si_state_draw.c revision 9f5ff5981cfc1b1eb613890e4f8cec6a448611cc
1/* 2 * Copyright 2012 Advanced Micro Devices, Inc. 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * on the rights to use, copy, modify, merge, publish, distribute, sub 8 * license, and/or sell copies of the Software, and to permit persons to whom 9 * the Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, 19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 21 * USE OR OTHER DEALINGS IN THE SOFTWARE. 22 * 23 * Authors: 24 * Christian König <christian.koenig@amd.com> 25 */ 26 27#include "util/u_memory.h" 28#include "util/u_framebuffer.h" 29#include "util/u_blitter.h" 30#include "tgsi/tgsi_parse.h" 31#include "radeonsi_pipe.h" 32#include "radeonsi_shader.h" 33#include "si_state.h" 34#include "sid.h" 35 36/* 37 * Shaders 38 */ 39 40static void si_pipe_shader_vs(struct pipe_context *ctx, struct si_pipe_shader *shader) 41{ 42 struct r600_context *rctx = (struct r600_context *)ctx; 43 struct si_pm4_state *pm4; 44 unsigned num_sgprs, num_user_sgprs; 45 unsigned nparams, i; 46 uint64_t va; 47 48 if (si_pipe_shader_create(ctx, shader)) 49 return; 50 51 si_pm4_delete_state(rctx, vs, shader->pm4); 52 pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state); 53 54 si_pm4_inval_shader_cache(pm4); 55 56 /* Certain attributes (position, psize, etc.) don't count as params. 57 * VS is required to export at least one param and r600_shader_from_tgsi() 58 * takes care of adding a dummy export. 59 */ 60 for (nparams = 0, i = 0 ; i < shader->shader.noutput; i++) { 61 if (shader->shader.output[i].name != TGSI_SEMANTIC_POSITION) 62 nparams++; 63 } 64 if (nparams < 1) 65 nparams = 1; 66 67 si_pm4_set_reg(pm4, R_0286C4_SPI_VS_OUT_CONFIG, 68 S_0286C4_VS_EXPORT_COUNT(nparams - 1)); 69 70 si_pm4_set_reg(pm4, R_02870C_SPI_SHADER_POS_FORMAT, 71 S_02870C_POS0_EXPORT_FORMAT(V_02870C_SPI_SHADER_4COMP) | 72 S_02870C_POS1_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) | 73 S_02870C_POS2_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) | 74 S_02870C_POS3_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE)); 75 76 va = r600_resource_va(ctx->screen, (void *)shader->bo); 77 si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ); 78 si_pm4_set_reg(pm4, R_00B120_SPI_SHADER_PGM_LO_VS, va >> 8); 79 si_pm4_set_reg(pm4, R_00B124_SPI_SHADER_PGM_HI_VS, va >> 40); 80 81 num_user_sgprs = 8; 82 num_sgprs = shader->num_sgprs; 83 if (num_user_sgprs > num_sgprs) 84 num_sgprs = num_user_sgprs; 85 /* Last 2 reserved SGPRs are used for VCC */ 86 num_sgprs += 2; 87 assert(num_sgprs <= 104); 88 89 si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS, 90 S_00B128_VGPRS((shader->num_vgprs - 1) / 4) | 91 S_00B128_SGPRS((num_sgprs - 1) / 8)); 92 si_pm4_set_reg(pm4, R_00B12C_SPI_SHADER_PGM_RSRC2_VS, 93 S_00B12C_USER_SGPR(num_user_sgprs)); 94 95 si_pm4_bind_state(rctx, vs, shader->pm4); 96} 97 98static void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *shader) 99{ 100 struct r600_context *rctx = (struct r600_context *)ctx; 101 struct si_pm4_state *pm4; 102 unsigned i, exports_ps, num_cout, spi_ps_in_control, db_shader_control; 103 unsigned num_sgprs, num_user_sgprs; 104 int ninterp = 0; 105 boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE; 