mesa: switch texel fetch functions from GLchan to GLfloat
[mesa.git] / src / mesa / tnl / t_vb_program.c
1 /*
2 * Mesa 3-D graphics library
3 * Version: 7.1
4 *
5 * Copyright (C) 1999-2007 Brian Paul All Rights Reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 */
24
25
26 /**
27 * \file tnl/t_vb_program.c
28 * \brief Pipeline stage for executing vertex programs.
29 * \author Brian Paul, Keith Whitwell
30 */
31
32
33 #include "main/glheader.h"
34 #include "main/colormac.h"
35 #include "main/context.h"
36 #include "main/macros.h"
37 #include "main/imports.h"
38 #include "shader/prog_instruction.h"
39 #include "shader/prog_statevars.h"
40 #include "shader/prog_execute.h"
41 #include "swrast/s_context.h"
42 #include "swrast/s_texfilter.h"
43
44 #include "tnl/tnl.h"
45 #include "tnl/t_context.h"
46 #include "tnl/t_pipeline.h"
47
48
49
50 /*!
51 * Private storage for the vertex program pipeline stage.
52 */
53 struct vp_stage_data {
54 /** The results of running the vertex program go into these arrays. */
55 GLvector4f results[VERT_RESULT_MAX];
56
57 GLvector4f ndcCoords; /**< normalized device coords */
58 GLubyte *clipmask; /**< clip flags */
59 GLubyte ormask, andmask; /**< for clipping */
60 };
61
62
63 #define VP_STAGE_DATA(stage) ((struct vp_stage_data *)(stage->privatePtr))
64
65
66 static void
67 userclip( GLcontext *ctx,
68 GLvector4f *clip,
69 GLubyte *clipmask,
70 GLubyte *clipormask,
71 GLubyte *clipandmask )
72 {
73 GLuint p;
74
75 for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
76 if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
77 GLuint nr, i;
78 const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
79 const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
80 const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
81 const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
82 GLfloat *coord = (GLfloat *)clip->data;
83 GLuint stride = clip->stride;
84 GLuint count = clip->count;
85
86 for (nr = 0, i = 0 ; i < count ; i++) {
87 GLfloat dp = (coord[0] * a +
88 coord[1] * b +
89 coord[2] * c +
90 coord[3] * d);
91
92 if (dp < 0) {
93 nr++;
94 clipmask[i] |= CLIP_USER_BIT;
95 }
96
97 STRIDE_F(coord, stride);
98 }
99
100 if (nr > 0) {
101 *clipormask |= CLIP_USER_BIT;
102 if (nr == count) {
103 *clipandmask |= CLIP_USER_BIT;
104 return;
105 }
106 }
107 }
108 }
109 }
110
111
112 static GLboolean
113 do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store)
114 {
115 TNLcontext *tnl = TNL_CONTEXT(ctx);
116 struct vertex_buffer *VB = &tnl->vb;
117 /* Cliptest and perspective divide. Clip functions must clear
118 * the clipmask.
119 */
120 store->ormask = 0;
121 store->andmask = CLIP_FRUSTUM_BITS;
122
123 if (tnl->NeedNdcCoords) {
124 VB->NdcPtr =
125 _mesa_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
126 &store->ndcCoords,
127 store->clipmask,
128 &store->ormask,
129 &store->andmask );
130 }
131 else {
132 VB->NdcPtr = NULL;
133 _mesa_clip_np_tab[VB->ClipPtr->size]( VB->ClipPtr,
134 NULL,
135 store->clipmask,
136 &store->ormask,
137 &store->andmask );
138 }
139
140 if (store->andmask) {
141 /* All vertices are outside the frustum */
142 return GL_FALSE;
143 }
144
145 /* Test userclip planes. This contributes to VB->ClipMask.
146 */
147 /** XXX NEW_SLANG _Enabled ??? */
148 if (ctx->Transform.ClipPlanesEnabled && (!ctx->VertexProgram._Enabled ||
149 ctx->VertexProgram.Current->IsPositionInvariant)) {
150 userclip( ctx,
151 VB->ClipPtr,
152 store->clipmask,
153 &store->ormask,
154 &store->andmask );
155
156 if (store->andmask) {
157 return GL_FALSE;
158 }
159 }
160
161 VB->ClipAndMask = store->andmask;
162 VB->ClipOrMask = store->ormask;
163 VB->ClipMask = store->clipmask;
164
165 return GL_TRUE;
166 }
167
168
169 /**
170 * XXX the texture sampling code in this module is a bit of a hack.
