diff options
Diffstat (limited to 'src/gallium/drivers/r300/r300_emit.c')
| -rw-r--r-- | src/gallium/drivers/r300/r300_emit.c | 478 |
1 files changed, 343 insertions, 135 deletions
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 01bac5f759..e0c38a06e4 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -24,6 +24,9 @@ #include "r300_emit.h" +#include "r300_fs.h" +#include "r300_vs.h" + void r300_emit_blend_state(struct r300_context* r300, struct r300_blend_state* blend) { @@ -56,6 +59,36 @@ void r300_emit_blend_color_state(struct r300_context* r300, } } +void r300_emit_clip_state(struct r300_context* r300, + struct pipe_clip_state* clip) +{ + int i; + struct r300_screen* r300screen = r300_screen(r300->context.screen); + CS_LOCALS(r300); + + if (r300screen->caps->has_tcl) { + BEGIN_CS(5 + (6 * 4)); + OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, + (r300screen->caps->is_r500 ? + R500_PVS_UCP_START : R300_PVS_UCP_START)); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, 6 * 4); + for (i = 0; i < 6; i++) { + OUT_CS_32F(clip->ucp[i][0]); + OUT_CS_32F(clip->ucp[i][1]); + OUT_CS_32F(clip->ucp[i][2]); + OUT_CS_32F(clip->ucp[i][3]); + } + OUT_CS_REG(R300_VAP_CLIP_CNTL, ((1 << clip->nr) - 1) | + R300_PS_UCP_MODE_CLIP_AS_TRIFAN); + END_CS; + } else { + BEGIN_CS(2); + OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE); + END_CS; + } + +} + void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa) { @@ -79,73 +112,158 @@ void r300_emit_dsa_state(struct r300_context* r300, END_CS; } -void r300_emit_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs) +static const float * get_shader_constant( + struct r300_context * r300, + struct rc_constant * constant, + struct r300_constant_buffer * externals) +{ + static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 }; + switch(constant->Type) { + case RC_CONSTANT_EXTERNAL: + return externals->constants[constant->u.External]; + + case RC_CONSTANT_IMMEDIATE: + return constant->u.Immediate; + + default: + debug_printf("r300: Implementation error: Unhandled constant type %i\n", + constant->Type); + return zero; + } +} + +/* Convert a normal single-precision float into the 7.16 format + * used by the R300 fragment shader. + */ +static uint32_t pack_float24(float f) { + union { + float fl; + uint32_t u; + } u; + float mantissa; + int exponent; + uint32_t float24 = 0; + + if (f == 0.0) + return 0; + + u.fl = f; + + mantissa = frexpf(f, &exponent); + + /* Handle -ve */ + if (mantissa < 0) { + float24 |= (1 << 23); + mantissa = mantissa * -1.0; + } + /* Handle exponent, bias of 63 */ + exponent += 62; + float24 |= (exponent << 16); + /* Kill 7 LSB of mantissa */ + float24 |= (u.u & 0x7FFFFF) >> 7; + + return float24; +} + +void r300_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals) +{ + struct r300_fragment_program_code * code = &generic_code->code.r300; + struct rc_constant_list * constants = &generic_code->constants; int i; CS_LOCALS(r300); - BEGIN_CS(22); - - OUT_CS_REG(R300_US_CONFIG, fs->indirections); - OUT_CS_REG(R300_US_PIXSIZE, fs->shader.stack_size); - /* XXX figure out exactly how big the sizes are on this reg */ - OUT_CS_REG(R300_US_CODE_OFFSET, 0x40); - /* XXX figure these ones out a bit better kthnx */ - OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_3, 0x40 | R300_RGBA_OUT); - - for (i = 0; i < fs->alu_instruction_count; i++) { - OUT_CS_REG(R300_US_ALU_RGB_INST_0 + (4 * i), - fs->instructions[i].alu_rgb_inst); - OUT_CS_REG(R300_US_ALU_RGB_ADDR_0 + (4 * i), - fs->instructions[i].alu_rgb_addr); - OUT_CS_REG(R300_US_ALU_ALPHA_INST_0 + (4 * i), - fs->instructions[i].alu_alpha_inst); - OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0 + (4 * i), - fs->instructions[i].alu_alpha_addr); + BEGIN_CS(15 + + code->alu.length * 4 + + (code->tex.length ? (1 + code->tex.length) : 0) + + (constants->Count ? (1 + constants->Count * 4) : 0)); + + OUT_CS_REG(R300_US_CONFIG, code->config); + OUT_CS_REG(R300_US_PIXSIZE, code->pixsize); + OUT_CS_REG(R300_US_CODE_OFFSET, code->code_offset); + + OUT_CS_REG_SEQ(R300_US_CODE_ADDR_0, 