1 /*
2 * Copyright 2011 Joakim Sindholt <[email protected]>
3 * Copyright 2013 Christoph Bumiller
4 * SPDX-License-Identifier: MIT
5 */
6
7 #define NINE_STATE
8
9 #include "device9.h"
10 #include "swapchain9.h"
11 #include "basetexture9.h"
12 #include "buffer9.h"
13 #include "indexbuffer9.h"
14 #include "surface9.h"
15 #include "vertexbuffer9.h"
16 #include "vertexdeclaration9.h"
17 #include "vertexshader9.h"
18 #include "pixelshader9.h"
19 #include "nine_pipe.h"
20 #include "nine_ff.h"
21 #include "nine_limits.h"
22 #include "pipe/p_context.h"
23 #include "pipe/p_state.h"
24 #include "cso_cache/cso_context.h"
25 #include "util/u_atomic.h"
26 #include "util/u_upload_mgr.h"
27 #include "util/u_math.h"
28 #include "util/box.h"
29 #include "util/u_simple_shaders.h"
30 #include "util/u_gen_mipmap.h"
31
32 /* CSMT headers */
33 #include "nine_queue.h"
34 #include "nine_csmt_helper.h"
35 #include "util/u_thread.h"
36
37 #define DBG_CHANNEL DBG_DEVICE
38
39 /* Nine CSMT */
40
41 struct csmt_instruction {
42 int (* func)(struct NineDevice9 *This, struct csmt_instruction *instr);
43 };
44
45 struct csmt_context {
46 thrd_t worker;
47 struct nine_queue_pool* pool;
48 BOOL terminate;
49 cnd_t event_processed;
50 mtx_t mutex_processed;
51 struct NineDevice9 *device;
52 BOOL processed;
53 BOOL toPause;
54 BOOL hasPaused;
55 mtx_t thread_running;
56 mtx_t thread_resume;
57 };
58
59 /* Wait for instruction to be processed.
60 * Caller has to ensure that only one thread waits at time.
61 */
62 static void
nine_csmt_wait_processed(struct csmt_context * ctx)63 nine_csmt_wait_processed(struct csmt_context *ctx)
64 {
65 mtx_lock(&ctx->mutex_processed);
66 while (!p_atomic_read(&ctx->processed)) {
67 cnd_wait(&ctx->event_processed, &ctx->mutex_processed);
68 }
69 mtx_unlock(&ctx->mutex_processed);
70 }
71
72 /* CSMT worker thread */
73 static
74 int
nine_csmt_worker(void * arg)75 nine_csmt_worker(void *arg)
76 {
77 struct csmt_context *ctx = arg;
78 struct csmt_instruction *instr;
79 DBG("CSMT worker spawned\n");
80
81 u_thread_setname("CSMT-Worker");
82
83 while (1) {
84 nine_queue_wait_flush(ctx->pool);
85 mtx_lock(&ctx->thread_running);
86
87 /* Get instruction. NULL on empty cmdbuf. */
88 while (!p_atomic_read(&ctx->terminate) &&
89 (instr = (struct csmt_instruction *)nine_queue_get(ctx->pool))) {
90
91 /* decode */
92 if (instr->func(ctx->device, instr)) {
93 mtx_lock(&ctx->mutex_processed);
94 p_atomic_set(&ctx->processed, true);
95 cnd_signal(&ctx->event_processed);
96 mtx_unlock(&ctx->mutex_processed);
97 }
98 if (p_atomic_read(&ctx->toPause)) {
99 mtx_unlock(&ctx->thread_running);
100 /* will wait here the thread can be resumed */
101 mtx_lock(&ctx->thread_resume);
102 mtx_lock(&ctx->thread_running);
103 mtx_unlock(&ctx->thread_resume);
104 }
105 }
106
107 mtx_unlock(&ctx->thread_running);
108 if (p_atomic_read(&ctx->terminate)) {
109 mtx_lock(&ctx->mutex_processed);
110 p_atomic_set(&ctx->processed, true);
111 cnd_signal(&ctx->event_processed);
112 mtx_unlock(&ctx->mutex_processed);
113 break;
114 }
115 }
116
117 DBG("CSMT worker destroyed\n");
118 return 0;
119 }
120
121 /* Create a CSMT context.
122 * Spawns a worker thread.
123 */
124 struct csmt_context *
nine_csmt_create(struct NineDevice9 * This)125 nine_csmt_create( struct NineDevice9 *This )
126 {
127 struct csmt_context *ctx;
128
129 ctx = CALLOC_STRUCT(csmt_context);
130 if (!ctx)
131 return NULL;
132
133 ctx->pool = nine_queue_create();
134 if (!ctx->pool) {
135 FREE(ctx);
136 return NULL;
137 }
138 cnd_init(&ctx->event_processed);
139 (void) mtx_init(&ctx->mutex_processed, mtx_plain);
140 (void) mtx_init(&ctx->thread_running, mtx_plain);
141 (void) mtx_init(&ctx->thread_resume, mtx_plain);
142
143 #if MESA_DEBUG || !defined(NDEBUG)
144 u_thread_setname("Main thread");
145 #endif
146
147 ctx->device = This;
148
149 if (thrd_success != u_thread_create(&ctx->worker, nine_csmt_worker, ctx)) {
150 nine_queue_delete(ctx->pool);
151 FREE(ctx);
152 return NULL;
153 }
154
155 DBG("Returning context %p\n", ctx);
156
157 return ctx;
158 }
159
160 static int
nop_func(struct NineDevice9 * This,struct csmt_instruction * instr)161 nop_func( struct NineDevice9 *This, struct csmt_instruction *instr )
162 {
163 (void) This;
164 (void) instr;
165
166 return 1;
167 }
168
169 /* Push nop instruction and flush the queue.
170 * Waits for the worker to complete. */
171 void
nine_csmt_process(struct NineDevice9 * device)172 nine_csmt_process( struct NineDevice9 *device )
173 {
174 struct csmt_instruction* instr;
175 struct csmt_context *ctx = device->csmt_ctx;
176
177 if (!device->csmt_active)
178 return;
179
180 if (nine_queue_isempty(ctx->pool))
181 return;
182
183 DBG("device=%p\n", device);
184
185 /* NOP */
186 instr = nine_queue_alloc(ctx->pool, sizeof(struct csmt_instruction));
187 assert(instr);
188 instr->func = nop_func;
189
190 p_atomic_set(&ctx->processed, false);
191 nine_queue_flush(ctx->pool);
192
193 nine_csmt_wait_processed(ctx);
194 }
195
196 void
nine_csmt_flush(struct NineDevice9 * device)197 nine_csmt_flush( struct NineDevice9* device )
198 {
199 if (!device->csmt_active)
200 return;
201
202 nine_queue_flush(device->csmt_ctx->pool);
203 }
204
205
206 /* Destroys a CSMT context.
207 * Waits for the worker thread to terminate.
208 */
209 void
nine_csmt_destroy(struct NineDevice9 * device,struct csmt_context * ctx)210 nine_csmt_destroy( struct NineDevice9 *device, struct csmt_context *ctx )
211 {
212 struct csmt_instruction* instr;
213 thrd_t render_thread = ctx->worker;
214
215 DBG("device=%p ctx=%p\n", device, ctx);
216
217 /* Push nop and flush the queue. */
218 instr = nine_queue_alloc(ctx->pool, sizeof(struct csmt_instruction));
219 assert(instr);
220 instr->func = nop_func;
221
222 p_atomic_set(&ctx->processed, false);
223 /* Signal worker to terminate. */
224 p_atomic_set(&ctx->terminate, true);
225 nine_queue_flush(ctx->pool);
226
227 nine_csmt_wait_processed(ctx);
228 nine_queue_delete(ctx->pool);
229
230 mtx_destroy(&ctx->thread_resume);
231 mtx_destroy(&ctx->thread_running);
232
233 mtx_destroy(&ctx->mutex_processed);
234 cnd_destroy(&ctx->event_processed);
235
236 FREE(ctx);
237
238 thrd_join(render_thread, NULL);
239 }
240
241 static void
nine_csmt_pause(struct NineDevice9 * device)242 nine_csmt_pause( struct NineDevice9 *device )
243 {
244 struct csmt_context *ctx = device->csmt_ctx;
245
246 if (!device->csmt_active)
247 return;
248
249 /* No need to pause the thread */
250 if (nine_queue_no_flushed_work(ctx->pool))
251 return;
252
253 mtx_lock(&ctx->thread_resume);
254 p_atomic_set(&ctx->toPause, true);
255
256 /* Wait the thread is paused */
257 mtx_lock(&ctx->thread_running);
258 ctx->hasPaused = true;
259 p_atomic_set(&ctx->toPause, false);
260 }
261
262 static void
nine_csmt_resume(struct NineDevice9 * device)263 nine_csmt_resume( struct NineDevice9 *device )
264 {
265 struct csmt_context *ctx = device->csmt_ctx;
266
267 if (!device->csmt_active)
268 return;
269
270 if (!ctx->hasPaused)
271 return;
272
273 ctx->hasPaused = false;
274 mtx_unlock(&ctx->thread_running);
275 mtx_unlock(&ctx->thread_resume);
276 }
277
278 struct pipe_context *
nine_context_get_pipe(struct NineDevice9 * device)279 nine_context_get_pipe( struct NineDevice9 *device )
280 {
281 nine_csmt_process(device);
282 return device->context.pipe;
283 }
284
285 struct pipe_context *
nine_context_get_pipe_multithread(struct NineDevice9 * device)286 nine_context_get_pipe_multithread( struct NineDevice9 *device )
287 {
288 struct csmt_context *ctx = device->csmt_ctx;
289
290 if (!device->csmt_active)
291 return device->context.pipe;
292
293 if (!u_thread_is_self(ctx->worker))
294 nine_csmt_process(device);
295
296 return device->context.pipe;
297 }
298
299 struct pipe_context *
nine_context_get_pipe_acquire(struct NineDevice9 * device)300 nine_context_get_pipe_acquire( struct NineDevice9 *device )
301 {
302 nine_csmt_pause(device);
303 return device->context.pipe;
304 }
305
306 void
nine_context_get_pipe_release(struct NineDevice9 * device)307 nine_context_get_pipe_release( struct NineDevice9 *device )
308 {
309 nine_csmt_resume(device);
310 }
311
312 bool
nine_context_is_worker(struct NineDevice9 * device)313 nine_context_is_worker( struct NineDevice9 *device )
314 {
315 struct csmt_context *ctx = device->csmt_ctx;
316
317 if (!device->csmt_active)
318 return false;
319
320 return u_thread_is_self(ctx->worker);
321 }
322
323 /* Nine state functions */
324
325 /* Check if some states need to be set dirty */
326
327 static inline DWORD
check_multisample(struct NineDevice9 * device)328 check_multisample(struct NineDevice9 *device)
329 {
330 struct nine_context *context = &device->context;
331 DWORD *rs = context->rs;
332 struct NineSurface9 *rt0 = context->rt[0];
333 bool multisampled_target;
334 DWORD new_value;
335
336 multisampled_target = rt0 && rt0->desc.MultiSampleType >= 1;
337 if (rt0 && rt0->desc.Format == D3DFMT_NULL && context->ds)
338 multisampled_target = context->ds->desc.MultiSampleType >= 1;
339 new_value = (multisampled_target && rs[D3DRS_MULTISAMPLEANTIALIAS]) ? 1 : 0;
340 if (rs[NINED3DRS_MULTISAMPLE] != new_value) {
341 rs[NINED3DRS_MULTISAMPLE] = new_value;
342 return NINE_STATE_RASTERIZER;
343 }
344 return 0;
345 }
346
347 /* State preparation only */
348
349 static inline void
prepare_blend(struct NineDevice9 * device)350 prepare_blend(struct NineDevice9 *device)
351 {
352 nine_convert_blend_state(&device->context.pipe_data.blend, device->context.rs);
353 device->context.commit |= NINE_STATE_COMMIT_BLEND;
354 }
355
356 static inline void
prepare_dsa(struct NineDevice9 * device)357 prepare_dsa(struct NineDevice9 *device)
358 {
359 nine_convert_dsa_state(&device->context.pipe_data.dsa, device->context.rs);
360 device->context.commit |= NINE_STATE_COMMIT_DSA;
361 }
362
363 static inline void
prepare_rasterizer(struct NineDevice9 * device)364 prepare_rasterizer(struct NineDevice9 *device)
365 {
366 nine_convert_rasterizer_state(device, &device->context.pipe_data.rast, device->context.rs);
367 device->context.commit |= NINE_STATE_COMMIT_RASTERIZER;
368 }
369
370 static void
prepare_vs_constants_userbuf_swvp(struct NineDevice9 * device)371 prepare_vs_constants_userbuf_swvp(struct NineDevice9 *device)
372 {
373 struct nine_context *context = &device->context;
374
375 if (device->driver_caps.emulate_ucp) {
376 /* TODO: Avoid memcpy all time by storing directly into the array */
377 memcpy(&context->vs_const_f[4 * NINE_MAX_CONST_SWVP_SPE_OFFSET], &context->clip.ucp, sizeof(context->clip));
378 context->changed.vs_const_f = 1; /* TODO optimize */
379 }
380
381 if (device->driver_caps.always_output_pointsize) {
382 context->vs_const_f[4 * (NINE_MAX_CONST_SWVP_SPE_OFFSET + 8)] =
383 CLAMP(asfloat(context->rs[D3DRS_POINTSIZE]),
384 asfloat(context->rs[D3DRS_POINTSIZE_MIN]),
385 asfloat(context->rs[D3DRS_POINTSIZE_MAX]));
386 context->changed.vs_const_f = 1; /* TODO optimize */
387 }
388
389 if (context->changed.vs_const_f || context->changed.group & NINE_STATE_SWVP) {
390 struct pipe_constant_buffer cb;
391
392 cb.buffer_offset = 0;
393 cb.buffer_size = 4096 * sizeof(float[4]);
394 cb.user_buffer = context->vs_const_f_swvp;
395
396 if (context->vs->lconstf.ranges) {
397 const struct nine_lconstf *lconstf = &(context->vs->lconstf);
398 const struct nine_range *r = lconstf->ranges;
399 unsigned n = 0;
400 float *dst = context->vs_lconstf_temp;
401 float *src = (float *)cb.user_buffer;
402 memcpy(dst, src, cb.buffer_size);
403 while (r) {
404 unsigned p = r->bgn;
405 unsigned c = r->end - r->bgn;
406 memcpy(&dst[p * 4], &lconstf->data[n * 4], c * 4 * sizeof(float));
407 n += c;
408 r = r->next;
409 }
410 cb.user_buffer = dst;
411 }
412
413 context->pipe_data.cb0_swvp.buffer_offset = cb.buffer_offset;
414 context->pipe_data.cb0_swvp.buffer_size = cb.buffer_size;
415 context->pipe_data.cb0_swvp.user_buffer = cb.user_buffer;
416
417 cb.user_buffer = (int8_t *)cb.user_buffer + 4096 * sizeof(float[4]);
418 context->pipe_data.cb1_swvp.buffer_offset = cb.buffer_offset;
419 context->pipe_data.cb1_swvp.buffer_size = cb.buffer_size;
420 context->pipe_data.cb1_swvp.user_buffer = cb.user_buffer;
421
422 context->changed.vs_const_f = 0;
423 }
424
425 if (context->changed.vs_const_i || context->changed.group & NINE_STATE_SWVP) {
426 struct pipe_constant_buffer cb;
427
428 cb.buffer_offset = 0;
429 cb.buffer_size = 2048 * sizeof(float[4]);
430 cb.user_buffer = context->vs_const_i;
431
432 context->pipe_data.cb2_swvp.buffer_offset = cb.buffer_offset;
433 context->pipe_data.cb2_swvp.buffer_size = cb.buffer_size;
434 context->pipe_data.cb2_swvp.user_buffer = cb.user_buffer;
435 context->changed.vs_const_i = 0;
436 }
437
438 if (context->changed.vs_const_b || context->changed.group & NINE_STATE_SWVP) {
439 struct pipe_constant_buffer cb;
440
441 cb.buffer_offset = 0;
442 cb.buffer_size = 512 * sizeof(float[4]);
443 cb.user_buffer = context->vs_const_b;
444
445 context->pipe_data.cb3_swvp.buffer_offset = cb.buffer_offset;
446 context->pipe_data.cb3_swvp.buffer_size = cb.buffer_size;
447 context->pipe_data.cb3_swvp.user_buffer = cb.user_buffer;
448 context->changed.vs_const_b = 0;
449 }
450
451 context->changed.group &= ~NINE_STATE_VS_CONST;
452 context->commit |= NINE_STATE_COMMIT_CONST_VS;
453 }
454
455 static void
prepare_vs_constants_userbuf(struct NineDevice9 * device)456 prepare_vs_constants_userbuf(struct NineDevice9 *device)
457 {
458 struct nine_context *context = &device->context;
459 uint8_t *upload_ptr = NULL;
460 struct pipe_constant_buffer cb;
461 cb.buffer = NULL;
462 cb.buffer_offset = 0;
463 cb.buffer_size = context->cso_shader.vs_const_used_size;
464 cb.user_buffer = context->vs_const_f;
465
466 if (context->swvp) {
467 prepare_vs_constants_userbuf_swvp(device);
468 return;
469 }
470
471 if (device->driver_caps.emulate_ucp) {
472 /* TODO: Avoid memcpy all time by storing directly into the array */
473 memcpy(&context->vs_const_f[4 * NINE_MAX_CONST_VS_SPE_OFFSET], &context->clip.ucp, sizeof(context->clip));
474 }
475 if (device->driver_caps.always_output_pointsize) {
476 context->vs_const_f[4 * (NINE_MAX_CONST_VS_SPE_OFFSET + 8)] =
477 CLAMP(asfloat(context->rs[D3DRS_POINTSIZE]),
478 asfloat(context->rs[D3DRS_POINTSIZE_MIN]),
479 asfloat(context->rs[D3DRS_POINTSIZE_MAX]));
480 }
481
482 if (context->changed.vs_const_i || context->changed.group & NINE_STATE_SWVP) {
483 int *idst = (int *)&context->vs_const_f[4 * device->max_vs_const_f];
484 memcpy(idst, context->vs_const_i, NINE_MAX_CONST_I * sizeof(int[4]));
485 context->changed.vs_const_i = 0;
486 }
487
488 if (context->changed.vs_const_b || context->changed.group & NINE_STATE_SWVP) {
489 int *idst = (int *)&context->vs_const_f[4 * device->max_vs_const_f];
490 uint32_t *bdst = (uint32_t *)&idst[4 * NINE_MAX_CONST_I];
491 memcpy(bdst, context->vs_const_b, NINE_MAX_CONST_B * sizeof(BOOL));
492 context->changed.vs_const_b = 0;
493 }
494
495 if (!cb.buffer_size)
496 return;
497
498 if (context->vs->lconstf.ranges) {
499 /* TODO: Can we make it so that we don't have to copy everything ? */
500 const struct nine_lconstf *lconstf = &(context->vs->lconstf);
501 const struct nine_range *r = lconstf->ranges;
502 unsigned n = 0;
503 float *dst = context->vs_lconstf_temp;
504 float *src = (float *)cb.user_buffer;
505 memcpy(dst, src, cb.buffer_size);
506 while (r) {
507 unsigned p = r->bgn;
508 unsigned c = r->end - r->bgn;
509 memcpy(&dst[p * 4], &lconstf->data[n * 4], c * 4 * sizeof(float));
510 n += c;
511 r = r->next;
512 }
513 cb.user_buffer = dst;
514 }
515
516 /* Note: We probably don't want to do separate memcpy to
517 * upload_ptr directly, if we have to copy some constants
518 * at random locations (context->vs->lconstf.ranges),
519 * to have efficient WC. Thus for this case we really want
520 * that intermediate buffer. */
521
522 u_upload_alloc(context->pipe->const_uploader,
523 0,
524 cb.buffer_size,
525 256, /* Be conservative about alignment */
526 &(cb.buffer_offset),
527 &(cb.buffer),
528 (void**)&upload_ptr);
529
530 assert(cb.buffer && upload_ptr);
531
532 if (!context->cso_shader.vs_const_ranges) {
533 memcpy(upload_ptr, cb.user_buffer, cb.buffer_size);
534 } else {
535 unsigned i = 0;
536 unsigned offset = 0;
537 while (context->cso_shader.vs_const_ranges[i*2+1] != 0) {
538 memcpy(upload_ptr+offset,
539 &((float*)cb.user_buffer)[4*context->cso_shader.vs_const_ranges[i*2]],
540 context->cso_shader.vs_const_ranges[i*2+1] * sizeof(float[4]));
541 offset += context->cso_shader.vs_const_ranges[i*2+1] * sizeof(float[4]);
542 i++;
543 }
544 }
545
546 u_upload_unmap(context->pipe->const_uploader);
547 cb.user_buffer = NULL;
548
549 /* Free previous resource */
550 pipe_resource_reference(&context->pipe_data.cb_vs.buffer, NULL);
551
552 context->pipe_data.cb_vs = cb;
553 context->changed.vs_const_f = 0;
554
555 context->changed.group &= ~NINE_STATE_VS_CONST;
556 context->commit |= NINE_STATE_COMMIT_CONST_VS;
557 }
558
559 static void
prepare_ps_constants_userbuf(struct NineDevice9 * device)560 prepare_ps_constants_userbuf(struct NineDevice9 *device)
561 {
562 struct nine_context *context = &device->context;
563 uint8_t *upload_ptr = NULL;
564 struct pipe_constant_buffer cb;
565 cb.buffer = NULL;
566 cb.buffer_offset = 0;
567 cb.buffer_size = context->cso_shader.ps_const_used_size;
568 cb.user_buffer = context->ps_const_f;
569
570 if (context->changed.ps_const_i) {
571 int *idst = (int *)&context->ps_const_f[4 * NINE_MAX_CONST_F_PS3];
572 memcpy(idst, context->ps_const_i, sizeof(context->ps_const_i));
573 context->changed.ps_const_i = 0;
574 }
575 if (context->changed.ps_const_b) {
576 int *idst = (int *)&context->ps_const_f[4 * NINE_MAX_CONST_F_PS3];
577 uint32_t *bdst = (uint32_t *)&idst[4 * NINE_MAX_CONST_I];
578 memcpy(bdst, context->ps_const_b, sizeof(context->ps_const_b));
579 context->changed.ps_const_b = 0;
580 }
581
582 /* Upload special constants needed to implement PS1.x instructions like TEXBEM,TEXBEML and BEM */
583 if (context->ps->bumpenvmat_needed)
584 memcpy(&context->ps_const_f[4 * NINE_MAX_CONST_PS_SPE_OFFSET], &device->context.bumpmap_vars, sizeof(device->context.bumpmap_vars));
585
586 if (context->ps->byte_code.version < 0x30 &&
587 context->rs[D3DRS_FOGENABLE]) {
588 float *dst = &context->ps_const_f[4 * (NINE_MAX_CONST_PS_SPE_OFFSET + 12)];
589
590 d3dcolor_to_rgba(dst, context->rs[D3DRS_FOGCOLOR]);
591 if (context->rs[D3DRS_FOGTABLEMODE] == D3DFOG_LINEAR) {
592 dst[4] = asfloat(context->rs[D3DRS_FOGEND]);
593 dst[5] = 1.0f / (asfloat(context->rs[D3DRS_FOGEND]) - asfloat(context->rs[D3DRS_FOGSTART]));
594 } else if (context->rs[D3DRS_FOGTABLEMODE] != D3DFOG_NONE) {
595 dst[4] = asfloat(context->rs[D3DRS_FOGDENSITY]);
596 }
597 }
598
599 context->ps_const_f[4 * (NINE_MAX_CONST_PS_SPE_OFFSET + 14)] = context->rs[D3DRS_ALPHAREF] / 255.f;
600
601 if (!cb.buffer_size)
602 return;
603
604 u_upload_alloc(context->pipe->const_uploader,
605 0,
606 cb.buffer_size,
607 256, /* Be conservative about alignment */
608 &(cb.buffer_offset),
609 &(cb.buffer),
610 (void**)&upload_ptr);
611
612 assert(cb.buffer && upload_ptr);
613
614 if (!context->cso_shader.ps_const_ranges) {
615 memcpy(upload_ptr, cb.user_buffer, cb.buffer_size);
616 } else {
617 unsigned i = 0;
618 unsigned offset = 0;
619 while (context->cso_shader.ps_const_ranges[i*2+1] != 0) {
620 memcpy(upload_ptr+offset,
621 &((float*)cb.user_buffer)[4*context->cso_shader.ps_const_ranges[i*2]],
622 context->cso_shader.ps_const_ranges[i*2+1] * sizeof(float[4]));
623 offset += context->cso_shader.ps_const_ranges[i*2+1] * sizeof(float[4]);
624 i++;
625 }
626 }
627
628 u_upload_unmap(context->pipe->const_uploader);
629 cb.user_buffer = NULL;
630
631 /* Free previous resource */
632 pipe_resource_reference(&context->pipe_data.cb_ps.buffer, NULL);
633
634 context->pipe_data.cb_ps = cb;
635 context->changed.ps_const_f = 0;
636
637 context->changed.group &= ~NINE_STATE_PS_CONST;
638 context->commit |= NINE_STATE_COMMIT_CONST_PS;
639 }
640
641 static inline uint32_t
prepare_vs(struct NineDevice9 * device,uint8_t shader_changed)642 prepare_vs(struct NineDevice9 *device, uint8_t shader_changed)
643 {
644 struct nine_context *context = &device->context;
645 struct NineVertexShader9 *vs = context->vs;
646 uint32_t changed_group = 0;
647 int has_key_changed = 0;
648
649 if (likely(context->programmable_vs))
650 has_key_changed = NineVertexShader9_UpdateKey(vs, device);
651
652 if (!shader_changed && !has_key_changed)
653 return 0;
654
655 /* likely because we dislike FF */
656 if (likely(context->programmable_vs)) {
657 context->cso_shader.vs = NineVertexShader9_GetVariant(vs,
658 &context->cso_shader.vs_const_ranges,
659 &context->cso_shader.vs_const_used_size);
660 } else {
661 vs = device->ff.vs;
662 context->cso_shader.vs = vs->ff_cso;
663 }
664
665 if (context->rs[NINED3DRS_VSPOINTSIZE] != vs->point_size) {
666 context->rs[NINED3DRS_VSPOINTSIZE] = vs->point_size;
667 changed_group |= NINE_STATE_RASTERIZER;
668 }
669 if (context->rs[NINED3DRS_POSITIONT] != vs->position_t) {
670 context->rs[NINED3DRS_POSITIONT] = vs->position_t;
671 if (!device->driver_caps.window_space_position_support &&
672 device->driver_caps.disabling_depth_clipping_support)
673 changed_group |= NINE_STATE_RASTERIZER;
674 }
675
676 if ((context->bound_samplers_mask_vs & vs->sampler_mask) != vs->sampler_mask)
677 /* Bound dummy sampler. */
678 changed_group |= NINE_STATE_SAMPLER;
679
680 context->commit |= NINE_STATE_COMMIT_VS;
681 return changed_group;
682 }
683
684 static inline uint32_t
prepare_ps(struct NineDevice9 * device,uint8_t shader_changed)685 prepare_ps(struct NineDevice9 *device, uint8_t shader_changed)
686 {
687 struct nine_context *context = &device->context;
688 struct NinePixelShader9 *ps = context->ps;
689 uint32_t changed_group = 0;
690 int has_key_changed = 0;
691
692 if (likely(ps))
693 has_key_changed = NinePixelShader9_UpdateKey(ps, context);
694
695 if (!shader_changed && !has_key_changed)
696 return 0;
697
698 if (likely(ps)) {
699 context->cso_shader.ps = NinePixelShader9_GetVariant(ps,
700 &context->cso_shader.ps_const_ranges,
701 &context->cso_shader.ps_const_used_size);
702 } else {
703 ps = device->ff.ps;
704 context->cso_shader.ps = ps->ff_cso;
705 }
706
707 if ((context->bound_samplers_mask_ps & ps->sampler_mask) != ps->sampler_mask)
708 /* Bound dummy sampler. */
709 changed_group |= NINE_STATE_SAMPLER;
710
711 context->commit |= NINE_STATE_COMMIT_PS;
712 return changed_group;
713 }
714
715 /* State preparation incremental */
716
717 /* State preparation + State commit */
718
719 static void
update_framebuffer(struct NineDevice9 * device,bool is_clear)720 update_framebuffer(struct NineDevice9 *device, bool is_clear)
721 {
722 struct nine_context *context = &device->context;
723 struct pipe_context *pipe = context->pipe;
724 struct pipe_framebuffer_state *fb = &context->pipe_data.fb;
725 unsigned i;
726 struct NineSurface9 *rt0 = context->rt[0];
727 unsigned w = rt0->desc.Width;
728 unsigned h = rt0->desc.Height;
729 unsigned nr_samples = rt0->base.info.nr_samples;
730 unsigned ps_mask = context->ps ? context->ps->rt_mask : 1;
731 unsigned mask = is_clear ? 0xf : ps_mask;
732 const int sRGB = context->rs[D3DRS_SRGBWRITEENABLE] ? 1 : 0;
733
734 DBG("\n");
735
736 context->rt_mask = 0x0;
737 fb->nr_cbufs = 0;
738
739 /* all render targets must have the same size and the depth buffer must be
740 * bigger. Multisample has to match, according to spec. But some apps do
741 * things wrong there, and no error is returned. The behaviour they get
742 * apparently is that depth buffer is disabled if it doesn't match.
