projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge branch 'glsl2'
[mesa.git] / src / mesa / program / 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/macros.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 TRUE for +0 and other positive values, FALSE otherwise.
85 * Used for RCC opcode.
86 */
87 static INLINE GLboolean
88 positive(float x)
89 {
90 fi_type fi;
91 fi.f = x;
92 if (fi.i & 0x80000000)
93 return GL_FALSE;
94 return GL_TRUE;
95 }
96
97
98
99 /**
100 * Return a pointer to the 4-element float vector specified by the given
101 * source register.
102 */
103 static INLINE const GLfloat *
104 get_src_register_pointer(const struct prog_src_register *source,
105 const struct gl_program_machine *machine)
106 {
107 const struct gl_program *prog = machine->CurProgram;
108 GLint reg = source->Index;
109
110 if (source->RelAddr) {
111 /* add address register value to src index/offset */
112 reg += machine->AddressReg[0][0];
113 if (reg < 0) {
114 return ZeroVec;
115 }
116 }
117
118 switch (source->File) {
119 case PROGRAM_TEMPORARY:
120 if (reg >= MAX_PROGRAM_TEMPS)
121 return ZeroVec;
122 return machine->Temporaries[reg];
123
124 case PROGRAM_INPUT:
125 if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
126 if (reg >= VERT_ATTRIB_MAX)
127 return ZeroVec;
128 return machine->VertAttribs[reg];
129 }
130 else {
131 if (reg >= FRAG_ATTRIB_MAX)
132 return ZeroVec;
133 return machine->Attribs[reg][machine->CurElement];
134 }
135
136 case PROGRAM_OUTPUT:
137 if (reg >= MAX_PROGRAM_OUTPUTS)
138 return ZeroVec;
139 return machine->Outputs[reg];
140
141 case PROGRAM_LOCAL_PARAM:
142 if (reg >= MAX_PROGRAM_LOCAL_PARAMS)
143 return ZeroVec;
144 return machine->CurProgram->LocalParams[reg];
145
146 case PROGRAM_ENV_PARAM:
147 if (reg >= MAX_PROGRAM_ENV_PARAMS)
148 return ZeroVec;
149 return machine->EnvParams[reg];
150
151 case PROGRAM_STATE_VAR:
152 /* Fallthrough */
153 case PROGRAM_CONSTANT:
154 /* Fallthrough */
155 case PROGRAM_UNIFORM:
156 /* Fallthrough */
157 case PROGRAM_NAMED_PARAM:
158 if (reg >= (GLint) prog->Parameters->NumParameters)
159 return ZeroVec;
160 return prog->Parameters->ParameterValues[reg];
161
162 default:
163 _mesa_problem(NULL,
164 "Invalid src register file %d in get_src_register_pointer()",
165 source->File);
166 return NULL;
167 }
168 }
169
170
171 /**
172 * Return a pointer to the 4-element float vector specified by the given
173 * destination register.
174 */
175 static INLINE GLfloat *
176 get_dst_register_pointer(const struct prog_dst_register *dest,
177 struct gl_program_machine *machine)
178 {
179 static GLfloat dummyReg[4];
180 GLint reg = dest->Index;
181
182 if (dest->RelAddr) {
183 /* add address register value to src index/offset */
184 reg += machine->AddressReg[0][0];
185 if (reg < 0) {
186 return dummyReg;
187 }
188 }
189
190 switch (dest->File) {
191 case PROGRAM_TEMPORARY:
192 if (reg >= MAX_PROGRAM_TEMPS)
193 return dummyReg;
194 return machine->Temporaries[reg];
195
196 case PROGRAM_OUTPUT:
197 if (reg >= MAX_PROGRAM_OUTPUTS)
198 return dummyReg;
199 return machine->Outputs[reg];
200
201 case PROGRAM_WRITE_ONLY:
202 return dummyReg;
203
204 default:
205 _mesa_problem(NULL,
206 "Invalid dest register file %d in get_dst_register_pointer()",
207 dest->File);
208 return NULL;
209 }
210 }
211
212
213
214 /**
215 * Fetch a 4-element float vector from the given source register.
216 * Apply swizzling and negating as needed.
217 */
218 static void
219 fetch_vector4(const struct prog_src_register *source,
220 const struct gl_program_machine *machine, GLfloat result[4])
221 {
222 const GLfloat *src = get_src_register_pointer(source, machine);
223 ASSERT(src);
224
225 if (source->Swizzle == SWIZZLE_NOOP) {
226 /* no swizzling */
227 COPY_4V(result, src);
228 }
229 else {
230 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
231 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
232 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
233 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
234 result[0] = src[GET_SWZ(source->Swizzle, 0)];
235 result[1] = src[GET_SWZ(source->Swizzle, 1)];
236 result[2] = src[GET_SWZ(source->Swizzle, 2)];
237 result[3] = src[GET_SWZ(source->Swizzle, 3)];
238 }
239
240 if (source->Abs) {
241 result[0] = FABSF(result[0]);
242 result[1] = FABSF(result[1]);
243 result[2] = FABSF(result[2]);
244 result[3] = FABSF(result[3]);
245 }
246 if (source->Negate) {
247 ASSERT(source->Negate == NEGATE_XYZW);
248 result[0] = -result[0];
249 result[1] = -result[1];
250 result[2] = -result[2];
251 result[3] = -result[3];
252 }
253
254 #ifdef NAN_CHECK
255 assert(!IS_INF_OR_NAN(result[0]));
256 assert(!IS_INF_OR_NAN(result[0]));
257 assert(!IS_INF_OR_NAN(result[0]));
258 assert(!IS_INF_OR_NAN(result[0]));
259 #endif
260 }
261
262
263 /**
264 * Fetch a 4-element uint vector from the given source register.
265 * Apply swizzling but not negation/abs.
266 */
267 static void
268 fetch_vector4ui(const struct prog_src_register *source,
269 const struct gl_program_machine *machine, GLuint result[4])
270 {
271 const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
272 ASSERT(src);
273
274 if (source->Swizzle == SWIZZLE_NOOP) {
275 /* no swizzling */
276 COPY_4V(result, src);
277 }
278 else {
279 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
280 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
281 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
282 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
283 result[0] = src[GET_SWZ(source->Swizzle, 0)];
284 result[1] = src[GET_SWZ(source->Swizzle, 1)];
285 result[2] = src[GET_SWZ(source->Swizzle, 2)];
286 result[3] = src[GET_SWZ(source->Swizzle, 3)];
287 }
288
289 /* Note: no Negate or Abs here */
290 }
291
292
293
294 /**
295 * Fetch the derivative with respect to X or Y for the given register.
296 * XXX this currently only works for fragment program input attribs.
297 */
298 static void
299 fetch_vector4_deriv(GLcontext * ctx,
300 const struct prog_src_register *source,
301 const struct gl_program_machine *machine,
302 char xOrY, GLfloat result[4])
303 {
304 if (source->File == PROGRAM_INPUT &&
305 source->Index < (GLint) machine->NumDeriv) {
306 const GLint col = machine->CurElement;
307 const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
308 const GLfloat invQ = 1.0f / w;
309 GLfloat deriv[4];
310
311 if (xOrY == 'X') {
312 deriv[0] = machine->DerivX[source->Index][0] * invQ;
313 deriv[1] = machine->DerivX[source->Index][1] * invQ;
314 deriv[2] = machine->DerivX[source->Index][2] * invQ;
315 deriv[3] = machine->DerivX[source->Index][3] * invQ;
316 }
317 else {
318 deriv[0] = machine->DerivY[source->Index][0] * invQ;
319 deriv[1] = machine->DerivY[source->Index][1] * invQ;
320 deriv[2] = machine->DerivY[source->Index][2] * invQ;
321 deriv[3] = machine->DerivY[source->Index][3] * invQ;
322 }
323
324 result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
325 result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
326 result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
327 result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
328
329 if (source->Abs) {
330 result[0] = FABSF(result[0]);
331 result[1] = FABSF(result[1]);
332 result[2] = FABSF(result[2]);
333 result[3] = FABSF(result[3]);
334 }
335 if (source->Negate) {
336 ASSERT(source->Negate == NEGATE_XYZW);
337 result[0] = -result[0];
338 result[1] = -result[1];
339 result[2] = -result[2];
340 result[3] = -result[3];
341 }
342 }
343 else {
344 ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
345 }
346 }
347
348
349 /**
350 * As above, but only return result[0] element.
