projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge branch 'mesa_7_7_branch'
[mesa.git] / src / mesa / shader / prog_execute.c
1 /*
2 * Mesa 3-D graphics library
3 * Version: 7.3
4 *
5 * Copyright (C) 1999-2008 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 * \file prog_execute.c
27 * Software interpreter for vertex/fragment programs.
28 * \author Brian Paul
29 */
30
31 /*
32 * NOTE: we do everything in single-precision floating point; we don't
33 * currently observe the single/half/fixed-precision qualifiers.
34 *
35 */
36
37
38 #include "main/glheader.h"
39 #include "main/colormac.h"
40 #include "main/context.h"
41 #include "prog_execute.h"
42 #include "prog_instruction.h"
43 #include "prog_parameter.h"
44 #include "prog_print.h"
45 #include "prog_noise.h"
46
47
48 /* debug predicate */
49 #define DEBUG_PROG 0
50
51
52 /**
53 * Set x to positive or negative infinity.
54 */
55 #if defined(USE_IEEE) || defined(_WIN32)
56 #define SET_POS_INFINITY(x) \
57 do { \
58 fi_type fi; \
59 fi.i = 0x7F800000; \
60 x = fi.f; \
61 } while (0)
62 #define SET_NEG_INFINITY(x) \
63 do { \
64 fi_type fi; \
65 fi.i = 0xFF800000; \
66 x = fi.f; \
67 } while (0)
68 #elif defined(VMS)
69 #define SET_POS_INFINITY(x) x = __MAXFLOAT
70 #define SET_NEG_INFINITY(x) x = -__MAXFLOAT
71 #else
72 #define SET_POS_INFINITY(x) x = (GLfloat) HUGE_VAL
73 #define SET_NEG_INFINITY(x) x = (GLfloat) -HUGE_VAL
74 #endif
75
76 #define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
77
78
79 static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
80
81
82
83 /**
84 * Return a pointer to the 4-element float vector specified by the given
85 * source register.
86 */
87 static INLINE const GLfloat *
88 get_src_register_pointer(const struct prog_src_register *source,
89 const struct gl_program_machine *machine)
90 {
91 const struct gl_program *prog = machine->CurProgram;
92 GLint reg = source->Index;
93
94 if (source->RelAddr) {
95 /* add address register value to src index/offset */
96 reg += machine->AddressReg[0][0];
97 if (reg < 0) {
98 return ZeroVec;
99 }
100 }
101
102 switch (source->File) {
103 case PROGRAM_TEMPORARY:
104 if (reg >= MAX_PROGRAM_TEMPS)
105 return ZeroVec;
106 return machine->Temporaries[reg];
107
108 case PROGRAM_INPUT:
109 if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
110 if (reg >= VERT_ATTRIB_MAX)
111 return ZeroVec;
112 return machine->VertAttribs[reg];
113 }
114 else {
115 if (reg >= FRAG_ATTRIB_MAX)
116 return ZeroVec;
117 return machine->Attribs[reg][machine->CurElement];
118 }
119
120 case PROGRAM_OUTPUT:
121 if (reg >= MAX_PROGRAM_OUTPUTS)
122 return ZeroVec;
123 return machine->Outputs[reg];
124
125 case PROGRAM_LOCAL_PARAM:
126 if (reg >= MAX_PROGRAM_LOCAL_PARAMS)
127 return ZeroVec;
128 return machine->CurProgram->LocalParams[reg];
129
130 case PROGRAM_ENV_PARAM:
131 if (reg >= MAX_PROGRAM_ENV_PARAMS)
132 return ZeroVec;
133 return machine->EnvParams[reg];
134
135 case PROGRAM_STATE_VAR:
136 /* Fallthrough */
137 case PROGRAM_CONSTANT:
138 /* Fallthrough */
139 case PROGRAM_UNIFORM:
140 /* Fallthrough */
141 case PROGRAM_NAMED_PARAM:
142 if (reg >= (GLint) prog->Parameters->NumParameters)
143 return ZeroVec;
144 return prog->Parameters->ParameterValues[reg];
145
146 default:
147 _mesa_problem(NULL,
148 "Invalid src register file %d in get_src_register_pointer()",
149 source->File);
150 return NULL;
151 }
152 }
153
154
155 /**
156 * Return a pointer to the 4-element float vector specified by the given
157 * destination register.
158 */
159 static INLINE GLfloat *
160 get_dst_register_pointer(const struct prog_dst_register *dest,
161 struct gl_program_machine *machine)
162 {
163 static GLfloat dummyReg[4];
164 GLint reg = dest->Index;
165
166 if (dest->RelAddr) {
167 /* add address register value to src index/offset */
168 reg += machine->AddressReg[0][0];
169 if (reg < 0) {
170 return dummyReg;
171 }
172 }
173
174 switch (dest->File) {
175 case PROGRAM_TEMPORARY:
176 if (reg >= MAX_PROGRAM_TEMPS)
177 return dummyReg;
178 return machine->Temporaries[reg];
179
180 case PROGRAM_OUTPUT:
181 if (reg >= MAX_PROGRAM_OUTPUTS)
182 return dummyReg;
183 return machine->Outputs[reg];
184
185 case PROGRAM_WRITE_ONLY:
186 return dummyReg;
187
188 default:
189 _mesa_problem(NULL,
190 "Invalid dest register file %d in get_dst_register_pointer()",
191 dest->File);
192 return NULL;
193 }
194 }
195
196
197
198 /**
199 * Fetch a 4-element float vector from the given source register.
200 * Apply swizzling and negating as needed.
201 */
202 static void
203 fetch_vector4(const struct prog_src_register *source,
204 const struct gl_program_machine *machine, GLfloat result[4])
205 {
206 const GLfloat *src = get_src_register_pointer(source, machine);
207 ASSERT(src);
208
209 if (source->Swizzle == SWIZZLE_NOOP) {
210 /* no swizzling */
211 COPY_4V(result, src);
212 }
213 else {
214 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
215 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
216 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
217 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
218 result[0] = src[GET_SWZ(source->Swizzle, 0)];
219 result[1] = src[GET_SWZ(source->Swizzle, 1)];
220 result[2] = src[GET_SWZ(source->Swizzle, 2)];
221 result[3] = src[GET_SWZ(source->Swizzle, 3)];
222 }
223
224 if (source->Abs) {
225 result[0] = FABSF(result[0]);
226 result[1] = FABSF(result[1]);
227 result[2] = FABSF(result[2]);
228 result[3] = FABSF(result[3]);
229 }
230 if (source->Negate) {
231 ASSERT(source->Negate == NEGATE_XYZW);
232 result[0] = -result[0];
233 result[1] = -result[1];
234 result[2] = -result[2];
235 result[3] = -result[3];
236 }
237
238 #ifdef NAN_CHECK
239 assert(!IS_INF_OR_NAN(result[0]));
240 assert(!IS_INF_OR_NAN(result[0]));
241 assert(!IS_INF_OR_NAN(result[0]));
242 assert(!IS_INF_OR_NAN(result[0]));
243 #endif
244 }
245
246
247 /**
248 * Fetch a 4-element uint vector from the given source register.
249 * Apply swizzling but not negation/abs.
250 */
251 static void
252 fetch_vector4ui(const struct prog_src_register *source,
253 const struct gl_program_machine *machine, GLuint result[4])
254 {
255 const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
256 ASSERT(src);
257
258 if (source->Swizzle == SWIZZLE_NOOP) {
259 /* no swizzling */
260 COPY_4V(result, src);
261 }
262 else {
263 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
264 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
265 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
266 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
267 result[0] = src[GET_SWZ(source->Swizzle, 0)];
268 result[1] = src[GET_SWZ(source->Swizzle, 1)];
269 result[2] = src[GET_SWZ(source->Swizzle, 2)];
270 result[3] = src[GET_SWZ(source->Swizzle, 3)];
271 }
272
273 /* Note: no Negate or Abs here */
274 }
275
276
277
278 /**
279 * Fetch the derivative with respect to X or Y for the given register.
