projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge commit 'origin/gallium-0.1'
[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 "program.h"
42 #include "prog_execute.h"
43 #include "prog_instruction.h"
44 #include "prog_parameter.h"
45 #include "prog_print.h"
46 #include "prog_noise.h"
47
48
49 /* debug predicate */
50 #define DEBUG_PROG 0
51
52
53 /**
54 * Set x to positive or negative infinity.
55 */
56 #if defined(USE_IEEE) || defined(_WIN32)
57 #define SET_POS_INFINITY(x) ( *((GLuint *) (void *)&x) = 0x7F800000 )
58 #define SET_NEG_INFINITY(x) ( *((GLuint *) (void *)&x) = 0xFF800000 )
59 #elif defined(VMS)
60 #define SET_POS_INFINITY(x) x = __MAXFLOAT
61 #define SET_NEG_INFINITY(x) x = -__MAXFLOAT
62 #else
63 #define SET_POS_INFINITY(x) x = (GLfloat) HUGE_VAL
64 #define SET_NEG_INFINITY(x) x = (GLfloat) -HUGE_VAL
65 #endif
66
67 #define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
68
69
70 static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
71
72
73
74 /**
75 * Return a pointer to the 4-element float vector specified by the given
76 * source register.
77 */
78 static INLINE const GLfloat *
79 get_src_register_pointer(const struct prog_src_register *source,
80 const struct gl_program_machine *machine)
81 {
82 const struct gl_program *prog = machine->CurProgram;
83 GLint reg = source->Index;
84
85 if (source->RelAddr) {
86 /* add address register value to src index/offset */
87 reg += machine->AddressReg[0][0];
88 if (reg < 0) {
89 return ZeroVec;
90 }
91 }
92
93 switch (source->File) {
94 case PROGRAM_TEMPORARY:
95 if (reg >= MAX_PROGRAM_TEMPS)
96 return ZeroVec;
97 return machine->Temporaries[reg];
98
99 case PROGRAM_INPUT:
100 if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
101 if (reg >= VERT_ATTRIB_MAX)
102 return ZeroVec;
103 return machine->VertAttribs[reg];
104 }
105 else {
106 if (reg >= FRAG_ATTRIB_MAX)
107 return ZeroVec;
108 return machine->Attribs[reg][machine->CurElement];
109 }
110
111 case PROGRAM_OUTPUT:
112 if (reg >= MAX_PROGRAM_OUTPUTS)
113 return ZeroVec;
114 return machine->Outputs[reg];
115
116 case PROGRAM_LOCAL_PARAM:
117 if (reg >= MAX_PROGRAM_LOCAL_PARAMS)
118 return ZeroVec;
119 return machine->CurProgram->LocalParams[reg];
120
121 case PROGRAM_ENV_PARAM:
122 if (reg >= MAX_PROGRAM_ENV_PARAMS)
123 return ZeroVec;
124 return machine->EnvParams[reg];
125
126 case PROGRAM_STATE_VAR:
127 /* Fallthrough */
128 case PROGRAM_CONSTANT:
129 /* Fallthrough */
130 case PROGRAM_UNIFORM:
131 /* Fallthrough */
132 case PROGRAM_NAMED_PARAM:
133 if (reg >= (GLint) prog->Parameters->NumParameters)
134 return ZeroVec;
135 return prog->Parameters->ParameterValues[reg];
136
137 default:
138 _mesa_problem(NULL,
139 "Invalid src register file %d in get_src_register_pointer()",
140 source->File);
141 return NULL;
142 }
143 }
144
145
146 /**
147 * Return a pointer to the 4-element float vector specified by the given
148 * destination register.
149 */
150 static INLINE GLfloat *
151 get_dst_register_pointer(const struct prog_dst_register *dest,
152 struct gl_program_machine *machine)
153 {
154 static GLfloat dummyReg[4];
155 GLint reg = dest->Index;
156
157 if (dest->RelAddr) {
158 /* add address register value to src index/offset */
159 reg += machine->AddressReg[0][0];
160 if (reg < 0) {
161 return dummyReg;
162 }
163 }
164
165 switch (dest->File) {
166 case PROGRAM_TEMPORARY:
167 if (reg >= MAX_PROGRAM_TEMPS)
168 return dummyReg;
169 return machine->Temporaries[reg];
170
171 case PROGRAM_OUTPUT:
172 if (reg >= MAX_PROGRAM_OUTPUTS)
173 return dummyReg;
174 return machine->Outputs[reg];
175
176 case PROGRAM_WRITE_ONLY:
177 return dummyReg;
178
179 default:
180 _mesa_problem(NULL,
181 "Invalid dest register file %d in get_dst_register_pointer()",
182 dest->File);
183 return NULL;
184 }
185 }
186
187
188
189 #if FEATURE_MESA_program_debug
190 static struct gl_program_machine *CurrentMachine = NULL;
191
192 /**
193 * For GL_MESA_program_debug.
194 * Return current value (4*GLfloat) of a program register.
195 * Called via ctx->Driver.GetProgramRegister().
196 */
197 void
198 _mesa_get_program_register(GLcontext *ctx, enum register_file file,
199 GLuint index, GLfloat val[4])
200 {
201 if (CurrentMachine) {
202 struct prog_src_register srcReg;
203 const GLfloat *src;
204 srcReg.File = file;
205 srcReg.Index = index;
206 src = get_src_register_pointer(&srcReg, CurrentMachine);
207 COPY_4V(val, src);
208 }
209 }
210 #endif /* FEATURE_MESA_program_debug */
211
212
213 /**
214 * Fetch a 4-element float vector from the given source register.
215 * Apply swizzling and negating as needed.
216 */
217 static void
218 fetch_vector4(const struct prog_src_register *source,
219 const struct gl_program_machine *machine, GLfloat result[4])
220 {
221 const GLfloat *src = get_src_register_pointer(source, machine);
222 ASSERT(src);
223
224 if (source->Swizzle == SWIZZLE_NOOP) {
225 /* no swizzling */
226 COPY_4V(result, src);
227 }
228 else {
229 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
230 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
231 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
232 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
233 result[0] = src[GET_SWZ(source->Swizzle, 0)];
234 result[1] = src[GET_SWZ(source->Swizzle, 1)];
235 result[2] = src[GET_SWZ(source->Swizzle, 2)];
236 result[3] = src[GET_SWZ(source->Swizzle, 3)];
237 }
238
239 if (source->NegateBase) {
240 result[0] = -result[0];
241 result[1] = -result[1];
242 result[2] = -result[2];
243 result[3] = -result[3];
244 }
245 if (source->Abs) {
246 result[0] = FABSF(result[0]);
247 result[1] = FABSF(result[1]);
248 result[2] = FABSF(result[2]);
249 result[3] = FABSF(result[3]);
250 }
251 if (source->NegateAbs) {
252 result[0] = -result[0];
253 result[1] = -result[1];
254 result[2] = -result[2];
255 result[3] = -result[3];
256 }
257 }
258
259
260 /**
261 * Fetch a 4-element uint vector from the given source register.
262 * Apply swizzling but not negation/abs.
263 */
264 static void
265 fetch_vector4ui(const struct prog_src_register *source,
266 const struct gl_program_machine *machine, GLuint result[4])
267 {
268 const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
269 ASSERT(src);
270
271 if (source->Swizzle == SWIZZLE_NOOP) {
272 /* no swizzling */
273 COPY_4V(result, src);
274 }
275 else {
276 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
277 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
278 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
279 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
280 result[0] = src[GET_SWZ(source->Swizzle, 0)];
281 result[1] = src[GET_SWZ(source->Swizzle, 1)];
282 result[2] = src[GET_SWZ(source->Swizzle, 2)];
283 result[3] = src[GET_SWZ(source->Swizzle, 3)];
284 }
285
286 /* Note: no NegateBase, Abs, NegateAbs here */
287 }
288
289
290
291 /**
292 * Fetch the derivative with respect to X or Y for the given register.
