1 /*
2 * Copyright 2010 Christoph Bumiller
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 OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 */
22
23 #include "pipe/p_defines.h"
24 #include "util/u_framebuffer.h"
25 #include "util/u_helpers.h"
26 #include "util/u_inlines.h"
27 #include "util/u_transfer.h"
28 #include "util/format_srgb.h"
29
30 #include "tgsi/tgsi_parse.h"
31 #include "compiler/nir/nir.h"
32
33 #include "nv50/nv50_stateobj.h"
34 #include "nv50/nv50_context.h"
35 #include "nv50/nv50_query_hw.h"
36
37 #include "nv50/nv50_3d.xml.h"
38 #include "nv50/g80_texture.xml.h"
39
40 #include "nouveau_gldefs.h"
41
42 /* Caveats:
43 * ! pipe_sampler_state.normalized_coords is ignored - rectangle textures will
44 * use non-normalized coordinates, everything else won't
45 * (The relevant bit is in the TIC entry and not the TSC entry.)
46 *
47 * ! pipe_sampler_state.seamless_cube_map is ignored - seamless filtering is
48 * always activated on NVA0 +
49 * (Give me the global bit, otherwise it's not worth the CPU work.)
50 *
51 * ! pipe_sampler_state.border_color is not swizzled according to the texture
52 * swizzle in pipe_sampler_view
53 * (This will be ugly with indirect independent texture/sampler access,
54 * we'd have to emulate the logic in the shader. GL doesn't have that,
55 * D3D doesn't have swizzle, if we knew what we were implementing we'd be
56 * good.)
57 *
58 * ! pipe_rasterizer_state.line_last_pixel is ignored - it is never drawn
59 *
60 * ! pipe_rasterizer_state.flatshade_first also applies to QUADS
61 * (There's a GL query for that, forcing an exception is just ridiculous.)
62 *
63 * ! pipe_rasterizer_state.sprite_coord_enable is masked with 0xff on NVC0
64 * (The hardware only has 8 slots meant for TexCoord and we have to assign
65 * in advance to maintain elegant separate shader objects.)
66 */
67
68 static inline uint32_t
nv50_colormask(unsigned mask)69 nv50_colormask(unsigned mask)
70 {
71 uint32_t ret = 0;
72
73 if (mask & PIPE_MASK_R)
74 ret |= 0x0001;
75 if (mask & PIPE_MASK_G)
76 ret |= 0x0010;
77 if (mask & PIPE_MASK_B)
78 ret |= 0x0100;
79 if (mask & PIPE_MASK_A)
80 ret |= 0x1000;
81
82 return ret;
83 }
84
85 #define NV50_BLEND_FACTOR_CASE(a, b) \
86 case PIPE_BLENDFACTOR_##a: return NV50_BLEND_FACTOR_##b
87
88 static inline uint32_t
nv50_blend_fac(unsigned factor)89 nv50_blend_fac(unsigned factor)
90 {
91 switch (factor) {
92 NV50_BLEND_FACTOR_CASE(ONE, ONE);
93 NV50_BLEND_FACTOR_CASE(SRC_COLOR, SRC_COLOR);
94 NV50_BLEND_FACTOR_CASE(SRC_ALPHA, SRC_ALPHA);
95 NV50_BLEND_FACTOR_CASE(DST_ALPHA, DST_ALPHA);
96 NV50_BLEND_FACTOR_CASE(DST_COLOR, DST_COLOR);
97 NV50_BLEND_FACTOR_CASE(SRC_ALPHA_SATURATE, SRC_ALPHA_SATURATE);
98 NV50_BLEND_FACTOR_CASE(CONST_COLOR, CONSTANT_COLOR);
99 NV50_BLEND_FACTOR_CASE(CONST_ALPHA, CONSTANT_ALPHA);
100 NV50_BLEND_FACTOR_CASE(SRC1_COLOR, SRC1_COLOR);
101 NV50_BLEND_FACTOR_CASE(SRC1_ALPHA, SRC1_ALPHA);
102 NV50_BLEND_FACTOR_CASE(ZERO, ZERO);
103 NV50_BLEND_FACTOR_CASE(INV_SRC_COLOR, ONE_MINUS_SRC_COLOR);
104 NV50_BLEND_FACTOR_CASE(INV_SRC_ALPHA, ONE_MINUS_SRC_ALPHA);
105 NV50_BLEND_FACTOR_CASE(INV_DST_ALPHA, ONE_MINUS_DST_ALPHA);
106 NV50_BLEND_FACTOR_CASE(INV_DST_COLOR, ONE_MINUS_DST_COLOR);
107 NV50_BLEND_FACTOR_CASE(INV_CONST_COLOR, ONE_MINUS_CONSTANT_COLOR);
108 NV50_BLEND_FACTOR_CASE(INV_CONST_ALPHA, ONE_MINUS_CONSTANT_ALPHA);
109 NV50_BLEND_FACTOR_CASE(INV_SRC1_COLOR, ONE_MINUS_SRC1_COLOR);
110 NV50_BLEND_FACTOR_CASE(INV_SRC1_ALPHA, ONE_MINUS_SRC1_ALPHA);
111 default:
112 return NV50_BLEND_FACTOR_ZERO;
113 }
114 }
115
116 static void *
nv50_blend_state_create(struct pipe_context * pipe,const struct pipe_blend_state * cso)117 nv50_blend_state_create(struct pipe_context *pipe,
118 const struct pipe_blend_state *cso)
119 {
120 struct nv50_blend_stateobj *so = CALLOC_STRUCT(nv50_blend_stateobj);
121 int i;
122 bool emit_common_func = cso->rt[0].blend_enable;
123 uint32_t ms;
124
125 if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
126 SB_BEGIN_3D(so, BLEND_INDEPENDENT, 1);
127 SB_DATA (so, cso->independent_blend_enable);
128 }
129
130 so->pipe = *cso;
131
132 SB_BEGIN_3D(so, COLOR_MASK_COMMON, 1);
133 SB_DATA (so, !cso->independent_blend_enable);
134
135 SB_BEGIN_3D(so, BLEND_ENABLE_COMMON, 1);
136 SB_DATA (so, !cso->independent_blend_enable);
137
138 if (cso->independent_blend_enable) {
139 SB_BEGIN_3D(so, BLEND_ENABLE(0), 8);
140 for (i = 0; i < 8; ++i) {
141 SB_DATA(so, cso->rt[i].blend_enable);
142 if (cso->rt[i].blend_enable)
143 emit_common_func = true;
144 }
145
146 if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
147 emit_common_func = false;
148
149 for (i = 0; i < 8; ++i) {
150 if (!cso->rt[i].blend_enable)
151 continue;
152 SB_BEGIN_3D_(so, NVA3_3D_IBLEND_EQUATION_RGB(i), 6);
153 SB_DATA (so, nvgl_blend_eqn(cso->rt[i].rgb_func));
154 SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_src_factor));
155 SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_dst_factor));
156 SB_DATA (so, nvgl_blend_eqn(cso->rt[i].alpha_func));
157 SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_src_factor));
158 SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_dst_factor));
159 }
160 }
161 } else {
162 SB_BEGIN_3D(so, BLEND_ENABLE(0), 1);
163 SB_DATA (so, cso->rt[0].blend_enable);
164 }
165
166 if (emit_common_func) {
167 SB_BEGIN_3D(so, BLEND_EQUATION_RGB, 5);
168 SB_DATA (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
169 SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_src_factor));
170 SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_dst_factor));
171 SB_DATA (so, nvgl_blend_eqn(cso->rt[0].alpha_func));
172 SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_src_factor));
173 SB_BEGIN_3D(so, BLEND_FUNC_DST_ALPHA, 1);
174 SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_dst_factor));
