summaryrefslogtreecommitdiff
path: root/src/mesa/pipe/softpipe/sp_quad_fs.c
blob: b5d7dfca1ca4b081f93d530342677327d6af2525 (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
/**************************************************************************
 * 
 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
 * 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, sub license, 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 (including the
 * next paragraph) 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 NON-INFRINGEMENT.
 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
 * 
 **************************************************************************/

/* Vertices are just an array of floats, with all the attributes
 * packed.  We currently assume a layout like:
 *
 * attr[0][0..3] - window position
 * attr[1..n][0..3] - remaining attributes.
 *
 * Attributes are assumed to be 4 floats wide but are packed so that
 * all the enabled attributes run contiguously.
 */

#include "pipe/p_util.h"
#include "pipe/p_defines.h"
#include "pipe/p_shader_tokens.h"

#include "x86/rtasm/x86sse.h"

#ifdef MESA_LLVM
#include "pipe/llvm/gallivm.h"
#endif

#include "sp_context.h"
#include "sp_state.h"
#include "sp_headers.h"
#include "sp_quad.h"
#include "sp_texture.h"
#include "sp_tex_sample.h"


struct quad_shade_stage
{
   struct quad_stage stage;
   struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS];
   struct tgsi_exec_machine machine;
   struct tgsi_exec_vector *inputs, *outputs;
   int colorOutSlot, depthOutSlot;
#ifdef MESA_LLVM
   struct gallivm_prog *llvm_prog;
#endif
};


/** cast wrapper */
static INLINE struct quad_shade_stage *
quad_shade_stage(struct quad_stage *qs)
{
   return (struct quad_shade_stage *) qs;
}


/**
 * Compute quad X,Y,Z,W for the four fragments in a quad.
 * Note that we only need to "compute" X and Y for the upper-left fragment.
 * We could do less work if we're not depth testing, or there's no
 * perspective-corrected attributes, but that's seldom.
 */
static void
setup_pos_vector(const struct tgsi_interp_coef *coef,
                 float x, float y,
                 struct tgsi_exec_vector *quadpos)
{
   uint chan;
   /* do X */
   quadpos->xyzw[0].f[0] = x;
   /* do Y */
   quadpos->xyzw[1].f[0] = y;
   /* do Z and W for all fragments in the quad */
   for (chan = 2; chan < 4; chan++) {
      const float dadx = coef->dadx[chan];
      const float dady = coef->dady[chan];
      const float a0 = coef->a0[chan] + dadx * x + dady * y;
      quadpos->xyzw[chan].f[0] = a0;
      quadpos->xyzw[chan].f[1] = a0 + dadx;
      quadpos->xyzw[chan].f[2] = a0 + dady;
      quadpos->xyzw[chan].f[3] = a0 + dadx + dady;
   }
}


typedef void (XSTDCALL *codegen_function)(
   const struct tgsi_exec_vector *input,
   struct tgsi_exec_vector *output,
   float (*constant)[4],
   struct tgsi_exec_vector *temporary,
   const struct tgsi_interp_coef *coef
#if 0
   ,const struct tgsi_exec_vector *quadPos
#endif
 );


/**
 * Execute fragment shader for the four fragments in the quad.
 */
static void
shade_quad(
   struct quad_stage *qs,
   struct quad_header *quad )
{
   struct quad_shade_stage *qss = quad_shade_stage( qs );
   struct softpipe_context *softpipe = qs->softpipe;
   struct tgsi_exec_machine *machine = &qss->machine;

   /* Consts do not require 16 byte alignment. */
   machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT];

   machine->InterpCoefs = quad->coef;

   /* Compute X, Y, Z, W vals for this quad */
   setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos);

   /* run shader */
#if defined(__i386__) || defined(__386__)
   if( softpipe->use_sse ) {
      codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program );
      func(
         machine->Inputs,
         machine->Outputs,
         machine->Consts,
         machine->Temps,
         machine->InterpCoefs
#if 0
         ,machine->QuadPos
#endif
           );
      quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]);
   }
   else
#endif
   {
      quad->mask &= tgsi_exec_machine_run( machine );
   }

   /* store result color */
   if (qss->colorOutSlot >= 0) {
      /* XXX need to handle multiple color outputs someday */
      assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot]
             == TGSI_SEMANTIC_COLOR);
      memcpy(
             quad->outputs.color,
             &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0],
             sizeof( quad->outputs.color ) );
   }

   /*
    * XXX the following code for updating quad->outputs.depth
    * isn't really needed if we did early z testing.
    */

   /* store result Z */
   if (qss->depthOutSlot >= 0) {
      /* output[slot] is new Z */
      uint i;
      for (i = 0; i < 4; i++) {
         quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i];
      }
   }
   else {
      /* copy input Z (which was interpolated by the executor) to output Z */
      uint i;
      for (i = 0; i < 4; i++) {
         quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i];
         /* XXX not sure the above line is always correct.  The following
          * might be better:
         quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i];
          */
      }
   }

   /* shader may cull fragments */
   if( quad->mask ) {
      qs->next->run( qs->next, quad );
   }
}

