projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge commit 'origin/gallium-0.1' into gallium-0.2
[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 "shader/slang/slang_library_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] = (GLint) FLOORF(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 result[0] = result[1] = result[2] = result[3] = LOG2(a[0]);
967 store_vector4(inst, machine, result);
968 }
969 break;
970 case OPCODE_LIT:
971 {
972 const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
973 GLfloat a[4], result[4];
974 fetch_vector4(&inst->SrcReg[0], machine, a);
975 a[0] = MAX2(a[0], 0.0F);
976 a[1] = MAX2(a[1], 0.0F);
977 /* XXX ARB version clamps a[3], NV version doesn't */
978 a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
979 result[0] = 1.0F;
980 result[1] = a[0];
981 /* XXX we could probably just use pow() here */
982 if (a[0] > 0.0F) {
983 if (a[1] == 0.0 && a[3] == 0.0)
984 result[2] = 1.0;
985 else
986 result[2] = EXPF(a[3] * LOGF(a[1]));
987 }
988 else {
989 result[2] = 0.0;
990 }
991 result[3] = 1.0F;
992 store_vector4(inst, machine, result);
993 if (DEBUG_PROG) {
994 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
995 result[0], result[1], result[2], result[3],
996 a[0], a[1], a[2], a[3]);
997 }
998 }
999 break;
1000 case OPCODE_LOG:
1001 {
1002 GLfloat t[4], q[4], abs_t0;
1003 fetch_vector1(&inst->SrcReg[0], machine, t);
1004 abs_t0 = FABSF(t[0]);
1005 if (abs_t0 != 0.0F) {
1006 /* Since we really can't handle infinite values on VMS
1007 * like other OSes we'll use __MAXFLOAT to represent
1008 * infinity. This may need some tweaking.
1009 */
1010 #ifdef VMS
1011 if (abs_t0 == __MAXFLOAT)
1012 #else
1013 if (IS_INF_OR_NAN(abs_t0))
1014 #endif
1015 {
1016 SET_POS_INFINITY(q[0]);
1017 q[1] = 1.0F;
1018 SET_POS_INFINITY(q[2]);
1019 }
1020 else {
1021 int exponent;
1022 GLfloat mantissa = FREXPF(t[0], &exponent);
1023 q[0] = (GLfloat) (exponent - 1);
1024 q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
1025 q[2] = (GLfloat) (q[0] + LOG2(q[1]));
1026 }
1027 }
1028 else {
1029 SET_NEG_INFINITY(q[0]);
1030 q[1] = 1.0F;
1031 SET_NEG_INFINITY(q[2]);
1032 }
1033 q[3] = 1.0;
1034 store_vector4(inst, machine, q);
1035 }
1036 break;
1037 case OPCODE_LRP:
1038 {
1039 GLfloat a[4], b[4], c[4], result[4];
1040 fetch_vector4(&inst->SrcReg[0], machine, a);
1041 fetch_vector4(&inst->SrcReg[1], machine, b);
1042 fetch_vector4(&inst->SrcReg[2], machine, c);
1043 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
1044 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
1045 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
1046 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
1047 store_vector4(inst, machine, result);
1048 if (DEBUG_PROG) {
1049 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1050 "(%g %g %g %g), (%g %g %g %g)\n",
1051 result[0], result[1], result[2], result[3],
1052 a[0], a[1], a[2], a[3],
1053 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1054 }
1055 }
1056 break;
1057 case OPCODE_MAD:
1058 {
1059 GLfloat a[4], b[4], c[4], result[4];
1060 fetch_vector4(&inst->SrcReg[0], machine, a);
1061 fetch_vector4(&inst->SrcReg[1], machine, b);
1062 fetch_vector4(&inst->SrcReg[2], machine, c);
1063 result[0] = a[0] * b[0] + c[0];
1064 result[1] = a[1] * b[1] + c[1];
1065 result[2] = a[2] * b[2] + c[2];
1066 result[3] = a[3] * b[3] + c[3];
1067 store_vector4(inst, machine, result);
1068 if (DEBUG_PROG) {
1069 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1070 "(%g %g %g %g) + (%g %g %g %g)\n",
1071 result[0], result[1], result[2], result[3],
1072 a[0], a[1], a[2], a[3],
1073 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
1074 }
1075 }
1076 break;
1077 case OPCODE_MAX:
1078 {
1079 GLfloat a[4], b[4], result[4];
1080 fetch_vector4(&inst->SrcReg[0], machine, a);
