nv50_state_validate.c revision 20443f969d78efb8437083b39b9ddcf4c08aa6cf
1/* 2 * Copyright 2008 Ben Skeggs 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 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice shall be included in 12 * all copies or substantial portions of the Software. 13 * 14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 17 * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, 18 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF 19 * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 20 * SOFTWARE. 21 */ 22 23#include "util/u_format.h" 24 25#include "nv50_context.h" 26#include "nouveau/nouveau_stateobj.h" 27 28static void 29nv50_state_validate_fb(struct nv50_context *nv50) 30{ 31 struct nouveau_grobj *tesla = nv50->screen->tesla; 32 struct nouveau_stateobj *so = so_new(32, 79, 18); 33 struct pipe_framebuffer_state *fb = &nv50->framebuffer; 34 unsigned i, w, h, gw = 0; 35 36 /* Set nr of active RTs and select RT for each colour output. 37 * FP result 0 always goes to RT[0], bits 4 - 6 are ignored. 38 * Ambiguous assignment results in no rendering (no DATA_ERROR). 39 */ 40 so_method(so, tesla, NV50TCL_RT_CONTROL, 1); 41 so_data (so, fb->nr_cbufs | 42 (0 << 4) | (1 << 7) | (2 << 10) | (3 << 13) | 43 (4 << 16) | (5 << 19) | (6 << 22) | (7 << 25)); 44 45 for (i = 0; i < fb->nr_cbufs; i++) { 46 struct pipe_texture *pt = fb->cbufs[i]->texture; 47 struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; 48 49 if (!gw) { 50 w = fb->cbufs[i]->width; 51 h = fb->cbufs[i]->height; 52 gw = 1; 53 } else { 54 assert(w == fb->cbufs[i]->width); 55 assert(h == fb->cbufs[i]->height); 56 } 57 58 so_method(so, tesla, NV50TCL_RT_HORIZ(i), 2); 59 so_data (so, fb->cbufs[i]->width); 60 so_data (so, fb->cbufs[i]->height); 61 62 so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5); 63 so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | 64 NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); 65 so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | 66 NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); 67 switch (fb->cbufs[i]->format) { 68 case PIPE_FORMAT_B8G8R8A8_UNORM: 69 so_data(so, NV50TCL_RT_FORMAT_A8R8G8B8_UNORM); 70 break; 71 case PIPE_FORMAT_B8G8R8X8_UNORM: 72 so_data(so, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM); 73 break; 74 case PIPE_FORMAT_B5G6R5_UNORM: 75 so_data(so, NV50TCL_RT_FORMAT_R5G6B5_UNORM); 76 break; 77 case PIPE_FORMAT_R16G16B16A16_SNORM: 78 so_data(so, NV50TCL_RT_FORMAT_R16G16B16A16_SNORM); 79 break; 80 case PIPE_FORMAT_R16G16B16A16_UNORM: 81 so_data(so, NV50TCL_RT_FORMAT_R16G16B16A16_UNORM); 82 break; 83 case PIPE_FORMAT_R32G32B32A32_FLOAT: 84 so_data(so, NV50TCL_RT_FORMAT_R32G32B32A32_FLOAT); 85 break; 86 case PIPE_FORMAT_R16G16_SNORM: 87 so_data(so, NV50TCL_RT_FORMAT_R16G16_SNORM); 88 break; 89 case PIPE_FORMAT_R16G16_UNORM: 90 so_data(so, NV50TCL_RT_FORMAT_R16G16_UNORM); 91 break; 92 default: 93 NOUVEAU_ERR("AIIII unknown format %s\n", 94 util_format_name(fb->cbufs[i]->format)); 95 so_data(so, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM); 96 break; 97 } 98 so_data(so, nv50_miptree(pt)-> 99 level[fb->cbufs[i]->level].tile_mode << 4); 100 so_data(so, 0x00000000); 101 102 so_method(so, tesla, NV50TCL_RT_ARRAY_MODE, 