#ifdef MESA_LLVM
#define DLLVM 0
static void
shade_quad_llvm(struct quad_stage *qs,
                struct quad_header *quad)
{
   struct quad_shade_stage *qss = quad_shade_stage(qs);
   struct softpipe_context *softpipe = qs->softpipe;
   float dests[4][16][4] ALIGN16_ATTRIB;
   float inputs[4][16][4] ALIGN16_ATTRIB;
   const float fx = (float) quad->x0;
   const float fy = (float) quad->y0;
   struct gallivm_prog *llvm = qss->llvm_prog;

   inputs[0][0][0] = fx;
   inputs[1][0][0] = fx + 1.0f;
   inputs[2][0][0] = fx;
   inputs[3][0][0] = fx + 1.0f;

   inputs[0][0][1] = fy;
   inputs[1][0][1] = fy;
   inputs[2][0][1] = fy + 1.0f;
   inputs[3][0][1] = fy + 1.0f;
#if DLLVM
   debug_printf("MASK = %d\n", quad->mask);
#endif
   gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef);
#if DLLVM
   for (int i = 0; i < 4; ++i) {
      for (int j = 0; j < 2; ++j) {
         debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, 
                inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]);
      }
   }
#endif

   quad->mask &=
      gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs,
                                   softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
                                   qss->samplers);
#if DLLVM
   debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n",
          dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], 
          dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]);
#endif

   /* store result color */
   if (qss->colorOutSlot >= 0) {
      unsigned i;
      /* XXX need to handle multiple color outputs someday */
      assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot]
             == TGSI_SEMANTIC_COLOR);
      for (i = 0; i < QUAD_SIZE; ++i) {
         quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0];
         quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1];
         quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2];
         quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3];
      }
   }
#if DLLVM
   for (int i = 0; i < QUAD_SIZE; ++i) {
      debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot,
             quad->outputs.color[0][i],
             quad->outputs.color[1][i],
             quad->outputs.color[2][i],
             quad->outputs.color[3][i]);
   }
#endif

   /* store result Z */
   if (qss->depthOutSlot >= 0) {
      /* output[slot] is new Z */
      uint i;
      for (i = 0; i < 4; i++) {
         quad->outputs.depth[i] = dests[i][0][2];
      }
   }
   else {
      /* copy input Z (which was interpolated by the executor) to output Z */
      uint i;
      for (i = 0; i < 4; i++) {
         quad->outputs.depth[i] = inputs[i][0][2];
      }
   }
#if DLLVM
   debug_printf("D [%f, %f, %f, %f] mask = %d\n",
             quad->outputs.depth[0],
             quad->outputs.depth[1],
             quad->outputs.depth[2],
             quad->outputs.depth[3], quad->mask);
#endif

   /* shader may cull fragments */
   if( quad->mask ) {
      qs->next->run( qs->next, quad );
   }
}
#endif /*MESA_LLVM*/

/**
 * Per-primitive (or per-begin?) setup
 */
static void shade_begin(struct quad_stage *qs)
{
   struct quad_shade_stage *qss = quad_shade_stage(qs);
   struct softpipe_context *softpipe = qs->softpipe;
   unsigned i;

   /* set TGSI sampler state that varies */
   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
      qss->samplers[i].state = softpipe->sampler[i];
      qss->samplers[i].texture = &softpipe->texture[i]->base;
   }

#ifdef MESA_LLVM
   qss->llvm_prog = softpipe->fs->llvm_prog;
#endif
   /* XXX only do this if the fragment shader changes... */
   tgsi_exec_machine_init(&qss->machine,
                          softpipe->fs->shader.tokens,
                          PIPE_MAX_SAMPLERS,
                          qss->samplers );

   /* find output slots for depth, color */
   qss->colorOutSlot = -1;
   qss->depthOutSlot = -1;
   for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) {
      switch (qss->stage.softpipe->fs->shader.output_semantic_name[i]) {
      case TGSI_SEMANTIC_POSITION:
         qss->depthOutSlot = i;
         break;
      case TGSI_SEMANTIC_COLOR:
         qss->colorOutSlot = i;
         break;
      }
   }

   qs->next->begin(qs->next);
}


static void shade_destroy(struct quad_stage *qs)
{
   struct quad_shade_stage *qss = (struct quad_shade_stage *) qs;

   tgsi_exec_machine_free_data(&qss->machine);
   FREE( qss->inputs );
   FREE( qss->outputs );
   FREE( qs );
}


struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
{
   struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage);
   uint i;

   /* allocate storage for program inputs/outputs, aligned to 16 bytes */
   qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16);
   qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16);
   qss->machine.Inputs = align16(qss->inputs);
   qss->machine.Outputs = align16(qss->outputs);

   qss->stage.softpipe = softpipe;
   qss->stage.begin = shade_begin;
#ifdef MESA_LLVM
   /* disable until ported to accept
    * x/y and soa layout
   qss->stage.run = shade_quad_llvm;
   */
   softpipe->use_sse = FALSE;
   qss->stage.run = shade_quad;
#else
   qss->stage.run = shade_quad;
#endif
   qss->stage.destroy = shade_destroy;

   /* set TGSI sampler state that's constant */
   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
      assert(softpipe->tex_cache[i]);
      qss->samplers[i].get_samples = sp_get_samples;
      qss->samplers[i].pipe = &softpipe->pipe;
      qss->samplers[i].cache = softpipe->tex_cache[i];
   }

   return &qss->stage;
}