743 * Surely the same for render targets. */
744
745 /* Special case: D3DFMT_NULL is used to bound no real render target,
746 * but render to depth buffer. We have to not take into account the render
747 * target info. TODO: know what should happen when there are several render targets
748 * and the first one is D3DFMT_NULL */
749 if (rt0->desc.Format == D3DFMT_NULL && context->ds) {
750 w = context->ds->desc.Width;
751 h = context->ds->desc.Height;
752 nr_samples = context->ds->base.info.nr_samples;
753 }
754
755 for (i = 0; i < device->caps.NumSimultaneousRTs; ++i) {
756 struct NineSurface9 *rt = context->rt[i];
757
758 if (rt && rt->desc.Format != D3DFMT_NULL && (mask & (1 << i)) &&
759 rt->desc.Width == w && rt->desc.Height == h &&
760 rt->base.info.nr_samples == nr_samples) {
761 fb->cbufs[i] = NineSurface9_GetSurface(rt, sRGB);
762 context->rt_mask |= 1 << i;
763 fb->nr_cbufs = i + 1;
764 } else {
765 /* Color outputs must match RT slot,
766 * drivers will have to handle NULL entries for GL, too.
767 */
768 fb->cbufs[i] = NULL;
769 }
770 }
771
772 if (context->ds && context->ds->desc.Width >= w &&
773 context->ds->desc.Height >= h &&
774 context->ds->base.info.nr_samples == nr_samples) {
775 fb->zsbuf = NineSurface9_GetSurface(context->ds, 0);
776 } else {
777 fb->zsbuf = NULL;
778 }
779
780 fb->width = w;
781 fb->height = h;
782
783 pipe->set_framebuffer_state(pipe, fb); /* XXX: cso ? */
784
785 if (is_clear && context->rt_mask == ps_mask)
786 context->changed.group &= ~NINE_STATE_FB;
787 }
788
789 static void
update_viewport(struct NineDevice9 * device)790 update_viewport(struct NineDevice9 *device)
791 {
792 struct nine_context *context = &device->context;
793 const D3DVIEWPORT9 *vport = &context->viewport;
794 struct pipe_viewport_state pvport;
795
796 /* D3D coordinates are:
797 * -1 .. +1 for X,Y and
798 * 0 .. +1 for Z (we use pipe_rasterizer_state.clip_halfz)
799 */
800 pvport.scale[0] = (float)vport->Width * 0.5f;
801 pvport.scale[1] = (float)vport->Height * -0.5f;
802 pvport.scale[2] = vport->MaxZ - vport->MinZ;
803 pvport.translate[0] = (float)vport->Width * 0.5f + (float)vport->X;
804 pvport.translate[1] = (float)vport->Height * 0.5f + (float)vport->Y;
805 pvport.translate[2] = vport->MinZ;
806 pvport.swizzle_x = PIPE_VIEWPORT_SWIZZLE_POSITIVE_X;
807 pvport.swizzle_y = PIPE_VIEWPORT_SWIZZLE_POSITIVE_Y;
808 pvport.swizzle_z = PIPE_VIEWPORT_SWIZZLE_POSITIVE_Z;
809 pvport.swizzle_w = PIPE_VIEWPORT_SWIZZLE_POSITIVE_W;
810
811 /* We found R600 and SI cards have some imprecision
812 * on the barycentric coordinates used for interpolation.
813 * Some shaders rely on having something precise.
814 * We found that the proprietary driver has the imprecision issue,
815 * except when the render target width and height are powers of two.
816 * It is using some sort of workaround for these cases
817 * which covers likely all the cases the applications rely
818 * on something precise.
819 * We haven't found the workaround, but it seems like it's better
820 * for applications if the imprecision is biased towards infinity
821 * instead of -infinity (which is what measured). So shift slightly
822 * the viewport: not enough to change rasterization result (in particular
823 * for multisampling), but enough to make the imprecision biased
824 * towards infinity. We do this shift only if render target width and
825 * height are powers of two.
826 * Solves 'red shadows' bug on UE3 games.
827 */
828 if (device->driver_bugs.buggy_barycentrics &&
829 ((vport->Width & (vport->Width-1)) == 0) &&
830 ((vport->Height & (vport->Height-1)) == 0)) {
831 pvport.translate[0] -= 1.0f / 128.0f;
832 pvport.translate[1] -= 1.0f / 128.0f;
833 }
834
835 cso_set_viewport(context->cso, &pvport);
836 }
837
838 /* Loop through VS inputs and pick the vertex elements with the declared
839 * usage from the vertex declaration, then insert the instance divisor from
840 * the stream source frequency setting.
841 */
842 static void
update_vertex_elements(struct NineDevice9 * device)843 update_vertex_elements(struct NineDevice9 *device)
844 {
845 struct nine_context *context = &device->context;
846 const struct NineVertexDeclaration9 *vdecl = device->context.vdecl;
847 const struct NineVertexShader9 *vs;
848 unsigned n, b, i;
849 int index;
850 int8_t vdecl_index_map[16]; /* vs->num_inputs <= 16 */
851 uint16_t used_streams = 0;
852 int dummy_vbo_stream = -1;
853 BOOL need_dummy_vbo = false;
854 struct cso_velems_state ve;
855 unsigned vtxbuf_mask;
856 unsigned vtxbuf_holes_map[PIPE_MAX_ATTRIBS];
857
858 context->stream_usage_mask = 0;
859 memset(vdecl_index_map, -1, 16);
860 vs = context->programmable_vs ? context->vs : device->ff.vs;
861
862 if (vdecl) {
863 for (n = 0; n < vs->num_inputs; ++n) {
864 DBG("looking up input %u (usage %u) from vdecl(%p)\n",
865 n, vs->input_map[n].ndecl, vdecl);
866
867 for (i = 0; i < vdecl->nelems; i++) {
868 if (vdecl->usage_map[i] == vs->input_map[n].ndecl) {
869 vdecl_index_map[n] = i;
870 used_streams |= BITFIELD_BIT(vdecl->elems[i].vertex_buffer_index);
871 break;
872 }
873 }
874 if (vdecl_index_map[n] < 0)
875 need_dummy_vbo = true;
876 }
877 } else {
878 /* No vertex declaration. Likely will never happen in practice,
879 * but we need not crash on this */
880 need_dummy_vbo = true;
881 }
882
883 if (need_dummy_vbo) {
884 u_foreach_bit(bit, BITFIELD_MASK(device->caps.MaxStreams) & ~used_streams) {
885 dummy_vbo_stream = bit;
886 break;
887 }
888 }
889 /* there are fewer vertex shader inputs than stream slots,
890 * so if we need a slot for the dummy vbo, we should have found one */
891 assert (!need_dummy_vbo || dummy_vbo_stream != -1);
892
893 /* calculate vtxbuf_holes_map to match for update_vertex_buffers() function */
894 i = 0;
895 vtxbuf_mask = context->vtxbuf_mask |
896 (need_dummy_vbo ? BITFIELD_BIT(dummy_vbo_stream) : 0);
897 u_foreach_bit(bit, vtxbuf_mask)
898 vtxbuf_holes_map[bit] = i++;
899
900 for (n = 0; n < vs->num_inputs; ++n) {
901 index = vdecl_index_map[n];
902 if (index >= 0) {
903 ve.velems[n] = vdecl->elems[index];
904 b = ve.velems[n].vertex_buffer_index;
905 ve.velems[n].vertex_buffer_index = vtxbuf_holes_map[b];
906 ve.velems[n].src_stride = context->vtxstride[b];
907 context->stream_usage_mask |= 1 << b;
908 /* XXX wine just uses 1 here: */
909 if (context->stream_freq[b] & D3DSTREAMSOURCE_INSTANCEDATA)
910 ve.velems[n].instance_divisor = context->stream_freq[b] & 0x7FFFFF;
911 } else {
912 /* if the vertex declaration is incomplete compared to what the
913 * vertex shader needs, we bind a dummy vbo with 0 0 0 0.
914 * This is not precised by the spec, but is the behaviour
915 * tested on win */
916 ve.velems[n].vertex_buffer_index = vtxbuf_holes_map[dummy_vbo_stream];
917 ve.velems[n].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
918 ve.velems[n].src_offset = 0;
919 ve.velems[n].src_stride = 0;
920 ve.velems[n].instance_divisor = 0;
921 ve.velems[n].dual_slot = false;
922 }
923 }
924
925 if (context->dummy_vbo_bound_at != dummy_vbo_stream) {
926 if (context->dummy_vbo_bound_at >= 0)
927 context->changed.vtxbuf |= 1 << context->dummy_vbo_bound_at;
928 if (dummy_vbo_stream >= 0)
929 context->changed.vtxbuf |= 1 << dummy_vbo_stream;
930 context->dummy_vbo_bound_at = dummy_vbo_stream;
931 }
932
933 ve.count = vs->num_inputs;
934 cso_set_vertex_elements(context->cso, &ve);
935 }
936
937 static void
update_vertex_buffers(struct NineDevice9 * device)938 update_vertex_buffers(struct NineDevice9 *device)
939 {
940 struct nine_context *context = &device->context;
941 struct pipe_context *pipe = context->pipe;
942 struct pipe_vertex_buffer vbuffer[PIPE_MAX_ATTRIBS];
943 unsigned vtxbuf_count;
944 unsigned mask, i, vtxbuf_i;
945
946 mask = context->vtxbuf_mask |
947 ((context->dummy_vbo_bound_at >= 0) ? BITFIELD_BIT(context->dummy_vbo_bound_at) : 0);
948 vtxbuf_count = util_bitcount(mask);
949
950 DBG("mask=%x\n", mask);
951 for (i = 0; mask; i++) {
952 vtxbuf_i = u_bit_scan(&mask);
953 if (vtxbuf_i == context->dummy_vbo_bound_at) {
954 vbuffer[i].buffer.resource = device->dummy_vbo;
955 vbuffer[i].is_user_buffer = false;
956 vbuffer[i].buffer_offset = 0;
957 } else {
958 memcpy(&vbuffer[i], &context->vtxbuf[vtxbuf_i], sizeof(struct pipe_vertex_buffer));
959 }
960 }
961
962 if (vtxbuf_count)
963 util_set_vertex_buffers(pipe, vtxbuf_count, false, vbuffer);
964 else
965 pipe->set_vertex_buffers(pipe, 0, NULL);
966
967 context->last_vtxbuf_count = vtxbuf_count;
968 context->changed.vtxbuf = 0;
969 }
970
971 static inline bool
update_sampler_derived(struct nine_context * context,unsigned s)972 update_sampler_derived(struct nine_context *context, unsigned s)
973 {
974 bool changed = false;
975
976 if (context->samp[s][NINED3DSAMP_SHADOW] != context->texture[s].shadow) {
977 changed = true;
978 context->samp[s][NINED3DSAMP_SHADOW] = context->texture[s].shadow;
979 }
980
981 if (context->samp[s][NINED3DSAMP_CUBETEX] !=
982 (context->texture[s].type == D3DRTYPE_CUBETEXTURE)) {
983 changed = true;
984 context->samp[s][NINED3DSAMP_CUBETEX] =
985 context->texture[s].type == D3DRTYPE_CUBETEXTURE;
986 }
987
988 if (context->samp[s][D3DSAMP_MIPFILTER] != D3DTEXF_NONE) {
989 int lod = context->samp[s][D3DSAMP_MAXMIPLEVEL] - context->texture[s].lod;
990 if (lod < 0)
991 lod = 0;
992 if (context->samp[s][NINED3DSAMP_MINLOD] != lod) {
993 changed = true;
994 context->samp[s][NINED3DSAMP_MINLOD] = lod;
995 }
996 } else {
997 context->changed.sampler[s] &= ~0x300; /* lod changes irrelevant */
998 }
999
1000 return changed;
1001 }
1002
1003 /* TODO: add sRGB override to pipe_sampler_state ? */
1004 static void
update_textures_and_samplers(struct NineDevice9 * device)1005 update_textures_and_samplers(struct NineDevice9 *device)
1006 {
1007 struct nine_context *context = &device->context;
1008 struct pipe_context *pipe = context->pipe;
1009 struct pipe_sampler_view *view[NINE_MAX_SAMPLERS];
1010 unsigned num_textures = 0;
1011 bool commit_samplers;
1012 uint16_t sampler_mask = context->ps ? context->ps->sampler_mask :
1013 device->ff.ps->sampler_mask;
1014
1015 commit_samplers = false;
1016 const uint16_t ps_mask = sampler_mask | context->enabled_samplers_mask_ps;
1017 context->bound_samplers_mask_ps = ps_mask;
1018 num_textures = util_last_bit(ps_mask);
1019 /* iterate over the enabled samplers */
1020 u_foreach_bit(i, context->enabled_samplers_mask_ps) {
1021 const unsigned s = NINE_SAMPLER_PS(i);
1022 int sRGB = context->samp[s][D3DSAMP_SRGBTEXTURE] ? 1 : 0;
1023
1024 view[i] = context->texture[s].view[sRGB];
1025
1026 if (update_sampler_derived(context, s) || (context->changed.sampler[s] & 0x05fe)) {
1027 context->changed.sampler[s] = 0;
1028 commit_samplers = true;
1029 nine_convert_sampler_state(context->cso, s, context->samp[s]);
1030 }
1031 }
1032 /* iterate over the dummy samplers */
1033 u_foreach_bit(i, sampler_mask & ~context->enabled_samplers_mask_ps) {
1034 const unsigned s = NINE_SAMPLER_PS(i);
1035 /* Bind dummy sampler. We do not bind dummy sampler when
1036 * it is not needed because it could add overhead. The
1037 * dummy sampler should have r=g=b=0 and a=1. We do not
1038 * unbind dummy sampler directly when they are not needed
1039 * anymore, but they're going to be removed as long as texture
1040 * or sampler states are changed. */
1041 view[i] = device->dummy_sampler_view;
1042
1043 cso_single_sampler(context->cso, PIPE_SHADER_FRAGMENT,
1044 s - NINE_SAMPLER_PS(0), &device->dummy_sampler_state);
1045
1046 commit_samplers = true;
1047 context->changed.sampler[s] = ~0;
1048 }
1049 /* fill in unused samplers */
1050 u_foreach_bit(i, BITFIELD_MASK(num_textures) & ~ps_mask)
1051 view[i] = NULL;
1052
1053 pipe->set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0, num_textures,
1054 num_textures < context->enabled_sampler_count_ps ? context->enabled_sampler_count_ps - num_textures : 0,
1055 false, view);
1056 context->enabled_sampler_count_ps = num_textures;
1057
1058 if (commit_samplers)
1059 cso_single_sampler_done(context->cso, PIPE_SHADER_FRAGMENT);
1060
1061 commit_samplers = false;
1062 sampler_mask = context->programmable_vs ? context->vs->sampler_mask : 0;
1063 const uint16_t vs_mask = sampler_mask | context->enabled_samplers_mask_vs;
1064 context->bound_samplers_mask_vs = vs_mask;
1065 num_textures = util_last_bit(vs_mask);
1066 u_foreach_bit(i, context->enabled_samplers_mask_vs) {
1067 const unsigned s = NINE_SAMPLER_VS(i);
1068 int sRGB = context->samp[s][D3DSAMP_SRGBTEXTURE] ? 1 : 0;
1069
1070 view[i] = context->texture[s].view[sRGB];
1071
1072 if (update_sampler_derived(context, s) || (context->changed.sampler[s] & 0x05fe)) {
1073 context->changed.sampler[s] = 0;
1074 commit_samplers = true;
1075 nine_convert_sampler_state(context->cso, s, context->samp[s]);
1076 }
1077 }
1078 u_foreach_bit(i, sampler_mask & ~context->enabled_samplers_mask_vs) {
1079 const unsigned s = NINE_SAMPLER_VS(i);
1080 /* Bind dummy sampler. We do not bind dummy sampler when
1081 * it is not needed because it could add overhead. The
1082 * dummy sampler should have r=g=b=0 and a=1. We do not
1083 * unbind dummy sampler directly when they are not needed
1084 * anymore, but they're going to be removed as long as texture
1085 * or sampler states are changed. */
1086 view[i] = device->dummy_sampler_view;
1087
1088 cso_single_sampler(context->cso, PIPE_SHADER_VERTEX,
1089 s - NINE_SAMPLER_VS(0), &device->dummy_sampler_state);
1090
1091 commit_samplers = true;
1092 context->changed.sampler[s] = ~0;
1093 }
1094 /* fill in unused samplers */
1095 u_foreach_bit(i, BITFIELD_MASK(num_textures) & ~vs_mask)
1096 view[i] = NULL;
1097
1098 pipe->set_sampler_views(pipe, PIPE_SHADER_VERTEX, 0, num_textures,
1099 num_textures < context->enabled_sampler_count_vs ? context->enabled_sampler_count_vs - num_textures : 0,
1100 false, view);
1101 context->enabled_sampler_count_vs = num_textures;
1102
1103 if (commit_samplers)
1104 cso_single_sampler_done(context->cso, PIPE_SHADER_VERTEX);
1105 }
1106
1107 /* State commit only */
1108
1109 static inline void
commit_blend(struct NineDevice9 * device)1110 commit_blend(struct NineDevice9 *device)
1111 {
1112 struct nine_context *context = &device->context;
1113
1114 cso_set_blend(context->cso, &context->pipe_data.blend);
1115 }
1116
1117 static inline void
commit_dsa(struct NineDevice9 * device)1118 commit_dsa(struct NineDevice9 *device)
1119 {
1120 struct nine_context *context = &device->context;
1121
1122 cso_set_depth_stencil_alpha(context->cso, &context->pipe_data.dsa);
1123 }
1124
1125 static inline void
commit_scissor(struct NineDevice9 * device)1126 commit_scissor(struct NineDevice9 *device)
1127 {
1128 struct nine_context *context = &device->context;
1129 struct pipe_context *pipe = context->pipe;
1130
1131 pipe->set_scissor_states(pipe, 0, 1, &context->scissor);
1132 }
1133
1134 static inline void
commit_rasterizer(struct NineDevice9 * device)1135 commit_rasterizer(struct NineDevice9 *device)
1136 {
1137 struct nine_context *context = &device->context;
1138
1139 cso_set_rasterizer(context->cso, &context->pipe_data.rast);
1140 }
1141
1142 static inline void
commit_vs_constants(struct NineDevice9 * device)1143 commit_vs_constants(struct NineDevice9 *device)
1144 {
1145 struct nine_context *context = &device->context;
1146 struct pipe_context *pipe = context->pipe;
1147
1148 if (unlikely(!context->programmable_vs))
1149 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, false, &context->pipe_data.cb_vs_ff);
1150 else {
1151 if (context->swvp) {
1152 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, false, &context->pipe_data.cb0_swvp);
1153 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 1, false, &context->pipe_data.cb1_swvp);
1154 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 2, false, &context->pipe_data.cb2_swvp);
1155 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 3, false, &context->pipe_data.cb3_swvp);
1156 } else {
1157 pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, false, &context->pipe_data.cb_vs);
1158 }
1159 }
1160 }
1161
1162 static inline void
commit_ps_constants(struct NineDevice9 * device)1163 commit_ps_constants(struct NineDevice9 *device)
1164 {
1165 struct nine_context *context = &device->context;
1166 struct pipe_context *pipe = context->pipe;
1167
1168 if (unlikely(!context->ps))
1169 pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, false, &context->pipe_data.cb_ps_ff);
1170 else
1171 pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, false, &context->pipe_data.cb_ps);
1172 }
1173
1174 static inline void
commit_vs(struct NineDevice9 * device)1175 commit_vs(struct NineDevice9 *device)
1176 {
1177 struct nine_context *context = &device->context;
1178 assert(context->cso_shader.vs);
1179
1180 context->pipe->bind_vs_state(context->pipe, context->cso_shader.vs);
1181 }
1182
1183
1184 static inline void
commit_ps(struct NineDevice9 * device)1185 commit_ps(struct NineDevice9 *device)
1186 {
1187 struct nine_context *context = &device->context;
1188
1189 context->pipe->bind_fs_state(context->pipe, context->cso_shader.ps);
1190 }
1191 /* State Update */
1192
1193 #define NINE_STATE_SHADER_CHANGE_VS \
1194 (NINE_STATE_VS | \
1195 NINE_STATE_TEXTURE | \
1196 NINE_STATE_VS_PARAMS_MISC | \
1197 NINE_STATE_SWVP)
1198
1199 #define NINE_STATE_SHADER_CHANGE_PS \
1200 (NINE_STATE_PS | \
1201 NINE_STATE_TEXTURE | \
1202 NINE_STATE_PS_PARAMS_MISC)
1203
1204 #define NINE_STATE_FREQUENT \
1205 (NINE_STATE_RASTERIZER | \
1206 NINE_STATE_TEXTURE | \
1207 NINE_STATE_SAMPLER | \
1208 NINE_STATE_VS_CONST | \
1209 NINE_STATE_PS_CONST | \
1210 NINE_STATE_MULTISAMPLE)
1211
1212 #define NINE_STATE_COMMON \
1213 (NINE_STATE_FB | \
1214 NINE_STATE_BLEND | \
1215 NINE_STATE_DSA | \
1216 NINE_STATE_VIEWPORT | \
1217 NINE_STATE_VDECL | \
1218 NINE_STATE_IDXBUF | \
1219 NINE_STATE_STREAMFREQ)
1220
1221 #define NINE_STATE_RARE \
1222 (NINE_STATE_SCISSOR | \
1223 NINE_STATE_BLEND_COLOR | \
1224 NINE_STATE_STENCIL_REF | \
1225 NINE_STATE_SAMPLE_MASK)
1226
1227 static void
nine_update_state(struct NineDevice9 * device)1228 nine_update_state(struct NineDevice9 *device)
1229 {
1230 struct nine_context *context = &device->context;
1231 struct pipe_context *pipe = context->pipe;
1232 uint32_t group;
1233
1234 DBG("changed state groups: %x\n", context->changed.group);
1235
1236 /* NOTE: We may want to use the cso cache for everything, or let
1237 * NineDevice9.RestoreNonCSOState actually set the states, then we wouldn't
1238 * have to care about state being clobbered here and could merge this back
1239 * into update_textures. Except, we also need to re-validate textures that
1240 * may be dirty anyway, even if no texture bindings changed.