175 }
176
177 if (cso->logicop_enable) {
178 SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 2);
179 SB_DATA (so, 1);
180 SB_DATA (so, nvgl_logicop_func(cso->logicop_func));
181 } else {
182 SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 1);
183 SB_DATA (so, 0);
184 }
185
186 if (cso->independent_blend_enable) {
187 SB_BEGIN_3D(so, COLOR_MASK(0), 8);
188 for (i = 0; i < 8; ++i)
189 SB_DATA(so, nv50_colormask(cso->rt[i].colormask));
190 } else {
191 SB_BEGIN_3D(so, COLOR_MASK(0), 1);
192 SB_DATA (so, nv50_colormask(cso->rt[0].colormask));
193 }
194
195 ms = 0;
196 if (cso->alpha_to_coverage)
197 ms |= NV50_3D_MULTISAMPLE_CTRL_ALPHA_TO_COVERAGE;
198 if (cso->alpha_to_one)
199 ms |= NV50_3D_MULTISAMPLE_CTRL_ALPHA_TO_ONE;
200
201 SB_BEGIN_3D(so, MULTISAMPLE_CTRL, 1);
202 SB_DATA (so, ms);
203
204 assert(so->size <= ARRAY_SIZE(so->state));
205 return so;
206 }
207
208 static void
nv50_blend_state_bind(struct pipe_context * pipe,void * hwcso)209 nv50_blend_state_bind(struct pipe_context *pipe, void *hwcso)
210 {
211 struct nv50_context *nv50 = nv50_context(pipe);
212
213 nv50->blend = hwcso;
214 nv50->dirty_3d |= NV50_NEW_3D_BLEND;
215 }
216
217 static void
nv50_blend_state_delete(struct pipe_context * pipe,void * hwcso)218 nv50_blend_state_delete(struct pipe_context *pipe, void *hwcso)
219 {
220 FREE(hwcso);
221 }
222
223 /* NOTE: ignoring line_last_pixel */
224 static void *
nv50_rasterizer_state_create(struct pipe_context * pipe,const struct pipe_rasterizer_state * cso)225 nv50_rasterizer_state_create(struct pipe_context *pipe,
226 const struct pipe_rasterizer_state *cso)
227 {
228 struct nv50_rasterizer_stateobj *so;
229 uint32_t reg;
230
231 so = CALLOC_STRUCT(nv50_rasterizer_stateobj);
232 if (!so)
233 return NULL;
234 so->pipe = *cso;
235
236 #ifndef NV50_SCISSORS_CLIPPING
237 for (int i = 0; i < NV50_MAX_VIEWPORTS; i++) {
238 SB_BEGIN_3D(so, SCISSOR_ENABLE(i), 1);
239 SB_DATA (so, cso->scissor);
240 }
241 #endif
242
243 SB_BEGIN_3D(so, SHADE_MODEL, 1);
244 SB_DATA (so, cso->flatshade ? NV50_3D_SHADE_MODEL_FLAT :
245 NV50_3D_SHADE_MODEL_SMOOTH);
246 SB_BEGIN_3D(so, PROVOKING_VERTEX_LAST, 1);
247 SB_DATA (so, !cso->flatshade_first);
248 SB_BEGIN_3D(so, VERTEX_TWO_SIDE_ENABLE, 1);
249 SB_DATA (so, cso->light_twoside);
250
251 SB_BEGIN_3D(so, FRAG_COLOR_CLAMP_EN, 1);
252 SB_DATA (so, cso->clamp_fragment_color ? 0x11111111 : 0x00000000);
253
254 SB_BEGIN_3D(so, MULTISAMPLE_ENABLE, 1);
255 SB_DATA (so, cso->multisample);
256
257 SB_BEGIN_3D(so, LINE_WIDTH, 1);
258 SB_DATA (so, fui(cso->line_width));
259 SB_BEGIN_3D(so, LINE_SMOOTH_ENABLE, 1);
260 SB_DATA (so, cso->line_smooth);
261
262 SB_BEGIN_3D(so, LINE_STIPPLE_ENABLE, 1);
263 if (cso->line_stipple_enable) {
264 SB_DATA (so, 1);
265 SB_BEGIN_3D(so, LINE_STIPPLE, 1);
266 SB_DATA (so, (cso->line_stipple_pattern << 8) |
267 cso->line_stipple_factor);
268 } else {
269 SB_DATA (so, 0);
270 }
271
272 if (!cso->point_size_per_vertex) {
273 SB_BEGIN_3D(so, POINT_SIZE, 1);
274 SB_DATA (so, fui(cso->point_size));
275 }
276 SB_BEGIN_3D(so, POINT_SPRITE_ENABLE, 1);
277 SB_DATA (so, cso->point_quad_rasterization);
278 SB_BEGIN_3D(so, POINT_SMOOTH_ENABLE, 1);
279 SB_DATA (so, cso->point_smooth);
280
281 SB_BEGIN_3D(so, POLYGON_MODE_FRONT, 3);
282 SB_DATA (so, nvgl_polygon_mode(cso->fill_front));
283 SB_DATA (so, nvgl_polygon_mode(cso->fill_back));
284 SB_DATA (so, cso->poly_smooth);
285
286 SB_BEGIN_3D(so, CULL_FACE_ENABLE, 3);
287 SB_DATA (so, cso->cull_face != PIPE_FACE_NONE);
288 SB_DATA (so, cso->front_ccw ? NV50_3D_FRONT_FACE_CCW :
289 NV50_3D_FRONT_FACE_CW);
290 switch (cso->cull_face) {
291 case PIPE_FACE_FRONT_AND_BACK:
292 SB_DATA(so, NV50_3D_CULL_FACE_FRONT_AND_BACK);
293 break;
294 case PIPE_FACE_FRONT:
295 SB_DATA(so, NV50_3D_CULL_FACE_FRONT);
296 break;
297 case PIPE_FACE_BACK:
298 default:
299 SB_DATA(so, NV50_3D_CULL_FACE_BACK);
300 break;
301 }
302
303 SB_BEGIN_3D(so, POLYGON_STIPPLE_ENABLE, 1);
304 SB_DATA (so, cso->poly_stipple_enable);
305 SB_BEGIN_3D(so, POLYGON_OFFSET_POINT_ENABLE, 3);
306 SB_DATA (so, cso->offset_point);
307 SB_DATA (so, cso->offset_line);
308 SB_DATA (so, cso->offset_tri);
309
310 if (cso->offset_point || cso->offset_line || cso->offset_tri) {
311 SB_BEGIN_3D(so, POLYGON_OFFSET_FACTOR, 1);
312 SB_DATA (so, fui(cso->offset_scale));
313 SB_BEGIN_3D(so, POLYGON_OFFSET_UNITS, 1);
314 SB_DATA (so, fui(cso->offset_units * 2.0f));
315 SB_BEGIN_3D(so, POLYGON_OFFSET_CLAMP, 1);
316 SB_DATA (so, fui(cso->offset_clamp));
317 }
318
319 if (cso->depth_clip_near) {
320 reg = 0;
321 } else {
322 reg =
323 NV50_3D_VIEW_VOLUME_CLIP_CTRL_DEPTH_CLAMP_NEAR |
324 NV50_3D_VIEW_VOLUME_CLIP_CTRL_DEPTH_CLAMP_FAR |
325 NV50_3D_VIEW_VOLUME_CLIP_CTRL_UNK12_UNK1;
326 }
327 #ifndef NV50_SCISSORS_CLIPPING
328 reg |=
329 NV50_3D_VIEW_VOLUME_CLIP_CTRL_UNK7 |
330 NV50_3D_VIEW_VOLUME_CLIP_CTRL_UNK12_UNK1;
331 #endif
332 SB_BEGIN_3D(so, VIEW_VOLUME_CLIP_CTRL, 1);
333 SB_DATA (so, reg);
334
335 SB_BEGIN_3D(so, DEPTH_CLIP_NEGATIVE_Z, 1);
336 SB_DATA (so, cso->clip_halfz);
337
338 SB_BEGIN_3D(so, PIXEL_CENTER_INTEGER, 1);
339 SB_DATA (so, !cso->half_pixel_center);
340
341 assert(so->size <= ARRAY_SIZE(so->state));
342 return (void *)so;
343 }
344
345 static void
nv50_rasterizer_state_bind(struct pipe_context * pipe,void * hwcso)346 nv50_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
347 {
348 struct nv50_context *nv50 = nv50_context(pipe);
349
350 nv50->rast = hwcso;
351 nv50->dirty_3d |= NV50_NEW_3D_RASTERIZER;
352 }
353
354 static void
nv50_rasterizer_state_delete(struct pipe_context * pipe,void * hwcso)355 nv50_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
356 {
357 FREE(hwcso);
358 }
359
360 static void *
nv50_zsa_state_create(struct pipe_context * pipe,const struct pipe_depth_stencil_alpha_state * cso)361 nv50_zsa_state_create(struct pipe_context *pipe,