293 * XXX this currently only works for fragment program input attribs.
294 */
295 static void
296 fetch_vector4_deriv(GLcontext * ctx,
297 const struct prog_src_register *source,
298 const struct gl_program_machine *machine,
299 char xOrY, GLfloat result[4])
300 {
301 if (source->File == PROGRAM_INPUT &&
302 source->Index < (GLint) machine->NumDeriv) {
303 const GLint col = machine->CurElement;
304 const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
305 const GLfloat invQ = 1.0f / w;
306 GLfloat deriv[4];
307
308 if (xOrY == 'X') {
309 deriv[0] = machine->DerivX[source->Index][0] * invQ;
310 deriv[1] = machine->DerivX[source->Index][1] * invQ;
311 deriv[2] = machine->DerivX[source->Index][2] * invQ;
312 deriv[3] = machine->DerivX[source->Index][3] * invQ;
313 }
314 else {
315 deriv[0] = machine->DerivY[source->Index][0] * invQ;
316 deriv[1] = machine->DerivY[source->Index][1] * invQ;
317 deriv[2] = machine->DerivY[source->Index][2] * invQ;
318 deriv[3] = machine->DerivY[source->Index][3] * invQ;
319 }
320
321 result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
322 result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
323 result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
324 result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
325
326 if (source->NegateBase) {
327 result[0] = -result[0];
328 result[1] = -result[1];
329 result[2] = -result[2];
330 result[3] = -result[3];
331 }
332 if (source->Abs) {
333 result[0] = FABSF(result[0]);
334 result[1] = FABSF(result[1]);
335 result[2] = FABSF(result[2]);
336 result[3] = FABSF(result[3]);
337 }
338 if (source->NegateAbs) {
339 result[0] = -result[0];
340 result[1] = -result[1];
341 result[2] = -result[2];
342 result[3] = -result[3];
343 }
344 }
345 else {
346 ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
347 }
348 }
349
350
351 /**
352 * As above, but only return result[0] element.
353 */
354 static void
355 fetch_vector1(const struct prog_src_register *source,
356 const struct gl_program_machine *machine, GLfloat result[4])
357 {
358 const GLfloat *src = get_src_register_pointer(source, machine);
359 ASSERT(src);
360
361 result[0] = src[GET_SWZ(source->Swizzle, 0)];
362
363 if (source->NegateBase) {
364 result[0] = -result[0];
365 }
366 if (source->Abs) {
367 result[0] = FABSF(result[0]);
368 }
369 if (source->NegateAbs) {
370 result[0] = -result[0];
371 }
372 }
373
374
375 /**
376 * Fetch texel from texture. Use partial derivatives when possible.
377 */
378 static INLINE void
379 fetch_texel(GLcontext *ctx,
380 const struct gl_program_machine *machine,
381 const struct prog_instruction *inst,
382 const GLfloat texcoord[4], GLfloat lodBias,
383 GLfloat color[4])
384 {
385 const GLuint unit = machine->Samplers[inst->TexSrcUnit];
386
387 /* Note: we only have the right derivatives for fragment input attribs.
388 */
389 if (machine->NumDeriv > 0 &&
390 inst->SrcReg[0].File == PROGRAM_INPUT &&
391 inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
392 /* simple texture fetch for which we should have derivatives */
393 GLuint attr = inst->SrcReg[0].Index;
394 machine->FetchTexelDeriv(ctx, texcoord,
395 machine->DerivX[attr],
396 machine->DerivY[attr],
397 lodBias, unit, color);
398 }
399 else {
400 machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
401 }
402 }
403
404
405 /**
406 * Test value against zero and return GT, LT, EQ or UN if NaN.
407 */
408 static INLINE GLuint
409 generate_cc(float value)
410 {
411 if (value != value)
412 return COND_UN; /* NaN */
413 if (value > 0.0F)
414 return COND_GT;
415 if (value < 0.0F)
416 return COND_LT;
417 return COND_EQ;
418 }
419
420
421 /**
422 * Test if the ccMaskRule is satisfied by the given condition code.
423 * Used to mask destination writes according to the current condition code.
424 */
425 static INLINE GLboolean
426 test_cc(GLuint condCode, GLuint ccMaskRule)
427 {
428 switch (ccMaskRule) {
429 case COND_EQ: return (condCode == COND_EQ);
430 case COND_NE: return (condCode != COND_EQ);
431 case COND_LT: return (condCode == COND_LT);
432 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
433 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
434 case COND_GT: return (condCode == COND_GT);
435 case COND_TR: return GL_TRUE;
436 case COND_FL: return GL_FALSE;
437 default: return GL_TRUE;
438 }
439 }
440
441
442 /**
443 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
444 * or GL_FALSE to indicate result.
445 */
446 static INLINE GLboolean
447 eval_condition(const struct gl_program_machine *machine,
448 const struct prog_instruction *inst)
449 {
450 const GLuint swizzle = inst->DstReg.CondSwizzle;
451 const GLuint condMask = inst->DstReg.CondMask;
452 if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
453 test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
454 test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
455 test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
456 return GL_TRUE;
457 }
458 else {
459 return GL_FALSE;
460 }
461 }
462
463
464
465 /**
466 * Store 4 floats into a register. Observe the instructions saturate and
467 * set-condition-code flags.
468 */
469 static void
470 store_vector4(const struct prog_instruction *inst,
471 struct gl_program_machine *machine, const GLfloat value[4])
472 {
473 const struct prog_dst_register *dstReg = &(inst->DstReg);
474 const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
475 GLuint writeMask = dstReg->WriteMask;
476 GLfloat clampedValue[4];
477 GLfloat *dst = get_dst_register_pointer(dstReg, machine);
478
479 #if 0
480 if (value[0] > 1.0e10 ||
481 IS_INF_OR_NAN(value[0]) ||
482 IS_INF_OR_NAN(value[1]) ||
483 IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
484 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
485 #endif
486
487 if (clamp) {
488 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
489 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
490 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
491 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
492 value = clampedValue;
493 }
494
495 if (dstReg->CondMask != COND_TR) {
496 /* condition codes may turn off some writes */
497 if (writeMask & WRITEMASK_X) {
498 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
499 dstReg->CondMask))
500 writeMask &= ~WRITEMASK_X;
501 }
502 if (writeMask & WRITEMASK_Y) {
503 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
504 dstReg->CondMask))
505 writeMask &= ~WRITEMASK_Y;
506 }
507 if (writeMask & WRITEMASK_Z) {
508 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
509 dstReg->CondMask))
510 writeMask &= ~WRITEMASK_Z;
511 }
512 if (writeMask & WRITEMASK_W) {
513 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
514 dstReg->CondMask))
515 writeMask &= ~WRITEMASK_W;
516 }
517 }
518
519 if (writeMask & WRITEMASK_X)
520 dst[0] = value[0];
521 if (writeMask & WRITEMASK_Y)
522 dst[1] = value[1];
523 if (writeMask & WRITEMASK_Z)
524 dst[2] = value[2];
525 if (writeMask & WRITEMASK_W)
526 dst[3] = value[3];
527
528 if (inst->CondUpdate) {
529 if (writeMask & WRITEMASK_X)
530 machine->CondCodes[0] = generate_cc(value[0]);
531 if (writeMask & WRITEMASK_Y)
532 machine->CondCodes[1] = generate_cc(value[1]);
533 if (writeMask & WRITEMASK_Z)
534 machine->CondCodes[2] = generate_cc(value[2]);
535 if (writeMask & WRITEMASK_W)
536 machine->CondCodes[3] = generate_cc(value[3]);
537 #if DEBUG_PROG
538 printf("CondCodes=(%s,%s,%s,%s) for:\n",
539 _mesa_condcode_string(machine->CondCodes[0]),
540 _mesa_condcode_string(machine->CondCodes[1]),
541 _mesa_condcode_string(machine->CondCodes[2]),
542 _mesa_condcode_string(machine->CondCodes[3]));
543 #endif
544 }
545 }
546
547
548 /**
549 * Store 4 uints into a register. Observe the set-condition-code flags.