1241 */
1242
1243 /* ff_update may change VS/PS dirty bits */
1244 if (unlikely(!context->programmable_vs || !context->ps))
1245 nine_ff_update(device);
1246 group = context->changed.group;
1247
1248 if (group & (NINE_STATE_SHADER_CHANGE_VS | NINE_STATE_SHADER_CHANGE_PS)) {
1249 if (group & NINE_STATE_SHADER_CHANGE_VS)
1250 group |= prepare_vs(device, (group & NINE_STATE_VS) != 0); /* may set NINE_STATE_RASTERIZER and NINE_STATE_SAMPLER*/
1251 if (group & NINE_STATE_SHADER_CHANGE_PS)
1252 group |= prepare_ps(device, (group & NINE_STATE_PS) != 0);
1253 }
1254
1255 if (group & (NINE_STATE_COMMON | NINE_STATE_VS)) {
1256 if (group & NINE_STATE_FB)
1257 update_framebuffer(device, false);
1258 if (group & NINE_STATE_BLEND)
1259 prepare_blend(device);
1260 if (group & NINE_STATE_DSA)
1261 prepare_dsa(device);
1262 if (group & NINE_STATE_VIEWPORT)
1263 update_viewport(device);
1264 if (group & (NINE_STATE_VDECL | NINE_STATE_VS | NINE_STATE_STREAMFREQ))
1265 update_vertex_elements(device);
1266 }
1267
1268 if (likely(group & (NINE_STATE_FREQUENT | NINE_STATE_VS | NINE_STATE_PS | NINE_STATE_SWVP))) {
1269 if (group & NINE_STATE_MULTISAMPLE)
1270 group |= check_multisample(device);
1271 if (group & NINE_STATE_RASTERIZER)
1272 prepare_rasterizer(device);
1273 if (group & (NINE_STATE_TEXTURE | NINE_STATE_SAMPLER))
1274 update_textures_and_samplers(device);
1275 if ((group & (NINE_STATE_VS_CONST | NINE_STATE_VS | NINE_STATE_SWVP)) && context->programmable_vs)
1276 prepare_vs_constants_userbuf(device);
1277 if ((group & (NINE_STATE_PS_CONST | NINE_STATE_PS)) && context->ps)
1278 prepare_ps_constants_userbuf(device);
1279 }
1280
1281 if (context->changed.vtxbuf)
1282 update_vertex_buffers(device);
1283
1284 if (context->commit & NINE_STATE_COMMIT_BLEND)
1285 commit_blend(device);
1286 if (context->commit & NINE_STATE_COMMIT_DSA)
1287 commit_dsa(device);
1288 if (context->commit & NINE_STATE_COMMIT_RASTERIZER)
1289 commit_rasterizer(device);
1290 if (context->commit & NINE_STATE_COMMIT_CONST_VS)
1291 commit_vs_constants(device);
1292 if (context->commit & NINE_STATE_COMMIT_CONST_PS)
1293 commit_ps_constants(device);
1294 if (context->commit & NINE_STATE_COMMIT_VS)
1295 commit_vs(device);
1296 if (context->commit & NINE_STATE_COMMIT_PS)
1297 commit_ps(device);
1298
1299 context->commit = 0;
1300
1301 if (unlikely(context->changed.ucp)) {
1302 pipe->set_clip_state(pipe, &context->clip);
1303 context->changed.ucp = false;
1304 }
1305
1306 if (unlikely(group & NINE_STATE_RARE)) {
1307 if (group & NINE_STATE_SCISSOR)
1308 commit_scissor(device);
1309 if (group & NINE_STATE_BLEND_COLOR) {
1310 struct pipe_blend_color color;
1311 d3dcolor_to_rgba(&color.color[0], context->rs[D3DRS_BLENDFACTOR]);
1312 pipe->set_blend_color(pipe, &color);
1313 }
1314 if (group & NINE_STATE_SAMPLE_MASK) {
1315 if (context->rt[0]->desc.MultiSampleType <= D3DMULTISAMPLE_NONMASKABLE) {
1316 pipe->set_sample_mask(pipe, ~0);
1317 } else {
1318 pipe->set_sample_mask(pipe, context->rs[D3DRS_MULTISAMPLEMASK]);
1319 }
1320 }
1321 if (group & NINE_STATE_STENCIL_REF) {
1322 struct pipe_stencil_ref ref;
1323 ref.ref_value[0] = context->rs[D3DRS_STENCILREF];
1324 ref.ref_value[1] = ref.ref_value[0];
1325 pipe->set_stencil_ref(pipe, ref);
1326 }
1327 }
1328
1329 context->changed.group &=
1330 (NINE_STATE_FF | NINE_STATE_VS_CONST | NINE_STATE_PS_CONST);
1331
1332 DBG("finished\n");
1333 }
1334
1335 #define RESZ_CODE 0x7fa05000
1336
1337 static void
NineDevice9_ResolveZ(struct NineDevice9 * device)1338 NineDevice9_ResolveZ( struct NineDevice9 *device )
1339 {
1340 struct nine_context *context = &device->context;
1341 const struct util_format_description *desc;
1342 struct NineSurface9 *source = context->ds;
1343 struct pipe_resource *src, *dst;
1344 struct pipe_blit_info blit;
1345
1346 DBG("RESZ resolve\n");
1347
1348 if (!source || !context->texture[0].enabled ||
1349 context->texture[0].type != D3DRTYPE_TEXTURE)
1350 return;
1351
1352 src = source->base.resource;
1353 dst = context->texture[0].resource;
1354
1355 if (!src || !dst)
1356 return;
1357
1358 /* check dst is depth format. we know already for src */
1359 desc = util_format_description(dst->format);
1360 if (desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS)
1361 return;
1362
1363 memset(&blit, 0, sizeof(blit));
1364 blit.src.resource = src;
1365 blit.src.level = 0;
1366 blit.src.format = src->format;
1367 blit.src.box.z = 0;
1368 blit.src.box.depth = 1;
1369 blit.src.box.x = 0;
1370 blit.src.box.y = 0;
1371 blit.src.box.width = src->width0;
1372 blit.src.box.height = src->height0;
1373
1374 blit.dst.resource = dst;
1375 blit.dst.level = 0;
1376 blit.dst.format = dst->format;
1377 blit.dst.box.z = 0;
1378 blit.dst.box.depth = 1;
1379 blit.dst.box.x = 0;
1380 blit.dst.box.y = 0;
1381 blit.dst.box.width = dst->width0;
1382 blit.dst.box.height = dst->height0;
1383
1384 blit.mask = PIPE_MASK_ZS;
1385 blit.filter = PIPE_TEX_FILTER_NEAREST;
1386 blit.scissor_enable = false;
1387
1388 context->pipe->blit(context->pipe, &blit);
1389 }
1390
1391 #define ALPHA_TO_COVERAGE_ENABLE MAKEFOURCC('A', '2', 'M', '1')
1392 #define ALPHA_TO_COVERAGE_DISABLE MAKEFOURCC('A', '2', 'M', '0')
1393 #define FETCH4_ENABLE MAKEFOURCC('G', 'E', 'T', '4')
1394 #define FETCH4_DISABLE MAKEFOURCC('G', 'E', 'T', '1')
1395
1396 /* Nine_context functions.
1397 * Serialized through CSMT macros.
1398 */
1399
1400 static void
1401 nine_context_set_texture_apply(struct NineDevice9 *device,
1402 DWORD stage,
1403 DWORD fetch4_shadow_enabled,
1404 DWORD lod,
1405 D3DRESOURCETYPE type,
1406 uint8_t pstype,
1407 struct pipe_resource *res,
1408 struct pipe_sampler_view *view0,
1409 struct pipe_sampler_view *view1);
1410
1411 static void
1412 nine_context_set_pixel_shader_constant_i_transformed(struct NineDevice9 *device,
1413 UINT StartRegister,
1414 const int *pConstantData,
1415 unsigned pConstantData_size,
1416 UINT Vector4iCount);
1417
CSMT_ITEM_NO_WAIT(nine_context_set_render_state,ARG_VAL (D3DRENDERSTATETYPE,State),ARG_VAL (DWORD,Value))1418 CSMT_ITEM_NO_WAIT(nine_context_set_render_state,
1419 ARG_VAL(D3DRENDERSTATETYPE, State),
1420 ARG_VAL(DWORD, Value))
1421 {
1422 struct nine_context *context = &device->context;
1423
1424 /* Amd hacks (equivalent to GL extensions) */
1425 if (unlikely(State == D3DRS_POINTSIZE)) {
1426 if (Value == RESZ_CODE) {
1427 NineDevice9_ResolveZ(device);
1428 return;
1429 }
1430
1431 /* NINED3DRS_ALPHACOVERAGE:
1432 * bit 0: NVIDIA alpha to coverage
1433 * bit 1: NVIDIA ATOC state active
1434 * bit 2: AMD alpha to coverage
1435 * These need to be separate else the set of states to
1436 * disable NVIDIA alpha to coverage can disable the AMD one */
1437 if (Value == ALPHA_TO_COVERAGE_ENABLE ||
1438 Value == ALPHA_TO_COVERAGE_DISABLE) {
1439 context->rs[NINED3DRS_ALPHACOVERAGE] &= 3;
1440 context->rs[NINED3DRS_ALPHACOVERAGE] |= (Value == ALPHA_TO_COVERAGE_ENABLE) ? 4 : 0;
1441 context->changed.group |= NINE_STATE_BLEND;
1442 return;
1443 }
1444 }
1445
1446 /* NV hack */
1447 if (unlikely(State == D3DRS_ADAPTIVETESS_Y)) {
1448 if (Value == D3DFMT_ATOC || (Value == D3DFMT_UNKNOWN && context->rs[NINED3DRS_ALPHACOVERAGE] & 3)) {
1449 context->rs[NINED3DRS_ALPHACOVERAGE] &= 4;
1450 context->rs[NINED3DRS_ALPHACOVERAGE] |=
1451 ((Value == D3DFMT_ATOC) ? 3 : 0) & (context->rs[D3DRS_ALPHATESTENABLE] ? 3 : 2);
1452 context->changed.group |= NINE_STATE_BLEND;
1453 return;
1454 }
1455 }
1456 if (unlikely(State == D3DRS_ALPHATESTENABLE && (context->rs[NINED3DRS_ALPHACOVERAGE] & 2))) {
1457 context->rs[NINED3DRS_ALPHACOVERAGE] &= 6;
1458 context->rs[NINED3DRS_ALPHACOVERAGE] |= (Value ? 1 : 0);
1459 }
1460
1461 context->rs[State] = nine_fix_render_state_value(State, Value);
1462 context->changed.group |= nine_render_state_group[State];
1463
1464 if (device->driver_caps.alpha_test_emulation) {
1465 if (State == D3DRS_ALPHATESTENABLE || State == D3DRS_ALPHAFUNC) {
1466 context->rs[NINED3DRS_EMULATED_ALPHATEST] = context->rs[D3DRS_ALPHATESTENABLE] ?
1467 d3dcmpfunc_to_pipe_func(context->rs[D3DRS_ALPHAFUNC]) : 7;
1468 context->changed.group |= NINE_STATE_PS_PARAMS_MISC | NINE_STATE_PS_CONST | NINE_STATE_FF_SHADER;
1469 }
1470 if (State == D3DRS_ALPHAREF)
1471 context->changed.group |= NINE_STATE_PS_CONST | NINE_STATE_FF_PS_CONSTS;
1472 }
1473
1474 if (device->driver_caps.always_output_pointsize) {
1475 if (State == D3DRS_POINTSIZE || State == D3DRS_POINTSIZE_MIN || State == D3DRS_POINTSIZE_MAX)
1476 context->changed.group |= NINE_STATE_VS_CONST;
1477 }
1478
1479 if (device->driver_caps.emulate_ucp && State == D3DRS_CLIPPLANEENABLE)
1480 context->changed.group |= NINE_STATE_VS_PARAMS_MISC | NINE_STATE_VS_CONST;
1481 }
1482
CSMT_ITEM_NO_WAIT(nine_context_set_texture_apply,ARG_VAL (DWORD,stage),ARG_VAL (DWORD,fetch4_shadow_enabled),ARG_VAL (DWORD,lod),ARG_VAL (D3DRESOURCETYPE,type),ARG_VAL (uint8_t,pstype),ARG_BIND_RES (struct pipe_resource,res),ARG_BIND_VIEW (struct pipe_sampler_view,view0),ARG_BIND_VIEW (struct pipe_sampler_view,view1))1483 CSMT_ITEM_NO_WAIT(nine_context_set_texture_apply,
1484 ARG_VAL(DWORD, stage),
1485 ARG_VAL(DWORD, fetch4_shadow_enabled),
1486 ARG_VAL(DWORD, lod),
1487 ARG_VAL(D3DRESOURCETYPE, type),
1488 ARG_VAL(uint8_t, pstype),
1489 ARG_BIND_RES(struct pipe_resource, res),
1490 ARG_BIND_VIEW(struct pipe_sampler_view, view0),
1491 ARG_BIND_VIEW(struct pipe_sampler_view, view1))
1492 {
1493 struct nine_context *context = &device->context;
1494 uint enabled = fetch4_shadow_enabled & 1;
1495 uint shadow = (fetch4_shadow_enabled >> 1) & 1;
1496 uint fetch4_compatible = (fetch4_shadow_enabled >> 2) & 1;
1497
1498 context->texture[stage].enabled = enabled;
1499 if (enabled) {
1500 if (stage < NINE_MAX_SAMPLERS_PS)
1501 context->enabled_samplers_mask_ps |= BITFIELD_BIT(stage - NINE_SAMPLER_PS(0));
1502 else if (stage >= NINE_SAMPLER_VS(0))
1503 context->enabled_samplers_mask_vs |= BITFIELD_BIT(stage - NINE_SAMPLER_VS(0));
1504 } else {
1505 if (stage < NINE_MAX_SAMPLERS_PS)
1506 context->enabled_samplers_mask_ps &= ~BITFIELD_BIT(stage - NINE_SAMPLER_PS(0));
1507 else if (stage >= NINE_SAMPLER_VS(0))
1508 context->enabled_samplers_mask_vs &= ~BITFIELD_BIT(stage - NINE_SAMPLER_VS(0));
1509 }
1510 context->samplers_shadow &= ~(1 << stage);
1511 context->samplers_shadow |= shadow << stage;
1512 context->samplers_fetch4 &= ~(1 << stage);
1513 context->samplers_fetch4 |= fetch4_compatible << stage;
1514 context->texture[stage].shadow = shadow;
1515 context->texture[stage].lod = lod;
1516 context->texture[stage].type = type;
1517 context->texture[stage].pstype = pstype;
1518 pipe_resource_reference(&context->texture[stage].resource, res);
1519 pipe_sampler_view_reference(&context->texture[stage].view[0], view0);
1520 pipe_sampler_view_reference(&context->texture[stage].view[1], view1);
1521
1522 context->changed.group |= NINE_STATE_TEXTURE;
1523 }
1524
1525 void
nine_context_set_texture(struct NineDevice9 * device,DWORD Stage,struct NineBaseTexture9 * tex)1526 nine_context_set_texture(struct NineDevice9 *device,
1527 DWORD Stage,
1528 struct NineBaseTexture9 *tex)
1529 {
1530 DWORD fetch4_shadow_enabled = 0;
1531 DWORD lod = 0;
1532 D3DRESOURCETYPE type = D3DRTYPE_TEXTURE;
1533 uint8_t pstype = 0;
1534 struct pipe_resource *res = NULL;
1535 struct pipe_sampler_view *view0 = NULL, *view1 = NULL;
1536
1537 /* For managed pool, the data can be initially incomplete.