362 const struct pipe_depth_stencil_alpha_state *cso)
363 {
364 struct nv50_zsa_stateobj *so = CALLOC_STRUCT(nv50_zsa_stateobj);
365
366 so->pipe = *cso;
367
368 SB_BEGIN_3D(so, DEPTH_WRITE_ENABLE, 1);
369 SB_DATA (so, cso->depth.writemask);
370 SB_BEGIN_3D(so, DEPTH_TEST_ENABLE, 1);
371 if (cso->depth.enabled) {
372 SB_DATA (so, 1);
373 SB_BEGIN_3D(so, DEPTH_TEST_FUNC, 1);
374 SB_DATA (so, nvgl_comparison_op(cso->depth.func));
375 } else {
376 SB_DATA (so, 0);
377 }
378
379 SB_BEGIN_3D(so, DEPTH_BOUNDS_EN, 1);
380 if (cso->depth.bounds_test) {
381 SB_DATA (so, 1);
382 SB_BEGIN_3D(so, DEPTH_BOUNDS(0), 2);
383 SB_DATA (so, fui(cso->depth.bounds_min));
384 SB_DATA (so, fui(cso->depth.bounds_max));
385 } else {
386 SB_DATA (so, 0);
387 }
388
389 if (cso->stencil[0].enabled) {
390 SB_BEGIN_3D(so, STENCIL_ENABLE, 5);
391 SB_DATA (so, 1);
392 SB_DATA (so, nvgl_stencil_op(cso->stencil[0].fail_op));
393 SB_DATA (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
394 SB_DATA (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
395 SB_DATA (so, nvgl_comparison_op(cso->stencil[0].func));
396 SB_BEGIN_3D(so, STENCIL_FRONT_MASK, 2);
397 SB_DATA (so, cso->stencil[0].writemask);
398 SB_DATA (so, cso->stencil[0].valuemask);
399 } else {
400 SB_BEGIN_3D(so, STENCIL_ENABLE, 1);
401 SB_DATA (so, 0);
402 }
403
404 if (cso->stencil[1].enabled) {
405 assert(cso->stencil[0].enabled);
406 SB_BEGIN_3D(so, STENCIL_TWO_SIDE_ENABLE, 5);
407 SB_DATA (so, 1);
408 SB_DATA (so, nvgl_stencil_op(cso->stencil[1].fail_op));
409 SB_DATA (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
410 SB_DATA (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
411 SB_DATA (so, nvgl_comparison_op(cso->stencil[1].func));
412 SB_BEGIN_3D(so, STENCIL_BACK_MASK, 2);
413 SB_DATA (so, cso->stencil[1].writemask);
414 SB_DATA (so, cso->stencil[1].valuemask);
415 } else {
416 SB_BEGIN_3D(so, STENCIL_TWO_SIDE_ENABLE, 1);
417 SB_DATA (so, 0);
418 }
419
420 SB_BEGIN_3D(so, ALPHA_TEST_ENABLE, 1);
421 if (cso->alpha.enabled) {
422 SB_DATA (so, 1);
423 SB_BEGIN_3D(so, ALPHA_TEST_REF, 2);
424 SB_DATA (so, fui(cso->alpha.ref_value));
425 SB_DATA (so, nvgl_comparison_op(cso->alpha.func));
426 } else {
427 SB_DATA (so, 0);
428 }
429
430 SB_BEGIN_3D(so, CB_ADDR, 1);
431 SB_DATA (so, NV50_CB_AUX_ALPHATEST_OFFSET << (8 - 2) | NV50_CB_AUX);
432 SB_BEGIN_3D(so, CB_DATA(0), 1);
433 SB_DATA (so, fui(cso->alpha.ref_value));
434
435 assert(so->size <= ARRAY_SIZE(so->state));
436 return (void *)so;
437 }
438
439 static void
nv50_zsa_state_bind(struct pipe_context * pipe,void * hwcso)440 nv50_zsa_state_bind(struct pipe_context *pipe, void *hwcso)
441 {
442 struct nv50_context *nv50 = nv50_context(pipe);
443
444 nv50->zsa = hwcso;
445 nv50->dirty_3d |= NV50_NEW_3D_ZSA;
446 }
447
448 static void
nv50_zsa_state_delete(struct pipe_context * pipe,void * hwcso)449 nv50_zsa_state_delete(struct pipe_context *pipe, void *hwcso)
450 {
451 FREE(hwcso);
452 }
453
454 /* ====================== SAMPLERS AND TEXTURES ================================
455 */
456
457 static inline unsigned
nv50_tsc_wrap_mode(unsigned wrap)458 nv50_tsc_wrap_mode(unsigned wrap)
459 {
460 switch (wrap) {
461 case PIPE_TEX_WRAP_REPEAT:
462 return G80_TSC_WRAP_WRAP;
463 case PIPE_TEX_WRAP_MIRROR_REPEAT:
464 return G80_TSC_WRAP_MIRROR;
465 case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
466 return G80_TSC_WRAP_CLAMP_TO_EDGE;
467 case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
468 return G80_TSC_WRAP_BORDER;
469 case PIPE_TEX_WRAP_CLAMP:
470 return G80_TSC_WRAP_CLAMP_OGL;
471 case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
472 return G80_TSC_WRAP_MIRROR_ONCE_CLAMP_TO_EDGE;
473 case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
474 return G80_TSC_WRAP_MIRROR_ONCE_BORDER;
475 case PIPE_TEX_WRAP_MIRROR_CLAMP:
476 return G80_TSC_WRAP_MIRROR_ONCE_CLAMP_OGL;
477 default:
478 NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
479 return G80_TSC_WRAP_WRAP;
480 }
481 }
482
483 void *
nv50_sampler_state_create(struct pipe_context * pipe,const struct pipe_sampler_state * cso)484 nv50_sampler_state_create(struct pipe_context *pipe,
485 const struct pipe_sampler_state *cso)
486 {
487 struct nv50_tsc_entry *so = MALLOC_STRUCT(nv50_tsc_entry);
488 float f[2];
489
490 so->id = -1;
491
492 so->tsc[0] = (0x00026000 |
493 (nv50_tsc_wrap_mode(cso->wrap_s) << 0) |
494 (nv50_tsc_wrap_mode(cso->wrap_t) << 3) |
495 (nv50_tsc_wrap_mode(cso->wrap_r) << 6));
496
497 switch (cso->mag_img_filter) {
498 case PIPE_TEX_FILTER_LINEAR:
499 so->tsc[1] = G80_TSC_1_MAG_FILTER_LINEAR;
500 break;
501 case PIPE_TEX_FILTER_NEAREST:
502 default:
503 so->tsc[1] = G80_TSC_1_MAG_FILTER_NEAREST;
504 break;
505 }
506
507 switch (cso->min_img_filter) {
508 case PIPE_TEX_FILTER_LINEAR:
509 so->tsc[1] |= G80_TSC_1_MIN_FILTER_LINEAR;
510 break;
511 case PIPE_TEX_FILTER_NEAREST:
512 default:
513 so->tsc[1] |= G80_TSC_1_MIN_FILTER_NEAREST;
514 break;
515 }
516
517 switch (cso->min_mip_filter) {
518 case PIPE_TEX_MIPFILTER_LINEAR:
519 so->tsc[1] |= G80_TSC_1_MIP_FILTER_LINEAR;
520 break;
521 case PIPE_TEX_MIPFILTER_NEAREST:
522 so->tsc[1] |= G80_TSC_1_MIP_FILTER_NEAREST;
523 break;
524 case PIPE_TEX_MIPFILTER_NONE:
525 default:
526 so->tsc[1] |= G80_TSC_1_MIP_FILTER_NONE;
527 break;
528 }
529
530 if (nouveau_screen(pipe->screen)->class_3d >= NVE4_3D_CLASS) {
531 if (cso->seamless_cube_map)
532 so->tsc[1] |= GK104_TSC_1_CUBEMAP_INTERFACE_FILTERING;
533 if (!cso->normalized_coords)
534 so->tsc[1] |= GK104_TSC_1_FLOAT_COORD_NORMALIZATION_FORCE_UNNORMALIZED_COORDS;
535 } else {
536 so->seamless_cube_map = cso->seamless_cube_map;
537 }
538
539 if (cso->max_anisotropy >= 16)
540 so->tsc[0] |= (7 << 20);
541 else
542 if (cso->max_anisotropy >= 12)