550 */
551 static void
552 store_vector4ui(const struct prog_instruction *inst,
553 struct gl_program_machine *machine, const GLuint value[4])
554 {
555 const struct prog_dst_register *dstReg = &(inst->DstReg);
556 GLuint writeMask = dstReg->WriteMask;
557 GLuint *dst = (GLuint *) get_dst_register_pointer(dstReg, machine);
558
559 if (dstReg->CondMask != COND_TR) {
560 /* condition codes may turn off some writes */
561 if (writeMask & WRITEMASK_X) {
562 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
563 dstReg->CondMask))
564 writeMask &= ~WRITEMASK_X;
565 }
566 if (writeMask & WRITEMASK_Y) {
567 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
568 dstReg->CondMask))
569 writeMask &= ~WRITEMASK_Y;
570 }
571 if (writeMask & WRITEMASK_Z) {
572 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
573 dstReg->CondMask))
574 writeMask &= ~WRITEMASK_Z;
575 }
576 if (writeMask & WRITEMASK_W) {
577 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
578 dstReg->CondMask))
579 writeMask &= ~WRITEMASK_W;
580 }
581 }
582
583 if (writeMask & WRITEMASK_X)
584 dst[0] = value[0];
585 if (writeMask & WRITEMASK_Y)
586 dst[1] = value[1];
587 if (writeMask & WRITEMASK_Z)
588 dst[2] = value[2];
589 if (writeMask & WRITEMASK_W)
590 dst[3] = value[3];
591
592 if (inst->CondUpdate) {
593 if (writeMask & WRITEMASK_X)
594 machine->CondCodes[0] = generate_cc(value[0]);
595 if (writeMask & WRITEMASK_Y)
596 machine->CondCodes[1] = generate_cc(value[1]);
597 if (writeMask & WRITEMASK_Z)
598 machine->CondCodes[2] = generate_cc(value[2]);
599 if (writeMask & WRITEMASK_W)
600 machine->CondCodes[3] = generate_cc(value[3]);
601 #if DEBUG_PROG
602 printf("CondCodes=(%s,%s,%s,%s) for:\n",
603 _mesa_condcode_string(machine->CondCodes[0]),
604 _mesa_condcode_string(machine->CondCodes[1]),
605 _mesa_condcode_string(machine->CondCodes[2]),
606 _mesa_condcode_string(machine->CondCodes[3]));
607 #endif
608 }
609 }
610
611
612
613 /**
614 * Execute the given vertex/fragment program.
615 *
616 * \param ctx rendering context
617 * \param program the program to execute
618 * \param machine machine state (must be initialized)
619 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
620 */
621 GLboolean
622 _mesa_execute_program(GLcontext * ctx,
623 const struct gl_program *program,
624 struct gl_program_machine *machine)
625 {
626 const GLuint numInst = program->NumInstructions;
627 const GLuint maxExec = 10000;
628 GLuint pc, numExec = 0;
629
630 machine->CurProgram = program;
631
632 if (DEBUG_PROG) {
633 printf("execute program %u --------------------\n", program->Id);
634 }
635
636 #if FEATURE_MESA_program_debug
637 CurrentMachine = machine;
638 #endif
639
640 if (program->Target == GL_VERTEX_PROGRAM_ARB) {
641 machine->EnvParams = ctx->VertexProgram.Parameters;
642 }
643 else {
644 machine->EnvParams = ctx->FragmentProgram.Parameters;
645 }
646
647 for (pc = 0; pc < numInst; pc++) {
648 const struct prog_instruction *inst = program->Instructions + pc;
649
650 #if FEATURE_MESA_program_debug
651 if (ctx->FragmentProgram.CallbackEnabled &&
652 ctx->FragmentProgram.Callback) {
653 ctx->FragmentProgram.CurrentPosition = inst->StringPos;
654 ctx->FragmentProgram.Callback(program->Target,
655 ctx->FragmentProgram.CallbackData);
656 }
657 #endif
658
659 if (DEBUG_PROG) {
660 _mesa_print_instruction(inst);
661 }
662
663 switch (inst->Opcode) {
664 case OPCODE_ABS:
665 {
666 GLfloat a[4], result[4];
667 fetch_vector4(&inst->SrcReg[0], machine, a);
668 result[0] = FABSF(a[0]);
669 result[1] = FABSF(a[1]);
670 result[2] = FABSF(a[2]);
671 result[3] = FABSF(a[3]);
672 store_vector4(inst, machine, result);
673 }
674 break;
675 case OPCODE_ADD:
676 {
677 GLfloat a[4], b[4], result[4];
678 fetch_vector4(&inst->SrcReg[0], machine, a);
679 fetch_vector4(&inst->SrcReg[1], machine, b);
680 result[0] = a[0] + b[0];
681 result[1] = a[1] + b[1];
682 result[2] = a[2] + b[2];
683 result[3] = a[3] + b[3];
684 store_vector4(inst, machine, result);
685 if (DEBUG_PROG) {
686 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
687 result[0], result[1], result[2], result[3],
688 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
689 }
690 }
691 break;
692 case OPCODE_AND: /* bitwise AND */
693 {
694 GLuint a[4], b[4], result[4];
695 fetch_vector4ui(&inst->SrcReg[0], machine, a);
696 fetch_vector4ui(&inst->SrcReg[1], machine, b);
697 result[0] = a[0] & b[0];
698 result[1] = a[1] & b[1];
699 result[2] = a[2] & b[2];
700 result[3] = a[3] & b[3];
701 store_vector4ui(inst, machine, result);
702 }
703 break;
704 case OPCODE_ARL:
705 {
706 GLfloat t[4];
707 fetch_vector4(&inst->SrcReg[0], machine, t);
708 machine->AddressReg[0][0] = IFLOOR(t[0]);
709 }
710 break;
711 case OPCODE_BGNLOOP:
712 /* no-op */
713 break;
714 case OPCODE_ENDLOOP:
715 /* subtract 1 here since pc is incremented by for(pc) loop */
716 pc = inst->BranchTarget - 1; /* go to matching BNGLOOP */
717 break;
718 case OPCODE_BGNSUB: /* begin subroutine */
719 break;
720 case OPCODE_ENDSUB: /* end subroutine */
721 break;
722 case OPCODE_BRA: /* branch (conditional) */
723 /* fall-through */
724 case OPCODE_BRK: /* break out of loop (conditional) */
725 /* fall-through */
726 case OPCODE_CONT: /* continue loop (conditional) */
727 if (eval_condition(machine, inst)) {
728 /* take branch */
729 /* Subtract 1 here since we'll do pc++ at end of for-loop */
730 pc = inst->BranchTarget - 1;
731 }
732 break;
733 case OPCODE_CAL: /* Call subroutine (conditional) */
734 if (eval_condition(machine, inst)) {
735 /* call the subroutine */
736 if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
737 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
738 }
739 machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
740 /* Subtract 1 here since we'll do pc++ at end of for-loop */
741 pc = inst->BranchTarget - 1;
742 }
743 break;
744 case OPCODE_CMP:
745 {
746 GLfloat a[4], b[4], c[4], result[4];
747 fetch_vector4(&inst->SrcReg[0], machine, a);
748 fetch_vector4(&inst->SrcReg[1], machine, b);
