summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/svga/svga_state_fs.c
blob: 8f419fa69ac8a320b3b979fe525a558113b6cbd8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
/**********************************************************
 * Copyright 2008-2009 VMware, Inc.  All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person
 * obtaining a copy of this software and associated documentation
 * files (the "Software"), to deal in the Software without
 * restriction, including without limitation the rights to use, copy,
 * modify, merge, publish, distribute, sublicense, and/or sell copies
 * of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 **********************************************************/

#include "util/u_inlines.h"
#include "pipe/p_defines.h"
#include "util/u_math.h"
#include "util/u_memory.h"
#include "util/u_bitmask.h"
#include "tgsi/tgsi_ureg.h"

#include "svga_context.h"
#include "svga_state.h"
#include "svga_cmd.h"
#include "svga_shader.h"
#include "svga_resource_texture.h"
#include "svga_tgsi.h"

#include "svga_hw_reg.h"



static INLINE int
compare_fs_keys(const struct svga_fs_compile_key *a,
                const struct svga_fs_compile_key *b)
{
   unsigned keysize_a = svga_fs_key_size( a );
   unsigned keysize_b = svga_fs_key_size( b );

   if (keysize_a != keysize_b) {
      return (int)(keysize_a - keysize_b);
   }
   return memcmp( a, b, keysize_a );
}


/** Search for a fragment shader variant */
static struct svga_shader_variant *
search_fs_key(const struct svga_fragment_shader *fs,
              const struct svga_fs_compile_key *key)
{
   struct svga_shader_variant *variant = fs->base.variants;

   assert(key);

   for ( ; variant; variant = variant->next) {
      if (compare_fs_keys( key, &variant->key.fkey ) == 0)
         return variant;
   }
   
   return NULL;
}


/**
 * If we fail to compile a fragment shader (because it uses too many
 * registers, for example) we'll use a dummy/fallback shader that
 * simply emits a constant color (red for debug, black for release).
 * We hit this with the Unigine/Heaven demo when Shaders = High.
 * With black, the demo still looks good.
 */
static const struct tgsi_token *
get_dummy_fragment_shader(void)
{
#ifdef DEBUG
   static const float color[4] = { 1.0, 0.0, 0.0, 0.0 }; /* red */
#else
   static const float color[4] = { 0.0, 0.0, 0.0, 0.0 }; /* black */
#endif
   struct ureg_program *ureg;
   const struct tgsi_token *tokens;
   struct ureg_src src;
   struct ureg_dst dst;
   unsigned num_tokens;

   ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
   if (!ureg)
      return NULL;

   dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
   src = ureg_DECL_immediate(ureg, color, 4);
   ureg_MOV(ureg, dst, src);
   ureg_END(ureg);

   tokens = ureg_get_tokens(ureg, &num_tokens);

   ureg_destroy(ureg);

   return tokens;
}


/**
 * Replace the given shader's instruction with a simple constant-color
 * shader.  We use this when normal shader translation fails.
 */
static struct svga_shader_variant *
get_compiled_dummy_shader(struct svga_fragment_shader *fs,
                          const struct svga_fs_compile_key *key)
{
   const struct tgsi_token *dummy = get_dummy_fragment_shader();
   struct svga_shader_variant *variant;

   if (!dummy) {
      return NULL;
   }

   FREE((void *) fs->base.tokens);
   fs->base.tokens = dummy;

   variant = svga_translate_fragment_program(fs, key);
   return variant;
}


/**
 * Translate TGSI shader into an svga shader variant.
 */
static enum pipe_error
compile_fs(struct svga_context *svga,
           struct svga_fragment_shader *fs,
           const struct svga_fs_compile_key *key,
           struct svga_shader_variant **out_variant)
{
   struct svga_shader_variant *variant;
   enum pipe_error ret = PIPE_ERROR;

   variant = svga_translate_fragment_program( fs, key );
   if (variant == NULL) {
      debug_printf("Failed to compile fragment shader,"
                   " using dummy shader instead.\n");
      variant = get_compiled_dummy_shader(fs, key);
      if (!variant) {
         ret = PIPE_ERROR;
         goto fail;
      }
   }

   if (svga_shader_too_large(svga, variant)) {
      /* too big, use dummy shader */
      debug_printf("Shader too large (%lu bytes),"
                   " using dummy shader instead.\n",
                   (unsigned long ) variant->nr_tokens * sizeof(variant->tokens[0]));
      variant = get_compiled_dummy_shader(fs, key);
      if (!variant) {
         ret = PIPE_ERROR;
         goto fail;
      }
   }

   ret = svga_define_shader(svga, SVGA3D_SHADERTYPE_PS, variant);
   if (ret != PIPE_OK)
      goto fail;

   *out_variant = variant;