106 unsigned spi_baryc_cntl; 107 uint64_t va; 108 109 if (si_pipe_shader_create(ctx, shader)) 110 return; 111 112 si_pm4_delete_state(rctx, ps, shader->pm4); 113 pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state); 114 115 si_pm4_inval_shader_cache(pm4); 116 117 db_shader_control = S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z); 118 for (i = 0; i < shader->shader.ninput; i++) { 119 ninterp++; 120 /* XXX: Flat shading hangs the GPU */ 121 if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_CONSTANT || 122 (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_COLOR && 123 rctx->queued.named.rasterizer->flatshade)) 124 have_linear = TRUE; 125 if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_LINEAR) 126 have_linear = TRUE; 127 if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE) 128 have_perspective = TRUE; 129 if (shader->shader.input[i].centroid) 130 have_centroid = TRUE; 131 } 132 133 for (i = 0; i < shader->shader.noutput; i++) { 134 if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION) 135 db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1); 136 if (shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL) 137 db_shader_control |= 0; // XXX OP_VAL or TEST_VAL? 138 } 139 if (shader->shader.uses_kill) 140 db_shader_control |= S_02880C_KILL_ENABLE(1); 141 142 exports_ps = 0; 143 num_cout = 0; 144 for (i = 0; i < shader->shader.noutput; i++) { 145 if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION || 146 shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL) 147 exports_ps |= 1; 148 else if (shader->shader.output[i].name == TGSI_SEMANTIC_COLOR) { 149 if (shader->shader.fs_write_all) 150 num_cout = shader->shader.nr_cbufs; 151 else 152 num_cout++; 153 } 154 } 155 if (!exports_ps) { 156 /* always at least export 1 component per pixel */ 157 exports_ps = 2; 158 } 159 160 spi_ps_in_control = S_0286D8_NUM_INTERP(ninterp); 161 162 spi_baryc_cntl = 0; 163 if (have_perspective) 164 spi_baryc_cntl |= have_centroid ? 165 S_0286E0_PERSP_CENTROID_CNTL(1) : S_0286E0_PERSP_CENTER_CNTL(1); 166 if (have_linear) 167 spi_baryc_cntl |= have_centroid ? 168 S_0286E0_LINEAR_CENTROID_CNTL(1) : S_0286E0_LINEAR_CENTER_CNTL(1); 169 170 si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl); 171 si_pm4_set_reg(pm4, R_0286CC_SPI_PS_INPUT_ENA, shader->spi_ps_input_ena); 172 si_pm4_set_reg(pm4, R_0286D0_SPI_PS_INPUT_ADDR, shader->spi_ps_input_ena); 173 si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control); 174 175 /* XXX: Depends on Z buffer format? */ 176 si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, 0); 177 178 /* XXX: Depends on color buffer format? */ 179 si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT, 180 S_028714_COL0_EXPORT_FORMAT(V_028714_SPI_SHADER_32_ABGR)); 181 182 va = r600_resource_va(ctx->screen, (void *)shader->bo); 183 si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ); 184 si_pm4_set_reg(pm4, R_00B020_SPI_SHADER_PGM_LO_PS, va >> 8); 185 si_pm4_set_reg(pm4, R_00B024_SPI_SHADER_PGM_HI_PS, va >> 40); 186 187 num_user_sgprs = 6; 188 num_sgprs = shader->num_sgprs; 189 if (num_user_sgprs > num_sgprs) 190 num_sgprs = num_user_sgprs; 191 /* Last 2 reserved SGPRs are used for VCC */ 192 