749 fetch_vector4(&inst->SrcReg[2], machine, c);
750 result[0] = a[0] < 0.0F ? b[0] : c[0];
751 result[1] = a[1] < 0.0F ? b[1] : c[1];
752 result[2] = a[2] < 0.0F ? b[2] : c[2];
753 result[3] = a[3] < 0.0F ? b[3] : c[3];
754 store_vector4(inst, machine, result);
755 }
756 break;
757 case OPCODE_COS:
758 {
759 GLfloat a[4], result[4];
760 fetch_vector1(&inst->SrcReg[0], machine, a);
761 result[0] = result[1] = result[2] = result[3]
762 = (GLfloat) _mesa_cos(a[0]);
763 store_vector4(inst, machine, result);
764 }
765 break;
766 case OPCODE_DDX: /* Partial derivative with respect to X */
767 {
768 GLfloat result[4];
769 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
770 'X', result);
771 store_vector4(inst, machine, result);
772 }
773 break;
774 case OPCODE_DDY: /* Partial derivative with respect to Y */
775 {
776 GLfloat result[4];
777 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
778 'Y', result);
779 store_vector4(inst, machine, result);
780 }
781 break;
782 case OPCODE_DP2:
783 {
784 GLfloat a[4], b[4], result[4];
785 fetch_vector4(&inst->SrcReg[0], machine, a);
786 fetch_vector4(&inst->SrcReg[1], machine, b);
787 result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
788 store_vector4(inst, machine, result);
789 if (DEBUG_PROG) {
790 printf("DP2 %g = (%g %g) . (%g %g)\n",
791 result[0], a[0], a[1], b[0], b[1]);
792 }
793 }
794 break;
795 case OPCODE_DP2A:
796 {
797 GLfloat a[4], b[4], c, result[4];
798 fetch_vector4(&inst->SrcReg[0], machine, a);
799 fetch_vector4(&inst->SrcReg[1], machine, b);
800 fetch_vector1(&inst->SrcReg[1], machine, &c);
801 result[0] = result[1] = result[2] = result[3] = DOT2(a, b) + c;
802 store_vector4(inst, machine, result);
803 if (DEBUG_PROG) {
804 printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
805 result[0], a[0], a[1], b[0], b[1], c);
806 }
807 }
808 break;
809 case OPCODE_DP3:
810 {
811 GLfloat a[4], b[4], result[4];
812 fetch_vector4(&inst->SrcReg[0], machine, a);
813 fetch_vector4(&inst->SrcReg[1], machine, b);
814 result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
815 store_vector4(inst, machine, result);
816 if (DEBUG_PROG) {
817 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
818 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
819 }
820 }
821 break;
822 case OPCODE_DP4:
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] = DOT4(a, b);
828 store_vector4(inst, machine, result);
829 if (DEBUG_PROG) {
830 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
831 result[0], a[0], a[1], a[2], a[3],
832 b[0], b[1], b[2], b[3]);
833 }
834 }
835 break;
836 case OPCODE_DPH:
837 {
838 GLfloat a[4], b[4], result[4];
839 fetch_vector4(&inst->SrcReg[0], machine, a);
840 fetch_vector4(&inst->SrcReg[1], machine, b);
841 result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
842 store_vector4(inst, machine, result);
843 }
844 break;
845 case OPCODE_DST: /* Distance vector */
846 {
847 GLfloat a[4], b[4], result[4];
848 fetch_vector4(&inst->SrcReg[0], machine, a);
849 fetch_vector4(&inst->SrcReg[1], machine, b);
850 result[0] = 1.0F;
851 result[1] = a[1] * b[1];
852 result[2] = a[2];
853 result[3] = b[3];
854 store_vector4(inst, machine, result);
855 }
856 break;
857 case OPCODE_EXP:
858 {
859 GLfloat t[4], q[4], floor_t0;
860 fetch_vector1(&inst->SrcReg[0], machine, t);
861 floor_t0 = FLOORF(t[0]);
862 if (floor_t0 > FLT_MAX_EXP) {
863 SET_POS_INFINITY(q[0]);
864 SET_POS_INFINITY(q[2]);
865 }
866 else if (floor_t0 < FLT_MIN_EXP) {
867 q[0] = 0.0F;
868 q[2] = 0.0F;
869 }
870 else {
871 q[0] = LDEXPF(1.0, (int) floor_t0);
872 /* Note: GL_NV_vertex_program expects
873 * result.z = result.x * APPX(result.y)
874 * We do what the ARB extension says.
875 */
876 q[2] = (GLfloat) pow(2.0, t[0]);
877 }
878 q[1] = t[0] - floor_t0;
879 q[3] = 1.0F;
880 store_vector4( inst, machine, q );
881 }
882 break;
883 case OPCODE_EX2: /* Exponential base 2 */
884 {
885 GLfloat a[4], result[4];
886 fetch_vector1(&inst->SrcReg[0], machine, a);
887 result[0] = result[1] = result[2] = result[3] =
888 (GLfloat) _mesa_pow(2.0, a[0]);
889 store_vector4(inst, machine, result);
890 }
891 break;
892 case OPCODE_FLR:
893 {
894 GLfloat a[4], result[4];
895 fetch_vector4(&inst->SrcReg[0], machine, a);
896 result[0] = FLOORF(a[0]);
897 result[1] = FLOORF(a[1]);
898 result[2] = FLOORF(a[2]);
899 result[3] = FLOORF(a[3]);
900 store_vector4(inst, machine, result);
901 }
902 break;
903 case OPCODE_FRC:
904 {
905 GLfloat a[4], result[4];
906 fetch_vector4(&inst->SrcReg[0], machine, a);
907 result[0] = a[0] - FLOORF(a[0]);
908 result[1] = a[1] - FLOORF(a[1]);
909 result[2] = a[2] - FLOORF(a[2]);
910 result[3] = a[3] - FLOORF(a[3]);
911 store_vector4(inst, machine, result);
912 }
913 break;
914 case OPCODE_IF:
915 {
916 GLboolean cond;
917 /* eval condition */
918 if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
919 GLfloat a[4];
920 fetch_vector1(&inst->SrcReg[0], machine, a);
921 cond = (a[0] != 0.0);
922 }
923 else {
924 cond = eval_condition(machine, inst);
925 }
926 if (DEBUG_PROG) {
927 printf("IF: %d\n", cond);
928 }
929 /* do if/else */
930 if (cond) {
931 /* do if-clause (just continue execution) */
932 }
933 else {
934 /* go to the instruction after ELSE or ENDIF */
935 assert(inst->BranchTarget >= 0);
936 pc = inst->BranchTarget - 1;
937 }
938 }
939 break;
940 case OPCODE_ELSE:
941 /* goto ENDIF */
942 assert(inst->BranchTarget >= 0);
943 pc = inst->BranchTarget - 1;
944 break;
945 case OPCODE_ENDIF:
946 /* nothing */
947 break;
948 case OPCODE_KIL_NV: /* NV_f_p only (conditional) */
949 if (eval_condition(machine, inst)) {
950 return GL_FALSE;
951 }
952 break;
953 case OPCODE_KIL: /* ARB_f_p only */
954 {
955 GLfloat a[4];
956 fetch_vector4(&inst->SrcReg[0], machine, a);
957 if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
958 return GL_FALSE;
959 }
960 }
961 break;
962 case OPCODE_LG2: /* log base 2 */
963 {
964 GLfloat a[4], result[4];
965 fetch_vector1(&inst->SrcReg[0], machine, a);
966 /* The fast LOG2 macro doesn't meet the precision requirements.