171 * The texture sampling code is in swrast, though it doesn't have any
172 * real dependencies on the rest of swrast. It should probably be
173 * moved into main/ someday.
174 */
175 static void
176 vp_fetch_texel(GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda,
177 GLuint unit, GLfloat color[4])
178 {
179 SWcontext *swrast = SWRAST_CONTEXT(ctx);
180
181 /* XXX use a float-valued TextureSample routine here!!! */
182 swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current,
183 1, (const GLfloat (*)[4]) texcoord,
184 &lambda, (GLfloat (*)[4]) color);
185 }
186
187
188 /**
189 * Called via ctx->Driver.ProgramStringNotify() after a new vertex program
190 * string has been parsed.
191 */
192 void
193 _tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program)
194 {
195 /* No-op.
196 * If we had derived anything from the program that was private to this
197 * stage we'd recompute/validate it here.
198 */
199 }
200
201
202 /**
203 * Initialize virtual machine state prior to executing vertex program.
204 */
205 static void
206 init_machine(GLcontext *ctx, struct gl_program_machine *machine)
207 {
208 /* Input registers get initialized from the current vertex attribs */
209 MEMCPY(machine->VertAttribs, ctx->Current.Attrib,
210 MAX_VERTEX_PROGRAM_ATTRIBS * 4 * sizeof(GLfloat));
211
212 if (ctx->VertexProgram._Current->IsNVProgram) {
213 GLuint i;
214 /* Output/result regs are initialized to [0,0,0,1] */
215 for (i = 0; i < MAX_NV_VERTEX_PROGRAM_OUTPUTS; i++) {
216 ASSIGN_4V(machine->Outputs[i], 0.0F, 0.0F, 0.0F, 1.0F);
217 }
218 /* Temp regs are initialized to [0,0,0,0] */
219 for (i = 0; i < MAX_NV_VERTEX_PROGRAM_TEMPS; i++) {
220 ASSIGN_4V(machine->Temporaries[i], 0.0F, 0.0F, 0.0F, 0.0F);
221 }
222 for (i = 0; i < MAX_VERTEX_PROGRAM_ADDRESS_REGS; i++) {
223 ASSIGN_4V(machine->AddressReg[i], 0, 0, 0, 0);
224 }
225 }
226
227 machine->NumDeriv = 0;
228
229 /* init condition codes */
230 machine->CondCodes[0] = COND_EQ;
231 machine->CondCodes[1] = COND_EQ;
232 machine->CondCodes[2] = COND_EQ;
233 machine->CondCodes[3] = COND_EQ;
234
235 /* init call stack */
236 machine->StackDepth = 0;
237
238 machine->FetchTexelLod = vp_fetch_texel;
239 machine->FetchTexelDeriv = NULL; /* not used by vertex programs */
240
241 machine->Samplers = ctx->VertexProgram._Current->Base.SamplerUnits;
242 }
243
244
245 /**
246 * Map the texture images which the vertex program will access (if any).
247 */
248 static void
249 map_textures(GLcontext *ctx, const struct gl_vertex_program *vp)
250 {
251 GLuint u;
252
253 if (!ctx->Driver.MapTexture)
254 return;
255
256 for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) {
257 if (vp->Base.TexturesUsed[u]) {
258 /* Note: _Current *should* correspond to the target indicated
259 * in TexturesUsed[u].
260 */
261 ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[u]._Current);
262 }
263 }
264 }
265
266
267 /**
268 * Unmap the texture images which were used by the vertex program (if any).
269 */
270 static void
271 unmap_textures(GLcontext *ctx, const struct gl_vertex_program *vp)
272 {
273 GLuint u;
274
275 if (!ctx->Driver.MapTexture)
276 return;
277
278 for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) {
279 if (vp->Base.TexturesUsed[u]) {
280 /* Note: _Current *should* correspond to the target indicated
281 * in TexturesUsed[u].