num_sgprs += 2; 193 assert(num_sgprs <= 104); 194 195 si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS, 196 S_00B028_VGPRS((shader->num_vgprs - 1) / 4) | 197 S_00B028_SGPRS((num_sgprs - 1) / 8)); 198 si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS, 199 S_00B02C_USER_SGPR(num_user_sgprs)); 200 201 si_pm4_set_reg(pm4, R_02880C_DB_SHADER_CONTROL, db_shader_control); 202 203 shader->sprite_coord_enable = rctx->sprite_coord_enable; 204 si_pm4_bind_state(rctx, ps, shader->pm4); 205} 206 207/* 208 * Drawing 209 */ 210 211static unsigned si_conv_pipe_prim(unsigned pprim) 212{ 213 static const unsigned prim_conv[] = { 214 [PIPE_PRIM_POINTS] = V_008958_DI_PT_POINTLIST, 215 [PIPE_PRIM_LINES] = V_008958_DI_PT_LINELIST, 216 [PIPE_PRIM_LINE_LOOP] = V_008958_DI_PT_LINELOOP, 217 [PIPE_PRIM_LINE_STRIP] = V_008958_DI_PT_LINESTRIP, 218 [PIPE_PRIM_TRIANGLES] = V_008958_DI_PT_TRILIST, 219 [PIPE_PRIM_TRIANGLE_STRIP] = V_008958_DI_PT_TRISTRIP, 220 [PIPE_PRIM_TRIANGLE_FAN] = V_008958_DI_PT_TRIFAN, 221 [PIPE_PRIM_QUADS] = V_008958_DI_PT_QUADLIST, 222 [PIPE_PRIM_QUAD_STRIP] = V_008958_DI_PT_QUADSTRIP, 223 [PIPE_PRIM_POLYGON] = V_008958_DI_PT_POLYGON, 224 [PIPE_PRIM_LINES_ADJACENCY] = ~0, 225 [PIPE_PRIM_LINE_STRIP_ADJACENCY] = ~0, 226 [PIPE_PRIM_TRIANGLES_ADJACENCY] = ~0, 227 [PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY] = ~0 228 }; 229 unsigned result = prim_conv[pprim]; 230 if (result == ~0) { 231 R600_ERR("unsupported primitive type %d\n", pprim); 232 } 233 return result; 234} 235 236static bool si_update_draw_info_state(struct r600_context *rctx, 237 const struct pipe_draw_info *info) 238{ 239 struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state); 240 unsigned prim = si_conv_pipe_prim(info->mode); 241 unsigned ls_mask = 0; 242 243 if (pm4 == NULL) 244 return false; 245 246 if (prim == ~0) { 247 FREE(pm4); 248 return false; 249 } 250 251 si_pm4_set_reg(pm4, R_008958_VGT_PRIMITIVE_TYPE, prim); 252 si_pm4_set_reg(pm4, R_028400_VGT_MAX_VTX_INDX, ~0); 253 si_pm4_set_reg(pm4, R_028404_VGT_MIN_VTX_INDX, 0); 254 si_pm4_set_reg(pm4, R_028408_VGT_INDX_OFFSET, 255 info->indexed ? info->index_bias : info->start); 256 si_pm4_set_reg(pm4, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, info->restart_index); 257 si_pm4_set_reg(pm4, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, info->primitive_restart); 258#if 0 259 si_pm4_set_reg(pm4, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0); 260 si_pm4_set_reg(pm4, R_03CFF4_SQ_VTX_START_INST_LOC, info->start_instance); 261#endif 262 263 if (prim == V_008958_DI_PT_LINELIST) 264 ls_mask = 1; 265 else if (prim == V_008958_DI_PT_LINESTRIP) 266 ls_mask = 2; 267 si_pm4_set_reg(pm4, R_028A0C_PA_SC_LINE_STIPPLE, 268 S_028A0C_AUTO_RESET_CNTL(ls_mask) | 269 rctx->pa_sc_line_stipple); 270 271 if (info->mode == PIPE_PRIM_QUADS || info->mode == PIPE_PRIM_QUAD_STRIP || info->mode == PIPE_PRIM_POLYGON) { 272 si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL, 273 S_028814_PROVOKING_VTX_LAST(1) | rctx->pa_su_sc_mode_cntl); 274 } else { 275 si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL, rctx->pa_su_sc_mode_cntl); 276 } 277 si_pm4_set_reg(pm4, R_02881C_PA_CL_VS_OUT_CNTL, 278 prim == PIPE_PRIM_POINTS ? rctx->pa_cl_vs_out_cntl : 0 279 /*| (rctx->rasterizer->clip_plane_enable & 280 rctx->vs_shader->shader.clip_dist_write)*/); 281 si_pm4_set_reg(pm4, R_028810_PA_CL_CLIP_CNTL, rctx->pa_cl_clip_cntl 282 /*| (rctx->vs_shader->shader.clip_dist_write || 283 rctx->vs_shader->shader.vs_prohibit_ucps ? 