967 */
968 result[0] = result[1] = result[2] = result[3] =
969 (log(a[0]) * 1.442695F);
970 store_vector4(inst, machine, result);
971 }
972 break;
973 case OPCODE_LIT:
974 {
975 const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
976 GLfloat a[4], result[4];
977 fetch_vector4(&inst->SrcReg[0], machine, a);
978 a[0] = MAX2(a[0], 0.0F);
979 a[1] = MAX2(a[1], 0.0F);
980 /* XXX ARB version clamps a[3], NV version doesn't */
981 a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
982 result[0] = 1.0F;
983 result[1] = a[0];
984 /* XXX we could probably just use pow() here */
985 if (a[0] > 0.0F) {
986 if (a[1] == 0.0 && a[3] == 0.0)
987 result[2] = 1.0;
988 else
989 result[2] = EXPF(a[3] * LOGF(a[1]));
990 }
991 else {
992 result[2] = 0.0;
993 }
994 result[3] = 1.0F;
995 store_vector4(inst, machine, result);
996 if (DEBUG_PROG) {
997 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
998 result[0], result[1], result[2], result[3],
999 a[0], a[1], a[2], a[3]);
1000 }
1001 }
1002 break;
1003 case OPCODE_LOG:
1004 {
1005 GLfloat t[4], q[4], abs_t0;
1006 fetch_vector1(&inst->SrcReg[0], machine, t);
1007 abs_t0 = FABSF(t[0]);
1008 if (abs_t0 != 0.0F) {
1009 /* Since we really can't handle infinite values on VMS
1010 * like other OSes we'll use __MAXFLOAT to represent
1011 * infinity. This may need some tweaking.
1012 */
1013 #ifdef VMS
1014 if (abs_t0 == __MAXFLOAT)
1015 #else
1016 if (IS_INF_OR_NAN(abs_t0))
1017 #endif
1018 {
1019 SET_POS_INFINITY(q[0]);
1020 q[1] = 1.0F;
1021 SET_POS_INFINITY(q[2]);
1022 }
1023 else {
1024 int exponent;
1025 GLfloat mantissa = FREXPF(t[0], &exponent);
1026 q[0] = (GLfloat) (exponent - 1);
1027 q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
1028
1029 /* The fast LOG2 macro doesn't meet the precision
1030 * requirements.
1031 */
1032 q[2] = (log(t[0]) * 1.442695F);
1033 }
1034 }
1035 else {
1036 SET_NEG_INFINITY(q[0]);
1037 q[1] = 1.0F;
1038 SET_NEG_INFINITY(q[2]);
1039 }
1040 q[3] = 1.0;
1041 store_vector4(inst, machine, q);
1042 }
1043 break;
1044 case OPCODE_LRP:
1045 {
1046 GLfloat a[4], b[4], c[4], result[4];
1047 fetch_vector4(&inst->SrcReg[0], machine, a);
1048 fetch_vector4(&inst->SrcReg[1], machine, b);
1049 fetch_vector4(&inst->SrcReg[2], machine, c);
1050 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
1051 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
1052 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
1053 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
1054 store_vector4(inst, machine, result);
1055 if (DEBUG_PROG) {
1056 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1057 "(%g %g %g %g), (%g %g %g %g)\n",
1058 result[0], result[1], result[2], result[3],
1059 a[0], a[1], a[2], a[3],
1060 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1061 }
1062 }
1063 break;
1064 case OPCODE_MAD:
1065 {
1066 GLfloat a[4], b[4], c[4], result[4];
1067 fetch_vector4(&inst->SrcReg[0], machine, a);
1068 fetch_vector4(&inst->SrcReg[1], machine, b);
1069 fetch_vector4(&inst->SrcReg[2], machine, c);
1070 result[0] = a[0] * b[0] + c[0];
1071 result[1] = a[1] * b[1] + c[1];
1072 result[2] = a[2] * b[2] + c[2];
1073 result[3] = a[3] * b[3] + c[3];
1074 store_vector4(inst, machine, result);
1075 if (DEBUG_PROG) {
1076 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1077 "(%g %g %g %g) + (%g %g %g %g)\n",
1078 result[0], result[1], result[2], result[3],
1079 a[0], a[1], a[2], a[3],
1080 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1081 }
1082 }
1083 break;
1084 case OPCODE_MAX:
1085 {
1086 GLfloat a[4], b[4], result[4];
1087 fetch_vector4(&inst->SrcReg[0], machine, a);
1088 fetch_vector4(&inst->SrcReg[1], machine, b);
1089 result[0] = MAX2(a[0], b[0]);
1090 result[1] = MAX2(a[1], b[1]);
1091 result[2] = MAX2(a[2], b[2]);
1092 result[3] = MAX2(a[3], b[3]);
1093 store_vector4(inst, machine, result);
1094 if (DEBUG_PROG) {
1095 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1096 result[0], result[1], result[2], result[3],
1097 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1098 }
1099 }
1100 break;
1101 case OPCODE_MIN:
1102 {
1103 GLfloat a[4], b[4], result[4];
1104 fetch_vector4(&inst->SrcReg[0], machine, a);
1105 fetch_vector4(&inst->SrcReg[1], machine, b);
1106 result[0] = MIN2(a[0], b[0]);
1107 result[1] = MIN2(a[1], b[1]);
1108 result[2] = MIN2(a[2], b[2]);
1109 result[3] = MIN2(a[3], b[3]);
1110 store_vector4(inst, machine, result);
1111 }
1112 break;
1113 case OPCODE_MOV:
1114 {
1115 GLfloat result[4];
1116 fetch_vector4(&inst->SrcReg[0], machine, result);
1117 store_vector4(inst, machine, result);
1118 if (DEBUG_PROG) {
1119 printf("MOV (%g %g %g %g)\n",
1120 result[0], result[1], result[2], result[3]);
1121 }
1122 }
1123 break;
1124 case OPCODE_MUL:
1125 {
1126 GLfloat a[4], b[4], result[4];
1127 fetch_vector4(&inst->SrcReg[0], machine, a);
1128 fetch_vector4(&inst->SrcReg[1], machine, b);
1129 result[0] = a[0] * b[0];
1130 result[1] = a[1] * b[1];
1131 result[2] = a[2] * b[2];
1132 result[3] = a[3] * b[3];
1133 store_vector4(inst, machine, result);
1134 if (DEBUG_PROG) {
1135 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1136 result[0], result[1], result[2], result[3],
1137 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1138 }
1139 }
1140 break;
1141 case OPCODE_NOISE1:
1142 {
1143 GLfloat a[4], result[4];
1144 fetch_vector1(&inst->SrcReg[0], machine, a);
1145 result[0] =
1146 result[1] =
1147 result[2] =
1148 result[3] = _mesa_noise1(a[0]);
1149 store_vector4(inst, machine, result);
1150 }
1151 break;
1152 case OPCODE_NOISE2:
1153 {
1154 GLfloat a[4], result[4];
1155 fetch_vector4(&inst->SrcReg[0], machine, a);
1156 result[0] =
1157 result[1] =
1158 result[2] = result[3] = _mesa_noise2(a[0], a[1]);
1159 store_vector4(inst, machine, result);
1160 }
1161 break;
1162 case OPCODE_NOISE3:
1163 {
1164 GLfloat a[4], result[4];
1165 fetch_vector4(&inst->SrcReg[0], machine, a);
1166 result[0] =
1167 result[1] =
1168 result[2] =
1169 result[3] = _mesa_noise3(a[0], a[1], a[2]);
1170 store_vector4(inst, machine, result);
1171 }
1172 break;
1173 case OPCODE_NOISE4:
1174 {
1175 GLfloat a[4], result[4];
1176 fetch_vector4(&inst->SrcReg[0], machine, a);
1177 result[0] =
1178 result[1] =
1179 result[2] =
1180 result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
1181 store_vector4(inst, machine, result);
1182 }
1183 break;
1184 case OPCODE_NOP:
1185 break;
1186 case OPCODE_NOT: /* bitwise NOT */
1187 {
1188 GLuint a[4], result[4];
1189 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1190 result[0] = ~a[0];
1191 result[1] = ~a[1];
1192 result[2] = ~a[2];
1193 result[3] = ~a[3];
1194 store_vector4ui(inst, machine, result);
1195 }
1196 break;
1197 case OPCODE_NRM3: /* 3-component normalization */
1198 {
1199 GLfloat a[4], result[4];
1200 GLfloat tmp;
1201 fetch_vector4(&inst->SrcReg[0], machine, a);
1202 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
1203 if (tmp != 0.0F)
1204 tmp = INV_SQRTF(tmp);
1205 result[0] = tmp * a[0];
1206 result[1] = tmp * a[1];
1207 result[2] = tmp * a[2];