280 * XXX this currently only works for fragment program input attribs.
281 */
282 static void
283 fetch_vector4_deriv(GLcontext * ctx,
284 const struct prog_src_register *source,
285 const struct gl_program_machine *machine,
286 char xOrY, GLfloat result[4])
287 {
288 if (source->File == PROGRAM_INPUT &&
289 source->Index < (GLint) machine->NumDeriv) {
290 const GLint col = machine->CurElement;
291 const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
292 const GLfloat invQ = 1.0f / w;
293 GLfloat deriv[4];
294
295 if (xOrY == 'X') {
296 deriv[0] = machine->DerivX[source->Index][0] * invQ;
297 deriv[1] = machine->DerivX[source->Index][1] * invQ;
298 deriv[2] = machine->DerivX[source->Index][2] * invQ;
299 deriv[3] = machine->DerivX[source->Index][3] * invQ;
300 }
301 else {
302 deriv[0] = machine->DerivY[source->Index][0] * invQ;
303 deriv[1] = machine->DerivY[source->Index][1] * invQ;
304 deriv[2] = machine->DerivY[source->Index][2] * invQ;
305 deriv[3] = machine->DerivY[source->Index][3] * invQ;
306 }
307
308 result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
309 result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
310 result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
311 result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
312
313 if (source->Abs) {
314 result[0] = FABSF(result[0]);
315 result[1] = FABSF(result[1]);
316 result[2] = FABSF(result[2]);
317 result[3] = FABSF(result[3]);
318 }
319 if (source->Negate) {
320 ASSERT(source->Negate == NEGATE_XYZW);
321 result[0] = -result[0];
322 result[1] = -result[1];
323 result[2] = -result[2];
324 result[3] = -result[3];
325 }
326 }
327 else {
328 ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
329 }
330 }
331
332
333 /**
334 * As above, but only return result[0] element.
335 */
336 static void
337 fetch_vector1(const struct prog_src_register *source,
338 const struct gl_program_machine *machine, GLfloat result[4])
339 {
340 const GLfloat *src = get_src_register_pointer(source, machine);
341 ASSERT(src);
342
343 result[0] = src[GET_SWZ(source->Swizzle, 0)];
344
345 if (source->Abs) {
346 result[0] = FABSF(result[0]);
347 }
348 if (source->Negate) {
349 result[0] = -result[0];
350 }
351 }
352
353
354 static GLuint
355 fetch_vector1ui(const struct prog_src_register *source,
356 const struct gl_program_machine *machine)
357 {
358 const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
359 GLuint result;
360
361 ASSERT(src);
362
363 result = src[GET_SWZ(source->Swizzle, 0)];
364
365 if (source->Abs) {
366 result = FABSF(result);
367 }
368 if (source->Negate) {
369 result = -result;
370 }
371
372 return result;
373 }
374
375
376 /**
377 * Fetch texel from texture. Use partial derivatives when possible.
378 */
379 static INLINE void
380 fetch_texel(GLcontext *ctx,
381 const struct gl_program_machine *machine,
382 const struct prog_instruction *inst,
383 const GLfloat texcoord[4], GLfloat lodBias,
384 GLfloat color[4])
385 {
386 const GLuint unit = machine->Samplers[inst->TexSrcUnit];
387
388 /* Note: we only have the right derivatives for fragment input attribs.
389 */
390 if (machine->NumDeriv > 0 &&
391 inst->SrcReg[0].File == PROGRAM_INPUT &&
392 inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
393 /* simple texture fetch for which we should have derivatives */
394 GLuint attr = inst->SrcReg[0].Index;
395 machine->FetchTexelDeriv(ctx, texcoord,
396 machine->DerivX[attr],
397 machine->DerivY[attr],
398 lodBias, unit, color);
399 }
400 else {
401 machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
402 }
403 }
404
405
406 /**
407 * Test value against zero and return GT, LT, EQ or UN if NaN.
408 */
409 static INLINE GLuint
410 generate_cc(float value)
411 {
412 if (value != value)
413 return COND_UN; /* NaN */
414 if (value > 0.0F)
415 return COND_GT;
416 if (value < 0.0F)
417 return COND_LT;
418 return COND_EQ;
419 }
420
421
422 /**
423 * Test if the ccMaskRule is satisfied by the given condition code.
424 * Used to mask destination writes according to the current condition code.
425 */
426 static INLINE GLboolean
427 test_cc(GLuint condCode, GLuint ccMaskRule)
428 {
429 switch (ccMaskRule) {
430 case COND_EQ: return (condCode == COND_EQ);
431 case COND_NE: return (condCode != COND_EQ);
432 case COND_LT: return (condCode == COND_LT);
433 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
434 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
435 case COND_GT: return (condCode == COND_GT);
436 case COND_TR: return GL_TRUE;
437 case COND_FL: return GL_FALSE;
438 default: return GL_TRUE;
439 }
440 }
441
442
443 /**
444 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
445 * or GL_FALSE to indicate result.
446 */
447 static INLINE GLboolean
448 eval_condition(const struct gl_program_machine *machine,
449 const struct prog_instruction *inst)
450 {
451 const GLuint swizzle = inst->DstReg.CondSwizzle;
452 const GLuint condMask = inst->DstReg.CondMask;
453 if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
454 test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
455 test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
456 test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
457 return GL_TRUE;
458 }
459 else {
460 return GL_FALSE;
461 }
462 }
463
464
465
466 /**
467 * Store 4 floats into a register. Observe the instructions saturate and
468 * set-condition-code flags.
469 */
470 static void
471 store_vector4(const struct prog_instruction *inst,
472 struct gl_program_machine *machine, const GLfloat value[4])
473 {
474 const struct prog_dst_register *dstReg = &(inst->DstReg);
475 const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
476 GLuint writeMask = dstReg->WriteMask;
477 GLfloat clampedValue[4];
478 GLfloat *dst = get_dst_register_pointer(dstReg, machine);
479
480 #if 0
481 if (value[0] > 1.0e10 ||
482 IS_INF_OR_NAN(value[0]) ||
483 IS_INF_OR_NAN(value[1]) ||
484 IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
485 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
486 #endif
487
488 if (clamp) {
489 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
490 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
491 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
492 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
493 value = clampedValue;
494 }
495
496 if (dstReg->CondMask != COND_TR) {
497 /* condition codes may turn off some writes */
498 if (writeMask & WRITEMASK_X) {
499 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
500 dstReg->CondMask))
501 writeMask &= ~WRITEMASK_X;
502 }
503 if (writeMask & WRITEMASK_Y) {
504 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
505 dstReg->CondMask))
506 writeMask &= ~WRITEMASK_Y;
507 }
508 if (writeMask & WRITEMASK_Z) {
509 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
510 dstReg->CondMask))
511 writeMask &= ~WRITEMASK_Z;
512 }
513 if (writeMask & WRITEMASK_W) {
514 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
515 dstReg->CondMask))
516 writeMask &= ~WRITEMASK_W;
517 }
518 }
519
520 #ifdef NAN_CHECK
521 assert(!IS_INF_OR_NAN(value[0]));
522 assert(!IS_INF_OR_NAN(value[0]));
523 assert(!IS_INF_OR_NAN(value[0]));
524 assert(!IS_INF_OR_NAN(value[0]));
525 #endif
526
527 if (writeMask & WRITEMASK_X)
528 dst[0] = value[0];
529 if (writeMask & WRITEMASK_Y)
530 dst[1] = value[1];
531 if (writeMask & WRITEMASK_Z)
532 dst[2] = value[2];
533 if (writeMask & WRITEMASK_W)
534 dst[3] = value[3];
535
536 if (inst->CondUpdate) {
537 if (writeMask & WRITEMASK_X)
538 machine->CondCodes[0] = generate_cc(value[0]);
539 if (writeMask & WRITEMASK_Y)
540 machine->CondCodes[1] = generate_cc(value[1]);
541 if (writeMask & WRITEMASK_Z)
542 machine->CondCodes[2] = generate_cc(value[2]);
543 if (writeMask & WRITEMASK_W)
544 machine->CondCodes[3] = generate_cc(value[3]);
545 #if DEBUG_PROG
546 printf("CondCodes=(%s,%s,%s,%s) for:\n",
547 _mesa_condcode_string(machine->CondCodes[0]),
548 _mesa_condcode_string(machine->CondCodes[1]),
549 _mesa_condcode_string(machine->CondCodes[2]),
550 _mesa_condcode_string(machine->CondCodes[3]));
551 #endif
552 }
553 }
554
555
556 /**
557 * Store 4 uints into a register. Observe the set-condition-code flags.
