diff options
Diffstat (limited to 'src/mesa/drivers/dri/r600')
-rw-r--r-- | src/mesa/drivers/dri/r600/Makefile | 2 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r600_cmdbuf.c | 216 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r600_cmdbuf.h | 16 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r600_context.c | 18 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_assembler.c | 540 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_assembler.h | 67 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_chip.c | 40 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_clear.c | 4 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_fragprog.c | 144 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_fragprog.h | 2 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_shaderinst.h | 7 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_state.c | 13 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r600/r700_vertprog.c | 25 | ||||
l--------- | src/mesa/drivers/dri/r600/radeon_bo.c | 1 | ||||
l--------- | src/mesa/drivers/dri/r600/radeon_bo_int_drm.h | 1 | ||||
l--------- | src/mesa/drivers/dri/r600/radeon_cs.c | 1 | ||||
l--------- | src/mesa/drivers/dri/r600/radeon_cs_int_drm.h | 1 |
17 files changed, 888 insertions, 210 deletions
diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile index 9b7c42042e..26f47b7268 100644 --- a/src/mesa/drivers/dri/r600/Makefile +++ b/src/mesa/drivers/dri/r600/Makefile @@ -14,7 +14,7 @@ EGL_SOURCES = server/radeon_egl.c endif ifeq ($(RADEON_LDFLAGS),) -CS_SOURCES = radeon_cs_space_drm.c +CS_SOURCES = radeon_cs_space_drm.c radeon_bo.c radeon_cs.c endif COMMON_SOURCES = \ diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index d27a3245a3..370bb04f93 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -52,29 +52,49 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "radeon_reg.h" +#ifdef HAVE_LIBDRM_RADEON +#include "radeon_cs_int.h" +#else +#include "radeon_cs_int_drm.h" +#endif +struct r600_cs_manager_legacy +{ + struct radeon_cs_manager base; + struct radeon_context *ctx; + /* hack for scratch stuff */ + uint32_t pending_age; + uint32_t pending_count; +}; + +struct r600_cs_reloc_legacy { + struct radeon_cs_reloc base; + uint32_t cindices; + uint32_t *indices; + uint32_t *reloc_indices; +}; -static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm, - uint32_t ndw) +static struct radeon_cs_int *r600_cs_create(struct radeon_cs_manager *csm, + uint32_t ndw) { - struct radeon_cs *cs; + struct radeon_cs_int *csi; - cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs)); - if (cs == NULL) { + csi = (struct radeon_cs_int*)calloc(1, sizeof(struct radeon_cs_int)); + if (csi == NULL) { return NULL; } - cs->csm = csm; - cs->ndw = (ndw + 0x3FF) & (~0x3FF); - cs->packets = (uint32_t*)malloc(4*cs->ndw); - if (cs->packets == NULL) { - free(cs); + csi->csm = csm; + csi->ndw = (ndw + 0x3FF) & (~0x3FF); + csi->packets = (uint32_t*)malloc(4*csi->ndw); + if (csi->packets == NULL) { + free(csi); return NULL; } - cs->relocs_total_size = 0; - return cs; + csi->relocs_total_size = 0; + return csi; } -static int r600_cs_write_reloc(struct radeon_cs *cs, +static int r600_cs_write_reloc(struct radeon_cs_int *csi, struct radeon_bo *bo, uint32_t read_domain, uint32_t write_domain, @@ -83,7 +103,7 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, struct r600_cs_reloc_legacy *relocs; int i; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; /* check domains */ if ((read_domain && write_domain) || (!read_domain && !write_domain)) { /* in one CS a bo can only be in read or write domain but not @@ -98,7 +118,7 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, return -EINVAL; } /* check if bo is already referenced */ - for(i = 0; i < cs->crelocs; i++) { + for(i = 0; i < csi->crelocs; i++) { uint32_t *indices; uint32_t *reloc_indices; @@ -129,109 +149,108 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, } relocs[i].indices = indices; relocs[i].reloc_indices = reloc_indices; - relocs[i].indices[relocs[i].cindices - 1] = cs->cdw; - relocs[i].reloc_indices[relocs[i].cindices - 1] = cs->cdw; - cs->section_cdw += 2; - cs->cdw += 2; + relocs[i].indices[relocs[i].cindices - 1] = csi->cdw; + relocs[i].reloc_indices[relocs[i].cindices - 1] = csi->cdw; + csi->section_cdw += 2; + csi->cdw += 2; return 0; } } /* add bo to reloc */ relocs = (struct r600_cs_reloc_legacy*) - realloc(cs->relocs, - sizeof(struct r600_cs_reloc_legacy) * (cs->crelocs + 1)); + realloc(csi->relocs, + sizeof(struct r600_cs_reloc_legacy) * (csi->crelocs + 1)); if (relocs == NULL) { return -ENOMEM; } - cs->relocs = relocs; - relocs[cs->crelocs].base.bo = bo; - relocs[cs->crelocs].base.read_domain = read_domain; - relocs[cs->crelocs].base.write_domain = write_domain; - relocs[cs->crelocs].base.flags = flags; - relocs[cs->crelocs].indices = (uint32_t*)malloc(4); - relocs[cs->crelocs].reloc_indices = (uint32_t*)malloc(4); - if ( (relocs[cs->crelocs].indices == NULL) || (relocs[cs->crelocs].reloc_indices == NULL) ) + csi->relocs = relocs; + relocs[csi->crelocs].base.bo = bo; + relocs[csi->crelocs].base.read_domain = read_domain; + relocs[csi->crelocs].base.write_domain = write_domain; + relocs[csi->crelocs].base.flags = flags; + relocs[csi->crelocs].indices = (uint32_t*)malloc(4); + relocs[csi->crelocs].reloc_indices = (uint32_t*)malloc(4); + if ( (relocs[csi->crelocs].indices == NULL) || (relocs[csi->crelocs].reloc_indices == NULL) ) { return -ENOMEM; } - relocs[cs->crelocs].indices[0] = cs->cdw; - relocs[cs->crelocs].reloc_indices[0] = cs->cdw; - cs->section_cdw += 2; - cs->cdw += 2; - relocs[cs->crelocs].cindices = 1; - cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo); - cs->crelocs++; + relocs[csi->crelocs].indices[0] = csi->cdw; + relocs[csi->crelocs].reloc_indices[0] = csi->cdw; + csi->section_cdw += 2; + csi->cdw += 2; + relocs[csi->crelocs].cindices = 1; + csi->relocs_total_size += radeon_bo_legacy_relocs_size(bo); + csi->crelocs++; radeon_bo_ref(bo); return 0; } -static int r600_cs_begin(struct radeon_cs *cs, +static int r600_cs_begin(struct radeon_cs_int *csi, uint32_t ndw, const char *file, const char *func, int line) { - if (cs->section) { + if (csi->section_ndw) { fprintf(stderr, "CS already in a section(%s,%s,%d)\n", - cs->section_file, cs->section_func, cs->section_line); + csi->section_file, csi->section_func, csi->section_line); fprintf(stderr, "CS can't start section(%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 1; - cs->section_ndw = ndw; - cs->section_cdw = 0; - cs->section_file = file; - cs->section_func = func; - cs->section_line = line; + csi->section_ndw = ndw; + csi->section_cdw = 0; + csi->section_file = file; + csi->section_func = func; + csi->section_line = line; - if (cs->cdw + ndw > cs->ndw) { + if (csi->cdw + ndw > csi->ndw) { uint32_t tmp, *ptr; int num = (ndw > 0x400) ? ndw : 0x400; - tmp = (cs->cdw + num + 0x3FF) & (~0x3FF); - ptr = (uint32_t*)realloc(cs->packets, 4 * tmp); + tmp = (csi->cdw + num + 0x3FF) & (~0x3FF); + ptr = (uint32_t*)realloc(csi->packets, 4 * tmp); if (ptr == NULL) { return -ENOMEM; } - cs->packets = ptr; - cs->ndw = tmp; + csi->packets = ptr; + csi->ndw = tmp; } return 0; } -static int r600_cs_end(struct radeon_cs *cs, +static int r600_cs_end(struct radeon_cs_int *csi, const char *file, const char *func, int line) { - if (!cs->section) { + if (!csi->section_ndw) { fprintf(stderr, "CS no section to end at (%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 0; - if ( cs->section_ndw != cs->section_cdw ) { + if ( csi->section_ndw != csi->section_cdw ) { fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n", - cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw); - fprintf(stderr, "cs->section_ndw = %d, cs->cdw = %d, cs->section_cdw = %d \n", - cs->section_ndw, cs->cdw, cs->section_cdw); + csi->section_file, csi->section_func, csi->section_line, csi->section_ndw, csi->section_cdw); + fprintf(stderr, "csi->section_ndw = %d, csi->cdw = %d, csi->section_cdw = %d \n", + csi->section_ndw, csi->cdw, csi->section_cdw); fprintf(stderr, "CS section end at (%s,%s,%d)\n", file, func, line); return -EPIPE; } + csi->section_ndw = 0; - if (cs->cdw > cs->ndw) { + if (csi->cdw > csi->ndw) { fprintf(stderr, "CS section overflow at (%s,%s,%d) cdw %d ndw %d\n", - cs->section_file, cs->section_func, cs->section_line,cs->cdw,cs->ndw); + csi->section_file, csi->section_func, csi->section_line,csi->cdw,csi->ndw); fprintf(stderr, "CS section end at (%s,%s,%d)\n", file, func, line); assert(0); @@ -240,20 +259,20 @@ static int r600_cs_end(struct radeon_cs *cs, return 0; } -static int r600_cs_process_relocs(struct radeon_cs *cs, +static int r600_cs_process_relocs(struct radeon_cs_int *csi, uint32_t * reloc_chunk, uint32_t * length_dw_reloc_chunk) { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct r600_cs_reloc_legacy *relocs; int i, j, r; uint32_t offset_dw = 0; - csm = (struct r600_cs_manager_legacy*)cs->csm; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + csm = (struct r600_cs_manager_legacy*)csi->csm; + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; restart: - for (i = 0; i < cs->crelocs; i++) { + for (i = 0; i < csi->crelocs; i++) { uint32_t soffset, eoffset; r = radeon_bo_legacy_validate(relocs[i].base.bo, @@ -269,9 +288,9 @@ restart: for (j = 0; j < relocs[i].cindices; j++) { /* pkt3 nop header in ib chunk */ - cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000; + csi->packets[relocs[i].reloc_indices[j]] = 0xC0001000; /* reloc index in ib chunk */ - cs->packets[relocs[i].reloc_indices[j] + 1] = offset_dw; + csi->packets[relocs[i].reloc_indices[j] + 1] = offset_dw; } /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */ @@ -286,14 +305,14 @@ restart: return 0; } -static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */ +static int r600_cs_set_age(struct radeon_cs_int *csi) /* -------------- */ { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct r600_cs_reloc_legacy *relocs; int i; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; - for (i = 0; i < cs->crelocs; i++) { + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; + for (i = 0; i < csi->crelocs; i++) { radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age); radeon_bo_unref(relocs[i].base.bo); } @@ -301,21 +320,21 @@ static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */ } #if 0 -static void dump_cmdbuf(struct radeon_cs *cs) +static void dump_cmdbuf(struct radeon_cs_int *csi) { int i; fprintf(stderr,"--start--\n"); - for (i = 0; i < cs->cdw; i++){ - fprintf(stderr,"0x%08x\n", cs->packets[i]); + for (i = 0; i < csi->cdw; i++){ + fprintf(stderr,"0x%08x\n", csi->packets[i]); } fprintf(stderr,"--end--\n"); } #endif -static int r600_cs_emit(struct radeon_cs *cs) +static int r600_cs_emit(struct radeon_cs_int *csi) { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct drm_radeon_cs cs_cmd; struct drm_radeon_cs_chunk cs_chunk[2]; uint32_t length_dw_reloc_chunk; @@ -329,9 +348,9 @@ static int r600_cs_emit(struct radeon_cs *cs) csm->pending_count = 1; - reloc_chunk = (uint32_t*)calloc(1, cs->crelocs * 4 * 4); + reloc_chunk = (uint32_t*)calloc(1, csi->crelocs * 4 * 4); - r = r600_cs_process_relocs(cs, reloc_chunk, &length_dw_reloc_chunk); + r = r600_cs_process_relocs(csi, reloc_chunk, &length_dw_reloc_chunk); if (r) { free(reloc_chunk); return 0; @@ -339,8 +358,8 @@ static int r600_cs_emit(struct radeon_cs *cs) /* raw ib chunk */ cs_chunk[0].chunk_id = RADEON_CHUNK_ID_IB; - cs_chunk[0].length_dw = cs->cdw; - cs_chunk[0].chunk_data = (unsigned long)(cs->packets); + cs_chunk[0].length_dw = csi->cdw; + cs_chunk[0].chunk_data = (unsigned long)(csi->packets); /* reloc chaunk */ cs_chunk[1].chunk_id = RADEON_CHUNK_ID_RELOCS; @@ -358,7 +377,7 @@ static int r600_cs_emit(struct radeon_cs *cs) do { - r = drmCommandWriteRead(cs->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd)); + r = drmCommandWriteRead(csi->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd)); retry++; } while (r == -EAGAIN && retry < 1000); @@ -369,11 +388,11 @@ static int r600_cs_emit(struct radeon_cs *cs) csm->pending_age = cs_cmd.cs_id; - r600_cs_set_age(cs); + r600_cs_set_age(csi); - cs->csm->read_used = 0; - cs->csm->vram_write_used = 0; - cs->csm->gart_write_used = 0; + csi->csm->read_used = 0; + csi->csm->vram_write_used = 0; + csi->csm->gart_write_used = 0; free(reloc_chunk); @@ -393,35 +412,34 @@ static void inline r600_cs_free_reloc(void *relocs_p, int crelocs) } } -static int r600_cs_destroy(struct radeon_cs *cs) +static int r600_cs_destroy(struct radeon_cs_int *csi) { - r600_cs_free_reloc(cs->relocs, cs->crelocs); - free(cs->relocs); - free(cs->packets); - free(cs); + r600_cs_free_reloc(csi->relocs, csi->crelocs); + free(csi->relocs); + free(csi->packets); + free(csi); return 0; } -static int r600_cs_erase(struct radeon_cs *cs) +static int r600_cs_erase(struct radeon_cs_int *csi) { - r600_cs_free_reloc(cs->relocs, cs->crelocs); - free(cs->relocs); - cs->relocs_total_size = 0; - cs->relocs = NULL; - cs->crelocs = 0; - cs->cdw = 0; - cs->section = 0; + r600_cs_free_reloc(csi->relocs, csi->crelocs); + free(csi->relocs); + csi->relocs_total_size = 0; + csi->relocs = NULL; + csi->crelocs = 0; + csi->cdw = 0; return 0; } -static int r600_cs_need_flush(struct radeon_cs *cs) +static int r600_cs_need_flush(struct radeon_cs_int *csi) { /* this function used to flush when the BO usage got to * a certain size, now the higher levels handle this better */ return 0; } -static void r600_cs_print(struct radeon_cs *cs, FILE *file) +static void r600_cs_print(struct radeon_cs_int *csi, FILE *file) { } diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h index eba43d37b6..dff0009699 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.h +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h @@ -118,22 +118,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R600_IT_SET_CTL_CONST 0x00006F00 #define R600_IT_SURFACE_BASE_UPDATE 0x00007300 -struct r600_cs_manager_legacy -{ - struct radeon_cs_manager base; - struct radeon_context *ctx; - /* hack for scratch stuff */ - uint32_t pending_age; - uint32_t pending_count; -}; - -struct r600_cs_reloc_legacy { - struct radeon_cs_reloc base; - uint32_t cindices; - uint32_t *indices; - uint32_t *reloc_indices; -}; - struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx); /** diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 25314eff56..45bbc3c071 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -74,7 +74,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -//#define R600_ENABLE_GLSL_TEST 1 +#define R600_ENABLE_GLSL_TEST 1 #define need_GL_VERSION_2_0 #define need_GL_ARB_occlusion_query @@ -317,20 +317,8 @@ static void r600InitGLExtensions(GLcontext *ctx) #ifdef R600_ENABLE_GLSL_TEST driInitExtensions(ctx, gl_20_extension, GL_TRUE); - //_mesa_enable_2_0_extensions(ctx); - //1.5 - ctx->Extensions.ARB_occlusion_query = GL_TRUE; - ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; - ctx->Extensions.EXT_shadow_funcs = GL_TRUE; - //2.0 - ctx->Extensions.ARB_draw_buffers = GL_TRUE; - ctx->Extensions.ARB_point_sprite = GL_TRUE; - ctx->Extensions.ARB_shader_objects = GL_TRUE; - ctx->Extensions.ARB_vertex_shader = GL_TRUE; - ctx->Extensions.ARB_fragment_shader = GL_TRUE; - ctx->Extensions.EXT_blend_equation_separate = GL_TRUE; - ctx->Extensions.ATI_separate_stencil = GL_TRUE; - + _mesa_enable_2_0_extensions(ctx); + /* glsl compiler has problem if this is not GL_TRUE */ ctx->Shader.EmitCondCodes = GL_TRUE; #endif /* R600_ENABLE_GLSL_TEST */ diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index aed84fc3bd..51692a11ff 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -32,6 +32,7 @@ #include "main/mtypes.h" #include "main/imports.h" +#include "shader/prog_parameter.h" #include "radeon_debug.h" #include "r600_context.h" @@ -41,6 +42,39 @@ #define USE_CF_FOR_CONTINUE_BREAK 1 #define USE_CF_FOR_POP_AFTER 1 +struct prog_instruction noise1_insts[12] = { + {OPCODE_BGNSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 2, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 4, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 585, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 8, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_SGT , {{0, 0, 585, 0, 0, 0}, {8, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 1, 1, 0, 8, 1672, 0}, 1, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_IF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 7, 0, 0}, 0, 0, 0, 1, 0, 0, 0, 15, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1755, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDIF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0} +}; +float noise1_const[2][4] = { + {0.300000f, 0.900000f, 0.500000f, 0.300000f} +}; + +COMPILED_SUB noise1_presub = { + &(noise1_insts[0]), + 12, + 2, + 1, + 0, + &(noise1_const[0]), + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + {0,0,0}, + 0 +}; + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -330,14 +364,14 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size) return(format); } -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3) { - if(pAsm->D.dst.op3) + if(nIsOp3 > 0) { return 3; } - switch (pAsm->D.dst.opcode) + switch (opcode) { case SQ_OP2_INST_ADD: case SQ_OP2_INST_KILLE: @@ -378,7 +412,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) return 1; default: radeon_error( - "Need instruction operand number for %x.\n", pAsm->D.dst.opcode); + "Need instruction operand number for %x.\n", opcode); }; return 3; @@ -500,12 +534,20 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->unCFflags = 0; + pAsm->presubs = NULL; + pAsm->unPresubArraySize = 0; + pAsm->unNumPresub = 0; + pAsm->unCurNumILInsts = 0; + + pAsm->unVetTexBits = 0; + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) { - if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) ) + if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) || + (OPCODE_DDX==Opcode) || (OPCODE_DDY==Opcode) ) { return GL_TRUE; } @@ -1220,7 +1262,15 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } pAsm->S[fld].src.rtype = SRC_REG_CONSTANT; - pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + if(pILInst->SrcReg[src].Index < 0) + { + WARN_ONCE("Negative register offsets not supported yet!