1538 * In that case, the texture is rebound later
1539 * (in NineBaseTexture9_Validate/NineBaseTexture9_UploadSelf). */
1540 if (tex && tex->base.resource) {
1541 fetch4_shadow_enabled = 1;
1542 fetch4_shadow_enabled |= tex->shadow << 1;
1543 fetch4_shadow_enabled |= tex->fetch4_compatible << 2;
1544 lod = tex->managed.lod;
1545 type = tex->base.type;
1546 pstype = tex->pstype;
1547 res = tex->base.resource;
1548 view0 = NineBaseTexture9_GetSamplerView(tex, 0);
1549 view1 = NineBaseTexture9_GetSamplerView(tex, 1);
1550 }
1551
1552 nine_context_set_texture_apply(device, Stage,
1553 fetch4_shadow_enabled,
1554 lod, type, pstype,
1555 res, view0, view1);
1556 }
1557
CSMT_ITEM_NO_WAIT(nine_context_set_sampler_state,ARG_VAL (DWORD,Sampler),ARG_VAL (D3DSAMPLERSTATETYPE,Type),ARG_VAL (DWORD,Value))1558 CSMT_ITEM_NO_WAIT(nine_context_set_sampler_state,
1559 ARG_VAL(DWORD, Sampler),
1560 ARG_VAL(D3DSAMPLERSTATETYPE, Type),
1561 ARG_VAL(DWORD, Value))
1562 {
1563 struct nine_context *context = &device->context;
1564
1565 if (unlikely(Type == D3DSAMP_MIPMAPLODBIAS)) {
1566 if (Value == FETCH4_ENABLE ||
1567 Value == FETCH4_DISABLE) {
1568 context->rs[NINED3DRS_FETCH4] &= ~(1 << Sampler);
1569 context->rs[NINED3DRS_FETCH4] |= (Value == FETCH4_ENABLE) << Sampler;
1570 context->changed.group |= NINE_STATE_PS_PARAMS_MISC;
1571 if (Value == FETCH4_ENABLE)
1572 WARN_ONCE("FETCH4 support is incomplete. Please report if buggy shadows.");
1573 return;
1574 }
1575 }
1576
1577 if (unlikely(!nine_check_sampler_state_value(Type, Value)))
1578 return;
1579
1580 context->samp[Sampler][Type] = Value;
1581 context->changed.group |= NINE_STATE_SAMPLER;
1582 context->changed.sampler[Sampler] |= 1 << Type;
1583 }
1584
CSMT_ITEM_NO_WAIT(nine_context_set_stream_source_apply,ARG_VAL (UINT,StreamNumber),ARG_BIND_RES (struct pipe_resource,res),ARG_VAL (UINT,OffsetInBytes),ARG_VAL (UINT,Stride))1585 CSMT_ITEM_NO_WAIT(nine_context_set_stream_source_apply,
1586 ARG_VAL(UINT, StreamNumber),
1587 ARG_BIND_RES(struct pipe_resource, res),
1588 ARG_VAL(UINT, OffsetInBytes),
1589 ARG_VAL(UINT, Stride))
1590 {
1591 struct nine_context *context = &device->context;
1592 const unsigned i = StreamNumber;
1593
1594 /* For normal draws, these tests are useless,
1595 * but not for *Up draws */
1596 if (context->vtxbuf[i].buffer.resource == res &&
1597 context->vtxbuf[i].buffer_offset == OffsetInBytes &&
1598 context->vtxstride[i] == Stride)
1599 return;
1600
1601 if (context->vtxstride[i] != Stride) {
1602 context->vtxstride[i] = Stride;
1603 context->changed.group |= NINE_STATE_VDECL;
1604 }
1605 context->vtxbuf[i].buffer_offset = OffsetInBytes;
1606 pipe_resource_reference(&context->vtxbuf[i].buffer.resource, res);
1607
1608 context->changed.vtxbuf |= 1 << StreamNumber;
1609 if (res)
1610 context->vtxbuf_mask |= 1 << StreamNumber;
1611 else
1612 context->vtxbuf_mask &= ~(1 << StreamNumber);
1613 }
1614
1615 void
nine_context_set_stream_source(struct NineDevice9 * device,UINT StreamNumber,struct NineVertexBuffer9 * pVBuf9,UINT OffsetInBytes,UINT Stride)1616 nine_context_set_stream_source(struct NineDevice9 *device,
1617 UINT StreamNumber,
1618 struct NineVertexBuffer9 *pVBuf9,
1619 UINT OffsetInBytes,
1620 UINT Stride)
1621 {
1622 struct pipe_resource *res = NULL;
1623 unsigned offset = 0;
1624
1625 if (pVBuf9)
1626 res = NineVertexBuffer9_GetResource(pVBuf9, &offset);
1627 /* in the future when there is internal offset, add it
1628 * to OffsetInBytes */
1629
1630 nine_context_set_stream_source_apply(device, StreamNumber,
1631 res, offset + OffsetInBytes,
1632 Stride);
1633 }
1634
CSMT_ITEM_NO_WAIT(nine_context_set_stream_source_freq,ARG_VAL (UINT,StreamNumber),ARG_VAL (UINT,Setting))1635 CSMT_ITEM_NO_WAIT(nine_context_set_stream_source_freq,
1636 ARG_VAL(UINT, StreamNumber),
1637 ARG_VAL(UINT, Setting))
1638 {
1639 struct nine_context *context = &device->context;
1640
1641 context->stream_freq[StreamNumber] = Setting;
1642
1643 if (Setting & D3DSTREAMSOURCE_INSTANCEDATA)
1644 context->stream_instancedata_mask |= 1 << StreamNumber;
1645 else
1646 context->stream_instancedata_mask &= ~(1 << StreamNumber);
1647
1648 if (StreamNumber != 0)
1649 context->changed.group |= NINE_STATE_STREAMFREQ;
1650 }
1651
CSMT_ITEM_NO_WAIT(nine_context_set_indices_apply,ARG_BIND_RES (struct pipe_resource,res),ARG_VAL (UINT,IndexSize),ARG_VAL (UINT,OffsetInBytes))1652 CSMT_ITEM_NO_WAIT(nine_context_set_indices_apply,
1653 ARG_BIND_RES(struct pipe_resource, res),
1654 ARG_VAL(UINT, IndexSize),
1655 ARG_VAL(UINT, OffsetInBytes))
1656 {
1657 struct nine_context *context = &device->context;
1658
1659 context->index_size = IndexSize;
1660 context->index_offset = OffsetInBytes;
1661 pipe_resource_reference(&context->idxbuf, res);
1662
1663 context->changed.group |= NINE_STATE_IDXBUF;
1664 }
1665
1666 void
nine_context_set_indices(struct NineDevice9 * device,struct NineIndexBuffer9 * idxbuf)1667 nine_context_set_indices(struct NineDevice9 *device,
1668 struct NineIndexBuffer9 *idxbuf)
1669 {
1670 struct pipe_resource *res = NULL;
1671 UINT IndexSize = 0;
1672 unsigned OffsetInBytes = 0;
1673
1674 if (idxbuf) {
1675 res = NineIndexBuffer9_GetBuffer(idxbuf, &OffsetInBytes);
1676 IndexSize = idxbuf->index_size;
1677 }
1678
1679 nine_context_set_indices_apply(device, res, IndexSize, OffsetInBytes);
1680 }
1681
CSMT_ITEM_NO_WAIT(nine_context_set_vertex_declaration,ARG_BIND_REF (struct NineVertexDeclaration9,vdecl))1682 CSMT_ITEM_NO_WAIT(nine_context_set_vertex_declaration,
1683 ARG_BIND_REF(struct NineVertexDeclaration9, vdecl))
1684 {
1685 struct nine_context *context = &device->context;
1686 BOOL was_programmable_vs = context->programmable_vs;
1687
1688 nine_bind(&context->vdecl, vdecl);
1689
1690 context->programmable_vs = context->vs && !(context->vdecl && context->vdecl->position_t);
1691 if (was_programmable_vs != context->programmable_vs) {
1692 context->commit |= NINE_STATE_COMMIT_CONST_VS;
1693 context->changed.group |= NINE_STATE_VS;
1694 }
1695
1696 context->changed.group |= NINE_STATE_VDECL;
1697 }
1698
CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader,ARG_BIND_REF (struct NineVertexShader9,pShader))1699 CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader,
1700 ARG_BIND_REF(struct NineVertexShader9, pShader))
1701 {
1702 struct nine_context *context = &device->context;
1703 BOOL was_programmable_vs = context->programmable_vs;
1704
1705 nine_bind(&context->vs, pShader);
1706
1707 context->programmable_vs = context->vs && !(context->vdecl && context->vdecl->position_t);
1708
1709 /* ff -> non-ff: commit back non-ff constants */
1710 if (!was_programmable_vs && context->programmable_vs)
1711 context->commit |= NINE_STATE_COMMIT_CONST_VS;
1712
1713 context->changed.group |= NINE_STATE_VS;
1714 }
1715
CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_f,ARG_VAL (UINT,StartRegister),ARG_MEM (float,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,Vector4fCount))1716 CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_f,
1717 ARG_VAL(UINT, StartRegister),
1718 ARG_MEM(float, pConstantData),
1719 ARG_MEM_SIZE(unsigned, pConstantData_size),
1720 ARG_VAL(UINT, Vector4fCount))
1721 {
1722 struct nine_context *context = &device->context;
1723 float *vs_const_f = device->may_swvp ? context->vs_const_f_swvp : context->vs_const_f;
1724
1725 memcpy(&vs_const_f[StartRegister * 4],
1726 pConstantData,
1727 pConstantData_size);
1728
1729 if (device->may_swvp) {
1730 Vector4fCount = MIN2(StartRegister + Vector4fCount, NINE_MAX_CONST_F) - StartRegister;
1731 if (StartRegister < NINE_MAX_CONST_F)
1732 memcpy(&context->vs_const_f[StartRegister * 4],
1733 pConstantData,
1734 Vector4fCount * 4 * sizeof(context->vs_const_f[0]));
1735 }
1736
1737 context->changed.vs_const_f = true;
1738 context->changed.group |= NINE_STATE_VS_CONST;
1739 }
1740
CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_i,ARG_VAL (UINT,StartRegister),ARG_MEM (int,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,Vector4iCount))1741 CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_i,
1742 ARG_VAL(UINT, StartRegister),
1743 ARG_MEM(int, pConstantData),
1744 ARG_MEM_SIZE(unsigned, pConstantData_size),
1745 ARG_VAL(UINT, Vector4iCount))
1746 {
1747 struct nine_context *context = &device->context;
1748 int i;
1749
1750 if (device->driver_caps.vs_integer) {
1751 memcpy(&context->vs_const_i[4 * StartRegister],
1752 pConstantData,
1753 pConstantData_size);
1754 } else {
1755 for (i = 0; i < Vector4iCount; i++) {
1756 context->vs_const_i[4 * (StartRegister + i)] = fui((float)(pConstantData[4 * i]));
1757 context->vs_const_i[4 * (StartRegister + i) + 1] = fui((float)(pConstantData[4 * i + 1]));
1758 context->vs_const_i[4 * (StartRegister + i) + 2] = fui((float)(pConstantData[4 * i + 2]));
1759 context->vs_const_i[4 * (StartRegister + i) + 3] = fui((float)(pConstantData[4 * i + 3]));
1760 }
1761 }
1762
1763 context->changed.vs_const_i = true;
1764 context->changed.group |= NINE_STATE_VS_CONST | NINE_STATE_VS_PARAMS_MISC;
1765 }
1766
CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_b,ARG_VAL (UINT,StartRegister),ARG_MEM (BOOL,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,BoolCount))1767 CSMT_ITEM_NO_WAIT(nine_context_set_vertex_shader_constant_b,
1768 ARG_VAL(UINT, StartRegister),
1769 ARG_MEM(BOOL, pConstantData),
1770 ARG_MEM_SIZE(unsigned, pConstantData_size),
1771 ARG_VAL(UINT, BoolCount))
1772 {
1773 struct nine_context *context = &device->context;
1774 int i;
1775 uint32_t bool_true = device->driver_caps.vs_integer ? 0xFFFFFFFF : fui(1.0f);
1776
1777 (void) pConstantData_size;
1778
1779 for (i = 0; i < BoolCount; i++)
1780 context->vs_const_b[StartRegister + i] = pConstantData[i] ? bool_true : 0;
1781
1782 context->changed.vs_const_b = true;
1783 context->changed.group |= NINE_STATE_VS_CONST | NINE_STATE_VS_PARAMS_MISC;
1784 }
1785
CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader,ARG_BIND_REF (struct NinePixelShader9,ps))1786 CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader,
1787 ARG_BIND_REF(struct NinePixelShader9, ps))
1788 {
1789 struct nine_context *context = &device->context;
1790 unsigned old_mask = context->ps ? context->ps->rt_mask : 1;
1791 unsigned mask;
1792
1793 /* ff -> non-ff: commit back non-ff constants */
1794 if (!context->ps && ps)
1795 context->commit |= NINE_STATE_COMMIT_CONST_PS;
1796
1797 nine_bind(&context->ps, ps);
1798
1799 context->changed.group |= NINE_STATE_PS;
1800
1801 mask = context->ps ? context->ps->rt_mask : 1;
1802 /* We need to update cbufs if the pixel shader would
1803 * write to different render targets */
1804 if (mask != old_mask)
1805 context->changed.group |= NINE_STATE_FB;
1806 }
1807
CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_f,ARG_VAL (UINT,StartRegister),ARG_MEM (float,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,Vector4fCount))1808 CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_f,
1809 ARG_VAL(UINT, StartRegister),
1810 ARG_MEM(float, pConstantData),
1811 ARG_MEM_SIZE(unsigned, pConstantData_size),
1812 ARG_VAL(UINT, Vector4fCount))
1813 {
1814 struct nine_context *context = &device->context;
1815
1816 memcpy(&context->ps_const_f[StartRegister * 4],
1817 pConstantData,
1818 pConstantData_size);
1819
1820 context->changed.ps_const_f = true;
1821 context->changed.group |= NINE_STATE_PS_CONST;
1822 }
1823
1824 /* For stateblocks */
CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_i_transformed,ARG_VAL (UINT,StartRegister),ARG_MEM (int,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,Vector4iCount))1825 CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_i_transformed,
1826 ARG_VAL(UINT, StartRegister),
1827 ARG_MEM(int, pConstantData),
1828 ARG_MEM_SIZE(unsigned, pConstantData_size),
1829 ARG_VAL(UINT, Vector4iCount))
1830 {
1831 struct nine_context *context = &device->context;
1832
1833 memcpy(&context->ps_const_i[StartRegister][0],
1834 pConstantData,
1835 Vector4iCount * sizeof(context->ps_const_i[0]));
1836
1837 context->changed.ps_const_i = true;
1838 context->changed.group |= NINE_STATE_PS_CONST | NINE_STATE_PS_PARAMS_MISC;
1839 }
1840
CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_i,ARG_VAL (UINT,StartRegister),ARG_MEM (int,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,Vector4iCount))1841 CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_i,
1842 ARG_VAL(UINT, StartRegister),
1843 ARG_MEM(int, pConstantData),
1844 ARG_MEM_SIZE(unsigned, pConstantData_size),
1845 ARG_VAL(UINT, Vector4iCount))
1846 {
1847 struct nine_context *context = &device->context;
1848 int i;
1849
1850 if (device->driver_caps.ps_integer) {
1851 memcpy(&context->ps_const_i[StartRegister][0],
1852 pConstantData,
1853 pConstantData_size);
1854 } else {
1855 for (i = 0; i < Vector4iCount; i++) {
1856 context->ps_const_i[StartRegister+i][0] = fui((float)(pConstantData[4*i]));
1857 context->ps_const_i[StartRegister+i][1] = fui((float)(pConstantData[4*i+1]));
1858 context->ps_const_i[StartRegister+i][2] = fui((float)(pConstantData[4*i+2]));
1859 context->ps_const_i[StartRegister+i][3] = fui((float)(pConstantData[4*i+3]));
1860 }
1861 }
1862 context->changed.ps_const_i = true;
1863 context->changed.group |= NINE_STATE_PS_CONST | NINE_STATE_PS_PARAMS_MISC;
1864 }
1865
CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_b,ARG_VAL (UINT,StartRegister),ARG_MEM (BOOL,pConstantData),ARG_MEM_SIZE (unsigned,pConstantData_size),ARG_VAL (UINT,BoolCount))1866 CSMT_ITEM_NO_WAIT(nine_context_set_pixel_shader_constant_b,
1867 ARG_VAL(UINT, StartRegister),
1868 ARG_MEM(BOOL, pConstantData),
1869 ARG_MEM_SIZE(unsigned, pConstantData_size),
1870 ARG_VAL(UINT, BoolCount))
1871 {
1872 struct nine_context *context = &device->context;
1873 int i;
1874 uint32_t bool_true = device->driver_caps.ps_integer ? 0xFFFFFFFF : fui(1.0f);
1875
1876 (void) pConstantData_size;
1877
1878 for (i = 0; i < BoolCount; i++)
1879 context->ps_const_b[StartRegister + i] = pConstantData[i] ? bool_true : 0;
1880
1881 context->changed.ps_const_b = true;
1882 context->changed.group |= NINE_STATE_PS_CONST | NINE_STATE_PS_PARAMS_MISC;
1883 }
1884
1885 /* XXX: use resource, as resource might change */
CSMT_ITEM_NO_WAIT(nine_context_set_render_target,ARG_VAL (DWORD,RenderTargetIndex),ARG_BIND_REF (struct NineSurface9,rt))1886 CSMT_ITEM_NO_WAIT(nine_context_set_render_target,
1887 ARG_VAL(DWORD, RenderTargetIndex),
1888 ARG_BIND_REF(struct NineSurface9, rt))
1889 {
1890 struct nine_context *context = &device->context;
1891 const unsigned i = RenderTargetIndex;
1892
1893 if (i == 0) {
1894 context->changed.group |= NINE_STATE_MULTISAMPLE;
1895
1896 if (context->rt[0] &&
1897 (context->rt[0]->desc.MultiSampleType <= D3DMULTISAMPLE_NONMASKABLE) !=
1898 (rt->desc.MultiSampleType <= D3DMULTISAMPLE_NONMASKABLE))
1899 context->changed.group |= NINE_STATE_SAMPLE_MASK;
1900 }
1901
1902 if (context->rt[i] != rt) {
1903 nine_bind(&context->rt[i], rt);
1904 context->changed.group |= NINE_STATE_FB;
1905 }
1906 }
1907
1908 /* XXX: use resource instead of ds, as resource might change */
CSMT_ITEM_NO_WAIT(nine_context_set_depth_stencil,ARG_BIND_REF (struct NineSurface9,ds))1909 CSMT_ITEM_NO_WAIT(nine_context_set_depth_stencil,
1910 ARG_BIND_REF(struct NineSurface9, ds))
1911 {
1912 struct nine_context *context = &device->context;
1913
1914 nine_bind(&context->ds, ds);
1915 context->changed.group |= NINE_STATE_FB;
1916 }
1917
CSMT_ITEM_NO_WAIT(nine_context_set_viewport,ARG_COPY_REF (D3DVIEWPORT9,viewport))1918 CSMT_ITEM_NO_WAIT(nine_context_set_viewport,
1919 ARG_COPY_REF(D3DVIEWPORT9, viewport))
1920 {
1921 struct nine_context *context = &device->context;
1922
1923 if (!memcmp(viewport, &context->viewport, sizeof(context->viewport)))
1924 return;
1925
1926 context->viewport = *viewport;
1927 context->changed.group |= NINE_STATE_VIEWPORT;
1928 }
1929
CSMT_ITEM_NO_WAIT(nine_context_set_scissor,ARG_COPY_REF (struct pipe_scissor_state,scissor))1930 CSMT_ITEM_NO_WAIT(nine_context_set_scissor,
1931 ARG_COPY_REF(struct pipe_scissor_state, scissor))
1932 {
1933 struct nine_context *context = &device->context;
1934
1935 if (!memcmp(scissor, &context->scissor, sizeof(context->scissor)))
1936 return;
1937
1938 context->scissor = *scissor;
1939 context->changed.group |= NINE_STATE_SCISSOR;
1940 }
1941
CSMT_ITEM_NO_WAIT(nine_context_set_transform,ARG_VAL (D3DTRANSFORMSTATETYPE,State),ARG_COPY_REF (D3DMATRIX,pMatrix))1942 CSMT_ITEM_NO_WAIT(nine_context_set_transform,
1943 ARG_VAL(D3DTRANSFORMSTATETYPE, State),
1944 ARG_COPY_REF(D3DMATRIX, pMatrix))
1945 {
1946 struct nine_context *context = &device->context;
1947 D3DMATRIX *M = nine_state_access_transform(&context->ff, State, true);
1948
1949 *M = *pMatrix;
1950 if (State == D3DTS_PROJECTION) {
1951 BOOL prev_zfog = context->zfog;
1952 /* Pixel fog (with WFOG advertised): source is either Z or W.
1953 * W is the source if the projection matrix is not orthogonal.
1954 * Tests on Win 10 seem to indicate _34
1955 * and _33 are checked against 0, 1. */
1956 context->zfog = (M->_34 == 0.0f &&
1957 M->_44 == 1.0f);
1958 if (context->zfog != prev_zfog)
1959 context->changed.group |= NINE_STATE_PS_PARAMS_MISC;
1960 }
1961 context->ff.changed.transform[State / 32] |= 1 << (State % 32);
1962 context->changed.group |= NINE_STATE_FF;
1963 }
1964
CSMT_ITEM_NO_WAIT(nine_context_set_material,ARG_COPY_REF (D3DMATERIAL9,pMaterial))1965 CSMT_ITEM_NO_WAIT(nine_context_set_material,
1966 ARG_COPY_REF(D3DMATERIAL9, pMaterial))
1967 {
1968 struct nine_context *context = &device->context;
1969
1970 context->ff.material = *pMaterial;
1971 context->changed.group |= NINE_STATE_FF_MATERIAL;
1972 }
1973
CSMT_ITEM_NO_WAIT(nine_context_set_light,ARG_VAL (DWORD,Index),ARG_COPY_REF (D3DLIGHT9,pLight))1974 CSMT_ITEM_NO_WAIT(nine_context_set_light,
1975 ARG_VAL(DWORD, Index),
1976 ARG_COPY_REF(D3DLIGHT9, pLight))
1977 {
1978 struct nine_context *context = &device->context;
1979
1980 (void)nine_state_set_light(&context->ff, Index, pLight);
1981 context->changed.group |= NINE_STATE_FF_LIGHTING;
1982 }
1983
1984
1985 /* For stateblocks */
1986 static void
nine_context_light_enable_stateblock(struct NineDevice9 * device,const uint16_t active_light[NINE_MAX_LIGHTS_ACTIVE],unsigned int num_lights_active)1987 nine_context_light_enable_stateblock(struct NineDevice9 *device,
1988 const uint16_t active_light[NINE_MAX_LIGHTS_ACTIVE], /* TODO: use pointer that convey size for csmt */
1989 unsigned int num_lights_active)
1990 {
1991 struct nine_context *context = &device->context;
1992
1993 /* TODO: Use CSMT_* to avoid calling nine_csmt_process */
1994 nine_csmt_process(device);
1995 memcpy(context->ff.active_light, active_light, NINE_MAX_LIGHTS_ACTIVE * sizeof(context->ff.active_light[0]));
1996 context->ff.num_lights_active = num_lights_active;
1997 context->changed.group |= NINE_STATE_FF_LIGHTING;
1998 }
1999
CSMT_ITEM_NO_WAIT(nine_context_light_enable,ARG_VAL (DWORD,Index),ARG_VAL (BOOL,Enable))2000 CSMT_ITEM_NO_WAIT(nine_context_light_enable,
2001 ARG_VAL(DWORD, Index),
2002 ARG_VAL(BOOL, Enable))
2003 {
2004 struct nine_context *context = &device->context;
2005
2006 nine_state_light_enable(&context->ff, Index, Enable);
2007 context->changed.group |= NINE_STATE_FF_LIGHTING;
2008 }
2009
CSMT_ITEM_NO_WAIT(nine_context_set_texture_stage_state,ARG_VAL (DWORD,Stage),ARG_VAL (D3DTEXTURESTAGESTATETYPE,Type),ARG_VAL (DWORD,Value))2010 CSMT_ITEM_NO_WAIT(nine_context_set_texture_stage_state,
2011 ARG_VAL(DWORD, Stage),
2012 ARG_VAL(D3DTEXTURESTAGESTATETYPE, Type),
2013 ARG_VAL(DWORD, Value))
2014 {
2015 struct nine_context *context = &device->context;
2016 int bumpmap_index = -1;
2017
2018 context->ff.tex_stage[Stage][Type] = Value;
2019 switch (Type) {
2020 case D3DTSS_BUMPENVMAT00:
2021 bumpmap_index = 4 * Stage;
2022 break;
2023 case D3DTSS_BUMPENVMAT01:
2024 bumpmap_index = 4 * Stage + 1;
2025 break;
2026 case D3DTSS_BUMPENVMAT10:
2027 bumpmap_index = 4 * Stage + 2;
2028 break;
2029 case D3DTSS_BUMPENVMAT11:
2030 bumpmap_index = 4 * Stage + 3;
2031 break;
2032 case D3DTSS_BUMPENVLSCALE:
2033 bumpmap_index = 4 * 8 + 2 * Stage;
2034 break;
2035 case D3DTSS_BUMPENVLOFFSET:
2036 bumpmap_index = 4 * 8 + 2 * Stage + 1;
2037 break;
2038 case D3DTSS_TEXTURETRANSFORMFLAGS:
2039 context->changed.group |= NINE_STATE_PS_PARAMS_MISC;
2040 break;
2041 default:
2042 break;
2043 }
2044
2045 if (bumpmap_index >= 0) {
2046 context->bumpmap_vars[bumpmap_index] = Value;
2047 context->changed.group |= NINE_STATE_PS_CONST;
2048 }
2049
2050 context->changed.group |= NINE_STATE_FF_PS_CONSTS;
2051 context->ff.changed.tex_stage[Stage][Type / 32] |= 1 << (Type % 32);
2052 }
2053
CSMT_ITEM_NO_WAIT(nine_context_set_clip_plane,ARG_VAL (DWORD,Index),ARG_COPY_REF (struct nine_clipplane,pPlane))2054 CSMT_ITEM_NO_WAIT(nine_context_set_clip_plane,
2055 ARG_VAL(DWORD, Index),
2056 ARG_COPY_REF(struct nine_clipplane, pPlane))
2057 {
2058 struct nine_context *context = &device->context;
2059
2060 memcpy(&context->clip.ucp[Index][0], pPlane, sizeof(context->clip.ucp[0]));
2061 if (!device->driver_caps.emulate_ucp)
2062 context->changed.ucp = true;
2063 else
2064 context->changed.group |= NINE_STATE_FF_VS_OTHER | NINE_STATE_VS_CONST;
2065 }
2066
CSMT_ITEM_NO_WAIT(nine_context_set_swvp,ARG_VAL (bool,swvp))2067 CSMT_ITEM_NO_WAIT(nine_context_set_swvp,
2068 ARG_VAL(bool, swvp))
2069 {
2070 struct nine_context *context = &device->context;
2071
2072 context->swvp = swvp;
2073 context->changed.group |= NINE_STATE_SWVP;
2074 }
2075
2076 /* Do not write to nine_context directly. Slower,
2077 * but works with csmt. TODO: write a special csmt version that
2078 * would record the list of commands as much as possible,
2079 * and use the version above else.