351 */
352 static void
353 fetch_vector1(const struct prog_src_register *source,
354 const struct gl_program_machine *machine, GLfloat result[4])
355 {
356 const GLfloat *src = get_src_register_pointer(source, machine);
357 ASSERT(src);
358
359 result[0] = src[GET_SWZ(source->Swizzle, 0)];
360
361 if (source->Abs) {
362 result[0] = FABSF(result[0]);
363 }
364 if (source->Negate) {
365 result[0] = -result[0];
366 }
367 }
368
369
370 static GLuint
371 fetch_vector1ui(const struct prog_src_register *source,
372 const struct gl_program_machine *machine)
373 {
374 const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
375 return src[GET_SWZ(source->Swizzle, 0)];
376 }
377
378
379 /**
380 * Fetch texel from texture. Use partial derivatives when possible.
381 */
382 static INLINE void
383 fetch_texel(GLcontext *ctx,
384 const struct gl_program_machine *machine,
385 const struct prog_instruction *inst,
386 const GLfloat texcoord[4], GLfloat lodBias,
387 GLfloat color[4])
388 {
389 const GLuint unit = machine->Samplers[inst->TexSrcUnit];
390
391 /* Note: we only have the right derivatives for fragment input attribs.
392 */
393 if (machine->NumDeriv > 0 &&
394 inst->SrcReg[0].File == PROGRAM_INPUT &&
395 inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
396 /* simple texture fetch for which we should have derivatives */
397 GLuint attr = inst->SrcReg[0].Index;
398 machine->FetchTexelDeriv(ctx, texcoord,
399 machine->DerivX[attr],
400 machine->DerivY[attr],
401 lodBias, unit, color);
402 }
403 else {
404 machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
405 }
406 }
407
408
409 /**
410 * Test value against zero and return GT, LT, EQ or UN if NaN.
411 */
412 static INLINE GLuint
413 generate_cc(float value)
414 {
415 if (value != value)
416 return COND_UN; /* NaN */
417 if (value > 0.0F)
418 return COND_GT;
419 if (value < 0.0F)
420 return COND_LT;
421 return COND_EQ;
422 }
423
424
425 /**
426 * Test if the ccMaskRule is satisfied by the given condition code.
427 * Used to mask destination writes according to the current condition code.
428 */
429 static INLINE GLboolean
430 test_cc(GLuint condCode, GLuint ccMaskRule)
431 {
432 switch (ccMaskRule) {
433 case COND_EQ: return (condCode == COND_EQ);
434 case COND_NE: return (condCode != COND_EQ);
435 case COND_LT: return (condCode == COND_LT);
436 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
437 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
438 case COND_GT: return (condCode == COND_GT);
439 case COND_TR: return GL_TRUE;
440 case COND_FL: return GL_FALSE;
441 default: return GL_TRUE;
442 }
443 }
444
445
446 /**
447 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
448 * or GL_FALSE to indicate result.
449 */
450 static INLINE GLboolean
451 eval_condition(const struct gl_program_machine *machine,
452 const struct prog_instruction *inst)
453 {
454 const GLuint swizzle = inst->DstReg.CondSwizzle;
455 const GLuint condMask = inst->DstReg.CondMask;
456 if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
457 test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
458 test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
459 test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
460 return GL_TRUE;
461 }
462 else {
463 return GL_FALSE;
464 }
465 }
466
467
468
469 /**
470 * Store 4 floats into a register. Observe the instructions saturate and
471 * set-condition-code flags.
472 */
473 static void
474 store_vector4(const struct prog_instruction *inst,
475 struct gl_program_machine *machine, const GLfloat value[4])
476 {
477 const struct prog_dst_register *dstReg = &(inst->DstReg);
478 const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
479 GLuint writeMask = dstReg->WriteMask;
480 GLfloat clampedValue[4];
481 GLfloat *dst = get_dst_register_pointer(dstReg, machine);
482
483 #if 0
484 if (value[0] > 1.0e10 ||
485 IS_INF_OR_NAN(value[0]) ||
486 IS_INF_OR_NAN(value[1]) ||
487 IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
488 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
489 #endif
490
491 if (clamp) {
492 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
493 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
494 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
495 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
496 value = clampedValue;
497 }
498
499 if (dstReg->CondMask != COND_TR) {
500 /* condition codes may turn off some writes */
501 if (writeMask & WRITEMASK_X) {
502 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
503 dstReg->CondMask))
504 writeMask &= ~WRITEMASK_X;
505 }
506 if (writeMask & WRITEMASK_Y) {
507 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
508 dstReg->CondMask))
509 writeMask &= ~WRITEMASK_Y;
510 }
511 if (writeMask & WRITEMASK_Z) {
512 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
513 dstReg->CondMask))
514 writeMask &= ~WRITEMASK_Z;
515 }
516 if (writeMask & WRITEMASK_W) {
517 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
518 dstReg->CondMask))
519 writeMask &= ~WRITEMASK_W;
520 }
521 }
522
523 #ifdef NAN_CHECK
524 assert(!IS_INF_OR_NAN(value[0]));
525 assert(!IS_INF_OR_NAN(value[0]));
526 assert(!IS_INF_OR_NAN(value[0]));
527 assert(!IS_INF_OR_NAN(value[0]));
528 #endif
529
530 if (writeMask & WRITEMASK_X)
531 dst[0] = value[0];
532 if (writeMask & WRITEMASK_Y)
533 dst[1] = value[1];
534 if (writeMask & WRITEMASK_Z)
535 dst[2] = value[2];
536 if (writeMask & WRITEMASK_W)
537 dst[3] = value[3];
538
539 if (inst->CondUpdate) {
540 if (writeMask & WRITEMASK_X)
541 machine->CondCodes[0] = generate_cc(value[0]);
542 if (writeMask & WRITEMASK_Y)
543 machine->CondCodes[1] = generate_cc(value[1]);
544 if (writeMask & WRITEMASK_Z)
545 machine->CondCodes[2] = generate_cc(value[2]);
546 if (writeMask & WRITEMASK_W)
547 machine->CondCodes[3] = generate_cc(value[3]);
548 #if DEBUG_PROG
549 printf("CondCodes=(%s,%s,%s,%s) for:\n",
550 _mesa_condcode_string(machine->CondCodes[0]),
551 _mesa_condcode_string(machine->CondCodes[1]),
552 _mesa_condcode_string(machine->CondCodes[2]),
553 _mesa_condcode_string(machine->CondCodes[3]));
554 #endif
555 }
556 }
557
558
559 /**
560 * Store 4 uints into a register. Observe the set-condition-code flags.