\n"); + pAsm->S[fld].src.reg = 0; + } + else + { + pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + } break; case PROGRAM_INPUT: setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); @@ -1373,43 +1423,65 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; break; case PROGRAM_INPUT: - switch (pILInst->SrcReg[0].Index) + if(SPT_VP == pAsm->currentShaderType) { - case FRAG_ATTRIB_WPOS: - case FRAG_ATTRIB_COL0: - case FRAG_ATTRIB_COL1: - case FRAG_ATTRIB_FOGC: - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - bValidTexCoord = GL_TRUE; + switch (pILInst->SrcReg[0].Index) + { + case VERT_ATTRIB_TEX0: + case VERT_ATTRIB_TEX1: + case VERT_ATTRIB_TEX2: + case VERT_ATTRIB_TEX3: + case VERT_ATTRIB_TEX4: + case VERT_ATTRIB_TEX5: + case VERT_ATTRIB_TEX6: + case VERT_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->ucVP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + } + } + else + { + switch (pILInst->SrcReg[0].Index) + { + case FRAG_ATTRIB_WPOS: + case FRAG_ATTRIB_COL0: + case FRAG_ATTRIB_COL1: + case FRAG_ATTRIB_FOGC: + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + case FRAG_ATTRIB_FACE: + fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); + break; + case FRAG_ATTRIB_PNTC: + fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); + break; + } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; pAsm->S[0].src.reg = pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; pAsm->S[0].src.rtype = SRC_REG_INPUT; - break; - case FRAG_ATTRIB_FACE: - fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); - break; - case FRAG_ATTRIB_PNTC: - fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); - break; - } - - if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || - (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) - { - bValidTexCoord = GL_TRUE; - pAsm->S[0].src.reg = - pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; - pAsm->S[0].src.rtype = SRC_REG_INPUT; + } } - break; + break; } } @@ -1454,8 +1526,17 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word0.f.tex_inst = pAsm->D.dst.opcode; tex_instruction_ptr->m_Word0.f.bc_frac_mode = 0x0; tex_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0; + tex_instruction_ptr->m_Word0.f.alt_const = 0; - tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + if(SPT_VP == pAsm->currentShaderType) + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg + VERT_ATTRIB_MAX; + pAsm->unVetTexBits |= 1 << texture_unit_source->reg; + } + else + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + } tex_instruction_ptr->m_Word1.f.lod_bias = 0x0; if (normalized) { @@ -1474,7 +1555,6 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word2.f.offset_x = 0x0; tex_instruction_ptr->m_Word2.f.offset_y = 0x0; tex_instruction_ptr->m_Word2.f.offset_z = 0x0; - tex_instruction_ptr->m_Word2.f.sampler_id = texture_unit_source->reg; // dst @@ -2010,7 +2090,7 @@ GLboolean check_scalar(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2099,7 +2179,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2180,7 +2260,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) int current_source_index; GLuint contiguous_slots_needed; - GLuint uNumSrc = r700GetNumOperands(pAsm); + GLuint uNumSrc = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); //GLuint channel_swizzle, j; //GLuint chan_counter[4] = {0, 0, 0, 0}; //PVSSRC * pSource[3]; @@ -4292,13 +4372,20 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) } - if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB) - { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; - } - else + switch(pAsm->pILInst[pAsm->uiCurInst].Opcode) { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; + case OPCODE_DDX: + /* will these need WQM(1) on CF inst ? */ + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_H; + break; + case OPCODE_DDY: + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_V; + break; + case OPCODE_TXB: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; + break; + default: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; } pAsm->is_tex = GL_TRUE; @@ -4968,7 +5055,7 @@ void add_return_inst(r700_AssemblerBase *pAsm) pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; } -GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift) { /* Put in sub */ if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) @@ -4983,7 +5070,7 @@ GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) pAsm->unSubArraySize += 10; } - pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex; + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex + uiIL_Shift; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; @@ -5074,9 +5161,13 @@ GLboolean assemble_RET(r700_AssemblerBase *pAsm) GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLint nILindex, + GLuint uiIL_Shift, GLuint uiNumberInsts, - struct prog_instruction *pILInst) + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc) { + GLint uiIL_Offset; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; if(GL_FALSE == add_cf_instruction(pAsm) ) @@ -5109,8 +5200,12 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, pAsm->unCallerArraySize += 10; } - pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = nILindex; - pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + uiIL_Offset = nILindex + uiIL_Shift; + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = uiIL_Offset; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->callers[pAsm->unCallerArrayPointer].finale_cf_ptr = NULL; + pAsm->callers[pAsm->unCallerArrayPointer].prelude_cf_ptr = NULL; pAsm->unCallerArrayPointer++; @@ -5120,7 +5215,7 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLboolean bRet; for(j=0; j<pAsm->unSubArrayPointer; j++) { - if(nILindex == pAsm->subs[j].subIL_Offset) + if(uiIL_Offset == pAsm->subs[j].subIL_Offset) { /* compiled before */ max = pAsm->subs[j].unStackDepthMax @@ -5138,7 +5233,7 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; unSubID = pAsm->unSubArrayPointer; - bRet = AssembleInstr(nILindex, uiNumberInsts, pILInst, pAsm); + bRet = AssembleInstr(nILindex, uiIL_Shift, uiNumberInsts, pILInst, pAsm); if(GL_TRUE == bRet) { @@ -5148,6 +5243,8 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, { pAsm->CALLSTACK[pAsm->CALLSP].max = max; } + + pAsm->subs[unSubID].pPresubDesc = pPresubDesc; } return bRet; @@ -5313,6 +5410,7 @@ GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) } GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) @@ -5468,6 +5566,26 @@ GLboolean AssembleInstr(GLuint uiFirstInst, case OPCODE_MUL: if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) return GL_FALSE; + break; + + case OPCODE_NOISE1: + { + callPreSub(pR700AsmCode, + GLSL_NOISE1, + &noise1_presub, + pILInst->DstReg.Index + pR700AsmCode->starting_temp_register_number, + 1); + radeon_error("noise1: not yet supported shader instruction\n"); + }; + break; + case OPCODE_NOISE2: + radeon_error("noise2: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE3: + radeon_error("noise3: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE4: + radeon_error("noise4: not yet supported shader instruction\n"); break; case OPCODE_POW: @@ -5580,7 +5698,8 @@ GLboolean AssembleInstr(GLuint uiFirstInst, } } break; - + case OPCODE_DDX: + case OPCODE_DDY: case OPCODE_TEX: case OPCODE_TXB: case OPCODE_TXP: @@ -5653,7 +5772,7 @@ GLboolean AssembleInstr(GLuint uiFirstInst, break; case OPCODE_BGNSUB: - if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i) ) + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i, uiIL_Shift) ) { return GL_FALSE; } @@ -5668,9 +5787,11 @@ GLboolean AssembleInstr(GLuint uiFirstInst, case OPCODE_CAL: if( GL_FALSE == assemble_CAL(pR700AsmCode, - pILInst[i].BranchTarget, + pILInst[i].BranchTarget, + uiIL_Shift, uiNumberInsts, - pILInst) ) + pILInst, + NULL) ) { return GL_FALSE; } @@ -5707,7 +5828,7 @@ GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) return GL_TRUE; } -GLboolean RelocProgram(r700_AssemblerBase * pAsm) +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg) { GLuint i; GLuint unCFoffset; @@ -5717,6 +5838,12 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) R700ShaderInstruction * pInst; R700ControlFlowGenericClause * pCFInst; + R700ControlFlowALUClause * pCF_ALU; + R700ALUInstruction * pALU; + GLuint unConstOffset = 0; + GLuint unRegOffset; + GLuint unMinRegIndex; + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; /* remove flags init if they are not used */ @@ -5762,6 +5889,11 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) unCFoffset = plstCFmain->uNumOfNode; + if(NULL != pILProg->Parameters) + { + unConstOffset = pILProg->Parameters->NumParameters; + } + /* Reloc subs */ for(i=0; i<pAsm->unSubArrayPointer; i++) { @@ -5799,6 +5931,84 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) pInst = pInst->pNextInst; }; + if(NULL != pAsm->subs[i].pPresubDesc) + { + GLuint uNumSrc; + + unMinRegIndex = pAsm->subs[i].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[i].pPresubDesc->maxStartReg; + unConstOffset += pAsm->subs[i].pPresubDesc->unConstantsStart; + + pInst = plstCFsub->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + + if(pALU->m_Word0.f.src0_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src0_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src0_sel += unConstOffset; + } + + if( ((pALU->m_Word1.val >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT) & 0x0000001F) + >= SQ_OP3_INST_MUL_LIT ) + { /* op3 : 3 srcs */ + if(pALU->m_Word1_OP3.f.src2_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word1_OP3.f.src2_sel = pALU->m_Word1_OP3.f.src2_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word1_OP3.f.src2_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word1_OP3.f.src2_sel += unConstOffset; + } + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + else + { + if(pAsm->bR6xx) + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f6.alu_inst, 0); + } + else + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f.alu_inst, 0); + } + if(2 == uNumSrc) + { /* 2 srcs */ + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + } + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + pInst = pInst->pNextInst; + }; + } + /* Put sub into main */ plstCFmain->pTail->pNextInst = plstCFsub->pHead; plstCFmain->pTail = plstCFsub->pTail; @@ -5812,11 +6022,216 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) { pAsm->callers[i].cf_ptr->m_Word0.f.addr = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + + if(NULL != pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc) + { + unMinRegIndex = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->maxStartReg; + + if(NULL != pAsm->callers[i].prelude_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].prelude_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + if(NULL != pAsm->callers[i].finale_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].finale_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + } } return GL_TRUE; } +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc) +{ + /* save assemble context */ + GLuint starting_temp_register_number_save; + GLuint number_used_registers_save; + GLuint uFirstHelpReg_save; + GLuint uHelpReg_save; + GLuint uiCurInst_save; + struct prog_instruction *pILInst_save; + PRESUB_DESC * pPresubDesc; + GLboolean bRet; + int i; + + R700ControlFlowGenericClause* prelude_cf_ptr = NULL; + + /* copy srcs to presub inputs */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + for(i=0; i<uNumValidSrc; i++) + { + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pCompiledSub->srcRegIndex[i]; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 1; + pAsm->D.dst.writew = 1; + + if( GL_FALSE == assemble_src(pAsm, i, 0) ) + { + return GL_FALSE; + } + + next_ins(pAsm); + } + if(uNumValidSrc > 0) + { + prelude_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + /* browse thro existing presubs. */ + for(i=0; i<pAsm->unNumPresub; i++) + { + if(pAsm->presubs[i].sptSigniture == scriptSigniture) + { + break; + } + } + + if(i == pAsm->unNumPresub) + { /* not loaded yet */ + /* save assemble context */ + number_used_registers_save = pAsm->number_used_registers; + uFirstHelpReg_save = pAsm->uFirstHelpReg; + uHelpReg_save = pAsm->uHelpReg; + starting_temp_register_number_save = pAsm->starting_temp_register_number; + pILInst_save = pAsm->pILInst; + uiCurInst_save = pAsm->uiCurInst; + + /* alloc in presub */ + if( (pAsm->unNumPresub + 1) > pAsm->unPresubArraySize ) + { + pAsm->presubs = (PRESUB_DESC*)_mesa_realloc( (void *)pAsm->presubs, + sizeof(PRESUB_DESC) * pAsm->unPresubArraySize, + sizeof(PRESUB_DESC) * (pAsm->unPresubArraySize + 4) ); + if(NULL == pAsm->presubs) + { + radeon_error("No memeory to allocate built in shader function description structures. \n"); + return GL_FALSE; + } + pAsm->unPresubArraySize += 4; + } + + pPresubDesc = &(pAsm->presubs[i]); + pPresubDesc->sptSigniture = scriptSigniture; + + /* constants offsets need to be final resolved at reloc. */ + if(0 == pAsm->unNumPresub) + { + pPresubDesc->unConstantsStart = 0; + } + else + { + pPresubDesc->unConstantsStart = pAsm->presubs[i-1].unConstantsStart + + pAsm->presubs[i-1].pCompiledSub->NumParameters; + } + + pPresubDesc->pCompiledSub = pCompiledSub; + + pPresubDesc->subIL_Shift = pAsm->unCurNumILInsts; + pPresubDesc->maxStartReg = uFirstHelpReg_save; + pAsm->unCurNumILInsts += pCompiledSub->NumInstructions; + + pAsm->unNumPresub++; + + /* setup new assemble context */ + pAsm->starting_temp_register_number = 0; + pAsm->number_used_registers = pCompiledSub->NumTemporaries; + pAsm->uFirstHelpReg = pAsm->number_used_registers; + pAsm->uHelpReg = pAsm->uFirstHelpReg; + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + + + pPresubDesc->number_used_registers = pAsm->number_used_registers; + + /* restore assemble context */ + pAsm->number_used_registers = number_used_registers_save; + pAsm->uFirstHelpReg = uFirstHelpReg_save; + pAsm->uHelpReg = uHelpReg_save; + pAsm->starting_temp_register_number = starting_temp_register_number_save; + pAsm->pILInst = pILInst_save; + pAsm->uiCurInst = uiCurInst_save; + } + else + { /* was loaded */ + pPresubDesc = &(pAsm->presubs[i]); + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + } + + if(GL_FALSE == bRet) + { + radeon_error("Shader presub assemble failed. \n"); + } + else + { + /* copy presub output to real dst */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pCompiledSub->dstRegIndex; + pAsm->S[0].src.swizzlex = pCompiledSub->outputSwizzleX; + pAsm->S[0].src.swizzley = pCompiledSub->outputSwizzleY; + pAsm->S[0].src.swizzlez = pCompiledSub->outputSwizzleZ; + pAsm->S[0].src.swizzlew = pCompiledSub->outputSwizzleW; + + next_ins(pAsm); + + pAsm->callers[pAsm->unCallerArrayPointer - 1].finale_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->callers[pAsm->unCallerArrayPointer - 1].prelude_cf_ptr = prelude_cf_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + if( (pPresubDesc->number_used_registers + pAsm->uFirstHelpReg) > pAsm->number_used_registers ) + { + pAsm->number_used_registers = pPresubDesc->number_used_registers + pAsm->uFirstHelpReg; + } + if(pAsm->uFirstHelpReg > pPresubDesc->maxStartReg) + { + pPresubDesc->maxStartReg = pAsm->uFirstHelpReg; + } + + return bRet; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -6174,6 +6589,11 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) FREE(pR700AsmCode->callers); } + if(NULL != pR700AsmCode->presubs) + { + FREE(pR700AsmCode->presubs); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index 6dc44017eb..dbd9860f7d 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -34,6 +34,45 @@ #include "r700_shaderinst.h" #include "r700_shader.h" +typedef enum LOADABLE_SCRIPT_SIGNITURE +{ + GLSL_NOISE1 = 0x10000001, + GLSL_NOISE2 = 0x10000002, + GLSL_NOISE3 = 0x10000003, + GLSL_NOISE4 = 0x10000004 +}LOADABLE_SCRIPT_SIGNITURE; + +typedef struct COMPILED_SUB +{ + struct prog_instruction *Instructions; + GLuint NumInstructions; + GLuint NumTemporaries; + GLuint NumParameters; + GLuint MinRegIndex; + GLfloat (*ParameterValues)[4]; + GLbyte outputSwizzleX; + GLbyte outputSwizzleY; + GLbyte outputSwizzleZ; + GLbyte outputSwizzleW; + GLshort srcRegIndex[3]; + GLushort dstRegIndex; +}COMPILED_SUB; + +typedef struct PRESUB_DESCtag +{ + LOADABLE_SCRIPT_SIGNITURE sptSigniture; + GLint subIL_Shift; + struct prog_src_register InReg[3]; + struct prog_dst_register OutReg; + + GLushort maxStartReg; + GLushort number_used_registers; + + GLuint unConstantsStart; + + COMPILED_SUB * pCompiledSub; +} PRESUB_DESC; + typedef enum SHADER_PIPE_TYPE { SPT_VP = 0, @@ -296,6 +335,7 @@ typedef struct SUB_OFFSET GLint subIL_Offset; GLuint unCFoffset; GLuint unStackDepthMax; + PRESUB_DESC * pPresubDesc; TypedShaderList lstCFInstructions_local; } SUB_OFFSET; @@ -304,6 +344,9 @@ typedef struct CALLER_POINTER GLint subIL_Offset; GLint subDescIndex; R700ControlFlowGenericClause* cf_ptr; + + R700ControlFlowGenericClause* prelude_cf_ptr; + R700ControlFlowGenericClause* finale_cf_ptr; } CALLER_POINTER; #define SQ_MAX_CALL_DEPTH 0x00000020 @@ -437,6 +480,13 @@ typedef struct r700_AssemblerBase GLuint unCFflags; + PRESUB_DESC * presubs; + GLuint unPresubArraySize; + GLuint unNumPresub; + GLuint unCurNumILInsts; + + GLuint unVetTexBits; + } r700_AssemblerBase; //Internal use @@ -458,7 +508,7 @@ BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ; GLboolean is_reduction_opcode(PVSDWORD * dest); GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size); -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm); +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3); GLboolean IsTex(gl_inst_opcode Opcode); GLboolean IsAlu(gl_inst_opcode Opcode); @@ -585,13 +635,15 @@ GLboolean assemble_BRK(r700_AssemblerBase *pAsm); GLboolean assemble_COND(r700_AssemblerBase *pAsm); GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); -GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex); +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift); GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); GLboolean assemble_RET(r700_AssemblerBase *pAsm); GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLint nILindex, + GLuint uiIL_Offest, GLuint uiNumberInsts, - struct prog_instruction *pILInst); + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc); GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, @@ -602,16 +654,23 @@ GLboolean Process_Export(r700_AssemblerBase* pAsm, GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select); +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + /* struct prog_instruction ** pILInstParent, */ + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc); //Interface GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); -GLboolean RelocProgram(r700_AssemblerBase * pAsm); +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg); GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index d702740014..c124e02184 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -45,6 +45,9 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + struct r700_vertex_program *vp = context->selected_vp; + struct radeon_bo *bo = NULL; unsigned int i; BATCH_LOCALS(&context->radeon); @@ -52,7 +55,7 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; uint32_t offset; if (t) { @@ -71,7 +74,16 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) BEGIN_BATCH_NO_AUTOSTATE(9 + 4); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - R600_OUT_BATCH(i * 7); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i + VERT_ATTRIB_MAX + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + } + else + { + R600_OUT_BATCH(i * 7); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2); @@ -95,21 +107,35 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) } } +#define SAMPLER_STRIDE 3 + static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); unsigned int i; + + struct r700_vertex_program *vp = context->selected_vp; + BATCH_LOCALS(&context->radeon); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; if (t) { BEGIN_BATCH_NO_AUTOSTATE(5); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3)); - R600_OUT_BATCH(i * 3); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i+SQ_TEX_SAMPLER_VS_OFFSET) * SAMPLER_STRIDE); //work 1 + } + else + { + R600_OUT_BATCH(i * 3); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2); @@ -1163,7 +1189,11 @@ static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom) count += 3; if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) { - for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) { + /* targets are enabled in r700SetRenderTarget but state + size is calculated before that. Until MRT's are done + hardcode target0 as enabled. */ + count += 3; + for (ui = 1; ui < R700_MAX_RENDER_TARGETS; ui++) { if (r700->render_target[ui].enabled) count += 3; } diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index c6546ab00c..526d3843d1 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -57,6 +57,10 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x\n", __func__, mask); + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + context->radeon.front_buffer_dirty = GL_TRUE; + } + if( GL_TRUE == r700ClearFast(context, mask) ) { return; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index 8eb439a951..ce2d9fdf79 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -34,6 +34,7 @@ #include "main/imports.h" #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" +#include "shader/program.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -42,6 +43,54 @@ #include "r700_debug.h" +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog) +{ + static const gl_state_index winstate[STATE_LENGTH] + = { STATE_INTERNAL, STATE_FB_SIZE, 0, 0, 0}; + struct prog_instruction *newInst, *inst; + GLint win_size; /* state reference */ + GLuint wpos_temp; /* temp register */ + int i, j; + + /* PARAM win_size = STATE_FB_SIZE */ + win_size = _mesa_add_state_reference(fprog->Base.Parameters, winstate); + + wpos_temp = fprog->Base.NumTemporaries++; + + /* scan program where WPOS is used and replace with wpos_temp */ + inst = fprog->Base.Instructions; + for (i = 0; i < fprog->Base.NumInstructions; i++) { + for (j=0; j < 3; j++) { + if(inst->SrcReg[j].File == PROGRAM_INPUT && + inst->SrcReg[j].Index == FRAG_ATTRIB_WPOS) { + inst->SrcReg[j].File = PROGRAM_TEMPORARY; + inst->SrcReg[j].Index = wpos_temp; + } + } + inst++; + } + + _mesa_insert_instructions(&(fprog->Base), 0, 1); + + newInst = fprog->Base.Instructions; + /* invert wpos.y + * wpos_temp.xyzw = wpos.x-yzw + winsize.0y00 */ + newInst[0].Opcode = OPCODE_ADD; + newInst[0].DstReg.File = PROGRAM_TEMPORARY; + newInst[0].DstReg.Index = wpos_temp; + newInst[0].DstReg.WriteMask = WRITEMASK_XYZW; + + newInst[0].SrcReg[0].File = PROGRAM_INPUT; + newInst[0].SrcReg[0].Index = FRAG_ATTRIB_WPOS; + newInst[0].SrcReg[0].Swizzle = SWIZZLE_XYZW; + newInst[0].SrcReg[0].Negate = NEGATE_Y; + + newInst[0].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[0].SrcReg[1].Index = win_size; + newInst[0].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ZERO); + +} + //TODO : Validate FP input with VP output. void Map_Fragment_Program(r700_AssemblerBase *pAsm, struct gl_fragment_program *mesa_fp, @@ -155,6 +204,12 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE] = pAsm->number_used_registers++; } + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC] = pAsm->number_used_registers++; + } + /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -312,7 +367,13 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, //Init_Program Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) ); - Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); + + if(mesa_fp->Base.InputsRead & FRAG_BIT_WPOS) + { + insert_wpos_code(ctx, mesa_fp); + } + + Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) ) { @@ -325,7 +386,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, { fp->r700AsmCode.SamplerUnits[i] = fp->mesa_program.Base.SamplerUnits[i]; } + + fp->r700AsmCode.unCurNumILInsts = mesa_fp->Base.NumInstructions; + if( GL_FALSE == AssembleInstr(0, + 0, mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) @@ -338,7 +403,7 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } - if( GL_FALSE == RelocProgram(&(fp->r700AsmCode)) ) + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode), &(mesa_fp->Base)) ) { return GL_FALSE; } @@ -408,6 +473,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unsigned int unNumOfReg; unsigned int unBit; GLuint exportCount; + GLboolean point_sprite = GL_FALSE; if(GL_FALSE == fp->loaded) { @@ -474,6 +540,36 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); } + /* see if we need any point_sprite replacements */ + for (i = VERT_RESULT_TEX0; i<= VERT_RESULT_TEX7; i++) + { + if(ctx->Point.CoordReplace[i - VERT_RESULT_TEX0] == GL_TRUE) + point_sprite = GL_TRUE; + } + + if ((mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) || point_sprite) + { + /* for FRAG_ATTRIB_PNTC we need to