1081 fetch_vector4(&inst->SrcReg[1], machine, b);
1082 result[0] = MAX2(a[0], b[0]);
1083 result[1] = MAX2(a[1], b[1]);
1084 result[2] = MAX2(a[2], b[2]);
1085 result[3] = MAX2(a[3], b[3]);
1086 store_vector4(inst, machine, result);
1087 if (DEBUG_PROG) {
1088 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1089 result[0], result[1], result[2], result[3],
1090 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1091 }
1092 }
1093 break;
1094 case OPCODE_MIN:
1095 {
1096 GLfloat a[4], b[4], result[4];
1097 fetch_vector4(&inst->SrcReg[0], machine, a);
1098 fetch_vector4(&inst->SrcReg[1], machine, b);
1099 result[0] = MIN2(a[0], b[0]);
1100 result[1] = MIN2(a[1], b[1]);
1101 result[2] = MIN2(a[2], b[2]);
1102 result[3] = MIN2(a[3], b[3]);
1103 store_vector4(inst, machine, result);
1104 }
1105 break;
1106 case OPCODE_MOV:
1107 {
1108 GLfloat result[4];
1109 fetch_vector4(&inst->SrcReg[0], machine, result);
1110 store_vector4(inst, machine, result);
1111 if (DEBUG_PROG) {
1112 printf("MOV (%g %g %g %g)\n",
1113 result[0], result[1], result[2], result[3]);
1114 }
1115 }
1116 break;
1117 case OPCODE_MUL:
1118 {
1119 GLfloat a[4], b[4], result[4];
1120 fetch_vector4(&inst->SrcReg[0], machine, a);
1121 fetch_vector4(&inst->SrcReg[1], machine, b);
1122 result[0] = a[0] * b[0];
1123 result[1] = a[1] * b[1];
1124 result[2] = a[2] * b[2];
1125 result[3] = a[3] * b[3];
1126 store_vector4(inst, machine, result);
1127 if (DEBUG_PROG) {
1128 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1129 result[0], result[1], result[2], result[3],
1130 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1131 }
1132 }
1133 break;
1134 case OPCODE_NOISE1:
1135 {
1136 GLfloat a[4], result[4];
1137 fetch_vector1(&inst->SrcReg[0], machine, a);
1138 result[0] =
1139 result[1] =
1140 result[2] = result[3] = _slang_library_noise1(a[0]);
1141 store_vector4(inst, machine, result);
1142 }
1143 break;
1144 case OPCODE_NOISE2:
1145 {
1146 GLfloat a[4], result[4];
1147 fetch_vector4(&inst->SrcReg[0], machine, a);
1148 result[0] =
1149 result[1] =
1150 result[2] = result[3] = _slang_library_noise2(a[0], a[1]);
1151 store_vector4(inst, machine, result);
1152 }
1153 break;
1154 case OPCODE_NOISE3:
1155 {
1156 GLfloat a[4], result[4];
1157 fetch_vector4(&inst->SrcReg[0], machine, a);
1158 result[0] =
1159 result[1] =
1160 result[2] =
1161 result[3] = _slang_library_noise3(a[0], a[1], a[2]);
1162 store_vector4(inst, machine, result);
1163 }
1164 break;
1165 case OPCODE_NOISE4:
1166 {
1167 GLfloat a[4], result[4];
1168 fetch_vector4(&inst->SrcReg[0], machine, a);
1169 result[0] =
1170 result[1] =
1171 result[2] =
1172 result[3] = _slang_library_noise4(a[0], a[1], a[2], a[3]);
1173 store_vector4(inst, machine, result);
1174 }
1175 break;
1176 case OPCODE_NOP:
1177 break;
1178 case OPCODE_NOT: /* bitwise NOT */
1179 {
1180 GLuint a[4], result[4];
1181 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1182 result[0] = ~a[0];
1183 result[1] = ~a[1];
1184 result[2] = ~a[2];
1185 result[3] = ~a[3];
1186 store_vector4ui(inst, machine, result);
1187 }
1188 break;
1189 case OPCODE_NRM3: /* 3-component normalization */
1190 {
1191 GLfloat a[4], result[4];
1192 GLfloat tmp;
1193 fetch_vector4(&inst->SrcReg[0], machine, a);
1194 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
1195 if (tmp != 0.0F)
1196 tmp = INV_SQRTF(tmp);
1197 result[0] = tmp * a[0];
1198 result[1] = tmp * a[1];
1199 result[2] = tmp * a[2];
1200 result[3] = 0.0; /* undefined, but prevent valgrind warnings */
1201 store_vector4(inst, machine, result);
1202 }
1203 break;
1204 case OPCODE_NRM4: /* 4-component normalization */
1205 {
1206 GLfloat a[4], result[4];
1207 GLfloat tmp;
1208 fetch_vector4(&inst->SrcReg[0], machine, a);
1209 tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
1210 if (tmp != 0.0F)
1211 tmp = INV_SQRTF(tmp);