1); 103 so_data (so, 1); 104 } 105 106 if (fb->zsbuf) { 107 struct pipe_texture *pt = fb->zsbuf->texture; 108 struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; 109 110 if (!gw) { 111 w = fb->zsbuf->width; 112 h = fb->zsbuf->height; 113 gw = 1; 114 } else { 115 assert(w == fb->zsbuf->width); 116 assert(h == fb->zsbuf->height); 117 } 118 119 so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5); 120 so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | 121 NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); 122 so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | 123 NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); 124 switch (fb->zsbuf->format) { 125 case PIPE_FORMAT_Z24S8_UNORM: 126 so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM); 127 break; 128 case PIPE_FORMAT_Z24X8_UNORM: 129 so_data(so, NV50TCL_ZETA_FORMAT_X8Z24_UNORM); 130 break; 131 case PIPE_FORMAT_S8Z24_UNORM: 132 so_data(so, NV50TCL_ZETA_FORMAT_Z24S8_UNORM); 133 break; 134 case PIPE_FORMAT_Z32_FLOAT: 135 so_data(so, NV50TCL_ZETA_FORMAT_Z32_FLOAT); 136 break; 137 default: 138 NOUVEAU_ERR("AIIII unknown format %s\n", 139 util_format_name(fb->zsbuf->format)); 140 so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM); 141 break; 142 } 143 so_data(so, nv50_miptree(pt)-> 144 level[fb->zsbuf->level].tile_mode << 4); 145 so_data(so, 0x00000000); 146 147 so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); 148 so_data (so, 1); 149 so_method(so, tesla, NV50TCL_ZETA_HORIZ, 3); 150 so_data (so, fb->zsbuf->width); 151 so_data (so, fb->zsbuf->height); 152 so_data (so, 0x00010001); 153 } else { 154 so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); 155 so_data (so, 0); 156 } 157 158 so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ(0), 2); 159 so_data (so, w << 16); 160 so_data (so, h << 16); 161 /* set window lower left corner */ 162 so_method(so, tesla, NV50TCL_WINDOW_OFFSET_X, 2); 163 so_data (so, 0); 164 so_data (so, 0); 165 /* set screen scissor rectangle */ 166 so_method(so, tesla, NV50TCL_SCREEN_SCISSOR_HORIZ, 2); 167 so_data (so, w << 16); 168 so_data (so, h << 16); 169 170 /* we set scissors to framebuffer size when they're 'turned off' */ 171 nv50->dirty |= NV50_NEW_SCISSOR; 172 so_ref(NULL, &nv50->state.scissor); 173 174 so_ref(so, &nv50->state.fb); 175 so_ref(NULL, &so); 176} 177 178static void 179nv50_validate_samplers(struct nv50_context *nv50, struct nouveau_stateobj *so, 180 unsigned p) 181{ 182 struct nouveau_grobj *eng2d = nv50->screen->eng2d; 183 unsigned i, j, dw = nv50->sampler_nr[p] * 8; 184 185 if (!dw) 186 return; 187 nv50_so_init_sifc(nv50, so, nv50->screen->tsc, NOUVEAU_BO_VRAM, 188 p * (32 * 8 * 4), dw * 4); 189 190 so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), dw); 191 192 for (i = 0; i < nv50->sampler_nr[p]; ++i) { 193 if (nv50->sampler[p][i]) 194 so_datap(so, nv50->sampler[p][i]->tsc, 8); 195 else { 196 for (j = 0; j < 8; ++j) /* you get punished */ 197 so_data(so, 0); /* ... for leaving holes */ 198 } 199 } 200} 201 202static void 203nv50_state_emit(struct nv50_context *nv50) 204{ 205 struct nv50_screen *screen = nv50->screen; 206 struct nouveau_channel *chan = screen->base.channel; 207 208 /* XXX: this is racy for multiple contexts active on separate 209 * threads. 