nv50_state_validate.c revision 67479eea365954396679714901c11175253bb13b
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 "nv50_context.h" 24#include "nouveau/nouveau_stateobj.h" 25 26#define NV50_CBUF_FORMAT_CASE(n) \ 27 case PIPE_FORMAT_##n: so_data(so, NV50TCL_RT_FORMAT_##n); break 28 29#define NV50_ZETA_FORMAT_CASE(n) \ 30 case PIPE_FORMAT_##n: so_data(so, NV50TCL_ZETA_FORMAT_##n); break 31 32static void 33nv50_state_validate_fb(struct nv50_context *nv50) 34{ 35 struct nouveau_grobj *tesla = nv50->screen->tesla; 36 struct nouveau_stateobj *so = so_new(32, 79, 18); 37 struct pipe_framebuffer_state *fb = &nv50->framebuffer; 38 unsigned i, w, h, gw = 0; 39 40 /* Set nr of active RTs and select RT for each colour output. 41 * FP result 0 always goes to RT[0], bits 4 - 6 are ignored. 42 * Ambiguous assignment results in no rendering (no DATA_ERROR). 43 */ 44 so_method(so, tesla, NV50TCL_RT_CONTROL, 1); 45 so_data (so, fb->nr_cbufs | 46 (0 << 4) | (1 << 7) | (2 << 10) | (3 << 13) | 47 (4 << 16) | (5 << 19) | (6 << 22) | (7 << 25)); 48 49 for (i = 0; i < fb->nr_cbufs; i++) { 50 struct pipe_texture *pt = fb->cbufs[i]->texture; 51 struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; 52 53 if (!gw) { 54 w = fb->cbufs[i]->width; 55 h = fb->cbufs[i]->height; 56 gw = 1; 57 } else { 58 assert(w == fb->cbufs[i]->width); 59 assert(h == fb->cbufs[i]->height); 60 } 61 62 so_method(so, tesla, NV50TCL_RT_HORIZ(i), 2); 63 so_data (so, fb->cbufs[i]->width); 64 so_data (so, fb->cbufs[i]->height); 65 66 so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5); 67 so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | 68 NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); 69 so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM | 70 NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); 71 switch (fb->cbufs[i]->format) { 72 NV50_CBUF_FORMAT_CASE(A8R8G8B8_UNORM); 73 NV50_CBUF_FORMAT_CASE(X8R8G8B8_UNORM); 74 NV50_CBUF_FORMAT_CASE(R5G6B5_UNORM); 75 NV50_CBUF_FORMAT_CASE(R16G16B16A16_SNORM); 76 NV50_CBUF_FORMAT_CASE(R16G16B16A16_UNORM); 77 NV50_CBUF_FORMAT_CASE(R32G32B32A32_FLOAT); 78 NV50_CBUF_FORMAT_CASE(R16G16_SNORM); 79 NV50_CBUF_FORMAT_CASE(R16G16_UNORM); 80 default: 81 NOUVEAU_ERR("AIIII unknown format %s\n", 82 pf_name(fb->cbufs[i]->format)); 83 so_data(so, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM); 84 break; 85 } 86 so_data(so, nv50_miptree(pt)-> 87 level[fb->cbufs[i]->level].tile_mode << 4); 88 so_data(so, 0x00000000); 89 90 so_method(so, tesla, NV50TCL_RT_ARRAY_MODE, 1); 91 so_data (so, 1); 92 } 93 94 if (fb->zsbuf) { 95 struct pipe_texture *pt = fb->zsbuf->texture; 96 struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; 97 98 if (!gw) { 99 w = fb->zsbuf->width; 100 h = fb->zsbuf->height; 101 gw = 1; 102 } else { 103 assert(w == fb->zsbuf->width); 104 assert(h == fb->zsbuf->height); 105 } 106 107 so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5); 108 so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | 109 NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0); 110 so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | 111 NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); 112 switch (fb->zsbuf->format) { 113 NV50_ZETA_FORMAT_CASE(S8Z24_UNORM); 114 NV50_ZETA_FORMAT_CASE(X8Z24_UNORM); 115 NV50_ZETA_FORMAT_CASE(Z24S8_UNORM); 116 NV50_ZETA_FORMAT_CASE(Z32_FLOAT); 117 default: 118 NOUVEAU_ERR("AIIII unknown format %s\n", 119 pf_name(fb->zsbuf->format)); 120 so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM); 121 break; 122 } 123 so_data(so, nv50_miptree(pt)-> 124 level[fb->zsbuf->level].tile_mode << 4); 125 so_data(so, 0x00000000); 126 127 so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); 128 so_data (so, 1); 129 so_method(so, tesla, NV50TCL_ZETA_HORIZ, 3); 130 so_data (so, fb->zsbuf->width); 131 so_data (so, fb->zsbuf->height); 132 so_data (so, 0x00010001); 133 } else { 134 so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); 135 so_data (so, 0); 136 } 137 138 so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ(0), 2); 139 so_data (so, w << 16); 140 so_data (so, h << 16); 141 /* set window lower left corner */ 142 so_method(so, tesla, NV50TCL_WINDOW_OFFSET_X, 2); 143 so_data (so, 0); 144 so_data (so, 0); 145 /* set screen scissor rectangle */ 146 so_method(so, tesla, NV50TCL_SCREEN_SCISSOR_HORIZ, 2); 147 so_data (so, w << 16); 148 so_data (so, h << 16); 149 150 /* we set scissors to framebuffer size when they're 'turned off' */ 151 nv50->dirty |= NV50_NEW_SCISSOR; 152 so_ref(NULL, &nv50->state.scissor); 153 154 so_ref(so, &nv50->state.fb); 155 so_ref(NULL, &so); 156} 157 158static void 159nv50_validate_samplers(struct nv50_context *nv50, struct nouveau_stateobj *so, 160 unsigned p) 161{ 162 struct nouveau_grobj *eng2d = nv50->screen->eng2d; 163 unsigned i, j, dw = nv50->sampler_nr[p] * 8; 164 165 if (!dw) 166 return; 167 nv50_so_init_sifc(nv50, so, nv50->screen->tsc, NOUVEAU_BO_VRAM, 168 p * (32 * 8 * 4), dw * 4); 169 170 so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), dw); 171 172 for (i = 0; i < nv50->sampler_nr[p]; ++i) { 173 if (nv50->sampler[p][i]) 174 so_datap(so, nv50->sampler[p][i]->tsc, 8); 175 else { 176 for (j = 0; j < 8; ++j) /* you get punished */ 177 so_data(so, 0); /* ... for leaving holes */ 178 } 179 } 180} 181 182static void 183nv50_state_emit(struct nv50_context *nv50) 184{ 185 struct nv50_screen *screen = nv50->screen; 186 struct nouveau_channel *chan = screen->base.channel; 187 188 /* XXX: this is racy for multiple contexts active on separate 189 * threads. 190 */ 191 if (screen->cur_ctx != nv50) { 192 if (nv50->state.fb) 193 nv50->state.dirty |= NV50_NEW_FRAMEBUFFER; 194 if (nv50->state.blend) 195 nv50->state.dirty |= NV50_NEW_BLEND; 196 if (nv50->state.zsa) 197 nv50->state.dirty |= NV50_NEW_ZSA; 198 if (nv50->state.vertprog) 199 nv50->state.dirty |= NV50_NEW_VERTPROG; 200 if (nv50->state.fragprog) 201 nv50->state.dirty |= NV50_NEW_FRAGPROG; 202 if (nv50->state.geomprog) 203 nv50->state.dirty |= NV50_NEW_GEOMPROG; 204 if (nv50->state.rast) 205 nv50->state.dirty |= NV50_NEW_RASTERIZER; 206 if (nv50->state.blend_colour) 207 nv50->state.dirty |= NV50_NEW_BLEND_COLOUR; 208 if (nv50->state.stencil_ref) 209 nv50->state.dirty |= NV50_NEW_STENCIL_REF; 210 if (nv50->state.stipple) 211 