1212 result[0] = tmp * a[0];
1213 result[1] = tmp * a[1];
1214 result[2] = tmp * a[2];
1215 result[3] = tmp * a[3];
1216 store_vector4(inst, machine, result);
1217 }
1218 break;
1219 case OPCODE_OR: /* bitwise OR */
1220 {
1221 GLuint a[4], b[4], result[4];
1222 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1223 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1224 result[0] = a[0] | b[0];
1225 result[1] = a[1] | b[1];
1226 result[2] = a[2] | b[2];
1227 result[3] = a[3] | b[3];
1228 store_vector4ui(inst, machine, result);
1229 }
1230 break;
1231 case OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
1232 {
1233 GLfloat a[4];
1234 GLuint result[4];
1235 GLhalfNV hx, hy;
1236 fetch_vector4(&inst->SrcReg[0], machine, a);
1237 hx = _mesa_float_to_half(a[0]);
1238 hy = _mesa_float_to_half(a[1]);
1239 result[0] =
1240 result[1] =
1241 result[2] =
1242 result[3] = hx | (hy << 16);
1243 store_vector4ui(inst, machine, result);
1244 }
1245 break;
1246 case OPCODE_PK2US: /* pack two GLushorts into one 32-bit float */
1247 {
1248 GLfloat a[4];
1249 GLuint result[4], usx, usy;
1250 fetch_vector4(&inst->SrcReg[0], machine, a);
1251 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1252 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1253 usx = IROUND(a[0] * 65535.0F);
1254 usy = IROUND(a[1] * 65535.0F);
1255 result[0] =
1256 result[1] =
1257 result[2] =
1258 result[3] = usx | (usy << 16);
1259 store_vector4ui(inst, machine, result);
1260 }
1261 break;
1262 case OPCODE_PK4B: /* pack four GLbytes into one 32-bit float */
1263 {
1264 GLfloat a[4];
1265 GLuint result[4], ubx, uby, ubz, ubw;
1266 fetch_vector4(&inst->SrcReg[0], machine, a);
1267 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
1268 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
1269 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
1270 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
1271 ubx = IROUND(127.0F * a[0] + 128.0F);
1272 uby = IROUND(127.0F * a[1] + 128.0F);
1273 ubz = IROUND(127.0F * a[2] + 128.0F);
1274 ubw = IROUND(127.0F * a[3] + 128.0F);
1275 result[0] =
1276 result[1] =
1277 result[2] =
1278 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1279 store_vector4ui(inst, machine, result);
1280 }
1281 break;
1282 case OPCODE_PK4UB: /* pack four GLubytes into one 32-bit float */
1283 {
1284 GLfloat a[4];
1285 GLuint result[4], ubx, uby, ubz, ubw;
1286 fetch_vector4(&inst->SrcReg[0], machine, a);
1287 a[0] = CLAMP(a[0], 0.0F, 1.0F);
1288 a[1] = CLAMP(a[1], 0.0F, 1.0F);
1289 a[2] = CLAMP(a[2], 0.0F, 1.0F);
1290 a[3] = CLAMP(a[3], 0.0F, 1.0F);
1291 ubx = IROUND(255.0F * a[0]);
1292 uby = IROUND(255.0F * a[1]);
1293 ubz = IROUND(255.0F * a[2]);
1294 ubw = IROUND(255.0F * a[3]);
1295 result[0] =
1296 result[1] =
1297 result[2] =
1298 result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
1299 store_vector4ui(inst, machine, result);
1300 }
1301 break;
1302 case OPCODE_POW:
1303 {
1304 GLfloat a[4], b[4], result[4];
1305 fetch_vector1(&inst->SrcReg[0], machine, a);
1306 fetch_vector1(&inst->SrcReg[1], machine, b);
1307 result[0] = result[1] = result[2] = result[3]
1308 = (GLfloat) _mesa_pow(a[0], b[0]);
1309 store_vector4(inst, machine, result);
1310 }
1311 break;
1312 case OPCODE_RCP:
1313 {
1314 GLfloat a[4], result[4];
1315 fetch_vector1(&inst->SrcReg[0], machine, a);
1316 if (DEBUG_PROG) {
1317 if (a[0] == 0)
1318 printf("RCP(0)\n");
1319 else if (IS_INF_OR_NAN(a[0]))
1320 printf("RCP(inf)\n");
1321 }
1322 result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
1323 store_vector4(inst, machine, result);
1324 }
1325 break;
1326 case OPCODE_RET: /* return from subroutine (conditional) */
1327 if (eval_condition(machine, inst)) {
1328 if (machine->StackDepth == 0) {
1329 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
1330 }