210 */ 211 if (screen->cur_ctx != nv50) { 212 if (nv50->state.fb) 213 nv50->state.dirty |= NV50_NEW_FRAMEBUFFER; 214 if (nv50->state.blend) 215 nv50->state.dirty |= NV50_NEW_BLEND; 216 if (nv50->state.zsa) 217 nv50->state.dirty |= NV50_NEW_ZSA; 218 if (nv50->state.vertprog) 219 nv50->state.dirty |= NV50_NEW_VERTPROG; 220 if (nv50->state.fragprog) 221 nv50->state.dirty |= NV50_NEW_FRAGPROG; 222 if (nv50->state.geomprog) 223 nv50->state.dirty |= NV50_NEW_GEOMPROG; 224 if (nv50->state.rast) 225 nv50->state.dirty |= NV50_NEW_RASTERIZER; 226 if (nv50->state.blend_colour) 227 nv50->state.dirty |= NV50_NEW_BLEND_COLOUR; 228 if (nv50->state.stencil_ref) 229 nv50->state.dirty |= NV50_NEW_STENCIL_REF; 230 if (nv50->state.stipple) 231 nv50->state.dirty |= NV50_NEW_STIPPLE; 232 if (nv50->state.scissor) 233 nv50->state.dirty |= NV50_NEW_SCISSOR; 234 if (nv50->state.viewport) 235 nv50->state.dirty |= NV50_NEW_VIEWPORT; 236 if (nv50->state.tsc_upload) 237 nv50->state.dirty |= NV50_NEW_SAMPLER; 238 if (nv50->state.tic_upload) 239 nv50->state.dirty |= NV50_NEW_TEXTURE; 240 if (nv50->state.vtxfmt && nv50->state.vtxbuf) 241 nv50->state.dirty |= NV50_NEW_ARRAYS; 242 screen->cur_ctx = nv50; 243 } 244 245 if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER) 246 so_emit(chan, nv50->state.fb); 247 if (nv50->state.dirty & NV50_NEW_BLEND) 248 so_emit(chan, nv50->state.blend); 249 if (nv50->state.dirty & NV50_NEW_ZSA) 250 so_emit(chan, nv50->state.zsa); 251 if (nv50->state.dirty & NV50_NEW_VERTPROG) 252 so_emit(chan, nv50->state.vertprog); 253 if (nv50->state.dirty & NV50_NEW_FRAGPROG) 254 so_emit(chan, nv50->state.fragprog); 255 if (nv50->state.dirty & NV50_NEW_GEOMPROG && nv50->state.geomprog) 256 so_emit(chan, nv50->state.geomprog); 257 if (nv50->state.dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | 258 NV50_NEW_GEOMPROG | NV50_NEW_RASTERIZER)) 259 so_emit(chan, nv50->state.fp_linkage); 260 if ((nv50->state.dirty & (NV50_NEW_VERTPROG | NV50_NEW_GEOMPROG)) 261 && nv50->state.gp_linkage) 262 so_emit(chan, nv50->state.gp_linkage); 263 if (nv50->state.dirty & NV50_NEW_RASTERIZER) 264 so_emit(chan, nv50->state.rast); 265 if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR) 266 so_emit(chan, nv50->state.blend_colour); 267 if (nv50->state.dirty & NV50_NEW_STENCIL_REF) 268 so_emit(chan, nv50->state.stencil_ref); 269 if (nv50->state.dirty & NV50_NEW_STIPPLE) 270 so_emit(chan, nv50->state.stipple); 271 if (nv50->state.dirty & NV50_NEW_SCISSOR) 272 so_emit(chan, nv50->state.scissor); 273 if (nv50->state.dirty & NV50_NEW_VIEWPORT) 274 so_emit(chan, nv50->state.viewport); 275 if (nv50->state.dirty & NV50_NEW_SAMPLER) 276 so_emit(chan, nv50->state.tsc_upload); 277 if (nv50->state.dirty & NV50_NEW_TEXTURE) 278 so_emit(chan, nv50->state.tic_upload); 279 if (nv50->state.dirty & NV50_NEW_ARRAYS) { 280 so_emit(chan, nv50->state.vtxfmt); 281 so_emit(chan, nv50->state.vtxbuf); 282 if (nv50->state.vtxattr) 283 so_emit(chan, nv50->state.vtxattr); 284 } 285 nv50->state.dirty = 0; 286} 287 288void 289nv50_state_flush_notify(struct nouveau_channel *chan) 290{ 291 struct nv50_context *nv50 = chan->user_private; 292 293 if (nv50->state.tic_upload && !