projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
mesa: Fix some signed-unsigned comparison warnings
[mesa.git] / src / mesa / program / prog_execute.c
1 /*
2 * Mesa 3-D graphics library
3 *
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 * OTHER DEALINGS IN THE SOFTWARE.
23 */
24
25 /**
26 * \file prog_execute.c
27 * Software interpreter for vertex/fragment programs.
28 * \author Brian Paul
29 */
30
31 /*
32 * NOTE: we do everything in single-precision floating point; we don't
33 * currently observe the single/half/fixed-precision qualifiers.
34 *
35 */
36
37
38 #include "main/glheader.h"
39 #include "main/colormac.h"
40 #include "main/macros.h"
41 #include "prog_execute.h"
42 #include "prog_instruction.h"
43 #include "prog_parameter.h"
44 #include "prog_print.h"
45 #include "prog_noise.h"
46
47
48 /* debug predicate */
49 #define DEBUG_PROG 0
50
51
52 /**
53 * Set x to positive or negative infinity.
54 */
55 #define SET_POS_INFINITY(x) \
56 do { \
57 fi_type fi; \
58 fi.i = 0x7F800000; \
59 x = fi.f; \
60 } while (0)
61 #define SET_NEG_INFINITY(x) \
62 do { \
63 fi_type fi; \
64 fi.i = 0xFF800000; \
65 x = fi.f; \
66 } while (0)
67
68 #define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
69
70
71 static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
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 >= VARYING_SLOT_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_STATE_VAR:
117 /* Fallthrough */
118 case PROGRAM_CONSTANT:
119 /* Fallthrough */
120 case PROGRAM_UNIFORM:
121 if (reg >= (GLint) prog->Parameters->NumParameters)
122 return ZeroVec;
123 return (GLfloat *) prog->Parameters->ParameterValues[reg];
124
125 case PROGRAM_SYSTEM_VALUE:
126 assert(reg < (GLint) Elements(machine->SystemValues));
127 return machine->SystemValues[reg];
128
129 default:
130 _mesa_problem(NULL,
131 "Invalid src register file %d in get_src_register_pointer()",
132 source->File);
133 return ZeroVec;
134 }
135 }
136
137
138 /**
139 * Return a pointer to the 4-element float vector specified by the given
140 * destination register.
141 */
142 static inline GLfloat *
143 get_dst_register_pointer(const struct prog_dst_register *dest,
144 struct gl_program_machine *machine)
145 {
146 static GLfloat dummyReg[4];
147 GLint reg = dest->Index;
148
149 if (dest->RelAddr) {
150 /* add address register value to src index/offset */
151 reg += machine->AddressReg[0][0];
152 if (reg < 0) {
153 return dummyReg;
154 }
155 }
156
157 switch (dest->File) {
158 case PROGRAM_TEMPORARY:
159 if (reg >= MAX_PROGRAM_TEMPS)
160 return dummyReg;
161 return machine->Temporaries[reg];
162
163 case PROGRAM_OUTPUT:
164 if (reg >= MAX_PROGRAM_OUTPUTS)
165 return dummyReg;
166 return machine->Outputs[reg];
167
168 default:
169 _mesa_problem(NULL,
170 "Invalid dest register file %d in get_dst_register_pointer()",
171 dest->File);
172 return dummyReg;
173 }
174 }
175
176
177
178 /**
179 * Fetch a 4-element float vector from the given source register.
180 * Apply swizzling and negating as needed.