increase num_interp */ + if(mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) + { + ui++; + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask); + } + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_S, PNT_SPRITE_OVRD_X_shift, PNT_SPRITE_OVRD_X_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_T, PNT_SPRITE_OVRD_Y_shift, PNT_SPRITE_OVRD_Y_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_0, PNT_SPRITE_OVRD_Z_shift, PNT_SPRITE_OVRD_Z_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_1, PNT_SPRITE_OVRD_W_shift, PNT_SPRITE_OVRD_W_mask); + if(ctx->Point.SpriteOrigin == GL_LOWER_LEFT) + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + else + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + } + else + { + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + } + ui = (unNumOfReg < ui) ? ui : unNumOfReg; @@ -494,6 +590,10 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) struct r700_vertex_program_cont *vpc = (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + + for(ui = 0; ui < R700_MAX_SHADER_EXPORTS; ui++) + r700->SPI_PS_INPUT_CNTL[ui].u32All = 0; + unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) { @@ -556,6 +656,11 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, SEMANTIC_shift, SEMANTIC_mask); CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + /* ARB_point_sprite */ + if(ctx->Point.CoordReplace[i] == GL_TRUE) + { + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } } } @@ -571,6 +676,22 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) else CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } + + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) { @@ -620,6 +741,25 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } else r700->ps.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->ps.num_consts; + for(ui=0; ui<pAsm->unNumPresub; ui++) + { + pCompiledSub = pAsm->presubs[ui].pCompiledSub; + + r700->ps.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->ps.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->ps.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->ps.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->ps.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index e562bfa478..39c59c9201 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -48,6 +48,8 @@ struct r700_fragment_program }; /* Internal */ +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog); + void Map_Fragment_Program(r700_AssemblerBase *pAsm, struct gl_fragment_program *mesa_fp, GLcontext *ctx); diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.h b/src/mesa/drivers/dri/r600/r700_shaderinst.h index 2829cca0a3..cdb9a570f7 100644 --- a/src/mesa/drivers/dri/r600/r700_shaderinst.h +++ b/src/mesa/drivers/dri/r600/r700_shaderinst.h @@ -42,6 +42,13 @@ #define SQ_FETCH_RESOURCE_VS_OFFSET 0x000000a0 #define SQ_FETCH_RESOURCE_VS_COUNT 0x000000b0 +//richard dec.10 glsl +#define SQ_TEX_SAMPLER_PS_OFFSET 0x00000000 +#define SQ_TEX_SAMPLER_PS_COUNT 0x00000012 +#define SQ_TEX_SAMPLER_VS_OFFSET 0x00000012 +#define SQ_TEX_SAMPLER_VS_COUNT 0x00000012 +//------------------- + #define SHADERINST_TYPEMASK_CF 0x10 #define SHADERINST_TYPEMASK_ALU 0x20 #define SHADERINST_TYPEMASK_TEX 0x40 diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index d7420678ff..16b05d5cd9 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -705,6 +705,10 @@ static void r700UpdateCulling(GLcontext * ctx) CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); /* default: ccw */ break; } + + /* Winding is inverted when rendering to FBO */ + if (ctx->DrawBuffer && ctx->DrawBuffer->Name) + r700->PA_SU_SC_MODE_CNTL.u32All ^= FACE_bit; } static void r700UpdateLineStipple(GLcontext * ctx) @@ -1227,13 +1231,8 @@ static void r700UpdatePolygonMode(GLcontext * ctx) /* Handle GL_CW (clock wise and GL_CCW (counter clock wise) * correctly by selecting the correct front and back face */ - if (ctx->Polygon.FrontFace == GL_CCW) { - f = ctx->Polygon.FrontMode; - b = ctx->Polygon.BackMode; - } else { - f = ctx->Polygon.BackMode; - b = ctx->Polygon.FrontMode; - } + f = ctx->Polygon.FrontMode; + b = ctx->Polygon.BackMode; /* Enable polygon mode */ SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DUAL_MODE, POLY_MODE_shift, POLY_MODE_mask); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 759b74dc7e..90fac078ff 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -341,7 +341,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, { vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i]; } + + vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions; + if(GL_FALSE == AssembleInstr(0, + 0, vp->mesa_program->Base.NumInstructions, &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) @@ -354,7 +358,7 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } - if( GL_FALSE == RelocProgram(&(vp->r700AsmCode)) ) + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) ) { return GL_FALSE; } @@ -671,5 +675,24 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) } else r700->vs.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->vs.num_consts; + for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++) + { + pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub; + + r700->vs.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/radeon_bo.c b/src/mesa/drivers/dri/r600/radeon_bo.c new file mode 120000 index 0000000000..9448ffee54 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_bo.c @@ -0,0 +1 @@ +../radeon/radeon_bo.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h b/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h new file mode 120000 index 0000000000..029450928b --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_bo_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_cs.c b/src/mesa/drivers/dri/r600/radeon_cs.c new file mode 120000 index 0000000000..66b7ad1eb0 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_cs.c @@ -0,0 +1 @@ +../radeon/radeon_cs.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h b/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h new file mode 120000 index 0000000000..462f5245d0 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_cs_int_drm.h
\ No newline at end of file |