1208 result[3] = 0.0; /* undefined, but prevent valgrind warnings */
1209 store_vector4(inst, machine, result);
1210 }
1211 break;
1212 case OPCODE_NRM4: /* 4-component normalization */
1213 {
1214 GLfloat a[4], result[4];
1215 GLfloat tmp;
1216 fetch_vector4(&inst->SrcReg[0], machine, a);
1217 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
1218 if (tmp != 0.0F)
1219 tmp = INV_SQRTF(tmp);
1220 result[0] = tmp * a[0];
1221 result[1] = tmp * a[1];
1222 result[2] = tmp * a[2];
1223 result[3] = tmp * a[3];
1224 store_vector4(inst, machine, result);
1225 }
1226 break;
1227 case OPCODE_OR: /* bitwise OR */
1228 {
1229 GLuint a[4], b[4], result[4];
1230 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1231 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1232 result[0] = a[0] | b[0];
1233 result[1] = a[1] | b[1];
1234 result[2] = a[2] | b[2];
1235 result[3] = a[3] | b[3];
1236 store_vector4ui(inst, machine, result);
1237 }
1238 break;
1239 case OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
1240 {
1241 GLfloat a[4];
1242 GLuint result[4];
1243 GLhalfNV hx, hy;
1244 fetch_vector4(&inst->SrcReg[0], machine, a);
1245 hx = _mesa_float_to_half(a[0]);
1246 hy = _mesa_float_to_half(a[1]);
1247 result[0] =
1248 result[1] =
1249 result[2] =
1250 result[3] = hx | (hy << 16);
1251 store_vector4ui(inst, machine, result);
1252 }
1253 break;
1254 case OPCODE_PK2US: /* pack two GLushorts into one 32-bit float */
1255 {
1256 GLfloat a[4];
1257 GLuint result[4], usx, usy;
1258 fetch_vector4(&inst->SrcReg[0], machine, a);
1259 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1260 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1261 usx = IROUND(a[0] * 65535.0F);
1262 usy = IROUND(a[1] * 65535.0F);
1263 result[0] =
1264 result[1] =
1265 result[2] =
1266 result[3] = usx | (usy << 16);
1267 store_vector4ui(inst, machine, result);
1268 }
1269 break;
1270 case OPCODE_PK4B: /* pack four GLbytes into one 32-bit float */
1271 {
1272 GLfloat a[4];
1273 GLuint result[4], ubx, uby, ubz, ubw;
1274 fetch_vector4(&inst->SrcReg[0], machine, a);
1275 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
1276 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
1277 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
1278 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
1279 ubx = IROUND(127.0F * a[0] + 128.0F);
1280 uby = IROUND(127.0F * a[1] + 128.0F);
1281 ubz = IROUND(127.0F * a[2] + 128.0F);
1282 ubw = IROUND(127.0F * a[3] + 128.0F);
1283 result[0] =
1284 result[1] =
1285 result[2] =
1286 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1287 store_vector4ui(inst, machine, result);
1288 }
1289 break;
1290 case OPCODE_PK4UB: /* pack four GLubytes into one 32-bit float */
1291 {
1292 GLfloat a[4];
1293 GLuint result[4], ubx, uby, ubz, ubw;
1294 fetch_vector4(&inst->SrcReg[0], machine, a);
1295 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1296 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1297 a[2] = CLAMP(a[2], 0.0F, 1.0F);
1298 a[3] = CLAMP(a[3], 0.0F, 1.0F);
1299 ubx = IROUND(255.0F * a[0]);
1300 uby = IROUND(255.0F * a[1]);
1301 ubz = IROUND(255.0F * a[2]);
1302 ubw = IROUND(255.0F * a[3]);
1303 result[0] =
1304 result[1] =
1305 result[2] =
1306 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1307 store_vector4ui(inst, machine, result);
1308 }
1309 break;
1310 case OPCODE_POW:
1311 {
1312 GLfloat a[4], b[4], result[4];
1313 fetch_vector1(&inst->SrcReg[0], machine, a);
1314 fetch_vector1(&inst->SrcReg[1], machine, b);
1315 result[0] = result[1] = result[2] = result[3]
1316 = (GLfloat) _mesa_pow(a[0], b[0]);
1317 store_vector4(inst, machine, result);
1318 }
1319 break;
1320 case OPCODE_RCP:
1321 {
1322 GLfloat a[4], result[4];
1323 fetch_vector1(&inst->SrcReg[0], machine, a);
1324 if (DEBUG_PROG) {
1325 if (a[0] == 0)
1326 printf("RCP(0)\n");
1327 else if (IS_INF_OR_NAN(a[0]))
1328 printf("RCP(inf)\n");
1329 }
1330 result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
1331 store_vector4(inst, machine, result);
1332 }
1333 break;
1334 case OPCODE_RET: /* return from subroutine (conditional) */
1335 if (eval_condition(machine, inst)) {
1336 if (machine->StackDepth == 0) {
1337 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
1338 }
1339 /* subtract one because of pc++ in the for loop */
1340 pc = machine->CallStack[--machine->StackDepth] - 1;
1341 }
1342 break;
1343 case OPCODE_RFL: /* reflection vector */
1344 {
1345 GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
1346 fetch_vector4(&inst->SrcReg[0], machine, axis);
1347 fetch_vector4(&inst->SrcReg[1], machine, dir);
1348 tmpW = DOT3(axis, axis);
1349 tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
1350 result[0] = tmpX * axis[0] - dir[0];
1351 result[1] = tmpX * axis[1] - dir[1];
1352 result[2] = tmpX * axis[2] - dir[2];
1353 /* result[3] is never written! XXX enforce in parser! */
1354 store_vector4(inst, machine, result);
1355 }
1356 break;
1357 case OPCODE_RSQ: /* 1 / sqrt() */
1358 {
1359 GLfloat a[4], result[4];
1360 fetch_vector1(&inst->SrcReg[0], machine, a);
1361 a[0] = FABSF(a[0]);
1362 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
1363 store_vector4(inst, machine, result);
1364 if (DEBUG_PROG) {
1365 printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
1366 }
1367 }
1368 break;
1369 case OPCODE_SCS: /* sine and cos */
1370 {
1371 GLfloat a[4], result[4];
1372 fetch_vector1(&inst->SrcReg[0], machine, a);
1373 result[0] = (GLfloat) _mesa_cos(a[0]);
1374 result[1] = (GLfloat) _mesa_sin(a[0]);
1375 result[2] = 0.0; /* undefined! */
1376 result[3] = 0.0; /* undefined! */
1377 store_vector4(inst, machine, result);
1378 }
1379 break;
1380 case OPCODE_SEQ: /* set on equal */
1381 {
1382 GLfloat a[4], b[4], result[4];
1383 fetch_vector4(&inst->SrcReg[0], machine, a);
1384 fetch_vector4(&inst->SrcReg[1], machine, b);
1385 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1386 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1387 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1388 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1389 store_vector4(inst, machine, result);
1390 if (DEBUG_PROG) {
1391 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1392 result[0], result[1], result[2], result[3],
1393 a[0], a[1], a[2], a[3],
1394 b[0], b[1], b[2], b[3]);
1395 }
1396 }
1397 break;
1398 case OPCODE_SFL: /* set false, operands ignored */
1399 {
1400 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1401 store_vector4(inst, machine, result);
1402 }
1403 break;
1404 case OPCODE_SGE: /* set on greater or equal */
1405 {
1406 GLfloat a[4], b[4], result[4];
1407 fetch_vector4(&inst->SrcReg[0], machine, a);
1408 fetch_vector4(&inst->SrcReg[1], machine, b);
1409 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1410 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1411 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1412 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1413 store_vector4(inst, machine, result);
1414 if (DEBUG_PROG) {