543 so->tsc[0] |= (6 << 20);
544 else {
545 so->tsc[0] |= (cso->max_anisotropy >> 1) << 20;
546
547 if (cso->max_anisotropy >= 4)
548 so->tsc[1] |= 6 << G80_TSC_1_TRILIN_OPT__SHIFT;
549 else
550 if (cso->max_anisotropy >= 2)
551 so->tsc[1] |= 4 << G80_TSC_1_TRILIN_OPT__SHIFT;
552 }
553
554 if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
555 /* NOTE: must be deactivated for non-shadow textures */
556 so->tsc[0] |= (1 << 9);
557 so->tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7) << 10;
558 }
559
560 f[0] = CLAMP(cso->lod_bias, -16.0f, 15.0f);
561 so->tsc[1] |= ((int)(f[0] * 256.0f) & 0x1fff) << 12;
562
563 f[0] = CLAMP(cso->min_lod, 0.0f, 15.0f);
564 f[1] = CLAMP(cso->max_lod, 0.0f, 15.0f);
565 so->tsc[2] =
566 (((int)(f[1] * 256.0f) & 0xfff) << 12) | ((int)(f[0] * 256.0f) & 0xfff);
567
568 so->tsc[2] |=
569 util_format_linear_float_to_srgb_8unorm(cso->border_color.f[0]) << 24;
570 so->tsc[3] =
571 util_format_linear_float_to_srgb_8unorm(cso->border_color.f[1]) << 12;
572 so->tsc[3] |=
573 util_format_linear_float_to_srgb_8unorm(cso->border_color.f[2]) << 20;
574
575 so->tsc[4] = fui(cso->border_color.f[0]);
576 so->tsc[5] = fui(cso->border_color.f[1]);
577 so->tsc[6] = fui(cso->border_color.f[2]);
578 so->tsc[7] = fui(cso->border_color.f[3]);
579
580 return (void *)so;
581 }
582
583 static void
nv50_sampler_state_delete(struct pipe_context * pipe,void * hwcso)584 nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
585 {
586 unsigned s, i;
587
588 for (s = 0; s < 3; ++s) {
589 assert(nv50_context(pipe)->num_samplers[s] <= PIPE_MAX_SAMPLERS);
590 for (i = 0; i < nv50_context(pipe)->num_samplers[s]; ++i)
591 if (nv50_context(pipe)->samplers[s][i] == hwcso)
592 nv50_context(pipe)->samplers[s][i] = NULL;
593 }
594
595 nv50_screen_tsc_free(nv50_context(pipe)->screen, nv50_tsc_entry(hwcso));
596
597 FREE(hwcso);
598 }
599
600 static inline void
nv50_stage_sampler_states_bind(struct nv50_context * nv50,int s,unsigned nr,void ** hwcsos)601 nv50_stage_sampler_states_bind(struct nv50_context *nv50, int s,
602 unsigned nr, void **hwcsos)
603 {
604 unsigned highest_found = 0;
605 unsigned i;
606
607 assert(nr <= PIPE_MAX_SAMPLERS);
608 for (i = 0; i < nr; ++i) {
609 struct nv50_tsc_entry *hwcso = hwcsos ? nv50_tsc_entry(hwcsos[i]) : NULL;
610 struct nv50_tsc_entry *old = nv50->samplers[s][i];
611
612 if (hwcso)
613 highest_found = i;
614
615 nv50->samplers[s][i] = hwcso;
616 if (old)
617 nv50_screen_tsc_unlock(nv50->screen, old);
618 }
619 assert(nv50->num_samplers[s] <= PIPE_MAX_SAMPLERS);
620 if (nr >= nv50->num_samplers[s])
621 nv50->num_samplers[s] = highest_found + 1;
622
623 nv50->dirty_3d |= NV50_NEW_3D_SAMPLERS;
624 }
625
626 static void
nv50_vp_sampler_states_bind(struct pipe_context * pipe,unsigned nr,void ** s)627 nv50_vp_sampler_states_bind(struct pipe_context *pipe, unsigned nr, void **s)
628 {
629 nv50_stage_sampler_states_bind(nv50_context(pipe), 0, nr, s);
630 }
631
632 static void
nv50_fp_sampler_states_bind(struct pipe_context * pipe,unsigned nr,void ** s)633 nv50_fp_sampler_states_bind(struct pipe_context *pipe, unsigned nr, void **s)
634 {
635 nv50_stage_sampler_states_bind(nv50_context(pipe), 2, nr, s);
636 }
637
638 static void
nv50_gp_sampler_states_bind(struct pipe_context * pipe,unsigned nr,void ** s)639 nv50_gp_sampler_states_bind(struct pipe_context *pipe, unsigned nr, void **s)
640 {
641 nv50_stage_sampler_states_bind(nv50_context(pipe), 1, nr, s);
642 }
643
644 static void
nv50_bind_sampler_states(struct pipe_context * pipe,enum pipe_shader_type shader,unsigned start,unsigned num_samplers,void ** samplers)645 nv50_bind_sampler_states(struct pipe_context *pipe,
646 enum pipe_shader_type shader, unsigned start,
647 unsigned num_samplers, void **samplers)
648 {
649 assert(start == 0);
650 switch (shader) {
651 case PIPE_SHADER_VERTEX:
652 nv50_vp_sampler_states_bind(pipe, num_samplers, samplers);
653 break;
654 case PIPE_SHADER_GEOMETRY:
655 nv50_gp_sampler_states_bind(pipe, num_samplers, samplers);
656 break;
657 case PIPE_SHADER_FRAGMENT:
658 nv50_fp_sampler_states_bind(pipe, num_samplers, samplers);
659 break;
660 default:
661 assert(!"unexpected shader type");
662 break;
663 }
664 }
665
666
667
668 /* NOTE: only called when not referenced anywhere, won't be bound */
669 static void
nv50_sampler_view_destroy(struct pipe_context * pipe,struct pipe_sampler_view * view)670 nv50_sampler_view_destroy(struct pipe_context *pipe,
671 struct pipe_sampler_view *view)
672 {
673 pipe_resource_reference(&view->texture, NULL);
674
675 nv50_screen_tic_free(nv50_context(pipe)->screen, nv50_tic_entry(view));
676
677 FREE(nv50_tic_entry(view));
678 }
679
680 static inline void
nv50_stage_set_sampler_views(struct nv50_context * nv50,int s,unsigned nr,struct pipe_sampler_view ** views)681 nv50_stage_set_sampler_views(struct nv50_context *nv50, int s,
682 unsigned nr,
683 struct pipe_sampler_view **views)
684 {
685 unsigned i;
686
687 assert(nr <= PIPE_MAX_SAMPLERS);
688 for (i = 0; i < nr; ++i) {
689 struct pipe_sampler_view *view = views ? views[i] : NULL;
690 struct nv50_tic_entry *old = nv50_tic_entry(nv50->textures[s][i]);
691 if (old)
692 nv50_screen_tic_unlock(nv50->screen, old);
693
694 if (view && view->texture) {
695 struct pipe_resource *res = view->texture;
696 if (res->target == PIPE_BUFFER &&
697 (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
698 nv50->textures_coherent[s] |= 1 << i;
699 else
700 nv50->textures_coherent[s] &= ~(1 << i);
701 } else {
702 nv50->textures_coherent[s] &= ~(1 << i);
703 }
704
705 pipe_sampler_view_reference(&nv50->textures[s][i], view);
706 }
707
708 assert(nv50->num_textures[s] <= PIPE_MAX_SAMPLERS);
709 for (i = nr; i < nv50->num_textures[s]; ++i) {
710 struct nv50_tic_entry *old = nv50_tic_entry(nv50->textures[s][i]);