282 */
283 ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[u]._Current);
284 }
285 }
286 }
287
288
289 /**
290 * This function executes vertex programs
291 */
292 static GLboolean
293 run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
294 {
295 TNLcontext *tnl = TNL_CONTEXT(ctx);
296 struct vp_stage_data *store = VP_STAGE_DATA(stage);
297 struct vertex_buffer *VB = &tnl->vb;
298 struct gl_vertex_program *program = ctx->VertexProgram._Current;
299 struct gl_program_machine machine;
300 GLuint outputs[VERT_RESULT_MAX], numOutputs;
301 GLuint i, j;
302
303 if (!program)
304 return GL_TRUE;
305
306 if (program->IsNVProgram) {
307 _mesa_load_tracked_matrices(ctx);
308 }
309 else {
310 /* ARB program or vertex shader */
311 _mesa_load_state_parameters(ctx, program->Base.Parameters);
312 }
313
314 /* make list of outputs to save some time below */
315 numOutputs = 0;
316 for (i = 0; i < VERT_RESULT_MAX; i++) {
317 if (program->Base.OutputsWritten & (1 << i)) {
318 outputs[numOutputs++] = i;
319 }
320 }
321
322 map_textures(ctx, program);
323
324 for (i = 0; i < VB->Count; i++) {
325 GLuint attr;
326
327 init_machine(ctx, &machine);
328
329 #if 0
330 printf("Input %d: %f, %f, %f, %f\n", i,
331 VB->AttribPtr[0]->data[i][0],
332 VB->AttribPtr[0]->data[i][1],
333 VB->AttribPtr[0]->data[i][2],
334 VB->AttribPtr[0]->data[i][3]);
335 printf(" color: %f, %f, %f, %f\n",
336 VB->AttribPtr[3]->data[i][0],
337 VB->AttribPtr[3]->data[i][1],
338 VB->AttribPtr[3]->data[i][2],
339 VB->AttribPtr[3]->data[i][3]);
340 printf(" normal: %f, %f, %f, %f\n",
341 VB->AttribPtr[2]->data[i][0],
342 VB->AttribPtr[2]->data[i][1],
343 VB->AttribPtr[2]->data[i][2],
344 VB->AttribPtr[2]->data[i][3]);
345 #endif
346
347 /* the vertex array case */
348 for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
349 if (program->Base.InputsRead & (1 << attr)) {
350 const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data;
351 const GLuint size = VB->AttribPtr[attr]->size;
352 const GLuint stride = VB->AttribPtr[attr]->stride;
353 const GLfloat *data = (GLfloat *) (ptr + stride * i);
354 COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
355 }
356 }
357
358 /* execute the program */
359 _mesa_execute_program(ctx, &program->Base, &machine);
360
361 /* copy the output registers into the VB->attribs arrays */
362 for (j = 0; j < numOutputs; j++) {
363 const GLuint attr = outputs[j];
364 COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
365 }
366 #if 0
367 printf("HPOS: %f %f %f %f\n",
368 machine.Outputs[0][0],
369 machine.Outputs[0][1],
370 machine.Outputs[0][2],
371 machine.Outputs[0][3]);
372 #endif
373 }
374
375 unmap_textures(ctx, program);
376
377 /* Fixup fog and point size results if needed */
378 if (program->IsNVProgram) {
379 if (ctx->Fog.Enabled &&
380 (program->Base.OutputsWritten & (1 << VERT_RESULT_FOGC)) == 0) {
381 for (i = 0; i < VB->Count; i++) {
382 store->results[VERT_RESULT_FOGC].data[i][0] = 1.0;
383 }
384 }
385
386 if (ctx->VertexProgram.PointSizeEnabled &&
387 (program->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) == 0) {
388 for (i = 0; i < VB->Count; i++) {
389 store->results[VERT_RESULT_PSIZ].data[i][0] = ctx->Point.Size;
390 }
391 }
392 }
393
394 if (program->IsPositionInvariant) {
395 /* We need the exact same transform as in the fixed function path here
396 * to guarantee invariance, depending on compiler optimization flags
397 * results could be different otherwise.
398 */
399 VB->ClipPtr = TransformRaw( &store->results[0],
400 &ctx->_ModelProjectMatrix,
401 VB->AttribPtr[0] );
402
403 /* Drivers expect this to be clean to element 4...