1415 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1416 result[0], result[1], result[2], result[3],
1417 a[0], a[1], a[2], a[3],
1418 b[0], b[1], b[2], b[3]);
1419 }
1420 }
1421 break;
1422 case OPCODE_SGT: /* set on greater */
1423 {
1424 GLfloat a[4], b[4], result[4];
1425 fetch_vector4(&inst->SrcReg[0], machine, a);
1426 fetch_vector4(&inst->SrcReg[1], machine, b);
1427 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1428 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1429 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1430 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1431 store_vector4(inst, machine, result);
1432 if (DEBUG_PROG) {
1433 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1434 result[0], result[1], result[2], result[3],
1435 a[0], a[1], a[2], a[3],
1436 b[0], b[1], b[2], b[3]);
1437 }
1438 }
1439 break;
1440 case OPCODE_SIN:
1441 {
1442 GLfloat a[4], result[4];
1443 fetch_vector1(&inst->SrcReg[0], machine, a);
1444 result[0] = result[1] = result[2] = result[3]
1445 = (GLfloat) _mesa_sin(a[0]);
1446 store_vector4(inst, machine, result);
1447 }
1448 break;
1449 case OPCODE_SLE: /* set on less or equal */
1450 {
1451 GLfloat a[4], b[4], result[4];
1452 fetch_vector4(&inst->SrcReg[0], machine, a);
1453 fetch_vector4(&inst->SrcReg[1], machine, b);
1454 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1455 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1456 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1457 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1458 store_vector4(inst, machine, result);
1459 if (DEBUG_PROG) {
1460 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1461 result[0], result[1], result[2], result[3],
1462 a[0], a[1], a[2], a[3],
1463 b[0], b[1], b[2], b[3]);
1464 }
1465 }
1466 break;
1467 case OPCODE_SLT: /* set on less */
1468 {
1469 GLfloat a[4], b[4], result[4];
1470 fetch_vector4(&inst->SrcReg[0], machine, a);
1471 fetch_vector4(&inst->SrcReg[1], machine, b);
1472 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1473 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1474 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1475 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1476 store_vector4(inst, machine, result);
1477 if (DEBUG_PROG) {
1478 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1479 result[0], result[1], result[2], result[3],
1480 a[0], a[1], a[2], a[3],
1481 b[0], b[1], b[2], b[3]);
1482 }
1483 }
1484 break;
1485 case OPCODE_SNE: /* set on not equal */
1486 {
1487 GLfloat a[4], b[4], result[4];
1488 fetch_vector4(&inst->SrcReg[0], machine, a);
1489 fetch_vector4(&inst->SrcReg[1], machine, b);
1490 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1491 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1492 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1493 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1494 store_vector4(inst, machine, result);
1495 if (DEBUG_PROG) {
1496 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1497 result[0], result[1], result[2], result[3],
1498 a[0], a[1], a[2], a[3],
1499 b[0], b[1], b[2], b[3]);
1500 }
1501 }
1502 break;
1503 case OPCODE_SSG: /* set sign (-1, 0 or +1) */
1504 {
1505 GLfloat a[4], result[4];
1506 fetch_vector4(&inst->SrcReg[0], machine, a);
1507 result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
1508 result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
1509 result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
1510 result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
1511 store_vector4(inst, machine, result);
1512 }
1513 break;
1514 case OPCODE_STR: /* set true, operands ignored */
1515 {
1516 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1517 store_vector4(inst, machine, result);
1518 }
1519 break;
1520 case OPCODE_SUB:
1521 {
1522 GLfloat a[4], b[4], result[4];
1523 fetch_vector4(&inst->SrcReg[0], machine, a);
1524 fetch_vector4(&inst->SrcReg[1], machine, b);
1525 result[0] = a[0] - b[0];
1526 result[1] = a[1] - b[1];
1527 result[2] = a[2] - b[2];
1528 result[3] = a[3] - b[3];
1529 store_vector4(inst, machine, result);
1530 if (DEBUG_PROG) {
1531 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1532 result[0], result[1], result[2], result[3],
1533 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1534 }
1535 }
1536 break;
1537 case OPCODE_SWZ: /* extended swizzle */
1538 {
1539 const struct prog_src_register *source = &inst->SrcReg[0];
1540 const GLfloat *src = get_src_register_pointer(source, machine);
1541 GLfloat result[4];
1542 GLuint i;
1543 for (i = 0; i < 4; i++) {
1544 const GLuint swz = GET_SWZ(source->Swizzle, i);
1545 if (swz == SWIZZLE_ZERO)
1546 result[i] = 0.0;
1547 else if (swz == SWIZZLE_ONE)
1548 result[i] = 1.0;
1549 else {
1550 ASSERT(swz >= 0);
1551 ASSERT(swz <= 3);
1552 result[i] = src[swz];
1553 }
1554 if (source->NegateBase & (1 << i))
1555 result[i] = -result[i];
1556 }
1557 store_vector4(inst, machine, result);
1558 }
1559 break;
1560 case OPCODE_TEX: /* Both ARB and NV frag prog */
1561 /* Simple texel lookup */
1562 {
1563 GLfloat texcoord[4], color[4];
1564 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1565
1566 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1567
1568 if (DEBUG_PROG) {
1569 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1570 color[0], color[1], color[2], color[3],
1571 inst->TexSrcUnit,
1572 texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
1573 }
1574 store_vector4(inst, machine, color);
1575 }
1576 break;
1577 case OPCODE_TXB: /* GL_ARB_fragment_program only */
1578 /* Texel lookup with LOD bias */
1579 {
1580 const struct gl_texture_unit *texUnit
1581 = &ctx->Texture.Unit[inst->TexSrcUnit];
1582 GLfloat texcoord[4], color[4], lodBias;
1583
1584 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1585
1586 /* texcoord[3] is the bias to add to lambda */
1587 lodBias = texUnit->LodBias + texcoord[3];
1588 if (texUnit->_Current) {
1589 lodBias += texUnit->_Current->LodBias;
1590 }
1591
1592 fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
1593
1594 store_vector4(inst, machine, color);
1595 }
1596 break;
1597 case OPCODE_TXD: /* GL_NV_fragment_program only */
1598 /* Texture lookup w/ partial derivatives for LOD */
1599 {
1600 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1601 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1602 fetch_vector4(&inst->SrcReg[1], machine, dtdx);
1603 fetch_vector4(&inst->SrcReg[2], machine, dtdy);
1604 machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
1605 0.0, /* lodBias */
1606 inst->TexSrcUnit, color);
1607 store_vector4(inst, machine, color);
1608 }
1609 break;
1610 case OPCODE_TXP: /* GL_ARB_fragment_program only */
1611 /* Texture lookup w/ projective divide */
1612 {
1613 GLfloat texcoord[4], color[4];
1614
1615 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1616 /* Not so sure about this test - if texcoord[3] is
1617 * zero, we'd probably be fine except for an ASSERT in
1618 * IROUND_POS() which gets triggered by the inf values created.