558 */
559 static void
560 store_vector4ui(const struct prog_instruction *inst,
561 struct gl_program_machine *machine, const GLuint value[4])
562 {
563 const struct prog_dst_register *dstReg = &(inst->DstReg);
564 GLuint writeMask = dstReg->WriteMask;
565 GLuint *dst = (GLuint *) get_dst_register_pointer(dstReg, machine);
566
567 if (dstReg->CondMask != COND_TR) {
568 /* condition codes may turn off some writes */
569 if (writeMask & WRITEMASK_X) {
570 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
571 dstReg->CondMask))
572 writeMask &= ~WRITEMASK_X;
573 }
574 if (writeMask & WRITEMASK_Y) {
575 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
576 dstReg->CondMask))
577 writeMask &= ~WRITEMASK_Y;
578 }
579 if (writeMask & WRITEMASK_Z) {
580 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
581 dstReg->CondMask))
582 writeMask &= ~WRITEMASK_Z;
583 }
584 if (writeMask & WRITEMASK_W) {
585 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
586 dstReg->CondMask))
587 writeMask &= ~WRITEMASK_W;
588 }
589 }
590
591 if (writeMask & WRITEMASK_X)
592 dst[0] = value[0];
593 if (writeMask & WRITEMASK_Y)
594 dst[1] = value[1];
595 if (writeMask & WRITEMASK_Z)
596 dst[2] = value[2];
597 if (writeMask & WRITEMASK_W)
598 dst[3] = value[3];
599
600 if (inst->CondUpdate) {
601 if (writeMask & WRITEMASK_X)
602 machine->CondCodes[0] = generate_cc(value[0]);
603 if (writeMask & WRITEMASK_Y)
604 machine->CondCodes[1] = generate_cc(value[1]);
605 if (writeMask & WRITEMASK_Z)
606 machine->CondCodes[2] = generate_cc(value[2]);
607 if (writeMask & WRITEMASK_W)
608 machine->CondCodes[3] = generate_cc(value[3]);
609 #if DEBUG_PROG
610 printf("CondCodes=(%s,%s,%s,%s) for:\n",
611 _mesa_condcode_string(machine->CondCodes[0]),
612 _mesa_condcode_string(machine->CondCodes[1]),
613 _mesa_condcode_string(machine->CondCodes[2]),
614 _mesa_condcode_string(machine->CondCodes[3]));
615 #endif
616 }
617 }
618
619
620
621 /**
622 * Execute the given vertex/fragment program.
623 *
624 * \param ctx rendering context
625 * \param program the program to execute
626 * \param machine machine state (must be initialized)
627 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
628 */
629 GLboolean
630 _mesa_execute_program(GLcontext * ctx,
631 const struct gl_program *program,
632 struct gl_program_machine *machine)
633 {
634 const GLuint numInst = program->NumInstructions;
635 const GLuint maxExec = 10000;
636 GLuint pc, numExec = 0;
637
638 machine->CurProgram = program;
639
640 if (DEBUG_PROG) {
641 printf("execute program %u --------------------\n", program->Id);
642 }
643
644 if (program->Target == GL_VERTEX_PROGRAM_ARB) {
645 machine->EnvParams = ctx->VertexProgram.Parameters;
646 }
647 else {
648 machine->EnvParams = ctx->FragmentProgram.Parameters;
649 }
650
651 for (pc = 0; pc < numInst; pc++) {
652 const struct prog_instruction *inst = program->Instructions + pc;
653
654 if (DEBUG_PROG) {
655 _mesa_print_instruction(inst);
656 }
657
658 switch (inst->Opcode) {
659 case OPCODE_ABS:
660 {
661 GLfloat a[4], result[4];
662 fetch_vector4(&inst->SrcReg[0], machine, a);
663 result[0] = FABSF(a[0]);
664 result[1] = FABSF(a[1]);
665 result[2] = FABSF(a[2]);
666 result[3] = FABSF(a[3]);
667 store_vector4(inst, machine, result);
668 }
669 break;
670 case OPCODE_ADD:
671 {
672 GLfloat a[4], b[4], result[4];
673 fetch_vector4(&inst->SrcReg[0], machine, a);
674 fetch_vector4(&inst->SrcReg[1], machine, b);
675 result[0] = a[0] + b[0];
676 result[1] = a[1] + b[1];
677 result[2] = a[2] + b[2];
678 result[3] = a[3] + b[3];
679 store_vector4(inst, machine, result);
680 if (DEBUG_PROG) {
681 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
682 result[0], result[1], result[2], result[3],
683 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
684 }
685 }
686 break;
687 case OPCODE_AND: /* bitwise AND */
688 {
689 GLuint a[4], b[4], result[4];
690 fetch_vector4ui(&inst->SrcReg[0], machine, a);
691 fetch_vector4ui(&inst->SrcReg[1], machine, b);
692 result[0] = a[0] & b[0];
693 result[1] = a[1] & b[1];
694 result[2] = a[2] & b[2];
695 result[3] = a[3] & b[3];
696 store_vector4ui(inst, machine, result);
697 }
698 break;
699 case OPCODE_ARL:
700 {
701 GLfloat t[4];
702 fetch_vector4(&inst->SrcReg[0], machine, t);
703 machine->AddressReg[0][0] = IFLOOR(t[0]);
704 }
705 break;
706 case OPCODE_BGNLOOP:
707 /* no-op */
708 ASSERT(program->Instructions[inst->BranchTarget].Opcode
709 == OPCODE_ENDLOOP);
710 break;
711 case OPCODE_ENDLOOP:
712 /* subtract 1 here since pc is incremented by for(pc) loop */
713 ASSERT(program->Instructions[inst->BranchTarget].Opcode
714 == OPCODE_BGNLOOP);
715 pc = inst->BranchTarget - 1; /* go to matching BNGLOOP */
716 break;
717 case OPCODE_BGNSUB: /* begin subroutine */
718 break;
719 case OPCODE_ENDSUB: /* end subroutine */
720 break;
721 case OPCODE_BRA: /* branch (conditional) */
722 if (eval_condition(machine, inst)) {
723 /* take branch */
724 /* Subtract 1 here since we'll do pc++ below */
725 pc = inst->BranchTarget - 1;
726 }
727 break;
728 case OPCODE_BRK: /* break out of loop (conditional) */
729 ASSERT(program->Instructions[inst->BranchTarget].Opcode
730 == OPCODE_ENDLOOP);
731 if (eval_condition(machine, inst)) {
732 /* break out of loop */
733 /* pc++ at end of for-loop will put us after the ENDLOOP inst */
734 pc = inst->BranchTarget;
735 }
736 break;
737 case OPCODE_CONT: /* continue loop (conditional) */
738 ASSERT(program->Instructions[inst->BranchTarget].Opcode
739 == OPCODE_ENDLOOP);
740 if (eval_condition(machine, inst)) {
741 /* continue at ENDLOOP */
742 /* Subtract 1 here since we'll do pc++ at end of for-loop */
743 pc = inst->BranchTarget - 1;
744 }
745 break;
746 case OPCODE_CAL: /* Call subroutine (conditional) */
747 if (eval_condition(machine, inst)) {
748 /* call the subroutine */
749 if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
750 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
751 }
752 machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
753 /* Subtract 1 here since we'll do pc++ at end of for-loop */
754 pc = inst->BranchTarget - 1;
755 }
756 break;
757 case OPCODE_CMP:
758 {
759 GLfloat a[4], b[4], c[4], result[4];
760 fetch_vector4(&inst->SrcReg[0], machine, a);
761 fetch_vector4(&inst->SrcReg[1], machine, b);
762 fetch_vector4(&inst->SrcReg[2], machine, c);
763 result[0] = a[0] < 0.0F ? b[0] : c[0];
764 result[1] = a[1] < 0.0F ? b[1] : c[1];