711 if (!old)
712 continue;
713 nv50_screen_tic_unlock(nv50->screen, old);
714
715 pipe_sampler_view_reference(&nv50->textures[s][i], NULL);
716 }
717
718 nv50->num_textures[s] = nr;
719
720 nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_TEXTURES);
721
722 nv50->dirty_3d |= NV50_NEW_3D_TEXTURES;
723 }
724
725 static void
nv50_set_sampler_views(struct pipe_context * pipe,enum pipe_shader_type shader,unsigned start,unsigned nr,struct pipe_sampler_view ** views)726 nv50_set_sampler_views(struct pipe_context *pipe, enum pipe_shader_type shader,
727 unsigned start, unsigned nr,
728 struct pipe_sampler_view **views)
729 {
730 assert(start == 0);
731 switch (shader) {
732 case PIPE_SHADER_VERTEX:
733 nv50_stage_set_sampler_views(nv50_context(pipe), 0, nr, views);
734 break;
735 case PIPE_SHADER_GEOMETRY:
736 nv50_stage_set_sampler_views(nv50_context(pipe), 1, nr, views);
737 break;
738 case PIPE_SHADER_FRAGMENT:
739 nv50_stage_set_sampler_views(nv50_context(pipe), 2, nr, views);
740 break;
741 default:
742 ;
743 }
744 }
745
746
747
748 /* ============================= SHADERS =======================================
749 */
750
751 static void *
nv50_sp_state_create(struct pipe_context * pipe,const struct pipe_shader_state * cso,unsigned type)752 nv50_sp_state_create(struct pipe_context *pipe,
753 const struct pipe_shader_state *cso, unsigned type)
754 {
755 struct nv50_program *prog;
756
757 prog = CALLOC_STRUCT(nv50_program);
758 if (!prog)
759 return NULL;
760
761 prog->type = type;
762 prog->pipe.type = cso->type;
763
764 switch (cso->type) {
765 case PIPE_SHADER_IR_TGSI:
766 prog->pipe.tokens = tgsi_dup_tokens(cso->tokens);
767 break;
768 case PIPE_SHADER_IR_NIR:
769 prog->pipe.ir.nir = cso->ir.nir;
770 break;
771 default:
772 assert(!"unsupported IR!");
773 free(prog);
774 return NULL;
775 }
776
777 if (cso->stream_output.num_outputs)
778 prog->pipe.stream_output = cso->stream_output;
779
780 prog->translated = nv50_program_translate(
781 prog, nv50_context(pipe)->screen->base.device->chipset,
782 &nouveau_context(pipe)->debug);
783
784 return (void *)prog;
785 }
786
787 static void
nv50_sp_state_delete(struct pipe_context * pipe,void * hwcso)788 nv50_sp_state_delete(struct pipe_context *pipe, void *hwcso)
789 {
790 struct nv50_program *prog = (struct nv50_program *)hwcso;
791
792 nv50_program_destroy(nv50_context(pipe), prog);
793
794 if (prog->pipe.type == PIPE_SHADER_IR_TGSI)
795 FREE((void *)prog->pipe.tokens);
796 else if (prog->pipe.type == PIPE_SHADER_IR_NIR)
797 ralloc_free(prog->pipe.ir.nir);
798 FREE(prog);
799 }
800
801 static void *
nv50_vp_state_create(struct pipe_context * pipe,const struct pipe_shader_state * cso)802 nv50_vp_state_create(struct pipe_context *pipe,
803 const struct pipe_shader_state *cso)
804 {
805 return nv50_sp_state_create(pipe, cso, PIPE_SHADER_VERTEX);
806 }
807
808 static void
nv50_vp_state_bind(struct pipe_context * pipe,void * hwcso)809 nv50_vp_state_bind(struct pipe_context *pipe, void *hwcso)
810 {
811 struct nv50_context *nv50 = nv50_context(pipe);
812
813 nv50->vertprog = hwcso;
814 nv50->dirty_3d |= NV50_NEW_3D_VERTPROG;
815 }
816
817 static void *
nv50_fp_state_create(struct pipe_context * pipe,const struct pipe_shader_state * cso)818 nv50_fp_state_create(struct pipe_context *pipe,
819 const struct pipe_shader_state *cso)
820 {
821 return nv50_sp_state_create(pipe, cso, PIPE_SHADER_FRAGMENT);
822 }
823
824 static void
nv50_fp_state_bind(struct pipe_context * pipe,void * hwcso)825 nv50_fp_state_bind(struct pipe_context *pipe, void *hwcso)
826 {
827 struct nv50_context *nv50 = nv50_context(pipe);
828
829 nv50->fragprog = hwcso;
830 nv50->dirty_3d |= NV50_NEW_3D_FRAGPROG;
831 }
832
833 static void *
nv50_gp_state_create(struct pipe_context * pipe,const struct pipe_shader_state * cso)834 nv50_gp_state_create(struct pipe_context *pipe,
835 const struct pipe_shader_state *cso)
836 {
837 return nv50_sp_state_create(pipe, cso, PIPE_SHADER_GEOMETRY);
838 }
839
840 static void
nv50_gp_state_bind(struct pipe_context * pipe,void * hwcso)841 nv50_gp_state_bind(struct pipe_context *pipe, void *hwcso)
842 {
843 struct nv50_context *nv50 = nv50_context(pipe);
844
845 nv50->gmtyprog = hwcso;
846 nv50->dirty_3d |= NV50_NEW_3D_GMTYPROG;
847 }
848
849 static void *
nv50_cp_state_create(struct pipe_context * pipe,const struct pipe_compute_state * cso)850 nv50_cp_state_create(struct pipe_context *pipe,
851 const struct pipe_compute_state *cso)
852 {
853 struct nv50_program *prog;
854
855 prog = CALLOC_STRUCT(nv50_program);
856 if (!prog)
857 return NULL;
858 prog->type = PIPE_SHADER_COMPUTE;
859 prog->pipe.type = cso->ir_type;
860
861 switch(cso->ir_type) {
862 case PIPE_SHADER_IR_TGSI:
863 prog->pipe.tokens = tgsi_dup_tokens((const struct tgsi_token *)cso->prog);
864 break;
865 case PIPE_SHADER_IR_NIR:
866 prog->pipe.ir.nir = (nir_shader *)cso->prog;
867 break;
868 default:
869 assert(!"unsupported IR!");
870 free(prog);
871 return NULL;
872 }
873
874 prog->cp.smem_size = cso->req_local_mem;
875 prog->cp.lmem_size = cso->req_private_mem;
876 prog->parm_size = cso->req_input_mem;
877
878 return (void *)prog;
879 }
880
881 static void
nv50_cp_state_bind(struct pipe_context * pipe,void * hwcso)882 nv50_cp_state_bind(struct pipe_context *pipe, void *hwcso)
883 {
884 struct nv50_context *nv50 = nv50_context(pipe);
885
886 nv50->compprog = hwcso;
887 nv50->dirty_cp |= NV50_NEW_CP_PROGRAM;
888 }
889
890 static void
nv50_set_constant_buffer(struct pipe_context * pipe,enum pipe_shader_type shader,uint index,const struct pipe_constant_buffer * cb)891 nv50_set_constant_buffer(struct pipe_context *pipe,
892 enum pipe_shader_type shader, uint index,
893 const struct pipe_constant_buffer *cb)
894 {
895 struct nv50_context *nv50 = nv50_context(pipe);