561 */
562 static void
563 store_vector4ui(const struct prog_instruction *inst,
564 struct gl_program_machine *machine, const GLuint value[4])
565 {
566 const struct prog_dst_register *dstReg = &(inst->DstReg);
567 GLuint writeMask = dstReg->WriteMask;
568 GLuint *dst = (GLuint *) get_dst_register_pointer(dstReg, machine);
569
570 if (dstReg->CondMask != COND_TR) {
571 /* condition codes may turn off some writes */
572 if (writeMask & WRITEMASK_X) {
573 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
574 dstReg->CondMask))
575 writeMask &= ~WRITEMASK_X;
576 }
577 if (writeMask & WRITEMASK_Y) {
578 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
579 dstReg->CondMask))
580 writeMask &= ~WRITEMASK_Y;
581 }
582 if (writeMask & WRITEMASK_Z) {
583 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
584 dstReg->CondMask))
585 writeMask &= ~WRITEMASK_Z;
586 }
587 if (writeMask & WRITEMASK_W) {
588 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
589 dstReg->CondMask))
590 writeMask &= ~WRITEMASK_W;
591 }
592 }
593
594 if (writeMask & WRITEMASK_X)
595 dst[0] = value[0];
596 if (writeMask & WRITEMASK_Y)
597 dst[1] = value[1];
598 if (writeMask & WRITEMASK_Z)
599 dst[2] = value[2];
600 if (writeMask & WRITEMASK_W)
601 dst[3] = value[3];
602
603 if (inst->CondUpdate) {
604 if (writeMask & WRITEMASK_X)
605 machine->CondCodes[0] = generate_cc((float)value[0]);
606 if (writeMask & WRITEMASK_Y)
607 machine->CondCodes[1] = generate_cc((float)value[1]);
608 if (writeMask & WRITEMASK_Z)
609 machine->CondCodes[2] = generate_cc((float)value[2]);
610 if (writeMask & WRITEMASK_W)
611 machine->CondCodes[3] = generate_cc((float)value[3]);
612 #if DEBUG_PROG
613 printf("CondCodes=(%s,%s,%s,%s) for:\n",
614 _mesa_condcode_string(machine->CondCodes[0]),
615 _mesa_condcode_string(machine->CondCodes[1]),
616 _mesa_condcode_string(machine->CondCodes[2]),
617 _mesa_condcode_string(machine->CondCodes[3]));
618 #endif
619 }
620 }
621
622
623
624 /**
625 * Execute the given vertex/fragment program.
626 *
627 * \param ctx rendering context
628 * \param program the program to execute
629 * \param machine machine state (must be initialized)
630 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
631 */
632 GLboolean
633 _mesa_execute_program(GLcontext * ctx,
634 const struct gl_program *program,
635 struct gl_program_machine *machine)
636 {
637 const GLuint numInst = program->NumInstructions;
638 const GLuint maxExec = 10000;
639 GLuint pc, numExec = 0;
640
641 machine->CurProgram = program;
642
643 if (DEBUG_PROG) {
644 printf("execute program %u --------------------\n", program->Id);
645 }
646
647 if (program->Target == GL_VERTEX_PROGRAM_ARB) {
648 machine->EnvParams = ctx->VertexProgram.Parameters;
649 }
650 else {
651 machine->EnvParams = ctx->FragmentProgram.Parameters;
652 }
653
654 for (pc = 0; pc < numInst; pc++) {
655 const struct prog_instruction *inst = program->Instructions + pc;
656
657 if (DEBUG_PROG) {
658 _mesa_print_instruction(inst);
659 }
660
661 switch (inst->Opcode) {
662 case OPCODE_ABS:
663 {
664 GLfloat a[4], result[4];
665 fetch_vector4(&inst->SrcReg[0], machine, a);
666 result[0] = FABSF(a[0]);
667 result[1] = FABSF(a[1]);
668 result[2] = FABSF(a[2]);
669 result[3] = FABSF(a[3]);
670 store_vector4(inst, machine, result);
671 }
672 break;
673 case OPCODE_ADD:
674 {
675 GLfloat a[4], b[4], result[4];
676 fetch_vector4(&inst->SrcReg[0], machine, a);
677 fetch_vector4(&inst->SrcReg[1], machine, b);
678 result[0] = a[0] + b[0];
679 result[1] = a[1] + b[1];
680 result[2] = a[2] + b[2];
681 result[3] = a[3] + b[3];
682 store_vector4(inst, machine, result);
683 if (DEBUG_PROG) {
684 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
685 result[0], result[1], result[2], result[3],
686 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
687 }
688 }
689 break;
690 case OPCODE_AND: /* bitwise AND */
691 {
692 GLuint a[4], b[4], result[4];
693 fetch_vector4ui(&inst->SrcReg[0], machine, a);
694 fetch_vector4ui(&inst->SrcReg[1], machine, b);
695 result[0] = a[0] & b[0];
696 result[1] = a[1] & b[1];
697 result[2] = a[2] & b[2];
698 result[3] = a[3] & b[3];
699 store_vector4ui(inst, machine, result);
700 }
701 break;
702 case OPCODE_ARL:
703 {
704 GLfloat t[4];
705 fetch_vector4(&inst->SrcReg[0], machine, t);
706 machine->AddressReg[0][0] = IFLOOR(t[0]);
707 if (DEBUG_PROG) {
708 printf("ARL %d\n", machine->AddressReg[0][0]);
709 }
710 }
711 break;
712 case OPCODE_BGNLOOP:
713 /* no-op */
714 ASSERT(program->Instructions[inst->BranchTarget].Opcode
715 == OPCODE_ENDLOOP);
716 break;
717 case OPCODE_ENDLOOP:
718 /* subtract 1 here since pc is incremented by for(pc) loop */
719 ASSERT(program->Instructions[inst->BranchTarget].Opcode
720 == OPCODE_BGNLOOP);
721 pc = inst->BranchTarget - 1; /* go to matching BNGLOOP */
722 break;
723 case OPCODE_BGNSUB: /* begin subroutine */
724 break;
725 case OPCODE_ENDSUB: /* end subroutine */
726 break;
727 case OPCODE_BRA: /* branch (conditional) */
728 if (eval_condition(machine, inst)) {
729 /* take branch */
730 /* Subtract 1 here since we'll do pc++ below */
731 pc = inst->BranchTarget - 1;
732 }
733 break;
734 case OPCODE_BRK: /* break out of loop (conditional) */
735 ASSERT(program->Instructions[inst->BranchTarget].Opcode
736 == OPCODE_ENDLOOP);
737 if (eval_condition(machine, inst)) {
738 /* break out of loop */
739 /* pc++ at end of for-loop will put us after the ENDLOOP inst */
740 pc = inst->BranchTarget;
741 }
742 break;
743 case OPCODE_CONT: /* continue loop (conditional) */
744 ASSERT(program->Instructions[inst->BranchTarget].Opcode
745 == OPCODE_ENDLOOP);
746 if (eval_condition(machine, inst)) {
747 /* continue at ENDLOOP */
748 /* Subtract 1 here since we'll do pc++ at end of for-loop */
749 pc = inst->BranchTarget - 1;
750 }
751 break;
752 case OPCODE_CAL: /* Call subroutine (conditional) */
753 if (eval_condition(machine, inst)) {
754 /* call the subroutine */
755 if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
756 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
757 }
758 machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
759 /* Subtract 1 here since we'll do pc++ at end of for-loop */
760 pc = inst->BranchTarget - 1;
761 }
762 break;
763 case OPCODE_CMP:
764 {
765 GLfloat a[4], b[4], c[4], result[4];
766 fetch_vector4(&inst->SrcReg[0], machine, a);
767 fetch_vector4(&inst->SrcReg[1], machine, b);
768 fetch_vector4(&inst->SrcReg[2], machine, c);
769 result[0] = a[0] < 0.0F ? b[0] : c[0];