1331 /* subtract one because of pc++ in the for loop */
1332 pc = machine->CallStack[--machine->StackDepth] - 1;
1333 }
1334 break;
1335 case OPCODE_RFL: /* reflection vector */
1336 {
1337 GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
1338 fetch_vector4(&inst->SrcReg[0], machine, axis);
1339 fetch_vector4(&inst->SrcReg[1], machine, dir);
1340 tmpW = DOT3(axis, axis);
1341 tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
1342 result[0] = tmpX * axis[0] - dir[0];
1343 result[1] = tmpX * axis[1] - dir[1];
1344 result[2] = tmpX * axis[2] - dir[2];
1345 /* result[3] is never written! XXX enforce in parser! */
1346 store_vector4(inst, machine, result);
1347 }
1348 break;
1349 case OPCODE_RSQ: /* 1 / sqrt() */
1350 {
1351 GLfloat a[4], result[4];
1352 fetch_vector1(&inst->SrcReg[0], machine, a);
1353 a[0] = FABSF(a[0]);
1354 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
1355 store_vector4(inst, machine, result);
1356 if (DEBUG_PROG) {
1357 printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
1358 }
1359 }
1360 break;
1361 case OPCODE_SCS: /* sine and cos */
1362 {
1363 GLfloat a[4], result[4];
1364 fetch_vector1(&inst->SrcReg[0], machine, a);
1365 result[0] = (GLfloat) _mesa_cos(a[0]);
1366 result[1] = (GLfloat) _mesa_sin(a[0]);
1367 result[2] = 0.0; /* undefined! */
1368 result[3] = 0.0; /* undefined! */
1369 store_vector4(inst, machine, result);
1370 }
1371 break;
1372 case OPCODE_SEQ: /* set on equal */
1373 {
1374 GLfloat a[4], b[4], result[4];
1375 fetch_vector4(&inst->SrcReg[0], machine, a);
1376 fetch_vector4(&inst->SrcReg[1], machine, b);
1377 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1378 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1379 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1380 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1381 store_vector4(inst, machine, result);
1382 if (DEBUG_PROG) {
1383 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1384 result[0], result[1], result[2], result[3],
1385 a[0], a[1], a[2], a[3],
1386 b[0], b[1], b[2], b[3]);
1387 }
1388 }
1389 break;
1390 case OPCODE_SFL: /* set false, operands ignored */
1391 {
1392 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1393 store_vector4(inst, machine, result);
1394 }
1395 break;
1396 case OPCODE_SGE: /* set on greater or equal */
1397 {
1398 GLfloat a[4], b[4], result[4];
1399 fetch_vector4(&inst->SrcReg[0], machine, a);
1400 fetch_vector4(&inst->SrcReg[1], machine, b);
1401 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1402 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1403 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1404 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1405 store_vector4(inst, machine, result);
1406 if (DEBUG_PROG) {
1407 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1408 result[0], result[1], result[2], result[3],
1409 a[0], a[1], a[2], a[3],
1410 b[0], b[1], b[2], b[3]);
1411 }
1412 }
1413 break;
1414 case OPCODE_SGT: /* set on greater */
1415 {
1416 GLfloat a[4], b[4], result[4];
1417 fetch_vector4(&inst->SrcReg[0], machine, a);
1418 fetch_vector4(&inst->SrcReg[1], machine, b);
1419 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1420 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1421 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1422 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1423 store_vector4(inst, machine, result);
1424 if (DEBUG_PROG) {
1425 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1426 result[0], result[1], result[2], result[3],
1427 a[0], a[1], a[2], a[3],
1428 b[0], b[1], b[2], b[3]);
1429 }
1430 }
1431 break;
1432 case OPCODE_SIN:
1433 {
1434 GLfloat a[4], result[4];
1435 fetch_vector1(&inst->SrcReg[0], machine, a);
1436 result[0] = result[1] = result[2] = result[3]
1437 = (GLfloat) _mesa_sin(a[0]);