4); + for(i = 0; i < 4; ++i) + OUT_CS(code->code_addr[i]); + + OUT_CS_REG_SEQ(R300_US_ALU_RGB_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].rgb_inst); + + OUT_CS_REG_SEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].rgb_addr); + + OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].alpha_inst); + + OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].alpha_addr); + + if (code->tex.length) { + OUT_CS_REG_SEQ(R300_US_TEX_INST_0, code->tex.length); + for(i = 0; i < code->tex.length; ++i) + OUT_CS(code->tex.inst[i]); + } + + if (constants->Count) { + OUT_CS_ONE_REG(R300_PFS_PARAM_0_X, constants->Count * 4); + for(i = 0; i < constants->Count; ++i) { + const float * data = get_shader_constant(r300, &constants->Constants[i], externals); + OUT_CS(pack_float24(data[0])); + OUT_CS(pack_float24(data[1])); + OUT_CS(pack_float24(data[2])); + OUT_CS(pack_float24(data[3])); + } } END_CS; } -void r500_emit_fragment_shader(struct r300_context* r300, - struct r500_fragment_shader* fs) +void r500_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals) { + struct r500_fragment_program_code * code = &generic_code->code.r500; + struct rc_constant_list * constants = &generic_code->constants; int i; - struct r300_constant_buffer* constants = - &r300->shader_constants[PIPE_SHADER_FRAGMENT]; CS_LOCALS(r300); - BEGIN_CS(9 + (fs->instruction_count * 6) + (constants->count ? 3 : 0) + - (constants->count * 4)); - OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_CS_REG(R500_US_PIXSIZE, fs->shader.stack_size); - OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) | - R500_US_CODE_END_ADDR(fs->instruction_count)); + BEGIN_CS(13 + + ((code->inst_end + 1) * 6) + + (constants->Count ? (3 + (constants->Count * 4)) : 0)); + OUT_CS_REG(R500_US_CONFIG, 0); + OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx); + OUT_CS_REG(R500_US_CODE_RANGE, + R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end)); + OUT_CS_REG(R500_US_CODE_OFFSET, 0); + OUT_CS_REG(R500_US_CODE_ADDR, + R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(code->inst_end)); OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR); - OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, fs->instruction_count * 6); - for (i = 0; i < fs->instruction_count; i++) { - OUT_CS(fs->instructions[i].inst0); - OUT_CS(fs->instructions[i].inst1); - OUT_CS(fs->instructions[i].inst2); - OUT_CS(fs->instructions[i].inst3); - OUT_CS(fs->instructions[i].inst4); - OUT_CS(fs->instructions[i].inst5); - } - - if (constants->count) { - OUT_CS_REG(R500_GA_US_VECTOR_INDEX, - R500_GA_US_VECTOR_INDEX_TYPE_CONST); - OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->count * 4); - for (i = 0; i < constants->count; i++) { - OUT_CS_32F(constants->constants[i][0]); - OUT_CS_32F(constants->constants[i][1]); - OUT_CS_32F(constants->constants[i][2]); - OUT_CS_32F(constants->constants[i][3]); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, (code->inst_end + 1) * 6); + for (i = 0; i <= code->inst_end; i++) { + OUT_CS(code->inst[i].inst0); + OUT_CS(code->inst[i].inst1); + OUT_CS(code->inst[i].inst2); + OUT_CS(code->inst[i].inst3); + OUT_CS(code->inst[i].inst4); + OUT_CS(code->inst[i].inst5); + } + + if (constants->Count) { + OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4); + for (i = 0; i < constants->Count; i++) { + const float * data = get_shader_constant(r300, &constants->Constants[i], externals); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); } } @@ -163,6 +281,7 @@ void r300_emit_fb_state(struct r300_context* r300, BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4); for (i = 0; i < fb->nr_cbufs; i++) { tex = (struct r300_texture*)fb->cbufs[i]->texture; + assert(tex && tex->buffer && "cbuf is marked, but NULL!"); pixpitch = tex->stride / tex->tex.block.size; OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); @@ -177,7 +296,8 @@ void r300_emit_fb_state(struct r300_context* r300, if (fb->zsbuf) { tex = (struct