896 struct pipe_resource *res = cb ? cb->buffer : NULL;
897 const unsigned s = nv50_context_shader_stage(shader);
898 const unsigned i = index;
899
900 if (shader == PIPE_SHADER_COMPUTE)
901 return;
902
903 assert(i < NV50_MAX_PIPE_CONSTBUFS);
904 if (nv50->constbuf[s][i].user)
905 nv50->constbuf[s][i].u.buf = NULL;
906 else
907 if (nv50->constbuf[s][i].u.buf) {
908 nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_CB(s, i));
909 nv04_resource(nv50->constbuf[s][i].u.buf)->cb_bindings[s] &= ~(1 << i);
910 }
911 pipe_resource_reference(&nv50->constbuf[s][i].u.buf, res);
912
913 nv50->constbuf[s][i].user = (cb && cb->user_buffer) ? true : false;
914 if (nv50->constbuf[s][i].user) {
915 nv50->constbuf[s][i].u.data = cb->user_buffer;
916 nv50->constbuf[s][i].size = MIN2(cb->buffer_size, 0x10000);
917 nv50->constbuf_valid[s] |= 1 << i;
918 nv50->constbuf_coherent[s] &= ~(1 << i);
919 } else
920 if (res) {
921 nv50->constbuf[s][i].offset = cb->buffer_offset;
922 nv50->constbuf[s][i].size = MIN2(align(cb->buffer_size, 0x100), 0x10000);
923 nv50->constbuf_valid[s] |= 1 << i;
924 if (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
925 nv50->constbuf_coherent[s] |= 1 << i;
926 else
927 nv50->constbuf_coherent[s] &= ~(1 << i);
928 } else {
929 nv50->constbuf_valid[s] &= ~(1 << i);
930 nv50->constbuf_coherent[s] &= ~(1 << i);
931 }
932 nv50->constbuf_dirty[s] |= 1 << i;
933
934 nv50->dirty_3d |= NV50_NEW_3D_CONSTBUF;
935 }
936
937 /* =============================================================================
938 */
939
940 static void
nv50_set_blend_color(struct pipe_context * pipe,const struct pipe_blend_color * bcol)941 nv50_set_blend_color(struct pipe_context *pipe,
942 const struct pipe_blend_color *bcol)
943 {
944 struct nv50_context *nv50 = nv50_context(pipe);
945
946 nv50->blend_colour = *bcol;
947 nv50->dirty_3d |= NV50_NEW_3D_BLEND_COLOUR;
948 }
949
950 static void
nv50_set_stencil_ref(struct pipe_context * pipe,const struct pipe_stencil_ref * sr)951 nv50_set_stencil_ref(struct pipe_context *pipe,
952 const struct pipe_stencil_ref *sr)
953 {
954 struct nv50_context *nv50 = nv50_context(pipe);
955
956 nv50->stencil_ref = *sr;
957 nv50->dirty_3d |= NV50_NEW_3D_STENCIL_REF;
958 }
959
960 static void
nv50_set_clip_state(struct pipe_context * pipe,const struct pipe_clip_state * clip)961 nv50_set_clip_state(struct pipe_context *pipe,
962 const struct pipe_clip_state *clip)
963 {
964 struct nv50_context *nv50 = nv50_context(pipe);
965
966 memcpy(nv50->clip.ucp, clip->ucp, sizeof(clip->ucp));
967
968 nv50->dirty_3d |= NV50_NEW_3D_CLIP;
969 }
970
971 static void
nv50_set_sample_mask(struct pipe_context * pipe,unsigned sample_mask)972 nv50_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
973 {
974 struct nv50_context *nv50 = nv50_context(pipe);
975
976 nv50->sample_mask = sample_mask;
977 nv50->dirty_3d |= NV50_NEW_3D_SAMPLE_MASK;
978 }
979
980 static void
nv50_set_min_samples(struct pipe_context * pipe,unsigned min_samples)981 nv50_set_min_samples(struct pipe_context *pipe, unsigned min_samples)
982 {
983 struct nv50_context *nv50 = nv50_context(pipe);
984
985 if (nv50->min_samples != min_samples) {
986 nv50->min_samples = min_samples;
987 nv50->dirty_3d |= NV50_NEW_3D_MIN_SAMPLES;
988 }
989 }
990
991 static void
nv50_set_framebuffer_state(struct pipe_context * pipe,const struct pipe_framebuffer_state * fb)992 nv50_set_framebuffer_state(struct pipe_context *pipe,
993 const struct pipe_framebuffer_state *fb)
994 {
995 struct nv50_context *nv50 = nv50_context(pipe);
996
997 nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_FB);
998
999 util_copy_framebuffer_state(&nv50->framebuffer, fb);
1000
1001 nv50->dirty_3d |= NV50_NEW_3D_FRAMEBUFFER | NV50_NEW_3D_TEXTURES;
1002 }
1003
1004 static void
nv50_set_polygon_stipple(struct pipe_context * pipe,const struct pipe_poly_stipple * stipple)1005 nv50_set_polygon_stipple(struct pipe_context *pipe,
1006 const struct pipe_poly_stipple *stipple)
1007 {
1008 struct nv50_context *nv50 = nv50_context(pipe);
1009
1010 nv50->stipple = *stipple;
1011 nv50->dirty_3d |= NV50_NEW_3D_STIPPLE;
1012 }
1013
1014 static void
nv50_set_scissor_states(struct pipe_context * pipe,unsigned start_slot,unsigned num_scissors,const struct pipe_scissor_state * scissor)1015 nv50_set_scissor_states(struct pipe_context *pipe,
1016 unsigned start_slot,
1017 unsigned num_scissors,
1018 const struct pipe_scissor_state *scissor)
1019 {
1020 struct nv50_context *nv50 = nv50_context(pipe);
1021 int i;
1022
1023 assert(start_slot + num_scissors <= NV50_MAX_VIEWPORTS);
1024 for (i = 0; i < num_scissors; i++) {
1025 if (!memcmp(&nv50->scissors[start_slot + i], &scissor[i], sizeof(*scissor)))
1026 continue;
1027 nv50->scissors[start_slot + i] = scissor[i];
1028 nv50->scissors_dirty |= 1 << (start_slot + i);
1029 nv50->dirty_3d |= NV50_NEW_3D_SCISSOR;
1030 }
1031 }
1032
1033 static void
nv50_set_viewport_states(struct pipe_context * pipe,unsigned start_slot,unsigned num_viewports,const struct pipe_viewport_state * vpt)1034 nv50_set_viewport_states(struct pipe_context *pipe,
1035 unsigned start_slot,
1036 unsigned num_viewports,
1037 const struct pipe_viewport_state *vpt)
1038 {
1039 struct nv50_context *nv50 = nv50_context(pipe);
1040 int i;
1041
1042 assert(start_slot + num_viewports <= NV50_MAX_VIEWPORTS);
1043 for (i = 0; i < num_viewports; i++) {
1044 if (!memcmp(&nv50->viewports[start_slot + i], &vpt[i], sizeof(*vpt)))
1045 continue;
1046 nv50->viewports[start_slot + i] = vpt[i];
1047 nv50->viewports_dirty |= 1 << (start_slot + i);
1048 nv50->dirty_3d |= NV50_NEW_3D_VIEWPORT;
1049 }
1050 }
1051
1052 static void
nv50_set_window_rectangles(struct pipe_context * pipe,bool include,unsigned num_rectangles,const struct pipe_scissor_state * rectangles)1053 nv50_set_window_rectangles(struct pipe_context *pipe,