770 result[1] = a[1] < 0.0F ? b[1] : c[1];
771 result[2] = a[2] < 0.0F ? b[2] : c[2];
772 result[3] = a[3] < 0.0F ? b[3] : c[3];
773 store_vector4(inst, machine, result);
774 }
775 break;
776 case OPCODE_COS:
777 {
778 GLfloat a[4], result[4];
779 fetch_vector1(&inst->SrcReg[0], machine, a);
780 result[0] = result[1] = result[2] = result[3]
781 = (GLfloat) cos(a[0]);
782 store_vector4(inst, machine, result);
783 }
784 break;
785 case OPCODE_DDX: /* Partial derivative with respect to X */
786 {
787 GLfloat result[4];
788 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
789 'X', result);
790 store_vector4(inst, machine, result);
791 }
792 break;
793 case OPCODE_DDY: /* Partial derivative with respect to Y */
794 {
795 GLfloat result[4];
796 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
797 'Y', result);
798 store_vector4(inst, machine, result);
799 }
800 break;
801 case OPCODE_DP2:
802 {
803 GLfloat a[4], b[4], result[4];
804 fetch_vector4(&inst->SrcReg[0], machine, a);
805 fetch_vector4(&inst->SrcReg[1], machine, b);
806 result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
807 store_vector4(inst, machine, result);
808 if (DEBUG_PROG) {
809 printf("DP2 %g = (%g %g) . (%g %g)\n",
810 result[0], a[0], a[1], b[0], b[1]);
811 }
812 }
813 break;
814 case OPCODE_DP2A:
815 {
816 GLfloat a[4], b[4], c, result[4];
817 fetch_vector4(&inst->SrcReg[0], machine, a);
818 fetch_vector4(&inst->SrcReg[1], machine, b);
819 fetch_vector1(&inst->SrcReg[1], machine, &c);
820 result[0] = result[1] = result[2] = result[3] = DOT2(a, b) + c;
821 store_vector4(inst, machine, result);
822 if (DEBUG_PROG) {
823 printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
824 result[0], a[0], a[1], b[0], b[1], c);
825 }
826 }
827 break;
828 case OPCODE_DP3:
829 {
830 GLfloat a[4], b[4], result[4];
831 fetch_vector4(&inst->SrcReg[0], machine, a);
832 fetch_vector4(&inst->SrcReg[1], machine, b);
833 result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
834 store_vector4(inst, machine, result);
835 if (DEBUG_PROG) {
836 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
837 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
838 }
839 }
840 break;
841 case OPCODE_DP4:
842 {
843 GLfloat a[4], b[4], result[4];
844 fetch_vector4(&inst->SrcReg[0], machine, a);
845 fetch_vector4(&inst->SrcReg[1], machine, b);
846 result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
847 store_vector4(inst, machine, result);
848 if (DEBUG_PROG) {
849 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
850 result[0], a[0], a[1], a[2], a[3],
851 b[0], b[1], b[2], b[3]);
852 }
853 }
854 break;
855 case OPCODE_DPH:
856 {
857 GLfloat a[4], b[4], result[4];
858 fetch_vector4(&inst->SrcReg[0], machine, a);
859 fetch_vector4(&inst->SrcReg[1], machine, b);
860 result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
861 store_vector4(inst, machine, result);
862 }
863 break;
864 case OPCODE_DST: /* Distance vector */
865 {
866 GLfloat a[4], b[4], result[4];
867 fetch_vector4(&inst->SrcReg[0], machine, a);
868 fetch_vector4(&inst->SrcReg[1], machine, b);
869 result[0] = 1.0F;
870 result[1] = a[1] * b[1];
871 result[2] = a[2];
872 result[3] = b[3];
873 store_vector4(inst, machine, result);
874 }
875 break;
876 case OPCODE_EXP:
877 {
878 GLfloat t[4], q[4], floor_t0;
879 fetch_vector1(&inst->SrcReg[0], machine, t);
880 floor_t0 = FLOORF(t[0]);
881 if (floor_t0 > FLT_MAX_EXP) {
882 SET_POS_INFINITY(q[0]);
883 SET_POS_INFINITY(q[2]);
884 }
885 else if (floor_t0 < FLT_MIN_EXP) {
886 q[0] = 0.0F;
887 q[2] = 0.0F;
888 }
889 else {
890 q[0] = LDEXPF(1.0, (int) floor_t0);
891 /* Note: GL_NV_vertex_program expects
892 * result.z = result.x * APPX(result.y)
893 * We do what the ARB extension says.
894 */
895 q[2] = (GLfloat) pow(2.0, t[0]);
896 }
897 q[1] = t[0] - floor_t0;
898 q[3] = 1.0F;
899 store_vector4( inst, machine, q );
900 }
901 break;
902 case OPCODE_EX2: /* Exponential base 2 */
903 {
904 GLfloat a[4], result[4], val;
905 fetch_vector1(&inst->SrcReg[0], machine, a);
906 val = (GLfloat) pow(2.0, a[0]);
907 /*
908 if (IS_INF_OR_NAN(val))
909 val = 1.0e10;
910 */
911 result[0] = result[1] = result[2] = result[3] = val;
912 store_vector4(inst, machine, result);
913 }
914 break;
915 case OPCODE_FLR:
916 {
917 GLfloat a[4], result[4];
918 fetch_vector4(&inst->SrcReg[0], machine, a);
919 result[0] = FLOORF(a[0]);
920 result[1] = FLOORF(a[1]);
921 result[2] = FLOORF(a[2]);
922 result[3] = FLOORF(a[3]);
923 store_vector4(inst, machine, result);
924 }
925 break;
926 case OPCODE_FRC:
927 {
928 GLfloat a[4], result[4];
929 fetch_vector4(&inst->SrcReg[0], machine, a);
930 result[0] = a[0] - FLOORF(a[0]);
931 result[1] = a[1] - FLOORF(a[1]);
932 result[2] = a[2] - FLOORF(a[2]);
933 result[3] = a[3] - FLOORF(a[3]);
934 store_vector4(inst, machine, result);
935 }
936 break;
937 case OPCODE_IF:
938 {
939 GLboolean cond;
940 ASSERT(program->Instructions[inst->BranchTarget].Opcode
941 == OPCODE_ELSE ||
942 program->Instructions[inst->BranchTarget].Opcode
943 == OPCODE_ENDIF);
944 /* eval condition */
945 if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
946 GLfloat a[4];
947 fetch_vector1(&inst->SrcReg[0], machine, a);
948 cond = (a[0] != 0.0);
949 }
950 else {
951 cond = eval_condition(machine, inst);
952 }
953 if (DEBUG_PROG) {
954 printf("IF: %d\n", cond);
955 }
956 /* do if/else */
957 if (cond) {
958 /* do if-clause (just continue execution) */
959 }
960 else {
961 /* go to the instruction after ELSE or ENDIF */
962 assert(inst->BranchTarget >= 0);
963 pc = inst->BranchTarget;
964 }
965 }
966 break;
967 case OPCODE_ELSE:
968 /* goto ENDIF */
969 ASSERT(program->Instructions[inst->BranchTarget].Opcode
970 == OPCODE_ENDIF);
971 assert(inst->BranchTarget >= 0);
972 pc = inst->BranchTarget;
973 break;
974 case OPCODE_ENDIF:
975 /* nothing */
976 break;
977 case OPCODE_KIL_NV: /* NV_f_p only (conditional) */
978 if (eval_condition(machine, inst)) {
979 return GL_FALSE;
980 }
981 break;
982 case OPCODE_KIL: /* ARB_f_p only */
983 {
984 GLfloat a[4];
985 fetch_vector4(&inst->SrcReg[0], machine, a);
986 if (DEBUG_PROG) {
987 printf("KIL if (%g %g %g %g) <= 0.0\n",
988 a[0], a[1], a[2], a[3]);
989 }
990
991 if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
992 return GL_FALSE;
993 }
994 }
995 break;
996 case OPCODE_LG2: /* log base 2 */
997 {
998 GLfloat a[4], result[4], val;
999 fetch_vector1(&inst->SrcReg[0], machine, a);
1000 /* The fast LOG2 macro doesn't meet the precision requirements.