181 */
182 static void
183 fetch_vector4(const struct prog_src_register *source,
184 const struct gl_program_machine *machine, GLfloat result[4])
185 {
186 const GLfloat *src = get_src_register_pointer(source, machine);
187
188 if (source->Swizzle == SWIZZLE_NOOP) {
189 /* no swizzling */
190 COPY_4V(result, src);
191 }
192 else {
193 ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
194 ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
195 ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
196 ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
197 result[0] = src[GET_SWZ(source->Swizzle, 0)];
198 result[1] = src[GET_SWZ(source->Swizzle, 1)];
199 result[2] = src[GET_SWZ(source->Swizzle, 2)];
200 result[3] = src[GET_SWZ(source->Swizzle, 3)];
201 }
202
203 if (source->Abs) {
204 result[0] = FABSF(result[0]);
205 result[1] = FABSF(result[1]);
206 result[2] = FABSF(result[2]);
207 result[3] = FABSF(result[3]);
208 }
209 if (source->Negate) {
210 ASSERT(source->Negate == NEGATE_XYZW);
211 result[0] = -result[0];
212 result[1] = -result[1];
213 result[2] = -result[2];
214 result[3] = -result[3];
215 }
216
217 #ifdef NAN_CHECK
218 assert(!IS_INF_OR_NAN(result[0]));
219 assert(!IS_INF_OR_NAN(result[0]));
220 assert(!IS_INF_OR_NAN(result[0]));
221 assert(!IS_INF_OR_NAN(result[0]));
222 #endif
223 }
224
225
226 /**
227 * Fetch the derivative with respect to X or Y for the given register.
228 * XXX this currently only works for fragment program input attribs.
229 */
230 static void
231 fetch_vector4_deriv(struct gl_context * ctx,
232 const struct prog_src_register *source,
233 const struct gl_program_machine *machine,
234 char xOrY, GLfloat result[4])
235 {
236 if (source->File == PROGRAM_INPUT &&
237 source->Index < (GLint) machine->NumDeriv) {
238 const GLint col = machine->CurElement;
239 const GLfloat w = machine->Attribs[VARYING_SLOT_POS][col][3];
240 const GLfloat invQ = 1.0f / w;
241 GLfloat deriv[4];
242
243 if (xOrY == 'X') {
244 deriv[0] = machine->DerivX[source->Index][0] * invQ;
245 deriv[1] = machine->DerivX[source->Index][1] * invQ;
246 deriv[2] = machine->DerivX[source->Index][2] * invQ;
247 deriv[3] = machine->DerivX[source->Index][3] * invQ;
248 }
249 else {
250 deriv[0] = machine->DerivY[source->Index][0] * invQ;
251 deriv[1] = machine->DerivY[source->Index][1] * invQ;
252 deriv[2] = machine->DerivY[source->Index][2] * invQ;
253 deriv[3] = machine->DerivY[source->Index][3] * invQ;
254 }
255
256 result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
257 result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
258 result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
259 result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
260
261 if (source->Abs) {
262 result[0] = FABSF(result[0]);
263 result[1] = FABSF(result[1]);
264 result[2] = FABSF(result[2]);
265 result[3] = FABSF(result[3]);
266 }
267 if (source->Negate) {
268 ASSERT(source->Negate == NEGATE_XYZW);
269 result[0] = -result[0];
270 result[1] = -result[1];
271 result[2] = -result[2];
272 result[3] = -result[3];
273 }
274 }
275 else {
276 ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
277 }
278 }
279
280
281 /**
282 * As above, but only return result[0] element.
283 */
284 static void
285 fetch_vector1(const struct prog_src_register *source,
286 const struct gl_program_machine *machine, GLfloat result[4])
287 {
288 const GLfloat *src = get_src_register_pointer(source, machine);
289
290 result[0] = src[GET_SWZ(source->Swizzle, 0)];
291
292 if (source->Abs) {
293 result[0] = FABSF(result[0]);
294 }
295 if (source->Negate) {
296 result[0] = -result[0];
297 }
298 }
299
300
301 /**
302 * Fetch texel from texture. Use partial derivatives when possible.
303 */
304 static inline void
305 fetch_texel(struct gl_context *ctx,
306 const struct gl_program_machine *machine,
307 const struct prog_instruction *inst,
308 const GLfloat texcoord[4], GLfloat lodBias,
309 GLfloat color[4])
310 {
311 const GLuint unit = machine->Samplers[inst->TexSrcUnit];
312
313 /* Note: we only have the right derivatives for fragment input attribs.
314 */
315 if (machine->NumDeriv > 0 &&
316 inst->SrcReg[0].File == PROGRAM_INPUT &&
317 inst->SrcReg[0].Index == VARYING_SLOT_TEX0 + inst->TexSrcUnit) {
318 /* simple texture fetch for which we should have derivatives */
319 GLuint attr = inst->SrcReg[0].Index;
320 machine->FetchTexelDeriv(ctx, texcoord,
321 machine->DerivX[attr],
322 machine->DerivY[attr],
323 lodBias, unit, color);
324 }
325 else {
326 machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
327 }
328 }
329
330
331 /**
332 * Test value against zero and return GT, LT, EQ or UN if NaN.