1054 bool include,
1055 unsigned num_rectangles,
1056 const struct pipe_scissor_state *rectangles)
1057 {
1058 struct nv50_context *nv50 = nv50_context(pipe);
1059
1060 nv50->window_rect.inclusive = include;
1061 nv50->window_rect.rects = MIN2(num_rectangles, NV50_MAX_WINDOW_RECTANGLES);
1062 memcpy(nv50->window_rect.rect, rectangles,
1063 sizeof(struct pipe_scissor_state) * nv50->window_rect.rects);
1064
1065 nv50->dirty_3d |= NV50_NEW_3D_WINDOW_RECTS;
1066 }
1067
1068 static void
nv50_set_vertex_buffers(struct pipe_context * pipe,unsigned start_slot,unsigned count,const struct pipe_vertex_buffer * vb)1069 nv50_set_vertex_buffers(struct pipe_context *pipe,
1070 unsigned start_slot, unsigned count,
1071 const struct pipe_vertex_buffer *vb)
1072 {
1073 struct nv50_context *nv50 = nv50_context(pipe);
1074 unsigned i;
1075
1076 nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_VERTEX);
1077 nv50->dirty_3d |= NV50_NEW_3D_ARRAYS;
1078
1079 util_set_vertex_buffers_count(nv50->vtxbuf, &nv50->num_vtxbufs, vb,
1080 start_slot, count);
1081
1082 if (!vb) {
1083 nv50->vbo_user &= ~(((1ull << count) - 1) << start_slot);
1084 nv50->vbo_constant &= ~(((1ull << count) - 1) << start_slot);
1085 nv50->vtxbufs_coherent &= ~(((1ull << count) - 1) << start_slot);
1086 return;
1087 }
1088
1089 for (i = 0; i < count; ++i) {
1090 unsigned dst_index = start_slot + i;
1091
1092 if (vb[i].is_user_buffer) {
1093 nv50->vbo_user |= 1 << dst_index;
1094 if (!vb[i].stride)
1095 nv50->vbo_constant |= 1 << dst_index;
1096 else
1097 nv50->vbo_constant &= ~(1 << dst_index);
1098 nv50->vtxbufs_coherent &= ~(1 << dst_index);
1099 } else {
1100 nv50->vbo_user &= ~(1 << dst_index);
1101 nv50->vbo_constant &= ~(1 << dst_index);
1102
1103 if (vb[i].buffer.resource &&
1104 vb[i].buffer.resource->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
1105 nv50->vtxbufs_coherent |= (1 << dst_index);
1106 else
1107 nv50->vtxbufs_coherent &= ~(1 << dst_index);
1108 }
1109 }
1110 }
1111
1112 static void
nv50_vertex_state_bind(struct pipe_context * pipe,void * hwcso)1113 nv50_vertex_state_bind(struct pipe_context *pipe, void *hwcso)
1114 {
1115 struct nv50_context *nv50 = nv50_context(pipe);
1116
1117 nv50->vertex = hwcso;
1118 nv50->dirty_3d |= NV50_NEW_3D_VERTEX;
1119 }
1120
1121 static struct pipe_stream_output_target *
nv50_so_target_create(struct pipe_context * pipe,struct pipe_resource * res,unsigned offset,unsigned size)1122 nv50_so_target_create(struct pipe_context *pipe,
1123 struct pipe_resource *res,
1124 unsigned offset, unsigned size)
1125 {
1126 struct nv04_resource *buf = (struct nv04_resource *)res;
1127 struct nv50_so_target *targ = MALLOC_STRUCT(nv50_so_target);
1128 if (!targ)
1129 return NULL;
1130
1131 if (nouveau_context(pipe)->screen->class_3d >= NVA0_3D_CLASS) {
1132 targ->pq = pipe->create_query(pipe,
1133 NVA0_HW_QUERY_STREAM_OUTPUT_BUFFER_OFFSET, 0);
1134 if (!targ->pq) {
1135 FREE(targ);
1136 return NULL;
1137 }
1138 } else {
1139 targ->pq = NULL;
1140 }
1141 targ->clean = true;
1142
1143 targ->pipe.buffer_size = size;
1144 targ->pipe.buffer_offset = offset;
1145 targ->pipe.context = pipe;
1146 targ->pipe.buffer = NULL;
1147 pipe_resource_reference(&targ->pipe.buffer, res);
1148 pipe_reference_init(&targ->pipe.reference, 1);
1149
1150 assert(buf->base.target == PIPE_BUFFER);
1151 util_range_add(&buf->base, &buf->valid_buffer_range, offset, offset + size);
1152
1153 return &targ->pipe;
1154 }
1155
1156 static void
nva0_so_target_save_offset(struct pipe_context * pipe,struct pipe_stream_output_target * ptarg,unsigned index,bool serialize)1157 nva0_so_target_save_offset(struct pipe_context *pipe,
1158 struct pipe_stream_output_target *ptarg,
1159 unsigned index, bool serialize)
1160 {
1161 struct nv50_so_target *targ = nv50_so_target(ptarg);
1162
1163 if (serialize) {
1164 struct nouveau_pushbuf *push = nv50_context(pipe)->base.pushbuf;
1165 PUSH_SPACE(push, 2);
1166 BEGIN_NV04(push, SUBC_3D(NV50_GRAPH_SERIALIZE), 1);
1167 PUSH_DATA (push, 0);
1168 }
1169
1170 nv50_query(targ->pq)->index = index;
1171 pipe->end_query(pipe, targ->pq);
1172 }
1173
1174 static void
nv50_so_target_destroy(struct pipe_context * pipe,struct pipe_stream_output_target * ptarg)1175 nv50_so_target_destroy(struct pipe_context *pipe,
1176 struct pipe_stream_output_target *ptarg)
1177 {
1178 struct nv50_so_target *targ = nv50_so_target(ptarg);
1179 if (targ->pq)
1180 pipe->destroy_query(pipe, targ->pq);
1181 pipe_resource_reference(&targ->pipe.buffer, NULL);
1182 FREE(targ);
1183 }
1184
1185 static void
nv50_set_stream_output_targets(struct pipe_context * pipe,unsigned num_targets,struct pipe_stream_output_target ** targets,const unsigned * offsets)1186 nv50_set_stream_output_targets(struct pipe_context *pipe,
1187 unsigned num_targets,
1188 struct pipe_stream_output_target **targets,
1189 const unsigned *offsets)
1190 {
1191 struct nv50_context *nv50 = nv50_context(pipe);
1192 unsigned i;
1193 bool serialize = true;
1194 const bool can_resume = nv50->screen->base.class_3d >= NVA0_3D_CLASS;
1195
1196 assert(num_targets <= 4);
1197
1198 for (i = 0; i < num_targets; ++i) {
1199 const bool changed = nv50->so_target[i] != targets[i];
1200 const bool append = (offsets[i] == (unsigned)-1);
1201 if (!changed && append)
1202 continue;
1203 nv50->so_targets_dirty |= 1 << i;
1204
1205 if (can_resume && changed && nv50->so_target[i]) {
1206 nva0_so_target_save_offset(pipe, nv50->so_target[i], i, serialize);
1207 serialize = false;
1208 }
1209
1210 if (targets[i] && !append)
1211 nv50_so_target(targets[i])->clean = true;
1212
1213 pipe_so_target_reference(&nv50->so_target[i], targets[i]);
1214 }
1215 for (; i < nv50->num_so_targets; ++i) {
1216 if (can_resume && nv50->so_target[i]) {
1217 nva0_so_target_save_offset(pipe, nv50->so_target[i], i, serialize);