1438 store_vector4(inst, machine, result);
1439 }
1440 break;
1441 case OPCODE_SLE: /* set on less or equal */
1442 {
1443 GLfloat a[4], b[4], result[4];
1444 fetch_vector4(&inst->SrcReg[0], machine, a);
1445 fetch_vector4(&inst->SrcReg[1], machine, b);
1446 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1447 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1448 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1449 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1450 store_vector4(inst, machine, result);
1451 if (DEBUG_PROG) {
1452 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1453 result[0], result[1], result[2], result[3],
1454 a[0], a[1], a[2], a[3],
1455 b[0], b[1], b[2], b[3]);
1456 }
1457 }
1458 break;
1459 case OPCODE_SLT: /* set on less */
1460 {
1461 GLfloat a[4], b[4], result[4];
1462 fetch_vector4(&inst->SrcReg[0], machine, a);
1463 fetch_vector4(&inst->SrcReg[1], machine, b);
1464 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1465 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1466 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1467 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1468 store_vector4(inst, machine, result);
1469 if (DEBUG_PROG) {
1470 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1471 result[0], result[1], result[2], result[3],
1472 a[0], a[1], a[2], a[3],
1473 b[0], b[1], b[2], b[3]);
1474 }
1475 }
1476 break;
1477 case OPCODE_SNE: /* set on not equal */
1478 {
1479 GLfloat a[4], b[4], result[4];
1480 fetch_vector4(&inst->SrcReg[0], machine, a);
1481 fetch_vector4(&inst->SrcReg[1], machine, b);
1482 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1483 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1484 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1485 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1486 store_vector4(inst, machine, result);
1487 if (DEBUG_PROG) {
1488 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1489 result[0], result[1], result[2], result[3],
1490 a[0], a[1], a[2], a[3],
1491 b[0], b[1], b[2], b[3]);
1492 }
1493 }
1494 break;
1495 case OPCODE_SSG: /* set sign (-1, 0 or +1) */
1496 {
1497 GLfloat a[4], result[4];
1498 fetch_vector4(&inst->SrcReg[0], machine, a);
1499 result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
1500 result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
1501 result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
1502 result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
1503 store_vector4(inst, machine, result);
1504 }
1505 break;
1506 case OPCODE_STR: /* set true, operands ignored */
1507 {
1508 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1509 store_vector4(inst, machine, result);
1510 }
1511 break;
1512 case OPCODE_SUB:
1513 {
1514 GLfloat a[4], b[4], result[4];
1515 fetch_vector4(&inst->SrcReg[0], machine, a);
1516 fetch_vector4(&inst->SrcReg[1], machine, b);
1517 result[0] = a[0] - b[0];
1518 result[1] = a[1] - b[1];
1519 result[2] = a[2] - b[2];
1520 result[3] = a[3] - b[3];
1521 store_vector4(inst, machine, result);
1522 if (DEBUG_PROG) {
1523 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1524 result[0], result[1], result[2], result[3],
1525 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1526 }
1527 }
1528 break;
1529 case OPCODE_SWZ: /* extended swizzle */
1530 {
1531 const struct prog_src_register *source = &inst->SrcReg[0];
1532 const GLfloat *src = get_src_register_pointer(source, machine);
1533 GLfloat result[4];
1534 GLuint i;
1535 for (i = 0; i < 4; i++) {
1536 const GLuint swz = GET_SWZ(source->Swizzle, i);
1537 if (swz == SWIZZLE_ZERO)
1538 result[i] = 0.0;
1539 else if (swz == SWIZZLE_ONE)
1540 result[i] = 1.0;
1541 else {
1542 ASSERT(swz >= 0);
1543 ASSERT(swz <= 3);
1544 result[i] = src[swz];
1545 }
1546 if (source->NegateBase & (1 << i))