1001 */
1002 if (a[0] == 0.0F) {
1003 val = -FLT_MAX;
1004 }
1005 else {
1006 val = (float)(log(a[0]) * 1.442695F);
1007 }
1008 result[0] = result[1] = result[2] = result[3] = val;
1009 store_vector4(inst, machine, result);
1010 }
1011 break;
1012 case OPCODE_LIT:
1013 {
1014 const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
1015 GLfloat a[4], result[4];
1016 fetch_vector4(&inst->SrcReg[0], machine, a);
1017 a[0] = MAX2(a[0], 0.0F);
1018 a[1] = MAX2(a[1], 0.0F);
1019 /* XXX ARB version clamps a[3], NV version doesn't */
1020 a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
1021 result[0] = 1.0F;
1022 result[1] = a[0];
1023 /* XXX we could probably just use pow() here */
1024 if (a[0] > 0.0F) {
1025 if (a[1] == 0.0 && a[3] == 0.0)
1026 result[2] = 1.0F;
1027 else
1028 result[2] = (GLfloat) pow(a[1], a[3]);
1029 }
1030 else {
1031 result[2] = 0.0F;
1032 }
1033 result[3] = 1.0F;
1034 store_vector4(inst, machine, result);
1035 if (DEBUG_PROG) {
1036 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
1037 result[0], result[1], result[2], result[3],
1038 a[0], a[1], a[2], a[3]);
1039 }
1040 }
1041 break;
1042 case OPCODE_LOG:
1043 {
1044 GLfloat t[4], q[4], abs_t0;
1045 fetch_vector1(&inst->SrcReg[0], machine, t);
1046 abs_t0 = FABSF(t[0]);
1047 if (abs_t0 != 0.0F) {
1048 /* Since we really can't handle infinite values on VMS
1049 * like other OSes we'll use __MAXFLOAT to represent
1050 * infinity. This may need some tweaking.
1051 */
1052 #ifdef VMS
1053 if (abs_t0 == __MAXFLOAT)
1054 #else
1055 if (IS_INF_OR_NAN(abs_t0))
1056 #endif
1057 {
1058 SET_POS_INFINITY(q[0]);
1059 q[1] = 1.0F;
1060 SET_POS_INFINITY(q[2]);
1061 }
1062 else {
1063 int exponent;
1064 GLfloat mantissa = FREXPF(t[0], &exponent);
1065 q[0] = (GLfloat) (exponent - 1);
1066 q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
1067
1068 /* The fast LOG2 macro doesn't meet the precision
1069 * requirements.
1070 */
1071 q[2] = (float)(log(t[0]) * 1.442695F);
1072 }
1073 }
1074 else {
1075 SET_NEG_INFINITY(q[0]);
1076 q[1] = 1.0F;
1077 SET_NEG_INFINITY(q[2]);
1078 }
1079 q[3] = 1.0;
1080 store_vector4(inst, machine, q);
1081 }
1082 break;
1083 case OPCODE_LRP:
1084 {
1085 GLfloat a[4], b[4], c[4], result[4];
1086 fetch_vector4(&inst->SrcReg[0], machine, a);
1087 fetch_vector4(&inst->SrcReg[1], machine, b);
1088 fetch_vector4(&inst->SrcReg[2], machine, c);
1089 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
1090 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
1091 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
1092 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
1093 store_vector4(inst, machine, result);
1094 if (DEBUG_PROG) {
1095 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1096 "(%g %g %g %g), (%g %g %g %g)\n",
1097 result[0], result[1], result[2], result[3],
1098 a[0], a[1], a[2], a[3],
1099 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1100 }
1101 }
1102 break;
1103 case OPCODE_MAD:
1104 {
1105 GLfloat a[4], b[4], c[4], result[4];
1106 fetch_vector4(&inst->SrcReg[0], machine, a);
1107 fetch_vector4(&inst->SrcReg[1], machine, b);
1108 fetch_vector4(&inst->SrcReg[2], machine, c);
1109 result[0] = a[0] * b[0] + c[0];
1110 result[1] = a[1] * b[1] + c[1];
1111 result[2] = a[2] * b[2] + c[2];
1112 result[3] = a[3] * b[3] + c[3];
1113 store_vector4(inst, machine, result);
1114 if (DEBUG_PROG) {
1115 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1116 "(%g %g %g %g) + (%g %g %g %g)\n",
1117 result[0], result[1], result[2], result[3],
1118 a[0], a[1], a[2], a[3],
1119 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1120 }
1121 }
1122 break;
1123 case OPCODE_MAX:
1124 {
1125 GLfloat a[4], b[4], result[4];
1126 fetch_vector4(&inst->SrcReg[0], machine, a);
1127 fetch_vector4(&inst->SrcReg[1], machine, b);
1128 result[0] = MAX2(a[0], b[0]);
1129 result[1] = MAX2(a[1], b[1]);
1130 result[2] = MAX2(a[2], b[2]);
1131 result[3] = MAX2(a[3], b[3]);
1132 store_vector4(inst, machine, result);
1133 if (DEBUG_PROG) {
1134 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1135 result[0], result[1], result[2], result[3],
1136 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1137 }
1138 }
1139 break;
1140 case OPCODE_MIN:
1141 {
1142 GLfloat a[4], b[4], result[4];
1143 fetch_vector4(&inst->SrcReg[0], machine, a);
1144 fetch_vector4(&inst->SrcReg[1], machine, b);
1145 result[0] = MIN2(a[0], b[0]);
1146 result[1] = MIN2(a[1], b[1]);
1147 result[2] = MIN2(a[2], b[2]);
1148 result[3] = MIN2(a[3], b[3]);
1149 store_vector4(inst, machine, result);
1150 }
1151 break;
1152 case OPCODE_MOV:
1153 {
1154 GLfloat result[4];
1155 fetch_vector4(&inst->SrcReg[0], machine, result);
1156 store_vector4(inst, machine, result);
1157 if (DEBUG_PROG) {
1158 printf("MOV (%g %g %g %g)\n",
1159 result[0], result[1], result[2], result[3]);
1160 }
1161 }
1162 break;
1163 case OPCODE_MUL:
1164 {
1165 GLfloat a[4], b[4], result[4];
1166 fetch_vector4(&inst->SrcReg[0], machine, a);
1167 fetch_vector4(&inst->SrcReg[1], machine, b);
1168 result[0] = a[0] * b[0];
1169 result[1] = a[1] * b[1];
1170 result[2] = a[2] * b[2];
1171 result[3] = a[3] * b[3];
1172 store_vector4(inst, machine, result);
1173 if (DEBUG_PROG) {
1174 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1175 result[0], result[1], result[2], result[3],
1176 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1177 }
1178 }
1179 break;
1180 case OPCODE_NOISE1:
1181 {
1182 GLfloat a[4], result[4];
1183 fetch_vector1(&inst->SrcReg[0], machine, a);
1184 result[0] =
1185 result[1] =
1186 result[2] =
1187 result[3] = _mesa_noise1(a[0]);
1188 store_vector4(inst, machine, result);
1189 }
1190 break;
1191 case OPCODE_NOISE2:
1192 {
1193 GLfloat a[4], result[4];
1194 fetch_vector4(&inst->SrcReg[0], machine, a);
1195 result[0] =
1196 result[1] =
1197 result[2] = result[3] = _mesa_noise2(a[0], a[1]);
1198 store_vector4(inst, machine, result);
1199 }
1200 break;
1201 case OPCODE_NOISE3:
1202 {
1203 GLfloat a[4], result[4];
1204 fetch_vector4(&inst->SrcReg[0], machine, a);
1205 result[0] =
1206 result[1] =
1207 result[2] =
1208 result[3] = _mesa_noise3(a[0], a[1], a[2]);
1209 store_vector4(inst, machine, result);
1210 }
1211 break;
1212 case OPCODE_NOISE4:
1213 {
1214 GLfloat a[4], result[4];
1215 fetch_vector4(&inst->SrcReg[0], machine, a);
1216 result[0] =
1217 result[1] =
1218 result[2] =
1219 result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
1220 store_vector4(inst, machine, result);
1221 }
1222 break;
1223 case OPCODE_NOP:
1224 break;
1225 case OPCODE_NOT: /* bitwise NOT */
1226 {
1227 GLuint a[4], result[4];
1228 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1229 result[0] = ~a[0];
1230 result[1] = ~a[1];
1231 result[2] = ~a[2];
1232 result[3] = ~a[3];
1233 store_vector4ui(inst, machine, result);
1234 }
1235 break;
1236 case OPCODE_NRM3: /* 3-component normalization */
1237 {
1238 GLfloat a[4], result[4];
1239 GLfloat tmp;
1240 fetch_vector4(&inst->SrcReg[0], machine, a);
1241 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
1242 if (tmp != 0.0F)
1243 tmp = INV_SQRTF(tmp);
1244 result[0] = tmp * a[0];
1245 result[1] = tmp * a[1];
1246 result[2] = tmp * a[2];
1247 result[3] = 0.0; /* undefined, but prevent valgrind warnings */
1248 store_vector4(inst, machine, result);