333 */
334 static inline GLuint
335 generate_cc(float value)
336 {
337 if (value != value)
338 return COND_UN; /* NaN */
339 if (value > 0.0F)
340 return COND_GT;
341 if (value < 0.0F)
342 return COND_LT;
343 return COND_EQ;
344 }
345
346
347 /**
348 * Test if the ccMaskRule is satisfied by the given condition code.
349 * Used to mask destination writes according to the current condition code.
350 */
351 static inline GLboolean
352 test_cc(GLuint condCode, GLuint ccMaskRule)
353 {
354 switch (ccMaskRule) {
355 case COND_EQ: return (condCode == COND_EQ);
356 case COND_NE: return (condCode != COND_EQ);
357 case COND_LT: return (condCode == COND_LT);
358 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
359 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
360 case COND_GT: return (condCode == COND_GT);
361 case COND_TR: return GL_TRUE;
362 case COND_FL: return GL_FALSE;
363 default: return GL_TRUE;
364 }
365 }
366
367
368 /**
369 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
370 * or GL_FALSE to indicate result.
371 */
372 static inline GLboolean
373 eval_condition(const struct gl_program_machine *machine,
374 const struct prog_instruction *inst)
375 {
376 const GLuint swizzle = inst->DstReg.CondSwizzle;
377 const GLuint condMask = inst->DstReg.CondMask;
378 if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
379 test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
380 test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
381 test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
382 return GL_TRUE;
383 }
384 else {
385 return GL_FALSE;
386 }
387 }
388
389
390
391 /**
392 * Store 4 floats into a register. Observe the instructions saturate and
393 * set-condition-code flags.
394 */
395 static void
396 store_vector4(const struct prog_instruction *inst,
397 struct gl_program_machine *machine, const GLfloat value[4])
398 {
399 const struct prog_dst_register *dstReg = &(inst->DstReg);
400 const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
401 GLuint writeMask = dstReg->WriteMask;
402 GLfloat clampedValue[4];
403 GLfloat *dst = get_dst_register_pointer(dstReg, machine);
404
405 #if 0
406 if (value[0] > 1.0e10 ||
407 IS_INF_OR_NAN(value[0]) ||
408 IS_INF_OR_NAN(value[1]) ||
409 IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
410 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
411 #endif
412
413 if (clamp) {
414 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
415 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
416 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
417 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
418 value = clampedValue;
419 }
420
421 if (dstReg->CondMask != COND_TR) {
422 /* condition codes may turn off some writes */
423 if (writeMask & WRITEMASK_X) {
424 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
425 dstReg->CondMask))
426 writeMask &= ~WRITEMASK_X;
427 }
428 if (writeMask & WRITEMASK_Y) {
429 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
430 dstReg->CondMask))
431 writeMask &= ~WRITEMASK_Y;
432 }
433 if (writeMask & WRITEMASK_Z) {
434 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
435 dstReg->CondMask))
436 writeMask &= ~WRITEMASK_Z;
437 }
438 if (writeMask & WRITEMASK_W) {
439 if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
440 dstReg->CondMask))
441 writeMask &= ~WRITEMASK_W;
442 }
443 }
444
445 #ifdef NAN_CHECK
446 assert(!IS_INF_OR_NAN(value[0]));
447 assert(!IS_INF_OR_NAN(value[0]));
448 assert(!IS_INF_OR_NAN(value[0]));
449 assert(!IS_INF_OR_NAN(value[0]));
450 #endif
451
452 if (writeMask & WRITEMASK_X)
453 dst[0] = value[0];
454 if (writeMask & WRITEMASK_Y)
455 dst[1] = value[1];
456 if (writeMask & WRITEMASK_Z)
457 dst[2] = value[2];
458 if (writeMask & WRITEMASK_W)
459 dst[3] = value[3];
460
461 if (inst->CondUpdate) {
462 if (writeMask & WRITEMASK_X)
463 machine->CondCodes[0] = generate_cc(value[0]);
464 if (writeMask & WRITEMASK_Y)
465 machine->CondCodes[1] = generate_cc(value[1]);
466 if (writeMask & WRITEMASK_Z)
467 machine->CondCodes[2] = generate_cc(value[2]);
468 if (writeMask & WRITEMASK_W)
469 machine->CondCodes[3] = generate_cc(value[3]);
470 #if DEBUG_PROG
471 printf("CondCodes=(%s,%s,%s,%s) for:\n",
472 _mesa_condcode_string(machine->CondCodes[0]),
473 _mesa_condcode_string(machine->CondCodes[1]),
474 _mesa_condcode_string(machine->CondCodes[2]),
475 _mesa_condcode_string(machine->CondCodes[3]));
476 #endif
477 }
478 }
479
480
481 /**
482 * Execute the given vertex/fragment program.