1547 result[i] = -result[i];
1548 }
1549 store_vector4(inst, machine, result);
1550 }
1551 break;
1552 case OPCODE_TEX: /* Both ARB and NV frag prog */
1553 /* Simple texel lookup */
1554 {
1555 GLfloat texcoord[4], color[4];
1556 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1557
1558 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1559
1560 if (DEBUG_PROG) {
1561 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1562 color[0], color[1], color[2], color[3],
1563 inst->TexSrcUnit,
1564 texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
1565 }
1566 store_vector4(inst, machine, color);
1567 }
1568 break;
1569 case OPCODE_TXB: /* GL_ARB_fragment_program only */
1570 /* Texel lookup with LOD bias */
1571 {
1572 const struct gl_texture_unit *texUnit
1573 = &ctx->Texture.Unit[inst->TexSrcUnit];
1574 GLfloat texcoord[4], color[4], lodBias;
1575
1576 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1577
1578 /* texcoord[3] is the bias to add to lambda */
1579 lodBias = texUnit->LodBias + texcoord[3];
1580 if (texUnit->_Current) {
1581 lodBias += texUnit->_Current->LodBias;
1582 }
1583
1584 fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
1585
1586 store_vector4(inst, machine, color);
1587 }
1588 break;
1589 case OPCODE_TXD: /* GL_NV_fragment_program only */
1590 /* Texture lookup w/ partial derivatives for LOD */
1591 {
1592 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1593 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1594 fetch_vector4(&inst->SrcReg[1], machine, dtdx);
1595 fetch_vector4(&inst->SrcReg[2], machine, dtdy);
1596 machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
1597 0.0, /* lodBias */
1598 inst->TexSrcUnit, color);
1599 store_vector4(inst, machine, color);
1600 }
1601 break;
1602 case OPCODE_TXP: /* GL_ARB_fragment_program only */
1603 /* Texture lookup w/ projective divide */
1604 {
1605 GLfloat texcoord[4], color[4];
1606
1607 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1608 /* Not so sure about this test - if texcoord[3] is
1609 * zero, we'd probably be fine except for an ASSERT in
1610 * IROUND_POS() which gets triggered by the inf values created.
1611 */
1612 if (texcoord[3] != 0.0) {
1613 texcoord[0] /= texcoord[3];
1614 texcoord[1] /= texcoord[3];
1615 texcoord[2] /= texcoord[3];
1616 }
1617
1618 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1619
1620 store_vector4(inst, machine, color);
1621 }
1622 break;
1623 case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
1624 /* Texture lookup w/ projective divide, as above, but do not
1625 * do the divide by w if sampling from a cube map.
1626 */
1627 {
1628 GLfloat texcoord[4], color[4];
1629
1630 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1631 if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
1632 texcoord[3] != 0.0) {
1633 texcoord[0] /= texcoord[3];
1634 texcoord[1] /= texcoord[3];
1635 texcoord[2] /= texcoord[3];
1636 }
1637
1638 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1639
1640 store_vector4(inst, machine, color);
1641 }
1642 break;
1643 case OPCODE_TRUNC: /* truncate toward zero */
1644 {
1645 GLfloat a[4], result[4];
1646 fetch_vector4(&inst->SrcReg[0], machine, a);
1647 result[0] = (GLfloat) (GLint) a[0];
1648 result[1] = (GLfloat) (GLint) a[1];
1649 result[2] = (GLfloat) (GLint) a[2];
1650 result[3] = (GLfloat) (GLint) a[3];
1651 store_vector4(inst, machine, result);
1652 }
1653 break;
1654 case OPCODE_UP2H: /* unpack two 16-bit floats */
1655 {
1656 GLfloat a[4], result[4];
1657 const GLuint *rawBits = (const GLuint *) a;
1658 GLhalfNV hx, hy;
1659 fetch_vector1(&inst->SrcReg[0], machine, a);
1660 hx = rawBits[0] & 0xffff;
1661 hy = rawBits[0] >> 16;
1662 result[0] = result[2] = _mesa_half_to_float(hx);
1663 result[1] = result[3] = _mesa_half_to_float(hy);