r300_texture*)fb->zsbuf->texture; - pixpitch = (tex->stride / tex->tex.block.size); + assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); + pixpitch = tex->stride / tex->tex.block.size; OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); @@ -234,7 +354,7 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->ip[i]); - debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); + /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */ } OUT_CS_REG_SEQ(R300_RS_COUNT, 2); @@ -248,27 +368,15 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->inst[i]); - debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); + /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */ } - debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, - rs->inst_count); + /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, + * rs->inst_count); */ END_CS; } -void r300_emit_sampler(struct r300_context* r300, - struct r300_sampler_state* sampler, unsigned offset) -{ - CS_LOCALS(r300); - - BEGIN_CS(6); - OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); - OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); - OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); - END_CS; -} - void r300_emit_scissor_state(struct r300_context* r300, struct r300_scissor_state* scissor) { @@ -282,11 +390,17 @@ void r300_emit_scissor_state(struct r300_context* r300, } void r300_emit_texture(struct r300_context* r300, - struct r300_texture* tex, unsigned offset) + struct r300_sampler_state* sampler, + struct r300_texture* tex, + unsigned offset) { CS_LOCALS(r300); - BEGIN_CS(10); + BEGIN_CS(16); + OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); + OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); + OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); + OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0); OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1); OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2); @@ -296,6 +410,30 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } +void r300_emit_vertex_buffer(struct r300_context* r300) +{ + CS_LOCALS(r300); + + debug_printf("r300: Preparing vertex buffer %p for render, " + "vertex size %d\n", r300->vbo, + r300->vertex_info.vinfo.size); + /* Set the pointer to our vertex buffer. The emitted values are this: + * PACKET3 [3D_LOAD_VBPNTR] + * COUNT [1] + * FORMAT [size | stride << 8] + * OFFSET [offset into BO] + * VBPNTR [relocated BO] + */ + BEGIN_CS(7); + OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3); + OUT_CS(1); + OUT_CS(r300->vertex_info.vinfo.size | + (r300->vertex_info.vinfo.size << 8)); + OUT_CS(r300->vbo_offset); + OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_CS; +} + void r300_emit_vertex_format_state(struct r300_context* r300) { int i; @@ -310,33 +448,33 @@ void r300_emit_vertex_format_state(struct r300_context* r300) OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); OUT_CS(r300->vertex_info.vinfo.hwfmt[2]); OUT_CS(r300->vertex_info.vinfo.hwfmt[3]); - for (i = 0; i < 4; i++) { - debug_printf("hwfmt%d: 0x%08x\n", i, - r300->vertex_info.vinfo.hwfmt[i]); - } + /* for (i = 0; i < 4; i++) { + * debug_printf("hwfmt%d: 0x%08x\n", i, + * r300->vertex_info.vinfo.hwfmt[i]); + * } */ OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]); - debug_printf("prog_stream_cntl%d: 0x%08x\n", i, - r300->vertex_info.vap_prog_stream_cntl[i]); + /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i, + * r300->vertex_info.vap_prog_stream_cntl[i]); */ } OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]); - debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, - r300->vertex_info.vap_prog_stream_cntl_ext[i]); + /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, + * r300->vertex_info.vap_prog_stream_cntl_ext[i]); */ } END_CS; } -void r300_emit_vertex_shader(struct r300_context* r300, - struct r300_vertex_shader* vs) +void r300_emit_vertex_program_code(struct r300_context* r300, + struct r300_vertex_program_code* code, + struct r300_constant_buffer* constants) { int i; struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct r300_constant_buffer* constants = - &r300->shader_constants[PIPE_SHADER_VERTEX]; + unsigned instruction_count = code->length / 4; CS_LOCALS(r300); if (!r300screen->caps->has_tcl) { @@ -345,38 +483,40 @@ void r300_emit_vertex_shader(struct r300_context* r300, return; } - if (constants->count) { - BEGIN_CS(16 + (vs->instruction_count * 4) + (constants->count * 4)); + if (code->constants.Count) { + BEGIN_CS(14 + code->length + (code->constants.Count * 4)); } else { - BEGIN_CS(13 + (vs->instruction_count * 4) + (constants->count * 4)); + BEGIN_CS(11 + code->length); } - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) | - R300_PVS_LAST_INST(vs->instruction_count - 1)); - OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, vs->instruction_count - 1); - - /* XXX */ - OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x0); + /* R300_VAP_PVS_CODE_CNTL_0 + * R300_VAP_PVS_CONST_CNTL + * R300_VAP_PVS_CODE_CNTL_1 + * See the r5xx docs for instructions on how to use these. + * XXX these could be optimized to select better values... */ + OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3); + OUT_CS(R300_PVS_FIRST_INST(0) | + R300_PVS_XYZW_VALID_INST(instruction_count - 1) | + R300_PVS_LAST_INST(instruction_count - 1)); + OUT_CS(R300_PVS_MAX_CONST_ADDR(code->constants.Count - 1)); + OUT_CS(instruction_count - 1); OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0); - OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4); - for (i = 0; i < vs->instruction_count; i++) { - OUT_CS(vs->instructions[i].inst0); - OUT_CS(vs->instructions[i].inst1); - OUT_CS(vs->instructions[i].inst2); - OUT_CS(vs->instructions[i].inst3); - } + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->length); + for (i = 0; i < code->length; i++) + OUT_CS(code->body.d[i]); - if (constants->count) { + if (code->constants.Count) { OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, (r300screen->caps->is_r500 ? R500_PVS_CONST_START : R300_PVS_CONST_START)); - OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->count * 4); - for (i = 0; i < constants->count; i++) { - OUT_CS_32F(constants->constants[i][0]); - OUT_CS_32F(constants->constants[i][1]); - OUT_CS_32F(constants->constants[i][2]); - OUT_CS_32F(constants->constants[i][3]); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->constants.Count * 4); + for (i = 0; i < code->constants.Count; i++) { + const float * data = get_shader_constant(r300, &code->constants.Constants[i], constants); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); } } @@ -386,7 +526,12 @@ void r300_emit_vertex_shader(struct r300_context* r300, R300_PVS_VF_MAX_VTX_NUM(12)); OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); END_CS; +} +void r300_emit_vertex_shader(struct r300_context* r300, + struct r300_vertex_shader* vs) +{ + r300_emit_vertex_program_code(r300, &vs->code, &r300->shader_constants[PIPE_SHADER_VERTEX]); } void r300_emit_viewport_state(struct r300_context* r300, @@ -403,7 +548,11 @@ void r300_emit_viewport_state(struct r300_context* r300, OUT_CS_32F(viewport->zscale); OUT_CS_32F(viewport->zoffset); - OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control); + if (r300->rs_state->enable_vte) { + OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control); + } else { + OUT_CS_REG(R300_VAP_VTE_CNTL, 0); + } END_CS; } @@ -421,23 +570,67 @@ void r300_flush_textures(struct r300_context* r300) void r300_emit_dirty_state(struct r300_context* r300) { struct r300_screen* r300screen = r300_screen(r300->context.screen); - int i; - int dirty_tex = 0; + struct r300_texture* tex; + int i, dirty_tex = 0; + boolean invalid = FALSE; - if (!(r300->dirty_hw)) { + if (!