765 result[2] = a[2] < 0.0F ? b[2] : c[2];
766 result[3] = a[3] < 0.0F ? b[3] : c[3];
767 store_vector4(inst, machine, result);
768 }
769 break;
770 case OPCODE_COS:
771 {
772 GLfloat a[4], result[4];
773 fetch_vector1(&inst->SrcReg[0], machine, a);
774 result[0] = result[1] = result[2] = result[3]
775 = (GLfloat) _mesa_cos(a[0]);
776 store_vector4(inst, machine, result);
777 }
778 break;
779 case OPCODE_DDX: /* Partial derivative with respect to X */
780 {
781 GLfloat result[4];
782 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
783 'X', result);
784 store_vector4(inst, machine, result);
785 }
786 break;
787 case OPCODE_DDY: /* Partial derivative with respect to Y */
788 {
789 GLfloat result[4];
790 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
791 'Y', result);
792 store_vector4(inst, machine, result);
793 }
794 break;
795 case OPCODE_DP2:
796 {
797 GLfloat a[4], b[4], result[4];
798 fetch_vector4(&inst->SrcReg[0], machine, a);
799 fetch_vector4(&inst->SrcReg[1], machine, b);
800 result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
801 store_vector4(inst, machine, result);
802 if (DEBUG_PROG) {
803 printf("DP2 %g = (%g %g) . (%g %g)\n",
804 result[0], a[0], a[1], b[0], b[1]);
805 }
806 }
807 break;
808 case OPCODE_DP2A:
809 {
810 GLfloat a[4], b[4], c, result[4];
811 fetch_vector4(&inst->SrcReg[0], machine, a);
812 fetch_vector4(&inst->SrcReg[1], machine, b);
813 fetch_vector1(&inst->SrcReg[1], machine, &c);
814 result[0] = result[1] = result[2] = result[3] = DOT2(a, b) + c;
815 store_vector4(inst, machine, result);
816 if (DEBUG_PROG) {
817 printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
818 result[0], a[0], a[1], b[0], b[1], c);
819 }
820 }
821 break;
822 case OPCODE_DP3:
823 {
824 GLfloat a[4], b[4], result[4];
825 fetch_vector4(&inst->SrcReg[0], machine, a);
826 fetch_vector4(&inst->SrcReg[1], machine, b);
827 result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
828 store_vector4(inst, machine, result);
829 if (DEBUG_PROG) {
830 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
831 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
832 }
833 }
834 break;
835 case OPCODE_DP4:
836 {
837 GLfloat a[4], b[4], result[4];
838 fetch_vector4(&inst->SrcReg[0], machine, a);
839 fetch_vector4(&inst->SrcReg[1], machine, b);
840 result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
841 store_vector4(inst, machine, result);
842 if (DEBUG_PROG) {
843 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
844 result[0], a[0], a[1], a[2], a[3],
845 b[0], b[1], b[2], b[3]);
846 }
847 }
848 break;
849 case OPCODE_DPH:
850 {
851 GLfloat a[4], b[4], result[4];
852 fetch_vector4(&inst->SrcReg[0], machine, a);
853 fetch_vector4(&inst->SrcReg[1], machine, b);
854 result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
855 store_vector4(inst, machine, result);
856 }
857 break;
858 case OPCODE_DST: /* Distance vector */
859 {
860 GLfloat a[4], b[4], result[4];
861 fetch_vector4(&inst->SrcReg[0], machine, a);
862 fetch_vector4(&inst->SrcReg[1], machine, b);
863 result[0] = 1.0F;
864 result[1] = a[1] * b[1];
865 result[2] = a[2];
866 result[3] = b[3];
867 store_vector4(inst, machine, result);
868 }
869 break;
870 case OPCODE_EXP:
871 {
872 GLfloat t[4], q[4], floor_t0;
873 fetch_vector1(&inst->SrcReg[0], machine, t);
874 floor_t0 = FLOORF(t[0]);
875 if (floor_t0 > FLT_MAX_EXP) {
876 SET_POS_INFINITY(q[0]);
877 SET_POS_INFINITY(q[2]);
878 }
879 else if (floor_t0 < FLT_MIN_EXP) {
880 q[0] = 0.0F;
881 q[2] = 0.0F;
882 }
883 else {
884 q[0] = LDEXPF(1.0, (int) floor_t0);
885 /* Note: GL_NV_vertex_program expects
886 * result.z = result.x * APPX(result.y)
887 * We do what the ARB extension says.
888 */
889 q[2] = (GLfloat) _mesa_pow(2.0, t[0]);
890 }
891 q[1] = t[0] - floor_t0;
892 q[3] = 1.0F;
893 store_vector4( inst, machine, q );
894 }
895 break;
896 case OPCODE_EX2: /* Exponential base 2 */
897 {
898 GLfloat a[4], result[4], val;
899 fetch_vector1(&inst->SrcReg[0], machine, a);
900 val = (GLfloat) _mesa_pow(2.0, a[0]);
901 /*
902 if (IS_INF_OR_NAN(val))
903 val = 1.0e10;
904 */
905 result[0] = result[1] = result[2] = result[3] = val;
906 store_vector4(inst, machine, result);
907 }
908 break;
909 case OPCODE_FLR:
910 {
911 GLfloat a[4], result[4];
912 fetch_vector4(&inst->SrcReg[0], machine, a);
913 result[0] = FLOORF(a[0]);
914 result[1] = FLOORF(a[1]);
915 result[2] = FLOORF(a[2]);
916 result[3] = FLOORF(a[3]);
917 store_vector4(inst, machine, result);
918 }
919 break;
920 case OPCODE_FRC:
921 {
922 GLfloat a[4], result[4];
923 fetch_vector4(&inst->SrcReg[0], machine, a);
924 result[0] = a[0] - FLOORF(a[0]);
925 result[1] = a[1] - FLOORF(a[1]);
926 result[2] = a[2] - FLOORF(a[2]);
927 result[3] = a[3] - FLOORF(a[3]);
928 store_vector4(inst, machine, result);
929 }
930 break;
931 case OPCODE_IF:
932 {
933 GLboolean cond;
934 ASSERT(program->Instructions[inst->BranchTarget].Opcode
935 == OPCODE_ELSE ||
936 program->Instructions[inst->BranchTarget].Opcode
937 == OPCODE_ENDIF);
938 /* eval condition */
939 if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
940 GLfloat a[4];
941 fetch_vector1(&inst->SrcReg[0], machine, a);
942 cond = (a[0] != 0.0);
943 }
944 else {
945 cond = eval_condition(machine, inst);
946 }
947 if (DEBUG_PROG) {
948 printf("IF: %d\n", cond);
949 }
950 /* do if/else */
951 if (cond) {
952 /* do if-clause (just continue execution) */
953 }
954 else {
955 /* go to the instruction after ELSE or ENDIF */
956 assert(inst->BranchTarget >= 0);
957 pc = inst->BranchTarget;
958 }
959 }
960 break;
961 case OPCODE_ELSE:
962 /* goto ENDIF */
963 ASSERT(program->Instructions[inst->BranchTarget].Opcode
964 == OPCODE_ENDIF);
965 assert(inst->BranchTarget >= 0);
966 pc = inst->BranchTarget;
967 break;
968 case OPCODE_ENDIF:
969 /* nothing */
970 break;
971 case OPCODE_KIL_NV: /* NV_f_p only (conditional) */
972 if (eval_condition(machine, inst)) {
973 return GL_FALSE;
974 }
975 break;
976 case OPCODE_KIL: /* ARB_f_p only */
977 {
978 GLfloat a[4];
979 fetch_vector4(&inst->SrcReg[0], machine, a);
980 if (DEBUG_PROG) {
981 printf("KIL if (%g %g %g %g) <= 0.0\n",
982 a[0], a[1], a[2], a[3]);
983 }
984
985 if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
986 return GL_FALSE;
987 }
988 }
989 break;
990 case OPCODE_LG2: /* log base 2 */
991 {
992 GLfloat a[4], result[4], val;
993 fetch_vector1(&inst->SrcReg[0], machine, a);
994 /* The fast LOG2 macro doesn't meet the precision requirements.