1664 store_vector4(inst, machine, result);
1665 }
1666 break;
1667 case OPCODE_UP2US: /* unpack two GLushorts */
1668 {
1669 GLfloat a[4], result[4];
1670 const GLuint *rawBits = (const GLuint *) a;
1671 GLushort usx, usy;
1672 fetch_vector1(&inst->SrcReg[0], machine, a);
1673 usx = rawBits[0] & 0xffff;
1674 usy = rawBits[0] >> 16;
1675 result[0] = result[2] = usx * (1.0f / 65535.0f);
1676 result[1] = result[3] = usy * (1.0f / 65535.0f);
1677 store_vector4(inst, machine, result);
1678 }
1679 break;
1680 case OPCODE_UP4B: /* unpack four GLbytes */
1681 {
1682 GLfloat a[4], result[4];
1683 const GLuint *rawBits = (const GLuint *) a;
1684 fetch_vector1(&inst->SrcReg[0], machine, a);
1685 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1686 result[1] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1687 result[2] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1688 result[3] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1689 store_vector4(inst, machine, result);
1690 }
1691 break;
1692 case OPCODE_UP4UB: /* unpack four GLubytes */
1693 {
1694 GLfloat a[4], result[4];
1695 const GLuint *rawBits = (const GLuint *) a;
1696 fetch_vector1(&inst->SrcReg[0], machine, a);
1697 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1698 result[1] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1699 result[2] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1700 result[3] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1701 store_vector4(inst, machine, result);
1702 }
1703 break;
1704 case OPCODE_XOR: /* bitwise XOR */
1705 {
1706 GLuint a[4], b[4], result[4];
1707 fetch_vector4ui(&inst->SrcReg[0], machine, a);
1708 fetch_vector4ui(&inst->SrcReg[1], machine, b);
1709 result[0] = a[0] ^ b[0];
1710 result[1] = a[1] ^ b[1];
1711 result[2] = a[2] ^ b[2];
1712 result[3] = a[3] ^ b[3];
1713 store_vector4ui(inst, machine, result);
1714 }
1715 break;
1716 case OPCODE_XPD: /* cross product */
1717 {
1718 GLfloat a[4], b[4], result[4];
1719 fetch_vector4(&inst->SrcReg[0], machine, a);
1720 fetch_vector4(&inst->SrcReg[1], machine, b);
1721 result[0] = a[1] * b[2] - a[2] * b[1];
1722 result[1] = a[2] * b[0] - a[0] * b[2];
1723 result[2] = a[0] * b[1] - a[1] * b[0];
1724 result[3] = 1.0;
1725 store_vector4(inst, machine, result);
1726 if (DEBUG_PROG) {
1727 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1728 result[0], result[1], result[2], result[3],
1729 a[0], a[1], a[2], b[0], b[1], b[2]);
1730 }
1731 }
1732 break;
1733 case OPCODE_X2D: /* 2-D matrix transform */
1734 {
1735 GLfloat a[4], b[4], c[4], result[4];
1736 fetch_vector4(&inst->SrcReg[0], machine, a);
1737 fetch_vector4(&inst->SrcReg[1], machine, b);
1738 fetch_vector4(&inst->SrcReg[2], machine, c);
1739 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1740 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1741 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1742 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1743 store_vector4(inst, machine, result);
1744 }
1745 break;
1746 case OPCODE_PRINT:
1747 {
1748 if (inst->SrcReg[0].File != -1) {
1749 GLfloat a[4];
1750 fetch_vector4(&inst->SrcReg[0], machine, a);
1751 _mesa_printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
1752 a[0], a[1], a[2], a[3]);
1753 }
1754 else {
1755 _mesa_printf("%s\n", (const char *) inst->Data);
1756 }
1757 }
1758 break;
1759 case OPCODE_END:
1760 return GL_TRUE;
1761 default:
1762 _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
1763 inst->Opcode);
1764 return GL_TRUE; /* return value doesn't matter */
1765 }
1766
1767 numExec++;
1768 if (numExec > maxExec) {
1769 _mesa_problem(ctx, "Infinite loop detected in fragment program");
1770 return GL_TRUE;
1771 }
1772
1773 } /* for pc */
1774
1775 #if FEATURE_MESA_program_debug
1776 CurrentMachine = NULL;
1777 #endif
1778
1779 return GL_TRUE;
1780 }