void brw_emit_anyprim_setup( struct brw_sf_compile *c ) { struct brw_compile *p = &c->func; struct brw_context *brw = p->brw; struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0); struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0); struct brw_reg primmask; int jmp; struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); c->nr_verts = 3; alloc_regs(c); primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD); brw_MOV(p, primmask, brw_imm_ud(1)); brw_SHL(p, primmask, primmask, payload_prim); brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) | (1<<_3DPRIM_TRISTRIP) | (1<<_3DPRIM_TRIFAN) | (1<<_3DPRIM_TRISTRIP_REVERSE) | (1<<_3DPRIM_POLYGON) | (1<<_3DPRIM_RECTLIST) | (1<<_3DPRIM_TRIFAN_NOSTIPPLE))); brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z); jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store; brw_emit_tri_setup(c, false); brw_land_fwd_jump(p, jmp); brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) | (1<<_3DPRIM_LINESTRIP) | (1<<_3DPRIM_LINELOOP) | (1<<_3DPRIM_LINESTRIP_CONT) | (1<<_3DPRIM_LINESTRIP_BF) | (1<<_3DPRIM_LINESTRIP_CONT_BF))); brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z); jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store; brw_emit_line_setup(c, false); brw_land_fwd_jump(p, jmp); brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE)); brw_inst_set_cond_modifier(brw, brw_last_inst, BRW_CONDITIONAL_Z); jmp = brw_JMPI(p, brw_imm_d(0), BRW_PREDICATE_NORMAL) - p->store; brw_emit_point_sprite_setup(c, false); brw_land_fwd_jump(p, jmp); brw_emit_point_setup( c, false ); }
static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) { struct brw_sf_compile c; const GLuint *program; void *mem_ctx; GLuint program_size; GLuint i; memset(&c, 0, sizeof(c)); mem_ctx = ralloc_context(NULL); /* Begin the compilation: */ brw_init_compile(brw, &c.func, mem_ctx); c.key = *key; c.vue_map = brw->vue_map_geom_out; if (c.key.do_point_coord) { /* * gl_PointCoord is a FS instead of VS builtin variable, thus it's * not included in c.vue_map generated in VS stage. Here we add * it manually to let SF shader generate the needed interpolation * coefficient for FS shader. */ c.vue_map.varying_to_slot[BRW_VARYING_SLOT_PNTC] = c.vue_map.num_slots; c.vue_map.slot_to_varying[c.vue_map.num_slots++] = BRW_VARYING_SLOT_PNTC; } c.urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET; c.nr_attr_regs = (c.vue_map.num_slots + 1)/2 - c.urb_entry_read_offset; c.nr_setup_regs = c.nr_attr_regs; c.prog_data.urb_read_length = c.nr_attr_regs; c.prog_data.urb_entry_size = c.nr_setup_regs * 2; c.has_flat_shading = brw_any_flat_varyings(&key->interpolation_mode); /* Which primitive? Or all three? */ switch (key->primitive) { case SF_TRIANGLES: c.nr_verts = 3; brw_emit_tri_setup( &c, true ); break; case SF_LINES: c.nr_verts = 2; brw_emit_line_setup( &c, true ); break; case SF_POINTS: c.nr_verts = 1; if (key->do_point_sprite) brw_emit_point_sprite_setup( &c, true ); else brw_emit_point_setup( &c, true ); break; case SF_UNFILLED_TRIS: c.nr_verts = 3; brw_emit_anyprim_setup( &c ); break; default: assert(0); return; } /* get the program */ program = brw_get_program(&c.func, &program_size); if (unlikely(INTEL_DEBUG & DEBUG_SF)) { printf("sf:\n"); for (i = 0; i < program_size / sizeof(struct brw_instruction); i++) brw_disasm(stdout, &((struct brw_instruction *)program)[i], brw->gen); printf("\n"); } brw_upload_cache(&brw->cache, BRW_SF_PROG, &c.key, sizeof(c.key), program, program_size, &c.prog_data, sizeof(c.prog_data), &brw->sf.prog_offset, &brw->sf.prog_data); ralloc_free(mem_ctx); }