1249 }
1250 break;
1251 case OPCODE_NRM4: /* 4-component normalization */
1252 {
1253 GLfloat a[4], result[4];
1254 GLfloat tmp;
1255 fetch_vector4(&inst->SrcReg[0], machine, a);
1256 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
1257 if (tmp != 0.0F)
1258 tmp = INV_SQRTF(tmp);
1259 result[0] = tmp * a[0];
1260 result[1] = tmp * a[1];
1261 result[2] = tmp * a[2];
1262 result[3] = tmp * a[3];
1263 store_vector4(inst, machine, result);
1264 }
1265 break;
1266 case OPCODE_OR: /* bitwise OR */
1267 {
1268 GLuint a[4], b[4], result[4];
1269 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1270 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1271 result[0] = a[0] | b[0];
1272 result[1] = a[1] | b[1];
1273 result[2] = a[2] | b[2];
1274 result[3] = a[3] | b[3];
1275 store_vector4ui(inst, machine, result);
1276 }
1277 break;
1278 case OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
1279 {
1280 GLfloat a[4];
1281 GLuint result[4];
1282 GLhalfNV hx, hy;
1283 fetch_vector4(&inst->SrcReg[0], machine, a);
1284 hx = _mesa_float_to_half(a[0]);
1285 hy = _mesa_float_to_half(a[1]);
1286 result[0] =
1287 result[1] =
1288 result[2] =
1289 result[3] = hx | (hy << 16);
1290 store_vector4ui(inst, machine, result);
1291 }
1292 break;
1293 case OPCODE_PK2US: /* pack two GLushorts into one 32-bit float */
1294 {
1295 GLfloat a[4];
1296 GLuint result[4], usx, usy;
1297 fetch_vector4(&inst->SrcReg[0], machine, a);
1298 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1299 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1300 usx = IROUND(a[0] * 65535.0F);
1301 usy = IROUND(a[1] * 65535.0F);
1302 result[0] =
1303 result[1] =
1304 result[2] =
1305 result[3] = usx | (usy << 16);
1306 store_vector4ui(inst, machine, result);
1307 }
1308 break;
1309 case OPCODE_PK4B: /* pack four GLbytes into one 32-bit float */
1310 {
1311 GLfloat a[4];
1312 GLuint result[4], ubx, uby, ubz, ubw;
1313 fetch_vector4(&inst->SrcReg[0], machine, a);
1314 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
1315 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
1316 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
1317 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
1318 ubx = IROUND(127.0F * a[0] + 128.0F);
1319 uby = IROUND(127.0F * a[1] + 128.0F);
1320 ubz = IROUND(127.0F * a[2] + 128.0F);
1321 ubw = IROUND(127.0F * a[3] + 128.0F);
1322 result[0] =
1323 result[1] =
1324 result[2] =
1325 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1326 store_vector4ui(inst, machine, result);
1327 }
1328 break;
1329 case OPCODE_PK4UB: /* pack four GLubytes into one 32-bit float */
1330 {
1331 GLfloat a[4];
1332 GLuint result[4], ubx, uby, ubz, ubw;
1333 fetch_vector4(&inst->SrcReg[0], machine, a);
1334 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1335 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1336 a[2] = CLAMP(a[2], 0.0F, 1.0F);
1337 a[3] = CLAMP(a[3], 0.0F, 1.0F);
1338 ubx = IROUND(255.0F * a[0]);
1339 uby = IROUND(255.0F * a[1]);
1340 ubz = IROUND(255.0F * a[2]);
1341 ubw = IROUND(255.0F * a[3]);
1342 result[0] =
1343 result[1] =
1344 result[2] =
1345 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1346 store_vector4ui(inst, machine, result);
1347 }
1348 break;
1349 case OPCODE_POW:
1350 {
1351 GLfloat a[4], b[4], result[4];
1352 fetch_vector1(&inst->SrcReg[0], machine, a);
1353 fetch_vector1(&inst->SrcReg[1], machine, b);
1354 result[0] = result[1] = result[2] = result[3]
1355 = (GLfloat) pow(a[0], b[0]);
1356 store_vector4(inst, machine, result);
1357 }
1358 break;
1359 case OPCODE_RCC: /* clamped riciprocal */
1360 {
1361 const float largest = 1.884467e+19, smallest = 5.42101e-20;
1362 GLfloat a[4], r, result[4];
1363 fetch_vector1(&inst->SrcReg[0], machine, a);
1364 if (DEBUG_PROG) {
1365 if (a[0] == 0)
1366 printf("RCC(0)\n");
1367 else if (IS_INF_OR_NAN(a[0]))
1368 printf("RCC(inf)\n");
1369 }
1370 if (a[0] == 1.0F) {
1371 r = 1.0F;
1372 }
1373 else {
1374 r = 1.0F / a[0];
1375 }
1376 if (positive(r)) {
1377 if (r > largest) {
1378 r = largest;
1379 }
1380 else if (r < smallest) {
1381 r = smallest;
1382 }
1383 }
1384 else {
1385 if (r < -largest) {
1386 r = -largest;
1387 }
1388 else if (r > -smallest) {
1389 r = -smallest;
1390 }
1391 }
1392 result[0] = result[1] = result[2] = result[3] = r;
1393 store_vector4(inst, machine, result);
1394 }
1395 break;
1396
1397 case OPCODE_RCP:
1398 {
1399 GLfloat a[4], result[4];
1400 fetch_vector1(&inst->SrcReg[0], machine, a);
1401 if (DEBUG_PROG) {
1402 if (a[0] == 0)
1403 printf("RCP(0)\n");
1404 else if (IS_INF_OR_NAN(a[0]))
1405 printf("RCP(inf)\n");
1406 }
1407 result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
1408 store_vector4(inst, machine, result);
1409 }
1410 break;
1411 case OPCODE_RET: /* return from subroutine (conditional) */
1412 if (eval_condition(machine, inst)) {
1413 if (machine->StackDepth == 0) {
1414 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
1415 }
1416 /* subtract one because of pc++ in the for loop */
1417 pc = machine->CallStack[--machine->StackDepth] - 1;
1418 }
1419 break;
1420 case OPCODE_RFL: /* reflection vector */
1421 {
1422 GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
1423 fetch_vector4(&inst->SrcReg[0], machine, axis);
1424 fetch_vector4(&inst->SrcReg[1], machine, dir);
1425 tmpW = DOT3(axis, axis);
1426 tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
1427 result[0] = tmpX * axis[0] - dir[0];
1428 result[1] = tmpX * axis[1] - dir[1];
1429 result[2] = tmpX * axis[2] - dir[2];
1430 /* result[3] is never written! XXX enforce in parser! */
1431 store_vector4(inst, machine, result);
1432 }
1433 break;
1434 case OPCODE_RSQ: /* 1 / sqrt() */
1435 {
1436 GLfloat a[4], result[4];
1437 fetch_vector1(&inst->SrcReg[0], machine, a);
1438 a[0] = FABSF(a[0]);
1439 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
1440 store_vector4(inst, machine, result);
1441 if (DEBUG_PROG) {
1442 printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
1443 }
1444 }
1445 break;
1446 case OPCODE_SCS: /* sine and cos */
1447 {
1448 GLfloat a[4], result[4];
1449 fetch_vector1(&inst->SrcReg[0], machine, a);
1450 result[0] = (GLfloat) cos(a[0]);
1451 result[1] = (GLfloat) sin(a[0]);
1452 result[2] = 0.0; /* undefined! */
1453 result[3] = 0.0; /* undefined! */
1454 store_vector4(inst, machine, result);
1455 }
1456 break;
1457 case OPCODE_SEQ: /* set on equal */
1458 {
1459 GLfloat a[4], b[4], result[4];
1460 fetch_vector4(&inst->SrcReg[0], machine, a);
1461 fetch_vector4(&inst->SrcReg[1], machine, b);
1462 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1463 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1464 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1465 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1466 store_vector4(inst, machine, result);
1467 if (DEBUG_PROG) {
1468 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1469 result[0], result[1], result[2], result[3],
1470 a[0], a[1], a[2], a[3],
1471 b[0], b[1], b[2], b[3]);
1472 }
1473 }
1474 break;
1475 case OPCODE_SFL: /* set false, operands ignored */
1476 {
1477 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1478 store_vector4(inst, machine, result);
1479 }
1480 break;
1481 case OPCODE_SGE: /* set on greater or equal */
1482 {