nv50->state.dirty |= NV50_NEW_STIPPLE; 212 if (nv50->state.scissor) 213 nv50->state.dirty |= NV50_NEW_SCISSOR; 214 if (nv50->state.viewport) 215 nv50->state.dirty |= NV50_NEW_VIEWPORT; 216 if (nv50->state.tsc_upload) 217 nv50->state.dirty |= NV50_NEW_SAMPLER; 218 if (nv50->state.tic_upload) 219 nv50->state.dirty |= NV50_NEW_TEXTURE; 220 if (nv50->state.vtxfmt && nv50->state.vtxbuf) 221 nv50->state.dirty |= NV50_NEW_ARRAYS; 222 screen->cur_ctx = nv50; 223 } 224 225 if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER) 226 so_emit(chan, nv50->state.fb); 227 if (nv50->state.dirty & NV50_NEW_BLEND) 228 so_emit(chan, nv50->state.blend); 229 if (nv50->state.dirty & NV50_NEW_ZSA) 230 so_emit(chan, nv50->state.zsa); 231 if (nv50->state.dirty & NV50_NEW_VERTPROG) 232 so_emit(chan, nv50->state.vertprog); 233 if (nv50->state.dirty & NV50_NEW_FRAGPROG) 234 so_emit(chan, nv50->state.fragprog); 235 if (nv50->state.dirty & NV50_NEW_GEOMPROG && nv50->state.geomprog) 236 so_emit(chan, nv50->state.geomprog); 237 if (nv50->state.dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | 238 NV50_NEW_GEOMPROG | NV50_NEW_RASTERIZER)) 239 so_emit(chan, nv50->state.fp_linkage); 240 if ((nv50->state.dirty & (NV50_NEW_VERTPROG | NV50_NEW_GEOMPROG)) 241 && nv50->state.gp_linkage) 242 so_emit(chan, nv50->state.gp_linkage); 243 if (nv50->state.dirty & NV50_NEW_RASTERIZER) 244 so_emit(chan, nv50->state.rast); 245 if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR) 246 so_emit(chan, nv50->state.blend_colour); 247 if (nv50->state.dirty & NV50_NEW_STENCIL_REF) 248 so_emit(chan, nv50->state.stencil_ref); 249 if (nv50->state.dirty & NV50_NEW_STIPPLE) 250 so_emit(chan, nv50->state.stipple); 251 if (nv50->state.dirty & NV50_NEW_SCISSOR) 252 so_emit(chan, nv50->state.scissor); 253 if (nv50->state.dirty & NV50_NEW_VIEWPORT) 254 so_emit(chan, nv50->state.viewport); 255 if (nv50->state.dirty & NV50_NEW_SAMPLER) 256 so_emit(chan, nv50->state.tsc_upload); 257 if (nv50->state.dirty & NV50_NEW_TEXTURE) 258 so_emit(chan, nv50->state.tic_upload); 259 if (nv50->state.dirty & NV50_NEW_ARRAYS) { 260 so_emit(chan, nv50->state.vtxfmt); 261 so_emit(chan, nv50->state.vtxbuf); 262 if (nv50->state.vtxattr) 263 so_emit(chan, nv50->state.vtxattr); 264 } 265 nv50->state.dirty = 0; 266} 267 268void 269nv50_state_flush_notify(struct nouveau_channel *chan) 270{ 271 struct nv50_context *nv50 = chan->user_private; 272 273 if (nv50->state.tic_upload && !(nv50->dirty & NV50_NEW_TEXTURE)) 274 so_emit(chan, nv50->state.tic_upload); 275 276 so_emit_reloc_markers(chan, nv50->state.fb); 277 so_emit_reloc_markers(chan, nv50->state.vertprog); 278 so_emit_reloc_markers(chan, nv50->state.fragprog); 279 so_emit_reloc_markers(chan, nv50->state.vtxbuf); 280 so_emit_reloc_markers(chan, nv50->screen->static_init); 281 282 if (nv50->state.instbuf) 283 so_emit_reloc_markers(chan, nv50->state.instbuf); 284} 285 286boolean 287nv50_state_validate(struct nv50_context *nv50) 288{ 289 struct nouveau_grobj *tesla = nv50->screen->tesla; 290 struct nouveau_stateobj *so; 291 unsigned i; 292 293 if (nv50->dirty & NV50_NEW_FRAMEBUFFER) 294 nv50_state_validate_fb(nv50); 295 296 if (nv50->dirty & NV50_NEW_BLEND) 297 so_ref(nv50->blend->so, &nv50->state.blend); 298 299 if (nv50->dirty & NV50_NEW_ZSA) 300 so_ref(nv50->zsa->so, &nv50->state.zsa); 301 302 if (nv50->dirty & (NV50_NEW_VERTPROG | NV50_NEW_VERTPROG_CB)) 303 nv50_vertprog_validate(nv50); 304 305 if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB)) 306 nv50_fragprog_validate(nv50); 307 308 if (nv50->dirty & (NV50_NEW_GEOMPROG | NV50_NEW_GEOMPROG_CB)) 309 nv50_geomprog_validate(nv50); 310 311 if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | 312 NV50_NEW_GEOMPROG | NV50_NEW_RASTERIZER)) 313 nv50_fp_linkage_validate(nv50); 314 315 if (nv50->dirty & (NV50_NEW_GEOMPROG | NV50_NEW_VERTPROG)) 316 nv50_gp_linkage_validate(nv50); 317 318 if (nv50->dirty & NV50_NEW_RASTERIZER) 319 so_ref(nv50->rasterizer->so, &nv50->state.rast); 320 321 if (nv50->dirty & NV50_NEW_BLEND_COLOUR) { 322 so = so_new(1, 4, 0); 323 so_method(so, tesla, NV50TCL_BLEND_COLOR(0), 4); 324 so_data (so, fui(nv50->blend_colour.color[0])); 325 so_data (so, fui(nv50->blend_colour.color[1])); 326 so_data (so, fui(nv50->blend_colour.color[2])); 327 so_data (so, fui(nv50->blend_colour.color[3])); 328 so_ref(so, &nv50->state.blend_colour); 329 so_ref(NULL, &so); 330 } 331 332 if (nv50->dirty & NV50_NEW_STENCIL_REF) { 333 so = so_new(2, 2, 0); 334 so_method(so, tesla, NV50TCL_STENCIL_FRONT_FUNC_REF, 1); 335 so_data (so, nv50->stencil_ref.ref_value[0]; 336 so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 1); 337 so_data (so, nv50->stencil_ref.ref_value[1]; 338 so_ref(so, &nv50->state.stencil_ref); 339 so_ref(NULL, &so); 340 } 341 342 if (nv50->dirty & NV50_NEW_STIPPLE) { 343 so = so_new(1, 32, 0); 344 so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32); 345 for (i = 0; i < 32; i++) 346 so_data(so, util_bswap32(nv50->stipple.stipple[i])); 347 so_ref(so, &nv50->state.stipple); 348 so_ref(NULL, &so); 349 } 350 351 if (nv50->dirty & (NV50_NEW_SCISSOR | NV50_NEW_RASTERIZER)) { 352 struct pipe_rasterizer_state *rast = &nv50->rasterizer->pipe; 353 struct pipe_scissor_state *s = &nv50->scissor; 354 355 if (nv50->state.scissor && 356 (rast->scissor == 0 && nv50->state.scissor_enabled == 0)) 357 goto scissor_uptodate; 358 nv50->state.scissor_enabled = rast->scissor; 359 360 so = so_new(1, 2, 0); 361 so_method(so, tesla, NV50TCL_SCISSOR_HORIZ(0), 2); 362 if (nv50->state.scissor_enabled) { 363 so_data(so, (s->maxx << 16) | s->minx); 364 so_data(so, (s->maxy << 16) | s->miny); 365 } else { 366 so_data(so, (nv50->framebuffer.width << 16)); 367 so_data(so, (nv50->framebuffer.height << 16)); 368 } 369 so_ref(so, &nv50->state.scissor); 370 so_ref(NULL, &so); 371 nv50->state.dirty |= NV50_NEW_SCISSOR; 372 } 373scissor_uptodate: 374 375 if (nv50->dirty & (NV50_NEW_VIEWPORT | NV50_NEW_RASTERIZER)) { 376 unsigned bypass; 377 378 if (!nv50->rasterizer->pipe.bypass_vs_clip_and_viewport) 379 bypass = 0; 380 else 381 bypass = 1; 382 383 if (nv50->state.viewport && 384 (bypass || !