483 *
484 * \param ctx rendering context
485 * \param program the program to execute
486 * \param machine machine state (must be initialized)
487 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
488 */
489 GLboolean
490 _mesa_execute_program(struct gl_context * ctx,
491 const struct gl_program *program,
492 struct gl_program_machine *machine)
493 {
494 const GLuint numInst = program->NumInstructions;
495 const GLuint maxExec = 65536;
496 GLuint pc, numExec = 0;
497
498 machine->CurProgram = program;
499
500 if (DEBUG_PROG) {
501 printf("execute program %u --------------------\n", program->Id);
502 }
503
504 if (program->Target == GL_VERTEX_PROGRAM_ARB) {
505 machine->EnvParams = ctx->VertexProgram.Parameters;
506 }
507 else {
508 machine->EnvParams = ctx->FragmentProgram.Parameters;
509 }
510
511 for (pc = 0; pc < numInst; pc++) {
512 const struct prog_instruction *inst = program->Instructions + pc;
513
514 if (DEBUG_PROG) {
515 _mesa_print_instruction(inst);
516 }
517
518 switch (inst->Opcode) {
519 case OPCODE_ABS:
520 {
521 GLfloat a[4], result[4];
522 fetch_vector4(&inst->SrcReg[0], machine, a);
523 result[0] = FABSF(a[0]);
524 result[1] = FABSF(a[1]);
525 result[2] = FABSF(a[2]);
526 result[3] = FABSF(a[3]);
527 store_vector4(inst, machine, result);
528 }
529 break;
530 case OPCODE_ADD:
531 {
532 GLfloat a[4], b[4], result[4];
533 fetch_vector4(&inst->SrcReg[0], machine, a);
534 fetch_vector4(&inst->SrcReg[1], machine, b);
535 result[0] = a[0] + b[0];
536 result[1] = a[1] + b[1];
537 result[2] = a[2] + b[2];
538 result[3] = a[3] + b[3];
539 store_vector4(inst, machine, result);
540 if (DEBUG_PROG) {
541 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
542 result[0], result[1], result[2], result[3],
543 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
544 }
545 }
546 break;
547 case OPCODE_ARL:
548 {
549 GLfloat t[4];
550 fetch_vector4(&inst->SrcReg[0], machine, t);
551 machine->AddressReg[0][0] = IFLOOR(t[0]);
552 if (DEBUG_PROG) {
553 printf("ARL %d\n", machine->AddressReg[0][0]);
554 }
555 }
556 break;
557 case OPCODE_BGNLOOP:
558 /* no-op */
559 ASSERT(program->Instructions[inst->BranchTarget].Opcode
560 == OPCODE_ENDLOOP);
561 break;
562 case OPCODE_ENDLOOP:
563 /* subtract 1 here since pc is incremented by for(pc) loop */
564 ASSERT(program->Instructions[inst->BranchTarget].Opcode
565 == OPCODE_BGNLOOP);
566 pc = inst->BranchTarget - 1; /* go to matching BNGLOOP */
567 break;
568 case OPCODE_BGNSUB: /* begin subroutine */
569 break;
570 case OPCODE_ENDSUB: /* end subroutine */
571 break;
572 case OPCODE_BRK: /* break out of loop (conditional) */
573 ASSERT(program->Instructions[inst->BranchTarget].Opcode
574 == OPCODE_ENDLOOP);
575 if (eval_condition(machine, inst)) {
576 /* break out of loop */
577 /* pc++ at end of for-loop will put us after the ENDLOOP inst */
578 pc = inst->BranchTarget;
579 }
580 break;
581 case OPCODE_CONT: /* continue loop (conditional) */
582 ASSERT(program->Instructions[inst->BranchTarget].Opcode
583 == OPCODE_ENDLOOP);
584 if (eval_condition(machine, inst)) {
585 /* continue at ENDLOOP */
586 /* Subtract 1 here since we'll do pc++ at end of for-loop */
587 pc = inst->BranchTarget - 1;
588 }
589 break;
590 case OPCODE_CAL: /* Call subroutine (conditional) */
591 if (eval_condition(machine, inst)) {
592 /* call the subroutine */
593 if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