   /* insert variants at head of linked list */
   variant->next = fs->base.variants;
   fs->base.variants = variant;

   return PIPE_OK;

fail:
   if (variant) {
      svga_destroy_shader_variant(svga, SVGA3D_SHADERTYPE_PS, variant);
   }
   return ret;
}


/* SVGA_NEW_TEXTURE_BINDING
 * SVGA_NEW_RAST
 * SVGA_NEW_NEED_SWTNL
 * SVGA_NEW_SAMPLER
 */
static enum pipe_error
make_fs_key(const struct svga_context *svga,
            struct svga_fragment_shader *fs,
            struct svga_fs_compile_key *key)
{
   unsigned i;
   int idx = 0;

   memset(key, 0, sizeof *key);

   /* Only need fragment shader fixup for twoside lighting if doing
    * hwtnl.  Otherwise the draw module does the whole job for us.
    *
    * SVGA_NEW_SWTNL
    */
   if (!svga->state.sw.need_swtnl) {
      /* SVGA_NEW_RAST
       */
      key->light_twoside = svga->curr.rast->templ.light_twoside;
      key->front_ccw = svga->curr.rast->templ.front_ccw;
   }

   /* The blend workaround for simulating logicop xor behaviour
    * requires that the incoming fragment color be white.  This change
    * achieves that by creating a variant of the current fragment
    * shader that overrides all output colors with 1,1,1,1
    *   
    * This will work for most shaders, including those containing
    * TEXKIL and/or depth-write.  However, it will break on the
    * combination of xor-logicop plus alphatest.
    *
    * Ultimately, we could implement alphatest in the shader using
    * texkil prior to overriding the outgoing fragment color.
    *   
    * SVGA_NEW_BLEND
    */
   if (svga->curr.blend->need_white_fragments) {
      key->white_fragments = 1;
   }

#ifdef DEBUG
   /*
    * We expect a consistent set of samplers and sampler views.
    * Do some debug checks/warnings here.
    */
   {
      static boolean warned = FALSE;
      unsigned i, n = MAX2(svga->curr.num_sampler_views,
                           svga->curr.num_samplers);
      /* Only warn once to prevent too much debug output */
      if (!warned) {
         if (svga->curr.num_sampler_views != svga->curr.num_samplers) {
            debug_printf("svga: mismatched number of sampler views (%u) "
                         "vs. samplers (%u)\n",
                         svga->curr.num_sampler_views,
                         svga->curr.num_samplers);
         }
         for (i = 0; i < n; i++) {
            if ((svga->curr.sampler_views[i] == NULL) !=
                (svga->curr.sampler[i] == NULL))
               debug_printf("sampler_view[%u] = %p but sampler[%u] = %p\n",
                            i, svga->curr.sampler_views[i],
                            i, svga->curr.sampler[i]);
         }
         warned = TRUE;
      }
   }
#endif

   /* XXX: want to limit this to the textures that the shader actually
    * refers to.
    *
    * SVGA_NEW_TEXTURE_BINDING | SVGA_NEW_SAMPLER
    */
   for (i = 0; i < svga->curr.num_sampler_views; i++) {
      if (svga->curr.sampler_views[i] && svga->curr.sampler[i]) {
         assert(svga->curr.sampler_views[i]->texture);
         key->tex[i].texture_target = svga->curr.sampler_views[i]->texture->target;
         if (!svga->curr.sampler[i]->normalized_coords) {
            key->tex[i].width_height_idx = idx++;
            key->tex[i].unnormalized = TRUE;
            ++key->num_unnormalized_coords;
         }

         key->tex[i].swizzle_r = svga->curr.sampler_views[i]->swizzle_r;
         key->tex[i].swizzle_g = svga->curr.sampler_views[i]->swizzle_g;
         key->tex[i].swizzle_b = svga->curr.sampler_views[i]->swizzle_b;
         key->tex[i].swizzle_a = svga->curr.sampler_views[i]->swizzle_a;
      }
   }
   key->num_textures = svga->curr.num_sampler_views;

   idx = 0;
   for (i = 0; i < svga->curr.num_samplers; ++i) {
      if (svga->curr.sampler_views[i] && svga->curr.sampler[i]) {
         struct pipe_resource *tex = svga->curr.sampler_views[i]->texture;
         struct svga_texture *stex = svga_texture(tex);
         SVGA3dSurfaceFormat format = stex->key.format;