284 0 : rctx->rasterizer->clip_plane_enable & 0x3F)*/); 285 286 si_pm4_set_state(rctx, draw_info, pm4); 287 return true; 288} 289 290static void si_update_alpha_ref(struct r600_context *rctx) 291{ 292#if 0 293 unsigned alpha_ref; 294 struct r600_pipe_state rstate; 295 296 alpha_ref = rctx->alpha_ref; 297 rstate.nregs = 0; 298 if (rctx->export_16bpc) 299 alpha_ref &= ~0x1FFF; 300 si_pm4_set_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref); 301 302 si_pm4_set_state(rctx, TODO, pm4); 303 rctx->alpha_ref_dirty = false; 304#endif 305} 306 307static void si_update_spi_map(struct r600_context *rctx) 308{ 309 struct si_shader *ps = &rctx->ps_shader->shader; 310 struct si_shader *vs = &rctx->vs_shader->shader; 311 struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state); 312 unsigned i, j, tmp; 313 314 for (i = 0; i < ps->ninput; i++) { 315 tmp = 0; 316 317#if 0 318 /* XXX: Flat shading hangs the GPU */ 319 if (ps->input[i].name == TGSI_SEMANTIC_POSITION || 320 ps->input[i].interpolate == TGSI_INTERPOLATE_CONSTANT || 321 (ps->input[i].interpolate == TGSI_INTERPOLATE_COLOR && 322 rctx->rasterizer && rctx->rasterizer->flatshade)) { 323 tmp |= S_028644_FLAT_SHADE(1); 324 } 325#endif 326 327 if (ps->input[i].name == TGSI_SEMANTIC_GENERIC && 328 rctx->sprite_coord_enable & (1 << ps->input[i].sid)) { 329 tmp |= S_028644_PT_SPRITE_TEX(1); 330 } 331 332 for (j = 0; j < vs->noutput; j++) { 333 if (ps->input[i].name == vs->output[j].name && 334 ps->input[i].sid == vs->output[j].sid) { 335 tmp |= S_028644_OFFSET(vs->output[j].param_offset); 336 break; 337 } 338 } 339 340 if (j == vs->noutput) { 341 /* No corresponding output found, load defaults into input */ 342 tmp |= S_028644_OFFSET(0x20); 343 } 344 345 si_pm4_set_reg(pm4, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, tmp); 346 } 347 348 si_pm4_set_state(rctx, spi, pm4); 349} 350 351static void si_update_derived_state(struct r600_context *rctx) 352{ 353 struct pipe_context * ctx = (struct pipe_context*)rctx; 354 355 if (!rctx->blitter->running) { 356 if (rctx->have_depth_fb || rctx->have_depth_texture) 357 r600_flush_depth_textures(rctx); 358 } 359 360 if ((rctx->ps_shader->shader.fs_write_all && 361 (rctx->ps_shader->shader.nr_cbufs != rctx->framebuffer.nr_cbufs)) || 362 (rctx->sprite_coord_enable && 363 (rctx->ps_shader->sprite_coord_enable != rctx->sprite_coord_enable))) { 364 si_pipe_shader_destroy(&rctx->context, rctx->ps_shader); 365 } 366 367 if (rctx->alpha_ref_dirty) { 368 si_update_alpha_ref(rctx); 369 } 370 371 if (!rctx->vs_shader->bo) { 372 si_pipe_shader_vs(ctx, rctx->vs_shader); 373 } 374 375 if (!rctx->ps_shader->bo) { 376 si_pipe_shader_ps(ctx, rctx->ps_shader); 377 } 378 if (!rctx->ps_shader->bo) { 379 if (!rctx->dummy_pixel_shader->bo) 380 si_pipe_shader_ps(ctx, rctx->dummy_pixel_shader); 381 382 if (rctx->dummy_pixel_shader->pm4) 383 si_pm4_bind_state(rctx, vs, rctx->dummy_pixel_shader->pm4); 384 } 385 386 if (rctx->shader_dirty) { 