594 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
595 }
596 machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
597 /* Subtract 1 here since we'll do pc++ at end of for-loop */
598 pc = inst->BranchTarget - 1;
599 }
600 break;
601 case OPCODE_CMP:
602 {
603 GLfloat a[4], b[4], c[4], result[4];
604 fetch_vector4(&inst->SrcReg[0], machine, a);
605 fetch_vector4(&inst->SrcReg[1], machine, b);
606 fetch_vector4(&inst->SrcReg[2], machine, c);
607 result[0] = a[0] < 0.0F ? b[0] : c[0];
608 result[1] = a[1] < 0.0F ? b[1] : c[1];
609 result[2] = a[2] < 0.0F ? b[2] : c[2];
610 result[3] = a[3] < 0.0F ? b[3] : c[3];
611 store_vector4(inst, machine, result);
612 if (DEBUG_PROG) {
613 printf("CMP (%g %g %g %g) = (%g %g %g %g) < 0 ? (%g %g %g %g) : (%g %g %g %g)\n",
614 result[0], result[1], result[2], result[3],
615 a[0], a[1], a[2], a[3],
616 b[0], b[1], b[2], b[3],
617 c[0], c[1], c[2], c[3]);
618 }
619 }
620 break;
621 case OPCODE_COS:
622 {
623 GLfloat a[4], result[4];
624 fetch_vector1(&inst->SrcReg[0], machine, a);
625 result[0] = result[1] = result[2] = result[3]
626 = (GLfloat) cos(a[0]);
627 store_vector4(inst, machine, result);
628 }
629 break;
630 case OPCODE_DDX: /* Partial derivative with respect to X */
631 {
632 GLfloat result[4];
633 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
634 'X', result);
635 store_vector4(inst, machine, result);
636 }
637 break;
638 case OPCODE_DDY: /* Partial derivative with respect to Y */
639 {
640 GLfloat result[4];
641 fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
642 'Y', result);
643 store_vector4(inst, machine, result);
644 }
645 break;
646 case OPCODE_DP2:
647 {
648 GLfloat a[4], b[4], result[4];
649 fetch_vector4(&inst->SrcReg[0], machine, a);
650 fetch_vector4(&inst->SrcReg[1], machine, b);
651 result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
652 store_vector4(inst, machine, result);
653 if (DEBUG_PROG) {
654 printf("DP2 %g = (%g %g) . (%g %g)\n",
655 result[0], a[0], a[1], b[0], b[1]);
656 }
657 }
658 break;
659 case OPCODE_DP3:
660 {
661 GLfloat a[4], b[4], result[4];
662 fetch_vector4(&inst->SrcReg[0], machine, a);
663 fetch_vector4(&inst->SrcReg[1], machine, b);
664 result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
665 store_vector4(inst, machine, result);
666 if (DEBUG_PROG) {
667 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
668 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
669 }
670 }
671 break;
672 case OPCODE_DP4:
673 {
674 GLfloat a[4], b[4], result[4];
675 fetch_vector4(&inst->SrcReg[0], machine, a);
676 fetch_vector4(&inst->SrcReg[1], machine, b);
677 result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
678 store_vector4(inst, machine, result);
679 if (DEBUG_PROG) {
680 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
681 result[0], a[0], a[1], a[2], a[3],
682 b[0], b[1], b[2], b[3]);
683 }
684 }
685 break;
686 case OPCODE_DPH:
687 {
688 GLfloat a[4], b[4], result[4];
689 fetch_vector4(&inst->SrcReg[0], machine, a);
690 fetch_vector4(&inst->SrcReg[1], machine, b);
691 result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
692 store_vector4(inst, machine, result);
693 }
694 break;
695 case OPCODE_DST: /* Distance vector */
696 {
697 GLfloat a[4], b[4], result[4];
698 fetch_vector4(&inst->SrcReg[0], machine, a);
699 fetch_vector4(&inst->SrcReg[1], machine, b);