2080 */
2081 void
nine_context_apply_stateblock(struct NineDevice9 * device,const struct nine_state * src)2082 nine_context_apply_stateblock(struct NineDevice9 *device,
2083 const struct nine_state *src)
2084 {
2085 int i;
2086
2087 /* No need to apply src->changed.group, since all calls do
2088 * set context->changed.group */
2089
2090 for (i = 0; i < ARRAY_SIZE(src->changed.rs); ++i) {
2091 uint32_t m = src->changed.rs[i];
2092 while (m) {
2093 const int r = ffs(m) - 1;
2094 m &= ~(1 << r);
2095 nine_context_set_render_state(device, i * 32 + r, src->rs_advertised[i * 32 + r]);
2096 }
2097 }
2098
2099 /* Textures */
2100 if (src->changed.texture) {
2101 uint32_t m = src->changed.texture;
2102 unsigned s;
2103
2104 for (s = 0; m; ++s, m >>= 1) {
2105 struct NineBaseTexture9 *tex = src->texture[s];
2106 if (!(m & 1))
2107 continue;
2108 nine_context_set_texture(device, s, tex);
2109 }
2110 }
2111
2112 /* Sampler state */
2113 if (src->changed.group & NINE_STATE_SAMPLER) {
2114 unsigned s;
2115
2116 for (s = 0; s < NINE_MAX_SAMPLERS; ++s) {
2117 uint32_t m = src->changed.sampler[s];
2118 while (m) {
2119 const int i = ffs(m) - 1;
2120 m &= ~(1 << i);
2121 nine_context_set_sampler_state(device, s, i, src->samp_advertised[s][i]);
2122 }
2123 }
2124 }
2125
2126 /* Vertex buffers */
2127 if (src->changed.vtxbuf | src->changed.stream_freq) {
2128 uint32_t m = src->changed.vtxbuf | src->changed.stream_freq;
2129 for (i = 0; m; ++i, m >>= 1) {
2130 if (src->changed.vtxbuf & (1 << i))
2131 nine_context_set_stream_source(device, i, src->stream[i], src->vtxbuf[i].buffer_offset, src->vtxstride[i]);
2132 if (src->changed.stream_freq & (1 << i))
2133 nine_context_set_stream_source_freq(device, i, src->stream_freq[i]);
2134 }
2135 }
2136
2137 /* Index buffer */
2138 if (src->changed.group & NINE_STATE_IDXBUF)
2139 nine_context_set_indices(device, src->idxbuf);
2140
2141 /* Vertex declaration */
2142 if ((src->changed.group & NINE_STATE_VDECL) && src->vdecl)
2143 nine_context_set_vertex_declaration(device, src->vdecl);
2144
2145 /* Vertex shader */
2146 if (src->changed.group & NINE_STATE_VS)
2147 nine_context_set_vertex_shader(device, src->vs);
2148
2149 /* Pixel shader */
2150 if (src->changed.group & NINE_STATE_PS)
2151 nine_context_set_pixel_shader(device, src->ps);
2152
2153 /* Vertex constants */
2154 if (src->changed.group & NINE_STATE_VS_CONST) {
2155 struct nine_range *r;
2156 for (r = src->changed.vs_const_f; r; r = r->next)
2157 nine_context_set_vertex_shader_constant_f(device, r->bgn,
2158 &src->vs_const_f[r->bgn * 4],
2159 sizeof(float[4]) * (r->end - r->bgn),
2160 r->end - r->bgn);
2161 for (r = src->changed.vs_const_i; r; r = r->next)
2162 nine_context_set_vertex_shader_constant_i(device, r->bgn,
2163 &src->vs_const_i[r->bgn * 4],
2164 sizeof(int[4]) * (r->end - r->bgn),
2165 r->end - r->bgn);
2166 for (r = src->changed.vs_const_b; r; r = r->next)
2167 nine_context_set_vertex_shader_constant_b(device, r->bgn,
2168 &src->vs_const_b[r->bgn * 4],
2169 sizeof(BOOL) * (r->end - r->bgn),
2170 r->end - r->bgn);
2171 }
2172
2173 /* Pixel constants */
2174 if (src->changed.group & NINE_STATE_PS_CONST) {
2175 struct nine_range *r;
2176 for (r = src->changed.ps_const_f; r; r = r->next)
2177 nine_context_set_pixel_shader_constant_f(device, r->bgn,
2178 &src->ps_const_f[r->bgn * 4],
2179 sizeof(float[4]) * (r->end - r->bgn),
2180 r->end - r->bgn);
2181 if (src->changed.ps_const_i) {
2182 uint16_t m = src->changed.ps_const_i;
2183 for (i = ffs(m) - 1, m >>= i; m; ++i, m >>= 1)
2184 if (m & 1)
2185 nine_context_set_pixel_shader_constant_i_transformed(device, i,
2186 src->ps_const_i[i], sizeof(int[4]), 1);
2187 }
2188 if (src->changed.ps_const_b) {
2189 uint16_t m = src->changed.ps_const_b;
2190 for (i = ffs(m) - 1, m >>= i; m; ++i, m >>= 1)
2191 if (m & 1)
2192 nine_context_set_pixel_shader_constant_b(device, i,
2193 &src->ps_const_b[i], sizeof(BOOL), 1);
2194 }
2195 }
2196
2197 /* Viewport */
2198 if (src->changed.group & NINE_STATE_VIEWPORT)
2199 nine_context_set_viewport(device, &src->viewport);
2200
2201 /* Scissor */
2202 if (src->changed.group & NINE_STATE_SCISSOR)
2203 nine_context_set_scissor(device, &src->scissor);
2204
2205 /* User Clip Planes */
2206 if (src->changed.ucp)
2207 for (i = 0; i < PIPE_MAX_CLIP_PLANES; ++i)
2208 if (src->changed.ucp & (1 << i))
2209 nine_context_set_clip_plane(device, i, (struct nine_clipplane*)&src->clip.ucp[i][0]);
2210
2211 if (!(src->changed.group & NINE_STATE_FF))
2212 return;
2213
2214 /* Fixed function state. */
2215
2216 if (src->changed.group & NINE_STATE_FF_MATERIAL)
2217 nine_context_set_material(device, &src->ff.material);
2218
2219 if (src->changed.group & NINE_STATE_FF_PS_CONSTS) {
2220 unsigned s;
2221 for (s = 0; s < NINE_MAX_TEXTURE_STAGES; ++s) {
2222 for (i = 0; i < NINED3DTSS_COUNT; ++i)
2223 if (src->ff.changed.tex_stage[s][i / 32] & (1 << (i % 32)))
2224 nine_context_set_texture_stage_state(device, s, i, src->ff.tex_stage[s][i]);
2225 }
2226 }
2227 if (src->changed.group & NINE_STATE_FF_LIGHTING) {
2228 for (i = 0; i < src->ff.num_lights; ++i)
2229 if (src->ff.light[i].Type != NINED3DLIGHT_INVALID)
2230 nine_context_set_light(device, i, &src->ff.light[i]);
2231
2232 nine_context_light_enable_stateblock(device, src->ff.active_light, src->ff.num_lights_active);
2233 }
2234 if (src->changed.group & NINE_STATE_FF_VSTRANSF) {
2235 for (i = 0; i < ARRAY_SIZE(src->ff.changed.transform); ++i) {
2236 unsigned s;
2237 if (!src->ff.changed.transform[i])
2238 continue;
2239 for (s = i * 32; s < (i * 32 + 32); ++s) {
2240 if (!(src->ff.changed.transform[i] & (1 << (s % 32))))
2241 continue;
2242 /* MaxVertexBlendMatrixIndex is 8, which means
2243 * we don't read past index D3DTS_WORLDMATRIX(8).
2244 * swvp is supposed to allow all 256, but we don't
2245 * implement it for now. */
2246 if (s > D3DTS_WORLDMATRIX(8))
2247 break;
2248 nine_context_set_transform(device, s,
2249 nine_state_access_transform(
2250 (struct nine_ff_state *)&src->ff,
2251 s, false));
2252 }
2253 }
2254 }
2255 }
2256
2257 static void
nine_update_state_framebuffer_clear(struct NineDevice9 * device)2258 nine_update_state_framebuffer_clear(struct NineDevice9 *device)
2259 {
2260 struct nine_context *context = &device->context;
2261
2262 if (context->changed.group & NINE_STATE_FB)
2263 update_framebuffer(device, true);
2264 }
2265
CSMT_ITEM_NO_WAIT(nine_context_clear_fb,ARG_VAL (DWORD,Count),ARG_COPY_REF (D3DRECT,pRects),ARG_VAL (DWORD,Flags),ARG_VAL (D3DCOLOR,Color),ARG_VAL (float,Z),ARG_VAL (DWORD,Stencil))2266 CSMT_ITEM_NO_WAIT(nine_context_clear_fb,
2267 ARG_VAL(DWORD, Count),
2268 ARG_COPY_REF(D3DRECT, pRects),
2269 ARG_VAL(DWORD, Flags),
2270 ARG_VAL(D3DCOLOR, Color),
2271 ARG_VAL(float, Z),
2272 ARG_VAL(DWORD, Stencil))
2273 {
2274 struct nine_context *context = &device->context;
2275 const int sRGB = context->rs[D3DRS_SRGBWRITEENABLE] ? 1 : 0;
2276 struct pipe_surface *cbuf, *zsbuf;
2277 struct pipe_context *pipe = context->pipe;
2278 struct NineSurface9 *zsbuf_surf = context->ds;
2279 struct NineSurface9 *rt;
2280 unsigned bufs = 0;
2281 unsigned r, i;
2282 union pipe_color_union rgba;
2283 unsigned rt_mask = 0;
2284 D3DRECT rect;
2285
2286 nine_update_state_framebuffer_clear(device);
2287
2288 if (Flags & D3DCLEAR_TARGET) bufs |= PIPE_CLEAR_COLOR;
2289 /* Ignore Z buffer if not bound */
2290 if (context->pipe_data.fb.zsbuf != NULL) {
2291 if (Flags & D3DCLEAR_ZBUFFER) bufs |= PIPE_CLEAR_DEPTH;
2292 if (Flags & D3DCLEAR_STENCIL) bufs |= PIPE_CLEAR_STENCIL;
2293 }
2294 if (!bufs)
2295 return;
2296 d3dcolor_to_pipe_color_union(&rgba, Color);
2297
2298 rect.x1 = context->viewport.X;
2299 rect.y1 = context->viewport.Y;
2300 rect.x2 = context->viewport.Width + rect.x1;
2301 rect.y2 = context->viewport.Height + rect.y1;
2302
2303 /* Both rectangles apply, which is weird, but that's D3D9. */
2304 if (context->rs[D3DRS_SCISSORTESTENABLE]) {
2305 rect.x1 = MAX2(rect.x1, context->scissor.minx);
2306 rect.y1 = MAX2(rect.y1, context->scissor.miny);
2307 rect.x2 = MIN2(rect.x2, context->scissor.maxx);
2308 rect.y2 = MIN2(rect.y2, context->scissor.maxy);
2309 }
2310
2311 if (Count) {
2312 /* Maybe apps like to specify a large rect ? */
2313 if (pRects[0].x1 <= rect.x1 && pRects[0].x2 >= rect.x2 &&
2314 pRects[0].y1 <= rect.y1 && pRects[0].y2 >= rect.y2) {
2315 DBG("First rect covers viewport.\n");
2316 Count = 0;
2317 pRects = NULL;
2318 }
2319 }
2320
2321 if (rect.x1 >= context->pipe_data.fb.width || rect.y1 >= context->pipe_data.fb.height)
2322 return;
2323
2324 for (i = 0; i < device->caps.NumSimultaneousRTs; ++i) {
2325 if (context->rt[i] && context->rt[i]->desc.Format != D3DFMT_NULL)
2326 rt_mask |= 1 << i;
2327 }
2328
2329 /* fast path, clears everything at once */
2330 if (!Count &&
2331 (!(bufs & PIPE_CLEAR_COLOR) || (rt_mask == context->rt_mask)) &&
2332 rect.x1 == 0 && rect.y1 == 0 &&
2333 /* Case we clear only render target. Check clear region vs rt. */
2334 ((!(bufs & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL)) &&
2335 rect.x2 >= context->pipe_data.fb.width &&
2336 rect.y2 >= context->pipe_data.fb.height) ||
2337 /* Case we clear depth buffer (and eventually rt too).
2338 * depth buffer size is always >= rt size. Compare to clear region */
2339 ((bufs & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL)) &&
2340 rect.x2 >= zsbuf_surf->desc.Width &&
2341 rect.y2 >= zsbuf_surf->desc.Height))) {
2342 DBG("Clear fast path\n");
2343 pipe->clear(pipe, bufs, NULL, &rgba, Z, Stencil);
2344 return;
2345 }
2346
2347 if (!Count) {
2348 Count = 1;
2349 pRects = ▭
2350 }
2351
2352 for (i = 0; i < device->caps.NumSimultaneousRTs; ++i) {
2353 rt = context->rt[i];
2354 if (!rt || rt->desc.Format == D3DFMT_NULL ||
2355 !(bufs & PIPE_CLEAR_COLOR))
2356 continue; /* save space, compiler should hoist this */
2357 cbuf = NineSurface9_GetSurface(rt, sRGB);
2358 for (r = 0; r < Count; ++r) {
2359 /* Don't trust users to pass these in the right order. */
2360 unsigned x1 = MIN2(pRects[r].x1, pRects[r].x2);
2361 unsigned y1 = MIN2(pRects[r].y1, pRects[r].y2);
2362 unsigned x2 = MAX2(pRects[r].x1, pRects[r].x2);
2363 unsigned y2 = MAX2(pRects[r].y1, pRects[r].y2);
2364 #ifndef NINE_LAX
2365 /* Drop negative rectangles (like wine expects). */
2366 if (pRects[r].x1 > pRects[r].x2) continue;
2367 if (pRects[r].y1 > pRects[r].y2) continue;
2368 #endif
2369
2370 x1 = MAX2(x1, rect.x1);
2371 y1 = MAX2(y1, rect.y1);
2372 x2 = MIN3(x2, rect.x2, rt->desc.Width);
2373 y2 = MIN3(y2, rect.y2, rt->desc.Height);
2374
2375 DBG("Clearing (%u..%u)x(%u..%u)\n", x1, x2, y1, y2);
2376 pipe->clear_render_target(pipe, cbuf, &rgba,
2377 x1, y1, x2 - x1, y2 - y1, false);
2378 }
2379 }
2380 if (!(bufs & PIPE_CLEAR_DEPTHSTENCIL))
2381 return;
2382
2383 bufs &= PIPE_CLEAR_DEPTHSTENCIL;
2384
2385 for (r = 0; r < Count; ++r) {
2386 unsigned x1 = MIN2(pRects[r].x1, pRects[r].x2);
2387 unsigned y1 = MIN2(pRects[r].y1, pRects[r].y2);
2388 unsigned x2 = MAX2(pRects[r].x1, pRects[r].x2);
2389 unsigned y2 = MAX2(pRects[r].y1, pRects[r].y2);
2390 #ifndef NINE_LAX
2391 /* Drop negative rectangles. */
2392 if (pRects[r].x1 > pRects[r].x2) continue;
2393 if (pRects[r].y1 > pRects[r].y2) continue;
2394 #endif
2395
2396 x1 = MIN2(x1, rect.x1);
2397 y1 = MIN2(y1, rect.y1);
2398 x2 = MIN3(x2, rect.x2, zsbuf_surf->desc.Width);
2399 y2 = MIN3(y2, rect.y2, zsbuf_surf->desc.Height);
2400
2401 zsbuf = NineSurface9_GetSurface(zsbuf_surf, 0);
2402 assert(zsbuf);
2403 pipe->clear_depth_stencil(pipe, zsbuf, bufs, Z, Stencil,
2404 x1, y1, x2 - x1, y2 - y1, false);
2405 }
2406 return;
2407 }
2408
2409
2410 static inline void
init_draw_info(struct pipe_draw_info * info,struct pipe_draw_start_count_bias * draw,struct NineDevice9 * dev,D3DPRIMITIVETYPE type,UINT count)2411 init_draw_info(struct pipe_draw_info *info,
2412 struct pipe_draw_start_count_bias *draw,
2413 struct NineDevice9 *dev, D3DPRIMITIVETYPE type, UINT count)
2414 {
2415 info->mode = d3dprimitivetype_to_pipe_prim(type);
2416 draw->count = prim_count_to_vertex_count(type, count);
2417 info->start_instance = 0;
2418 info->instance_count = 1;
2419 if (dev->context.stream_instancedata_mask & dev->context.stream_usage_mask)
2420 info->instance_count = MAX2(dev->context.stream_freq[0] & 0x7FFFFF, 1);
2421 info->primitive_restart = false;
2422 info->has_user_indices = false;
2423 info->take_index_buffer_ownership = false;
2424 info->index_bias_varies = false;
2425 info->increment_draw_id = false;
2426 info->was_line_loop = false;
2427 info->restart_index = 0;
2428 info->view_mask = 0;
2429 }
2430
CSMT_ITEM_NO_WAIT(nine_context_draw_primitive,ARG_VAL (D3DPRIMITIVETYPE,PrimitiveType),ARG_VAL (UINT,StartVertex),ARG_VAL (UINT,PrimitiveCount))2431 CSMT_ITEM_NO_WAIT(nine_context_draw_primitive,
2432 ARG_VAL(D3DPRIMITIVETYPE, PrimitiveType),
2433 ARG_VAL(UINT, StartVertex),
2434 ARG_VAL(UINT, PrimitiveCount))
2435 {
2436 struct nine_context *context = &device->context;
2437 struct pipe_draw_info info;
2438 struct pipe_draw_start_count_bias draw;
2439
2440 if (context->vs && context->vs->swvp_only && !context->swvp)
2441 return;
2442
2443 nine_update_state(device);
2444
2445 init_draw_info(&info, &draw, device, PrimitiveType, PrimitiveCount);
2446 info.index_size = 0;
2447 draw.start = StartVertex;
2448 draw.index_bias = 0;
2449 info.min_index = draw.start;
2450 info.max_index = draw.start + draw.count - 1;
2451 info.index.resource = NULL;
2452
2453 context->pipe->draw_vbo(context->pipe, &info, 0, NULL, &draw, 1);
2454 }
2455
CSMT_ITEM_NO_WAIT(nine_context_draw_indexed_primitive,ARG_VAL (D3DPRIMITIVETYPE,PrimitiveType),ARG_VAL (INT,BaseVertexIndex),ARG_VAL (UINT,MinVertexIndex),ARG_VAL (UINT,NumVertices),ARG_VAL (UINT,StartIndex),ARG_VAL (UINT,PrimitiveCount))2456 CSMT_ITEM_NO_WAIT(nine_context_draw_indexed_primitive,
2457 ARG_VAL(D3DPRIMITIVETYPE, PrimitiveType),
2458 ARG_VAL(INT, BaseVertexIndex),
2459 ARG_VAL(UINT, MinVertexIndex),
2460 ARG_VAL(UINT, NumVertices),
2461 ARG_VAL(UINT, StartIndex),
2462 ARG_VAL(UINT, PrimitiveCount))
2463 {
2464 struct nine_context *context = &device->context;
2465 struct pipe_draw_info info;
2466 struct pipe_draw_start_count_bias draw;
2467
2468 if (context->vs && context->vs->swvp_only && !context->swvp)
2469 return;
2470
2471 nine_update_state(device);
2472
2473 init_draw_info(&info, &draw, device, PrimitiveType, PrimitiveCount);
2474 info.index_size = context->index_size;
2475 draw.start = context->index_offset / context->index_size + StartIndex;
2476 draw.index_bias = BaseVertexIndex;
2477 info.index_bounds_valid = true;
2478 /* These don't include index bias: */
2479 info.min_index = MinVertexIndex;
2480 info.max_index = MinVertexIndex + NumVertices - 1;
2481 info.index.resource = context->idxbuf;
2482
2483 context->pipe->draw_vbo(context->pipe, &info, 0, NULL, &draw, 1);
2484 }
2485
CSMT_ITEM_NO_WAIT(nine_context_draw_indexed_primitive_from_vtxbuf_idxbuf,ARG_VAL (D3DPRIMITIVETYPE,PrimitiveType),ARG_VAL (UINT,MinVertexIndex),ARG_VAL (UINT,NumVertices),ARG_VAL (UINT,PrimitiveCount),ARG_VAL (UINT,vbuf_stride),ARG_BIND_VBUF (struct pipe_vertex_buffer,vbuf),ARG_BIND_RES (struct pipe_resource,ibuf),ARG_VAL (void *,user_ibuf),ARG_VAL (UINT,index_offset),ARG_VAL (UINT,index_size))2486 CSMT_ITEM_NO_WAIT(nine_context_draw_indexed_primitive_from_vtxbuf_idxbuf,
2487 ARG_VAL(D3DPRIMITIVETYPE, PrimitiveType),
2488 ARG_VAL(UINT, MinVertexIndex),
2489 ARG_VAL(UINT, NumVertices),
2490 ARG_VAL(UINT, PrimitiveCount),
2491 ARG_VAL(UINT, vbuf_stride),
2492 ARG_BIND_VBUF(struct pipe_vertex_buffer, vbuf),
2493 ARG_BIND_RES(struct pipe_resource, ibuf),
2494 ARG_VAL(void *, user_ibuf),
2495 ARG_VAL(UINT, index_offset),
2496 ARG_VAL(UINT, index_size))
2497 {
2498 struct nine_context *context = &device->context;
2499 struct pipe_draw_info info;
2500 struct pipe_draw_start_count_bias draw;
2501
2502 if (context->vs && context->vs->swvp_only && !context->swvp)
2503 return;
2504
2505 if (context->vtxstride[0] != vbuf_stride) {
2506 context->vtxstride[0] = vbuf_stride;
2507 /* force elements update for stride.
2508 * We don't need to restore the old value,
2509 * as the caller set its to 0 after the call */
2510 context->changed.group |= NINE_STATE_VDECL;
2511 }
2512
2513 nine_update_state(device);
2514
2515 init_draw_info(&info, &draw, device, PrimitiveType, PrimitiveCount);
2516 info.index_size = index_size;
2517 draw.start = index_offset / info.index_size;
2518 draw.index_bias = 0;
2519 info.index_bounds_valid = true;
2520 info.min_index = MinVertexIndex;
2521 info.max_index = MinVertexIndex + NumVertices - 1;
2522 info.has_user_indices = ibuf == NULL;
2523 if (ibuf)
2524 info.index.resource = ibuf;
2525 else
2526 info.index.user = user_ibuf;
2527
2528 util_set_vertex_buffers(context->pipe, 1, false, vbuf);
2529 context->changed.vtxbuf |= 1;
2530
2531 context->pipe->draw_vbo(context->pipe, &info, 0, NULL, &draw, 1);
2532 }
2533
CSMT_ITEM_NO_WAIT(nine_context_resource_copy_region,ARG_BIND_REF (struct NineUnknown,dst),ARG_BIND_REF (struct NineUnknown,src),ARG_BIND_RES (struct pipe_resource,dst_res),ARG_VAL (unsigned,dst_level),ARG_COPY_REF (struct pipe_box,dst_box),ARG_BIND_RES (struct pipe_resource,src_res),ARG_VAL (unsigned,src_level),ARG_COPY_REF (struct pipe_box,src_box))2534 CSMT_ITEM_NO_WAIT(nine_context_resource_copy_region,
2535 ARG_BIND_REF(struct NineUnknown, dst),
2536 ARG_BIND_REF(struct NineUnknown, src),
2537 ARG_BIND_RES(struct pipe_resource, dst_res),
2538 ARG_VAL(unsigned, dst_level),
2539 ARG_COPY_REF(struct pipe_box, dst_box),
2540 ARG_BIND_RES(struct pipe_resource, src_res),
2541 ARG_VAL(unsigned, src_level),
2542 ARG_COPY_REF(struct pipe_box, src_box))
2543 {
2544 struct nine_context *context = &device->context;
2545
2546 (void) dst;
2547 (void) src;
2548
2549 context->pipe->resource_copy_region(context->pipe,
2550 dst_res, dst_level,
2551 dst_box->x, dst_box->y, dst_box->z,
2552 src_res, src_level,
2553 src_box);
2554 }
2555
CSMT_ITEM_NO_WAIT(nine_context_blit,ARG_BIND_REF (struct NineUnknown,dst),ARG_BIND_REF (struct NineUnknown,src),ARG_BIND_BLIT (struct pipe_blit_info,blit))2556 CSMT_ITEM_NO_WAIT(nine_context_blit,
2557 ARG_BIND_REF(struct NineUnknown, dst),
2558 ARG_BIND_REF(struct NineUnknown, src),
2559 ARG_BIND_BLIT(struct pipe_blit_info, blit))
2560 {
2561 struct nine_context *context = &device->context;
2562
2563 (void) dst;
2564 (void) src;
2565
2566 context->pipe->blit(context->pipe, blit);
2567 }
2568
CSMT_ITEM_NO_WAIT(nine_context_clear_render_target,ARG_BIND_REF (struct NineSurface9,surface),ARG_VAL (D3DCOLOR,color),ARG_VAL (UINT,x),ARG_VAL (UINT,y),ARG_VAL (UINT,width),ARG_VAL (UINT,height))2569 CSMT_ITEM_NO_WAIT(nine_context_clear_render_target,
2570 ARG_BIND_REF(struct NineSurface9, surface),
2571 ARG_VAL(D3DCOLOR, color),
2572 ARG_VAL(UINT, x),
2573 ARG_VAL(UINT, y),
2574 ARG_VAL(UINT, width),
2575 ARG_VAL(UINT, height))
2576 {
2577 struct nine_context *context = &device->context;
2578 struct pipe_surface *surf;
2579 union pipe_color_union rgba;
2580
2581 d3dcolor_to_pipe_color_union(&rgba, color);
2582 surf = NineSurface9_GetSurface(surface, 0);
2583 context->pipe->clear_render_target(context->pipe, surf, &rgba, x, y, width, height, false);
2584 }
2585
CSMT_ITEM_NO_WAIT(nine_context_gen_mipmap,ARG_BIND_REF (struct NineUnknown,dst),ARG_BIND_RES (struct pipe_resource,res),ARG_VAL (UINT,base_level),ARG_VAL (UINT,last_level),ARG_VAL (UINT,first_layer),ARG_VAL (UINT,last_layer),ARG_VAL (UINT,filter))2586 CSMT_ITEM_NO_WAIT(nine_context_gen_mipmap,
2587 ARG_BIND_REF(struct NineUnknown, dst),
2588 ARG_BIND_RES(struct pipe_resource, res),
2589 ARG_VAL(UINT, base_level),
2590 ARG_VAL(UINT, last_level),
2591 ARG_VAL(UINT, first_layer),
2592 ARG_VAL(UINT, last_layer),
2593 ARG_VAL(UINT, filter))
2594 {
2595 struct nine_context *context = &device->context;
2596
2597 /* We just bind dst for the bind count */
2598 (void)dst;
2599
2600 util_gen_mipmap(context->pipe, res, res->format, base_level,
2601 last_level, first_layer, last_layer, filter);
2602 }
2603
CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_range_upload,ARG_BIND_REF (struct NineUnknown,src_ref),ARG_BIND_RES (struct pipe_resource,res),ARG_VAL (unsigned,offset),ARG_VAL (unsigned,size),ARG_VAL (unsigned,usage),ARG_VAL (const void *,data))2604 CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_range_upload,
2605 ARG_BIND_REF(struct NineUnknown, src_ref),
2606 ARG_BIND_RES(struct pipe_resource, res),
2607 ARG_VAL(unsigned, offset),
2608 ARG_VAL(unsigned, size),
2609 ARG_VAL(unsigned, usage),
2610 ARG_VAL(const void *, data))
2611 {
2612 struct nine_context *context = &device->context;
2613
2614 /* Binding src_ref avoids release before upload */
2615 (void)src_ref;
2616
2617 context->pipe->buffer_subdata(context->pipe, res, usage, offset, size, data);
2618 }
2619
CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_box_upload,ARG_BIND_REF (struct NineUnknown,src_ref),ARG_BIND_RES (struct pipe_resource,res),ARG_VAL (unsigned,level),ARG_COPY_REF (struct pipe_box,dst_box),ARG_VAL (enum pipe_format,src_format),ARG_VAL (const void *,src),ARG_VAL (unsigned,src_stride),ARG_VAL (unsigned,src_layer_stride),ARG_COPY_REF (struct pipe_box,src_box))2620 CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_box_upload,
2621 ARG_BIND_REF(struct NineUnknown, src_ref),
2622 ARG_BIND_RES(struct pipe_resource, res),
2623 ARG_VAL(unsigned, level),
2624 ARG_COPY_REF(struct pipe_box, dst_box),
2625 ARG_VAL(enum pipe_format, src_format),
2626 ARG_VAL(const void *, src),
2627 ARG_VAL(unsigned, src_stride),
2628 ARG_VAL(unsigned, src_layer_stride),
2629 ARG_COPY_REF(struct pipe_box, src_box))
2630 {
2631 struct nine_context *context = &device->context;
2632 struct pipe_context *pipe = context->pipe;
2633 struct pipe_transfer *transfer = NULL;
2634 uint8_t *map;
2635
2636 /* Binding src_ref avoids release before upload */
2637 (void)src_ref;
2638
2639 if (is_ATI1_ATI2(src_format)) {
2640 const unsigned bw = util_format_get_blockwidth(src_format);
2641 const unsigned bh = util_format_get_blockheight(src_format);
2642 /* For these formats, the allocate surface can be too small to contain
2643 * a block. Yet we can be asked to upload such surfaces.