995 */
996 if (a[0] == 0.0F) {
997 val = -FLT_MAX;
998 }
999 else {
1000 val = log(a[0]) * 1.442695F;
1001 }
1002 result[0] = result[1] = result[2] = result[3] = val;
1003 store_vector4(inst, machine, result);
1004 }
1005 break;
1006 case OPCODE_LIT:
1007 {
1008 const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
1009 GLfloat a[4], result[4];
1010 fetch_vector4(&inst->SrcReg[0], machine, a);
1011 a[0] = MAX2(a[0], 0.0F);
1012 a[1] = MAX2(a[1], 0.0F);
1013 /* XXX ARB version clamps a[3], NV version doesn't */
1014 a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
1015 result[0] = 1.0F;
1016 result[1] = a[0];
1017 /* XXX we could probably just use pow() here */
1018 if (a[0] > 0.0F) {
1019 if (a[1] == 0.0 && a[3] == 0.0)
1020 result[2] = 1.0;
1021 else
1022 result[2] = (GLfloat) _mesa_pow(a[1], a[3]);
1023 }
1024 else {
1025 result[2] = 0.0;
1026 }
1027 result[3] = 1.0F;
1028 store_vector4(inst, machine, result);
1029 if (DEBUG_PROG) {
1030 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
1031 result[0], result[1], result[2], result[3],
1032 a[0], a[1], a[2], a[3]);
1033 }
1034 }
1035 break;
1036 case OPCODE_LOG:
1037 {
1038 GLfloat t[4], q[4], abs_t0;
1039 fetch_vector1(&inst->SrcReg[0], machine, t);
1040 abs_t0 = FABSF(t[0]);
1041 if (abs_t0 != 0.0F) {
1042 /* Since we really can't handle infinite values on VMS
1043 * like other OSes we'll use __MAXFLOAT to represent
1044 * infinity. This may need some tweaking.
1045 */
1046 #ifdef VMS
1047 if (abs_t0 == __MAXFLOAT)
1048 #else
1049 if (IS_INF_OR_NAN(abs_t0))
1050 #endif
1051 {
1052 SET_POS_INFINITY(q[0]);
1053 q[1] = 1.0F;
1054 SET_POS_INFINITY(q[2]);
1055 }
1056 else {
1057 int exponent;
1058 GLfloat mantissa = FREXPF(t[0], &exponent);
1059 q[0] = (GLfloat) (exponent - 1);
1060 q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
1061
1062 /* The fast LOG2 macro doesn't meet the precision
1063 * requirements.
1064 */
1065 q[2] = (log(t[0]) * 1.442695F);
1066 }
1067 }
1068 else {
1069 SET_NEG_INFINITY(q[0]);
1070 q[1] = 1.0F;
1071 SET_NEG_INFINITY(q[2]);
1072 }
1073 q[3] = 1.0;
1074 store_vector4(inst, machine, q);
1075 }
1076 break;
1077 case OPCODE_LRP:
1078 {
1079 GLfloat a[4], b[4], c[4], result[4];
1080 fetch_vector4(&inst->SrcReg[0], machine, a);
1081 fetch_vector4(&inst->SrcReg[1], machine, b);
1082 fetch_vector4(&inst->SrcReg[2], machine, c);
1083 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
1084 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
1085 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
1086 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
1087 store_vector4(inst, machine, result);
1088 if (DEBUG_PROG) {
1089 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1090 "(%g %g %g %g), (%g %g %g %g)\n",
1091 result[0], result[1], result[2], result[3],
1092 a[0], a[1], a[2], a[3],
1093 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1094 }
1095 }
1096 break;
1097 case OPCODE_MAD:
1098 {
1099 GLfloat a[4], b[4], c[4], result[4];
1100 fetch_vector4(&inst->SrcReg[0], machine, a);
1101 fetch_vector4(&inst->SrcReg[1], machine, b);
1102 fetch_vector4(&inst->SrcReg[2], machine, c);
1103 result[0] = a[0] * b[0] + c[0];
1104 result[1] = a[1] * b[1] + c[1];
1105 result[2] = a[2] * b[2] + c[2];
1106 result[3] = a[3] * b[3] + c[3];
1107 store_vector4(inst, machine, result);
1108 if (DEBUG_PROG) {
1109 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1110 "(%g %g %g %g) + (%g %g %g %g)\n",
1111 result[0], result[1], result[2], result[3],
1112 a[0], a[1], a[2], a[3],
1113 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1114 }
1115 }
1116 break;
1117 case OPCODE_MAX:
1118 {
1119 GLfloat a[4], b[4], result[4];
1120 fetch_vector4(&inst->SrcReg[0], machine, a);
1121 fetch_vector4(&inst->SrcReg[1], machine, b);
1122 result[0] = MAX2(a[0], b[0]);
1123 result[1] = MAX2(a[1], b[1]);
1124 result[2] = MAX2(a[2], b[2]);
1125 result[3] = MAX2(a[3], b[3]);
1126 store_vector4(inst, machine, result);
1127 if (DEBUG_PROG) {
1128 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1129 result[0], result[1], result[2], result[3],
1130 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1131 }
1132 }
1133 break;
1134 case OPCODE_MIN:
1135 {
1136 GLfloat a[4], b[4], result[4];
1137 fetch_vector4(&inst->SrcReg[0], machine, a);
1138 fetch_vector4(&inst->SrcReg[1], machine, b);
1139 result[0] = MIN2(a[0], b[0]);
1140 result[1] = MIN2(a[1], b[1]);
1141 result[2] = MIN2(a[2], b[2]);
1142 result[3] = MIN2(a[3], b[3]);
1143 store_vector4(inst, machine, result);
1144 }
1145 break;
1146 case OPCODE_MOV:
1147 {
1148 GLfloat result[4];
1149 fetch_vector4(&inst->SrcReg[0], machine, result);
1150 store_vector4(inst, machine, result);
1151 if (DEBUG_PROG) {
1152 printf("MOV (%g %g %g %g)\n",
1153 result[0], result[1], result[2], result[3]);
1154 }
1155 }
1156 break;
1157 case OPCODE_MUL:
1158 {
1159 GLfloat a[4], b[4], result[4];
1160 fetch_vector4(&inst->SrcReg[0], machine, a);
1161 fetch_vector4(&inst->SrcReg[1], machine, b);
1162 result[0] = a[0] * b[0];
1163 result[1] = a[1] * b[1];
1164 result[2] = a[2] * b[2];
1165 result[3] = a[3] * b[3];
1166 store_vector4(inst, machine, result);
1167 if (DEBUG_PROG) {
1168 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1169 result[0], result[1], result[2], result[3],
1170 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1171 }
1172 }
1173 break;
1174 case OPCODE_NOISE1:
1175 {
1176 GLfloat a[4], result[4];
1177 fetch_vector1(&inst->SrcReg[0], machine, a);
1178 result[0] =
1179 result[1] =
1180 result[2] =
1181 result[3] = _mesa_noise1(a[0]);
1182 store_vector4(inst, machine, result);
1183 }
1184 break;
1185 case OPCODE_NOISE2:
1186 {
1187 GLfloat a[4], result[4];
1188 fetch_vector4(&inst->SrcReg[0], machine, a);
1189 result[0] =
1190 result[1] =
1191 result[2] = result[3] = _mesa_noise2(a[0], a[1]);
1192 store_vector4(inst, machine, result);
1193 }
1194 break;
1195 case OPCODE_NOISE3:
1196 {
1197 GLfloat a[4], result[4];
1198 fetch_vector4(&inst->SrcReg[0], machine, a);
1199 result[0] =
1200 result[1] =
1201 result[2] =
1202 result[3] = _mesa_noise3(a[0], a[1], a[2]);
1203 store_vector4(inst, machine, result);
1204 }
1205 break;
1206 case OPCODE_NOISE4:
1207 {
1208 GLfloat a[4], result[4];
1209 fetch_vector4(&inst->SrcReg[0], machine, a);
1210 result[0] =
1211 result[1] =
1212 result[2] =
1213 result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
1214 store_vector4(inst, machine, result);
1215 }
1216 break;
1217 case OPCODE_NOP:
1218 break;
1219 case OPCODE_NOT: /* bitwise NOT */
1220 {
1221 GLuint a[4], result[4];
1222 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1223 result[0] = ~a[0];
1224 result[1] = ~a[1];
1225 result[2] = ~a[2];
1226 result[3] = ~a[3];
1227 store_vector4ui(inst, machine, result);
1228 }
1229 break;
1230 case OPCODE_NRM3: /* 3-component normalization */
1231 {
1232 GLfloat a[4], result[4];
1233 GLfloat tmp;
1234 fetch_vector4(&inst->SrcReg[0], machine, a);
1235 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