1619 */
1620 if (texcoord[3] != 0.0) {
1621 texcoord[0] /= texcoord[3];
1622 texcoord[1] /= texcoord[3];
1623 texcoord[2] /= texcoord[3];
1624 }
1625
1626 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1627
1628 store_vector4(inst, machine, color);
1629 }
1630 break;
1631 case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
1632 /* Texture lookup w/ projective divide, as above, but do not
1633 * do the divide by w if sampling from a cube map.
1634 */
1635 {
1636 GLfloat texcoord[4], color[4];
1637
1638 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1639 if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
1640 texcoord[3] != 0.0) {
1641 texcoord[0] /= texcoord[3];
1642 texcoord[1] /= texcoord[3];
1643 texcoord[2] /= texcoord[3];
1644 }
1645
1646 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1647
1648 store_vector4(inst, machine, color);
1649 }
1650 break;
1651 case OPCODE_TRUNC: /* truncate toward zero */
1652 {
1653 GLfloat a[4], result[4];
1654 fetch_vector4(&inst->SrcReg[0], machine, a);
1655 result[0] = (GLfloat) (GLint) a[0];
1656 result[1] = (GLfloat) (GLint) a[1];
1657 result[2] = (GLfloat) (GLint) a[2];
1658 result[3] = (GLfloat) (GLint) a[3];
1659 store_vector4(inst, machine, result);
1660 }
1661 break;
1662 case OPCODE_UP2H: /* unpack two 16-bit floats */
1663 {
1664 GLfloat a[4], result[4];
1665 const GLuint *rawBits = (const GLuint *) a;
1666 GLhalfNV hx, hy;
1667 fetch_vector1(&inst->SrcReg[0], machine, a);
1668 hx = rawBits[0] & 0xffff;
1669 hy = rawBits[0] >> 16;
1670 result[0] = result[2] = _mesa_half_to_float(hx);
1671 result[1] = result[3] = _mesa_half_to_float(hy);
1672 store_vector4(inst, machine, result);
1673 }
1674 break;
1675 case OPCODE_UP2US: /* unpack two GLushorts */
1676 {
1677 GLfloat a[4], result[4];
1678 const GLuint *rawBits = (const GLuint *) a;
1679 GLushort usx, usy;
1680 fetch_vector1(&inst->SrcReg[0], machine, a);
1681 usx = rawBits[0] & 0xffff;
1682 usy = rawBits[0] >> 16;
1683 result[0] = result[2] = usx * (1.0f / 65535.0f);
1684 result[1] = result[3] = usy * (1.0f / 65535.0f);
1685 store_vector4(inst, machine, result);
1686 }
1687 break;
1688 case OPCODE_UP4B: /* unpack four GLbytes */
1689 {
1690 GLfloat a[4], result[4];
1691 const GLuint *rawBits = (const GLuint *) a;
1692 fetch_vector1(&inst->SrcReg[0], machine, a);
1693 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1694 result[1] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1695 result[2] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1696 result[3] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1697 store_vector4(inst, machine, result);
1698 }
1699 break;
1700 case OPCODE_UP4UB: /* unpack four GLubytes */
1701 {
1702 GLfloat a[4], result[4];
1703 const GLuint *rawBits = (const GLuint *) a;
1704 fetch_vector1(&inst->SrcReg[0], machine, a);
1705 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1706 result[1] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1707 result[2] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1708 result[3] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1709 store_vector4(inst, machine, result);
1710 }
1711 break;
1712 case OPCODE_XOR: /* bitwise XOR */
1713 {
1714 GLuint a[4], b[4], result[4];
1715 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1716 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1717 result[0] = a[0] ^ b[0];
1718 result[1] = a[1] ^ b[1];
1719 result[2] = a[2] ^ b[2];
1720 result[3] = a[3] ^ b[3];
1721 store_vector4ui(inst, machine, result);
1722 }
1723 break;
1724 case OPCODE_XPD: /* cross product */
1725 {
1726 GLfloat a[4], b[4], result[4];
1727 fetch_vector4(&inst->SrcReg[0], machine, a);
1728 fetch_vector4(&inst->SrcReg[1], machine, b);
1729 result[0] = a[1] * b[2] - a[2] * b[1];
1730 result[1] = a[2] * b[0] - a[0] * b[2];
1731 result[2] = a[0] * b[1] - a[1] * b[0];
1732 result[3] = 1.0;
1733 store_vector4(inst, machine, result);
1734 if (DEBUG_PROG) {
1735 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1736 result[0], result[1], result[2], result[3],
1737 a[0], a[1], a[2], b[0], b[1], b[2]);
1738 }
1739 }
1740 break;
1741 case OPCODE_X2D: /* 2-D matrix transform */
1742 {
1743 GLfloat a[4], b[4], c[4], result[4];
1744 fetch_vector4(&inst->SrcReg[0], machine, a);
1745 fetch_vector4(&inst->SrcReg[1], machine, b);
1746 fetch_vector4(&inst->SrcReg[2], machine, c);
1747 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1748 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1749 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1750 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1751 store_vector4(inst, machine, result);
1752 }
1753 break;
1754 case OPCODE_PRINT:
1755 {
1756 if (inst->SrcReg[0].File != -1) {
1757 GLfloat a[4];
1758 fetch_vector4(&inst->SrcReg[0], machine, a);
1759 _mesa_printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
1760 a[0], a[1], a[2], a[3]);
1761 }
1762 else {
1763 _mesa_printf("%s\n", (const char *) inst->Data);
1764 }
1765 }
1766 break;
1767 case OPCODE_END:
1768 return GL_TRUE;
1769 default:
1770 _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
1771 inst->Opcode);
1772 return GL_TRUE; /* return value doesn't matter */
1773 }
1774
1775 numExec++;
1776 if (numExec > maxExec) {
1777 _mesa_problem(ctx, "Infinite loop detected in fragment program");
1778 return GL_TRUE;
1779 }
1780
1781 } /* for pc */
1782
1783 #if FEATURE_MESA_program_debug
1784 CurrentMachine = NULL;
1785 #endif
1786
1787 return GL_TRUE;
1788 }