700 result[0] = 1.0F;
701 result[1] = a[1] * b[1];
702 result[2] = a[2];
703 result[3] = b[3];
704 store_vector4(inst, machine, result);
705 }
706 break;
707 case OPCODE_EXP:
708 {
709 GLfloat t[4], q[4], floor_t0;
710 fetch_vector1(&inst->SrcReg[0], machine, t);
711 floor_t0 = FLOORF(t[0]);
712 if (floor_t0 > FLT_MAX_EXP) {
713 SET_POS_INFINITY(q[0]);
714 SET_POS_INFINITY(q[2]);
715 }
716 else if (floor_t0 < FLT_MIN_EXP) {
717 q[0] = 0.0F;
718 q[2] = 0.0F;
719 }
720 else {
721 q[0] = LDEXPF(1.0, (int) floor_t0);
722 /* Note: GL_NV_vertex_program expects
723 * result.z = result.x * APPX(result.y)
724 * We do what the ARB extension says.
725 */
726 q[2] = (GLfloat) pow(2.0, t[0]);
727 }
728 q[1] = t[0] - floor_t0;
729 q[3] = 1.0F;
730 store_vector4( inst, machine, q );
731 }
732 break;
733 case OPCODE_EX2: /* Exponential base 2 */
734 {
735 GLfloat a[4], result[4], val;
736 fetch_vector1(&inst->SrcReg[0], machine, a);
737 val = (GLfloat) pow(2.0, a[0]);
738 /*
739 if (IS_INF_OR_NAN(val))
740 val = 1.0e10;
741 */
742 result[0] = result[1] = result[2] = result[3] = val;
743 store_vector4(inst, machine, result);
744 }
745 break;
746 case OPCODE_FLR:
747 {
748 GLfloat a[4], result[4];
749 fetch_vector4(&inst->SrcReg[0], machine, a);
750 result[0] = FLOORF(a[0]);
751 result[1] = FLOORF(a[1]);
752 result[2] = FLOORF(a[2]);
753 result[3] = FLOORF(a[3]);
754 store_vector4(inst, machine, result);
755 }
756 break;
757 case OPCODE_FRC:
758 {
759 GLfloat a[4], result[4];
760 fetch_vector4(&inst->SrcReg[0], machine, a);
761 result[0] = a[0] - FLOORF(a[0]);
762 result[1] = a[1] - FLOORF(a[1]);
763 result[2] = a[2] - FLOORF(a[2]);
764 result[3] = a[3] - FLOORF(a[3]);
765 store_vector4(inst, machine, result);
766 }
767 break;
768 case OPCODE_IF:
769 {
770 GLboolean cond;
771 ASSERT(program->Instructions[inst->BranchTarget].Opcode
772 == OPCODE_ELSE ||
773 program->Instructions[inst->BranchTarget].Opcode
774 == OPCODE_ENDIF);
775 /* eval condition */
776 if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
777 GLfloat a[4];
778 fetch_vector1(&inst->SrcReg[0], machine, a);
779 cond = (a[0] != 0.0);
780 }
781 else {
782 cond = eval_condition(machine, inst);
783 }
784 if (DEBUG_PROG) {
785 printf("IF: %d\n", cond);
786 }
787 /* do if/else */
788 if (cond) {
789 /* do if-clause (just continue execution) */
790 }
791 else {
792 /* go to the instruction after ELSE or ENDIF */
793 assert(inst->BranchTarget >= 0);
794 pc = inst->BranchTarget;
795 }
796 }
797 break;
798 case OPCODE_ELSE:
799 /* goto ENDIF */
800 ASSERT(program->Instructions[inst->BranchTarget].Opcode
801 == OPCODE_ENDIF);
802 assert(inst->BranchTarget >= 0);
803 pc = inst->BranchTarget;
804 break;
805 case OPCODE_ENDIF:
806 /* nothing */
807 break;
808 case OPCODE_KIL_NV: /* NV_f_p only (conditional) */
809 if (eval_condition(machine, inst)) {
810 return GL_FALSE;
811 }
812 break;
813 case OPCODE_KIL: /* ARB_f_p only */
814 {
815 GLfloat a[4];
816 fetch_vector4(&inst->SrcReg[0], machine, a);
817 if (DEBUG_PROG) {
818 printf("KIL if (%g %g %g %g) <= 0.0\n",
819 a[0], a[1], a[2], a[3]);
820 }
821
822 if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
823 return GL_FALSE;
824 }
825 }
826 break;
827 case OPCODE_LG2: /* log base 2 */
828 {
829 GLfloat a[4], result[4], val;
830 fetch_vector1(&inst->SrcReg[0], machine, a);
831 /* The fast LOG2 macro doesn't meet the precision requirements.