387 si_update_spi_map(rctx); 388 rctx->shader_dirty = false; 389 } 390} 391 392static void si_vertex_buffer_update(struct r600_context *rctx) 393{ 394 struct pipe_context *ctx = &rctx->context; 395 struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state); 396 bool bound[PIPE_MAX_ATTRIBS] = {}; 397 struct si_resource *t_list_buffer; 398 unsigned i, count; 399 uint32_t *ptr; 400 uint64_t va; 401 402 si_pm4_inval_vertex_cache(pm4); 403 404 /* bind vertex buffer once */ 405 count = rctx->vertex_elements->count; 406 assert(count <= 256 / 4); 407 408 t_list_buffer = si_resource_create_custom(ctx->screen, PIPE_USAGE_IMMUTABLE, 409 4 * 4 * count); 410 if (t_list_buffer == NULL) { 411 FREE(pm4); 412 return; 413 } 414 si_pm4_add_bo(pm4, t_list_buffer, RADEON_USAGE_READ); 415 416 ptr = (uint32_t*)rctx->ws->buffer_map(t_list_buffer->cs_buf, 417 rctx->cs, 418 PIPE_TRANSFER_WRITE); 419 420 for (i = 0 ; i < count; i++, ptr += 4) { 421 struct pipe_vertex_element *ve = &rctx->vertex_elements->elements[i]; 422 struct pipe_vertex_buffer *vb; 423 struct si_resource *rbuffer; 424 unsigned offset; 425 426 if (ve->vertex_buffer_index >= rctx->nr_vertex_buffers) 427 continue; 428 429 vb = &rctx->vertex_buffer[ve->vertex_buffer_index]; 430 rbuffer = (struct si_resource*)vb->buffer; 431 if (rbuffer == NULL) 432 continue; 433 434 offset = 0; 435 offset += vb->buffer_offset; 436 offset += ve->src_offset; 437 438 va = r600_resource_va(ctx->screen, (void*)rbuffer); 439 va += offset; 440 441 /* Fill in T# buffer resource description */ 442 ptr[0] = va & 0xFFFFFFFF; 443 ptr[1] = (S_008F04_BASE_ADDRESS_HI(va >> 32) | 444 S_008F04_STRIDE(vb->stride)); 445 if (vb->stride > 0) 446 ptr[2] = (vb->buffer->width0 - offset) / vb->stride; 447 else 448 ptr[2] = vb->buffer->width0 - offset; 449 ptr[3] = rctx->vertex_elements->rsrc_word3[i]; 450 451 if (!bound[ve->vertex_buffer_index]) { 452 si_pm4_add_bo(pm4, rbuffer, RADEON_USAGE_READ); 453 bound[ve->vertex_buffer_index] = true; 454 } 455 } 456 457 va = r600_resource_va(ctx->screen, (void*)t_list_buffer); 458 si_pm4_set_reg(pm4, R_00B148_SPI_SHADER_USER_DATA_VS_6, va); 459 si_pm4_set_reg(pm4, R_00B14C_SPI_SHADER_USER_DATA_VS_7, va >> 32); 460 si_pm4_set_state(rctx, vertex_buffers, pm4); 461} 462 463static void si_state_draw(struct r600_context *rctx, 464 const struct pipe_draw_info *info, 465 const struct pipe_index_buffer *ib) 466{ 467 struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state); 468 469 /* queries need some special values 470 * (this is non-zero if any query is active) */ 471 if (rctx->num_cs_dw_queries_suspend) { 472 struct si_state_dsa *dsa = rctx->queued.named.dsa; 473 474 si_pm4_set_reg(pm4, R_028004_DB_COUNT_CONTROL, 475 S_028004_PERFECT_ZPASS_COUNTS(1)); 476 si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, 477 dsa->db_render_override | 478 S_02800C_NOOP_CULL_DISABLE(1)); 479 } 480 481 /* draw packet */ 482 si_pm4_cmd_begin(pm4, PKT3_INDEX_TYPE); 483 if (ib->index_size == 4) { 484 si_pm4_cmd_add(pm4, V_028A7C_VGT_INDEX_32 | (R600_BIG_ENDIAN ? 485 V_028A7C_VGT_DMA_SWAP_32_BIT : 0)); 486 } else { 487 si_pm4_cmd_add(pm4, V_028A7C_VGT_INDEX_16 | (R600_BIG_ENDIAN ? 