(r300->dirty_state)) { return; } r300_update_derived_state(r300); /* XXX check size */ - struct r300_texture* fb_tex = - (struct r300_texture*)r300->framebuffer_state.cbufs[0]; - r300->winsys->add_buffer(r300->winsys, fb_tex->buffer, - 0, RADEON_GEM_DOMAIN_VRAM); - if (r300->winsys->validate(r300->winsys)) { - /* XXX */ +validate: + /* Color buffers... */ + for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) { + tex = (struct r300_texture*)r300->framebuffer_state.cbufs[i]->texture; + assert(tex && tex->buffer && "cbuf is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...depth buffer... */ + if (r300->framebuffer_state.zsbuf) { + tex = (struct r300_texture*)r300->framebuffer_state.zsbuf->texture; + assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + 0, RADEON_GEM_DOMAIN_VRAM)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...textures... */ + for (i = 0; i < r300->texture_count; i++) { + tex = r300->textures[i]; + assert(tex && tex->buffer && "texture is marked, but NULL!"); + if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, + RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } + /* ...and vertex buffer. */ + if (r300->vbo) { + if (!r300->winsys->add_buffer(r300->winsys, r300->vbo, + RADEON_GEM_DOMAIN_GTT, 0)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } + } else { + debug_printf("No VBO while emitting dirty state!\n"); + } + if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); + if (invalid) { + /* Well, hell. */ + debug_printf("r300: Stuck in validation loop, gonna quit now."); + exit(1); + } + invalid = TRUE; + goto validate; } if (r300->dirty_state & R300_NEW_BLEND) { @@ -450,6 +643,11 @@ void r300_emit_dirty_state(struct r300_context* r300) r300->dirty_state &= ~R300_NEW_BLEND_COLOR; } + if (r300->dirty_state & R300_NEW_CLIP) { + r300_emit_clip_state(r300, &r300->clip_state); + r300->dirty_state &= ~R300_NEW_CLIP; + } + if (r300->dirty_state & R300_NEW_DSA) { r300_emit_dsa_state(r300, r300->dsa_state); r300->dirty_state &= ~R300_NEW_DSA; @@ -457,11 +655,9 @@ void r300_emit_dirty_state(struct r300_context* r300) if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { if (r300screen->caps->is_r500) { - r500_emit_fragment_shader(r300, - (struct r500_fragment_shader*)r300->fs); + r500_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); } else { - r300_emit_fragment_shader(r300, - (struct r300_fragment_shader*)r300->fs); + r300_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; } @@ -481,29 +677,27 @@ void r300_emit_dirty_state(struct r300_context* r300) r300->dirty_state &= ~R300_NEW_RS_BLOCK; } - if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) { - for (i = 0; i < r300->sampler_count; i++) { - if (r300->dirty_state & (R300_NEW_SAMPLER << i)) { - r300_emit_sampler(r300, r300->sampler_states[i], i); - r300->dirty_state &= ~(R300_NEW_SAMPLER << i); - dirty_tex++; - } - } - } - if (r300->dirty_state & R300_NEW_SCISSOR) { r300_emit_scissor_state(r300, r300->scissor_state); r300->dirty_state &= ~R300_NEW_SCISSOR; } - if (r300->dirty_state & R300_ANY_NEW_TEXTURES) { - for (i = 0; i < r300->texture_count; i++) { - if (r300->dirty_state & (R300_NEW_TEXTURE << i)) { - r300_emit_texture(r300, r300->textures[i], i); - r300->dirty_state &= ~(R300_NEW_TEXTURE << i); + /* Samplers and textures are tracked separately but emitted together. */ + if (r300->dirty_state & + (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) { + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { + if (r300->dirty_state & + ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { + r300_emit_texture(r300, + r300->sampler_states[i], + r300->textures[i], + i); + r300->dirty_state &= + ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i)); dirty_tex++; } } + r300->dirty_state &= ~(R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES); } if (r300->dirty_state & R300_NEW_VIEWPORT) { @@ -519,4 +713,18 @@ void r300_emit_dirty_state(struct r300_context* r300) r300_emit_vertex_format_state(r300); r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT; } + + if (r300->dirty_state & R300_NEW_VERTEX_SHADER) { + r300_emit_vertex_shader(r300, r300->vs); + r300->dirty_state &= ~R300_NEW_VERTEX_SHADER; + } + + /* XXX + assert(r300->dirty_state == 0); + */ + + /* Finally, emit the VBO. */ + r300_emit_vertex_buffer(r300); + + r300->dirty_hw++; } |