static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) { GLcontext *ctx = &brw->intel.ctx; struct brw_sf_compile c; const GLuint *program; GLuint program_size; GLuint i, idx; memset(&c, 0, sizeof(c)); /* Begin the compilation: */ brw_init_compile(brw, &c.func); c.key = *key; c.nr_attrs = brw_count_bits(c.key.attrs); c.nr_attr_regs = (c.nr_attrs+1)/2; c.nr_setup_attrs = brw_count_bits(c.key.attrs & DO_SETUP_BITS); c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; c.prog_data.urb_entry_size = c.nr_setup_regs * 2; /* Construct map from attribute number to position in the vertex. */ for (i = idx = 0; i < VERT_RESULT_MAX; i++) if (c.key.attrs & (1<<i)) { c.attr_to_idx[i] = idx; c.idx_to_attr[idx] = i; if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { c.point_attrs[i].CoordReplace = ctx->Point.CoordReplace[i - VERT_RESULT_TEX0]; } else { c.point_attrs[i].CoordReplace = GL_FALSE; } idx++; } /* Which primitive? Or all three? */ switch (key->primitive) { case SF_TRIANGLES: c.nr_verts = 3; brw_emit_tri_setup( &c, GL_TRUE ); break; case SF_LINES: c.nr_verts = 2; brw_emit_line_setup( &c, GL_TRUE ); break; case SF_POINTS: c.nr_verts = 1; if (key->do_point_sprite) brw_emit_point_sprite_setup( &c, GL_TRUE ); else brw_emit_point_setup( &c, GL_TRUE ); break; case SF_UNFILLED_TRIS: c.nr_verts = 3; brw_emit_anyprim_setup( &c ); break; default: assert(0); return; } /* get the program */ program = brw_get_program(&c.func, &program_size); /* Upload */ dri_bo_unreference(brw->sf.prog_bo); brw->sf.prog_bo = brw_upload_cache( &brw->cache, BRW_SF_PROG, &c.key, sizeof(c.key), NULL, 0, program, program_size, &c.prog_data, &brw->sf.prog_data ); }
static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) { struct brw_sf_compile c; const unsigned *program; unsigned program_size; memset(&c, 0, sizeof(c)); /* Begin the compilation: */ brw_init_compile(&c.func); c.key = *key; c.nr_attrs = c.key.vp_output_count; c.nr_attr_regs = (c.nr_attrs+1)/2; c.nr_setup_attrs = c.key.fp_input_count + 1; /* +1 for position */ c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; c.prog_data.urb_entry_size = c.nr_setup_regs * 2; /* Which primitive? Or all three? */ switch (key->primitive) { case SF_TRIANGLES: c.nr_verts = 3; brw_emit_tri_setup( &c ); break; case SF_LINES: c.nr_verts = 2; brw_emit_line_setup( &c ); break; case SF_POINTS: c.nr_verts = 1; brw_emit_point_setup( &c ); break; case SF_UNFILLED_TRIS: default: assert(0); return; } /* get the program */ program = brw_get_program(&c.func, &program_size); /* Upload */ brw->sf.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_SF_PROG], &c.key, sizeof(c.key), program, program_size, &c.prog_data, &brw->sf.prog_data ); }
static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) { struct brw_sf_compile c; const GLuint *program; void *mem_ctx; GLuint program_size; memset(&c, 0, sizeof(c)); mem_ctx = ralloc_context(NULL); /* Begin the compilation: */ brw_init_codegen(brw->intelScreen->devinfo, &c.func, mem_ctx); c.key = *key; c.vue_map = brw->vue_map_geom_out; if (c.key.do_point_coord) { /* * gl_PointCoord is a FS instead of VS builtin variable, thus it's * not included in c.vue_map generated in VS stage. Here we add * it manually to let SF shader generate the needed interpolation * coefficient for FS shader. */ c.vue_map.varying_to_slot[BRW_VARYING_SLOT_PNTC] = c.vue_map.num_slots; c.vue_map.slot_to_varying[c.vue_map.num_slots++] = BRW_VARYING_SLOT_PNTC; } c.urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET; c.nr_attr_regs = (c.vue_map.num_slots + 1)/2 - c.urb_entry_read_offset; c.nr_setup_regs = c.nr_attr_regs; c.prog_data.urb_read_length = c.nr_attr_regs; c.prog_data.urb_entry_size = c.nr_setup_regs * 2; c.has_flat_shading = brw_any_flat_varyings(&key->interpolation_mode); /* Which primitive? Or all three? */ switch (key->primitive) { case SF_TRIANGLES: c.nr_verts = 3; brw_emit_tri_setup( &c, true ); break; case SF_LINES: c.nr_verts = 2; brw_emit_line_setup( &c, true ); break; case SF_POINTS: c.nr_verts = 1; if (key->do_point_sprite) brw_emit_point_sprite_setup( &c, true ); else brw_emit_point_setup( &c, true ); break; case SF_UNFILLED_TRIS: c.nr_verts = 3; brw_emit_anyprim_setup( &c ); break; default: unreachable("not reached"); } /* FINISHME: SF programs use calculated jumps (i.e., JMPI with a register * source). Compacting would be difficult. */ /* brw_compact_instructions(&c.func, 0, 0, NULL); */ /* get the program */ program = brw_get_program(&c.func, &program_size); if (unlikely(INTEL_DEBUG & DEBUG_SF)) { fprintf(stderr, "sf:\n"); brw_disassemble(brw->intelScreen->devinfo, c.func.store, 0, program_size, stderr); fprintf(stderr, "\n"); } brw_upload_cache(&brw->cache, BRW_CACHE_SF_PROG, &c.key, sizeof(c.key), program, program_size, &c.prog_data, sizeof(c.prog_data), &brw->sf.prog_offset, &brw->sf.prog_data); ralloc_free(mem_ctx); }
void brw_emit_anyprim_setup( struct brw_sf_compile *c ) { struct brw_compile *p = &c->func; struct brw_reg ip = brw_ip_reg(); struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0); struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0); struct brw_reg primmask; struct brw_instruction *jmp; struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); GLuint saveflag; c->nr_verts = 3; alloc_regs(c); primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD); brw_MOV(p, primmask, brw_imm_ud(1)); brw_SHL(p, primmask, primmask, payload_prim); brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) | (1<<_3DPRIM_TRISTRIP) | (1<<_3DPRIM_TRIFAN) | (1<<_3DPRIM_TRISTRIP_REVERSE) | (1<<_3DPRIM_POLYGON) | (1<<_3DPRIM_RECTLIST) | (1<<_3DPRIM_TRIFAN_NOSTIPPLE))); jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); { saveflag = p->flag_value; brw_push_insn_state(p); brw_emit_tri_setup( c, GL_FALSE ); brw_pop_insn_state(p); p->flag_value = saveflag; /* note - thread killed in subroutine, so must * restore the flag which is changed when building * the subroutine. fix #13240 */ } brw_land_fwd_jump(p, jmp); brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) | (1<<_3DPRIM_LINESTRIP) | (1<<_3DPRIM_LINELOOP) | (1<<_3DPRIM_LINESTRIP_CONT) | (1<<_3DPRIM_LINESTRIP_BF) | (1<<_3DPRIM_LINESTRIP_CONT_BF))); jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); { saveflag = p->flag_value; brw_push_insn_state(p); brw_emit_line_setup( c, GL_FALSE ); brw_pop_insn_state(p); p->flag_value = saveflag; /* note - thread killed in subroutine */ } brw_land_fwd_jump(p, jmp); brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE)); jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); { saveflag = p->flag_value; brw_push_insn_state(p); brw_emit_point_sprite_setup( c, GL_FALSE ); brw_pop_insn_state(p); p->flag_value = saveflag; } brw_land_fwd_jump(p, jmp); brw_emit_point_setup( c, GL_FALSE ); }