1236 if (tmp != 0.0F)
1237 tmp = INV_SQRTF(tmp);
1238 result[0] = tmp * a[0];
1239 result[1] = tmp * a[1];
1240 result[2] = tmp * a[2];
1241 result[3] = 0.0; /* undefined, but prevent valgrind warnings */
1242 store_vector4(inst, machine, result);
1243 }
1244 break;
1245 case OPCODE_NRM4: /* 4-component normalization */
1246 {
1247 GLfloat a[4], result[4];
1248 GLfloat tmp;
1249 fetch_vector4(&inst->SrcReg[0], machine, a);
1250 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
1251 if (tmp != 0.0F)
1252 tmp = INV_SQRTF(tmp);
1253 result[0] = tmp * a[0];
1254 result[1] = tmp * a[1];
1255 result[2] = tmp * a[2];
1256 result[3] = tmp * a[3];
1257 store_vector4(inst, machine, result);
1258 }
1259 break;
1260 case OPCODE_OR: /* bitwise OR */
1261 {
1262 GLuint a[4], b[4], result[4];
1263 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1264 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1265 result[0] = a[0] | b[0];
1266 result[1] = a[1] | b[1];
1267 result[2] = a[2] | b[2];
1268 result[3] = a[3] | b[3];
1269 store_vector4ui(inst, machine, result);
1270 }
1271 break;
1272 case OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
1273 {
1274 GLfloat a[4];
1275 GLuint result[4];
1276 GLhalfNV hx, hy;
1277 fetch_vector4(&inst->SrcReg[0], machine, a);
1278 hx = _mesa_float_to_half(a[0]);
1279 hy = _mesa_float_to_half(a[1]);
1280 result[0] =
1281 result[1] =
1282 result[2] =
1283 result[3] = hx | (hy << 16);
1284 store_vector4ui(inst, machine, result);
1285 }
1286 break;
1287 case OPCODE_PK2US: /* pack two GLushorts into one 32-bit float */
1288 {
1289 GLfloat a[4];
1290 GLuint result[4], usx, usy;
1291 fetch_vector4(&inst->SrcReg[0], machine, a);
1292 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1293 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1294 usx = IROUND(a[0] * 65535.0F);
1295 usy = IROUND(a[1] * 65535.0F);
1296 result[0] =
1297 result[1] =
1298 result[2] =
1299 result[3] = usx | (usy << 16);
1300 store_vector4ui(inst, machine, result);
1301 }
1302 break;
1303 case OPCODE_PK4B: /* pack four GLbytes into one 32-bit float */
1304 {
1305 GLfloat a[4];
1306 GLuint result[4], ubx, uby, ubz, ubw;
1307 fetch_vector4(&inst->SrcReg[0], machine, a);
1308 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
1309 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
1310 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
1311 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
1312 ubx = IROUND(127.0F * a[0] + 128.0F);
1313 uby = IROUND(127.0F * a[1] + 128.0F);
1314 ubz = IROUND(127.0F * a[2] + 128.0F);
1315 ubw = IROUND(127.0F * a[3] + 128.0F);
1316 result[0] =
1317 result[1] =
1318 result[2] =
1319 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1320 store_vector4ui(inst, machine, result);
1321 }
1322 break;
1323 case OPCODE_PK4UB: /* pack four GLubytes into one 32-bit float */
1324 {
1325 GLfloat a[4];
1326 GLuint result[4], ubx, uby, ubz, ubw;
1327 fetch_vector4(&inst->SrcReg[0], machine, a);
1328 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1329 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1330 a[2] = CLAMP(a[2], 0.0F, 1.0F);
1331 a[3] = CLAMP(a[3], 0.0F, 1.0F);
1332 ubx = IROUND(255.0F * a[0]);
1333 uby = IROUND(255.0F * a[1]);
1334 ubz = IROUND(255.0F * a[2]);
1335 ubw = IROUND(255.0F * a[3]);
1336 result[0] =
1337 result[1] =
1338 result[2] =
1339 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1340 store_vector4ui(inst, machine, result);
1341 }
1342 break;
1343 case OPCODE_POW:
1344 {
1345 GLfloat a[4], b[4], result[4];
1346 fetch_vector1(&inst->SrcReg[0], machine, a);
1347 fetch_vector1(&inst->SrcReg[1], machine, b);
1348 result[0] = result[1] = result[2] = result[3]
1349 = (GLfloat) _mesa_pow(a[0], b[0]);
1350 store_vector4(inst, machine, result);
1351 }
1352 break;
1353 case OPCODE_RCP:
1354 {
1355 GLfloat a[4], result[4];
1356 fetch_vector1(&inst->SrcReg[0], machine, a);
1357 if (DEBUG_PROG) {
1358 if (a[0] == 0)
1359 printf("RCP(0)\n");
1360 else if (IS_INF_OR_NAN(a[0]))
1361 printf("RCP(inf)\n");
1362 }
1363 result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
1364 store_vector4(inst, machine, result);
1365 }
1366 break;
1367 case OPCODE_RET: /* return from subroutine (conditional) */
1368 if (eval_condition(machine, inst)) {
1369 if (machine->StackDepth == 0) {
1370 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
1371 }
1372 /* subtract one because of pc++ in the for loop */
1373 pc = machine->CallStack[--machine->StackDepth] - 1;
1374 }
1375 break;
1376 case OPCODE_RFL: /* reflection vector */
1377 {
1378 GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
1379 fetch_vector4(&inst->SrcReg[0], machine, axis);
1380 fetch_vector4(&inst->SrcReg[1], machine, dir);
1381 tmpW = DOT3(axis, axis);
1382 tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
1383 result[0] = tmpX * axis[0] - dir[0];
1384 result[1] = tmpX * axis[1] - dir[1];
1385 result[2] = tmpX * axis[2] - dir[2];
1386 /* result[3] is never written! XXX enforce in parser! */
1387 store_vector4(inst, machine, result);
1388 }
1389 break;
1390 case OPCODE_RSQ: /* 1 / sqrt() */
1391 {
1392 GLfloat a[4], result[4];
1393 fetch_vector1(&inst->SrcReg[0], machine, a);
1394 a[0] = FABSF(a[0]);
1395 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
1396 store_vector4(inst, machine, result);
1397 if (DEBUG_PROG) {
1398 printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
1399 }
1400 }
1401 break;
1402 case OPCODE_SCS: /* sine and cos */
1403 {
1404 GLfloat a[4], result[4];
1405 fetch_vector1(&inst->SrcReg[0], machine, a);
1406 result[0] = (GLfloat) _mesa_cos(a[0]);
1407 result[1] = (GLfloat) _mesa_sin(a[0]);
1408 result[2] = 0.0; /* undefined! */
1409 result[3] = 0.0; /* undefined! */
1410 store_vector4(inst, machine, result);
1411 }
1412 break;
1413 case OPCODE_SEQ: /* set on equal */
1414 {
1415 GLfloat a[4], b[4], result[4];
1416 fetch_vector4(&inst->SrcReg[0], machine, a);
1417 fetch_vector4(&inst->SrcReg[1], machine, b);
1418 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1419 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1420 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1421 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1422 store_vector4(inst, machine, result);
1423 if (DEBUG_PROG) {
1424 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1425 result[0], result[1], result[2], result[3],
1426 a[0], a[1], a[2], a[3],
1427 b[0], b[1], b[2], b[3]);
1428 }
1429 }
1430 break;
1431 case OPCODE_SFL: /* set false, operands ignored */
1432 {
1433 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1434 store_vector4(inst, machine, result);
1435 }
1436 break;
1437 case OPCODE_SGE: /* set on greater or equal */
1438 {
1439 GLfloat a[4], b[4], result[4];
1440 fetch_vector4(&inst->SrcReg[0], machine, a);
1441 fetch_vector4(&inst->SrcReg[1], machine, b);
1442 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1443 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1444 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1445 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1446 store_vector4(inst, machine, result);