832 */
833 if (a[0] == 0.0F) {
834 val = -FLT_MAX;
835 }
836 else {
837 val = (float)(log(a[0]) * 1.442695F);
838 }
839 result[0] = result[1] = result[2] = result[3] = val;
840 store_vector4(inst, machine, result);
841 }
842 break;
843 case OPCODE_LIT:
844 {
845 const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
846 GLfloat a[4], result[4];
847 fetch_vector4(&inst->SrcReg[0], machine, a);
848 a[0] = MAX2(a[0], 0.0F);
849 a[1] = MAX2(a[1], 0.0F);
850 /* XXX ARB version clamps a[3], NV version doesn't */
851 a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
852 result[0] = 1.0F;
853 result[1] = a[0];
854 /* XXX we could probably just use pow() here */
855 if (a[0] > 0.0F) {
856 if (a[1] == 0.0 && a[3] == 0.0)
857 result[2] = 1.0F;
858 else
859 result[2] = (GLfloat) pow(a[1], a[3]);
860 }
861 else {
862 result[2] = 0.0F;
863 }
864 result[3] = 1.0F;
865 store_vector4(inst, machine, result);
866 if (DEBUG_PROG) {
867 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
868 result[0], result[1], result[2], result[3],
869 a[0], a[1], a[2], a[3]);
870 }
871 }
872 break;
873 case OPCODE_LOG:
874 {
875 GLfloat t[4], q[4], abs_t0;
876 fetch_vector1(&inst->SrcReg[0], machine, t);
877 abs_t0 = FABSF(t[0]);
878 if (abs_t0 != 0.0F) {
879 if (IS_INF_OR_NAN(abs_t0))
880 {
881 SET_POS_INFINITY(q[0]);
882 q[1] = 1.0F;
883 SET_POS_INFINITY(q[2]);
884 }
885 else {
886 int exponent;
887 GLfloat mantissa = FREXPF(t[0], &exponent);
888 q[0] = (GLfloat) (exponent - 1);
889 q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
890
891 /* The fast LOG2 macro doesn't meet the precision
892 * requirements.