(nv50->dirty & NV50_NEW_TEXTURE)) 294 so_emit(chan, nv50->state.tic_upload); 295 296 so_emit_reloc_markers(chan, nv50->state.fb); 297 so_emit_reloc_markers(chan, nv50->state.vertprog); 298 so_emit_reloc_markers(chan, nv50->state.fragprog); 299 so_emit_reloc_markers(chan, nv50->state.vtxbuf); 300 so_emit_reloc_markers(chan, nv50->screen->static_init); 301 302 if (nv50->state.instbuf) 303 so_emit_reloc_markers(chan, nv50->state.instbuf); 304} 305 306boolean 307nv50_state_validate(struct nv50_context *nv50) 308{ 309 struct nouveau_grobj *tesla = nv50->screen->tesla; 310 struct nouveau_stateobj *so; 311 unsigned i; 312 313 if (nv50->dirty & NV50_NEW_FRAMEBUFFER) 314 nv50_state_validate_fb(nv50); 315 316 if (nv50->dirty & NV50_NEW_BLEND) 317 so_ref(nv50->blend->so, &nv50->state.blend); 318 319 if (nv50->dirty & NV50_NEW_ZSA) 320 so_ref(nv50->zsa->so, &nv50->state.zsa); 321 322 if (nv50->dirty & (NV50_NEW_VERTPROG | NV50_NEW_VERTPROG_CB)) 323 nv50_vertprog_validate(nv50); 324 325 if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB)) 326 nv50_fragprog_validate(nv50); 327 328 if (nv50->dirty & (NV50_NEW_GEOMPROG | NV50_NEW_GEOMPROG_CB)) 329 nv50_geomprog_validate(nv50); 330 331 if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | 332 NV50_NEW_GEOMPROG | NV50_NEW_RASTERIZER)) 333 nv50_fp_linkage_validate(nv50); 334 335 if (nv50->dirty & (NV50_NEW_GEOMPROG | NV50_NEW_VERTPROG)) 336 nv50_gp_linkage_validate(nv50); 337 338 if (nv50->dirty & NV50_NEW_RASTERIZER) 339 so_ref(nv50->rasterizer->so, &nv50->state.rast); 340 341 if (nv50->dirty & NV50_NEW_BLEND_COLOUR) { 342 so = so_new(1, 4, 0); 343 so_method(so, tesla, NV50TCL_BLEND_COLOR(0), 4); 344 so_data (so, fui(nv50->blend_colour.color[0])); 345 so_data (so, fui(nv50->blend_colour.color[1])); 346 so_data (so, fui(nv50->blend_colour.color[2])); 347 so_data (so, fui(nv50->blend_colour.color[3])); 348 so_ref(so, &nv50->state.blend_colour); 349 so_ref(NULL, &so); 350 } 351 352 if (nv50->dirty & NV50_NEW_STENCIL_REF) { 353 so = so_new(2, 2, 0); 354 so_method(so, tesla, NV50TCL_STENCIL_FRONT_FUNC_REF, 1); 355 so_data (so, nv50->stencil_ref.ref_value[0]); 356 so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 1); 357 so_data (so, nv50->stencil_ref.ref_value[1]); 358 so_ref(so, &nv50->state.stencil_ref); 359 so_ref(NULL, &so); 360 } 361 362 if (nv50->dirty & NV50_NEW_STIPPLE) { 363 so = so_new(1, 32, 0); 364 so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32); 365 for (i = 0; i < 32; i++) 366 so_data(so, util_bswap32(nv50->stipple.stipple[i])); 367 so_ref(so, &nv50->state.stipple); 368 so_ref(NULL, &so); 369 } 370 371 if (nv50->dirty & (NV50_NEW_SCISSOR | NV50_NEW_RASTERIZER)) { 372 struct pipe_rasterizer_state *rast = &nv50->rasterizer->pipe; 373 struct pipe_scissor_state *s = &nv50->scissor; 374 375 if (nv50->state.scissor && 376 (rast->scissor == 0 && nv50->state.scissor_enabled == 0)) 377 goto scissor_uptodate; 378 nv50->state.scissor_enabled = rast->scissor; 379 380 so = so_new(1, 2, 0); 381 so_method(so, tesla, NV50TCL_SCISSOR_HORIZ(0), 2); 382 if (nv50->state.scissor_enabled) { 383 so_data(so, (s->maxx << 16) | s->minx); 384 so_data(so, (s->maxy << 16) | s->miny); 385 } else { 386 so_data(so, (nv50->framebuffer.width << 16)); 387 so_data(so, (nv50->framebuffer.height << 16)); 388 } 389 so_ref(so, &nv50->state.scissor); 390 so_ref(NULL, &so); 391 nv50->state.dirty |= NV50_NEW_SCISSOR; 392 } 393scissor_uptodate: 394 395 if (nv50->dirty & (NV50_NEW_VIEWPORT | NV50_NEW_RASTERIZER)) { 396 if (nv50->state.viewport && 397 !