1447 if (DEBUG_PROG) {
1448 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1449 result[0], result[1], result[2], result[3],
1450 a[0], a[1], a[2], a[3],
1451 b[0], b[1], b[2], b[3]);
1452 }
1453 }
1454 break;
1455 case OPCODE_SGT: /* set on greater */
1456 {
1457 GLfloat a[4], b[4], result[4];
1458 fetch_vector4(&inst->SrcReg[0], machine, a);
1459 fetch_vector4(&inst->SrcReg[1], machine, b);
1460 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1461 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1462 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1463 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1464 store_vector4(inst, machine, result);
1465 if (DEBUG_PROG) {
1466 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1467 result[0], result[1], result[2], result[3],
1468 a[0], a[1], a[2], a[3],
1469 b[0], b[1], b[2], b[3]);
1470 }
1471 }
1472 break;
1473 case OPCODE_SIN:
1474 {
1475 GLfloat a[4], result[4];
1476 fetch_vector1(&inst->SrcReg[0], machine, a);
1477 result[0] = result[1] = result[2] = result[3]
1478 = (GLfloat) _mesa_sin(a[0]);
1479 store_vector4(inst, machine, result);
1480 }
1481 break;
1482 case OPCODE_SLE: /* set on less or equal */
1483 {
1484 GLfloat a[4], b[4], result[4];
1485 fetch_vector4(&inst->SrcReg[0], machine, a);
1486 fetch_vector4(&inst->SrcReg[1], machine, b);
1487 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1488 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1489 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1490 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1491 store_vector4(inst, machine, result);
1492 if (DEBUG_PROG) {
1493 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1494 result[0], result[1], result[2], result[3],
1495 a[0], a[1], a[2], a[3],
1496 b[0], b[1], b[2], b[3]);
1497 }
1498 }
1499 break;
1500 case OPCODE_SLT: /* set on less */
1501 {
1502 GLfloat a[4], b[4], result[4];
1503 fetch_vector4(&inst->SrcReg[0], machine, a);
1504 fetch_vector4(&inst->SrcReg[1], machine, b);
1505 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1506 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1507 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1508 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1509 store_vector4(inst, machine, result);
1510 if (DEBUG_PROG) {
1511 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1512 result[0], result[1], result[2], result[3],
1513 a[0], a[1], a[2], a[3],
1514 b[0], b[1], b[2], b[3]);
1515 }
1516 }
1517 break;
1518 case OPCODE_SNE: /* set on not equal */
1519 {
1520 GLfloat a[4], b[4], result[4];
1521 fetch_vector4(&inst->SrcReg[0], machine, a);
1522 fetch_vector4(&inst->SrcReg[1], machine, b);
1523 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1524 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1525 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1526 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1527 store_vector4(inst, machine, result);
1528 if (DEBUG_PROG) {
1529 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1530 result[0], result[1], result[2], result[3],
1531 a[0], a[1], a[2], a[3],
1532 b[0], b[1], b[2], b[3]);
1533 }
1534 }
1535 break;
1536 case OPCODE_SSG: /* set sign (-1, 0 or +1) */
1537 {
1538 GLfloat a[4], result[4];
1539 fetch_vector4(&inst->SrcReg[0], machine, a);
1540 result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
1541 result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
1542 result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
1543 result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
1544 store_vector4(inst, machine, result);
1545 }
1546 break;
1547 case OPCODE_STR: /* set true, operands ignored */
1548 {
1549 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1550 store_vector4(inst, machine, result);
1551 }
1552 break;
1553 case OPCODE_SUB:
1554 {
1555 GLfloat a[4], b[4], result[4];
1556 fetch_vector4(&inst->SrcReg[0], machine, a);
1557 fetch_vector4(&inst->SrcReg[1], machine, b);
1558 result[0] = a[0] - b[0];
1559 result[1] = a[1] - b[1];
1560 result[2] = a[2] - b[2];
1561 result[3] = a[3] - b[3];
1562 store_vector4(inst, machine, result);
1563 if (DEBUG_PROG) {
1564 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1565 result[0], result[1], result[2], result[3],
1566 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1567 }
1568 }
1569 break;
1570 case OPCODE_SWZ: /* extended swizzle */
1571 {
1572 const struct prog_src_register *source = &inst->SrcReg[0];
1573 const GLfloat *src = get_src_register_pointer(source, machine);
1574 GLfloat result[4];
1575 GLuint i;
1576 for (i = 0; i < 4; i++) {
1577 const GLuint swz = GET_SWZ(source->Swizzle, i);
1578 if (swz == SWIZZLE_ZERO)
1579 result[i] = 0.0;
1580 else if (swz == SWIZZLE_ONE)
1581 result[i] = 1.0;
1582 else {
1583 ASSERT(swz >= 0);
1584 ASSERT(swz <= 3);
1585 result[i] = src[swz];
1586 }
1587 if (source->Negate & (1 << i))
1588 result[i] = -result[i];
1589 }
1590 store_vector4(inst, machine, result);
1591 }
1592 break;
1593 case OPCODE_TEX: /* Both ARB and NV frag prog */
1594 /* Simple texel lookup */
1595 {
1596 GLfloat texcoord[4], color[4];
1597 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1598
1599 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1600
1601 if (DEBUG_PROG) {
1602 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1603 color[0], color[1], color[2], color[3],
1604 inst->TexSrcUnit,
1605 texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
1606 }
1607 store_vector4(inst, machine, color);
1608 }
1609 break;
1610 case OPCODE_TXB: /* GL_ARB_fragment_program only */
1611 /* Texel lookup with LOD bias */
1612 {
1613 GLfloat texcoord[4], color[4], lodBias;
1614
1615 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1616
1617 /* texcoord[3] is the bias to add to lambda */
1618 lodBias = texcoord[3];
1619
1620 fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
1621
1622 store_vector4(inst, machine, color);
1623 }
1624 break;
1625 case OPCODE_TXD: /* GL_NV_fragment_program only */
1626 /* Texture lookup w/ partial derivatives for LOD */
1627 {
1628 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1629 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1630 fetch_vector4(&inst->SrcReg[1], machine, dtdx);
1631 fetch_vector4(&inst->SrcReg[2], machine, dtdy);
1632 machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
1633 0.0, /* lodBias */
1634 inst->TexSrcUnit, color);
1635 store_vector4(inst, machine, color);
1636 }
1637 break;
1638 case OPCODE_TXP: /* GL_ARB_fragment_program only */
1639 /* Texture lookup w/ projective divide */
1640 {
1641 GLfloat texcoord[4], color[4];
1642
1643 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1644 /* Not so sure about this test - if texcoord[3] is
1645 * zero, we'd probably be fine except for an ASSERT in
1646 * IROUND_POS() which gets triggered by the inf values created.
