state(rmesa, t); + + if (texObj->Target == GL_TEXTURE_RECTANGLE_NV || + texObj->Target == GL_TEXTURE_2D || + texObj->Target == GL_TEXTURE_1D) + set_re_cntl_d3d( ctx, unit, GL_FALSE ); + else + set_re_cntl_d3d( ctx, unit, GL_TRUE ); + R200_STATECHANGE( rmesa, ctx ); + rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit; + + R200_STATECHANGE( rmesa, vtx ); + rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3)); + rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3); - if (t->dirty_state & (1<<unit)) { - import_tex_obj_state( rmesa, unit, t ); - } + rmesa->recheck_texgen[unit] = GL_TRUE; + import_tex_obj_state( rmesa, unit, t ); if (rmesa->recheck_texgen[unit]) { GLboolean fallback = !r200_validate_texgen( ctx, unit ); TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), fallback); rmesa->recheck_texgen[unit] = 0; - rmesa->NewGLState |= _NEW_TEXTURE_MATRIX; + rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX; } - FALLBACK( rmesa, R200_FALLBACK_BORDER_MODE, t->border_fallback ); - return !t->border_fallback; -} + t->validated = GL_TRUE; + FALLBACK( rmesa, RADEON_FALLBACK_BORDER_MODE, t->border_fallback ); + return !t->border_fallback; +} -static GLboolean r200UpdateTextureUnit( GLcontext *ctx, int unit ) +static GLboolean r200UpdateTextureUnit(GLcontext *ctx, int unit) { r200ContextPtr rmesa = R200_CONTEXT(ctx); GLuint unitneeded = rmesa->state.texture.unit[unit].unitneeded; - if ( unitneeded & (TEXTURE_RECT_BIT) ) { - return (enable_tex_rect( ctx, unit ) && - update_tex_common( ctx, unit )); - } - else if ( unitneeded & (TEXTURE_1D_BIT | TEXTURE_2D_BIT) ) { - return (enable_tex_2d( ctx, unit ) && - update_tex_common( ctx, unit )); - } -#if ENABLE_HW_3D_TEXTURE - else if ( unitneeded & (TEXTURE_3D_BIT) ) { - return (enable_tex_3d( ctx, unit ) && - update_tex_common( ctx, unit )); - } -#endif - else if ( unitneeded & (TEXTURE_CUBE_BIT) ) { - return (enable_tex_cube( ctx, unit ) && - update_tex_common( ctx, unit )); - } - else if ( unitneeded ) { - return GL_FALSE; - } - else { - disable_tex( ctx, unit ); - return GL_TRUE; + if (!unitneeded) { + /* disable the unit */ + disable_tex_obj_state(rmesa, unit); + return GL_TRUE; } + + if (!r200_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) { + _mesa_warning(ctx, + "failed to validate texture for unit %d.\n", + unit); + rmesa->state.texture.unit[unit].texobj = NULL; + return GL_FALSE; + } + + rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current); + return GL_TRUE; } @@ -1850,11 +1603,11 @@ void r200UpdateTextureState( GLcontext *ctx ) FALLBACK( rmesa, R200_FALLBACK_TEXTURE, !ok ); - if (rmesa->TclFallback) + if (rmesa->radeon.TclFallback) r200ChooseVertexState( ctx ); - if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) { + if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) { /* * T0 hang workaround ------------- @@ -1867,7 +1620,7 @@ void r200UpdateTextureState( GLcontext *ctx ) R200_STATECHANGE(rmesa, tex[1]); rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_1_ENABLE; if (!(rmesa->hw.cst.cmd[CST_PP_CNTL_X] & R200_PPX_TEX_1_ENABLE)) - rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK; + rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK; rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] |= R200_TXFORMAT_LOOKUP_DISABLE; } else if (!ctx->ATIFragmentShader._Enabled) { diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c index a2561df579..aadd1443ad 100644 --- a/src/mesa/drivers/dri/r200/r200_vertprog.c +++ b/src/mesa/drivers/dri/r200/r200_vertprog.c @@ -1110,9 +1110,9 @@ void r200SetupVertexProg( GLcontext *ctx ) { } /* could optimize setting up vertex progs away for non-tcl hw */ fallback = !(vp->native && r200VertexProgUpdateParams(ctx, vp) && - rmesa->r200Screen->drmSupportsVertexProgram); + rmesa->radeon.radeonScreen->drmSupportsVertexProgram); TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, fallback); - if (rmesa->TclFallback) return; + if (rmesa->radeon.TclFallback) return; R200_STATECHANGE( rmesa, vap ); /* FIXME: fglrx sets R200_VAP_SINGLE_BUF_STATE_ENABLE too. Do we need it? |