         if (format == SVGA3D_Z_D16 ||
             format == SVGA3D_Z_D24X8 ||
             format == SVGA3D_Z_D24S8) {
            /* If we're sampling from a SVGA3D_Z_D16, SVGA3D_Z_D24X8,
             * or SVGA3D_Z_D24S8 surface, we'll automatically get
             * shadow comparison.  But we only get LEQUAL mode.
             * Set TEX_COMPARE_NONE here so we don't emit the extra FS
             * code for shadow comparison.
             */
            key->tex[i].compare_mode = PIPE_TEX_COMPARE_NONE;
            key->tex[i].compare_func = PIPE_FUNC_NEVER;
            /* These depth formats _only_ support comparison mode and
             * not ordinary sampling so warn if the later is expected.
             */
            if (svga->curr.sampler[i]->compare_mode !=
                PIPE_TEX_COMPARE_R_TO_TEXTURE) {
               debug_warn_once("Unsupported shadow compare mode");
            }                   
            /* The only supported comparison mode is LEQUAL */
            if (svga->curr.sampler[i]->compare_func != PIPE_FUNC_LEQUAL) {
               debug_warn_once("Unsupported shadow compare function");
            }
         }
         else {
            /* For other texture formats, just use the compare func/mode
             * as-is.  Should be no-ops for color textures.  For depth
             * textures, we do not get automatic depth compare.  We have
             * to do it ourselves in the shader.  And we don't get PCF.
             */
            key->tex[i].compare_mode = svga->curr.sampler[i]->compare_mode;
            key->tex[i].compare_func = svga->curr.sampler[i]->compare_func;
         }
      }
   }

   /* sprite coord gen state */
   for (i = 0; i < svga->curr.num_samplers; ++i) {
      key->tex[i].sprite_texgen =
         svga->curr.rast->templ.sprite_coord_enable & (1 << i);
   }

   key->sprite_origin_lower_left = (svga->curr.rast->templ.sprite_coord_mode
                                    == PIPE_SPRITE_COORD_LOWER_LEFT);

   /* SVGA_NEW_FRAME_BUFFER */
   if (fs->base.info.color0_writes_all_cbufs) {
      /* Replicate color0 output to N colorbuffers */
      key->write_color0_to_n_cbufs = svga->curr.framebuffer.nr_cbufs;
   }

   return PIPE_OK;
}


/**
 * svga_reemit_fs_bindings - Reemit the fragment shader bindings
 */
enum pipe_error
svga_reemit_fs_bindings(struct svga_context *svga)
{
   enum pipe_error ret;

   assert(svga->rebind.fs);
   assert(svga_have_gb_objects(svga));

   if (!svga->state.hw_draw.fs)
      return PIPE_OK;

   ret = SVGA3D_SetGBShader(svga->swc, SVGA3D_SHADERTYPE_PS,
                            svga->state.hw_draw.fs->gb_shader);
   if (ret != PIPE_OK)
      return ret;

   svga->rebind.fs = FALSE;
   return PIPE_OK;
}



static enum pipe_error
emit_hw_fs(struct svga_context *svga, unsigned dirty)
{
   struct svga_shader_variant *variant = NULL;
   enum pipe_error ret = PIPE_OK;
   struct svga_fragment_shader *fs = svga->curr.fs;
   struct svga_fs_compile_key key;

   /* SVGA_NEW_BLEND
    * SVGA_NEW_TEXTURE_BINDING
    * SVGA_NEW_RAST
    * SVGA_NEW_NEED_SWTNL
    * SVGA_NEW_SAMPLER
    * SVGA_NEW_FRAME_BUFFER
    */
   ret = make_fs_key( svga, fs, &key );
   if (ret != PIPE_OK)
      return ret;

   variant = search_fs_key( fs, &key );
   if (!variant) {
      ret = compile_fs( svga, fs, &key, &variant );
      if (ret != PIPE_OK)
         return ret;
   }

   assert(variant);

   if (variant != svga->state.hw_draw.fs) {
      if (svga_have_gb_objects(svga)) {
         /*
          * Bind is necessary here only because pipebuffer_fenced may move
          * the shader contents around....
          */
         ret = SVGA3D_BindGBShader(svga->swc, variant->gb_shader);
         if (ret != PIPE_OK)
            return ret;

         ret = SVGA3D_SetGBShader(svga->swc, SVGA3D_SHADERTYPE_PS,
                                  variant->gb_shader);
         if (ret != PIPE_OK)
            return ret;

         svga->rebind.fs = FALSE;
      }
      else {
         ret = SVGA3D_SetShader(svga->swc, SVGA3D_SHADERTYPE_PS, variant->id);
         if (ret != PIPE_OK)
            return ret;
      }

      svga->dirty |= SVGA_NEW_FS_VARIANT;
      svga->state.hw_draw.fs = variant;      
   }

   return PIPE_OK;
}

struct svga_tracked_state svga_hw_fs = 
{
   "fragment shader (hwtnl)",
   (SVGA_NEW_FS |
    SVGA_NEW_TEXTURE_BINDING |
    SVGA_NEW_NEED_SWTNL |
    SVGA_NEW_RAST |
    SVGA_NEW_SAMPLER |
    SVGA_NEW_FRAME_BUFFER |
    SVGA_NEW_BLEND),
   emit_hw_fs
};