(nv50->dirty & NV50_NEW_VIEWPORT)) 398 goto viewport_uptodate; 399 400 so = so_new(5, 9, 0); 401 so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE_X(0), 3); 402 so_data (so, fui(nv50->viewport.translate[0])); 403 so_data (so, fui(nv50->viewport.translate[1])); 404 so_data (so, fui(nv50->viewport.translate[2])); 405 so_method(so, tesla, NV50TCL_VIEWPORT_SCALE_X(0), 3); 406 so_data (so, fui(nv50->viewport.scale[0])); 407 so_data (so, fui(nv50->viewport.scale[1])); 408 so_data (so, fui(nv50->viewport.scale[2])); 409 410 so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1); 411 so_data (so, 1); 412 /* 0x0000 = remove whole primitive only (xyz) 413 * 0x1018 = remove whole primitive only (xy), clamp z 414 * 0x1080 = clip primitive (xyz) 415 * 0x1098 = clip primitive (xy), clamp z 416 */ 417 so_method(so, tesla, NV50TCL_VIEW_VOLUME_CLIP_CTRL, 1); 418 so_data (so, 0x1080); 419 /* no idea what 0f90 does */ 420 so_method(so, tesla, 0x0f90, 1); 421 so_data (so, 0); 422 423 so_ref(so, &nv50->state.viewport); 424 so_ref(NULL, &so); 425 nv50->state.dirty |= NV50_NEW_VIEWPORT; 426 } 427viewport_uptodate: 428 429 if (nv50->dirty & NV50_NEW_SAMPLER) { 430 unsigned nr = 0; 431 432 for (i = 0; i < PIPE_SHADER_TYPES; ++i) 433 nr += nv50->sampler_nr[i]; 434 435 so = so_new(1 + 5 * PIPE_SHADER_TYPES, 436 1 + 19 * PIPE_SHADER_TYPES + nr * 8, 437 PIPE_SHADER_TYPES * 2); 438 439 nv50_validate_samplers(nv50, so, PIPE_SHADER_VERTEX); 440 nv50_validate_samplers(nv50, so, PIPE_SHADER_FRAGMENT); 441 442 so_method(so, tesla, 0x1334, 1); /* flush TSC */ 443 so_data (so, 0); 444 445 so_ref(so, &nv50->state.tsc_upload); 446 so_ref(NULL, &so); 447 } 448 449 if (nv50->dirty & (NV50_NEW_TEXTURE | NV50_NEW_SAMPLER)) 450 nv50_tex_validate(nv50); 451 452 if (nv50->dirty & NV50_NEW_ARRAYS) 453 nv50_vbo_validate(nv50); 454 455 nv50->state.dirty |= nv50->dirty; 456 nv50->dirty = 0; 457 nv50_state_emit(nv50); 458 459 return TRUE; 460} 461 462void nv50_so_init_sifc(struct nv50_context *nv50, 463 struct nouveau_stateobj *so, 464 struct nouveau_bo *bo, unsigned reloc, 465 unsigned offset, unsigned size) 466{ 467 struct nouveau_grobj *eng2d = nv50->screen->eng2d; 468 469 reloc |= NOUVEAU_BO_WR; 470 471 so_method(so, eng2d, NV50_2D_DST_FORMAT, 2); 472 so_data (so, NV50_2D_DST_FORMAT_R8_UNORM); 473 so_data (so, 1); 474 so_method(so, eng2d, NV50_2D_DST_PITCH, 5); 475 so_data (so, 262144); 476 so_data (so, 65536); 477 so_data (so, 1); 478 so_reloc (so, bo, offset, reloc | NOUVEAU_BO_HIGH, 0, 0); 479 so_reloc (so, bo, offset, reloc | NOUVEAU_BO_LOW, 0, 0); 480 so_method(so, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2); 481 so_data (so, 0); 482 so_data (so, NV50_2D_SIFC_FORMAT_R8_UNORM); 483 so_method(so, eng2d, NV50_2D_SIFC_WIDTH, 10); 484 so_data (so, size); 485 so_data (so, 1); 486 so_data (so, 0); 487 so_data (so, 1); 488 so_data (so, 0); 489 so_data (so, 1); 490 so_data (so, 0); 491 so_data (so, 0); 492 so_data (so, 0); 493 so_data (so, 0); 494} 495