1647 */
1648 if (texcoord[3] != 0.0) {
1649 texcoord[0] /= texcoord[3];
1650 texcoord[1] /= texcoord[3];
1651 texcoord[2] /= texcoord[3];
1652 }
1653
1654 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1655
1656 store_vector4(inst, machine, color);
1657 }
1658 break;
1659 case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
1660 /* Texture lookup w/ projective divide, as above, but do not
1661 * do the divide by w if sampling from a cube map.
1662 */
1663 {
1664 GLfloat texcoord[4], color[4];
1665
1666 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1667 if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
1668 texcoord[3] != 0.0) {
1669 texcoord[0] /= texcoord[3];
1670 texcoord[1] /= texcoord[3];
1671 texcoord[2] /= texcoord[3];
1672 }
1673
1674 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1675
1676 store_vector4(inst, machine, color);
1677 }
1678 break;
1679 case OPCODE_TRUNC: /* truncate toward zero */
1680 {
1681 GLfloat a[4], result[4];
1682 fetch_vector4(&inst->SrcReg[0], machine, a);
1683 result[0] = (GLfloat) (GLint) a[0];
1684 result[1] = (GLfloat) (GLint) a[1];
1685 result[2] = (GLfloat) (GLint) a[2];
1686 result[3] = (GLfloat) (GLint) a[3];
1687 store_vector4(inst, machine, result);
1688 }
1689 break;
1690 case OPCODE_UP2H: /* unpack two 16-bit floats */
1691 {
1692 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1693 GLfloat result[4];
1694 GLushort hx, hy;
1695 hx = raw & 0xffff;
1696 hy = raw >> 16;
1697 result[0] = result[2] = _mesa_half_to_float(hx);
1698 result[1] = result[3] = _mesa_half_to_float(hy);
1699 store_vector4(inst, machine, result);
1700 }
1701 break;
1702 case OPCODE_UP2US: /* unpack two GLushorts */
1703 {
1704 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1705 GLfloat result[4];
1706 GLushort usx, usy;
1707 usx = raw & 0xffff;
1708 usy = raw >> 16;
1709 result[0] = result[2] = usx * (1.0f / 65535.0f);
1710 result[1] = result[3] = usy * (1.0f / 65535.0f);
1711 store_vector4(inst, machine, result);
1712 }
1713 break;
1714 case OPCODE_UP4B: /* unpack four GLbytes */
1715 {
1716 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1717 GLfloat result[4];
1718 result[0] = (((raw >> 0) & 0xff) - 128) / 127.0F;
1719 result[1] = (((raw >> 8) & 0xff) - 128) / 127.0F;
1720 result[2] = (((raw >> 16) & 0xff) - 128) / 127.0F;
1721 result[3] = (((raw >> 24) & 0xff) - 128) / 127.0F;
1722 store_vector4(inst, machine, result);
1723 }
1724 break;
1725 case OPCODE_UP4UB: /* unpack four GLubytes */
1726 {
1727 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1728 GLfloat result[4];
1729 result[0] = ((raw >> 0) & 0xff) / 255.0F;
1730 result[1] = ((raw >> 8) & 0xff) / 255.0F;
1731 result[2] = ((raw >> 16) & 0xff) / 255.0F;
1732 result[3] = ((raw >> 24) & 0xff) / 255.0F;
1733 store_vector4(inst, machine, result);
1734 }
1735 break;
1736 case OPCODE_XOR: /* bitwise XOR */
1737 {
1738 GLuint a[4], b[4], result[4];
1739 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1740 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1741 result[0] = a[0] ^ b[0];
1742 result[1] = a[1] ^ b[1];
1743 result[2] = a[2] ^ b[2];
1744 result[3] = a[3] ^ b[3];
1745 store_vector4ui(inst, machine, result);
1746 }
1747 break;
1748 case OPCODE_XPD: /* cross product */
1749 {
1750 GLfloat a[4], b[4], result[4];
1751 fetch_vector4(&inst->SrcReg[0], machine, a);
1752 fetch_vector4(&inst->SrcReg[1], machine, b);
1753 result[0] = a[1] * b[2] - a[2] * b[1];
1754 result[1] = a[2] * b[0] - a[0] * b[2];
1755 result[2] = a[0] * b[1] - a[1] * b[0];
1756 result[3] = 1.0;
1757 store_vector4(inst, machine, result);
1758 if (DEBUG_PROG) {
1759 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1760 result[0], result[1], result[2], result[3],
1761 a[0], a[1], a[2], b[0], b[1], b[2]);
1762 }
1763 }
1764 break;
1765 case OPCODE_X2D: /* 2-D matrix transform */
1766 {
1767 GLfloat a[4], b[4], c[4], result[4];
1768 fetch_vector4(&inst->SrcReg[0], machine, a);
1769 fetch_vector4(&inst->SrcReg[1], machine, b);
1770 fetch_vector4(&inst->SrcReg[2], machine, c);
1771 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1772 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1773 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1774 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1775 store_vector4(inst, machine, result);
1776 }
1777 break;
1778 case OPCODE_PRINT:
1779 {
1780 if (inst->SrcReg[0].File != -1) {
1781 GLfloat a[4];
1782 fetch_vector4(&inst->SrcReg[0], machine, a);
1783 _mesa_printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
1784 a[0], a[1], a[2], a[3]);
1785 }
1786 else {
1787 _mesa_printf("%s\n", (const char *) inst->Data);
1788 }
1789 }
1790 break;
1791 case OPCODE_END:
1792 return GL_TRUE;
1793 default:
1794 _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
1795 inst->Opcode);
1796 return GL_TRUE; /* return value doesn't matter */
1797 }
1798
1799 numExec++;
1800 if (numExec > maxExec) {
1801 _mesa_problem(ctx, "Infinite loop detected in fragment program");
1802 return GL_TRUE;
1803 }
1804
1805 } /* for pc */
1806
1807 return GL_TRUE;
1808 }