1483 GLfloat a[4], b[4], result[4];
1484 fetch_vector4(&inst->SrcReg[0], machine, a);
1485 fetch_vector4(&inst->SrcReg[1], machine, b);
1486 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1487 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1488 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1489 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1490 store_vector4(inst, machine, result);
1491 if (DEBUG_PROG) {
1492 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1493 result[0], result[1], result[2], result[3],
1494 a[0], a[1], a[2], a[3],
1495 b[0], b[1], b[2], b[3]);
1496 }
1497 }
1498 break;
1499 case OPCODE_SGT: /* set on greater */
1500 {
1501 GLfloat a[4], b[4], result[4];
1502 fetch_vector4(&inst->SrcReg[0], machine, a);
1503 fetch_vector4(&inst->SrcReg[1], machine, b);
1504 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1505 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1506 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1507 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1508 store_vector4(inst, machine, result);
1509 if (DEBUG_PROG) {
1510 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1511 result[0], result[1], result[2], result[3],
1512 a[0], a[1], a[2], a[3],
1513 b[0], b[1], b[2], b[3]);
1514 }
1515 }
1516 break;
1517 case OPCODE_SIN:
1518 {
1519 GLfloat a[4], result[4];
1520 fetch_vector1(&inst->SrcReg[0], machine, a);
1521 result[0] = result[1] = result[2] = result[3]
1522 = (GLfloat) sin(a[0]);
1523 store_vector4(inst, machine, result);
1524 }
1525 break;
1526 case OPCODE_SLE: /* set on less or equal */
1527 {
1528 GLfloat a[4], b[4], result[4];
1529 fetch_vector4(&inst->SrcReg[0], machine, a);
1530 fetch_vector4(&inst->SrcReg[1], machine, b);
1531 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1532 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1533 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1534 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1535 store_vector4(inst, machine, result);
1536 if (DEBUG_PROG) {
1537 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1538 result[0], result[1], result[2], result[3],
1539 a[0], a[1], a[2], a[3],
1540 b[0], b[1], b[2], b[3]);
1541 }
1542 }
1543 break;
1544 case OPCODE_SLT: /* set on less */
1545 {
1546 GLfloat a[4], b[4], result[4];
1547 fetch_vector4(&inst->SrcReg[0], machine, a);
1548 fetch_vector4(&inst->SrcReg[1], machine, b);
1549 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1550 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1551 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1552 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1553 store_vector4(inst, machine, result);
1554 if (DEBUG_PROG) {
1555 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1556 result[0], result[1], result[2], result[3],
1557 a[0], a[1], a[2], a[3],
1558 b[0], b[1], b[2], b[3]);
1559 }
1560 }
1561 break;
1562 case OPCODE_SNE: /* set on not equal */
1563 {
1564 GLfloat a[4], b[4], result[4];
1565 fetch_vector4(&inst->SrcReg[0], machine, a);
1566 fetch_vector4(&inst->SrcReg[1], machine, b);
1567 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1568 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1569 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1570 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1571 store_vector4(inst, machine, result);
1572 if (DEBUG_PROG) {
1573 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1574 result[0], result[1], result[2], result[3],
1575 a[0], a[1], a[2], a[3],
1576 b[0], b[1], b[2], b[3]);
1577 }
1578 }
1579 break;
1580 case OPCODE_SSG: /* set sign (-1, 0 or +1) */
1581 {
1582 GLfloat a[4], result[4];
1583 fetch_vector4(&inst->SrcReg[0], machine, a);
1584 result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
1585 result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
1586 result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
1587 result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
1588 store_vector4(inst, machine, result);
1589 }
1590 break;
1591 case OPCODE_STR: /* set true, operands ignored */
1592 {
1593 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1594 store_vector4(inst, machine, result);
1595 }
1596 break;
1597 case OPCODE_SUB:
1598 {
1599 GLfloat a[4], b[4], result[4];
1600 fetch_vector4(&inst->SrcReg[0], machine, a);
1601 fetch_vector4(&inst->SrcReg[1], machine, b);
1602 result[0] = a[0] - b[0];
1603 result[1] = a[1] - b[1];
1604 result[2] = a[2] - b[2];
1605 result[3] = a[3] - b[3];
1606 store_vector4(inst, machine, result);
1607 if (DEBUG_PROG) {
1608 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1609 result[0], result[1], result[2], result[3],
1610 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1611 }
1612 }
1613 break;
1614 case OPCODE_SWZ: /* extended swizzle */
1615 {
1616 const struct prog_src_register *source = &inst->SrcReg[0];
1617 const GLfloat *src = get_src_register_pointer(source, machine);
1618 GLfloat result[4];
1619 GLuint i;
1620 for (i = 0; i < 4; i++) {
1621 const GLuint swz = GET_SWZ(source->Swizzle, i);
1622 if (swz == SWIZZLE_ZERO)
1623 result[i] = 0.0;
1624 else if (swz == SWIZZLE_ONE)
1625 result[i] = 1.0;
1626 else {
1627 ASSERT(swz >= 0);
1628 ASSERT(swz <= 3);
1629 result[i] = src[swz];
1630 }
1631 if (source->Negate & (1 << i))
1632 result[i] = -result[i];
1633 }
1634 store_vector4(inst, machine, result);
1635 }
1636 break;
1637 case OPCODE_TEX: /* Both ARB and NV frag prog */
1638 /* Simple texel lookup */
1639 {
1640 GLfloat texcoord[4], color[4];
1641 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1642
1643 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1644
1645 if (DEBUG_PROG) {
1646 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1647 color[0], color[1], color[2], color[3],
1648 inst->TexSrcUnit,
1649 texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
1650 }
1651 store_vector4(inst, machine, color);
1652 }
1653 break;
1654 case OPCODE_TXB: /* GL_ARB_fragment_program only */
1655 /* Texel lookup with LOD bias */
1656 {
1657 GLfloat texcoord[4], color[4], lodBias;
1658
1659 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1660
1661 /* texcoord[3] is the bias to add to lambda */
1662 lodBias = texcoord[3];
1663
1664 fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
1665
1666 store_vector4(inst, machine, color);
1667 }
1668 break;
1669 case OPCODE_TXD: /* GL_NV_fragment_program only */
1670 /* Texture lookup w/ partial derivatives for LOD */
1671 {
1672 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1673 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1674 fetch_vector4(&inst->SrcReg[1], machine, dtdx);
1675 fetch_vector4(&inst->SrcReg[2], machine, dtdy);
1676 machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
1677 0.0, /* lodBias */
1678 inst->TexSrcUnit, color);
1679 store_vector4(inst, machine, color);
1680 }
1681 break;
1682 case OPCODE_TXP: /* GL_ARB_fragment_program only */
1683 /* Texture lookup w/ projective divide */
1684 {
1685 GLfloat texcoord[4], color[4];
1686
1687 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1688 /* Not so sure about this test - if texcoord[3] is
1689 * zero, we'd probably be fine except for an ASSERT in
1690 * IROUND_POS() which gets triggered by the inf values created.