404 */
405 switch (VB->ClipPtr->size) {
406 case 1:
407 /* impossible */
408 case 2:
409 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
410 /* fall-through */
411 case 3:
412 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
413 /* fall-through */
414 case 4:
415 break;
416 }
417 }
418 else {
419 /* Setup the VB pointers so that the next pipeline stages get
420 * their data from the right place (the program output arrays).
421 */
422 VB->ClipPtr = &store->results[VERT_RESULT_HPOS];
423 VB->ClipPtr->size = 4;
424 VB->ClipPtr->count = VB->Count;
425 }
426
427 VB->ColorPtr[0] = &store->results[VERT_RESULT_COL0];
428 VB->ColorPtr[1] = &store->results[VERT_RESULT_BFC0];
429 VB->SecondaryColorPtr[0] = &store->results[VERT_RESULT_COL1];
430 VB->SecondaryColorPtr[1] = &store->results[VERT_RESULT_BFC1];
431 VB->FogCoordPtr = &store->results[VERT_RESULT_FOGC];
432
433 VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VERT_RESULT_COL0];
434 VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VERT_RESULT_COL1];
435 VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VERT_RESULT_FOGC];
436 VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VERT_RESULT_PSIZ];
437
438 for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
439 VB->TexCoordPtr[i] =
440 VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]
441 = &store->results[VERT_RESULT_TEX0 + i];
442 }
443
444 for (i = 0; i < ctx->Const.MaxVarying; i++) {
445 if (program->Base.OutputsWritten & (1 << (VERT_RESULT_VAR0 + i))) {
446 /* Note: varying results get put into the generic attributes */
447 VB->AttribPtr[VERT_ATTRIB_GENERIC0+i]
448 = &store->results[VERT_RESULT_VAR0 + i];
449 }
450 }
451
452
453 /* Perform NDC and cliptest operations:
454 */
455 return do_ndc_cliptest(ctx, store);
456 }
457
458
459 /**
460 * Called the first time stage->run is called. In effect, don't
461 * allocate data until the first time the stage is run.
462 */
463 static GLboolean
464 init_vp(GLcontext *ctx, struct tnl_pipeline_stage *stage)
465 {
466 TNLcontext *tnl = TNL_CONTEXT(ctx);
467 struct vertex_buffer *VB = &(tnl->vb);
468 struct vp_stage_data *store;
469 const GLuint size = VB->Size;
470 GLuint i;
471
472 stage->privatePtr = MALLOC(sizeof(*store));
473 store = VP_STAGE_DATA(stage);
474 if (!store)
475 return GL_FALSE;
476
477 /* Allocate arrays of vertex output values */
478 for (i = 0; i < VERT_RESULT_MAX; i++) {
479 _mesa_vector4f_alloc( &store->results[i], 0, size, 32 );
480 store->results[i].size = 4;
481 }
482
483 /* a few other misc allocations */
484 _mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
485 store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 );
486
487 return GL_TRUE;
488 }
489
490
491 /**
492 * Destructor for this pipeline stage.
493 */
494 static void
495 dtr(struct tnl_pipeline_stage *stage)
496 {
497 struct vp_stage_data *store = VP_STAGE_DATA(stage);
498
499 if (store) {
500 GLuint i;
501
502 /* free the vertex program result arrays */
503 for (i = 0; i < VERT_RESULT_MAX; i++)
504 _mesa_vector4f_free( &store->results[i] );
505
506 /* free misc arrays */
507 _mesa_vector4f_free( &store->ndcCoords );
508 ALIGN_FREE( store->clipmask );
509
510 FREE( store );
511 stage->privatePtr = NULL;
512 }
513 }
514
515
516 static void
517 validate_vp_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
518 {
519 if (ctx->VertexProgram._Current) {
520 _swrast_update_texture_samplers(ctx);
521 }
522 }
523
524
525
526 /**
527 * Public description of this pipeline stage.
528 */
529 const struct tnl_pipeline_stage _tnl_vertex_program_stage =
530 {
531 "vertex-program",
532 NULL, /* private_data */
533 init_vp, /* create */
534 dtr, /* destroy */
535 validate_vp_stage, /* validate */
536 run_vp /* run -- initially set to ctr */
537 };