2644 * It is ok for these surfaces to have buggy content,
2645 * but we should avoid crashing.
2646 * Calling util_format_translate_3d would read out of bounds. */
2647 if (dst_box->width < bw || dst_box->height < bh)
2648 return;
2649 }
2650
2651 map = pipe->texture_map(pipe,
2652 res,
2653 level,
2654 PIPE_MAP_WRITE | PIPE_MAP_DISCARD_RANGE,
2655 dst_box, &transfer);
2656 if (!map)
2657 return;
2658
2659 /* Note: if formats are the sames, it will revert
2660 * to normal memcpy */
2661 (void) util_format_translate_3d(res->format,
2662 map, transfer->stride,
2663 transfer->layer_stride,
2664 0, 0, 0,
2665 src_format,
2666 src, src_stride,
2667 src_layer_stride,
2668 src_box->x, src_box->y, src_box->z,
2669 dst_box->width, dst_box->height,
2670 dst_box->depth);
2671
2672 pipe_texture_unmap(pipe, transfer);
2673 }
2674
2675 struct pipe_query *
nine_context_create_query(struct NineDevice9 * device,unsigned query_type)2676 nine_context_create_query(struct NineDevice9 *device, unsigned query_type)
2677 {
2678 struct pipe_context *pipe;
2679 struct pipe_query *res;
2680
2681 pipe = nine_context_get_pipe_acquire(device);
2682 res = pipe->create_query(pipe, query_type, 0);
2683 nine_context_get_pipe_release(device);
2684 return res;
2685 }
2686
CSMT_ITEM_DO_WAIT(nine_context_destroy_query,ARG_REF (struct pipe_query,query))2687 CSMT_ITEM_DO_WAIT(nine_context_destroy_query,
2688 ARG_REF(struct pipe_query, query))
2689 {
2690 struct nine_context *context = &device->context;
2691
2692 context->pipe->destroy_query(context->pipe, query);
2693 }
2694
CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_begin_query,ARG_REF (struct pipe_query,query))2695 CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_begin_query,
2696 ARG_REF(struct pipe_query, query))
2697 {
2698 struct nine_context *context = &device->context;
2699
2700 (void) context->pipe->begin_query(context->pipe, query);
2701 }
2702
CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_end_query,ARG_REF (struct pipe_query,query))2703 CSMT_ITEM_NO_WAIT_WITH_COUNTER(nine_context_end_query,
2704 ARG_REF(struct pipe_query, query))
2705 {
2706 struct nine_context *context = &device->context;
2707
2708 (void) context->pipe->end_query(context->pipe, query);
2709 }
2710
2711 bool
nine_context_get_query_result(struct NineDevice9 * device,struct pipe_query * query,unsigned * counter,bool flush,bool wait,union pipe_query_result * result)2712 nine_context_get_query_result(struct NineDevice9 *device, struct pipe_query *query,
2713 unsigned *counter, bool flush, bool wait,
2714 union pipe_query_result *result)
2715 {
2716 struct pipe_context *pipe;
2717 bool ret;
2718
2719 if (wait)
2720 nine_csmt_process(device);
2721 else if (p_atomic_read(counter) > 0) {
2722 if (flush && device->csmt_active)
2723 nine_queue_flush(device->csmt_ctx->pool);
2724 DBG("Pending begin/end. Returning\n");
2725 return false;
2726 }
2727
2728 pipe = nine_context_get_pipe_acquire(device);
2729 ret = pipe->get_query_result(pipe, query, wait, result);
2730 nine_context_get_pipe_release(device);
2731
2732 DBG("Query result %s\n", ret ? "found" : "not yet available");
2733 return ret;
2734 }
2735
CSMT_ITEM_NO_WAIT(nine_context_pipe_flush)2736 CSMT_ITEM_NO_WAIT(nine_context_pipe_flush)
2737 {
2738 struct nine_context *context = &device->context;
2739
2740 context->pipe->flush(context->pipe, NULL, PIPE_FLUSH_ASYNC);
2741 }
2742
2743 /* State defaults */
2744
2745 static const DWORD nine_render_state_defaults[NINED3DRS_LAST + 1] =
2746 {
2747 /* [D3DRS_ZENABLE] = D3DZB_TRUE; wine: auto_depth_stencil */
2748 [D3DRS_ZENABLE] = D3DZB_FALSE,
2749 [D3DRS_FILLMODE] = D3DFILL_SOLID,
2750 [D3DRS_SHADEMODE] = D3DSHADE_GOURAUD,
2751 /* [D3DRS_LINEPATTERN] = 0x00000000, */
2752 [D3DRS_ZWRITEENABLE] = true,
2753 [D3DRS_ALPHATESTENABLE] = false,
2754 [D3DRS_LASTPIXEL] = true,
2755 [D3DRS_SRCBLEND] = D3DBLEND_ONE,
2756 [D3DRS_DESTBLEND] = D3DBLEND_ZERO,
2757 [D3DRS_CULLMODE] = D3DCULL_CCW,
2758 [D3DRS_ZFUNC] = D3DCMP_LESSEQUAL,
2759 [D3DRS_ALPHAFUNC] = D3DCMP_ALWAYS,
2760 [D3DRS_ALPHAREF] = 0,
2761 [D3DRS_DITHERENABLE] = false,
2762 [D3DRS_ALPHABLENDENABLE] = false,
2763 [D3DRS_FOGENABLE] = false,
2764 [D3DRS_SPECULARENABLE] = false,
2765 /* [D3DRS_ZVISIBLE] = 0, */
2766 [D3DRS_FOGCOLOR] = 0,
2767 [D3DRS_FOGTABLEMODE] = D3DFOG_NONE,
2768 [D3DRS_FOGSTART] = 0x00000000,
2769 [D3DRS_FOGEND] = 0x3F800000,
2770 [D3DRS_FOGDENSITY] = 0x3F800000,
2771 /* [D3DRS_EDGEANTIALIAS] = FALSE, */
2772 [D3DRS_RANGEFOGENABLE] = false,
2773 [D3DRS_STENCILENABLE] = false,
2774 [D3DRS_STENCILFAIL] = D3DSTENCILOP_KEEP,
2775 [D3DRS_STENCILZFAIL] = D3DSTENCILOP_KEEP,
2776 [D3DRS_STENCILPASS] = D3DSTENCILOP_KEEP,
2777 [D3DRS_STENCILREF] = 0,
2778 [D3DRS_STENCILMASK] = 0xFFFFFFFF,
2779 [D3DRS_STENCILFUNC] = D3DCMP_ALWAYS,
2780 [D3DRS_STENCILWRITEMASK] = 0xFFFFFFFF,
2781 [D3DRS_TEXTUREFACTOR] = 0xFFFFFFFF,
2782 [D3DRS_WRAP0] = 0,
2783 [D3DRS_WRAP1] = 0,
2784 [D3DRS_WRAP2] = 0,
2785 [D3DRS_WRAP3] = 0,
2786 [D3DRS_WRAP4] = 0,
2787 [D3DRS_WRAP5] = 0,
2788 [D3DRS_WRAP6] = 0,
2789 [D3DRS_WRAP7] = 0,
2790 [D3DRS_CLIPPING] = true,
2791 [D3DRS_LIGHTING] = true,
2792 [D3DRS_AMBIENT] = 0,
2793 [D3DRS_FOGVERTEXMODE] = D3DFOG_NONE,
2794 [D3DRS_COLORVERTEX] = true,
2795 [D3DRS_LOCALVIEWER] = true,
2796 [D3DRS_NORMALIZENORMALS] = false,
2797 [D3DRS_DIFFUSEMATERIALSOURCE] = D3DMCS_COLOR1,
2798 [D3DRS_SPECULARMATERIALSOURCE] = D3DMCS_COLOR2,
2799 [D3DRS_AMBIENTMATERIALSOURCE] = D3DMCS_MATERIAL,
2800 [D3DRS_EMISSIVEMATERIALSOURCE] = D3DMCS_MATERIAL,
2801 [D3DRS_VERTEXBLEND] = D3DVBF_DISABLE,
2802 [D3DRS_CLIPPLANEENABLE] = 0,
2803 /* [D3DRS_SOFTWAREVERTEXPROCESSING] = FALSE, */
2804 [D3DRS_POINTSIZE] = 0x3F800000,
2805 [D3DRS_POINTSIZE_MIN] = 0x3F800000,
2806 [D3DRS_POINTSPRITEENABLE] = false,
2807 [D3DRS_POINTSCALEENABLE] = false,
2808 [D3DRS_POINTSCALE_A] = 0x3F800000,
2809 [D3DRS_POINTSCALE_B] = 0x00000000,
2810 [D3DRS_POINTSCALE_C] = 0x00000000,
2811 [D3DRS_MULTISAMPLEANTIALIAS] = true,
2812 [D3DRS_MULTISAMPLEMASK] = 0xFFFFFFFF,
2813 [D3DRS_PATCHEDGESTYLE] = D3DPATCHEDGE_DISCRETE,
2814 /* [D3DRS_PATCHSEGMENTS] = 0x3F800000, */
2815 [D3DRS_DEBUGMONITORTOKEN] = 0xDEADCAFE,
2816 [D3DRS_POINTSIZE_MAX] = 0x3F800000, /* depends on cap */
2817 [D3DRS_INDEXEDVERTEXBLENDENABLE] = false,
2818 [D3DRS_COLORWRITEENABLE] = 0x0000000f,
2819 [D3DRS_TWEENFACTOR] = 0x00000000,
2820 [D3DRS_BLENDOP] = D3DBLENDOP_ADD,
2821 [D3DRS_POSITIONDEGREE] = D3DDEGREE_CUBIC,
2822 [D3DRS_NORMALDEGREE] = D3DDEGREE_LINEAR,
2823 [D3DRS_SCISSORTESTENABLE] = false,
2824 [D3DRS_SLOPESCALEDEPTHBIAS] = 0,
2825 [D3DRS_MINTESSELLATIONLEVEL] = 0x3F800000,
2826 [D3DRS_MAXTESSELLATIONLEVEL] = 0x3F800000,
2827 [D3DRS_ANTIALIASEDLINEENABLE] = false,
2828 [D3DRS_ADAPTIVETESS_X] = 0x00000000,
2829 [D3DRS_ADAPTIVETESS_Y] = 0x00000000,
2830 [D3DRS_ADAPTIVETESS_Z] = 0x3F800000,
2831 [D3DRS_ADAPTIVETESS_W] = 0x00000000,
2832 [D3DRS_ENABLEADAPTIVETESSELLATION] = false,
2833 [D3DRS_TWOSIDEDSTENCILMODE] = false,
2834 [D3DRS_CCW_STENCILFAIL] = D3DSTENCILOP_KEEP,
2835 [D3DRS_CCW_STENCILZFAIL] = D3DSTENCILOP_KEEP,
2836 [D3DRS_CCW_STENCILPASS] = D3DSTENCILOP_KEEP,
2837 [D3DRS_CCW_STENCILFUNC] = D3DCMP_ALWAYS,
2838 [D3DRS_COLORWRITEENABLE1] = 0x0000000F,
2839 [D3DRS_COLORWRITEENABLE2] = 0x0000000F,
2840 [D3DRS_COLORWRITEENABLE3] = 0x0000000F,
2841 [D3DRS_BLENDFACTOR] = 0xFFFFFFFF,
2842 [D3DRS_SRGBWRITEENABLE] = 0,
2843 [D3DRS_DEPTHBIAS] = 0,
2844 [D3DRS_WRAP8] = 0,
2845 [D3DRS_WRAP9] = 0,
2846 [D3DRS_WRAP10] = 0,
2847 [D3DRS_WRAP11] = 0,
2848 [D3DRS_WRAP12] = 0,
2849 [D3DRS_WRAP13] = 0,
2850 [D3DRS_WRAP14] = 0,
2851 [D3DRS_WRAP15] = 0,
2852 [D3DRS_SEPARATEALPHABLENDENABLE] = false,
2853 [D3DRS_SRCBLENDALPHA] = D3DBLEND_ONE,
2854 [D3DRS_DESTBLENDALPHA] = D3DBLEND_ZERO,
2855 [D3DRS_BLENDOPALPHA] = D3DBLENDOP_ADD,
2856 [NINED3DRS_VSPOINTSIZE] = false,
2857 [NINED3DRS_RTMASK] = 0xf,
2858 [NINED3DRS_ALPHACOVERAGE] = false,
2859 [NINED3DRS_MULTISAMPLE] = false,
2860 [NINED3DRS_FETCH4] = 0,
2861 [NINED3DRS_EMULATED_ALPHATEST] = 7 /* ALWAYS pass */
2862 };
2863 static const DWORD nine_tex_stage_state_defaults[NINED3DTSS_LAST + 1] =
2864 {
2865 [D3DTSS_COLOROP] = D3DTOP_DISABLE,
2866 [D3DTSS_ALPHAOP] = D3DTOP_DISABLE,
2867 [D3DTSS_COLORARG1] = D3DTA_TEXTURE,
2868 [D3DTSS_COLORARG2] = D3DTA_CURRENT,
2869 [D3DTSS_COLORARG0] = D3DTA_CURRENT,
2870 [D3DTSS_ALPHAARG1] = D3DTA_TEXTURE,
2871 [D3DTSS_ALPHAARG2] = D3DTA_CURRENT,
2872 [D3DTSS_ALPHAARG0] = D3DTA_CURRENT,
2873 [D3DTSS_RESULTARG] = D3DTA_CURRENT,
2874 [D3DTSS_BUMPENVMAT00] = 0,
2875 [D3DTSS_BUMPENVMAT01] = 0,
2876 [D3DTSS_BUMPENVMAT10] = 0,
2877 [D3DTSS_BUMPENVMAT11] = 0,
2878 [D3DTSS_BUMPENVLSCALE] = 0,
2879 [D3DTSS_BUMPENVLOFFSET] = 0,
2880 [D3DTSS_TEXCOORDINDEX] = 0,
2881 [D3DTSS_TEXTURETRANSFORMFLAGS] = D3DTTFF_DISABLE,
2882 };
2883 static const DWORD nine_samp_state_defaults[NINED3DSAMP_LAST + 1] =
2884 {
2885 [D3DSAMP_ADDRESSU] = D3DTADDRESS_WRAP,
2886 [D3DSAMP_ADDRESSV] = D3DTADDRESS_WRAP,
2887 [D3DSAMP_ADDRESSW] = D3DTADDRESS_WRAP,
2888 [D3DSAMP_BORDERCOLOR] = 0,
2889 [D3DSAMP_MAGFILTER] = D3DTEXF_POINT,
2890 [D3DSAMP_MINFILTER] = D3DTEXF_POINT,
2891 [D3DSAMP_MIPFILTER] = D3DTEXF_NONE,
2892 [D3DSAMP_MIPMAPLODBIAS] = 0,
2893 [D3DSAMP_MAXMIPLEVEL] = 0,
2894 [D3DSAMP_MAXANISOTROPY] = 1,
2895 [D3DSAMP_SRGBTEXTURE] = 0,
2896 [D3DSAMP_ELEMENTINDEX] = 0,
2897 [D3DSAMP_DMAPOFFSET] = 0,
2898 [NINED3DSAMP_MINLOD] = 0,
2899 [NINED3DSAMP_SHADOW] = 0,
2900 [NINED3DSAMP_CUBETEX] = 0
2901 };
2902
2903 /* Note: The following 4 functions assume there is no
2904 * pending commands */
2905
nine_state_restore_non_cso(struct NineDevice9 * device)2906 void nine_state_restore_non_cso(struct NineDevice9 *device)
2907 {
2908 struct nine_context *context = &device->context;
2909
2910 context->changed.group = NINE_STATE_ALL; /* TODO: we can remove states that have prepared commits */
2911 context->changed.vtxbuf = (1ULL << device->caps.MaxStreams) - 1;
2912 context->changed.ucp = true;
2913 context->commit |= 0xffffffff; /* re-commit everything */
2914 context->enabled_sampler_count_vs = 0;
2915 context->enabled_sampler_count_ps = 0;
2916 }
2917
2918 void
nine_state_set_defaults(struct NineDevice9 * device,const D3DCAPS9 * caps,bool is_reset)2919 nine_state_set_defaults(struct NineDevice9 *device, const D3DCAPS9 *caps,
2920 bool is_reset)
2921 {
2922 struct nine_state *state = &device->state;
2923 struct nine_context *context = &device->context;
2924 unsigned s;
2925
2926 /* Initialize defaults.
2927 */
2928 memcpy(context->rs, nine_render_state_defaults, sizeof(context->rs));
2929
2930 for (s = 0; s < ARRAY_SIZE(state->ff.tex_stage); ++s) {
2931 memcpy(&state->ff.tex_stage[s], nine_tex_stage_state_defaults,
2932 sizeof(state->ff.tex_stage[s]));
2933 state->ff.tex_stage[s][D3DTSS_TEXCOORDINDEX] = s;
2934 }
2935 state->ff.tex_stage[0][D3DTSS_COLOROP] = D3DTOP_MODULATE;
2936 state->ff.tex_stage[0][D3DTSS_ALPHAOP] = D3DTOP_SELECTARG1;
2937
2938 for (s = 0; s < ARRAY_SIZE(state->ff.tex_stage); ++s)
2939 memcpy(&context->ff.tex_stage[s], state->ff.tex_stage[s],
2940 sizeof(state->ff.tex_stage[s]));
2941
2942 memset(&context->bumpmap_vars, 0, sizeof(context->bumpmap_vars));
2943
2944 for (s = 0; s < NINE_MAX_SAMPLERS; ++s) {
2945 memcpy(&context->samp[s], nine_samp_state_defaults,
2946 sizeof(context->samp[s]));
2947 memcpy(&state->samp_advertised[s], nine_samp_state_defaults,
2948 sizeof(state->samp_advertised[s]));
2949 }
2950
2951 memset(state->vs_const_f, 0, VS_CONST_F_SIZE(device));
2952 memset(context->vs_const_f, 0, device->vs_const_size);
2953 if (context->vs_const_f_swvp)
2954 memset(context->vs_const_f_swvp, 0, NINE_MAX_CONST_F_SWVP * sizeof(float[4]));
2955 memset(state->vs_const_i, 0, VS_CONST_I_SIZE(device));
2956 memset(context->vs_const_i, 0, VS_CONST_I_SIZE(device));
2957 memset(state->vs_const_b, 0, VS_CONST_B_SIZE(device));
2958 memset(context->vs_const_b, 0, VS_CONST_B_SIZE(device));
2959 memset(state->ps_const_f, 0, device->ps_const_size);
2960 memset(context->ps_const_f, 0, device->ps_const_size);
2961 memset(state->ps_const_i, 0, sizeof(state->ps_const_i));
2962 memset(context->ps_const_i, 0, sizeof(context->ps_const_i));
2963 memset(state->ps_const_b, 0, sizeof(state->ps_const_b));
2964 memset(context->ps_const_b, 0, sizeof(context->ps_const_b));
2965 context->zfog = false; /* Guess from wine tests: both true or false are ok */
2966
2967 /* Cap dependent initial state:
2968 */
2969 context->rs[D3DRS_POINTSIZE_MAX] = fui(caps->MaxPointSize);
2970
2971 memcpy(state->rs_advertised, context->rs, sizeof(context->rs));
2972
2973 /* Set changed flags to initialize driver.