1218 serialize = false;
1219 }
1220 pipe_so_target_reference(&nv50->so_target[i], NULL);
1221 nv50->so_targets_dirty |= 1 << i;
1222 }
1223 nv50->num_so_targets = num_targets;
1224
1225 if (nv50->so_targets_dirty) {
1226 nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_SO);
1227 nv50->dirty_3d |= NV50_NEW_3D_STRMOUT;
1228 }
1229 }
1230
1231 static void
nv50_set_compute_resources(struct pipe_context * pipe,unsigned start,unsigned nr,struct pipe_surface ** resources)1232 nv50_set_compute_resources(struct pipe_context *pipe,
1233 unsigned start, unsigned nr,
1234 struct pipe_surface **resources)
1235 {
1236 /* TODO: bind surfaces */
1237 }
1238
1239 static inline void
nv50_set_global_handle(uint32_t * phandle,struct pipe_resource * res)1240 nv50_set_global_handle(uint32_t *phandle, struct pipe_resource *res)
1241 {
1242 struct nv04_resource *buf = nv04_resource(res);
1243 if (buf) {
1244 uint64_t limit = (buf->address + buf->base.width0) - 1;
1245 if (limit < (1ULL << 32)) {
1246 *phandle = (uint32_t)buf->address;
1247 } else {
1248 NOUVEAU_ERR("Cannot map into TGSI_RESOURCE_GLOBAL: "
1249 "resource not contained within 32-bit address space !\n");
1250 *phandle = 0;
1251 }
1252 } else {
1253 *phandle = 0;
1254 }
1255 }
1256
1257 static void
nv50_set_global_bindings(struct pipe_context * pipe,unsigned start,unsigned nr,struct pipe_resource ** resources,uint32_t ** handles)1258 nv50_set_global_bindings(struct pipe_context *pipe,
1259 unsigned start, unsigned nr,
1260 struct pipe_resource **resources,
1261 uint32_t **handles)
1262 {
1263 struct nv50_context *nv50 = nv50_context(pipe);
1264 struct pipe_resource **ptr;
1265 unsigned i;
1266 const unsigned end = start + nr;
1267
1268 if (nv50->global_residents.size <= (end * sizeof(struct pipe_resource *))) {
1269 const unsigned old_size = nv50->global_residents.size;
1270 if (util_dynarray_resize(&nv50->global_residents, struct pipe_resource *, end)) {
1271 memset((uint8_t *)nv50->global_residents.data + old_size, 0,
1272 nv50->global_residents.size - old_size);
1273 } else {
1274 NOUVEAU_ERR("Could not resize global residents array\n");
1275 return;
1276 }
1277 }
1278
1279 if (resources) {
1280 ptr = util_dynarray_element(
1281 &nv50->global_residents, struct pipe_resource *, start);
1282 for (i = 0; i < nr; ++i) {
1283 pipe_resource_reference(&ptr[i], resources[i]);
1284 nv50_set_global_handle(handles[i], resources[i]);
1285 }
1286 } else {
1287 ptr = util_dynarray_element(
1288 &nv50->global_residents, struct pipe_resource *, start);
1289 for (i = 0; i < nr; ++i)
1290 pipe_resource_reference(&ptr[i], NULL);
1291 }
1292
1293 nouveau_bufctx_reset(nv50->bufctx_cp, NV50_BIND_CP_GLOBAL);
1294
1295 nv50->dirty_cp |= NV50_NEW_CP_GLOBALS;
1296 }
1297
1298 void
nv50_init_state_functions(struct nv50_context * nv50)1299 nv50_init_state_functions(struct nv50_context *nv50)
1300 {
1301 struct pipe_context *pipe = &nv50->base.pipe;
1302
1303 pipe->create_blend_state = nv50_blend_state_create;
1304 pipe->bind_blend_state = nv50_blend_state_bind;
1305 pipe->delete_blend_state = nv50_blend_state_delete;
1306
1307 pipe->create_rasterizer_state = nv50_rasterizer_state_create;
1308 pipe->bind_rasterizer_state = nv50_rasterizer_state_bind;
1309 pipe->delete_rasterizer_state = nv50_rasterizer_state_delete;
1310
1311 pipe->create_depth_stencil_alpha_state = nv50_zsa_state_create;
1312 pipe->bind_depth_stencil_alpha_state = nv50_zsa_state_bind;
1313 pipe->delete_depth_stencil_alpha_state = nv50_zsa_state_delete;
1314
1315 pipe->create_sampler_state = nv50_sampler_state_create;
1316 pipe->delete_sampler_state = nv50_sampler_state_delete;
1317 pipe->bind_sampler_states = nv50_bind_sampler_states;
1318
1319 pipe->create_sampler_view = nv50_create_sampler_view;
1320 pipe->sampler_view_destroy = nv50_sampler_view_destroy;
1321 pipe->set_sampler_views = nv50_set_sampler_views;
1322
1323 pipe->create_vs_state = nv50_vp_state_create;
1324 pipe->create_fs_state = nv50_fp_state_create;
1325 pipe->create_gs_state = nv50_gp_state_create;
1326 pipe->create_compute_state = nv50_cp_state_create;
1327 pipe->bind_vs_state = nv50_vp_state_bind;
1328 pipe->bind_fs_state = nv50_fp_state_bind;
1329 pipe->bind_gs_state = nv50_gp_state_bind;
1330 pipe->bind_compute_state = nv50_cp_state_bind;
1331 pipe->delete_vs_state = nv50_sp_state_delete;
1332 pipe->delete_fs_state = nv50_sp_state_delete;
1333 pipe->delete_gs_state = nv50_sp_state_delete;
1334 pipe->delete_compute_state = nv50_sp_state_delete;
1335
1336 pipe->set_blend_color = nv50_set_blend_color;
1337 pipe->set_stencil_ref = nv50_set_stencil_ref;
1338 pipe->set_clip_state = nv50_set_clip_state;
1339 pipe->set_sample_mask = nv50_set_sample_mask;
1340 pipe->set_min_samples = nv50_set_min_samples;
1341 pipe->set_constant_buffer = nv50_set_constant_buffer;
1342 pipe->set_framebuffer_state = nv50_set_framebuffer_state;
1343 pipe->set_polygon_stipple = nv50_set_polygon_stipple;
1344 pipe->set_scissor_states = nv50_set_scissor_states;
1345 pipe->set_viewport_states = nv50_set_viewport_states;
1346 pipe->set_window_rectangles = nv50_set_window_rectangles;
1347
1348 pipe->create_vertex_elements_state = nv50_vertex_state_create;
1349 pipe->delete_vertex_elements_state = nv50_vertex_state_delete;
1350 pipe->bind_vertex_elements_state = nv50_vertex_state_bind;
1351
1352 pipe->set_vertex_buffers = nv50_set_vertex_buffers;
1353
1354 pipe->create_stream_output_target = nv50_so_target_create;
1355 pipe->stream_output_target_destroy = nv50_so_target_destroy;
1356 pipe->set_stream_output_targets = nv50_set_stream_output_targets;
1357
1358 pipe->set_global_binding = nv50_set_global_bindings;
1359 pipe->set_compute_resources = nv50_set_compute_resources;
1360
1361 nv50->sample_mask = ~0;
1362 nv50->min_samples = 1;
1363 }
1364