488 V_028A7C_VGT_DMA_SWAP_16_BIT : 0)); 489 } 490 si_pm4_cmd_end(pm4, rctx->predicate_drawing); 491 492 si_pm4_cmd_begin(pm4, PKT3_NUM_INSTANCES); 493 si_pm4_cmd_add(pm4, info->instance_count); 494 si_pm4_cmd_end(pm4, rctx->predicate_drawing); 495 496 if (info->indexed) { 497 uint64_t va; 498 va = r600_resource_va(&rctx->screen->screen, ib->buffer); 499 va += ib->offset; 500 501 si_pm4_add_bo(pm4, (struct si_resource *)ib->buffer, RADEON_USAGE_READ); 502 si_pm4_cmd_begin(pm4, PKT3_DRAW_INDEX_2); 503 si_pm4_cmd_add(pm4, (ib->buffer->width0 - ib->offset) / 504 rctx->index_buffer.index_size); 505 si_pm4_cmd_add(pm4, va); 506 si_pm4_cmd_add(pm4, (va >> 32UL) & 0xFF); 507 si_pm4_cmd_add(pm4, info->count); 508 si_pm4_cmd_add(pm4, V_0287F0_DI_SRC_SEL_DMA); 509 si_pm4_cmd_end(pm4, rctx->predicate_drawing); 510 } else { 511 si_pm4_cmd_begin(pm4, PKT3_DRAW_INDEX_AUTO); 512 si_pm4_cmd_add(pm4, info->count); 513 si_pm4_cmd_add(pm4, V_0287F0_DI_SRC_SEL_AUTO_INDEX | 514 (info->count_from_stream_output ? 515 S_0287F0_USE_OPAQUE(1) : 0)); 516 si_pm4_cmd_end(pm4, rctx->predicate_drawing); 517 } 518 si_pm4_set_state(rctx, draw, pm4); 519} 520 521void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) 522{ 523 struct r600_context *rctx = (struct r600_context *)ctx; 524 struct pipe_index_buffer ib = {}; 525 uint32_t cp_coher_cntl; 526 527 if ((!info->count && (info->indexed || !info->count_from_stream_output)) || 528 (info->indexed && !rctx->index_buffer.buffer)) { 529 return; 530 } 531 532 if (!rctx->ps_shader || !rctx->vs_shader) 533 return; 534 535 si_update_derived_state(rctx); 536 si_vertex_buffer_update(rctx); 537 538 if (info->indexed) { 539 /* Initialize the index buffer struct. */ 540 pipe_resource_reference(&ib.buffer, rctx->index_buffer.buffer); 541 ib.index_size = rctx->index_buffer.index_size; 542 ib.offset = rctx->index_buffer.offset + info->start * ib.index_size; 543 544 /* Translate or upload, if needed. */ 545 r600_translate_index_buffer(rctx, &ib, info->count); 546 547 if (ib.user_buffer) { 548 r600_upload_index_buffer(rctx, &ib, info->count); 549 } 550 551 } else if (info->count_from_stream_output) { 552 r600_context_draw_opaque_count(rctx, (struct r600_so_target*)info->count_from_stream_output); 553 } 554 555 rctx->vs_shader_so_strides = rctx->vs_shader->so_strides; 556 557 if (!si_update_draw_info_state(rctx, info)) 558 return; 559 560 si_state_draw(rctx, info, &ib); 561 562 cp_coher_cntl = si_pm4_sync_flags(rctx); 563 if (cp_coher_cntl) { 564 struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state); 565 si_cmd_surface_sync(pm4, cp_coher_cntl); 566 si_pm4_set_state(rctx, sync, pm4); 567 } 568 569 /* Emit states. */ 570 rctx->pm4_dirty_cdwords += si_pm4_dirty_dw(rctx); 571 572 r600_need_cs_space(rctx, 0, TRUE); 573 574 si_pm4_emit_dirty(rctx); 575 rctx->pm4_dirty_cdwords = 0; 576 577#if 0 578 /* Enable stream out if needed. */ 579 if (rctx->streamout_start) { 580 r600_context_streamout_begin(rctx); 581 rctx->streamout_start = FALSE; 582 } 583#endif 584 585 586 rctx->flags |= R600_CONTEXT_DST_CACHES_DIRTY; 587 588 if (rctx->framebuffer.zsbuf) 589 { 590 struct pipe_resource *tex = rctx->framebuffer.zsbuf->texture; 591 ((struct r600_resource_texture *)tex)->dirty_db = TRUE; 592 } 593 594 pipe_resource_reference(&ib.buffer, NULL); 595} 596