2974 */
2975 context->changed.group = NINE_STATE_ALL;
2976 context->changed.vtxbuf = (1ULL << device->caps.MaxStreams) - 1;
2977 context->changed.ucp = true;
2978
2979 context->ff.changed.transform[0] = ~0;
2980 context->ff.changed.transform[D3DTS_WORLD / 32] |= 1 << (D3DTS_WORLD % 32);
2981
2982 if (!is_reset) {
2983 state->viewport.MinZ = context->viewport.MinZ = 0.0f;
2984 state->viewport.MaxZ = context->viewport.MaxZ = 1.0f;
2985 }
2986
2987 for (s = 0; s < NINE_MAX_SAMPLERS; ++s)
2988 context->changed.sampler[s] = ~0;
2989
2990 if (!is_reset)
2991 context->dummy_vbo_bound_at = -1;
2992 }
2993
2994 void
nine_device_state_clear(struct NineDevice9 * device)2995 nine_device_state_clear(struct NineDevice9 *device)
2996 {
2997 struct nine_state *state = &device->state;
2998 unsigned i;
2999
3000 for (i = 0; i < ARRAY_SIZE(state->rt); ++i)
3001 nine_bind(&state->rt[i], NULL);
3002 nine_bind(&state->ds, NULL);
3003 nine_bind(&state->vs, NULL);
3004 nine_bind(&state->ps, NULL);
3005 nine_bind(&state->vdecl, NULL);
3006 for (i = 0; i < PIPE_MAX_ATTRIBS; ++i)
3007 NineBindBufferToDevice(device,
3008 (struct NineBuffer9 **)&state->stream[i],
3009 NULL);
3010 NineBindBufferToDevice(device,
3011 (struct NineBuffer9 **)&state->idxbuf,
3012 NULL);
3013
3014 for (i = 0; i < NINE_MAX_SAMPLERS; ++i)
3015 NineBindTextureToDevice(device, &state->texture[i], NULL);
3016 }
3017
3018 void
nine_context_clear(struct NineDevice9 * device)3019 nine_context_clear(struct NineDevice9 *device)
3020 {
3021 struct nine_context *context = &device->context;
3022 struct pipe_context *pipe = context->pipe;
3023 struct cso_context *cso = context->cso;
3024 unsigned i;
3025
3026 /* Early device ctor failure. Nothing to do */
3027 if (!pipe || !cso)
3028 return;
3029
3030 context->vtxbuf_mask = 0;
3031
3032 pipe->bind_vs_state(pipe, NULL);
3033 pipe->bind_fs_state(pipe, NULL);
3034
3035 /* Don't unbind constant buffers, they're device-private and
3036 * do not change on Reset.
3037 */
3038
3039 cso_set_samplers(cso, PIPE_SHADER_VERTEX, 0, NULL);
3040 cso_set_samplers(cso, PIPE_SHADER_FRAGMENT, 0, NULL);
3041 context->enabled_sampler_count_vs = 0;
3042 context->enabled_sampler_count_ps = 0;
3043
3044 pipe->set_sampler_views(pipe, PIPE_SHADER_VERTEX, 0, 0,
3045 NINE_MAX_SAMPLERS_VS, false, NULL);
3046 pipe->set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0, 0,
3047 NINE_MAX_SAMPLERS_PS, false, NULL);
3048
3049 pipe->set_vertex_buffers(pipe, 0, NULL);
3050
3051 for (i = 0; i < ARRAY_SIZE(context->rt); ++i)
3052 nine_bind(&context->rt[i], NULL);
3053 nine_bind(&context->ds, NULL);
3054 nine_bind(&context->vs, NULL);
3055 nine_bind(&context->ps, NULL);
3056 nine_bind(&context->vdecl, NULL);
3057 for (i = 0; i < PIPE_MAX_ATTRIBS; ++i)
3058 pipe_vertex_buffer_unreference(&context->vtxbuf[i]);
3059 pipe_resource_reference(&context->idxbuf, NULL);
3060 pipe_resource_reference(&context->pipe_data.cb_vs.buffer, NULL);
3061 pipe_resource_reference(&context->pipe_data.cb_ps.buffer, NULL);
3062
3063 for (i = 0; i < NINE_MAX_SAMPLERS; ++i) {
3064 context->texture[i].enabled = false;
3065 pipe_resource_reference(&context->texture[i].resource,
3066 NULL);
3067 pipe_sampler_view_reference(&context->texture[i].view[0],
3068 NULL);
3069 pipe_sampler_view_reference(&context->texture[i].view[1],
3070 NULL);
3071 }
3072 }
3073
3074 void
nine_context_update_state(struct NineDevice9 * device)3075 nine_context_update_state(struct NineDevice9 *device)
3076 {
3077 nine_update_state(device);
3078 }
3079
3080 void
nine_state_init_sw(struct NineDevice9 * device)3081 nine_state_init_sw(struct NineDevice9 *device)
3082 {
3083 struct pipe_context *pipe_sw = device->pipe_sw;
3084 struct pipe_rasterizer_state rast;
3085 struct pipe_blend_state blend;
3086 struct pipe_depth_stencil_alpha_state dsa;
3087 struct pipe_framebuffer_state fb;
3088
3089 /* Only used with Streamout */
3090 memset(&rast, 0, sizeof(rast));
3091 rast.rasterizer_discard = true;
3092 rast.point_quad_rasterization = 1; /* to make llvmpipe happy */
3093 cso_set_rasterizer(device->cso_sw, &rast);
3094
3095 /* dummy settings */
3096 memset(&blend, 0, sizeof(blend));
3097 memset(&dsa, 0, sizeof(dsa));
3098 memset(&fb, 0, sizeof(fb));
3099 cso_set_blend(device->cso_sw, &blend);
3100 cso_set_depth_stencil_alpha(device->cso_sw, &dsa);
3101 cso_set_framebuffer(device->cso_sw, &fb);
3102 cso_set_viewport_dims(device->cso_sw, 1.0, 1.0, false);
3103 cso_set_fragment_shader_handle(device->cso_sw, util_make_empty_fragment_shader(pipe_sw));
3104 }
3105
3106 /* There is duplication with update_vertex_elements.
3107 * TODO: Share the code */
3108
3109 static int
update_vertex_elements_sw(struct NineDevice9 * device)3110 update_vertex_elements_sw(struct NineDevice9 *device)
3111 {
3112 struct nine_state *state = &device->state;
3113 const struct NineVertexDeclaration9 *vdecl = device->state.vdecl;
3114 const struct NineVertexShader9 *vs;
3115 unsigned n, b, i, j;
3116 int index;
3117 int8_t vdecl_index_map[16]; /* vs->num_inputs <= 16 */
3118 int8_t used_streams[device->caps.MaxStreams];
3119 BOOL need_dummy_vbo = false;
3120 int dummy_vbo_stream = -1;
3121 struct cso_velems_state ve;
3122 bool programmable_vs = state->vs && !(state->vdecl && state->vdecl->position_t);
3123 unsigned vtxbuf_holes_map[PIPE_MAX_ATTRIBS];
3124
3125 memset(vdecl_index_map, -1, 16);
3126 memset(used_streams, 0, device->caps.MaxStreams);
3127 vs = programmable_vs ? device->state.vs : device->ff.vs;
3128
3129 if (vdecl) {
3130 for (n = 0; n < vs->num_inputs; ++n) {
3131 DBG("looking up input %u (usage %u) from vdecl(%p)\n",
3132 n, vs->input_map[n].ndecl, vdecl);
3133
3134 for (i = 0; i < vdecl->nelems; i++) {
3135 if (vdecl->usage_map[i] == vs->input_map[n].ndecl) {
3136 vdecl_index_map[n] = i;
3137 used_streams[vdecl->elems[i].vertex_buffer_index] = 1;
3138 break;
3139 }
3140 }
3141 if (vdecl_index_map[n] < 0)
3142 need_dummy_vbo = true;
3143 }
3144 } else {
3145 /* No vertex declaration. Likely will never happen in practice,
3146 * but we need not crash on this */
3147 need_dummy_vbo = true;
3148 }
3149
3150 if (need_dummy_vbo) {
3151 for (i = 0; i < device->caps.MaxStreams; i++) {
3152 if (!used_streams[i]) {
3153 dummy_vbo_stream = i;
3154 break;
3155 }
3156 }
3157 }
3158 /* there are less vertex shader inputs than stream slots,
3159 * so if we need a slot for the dummy vbo, we should have found one */
3160 assert (!need_dummy_vbo || dummy_vbo_stream != -1);
3161
3162 /* calculate vtxbuf_holes_map to match for update_vertex_buffers_sw() function */
3163 j = 0;
3164 for (i = 0; i < device->caps.MaxStreams; i++) {
3165 if (state->stream[i] || (need_dummy_vbo && dummy_vbo_stream == i)) {
3166 vtxbuf_holes_map[i] = j;
3167 j++;
3168 }
3169 }
3170
3171 for (n = 0; n < vs->num_inputs; ++n) {
3172 index = vdecl_index_map[n];
3173 if (index >= 0) {
3174 ve.velems[n] = vdecl->elems[index];
3175 b = ve.velems[n].vertex_buffer_index;
3176 ve.velems[n].vertex_buffer_index = vtxbuf_holes_map[b];
3177 /* XXX wine just uses 1 here: */
3178 if (state->stream_freq[b] & D3DSTREAMSOURCE_INSTANCEDATA)
3179 ve.velems[n].instance_divisor = state->stream_freq[b] & 0x7FFFFF;
3180 } else {
3181 /* if the vertex declaration is incomplete compared to what the
3182 * vertex shader needs, we bind a dummy vbo with 0 0 0 0.
3183 * This is not precised by the spec, but is the behaviour
3184 * tested on win */
3185 ve.velems[n].vertex_buffer_index = vtxbuf_holes_map[dummy_vbo_stream];
3186 ve.velems[n].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
3187 ve.velems[n].src_offset = 0;
3188 ve.velems[n].instance_divisor = 0;
3189 ve.velems[n].dual_slot = false;
3190 }
3191 }
3192
3193 ve.count = vs->num_inputs;
3194 cso_set_vertex_elements(device->cso_sw, &ve);
3195 return dummy_vbo_stream;
3196 }
3197
3198 static void
update_vertex_buffers_sw(struct NineDevice9 * device,int dummy_vbo_stream,int start_vertice,int num_vertices)3199 update_vertex_buffers_sw(struct NineDevice9 *device, int dummy_vbo_stream,
3200 int start_vertice, int num_vertices)
3201 {
3202 struct pipe_context *pipe = nine_context_get_pipe_acquire(device);
3203 struct pipe_context *pipe_sw = device->pipe_sw;
3204 struct nine_state *state = &device->state;
3205 struct nine_state_sw_internal *sw_internal = &device->state_sw_internal;
3206 struct pipe_vertex_buffer vbuffer[PIPE_MAX_ATTRIBS];
3207 unsigned vtxbuf_count = 0;
3208 unsigned i, j = 0;
3209
3210 for (i = 0; i < device->caps.MaxStreams; i++) {
3211 if (dummy_vbo_stream == i) {
3212 vbuffer[j].buffer.resource = NULL;
3213 pipe_resource_reference(&vbuffer[j].buffer.resource, device->dummy_vbo_sw);
3214 vbuffer[j].is_user_buffer = false;
3215 vbuffer[j].buffer_offset = 0;
3216 j++;
3217 } else if (state->stream[i]) {
3218 unsigned offset;
3219 struct pipe_resource *buf;
3220 struct pipe_box box;
3221 void *userbuf;
3222
3223 vbuffer[j] = state->vtxbuf[i];
3224 buf = NineVertexBuffer9_GetResource(state->stream[i], &offset);
3225
3226 DBG("Locking %p (offset %d, length %d)\n", buf,
3227 vbuffer[j].buffer_offset, num_vertices * state->vtxstride[i]);
3228
3229 u_box_1d(vbuffer[j].buffer_offset + offset + start_vertice *
3230 state->vtxstride[i], num_vertices * state->vtxstride[i], &box);
3231
3232 userbuf = pipe->buffer_map(pipe, buf, 0, PIPE_MAP_READ, &box,
3233 &(sw_internal->transfers_so[i]));
3234 vbuffer[j].is_user_buffer = true;
3235 vbuffer[j].buffer.user = userbuf;
3236
3237 if (!device->driver_caps.user_sw_vbufs) {
3238 vbuffer[j].buffer.resource = NULL;
3239 vbuffer[j].is_user_buffer = false;
3240 u_upload_data(device->pipe_sw->stream_uploader,
3241 0,
3242 box.width,
3243 16,
3244 userbuf,
3245 &(vbuffer[j].buffer_offset),
3246 &(vbuffer[j].buffer.resource));
3247 u_upload_unmap(device->pipe_sw->stream_uploader);
3248 }
3249 j++;
3250 }
3251 }
3252
3253 vtxbuf_count = j;
3254 pipe_sw->set_vertex_buffers(pipe_sw, vtxbuf_count, vbuffer);
3255
3256 nine_context_get_pipe_release(device);
3257 }
3258
3259 static void
update_vs_constants_sw(struct NineDevice9 * device)3260 update_vs_constants_sw(struct NineDevice9 *device)
3261 {
3262 struct nine_state *state = &device->state;
3263 struct pipe_context *pipe_sw = device->pipe_sw;
3264
3265 DBG("updating\n");
3266
3267 {
3268 struct pipe_constant_buffer cb;
3269 const void *buf;
3270
3271 cb.buffer = NULL;
3272 cb.buffer_offset = 0;
3273 cb.buffer_size = 4096 * sizeof(float[4]);
3274 cb.user_buffer = state->vs_const_f;
3275
3276 if (state->vs->lconstf.ranges) {
3277 const struct nine_lconstf *lconstf = &device->state.vs->lconstf;
3278 const struct nine_range *r = lconstf->ranges;
3279 unsigned n = 0;
3280 float *dst = device->state.vs_lconstf_temp;
3281 float *src = (float *)cb.user_buffer;
3282 memcpy(dst, src, 8192 * sizeof(float[4]));
3283 while (r) {
3284 unsigned p = r->bgn;
3285 unsigned c = r->end - r->bgn;
3286 memcpy(&dst[p * 4], &lconstf->data[n * 4], c * 4 * sizeof(float));
3287 n += c;
3288 r = r->next;
3289 }
3290 cb.user_buffer = dst;
3291 }
3292
3293 buf = cb.user_buffer;
3294
3295 pipe_sw->set_constant_buffer(pipe_sw, PIPE_SHADER_VERTEX, 0, false, &cb);
3296 if (cb.buffer)
3297 pipe_resource_reference(&cb.buffer, NULL);
3298
3299 cb.user_buffer = (int8_t *)buf + 4096 * sizeof(float[4]);
3300
3301 pipe_sw->set_constant_buffer(pipe_sw, PIPE_SHADER_VERTEX, 1, false, &cb);
3302 if (cb.buffer)
3303 pipe_resource_reference(&cb.buffer, NULL);
3304 }
3305
3306 {
3307 struct pipe_constant_buffer cb;
3308
3309 cb.buffer = NULL;
3310 cb.buffer_offset = 0;
3311 cb.buffer_size = 2048 * sizeof(float[4]);
3312 cb.user_buffer = state->vs_const_i;
3313
3314 pipe_sw->set_constant_buffer(pipe_sw, PIPE_SHADER_VERTEX, 2, false, &cb);
3315 if (cb.buffer)
3316 pipe_resource_reference(&cb.buffer, NULL);
3317 }
3318
3319 {
3320 struct pipe_constant_buffer cb;
3321
3322 cb.buffer = NULL;
3323 cb.buffer_offset = 0;
3324 cb.buffer_size = 512 * sizeof(float[4]);
3325 cb.user_buffer = state->vs_const_b;
3326
3327 pipe_sw->set_constant_buffer(pipe_sw, PIPE_SHADER_VERTEX, 3, false, &cb);
3328 if (cb.buffer)
3329 pipe_resource_reference(&cb.buffer, NULL);
3330 }
3331
3332 {
3333 struct pipe_constant_buffer cb;
3334 const D3DVIEWPORT9 *vport = &device->state.viewport;
3335 float viewport_data[8] = {(float)vport->Width * 0.5f,
3336 (float)vport->Height * -0.5f, vport->MaxZ - vport->MinZ, 0.f,
3337 (float)vport->Width * 0.5f + (float)vport->X,
3338 (float)vport->Height * 0.5f + (float)vport->Y,
3339 vport->MinZ, 0.f};
3340
3341 cb.buffer = NULL;
3342 cb.buffer_offset = 0;
3343 cb.buffer_size = 2 * sizeof(float[4]);
3344 cb.user_buffer = viewport_data;
3345
3346 {
3347 u_upload_data(device->pipe_sw->const_uploader,
3348 0,
3349 cb.buffer_size,
3350 16,
3351 cb.user_buffer,
3352 &(cb.buffer_offset),
3353 &(cb.buffer));
3354 u_upload_unmap(device->pipe_sw->const_uploader);
3355 cb.user_buffer = NULL;
3356 }
3357
3358 pipe_sw->set_constant_buffer(pipe_sw, PIPE_SHADER_VERTEX, 4, false, &cb);
3359 if (cb.buffer)
3360 pipe_resource_reference(&cb.buffer, NULL);
3361 }
3362
3363 }
3364
3365 void
nine_state_prepare_draw_sw(struct NineDevice9 * device,struct NineVertexDeclaration9 * vdecl_out,int start_vertice,int num_vertices,struct pipe_stream_output_info * so)3366 nine_state_prepare_draw_sw(struct NineDevice9 *device, struct NineVertexDeclaration9 *vdecl_out,
3367 int start_vertice, int num_vertices, struct pipe_stream_output_info *so)
3368 {
3369 struct nine_state *state = &device->state;
3370 bool programmable_vs = state->vs && !(state->vdecl && state->vdecl->position_t);
3371 struct NineVertexShader9 *vs = programmable_vs ? device->state.vs : device->ff.vs;
3372 int dummy_vbo_stream;
3373
3374 assert(programmable_vs);
3375
3376 DBG("Preparing draw\n");
3377 cso_set_vertex_shader_handle(device->cso_sw,
3378 NineVertexShader9_GetVariantProcessVertices(vs, vdecl_out, so));
3379 dummy_vbo_stream = update_vertex_elements_sw(device);
3380 update_vertex_buffers_sw(device, dummy_vbo_stream, start_vertice, num_vertices);
3381 update_vs_constants_sw(device);
3382 DBG("Preparation succeeded\n");
3383 }
3384
3385 void
nine_state_after_draw_sw(struct NineDevice9 * device)3386 nine_state_after_draw_sw(struct NineDevice9 *device)
3387 {
3388 struct nine_state_sw_internal *sw_internal = &device->state_sw_internal;
3389 struct pipe_context *pipe = nine_context_get_pipe_acquire(device);
3390 struct pipe_context *pipe_sw = device->pipe_sw;
3391 int i;
3392
3393 pipe_sw->set_vertex_buffers(pipe_sw, 0, NULL);
3394 for (i = 0; i < 4; i++) {
3395 if (sw_internal->transfers_so[i])
3396 pipe->buffer_unmap(pipe, sw_internal->transfers_so[i]);
3397 sw_internal->transfers_so[i] = NULL;
3398 }
3399 nine_context_get_pipe_release(device);
3400 }
3401
3402 void
nine_state_destroy_sw(struct NineDevice9 * device)3403 nine_state_destroy_sw(struct NineDevice9 *device)
3404 {
3405 (void) device;
3406 /* Everything destroyed with cso */
3407 }
3408
3409 /*
3410 static const DWORD nine_render_states_pixel[] =
3411 {
3412 D3DRS_ALPHABLENDENABLE,
3413 D3DRS_ALPHAFUNC,
3414 D3DRS_ALPHAREF,
3415 D3DRS_ALPHATESTENABLE,
3416 D3DRS_ANTIALIASEDLINEENABLE,
3417 D3DRS_BLENDFACTOR,
3418 D3DRS_BLENDOP,
3419 D3DRS_BLENDOPALPHA,
3420 D3DRS_CCW_STENCILFAIL,
3421 D3DRS_CCW_STENCILPASS,
3422 D3DRS_CCW_STENCILZFAIL,
3423 D3DRS_COLORWRITEENABLE,
3424 D3DRS_COLORWRITEENABLE1,
3425 D3DRS_COLORWRITEENABLE2,
3426 D3DRS_COLORWRITEENABLE3,
3427 D3DRS_DEPTHBIAS,
3428 D3DRS_DESTBLEND,
3429 D3DRS_DESTBLENDALPHA,
3430 D3DRS_DITHERENABLE,
3431 D3DRS_FILLMODE,
3432 D3DRS_FOGDENSITY,
3433 D3DRS_FOGEND,
3434 D3DRS_FOGSTART,
3435 D3DRS_LASTPIXEL,
3436 D3DRS_SCISSORTESTENABLE,
3437 D3DRS_SEPARATEALPHABLENDENABLE,
3438 D3DRS_SHADEMODE,
3439 D3DRS_SLOPESCALEDEPTHBIAS,
3440 D3DRS_SRCBLEND,
3441 D3DRS_SRCBLENDALPHA,
3442 D3DRS_SRGBWRITEENABLE,
3443 D3DRS_STENCILENABLE,
3444 D3DRS_STENCILFAIL,
3445 D3DRS_STENCILFUNC,
3446 D3DRS_STENCILMASK,
3447 D3DRS_STENCILPASS,
3448 D3DRS_STENCILREF,
3449 D3DRS_STENCILWRITEMASK,
3450 D3DRS_STENCILZFAIL,
3451 D3DRS_TEXTUREFACTOR,
3452 D3DRS_TWOSIDEDSTENCILMODE,