(nv50->dirty & NV50_NEW_VIEWPORT)) && 385 nv50->state.viewport_bypass == bypass) 386 goto viewport_uptodate; 387 nv50->state.viewport_bypass = bypass; 388 389 so = so_new(5, 9, 0); 390 if (!bypass) { 391 so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE_X(0), 3); 392 so_data (so, fui(nv50->viewport.translate[0])); 393 so_data (so, fui(nv50->viewport.translate[1])); 394 so_data (so, fui(nv50->viewport.translate[2])); 395 so_method(so, tesla, NV50TCL_VIEWPORT_SCALE_X(0), 3); 396 so_data (so, fui(nv50->viewport.scale[0])); 397 so_data (so, fui(nv50->viewport.scale[1])); 398 so_data (so, fui(nv50->viewport.scale[2])); 399 400 so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1); 401 so_data (so, 1); 402 /* 0x0000 = remove whole primitive only (xyz) 403 * 0x1018 = remove whole primitive only (xy), clamp z 404 * 0x1080 = clip primitive (xyz) 405 * 0x1098 = clip primitive (xy), clamp z 406 */ 407 so_method(so, tesla, NV50TCL_VIEW_VOLUME_CLIP_CTRL, 1); 408 so_data (so, 0x1080); 409 /* no idea what 0f90 does */ 410 so_method(so, tesla, 0x0f90, 1); 411 so_data (so, 0); 412 } else { 413 so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1); 414 so_data (so, 0); 415 so_method(so, tesla, NV50TCL_VIEW_VOLUME_CLIP_CTRL, 1); 416 so_data (so, 0x0000); 417 so_method(so, tesla, 0x0f90, 1); 418 so_data (so, 1); 419 } 420 421 so_ref(so, &nv50->state.viewport); 422 so_ref(NULL, &so); 423 nv50->state.dirty |= NV50_NEW_VIEWPORT; 424 } 425viewport_uptodate: 426 427 if (nv50->dirty & NV50_NEW_SAMPLER) { 428 unsigned nr = 0; 429 430 for (i = 0; i < PIPE_SHADER_TYPES; ++i) 431 nr += nv50->sampler_nr[i]; 432 433 so = so_new(1 + 5 * PIPE_SHADER_TYPES, 434 1 + 19 * PIPE_SHADER_TYPES + nr * 8, 435 PIPE_SHADER_TYPES * 2); 436 437 nv50_validate_samplers(nv50, so, PIPE_SHADER_VERTEX); 438 nv50_validate_samplers(nv50, so, PIPE_SHADER_FRAGMENT); 439 440 so_method(so, tesla, 0x1334, 1); /* flush TSC */ 441 so_data (so, 0); 442 443 so_ref(so, &nv50->state.tsc_upload); 444 so_ref(NULL, &so); 445 } 446 447 if (nv50->dirty & (NV50_NEW_TEXTURE | NV50_NEW_SAMPLER)) 448 nv50_tex_validate(nv50); 449 450 if (nv50->dirty & NV50_NEW_ARRAYS) 451 nv50_vbo_validate(nv50); 452 453 nv50->state.dirty |= nv50->dirty; 454 nv50->dirty = 0; 455 nv50_state_emit(nv50); 456 457 return TRUE; 458} 459 460void nv50_so_init_sifc(struct nv50_context *nv50, 461 struct nouveau_stateobj *so, 462 struct nouveau_bo *bo, unsigned reloc, 463 unsigned offset, unsigned size) 464{ 465 struct nouveau_grobj *eng2d = nv50->screen->eng2d; 466 467 reloc |= NOUVEAU_BO_WR; 468 469 so_method(so, eng2d, NV50_2D_DST_FORMAT, 2); 470 so_data (so, NV50_2D_DST_FORMAT_R8_UNORM); 471 so_data (so, 1); 472 so_method(so, eng2d, NV50_2D_DST_PITCH, 5); 473 so_data (so, 262144); 474 so_data (so, 65536); 475 so_data (so, 1); 476 so_reloc (so, bo, offset, reloc | NOUVEAU_BO_HIGH, 0, 0); 477 so_reloc (so, bo, offset, reloc | NOUVEAU_BO_LOW, 0, 0); 478 so_method(so, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2); 479 so_data (so, 0); 480 so_data (so, NV50_2D_SIFC_FORMAT_R8_UNORM); 481 so_method(so, eng2d, NV50_2D_SIFC_WIDTH, 10); 482 so_data (so, size); 483 so_data (so, 1); 484 so_data (so, 0); 485 so_data (so, 1); 486 so_data (so, 0); 487 so_data (so, 1); 488 so_data (so, 0); 489 so_data (so, 0); 490 so_data (so, 0); 491 so_data (so, 0); 492} 493