1691 */
1692 if (texcoord[3] != 0.0) {
1693 texcoord[0] /= texcoord[3];
1694 texcoord[1] /= texcoord[3];
1695 texcoord[2] /= texcoord[3];
1696 }
1697
1698 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1699
1700 store_vector4(inst, machine, color);
1701 }
1702 break;
1703 case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
1704 /* Texture lookup w/ projective divide, as above, but do not
1705 * do the divide by w if sampling from a cube map.
1706 */
1707 {
1708 GLfloat texcoord[4], color[4];
1709
1710 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1711 if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
1712 texcoord[3] != 0.0) {
1713 texcoord[0] /= texcoord[3];
1714 texcoord[1] /= texcoord[3];
1715 texcoord[2] /= texcoord[3];
1716 }
1717
1718 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1719
1720 store_vector4(inst, machine, color);
1721 }
1722 break;
1723 case OPCODE_TRUNC: /* truncate toward zero */
1724 {
1725 GLfloat a[4], result[4];
1726 fetch_vector4(&inst->SrcReg[0], machine, a);
1727 result[0] = (GLfloat) (GLint) a[0];
1728 result[1] = (GLfloat) (GLint) a[1];
1729 result[2] = (GLfloat) (GLint) a[2];
1730 result[3] = (GLfloat) (GLint) a[3];
1731 store_vector4(inst, machine, result);
1732 }
1733 break;
1734 case OPCODE_UP2H: /* unpack two 16-bit floats */
1735 {
1736 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1737 GLfloat result[4];
1738 GLushort hx, hy;
1739 hx = raw & 0xffff;
1740 hy = raw >> 16;
1741 result[0] = result[2] = _mesa_half_to_float(hx);
1742 result[1] = result[3] = _mesa_half_to_float(hy);
1743 store_vector4(inst, machine, result);
1744 }
1745 break;
1746 case OPCODE_UP2US: /* unpack two GLushorts */
1747 {
1748 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1749 GLfloat result[4];
1750 GLushort usx, usy;
1751 usx = raw & 0xffff;
1752 usy = raw >> 16;
1753 result[0] = result[2] = usx * (1.0f / 65535.0f);
1754 result[1] = result[3] = usy * (1.0f / 65535.0f);
1755 store_vector4(inst, machine, result);
1756 }
1757 break;
1758 case OPCODE_UP4B: /* unpack four GLbytes */
1759 {
1760 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1761 GLfloat result[4];
1762 result[0] = (((raw >> 0) & 0xff) - 128) / 127.0F;
1763 result[1] = (((raw >> 8) & 0xff) - 128) / 127.0F;
1764 result[2] = (((raw >> 16) & 0xff) - 128) / 127.0F;
1765 result[3] = (((raw >> 24) & 0xff) - 128) / 127.0F;
1766 store_vector4(inst, machine, result);
1767 }
1768 break;
1769 case OPCODE_UP4UB: /* unpack four GLubytes */
1770 {
1771 const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
1772 GLfloat result[4];
1773 result[0] = ((raw >> 0) & 0xff) / 255.0F;
1774 result[1] = ((raw >> 8) & 0xff) / 255.0F;
1775 result[2] = ((raw >> 16) & 0xff) / 255.0F;
1776 result[3] = ((raw >> 24) & 0xff) / 255.0F;
1777 store_vector4(inst, machine, result);
1778 }
1779 break;
1780 case OPCODE_XOR: /* bitwise XOR */
1781 {
1782 GLuint a[4], b[4], result[4];
1783 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1784 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1785 result[0] = a[0] ^ b[0];
1786 result[1] = a[1] ^ b[1];
1787 result[2] = a[2] ^ b[2];
1788 result[3] = a[3] ^ b[3];
1789 store_vector4ui(inst, machine, result);
1790 }
1791 break;
1792 case OPCODE_XPD: /* cross product */
1793 {
1794 GLfloat a[4], b[4], result[4];
1795 fetch_vector4(&inst->SrcReg[0], machine, a);
1796 fetch_vector4(&inst->SrcReg[1], machine, b);
1797 result[0] = a[1] * b[2] - a[2] * b[1];
1798 result[1] = a[2] * b[0] - a[0] * b[2];
1799 result[2] = a[0] * b[1] - a[1] * b[0];
1800 result[3] = 1.0;
1801 store_vector4(inst, machine, result);
1802 if (DEBUG_PROG) {
1803 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1804 result[0], result[1], result[2], result[3],
1805 a[0], a[1], a[2], b[0], b[1], b[2]);
1806 }
1807 }
1808 break;
1809 case OPCODE_X2D: /* 2-D matrix transform */
1810 {
1811 GLfloat a[4], b[4], c[4], result[4];
1812 fetch_vector4(&inst->SrcReg[0], machine, a);
1813 fetch_vector4(&inst->SrcReg[1], machine, b);
1814 fetch_vector4(&inst->SrcReg[2], machine, c);
1815 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1816 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1817 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1818 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1819 store_vector4(inst, machine, result);
1820 }
1821 break;
1822 case OPCODE_PRINT:
1823 {
1824 if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
1825 GLfloat a[4];
1826 fetch_vector4(&inst->SrcReg[0], machine, a);
1827 printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
1828 a[0], a[1], a[2], a[3]);
1829 }
1830 else {
1831 printf("%s\n", (const char *) inst->Data);
1832 }
1833 }
1834 break;
1835 case OPCODE_END:
1836 return GL_TRUE;
1837 default:
1838 _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
1839 inst->Opcode);
1840 return GL_TRUE; /* return value doesn't matter */
1841 }
1842
1843 numExec++;
1844 if (numExec > maxExec) {
1845 static GLboolean reported = GL_FALSE;
1846 if (!reported) {
1847 _mesa_problem(ctx, "Infinite loop detected in fragment program");
1848 reported = GL_TRUE;
1849 }
1850 return GL_TRUE;
1851 }
1852
1853 } /* for pc */
1854
1855 return GL_TRUE;
1856 }