3453 D3DRS_WRAP0,
3454 D3DRS_WRAP1,
3455 D3DRS_WRAP10,
3456 D3DRS_WRAP11,
3457 D3DRS_WRAP12,
3458 D3DRS_WRAP13,
3459 D3DRS_WRAP14,
3460 D3DRS_WRAP15,
3461 D3DRS_WRAP2,
3462 D3DRS_WRAP3,
3463 D3DRS_WRAP4,
3464 D3DRS_WRAP5,
3465 D3DRS_WRAP6,
3466 D3DRS_WRAP7,
3467 D3DRS_WRAP8,
3468 D3DRS_WRAP9,
3469 D3DRS_ZENABLE,
3470 D3DRS_ZFUNC,
3471 D3DRS_ZWRITEENABLE
3472 };
3473 */
3474 const uint32_t nine_render_states_pixel[(NINED3DRS_LAST + 31) / 32] =
3475 {
3476 0x0f99c380, 0x1ff00070, 0x00000000, 0x00000000,
3477 0x000000ff, 0xde01c900, 0x0003ffcf
3478 };
3479
3480 /*
3481 static const DWORD nine_render_states_vertex[] =
3482 {
3483 D3DRS_ADAPTIVETESS_W,
3484 D3DRS_ADAPTIVETESS_X,
3485 D3DRS_ADAPTIVETESS_Y,
3486 D3DRS_ADAPTIVETESS_Z,
3487 D3DRS_AMBIENT,
3488 D3DRS_AMBIENTMATERIALSOURCE,
3489 D3DRS_CLIPPING,
3490 D3DRS_CLIPPLANEENABLE,
3491 D3DRS_COLORVERTEX,
3492 D3DRS_CULLMODE,
3493 D3DRS_DIFFUSEMATERIALSOURCE,
3494 D3DRS_EMISSIVEMATERIALSOURCE,
3495 D3DRS_ENABLEADAPTIVETESSELLATION,
3496 D3DRS_FOGCOLOR,
3497 D3DRS_FOGDENSITY,
3498 D3DRS_FOGENABLE,
3499 D3DRS_FOGEND,
3500 D3DRS_FOGSTART,
3501 D3DRS_FOGTABLEMODE,
3502 D3DRS_FOGVERTEXMODE,
3503 D3DRS_INDEXEDVERTEXBLENDENABLE,
3504 D3DRS_LIGHTING,
3505 D3DRS_LOCALVIEWER,
3506 D3DRS_MAXTESSELLATIONLEVEL,
3507 D3DRS_MINTESSELLATIONLEVEL,
3508 D3DRS_MULTISAMPLEANTIALIAS,
3509 D3DRS_MULTISAMPLEMASK,
3510 D3DRS_NORMALDEGREE,
3511 D3DRS_NORMALIZENORMALS,
3512 D3DRS_PATCHEDGESTYLE,
3513 D3DRS_POINTSCALE_A,
3514 D3DRS_POINTSCALE_B,
3515 D3DRS_POINTSCALE_C,
3516 D3DRS_POINTSCALEENABLE,
3517 D3DRS_POINTSIZE,
3518 D3DRS_POINTSIZE_MAX,
3519 D3DRS_POINTSIZE_MIN,
3520 D3DRS_POINTSPRITEENABLE,
3521 D3DRS_POSITIONDEGREE,
3522 D3DRS_RANGEFOGENABLE,
3523 D3DRS_SHADEMODE,
3524 D3DRS_SPECULARENABLE,
3525 D3DRS_SPECULARMATERIALSOURCE,
3526 D3DRS_TWEENFACTOR,
3527 D3DRS_VERTEXBLEND
3528 };
3529 */
3530 const uint32_t nine_render_states_vertex[(NINED3DRS_LAST + 31) / 32] =
3531 {
3532 0x30400200, 0x0001007c, 0x00000000, 0x00000000,
3533 0xfd9efb00, 0x01fc34cf, 0x00000000
3534 };
3535
3536 /* TODO: put in the right values */
3537 const uint32_t nine_render_state_group[NINED3DRS_LAST + 1] =
3538 {
3539 [D3DRS_ZENABLE] = NINE_STATE_DSA | NINE_STATE_MULTISAMPLE,
3540 [D3DRS_FILLMODE] = NINE_STATE_RASTERIZER,
3541 [D3DRS_SHADEMODE] = NINE_STATE_RASTERIZER | NINE_STATE_PS_PARAMS_MISC,
3542 [D3DRS_ZWRITEENABLE] = NINE_STATE_DSA,
3543 [D3DRS_ALPHATESTENABLE] = NINE_STATE_DSA,
3544 [D3DRS_LASTPIXEL] = NINE_STATE_RASTERIZER,
3545 [D3DRS_SRCBLEND] = NINE_STATE_BLEND,
3546 [D3DRS_DESTBLEND] = NINE_STATE_BLEND,
3547 [D3DRS_CULLMODE] = NINE_STATE_RASTERIZER,
3548 [D3DRS_ZFUNC] = NINE_STATE_DSA,
3549 [D3DRS_ALPHAREF] = NINE_STATE_DSA,
3550 [D3DRS_ALPHAFUNC] = NINE_STATE_DSA,
3551 [D3DRS_DITHERENABLE] = NINE_STATE_BLEND,
3552 [D3DRS_ALPHABLENDENABLE] = NINE_STATE_BLEND,
3553 [D3DRS_FOGENABLE] = NINE_STATE_FF_SHADER | NINE_STATE_VS_PARAMS_MISC | NINE_STATE_PS_PARAMS_MISC | NINE_STATE_PS_CONST,
3554 [D3DRS_SPECULARENABLE] = NINE_STATE_FF_LIGHTING,
3555 [D3DRS_FOGCOLOR] = NINE_STATE_FF_PS_CONSTS | NINE_STATE_PS_CONST,
3556 [D3DRS_FOGTABLEMODE] = NINE_STATE_FF_SHADER | NINE_STATE_PS_PARAMS_MISC | NINE_STATE_PS_CONST,
3557 [D3DRS_FOGSTART] = NINE_STATE_FF_VS_OTHER | NINE_STATE_FF_PS_CONSTS | NINE_STATE_PS_CONST,
3558 [D3DRS_FOGEND] = NINE_STATE_FF_VS_OTHER | NINE_STATE_FF_PS_CONSTS | NINE_STATE_PS_CONST,
3559 [D3DRS_FOGDENSITY] = NINE_STATE_FF_VS_OTHER | NINE_STATE_FF_PS_CONSTS | NINE_STATE_PS_CONST,
3560 [D3DRS_RANGEFOGENABLE] = NINE_STATE_FF_SHADER,
3561 [D3DRS_STENCILENABLE] = NINE_STATE_DSA | NINE_STATE_MULTISAMPLE,
3562 [D3DRS_STENCILFAIL] = NINE_STATE_DSA,
3563 [D3DRS_STENCILZFAIL] = NINE_STATE_DSA,
3564 [D3DRS_STENCILPASS] = NINE_STATE_DSA,
3565 [D3DRS_STENCILFUNC] = NINE_STATE_DSA,
3566 [D3DRS_STENCILREF] = NINE_STATE_STENCIL_REF,
3567 [D3DRS_STENCILMASK] = NINE_STATE_DSA,
3568 [D3DRS_STENCILWRITEMASK] = NINE_STATE_DSA,
3569 [D3DRS_TEXTUREFACTOR] = NINE_STATE_FF_PS_CONSTS,
3570 [D3DRS_WRAP0] = NINE_STATE_UNHANDLED, /* cylindrical wrap is crazy */
3571 [D3DRS_WRAP1] = NINE_STATE_UNHANDLED,
3572 [D3DRS_WRAP2] = NINE_STATE_UNHANDLED,
3573 [D3DRS_WRAP3] = NINE_STATE_UNHANDLED,
3574 [D3DRS_WRAP4] = NINE_STATE_UNHANDLED,
3575 [D3DRS_WRAP5] = NINE_STATE_UNHANDLED,
3576 [D3DRS_WRAP6] = NINE_STATE_UNHANDLED,
3577 [D3DRS_WRAP7] = NINE_STATE_UNHANDLED,
3578 [D3DRS_CLIPPING] = 0, /* software vertex processing only */
3579 [D3DRS_LIGHTING] = NINE_STATE_FF_LIGHTING,
3580 [D3DRS_AMBIENT] = NINE_STATE_FF_LIGHTING | NINE_STATE_FF_MATERIAL,
3581 [D3DRS_FOGVERTEXMODE] = NINE_STATE_FF_SHADER,
3582 [D3DRS_COLORVERTEX] = NINE_STATE_FF_LIGHTING,
3583 [D3DRS_LOCALVIEWER] = NINE_STATE_FF_LIGHTING,
3584 [D3DRS_NORMALIZENORMALS] = NINE_STATE_FF_SHADER,
3585 [D3DRS_DIFFUSEMATERIALSOURCE] = NINE_STATE_FF_LIGHTING,
3586 [D3DRS_SPECULARMATERIALSOURCE] = NINE_STATE_FF_LIGHTING,
3587 [D3DRS_AMBIENTMATERIALSOURCE] = NINE_STATE_FF_LIGHTING,
3588 [D3DRS_EMISSIVEMATERIALSOURCE] = NINE_STATE_FF_LIGHTING,
3589 [D3DRS_VERTEXBLEND] = NINE_STATE_FF_SHADER,
3590 [D3DRS_CLIPPLANEENABLE] = NINE_STATE_RASTERIZER,
3591 [D3DRS_POINTSIZE] = NINE_STATE_RASTERIZER | NINE_STATE_FF_VS_OTHER,
3592 [D3DRS_POINTSIZE_MIN] = NINE_STATE_RASTERIZER | NINE_STATE_FF_VS_OTHER | NINE_STATE_VS_PARAMS_MISC,
3593 [D3DRS_POINTSPRITEENABLE] = NINE_STATE_RASTERIZER,
3594 [D3DRS_POINTSCALEENABLE] = NINE_STATE_FF_SHADER,
3595 [D3DRS_POINTSCALE_A] = NINE_STATE_FF_VS_OTHER,
3596 [D3DRS_POINTSCALE_B] = NINE_STATE_FF_VS_OTHER,
3597 [D3DRS_POINTSCALE_C] = NINE_STATE_FF_VS_OTHER,
3598 [D3DRS_MULTISAMPLEANTIALIAS] = NINE_STATE_MULTISAMPLE,
3599 [D3DRS_MULTISAMPLEMASK] = NINE_STATE_SAMPLE_MASK,
3600 [D3DRS_PATCHEDGESTYLE] = NINE_STATE_UNHANDLED,
3601 [D3DRS_DEBUGMONITORTOKEN] = NINE_STATE_UNHANDLED,
3602 [D3DRS_POINTSIZE_MAX] = NINE_STATE_RASTERIZER | NINE_STATE_FF_VS_OTHER | NINE_STATE_VS_PARAMS_MISC,
3603 [D3DRS_INDEXEDVERTEXBLENDENABLE] = NINE_STATE_FF_SHADER,
3604 [D3DRS_COLORWRITEENABLE] = NINE_STATE_BLEND,
3605 [D3DRS_TWEENFACTOR] = NINE_STATE_FF_VS_OTHER,
3606 [D3DRS_BLENDOP] = NINE_STATE_BLEND,
3607 [D3DRS_POSITIONDEGREE] = NINE_STATE_UNHANDLED,
3608 [D3DRS_NORMALDEGREE] = NINE_STATE_UNHANDLED,
3609 [D3DRS_SCISSORTESTENABLE] = NINE_STATE_RASTERIZER,
3610 [D3DRS_SLOPESCALEDEPTHBIAS] = NINE_STATE_RASTERIZER,
3611 [D3DRS_ANTIALIASEDLINEENABLE] = NINE_STATE_RASTERIZER,
3612 [D3DRS_MINTESSELLATIONLEVEL] = NINE_STATE_UNHANDLED,
3613 [D3DRS_MAXTESSELLATIONLEVEL] = NINE_STATE_UNHANDLED,
3614 [D3DRS_ADAPTIVETESS_X] = NINE_STATE_UNHANDLED,
3615 [D3DRS_ADAPTIVETESS_Y] = NINE_STATE_UNHANDLED,
3616 [D3DRS_ADAPTIVETESS_Z] = NINE_STATE_UNHANDLED,
3617 [D3DRS_ADAPTIVETESS_W] = NINE_STATE_UNHANDLED,
3618 [D3DRS_ENABLEADAPTIVETESSELLATION] = NINE_STATE_UNHANDLED,
3619 [D3DRS_TWOSIDEDSTENCILMODE] = NINE_STATE_DSA,
3620 [D3DRS_CCW_STENCILFAIL] = NINE_STATE_DSA,
3621 [D3DRS_CCW_STENCILZFAIL] = NINE_STATE_DSA,
3622 [D3DRS_CCW_STENCILPASS] = NINE_STATE_DSA,
3623 [D3DRS_CCW_STENCILFUNC] = NINE_STATE_DSA,
3624 [D3DRS_COLORWRITEENABLE1] = NINE_STATE_BLEND,
3625 [D3DRS_COLORWRITEENABLE2] = NINE_STATE_BLEND,
3626 [D3DRS_COLORWRITEENABLE3] = NINE_STATE_BLEND,
3627 [D3DRS_BLENDFACTOR] = NINE_STATE_BLEND_COLOR,
3628 [D3DRS_SRGBWRITEENABLE] = NINE_STATE_FB,
3629 [D3DRS_DEPTHBIAS] = NINE_STATE_RASTERIZER,
3630 [D3DRS_WRAP8] = NINE_STATE_UNHANDLED, /* cylwrap has to be done via GP */
3631 [D3DRS_WRAP9] = NINE_STATE_UNHANDLED,
3632 [D3DRS_WRAP10] = NINE_STATE_UNHANDLED,
3633 [D3DRS_WRAP11] = NINE_STATE_UNHANDLED,
3634 [D3DRS_WRAP12] = NINE_STATE_UNHANDLED,
3635 [D3DRS_WRAP13] = NINE_STATE_UNHANDLED,
3636 [D3DRS_WRAP14] = NINE_STATE_UNHANDLED,
3637 [D3DRS_WRAP15] = NINE_STATE_UNHANDLED,
3638 [D3DRS_SEPARATEALPHABLENDENABLE] = NINE_STATE_BLEND,
3639 [D3DRS_SRCBLENDALPHA] = NINE_STATE_BLEND,
3640 [D3DRS_DESTBLENDALPHA] = NINE_STATE_BLEND,
3641 [D3DRS_BLENDOPALPHA] = NINE_STATE_BLEND
3642 };
3643
3644 /* Misc */
3645
3646 static D3DMATRIX nine_state_identity = { .m[0] = { 1, 0, 0, 0 },
3647 .m[1] = { 0, 1, 0, 0 },
3648 .m[2] = { 0, 0, 1, 0 },
3649 .m[3] = { 0, 0, 0, 1 } };
3650
3651 void
nine_state_resize_transform(struct nine_ff_state * ff_state,unsigned N)3652 nine_state_resize_transform(struct nine_ff_state *ff_state, unsigned N)
3653 {
3654 unsigned n = ff_state->num_transforms;
3655
3656 if (N <= n)
3657 return;
3658
3659 ff_state->transform = REALLOC(ff_state->transform,
3660 n * sizeof(D3DMATRIX),
3661 N * sizeof(D3DMATRIX));
3662 for (; n < N; ++n)
3663 ff_state->transform[n] = nine_state_identity;
3664 ff_state->num_transforms = N;
3665 }
3666
3667 D3DMATRIX *
nine_state_access_transform(struct nine_ff_state * ff_state,D3DTRANSFORMSTATETYPE t,bool alloc)3668 nine_state_access_transform(struct nine_ff_state *ff_state, D3DTRANSFORMSTATETYPE t,
3669 bool alloc)
3670 {
3671 unsigned index;
3672
3673 switch (t) {
3674 case D3DTS_VIEW: index = 0; break;
3675 case D3DTS_PROJECTION: index = 1; break;
3676 case D3DTS_TEXTURE0: index = 2; break;
3677 case D3DTS_TEXTURE1: index = 3; break;
3678 case D3DTS_TEXTURE2: index = 4; break;
3679 case D3DTS_TEXTURE3: index = 5; break;
3680 case D3DTS_TEXTURE4: index = 6; break;
3681 case D3DTS_TEXTURE5: index = 7; break;
3682 case D3DTS_TEXTURE6: index = 8; break;
3683 case D3DTS_TEXTURE7: index = 9; break;
3684 default:
3685 if (!(t >= D3DTS_WORLDMATRIX(0) && t <= D3DTS_WORLDMATRIX(255)))
3686 return NULL;
3687 index = 10 + (t - D3DTS_WORLDMATRIX(0));
3688 break;
3689 }
3690
3691 if (index >= ff_state->num_transforms) {
3692 if (!alloc)
3693 return &nine_state_identity;
3694 nine_state_resize_transform(ff_state, index + 1);
3695 }
3696 return &ff_state->transform[index];
3697 }
3698
3699 HRESULT
nine_state_set_light(struct nine_ff_state * ff_state,DWORD Index,const D3DLIGHT9 * pLight)3700 nine_state_set_light(struct nine_ff_state *ff_state, DWORD Index,
3701 const D3DLIGHT9 *pLight)
3702 {
3703 if (Index >= ff_state->num_lights) {
3704 unsigned n = ff_state->num_lights;
3705 unsigned N = Index + 1;
3706
3707 ff_state->light = REALLOC(ff_state->light, n * sizeof(D3DLIGHT9),
3708 N * sizeof(D3DLIGHT9));
3709 if (!ff_state->light)
3710 return E_OUTOFMEMORY;
3711 ff_state->num_lights = N;
3712
3713 for (; n < Index; ++n) {
3714 memset(&ff_state->light[n], 0, sizeof(D3DLIGHT9));
3715 ff_state->light[n].Type = (D3DLIGHTTYPE)NINED3DLIGHT_INVALID;
3716 }
3717 }
3718 ff_state->light[Index] = *pLight;
3719
3720 if (pLight->Type == D3DLIGHT_SPOT && pLight->Theta >= pLight->Phi) {
3721 DBG("Warning: clamping D3DLIGHT9.Theta\n");
3722 ff_state->light[Index].Theta = ff_state->light[Index].Phi;
3723 }
3724 return D3D_OK;
3725 }
3726
3727 HRESULT
nine_state_light_enable(struct nine_ff_state * ff_state,DWORD Index,BOOL Enable)3728 nine_state_light_enable(struct nine_ff_state *ff_state,
3729 DWORD Index, BOOL Enable)
3730 {
3731 unsigned i;
3732
3733 user_assert(Index < ff_state->num_lights, D3DERR_INVALIDCALL);
3734
3735 for (i = 0; i < ff_state->num_lights_active; ++i) {
3736 if (ff_state->active_light[i] == Index)
3737 break;
3738 }
3739
3740 if (Enable) {
3741 if (i < ff_state->num_lights_active)
3742 return D3D_OK;
3743 /* XXX wine thinks this should still succeed:
3744 */
3745 user_assert(i < NINE_MAX_LIGHTS_ACTIVE, D3DERR_INVALIDCALL);
3746
3747 ff_state->active_light[i] = Index;
3748 ff_state->num_lights_active++;
3749 } else {
3750 if (i == ff_state->num_lights_active)
3751 return D3D_OK;
3752 --ff_state->num_lights_active;
3753 for (; i < ff_state->num_lights_active; ++i)
3754 ff_state->active_light[i] = ff_state->active_light[i + 1];
3755 }
3756
3757 return D3D_OK;
3758 }
3759
3760 #define D3DRS_TO_STRING_CASE(n) case D3DRS_##n: return "D3DRS_"#n
nine_d3drs_to_string(DWORD State)3761 const char *nine_d3drs_to_string(DWORD State)
3762 {
3763 switch (State) {
3764 D3DRS_TO_STRING_CASE(ZENABLE);
3765 D3DRS_TO_STRING_CASE(FILLMODE);
3766 D3DRS_TO_STRING_CASE(SHADEMODE);
3767 D3DRS_TO_STRING_CASE(ZWRITEENABLE);
3768 D3DRS_TO_STRING_CASE(ALPHATESTENABLE);
3769 D3DRS_TO_STRING_CASE(LASTPIXEL);
3770 D3DRS_TO_STRING_CASE(SRCBLEND);
3771 D3DRS_TO_STRING_CASE(DESTBLEND);
3772 D3DRS_TO_STRING_CASE(CULLMODE);
3773 D3DRS_TO_STRING_CASE(ZFUNC);
3774 D3DRS_TO_STRING_CASE(ALPHAREF);
3775 D3DRS_TO_STRING_CASE(ALPHAFUNC);
3776 D3DRS_TO_STRING_CASE(DITHERENABLE);
3777 D3DRS_TO_STRING_CASE(ALPHABLENDENABLE);
3778 D3DRS_TO_STRING_CASE(FOGENABLE);
3779 D3DRS_TO_STRING_CASE(SPECULARENABLE);
3780 D3DRS_TO_STRING_CASE(FOGCOLOR);
3781 D3DRS_TO_STRING_CASE(FOGTABLEMODE);
3782 D3DRS_TO_STRING_CASE(FOGSTART);
3783 D3DRS_TO_STRING_CASE(FOGEND);
3784 D3DRS_TO_STRING_CASE(FOGDENSITY);
3785 D3DRS_TO_STRING_CASE(RANGEFOGENABLE);
3786 D3DRS_TO_STRING_CASE(STENCILENABLE);
3787 D3DRS_TO_STRING_CASE(STENCILFAIL);
3788 D3DRS_TO_STRING_CASE(STENCILZFAIL);
3789 D3DRS_TO_STRING_CASE(STENCILPASS);
3790 D3DRS_TO_STRING_CASE(STENCILFUNC);
3791 D3DRS_TO_STRING_CASE(STENCILREF);
3792 D3DRS_TO_STRING_CASE(STENCILMASK);
3793 D3DRS_TO_STRING_CASE(STENCILWRITEMASK);
3794 D3DRS_TO_STRING_CASE(TEXTUREFACTOR);
3795 D3DRS_TO_STRING_CASE(WRAP0);
3796 D3DRS_TO_STRING_CASE(WRAP1);
3797 D3DRS_TO_STRING_CASE(WRAP2);
3798 D3DRS_TO_STRING_CASE(WRAP3);
3799 D3DRS_TO_STRING_CASE(WRAP4);
3800 D3DRS_TO_STRING_CASE(WRAP5);
3801 D3DRS_TO_STRING_CASE(WRAP6);
3802 D3DRS_TO_STRING_CASE(WRAP7);
3803 D3DRS_TO_STRING_CASE(CLIPPING);
3804 D3DRS_TO_STRING_CASE(LIGHTING);
3805 D3DRS_TO_STRING_CASE(AMBIENT);
3806 D3DRS_TO_STRING_CASE(FOGVERTEXMODE);
3807 D3DRS_TO_STRING_CASE(COLORVERTEX);
3808 D3DRS_TO_STRING_CASE(LOCALVIEWER);
3809 D3DRS_TO_STRING_CASE(NORMALIZENORMALS);
3810 D3DRS_TO_STRING_CASE(DIFFUSEMATERIALSOURCE);
3811 D3DRS_TO_STRING_CASE(SPECULARMATERIALSOURCE);
3812 D3DRS_TO_STRING_CASE(AMBIENTMATERIALSOURCE);
3813 D3DRS_TO_STRING_CASE(EMISSIVEMATERIALSOURCE);
3814 D3DRS_TO_STRING_CASE(VERTEXBLEND);
3815 D3DRS_TO_STRING_CASE(CLIPPLANEENABLE);
3816 D3DRS_TO_STRING_CASE(POINTSIZE);
3817 D3DRS_TO_STRING_CASE(POINTSIZE_MIN);
3818 D3DRS_TO_STRING_CASE(POINTSPRITEENABLE);
3819 D3DRS_TO_STRING_CASE(POINTSCALEENABLE);
3820 D3DRS_TO_STRING_CASE(POINTSCALE_A);
3821 D3DRS_TO_STRING_CASE(POINTSCALE_B);
3822 D3DRS_TO_STRING_CASE(POINTSCALE_C);
3823 D3DRS_TO_STRING_CASE(MULTISAMPLEANTIALIAS);
3824 D3DRS_TO_STRING_CASE(MULTISAMPLEMASK);
3825 D3DRS_TO_STRING_CASE(PATCHEDGESTYLE);
3826 D3DRS_TO_STRING_CASE(DEBUGMONITORTOKEN);
3827 D3DRS_TO_STRING_CASE(POINTSIZE_MAX);
3828 D3DRS_TO_STRING_CASE(INDEXEDVERTEXBLENDENABLE);
3829 D3DRS_TO_STRING_CASE(COLORWRITEENABLE);
3830 D3DRS_TO_STRING_CASE(TWEENFACTOR);
3831 D3DRS_TO_STRING_CASE(BLENDOP);
3832 D3DRS_TO_STRING_CASE(POSITIONDEGREE);
3833 D3DRS_TO_STRING_CASE(NORMALDEGREE);
3834 D3DRS_TO_STRING_CASE(SCISSORTESTENABLE);
3835 D3DRS_TO_STRING_CASE(SLOPESCALEDEPTHBIAS);
3836 D3DRS_TO_STRING_CASE(ANTIALIASEDLINEENABLE);
3837 D3DRS_TO_STRING_CASE(MINTESSELLATIONLEVEL);
3838 D3DRS_TO_STRING_CASE(MAXTESSELLATIONLEVEL);
3839 D3DRS_TO_STRING_CASE(ADAPTIVETESS_X);
3840 D3DRS_TO_STRING_CASE(ADAPTIVETESS_Y);
3841 D3DRS_TO_STRING_CASE(ADAPTIVETESS_Z);
3842 D3DRS_TO_STRING_CASE(ADAPTIVETESS_W);
3843 D3DRS_TO_STRING_CASE(ENABLEADAPTIVETESSELLATION);
3844 D3DRS_TO_STRING_CASE(TWOSIDEDSTENCILMODE);
3845 D3DRS_TO_STRING_CASE(CCW_STENCILFAIL);
3846 D3DRS_TO_STRING_CASE(CCW_STENCILZFAIL);
3847 D3DRS_TO_STRING_CASE(CCW_STENCILPASS);
3848 D3DRS_TO_STRING_CASE(CCW_STENCILFUNC);
3849 D3DRS_TO_STRING_CASE(COLORWRITEENABLE1);
3850 D3DRS_TO_STRING_CASE(COLORWRITEENABLE2);
3851 D3DRS_TO_STRING_CASE(COLORWRITEENABLE3);
3852 D3DRS_TO_STRING_CASE(BLENDFACTOR);
3853 D3DRS_TO_STRING_CASE(SRGBWRITEENABLE);
3854 D3DRS_TO_STRING_CASE(DEPTHBIAS);
3855 D3DRS_TO_STRING_CASE(WRAP8);
3856 D3DRS_TO_STRING_CASE(WRAP9);
3857 D3DRS_TO_STRING_CASE(WRAP10);
3858 D3DRS_TO_STRING_CASE(WRAP11);
3859 D3DRS_TO_STRING_CASE(WRAP12);
3860 D3DRS_TO_STRING_CASE(WRAP13);
3861 D3DRS_TO_STRING_CASE(WRAP14);
3862 D3DRS_TO_STRING_CASE(WRAP15);
3863 D3DRS_TO_STRING_CASE(SEPARATEALPHABLENDENABLE);
3864 D3DRS_TO_STRING_CASE(SRCBLENDALPHA);
3865 D3DRS_TO_STRING_CASE(DESTBLENDALPHA);
3866 D3DRS_TO_STRING_CASE(BLENDOPALPHA);
3867 default:
3868 return "(invalid)";
3869 }
3870 }
3871