893 */
894 q[2] = (float)(log(t[0]) * 1.442695F);
895 }
896 }
897 else {
898 SET_NEG_INFINITY(q[0]);
899 q[1] = 1.0F;
900 SET_NEG_INFINITY(q[2]);
901 }
902 q[3] = 1.0;
903 store_vector4(inst, machine, q);
904 }
905 break;
906 case OPCODE_LRP:
907 {
908 GLfloat a[4], b[4], c[4], result[4];
909 fetch_vector4(&inst->SrcReg[0], machine, a);
910 fetch_vector4(&inst->SrcReg[1], machine, b);
911 fetch_vector4(&inst->SrcReg[2], machine, c);
912 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
913 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
914 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
915 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
916 store_vector4(inst, machine, result);
917 if (DEBUG_PROG) {
918 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
919 "(%g %g %g %g), (%g %g %g %g)\n",
920 result[0], result[1], result[2], result[3],
921 a[0], a[1], a[2], a[3],
922 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
923 }
924 }
925 break;
926 case OPCODE_MAD:
927 {
928 GLfloat a[4], b[4], c[4], result[4];
929 fetch_vector4(&inst->SrcReg[0], machine, a);
930 fetch_vector4(&inst->SrcReg[1], machine, b);
931 fetch_vector4(&inst->SrcReg[2], machine, c);
932 result[0] = a[0] * b[0] + c[0];
933 result[1] = a[1] * b[1] + c[1];
934 result[2] = a[2] * b[2] + c[2];
935 result[3] = a[3] * b[3] + c[3];
936 store_vector4(inst, machine, result);
937 if (DEBUG_PROG) {
938 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
939 "(%g %g %g %g) + (%g %g %g %g)\n",
940 result[0], result[1], result[2], result[3],
941 a[0], a[1], a[2], a[3],
942 b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
943 }
944 }
945 break;
946 case OPCODE_MAX:
947 {
948 GLfloat a[4], b[4], result[4];
949 fetch_vector4(&inst->SrcReg[0], machine, a);
950 fetch_vector4(&inst->SrcReg[1], machine, b);
951 result[0] = MAX2(a[0], b[0]);
952 result[1] = MAX2(a[1], b[1]);
953 result[2] = MAX2(a[2], b[2]);
954 result[3] = MAX2(a[3], b[3]);
955 store_vector4(inst, machine, result);
956 if (DEBUG_PROG) {
957 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
958 result[0], result[1], result[2], result[3],
959 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
960 }
961 }
962 break;
963 case OPCODE_MIN:
964 {
965 GLfloat a[4], b[4], result[4];
966 fetch_vector4(&inst->SrcReg[0], machine, a);
967 fetch_vector4(&inst->SrcReg[1], machine, b);
968 result[0] = MIN2(a[0], b[0]);
969 result[1] = MIN2(a[1], b[1]);
970 result[2] = MIN2(a[2], b[2]);
971 result[3] = MIN2(a[3], b[3]);
972 store_vector4(inst, machine, result);
973 }
974 break;
975 case OPCODE_MOV:
976 {
977 GLfloat result[4];
978 fetch_vector4(&inst->SrcReg[0], machine, result);
979 store_vector4(inst, machine, result);
980 if (DEBUG_PROG) {
981 printf("MOV (%g %g %g %g)\n",
982 result[0], result[1], result[2], result[3]);
983 }
984 }
985 break;
986 case OPCODE_MUL:
987 {
988 GLfloat a[4], b[4], result[4];
989 fetch_vector4(&inst->SrcReg[0], machine, a);
990 fetch_vector4(&inst->SrcReg[1], machine, b);
991 result[0] = a[0] * b[0];
992 result[1] = a[1] * b[1];
993 result[2] = a[2] * b[2];
994 result[3] = a[3] * b[3];
995 store_vector4(inst, machine, result);
996 if (DEBUG_PROG) {
997 printf("MUL (%g %g %g %g) = (%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], b[0], b[1], b[2], b[3]);
1000 }
1001 }
1002 break;
1003 case OPCODE_NOISE1:
1004 {
1005 GLfloat a[4], result[4];
1006 fetch_vector1(&inst->SrcReg[0], machine, a);
1007 result[0] =
1008 result[1] =
1009 result[2] =
1010 result[3] = _mesa_noise1(a[0]);
1011 store_vector4(inst, machine, result);
1012 }
1013 break;
1014 case OPCODE_NOISE2:
1015 {
1016 GLfloat a[4], result[4];
1017 fetch_vector4(&inst->SrcReg[0], machine, a);
1018 result[0] =
1019 result[1] =
1020 result[2] = result[3] = _mesa_noise2(a[0], a[1]);
1021 store_vector4(inst, machine, result);
1022 }
1023 break;
1024 case OPCODE_NOISE3:
1025 {
1026 GLfloat a[4], result[4];
1027 fetch_vector4(&inst->SrcReg[0], machine, a);
1028 result[0] =
1029 result[1] =
1030 result[2] =
1031 result[3] = _mesa_noise3(a[0], a[1], a[2]);
1032 store_vector4(inst, machine, result);
1033 }
1034 break;
1035 case OPCODE_NOISE4:
1036 {
1037 GLfloat a[4], result[4];
1038 fetch_vector4(&inst->SrcReg[0], machine, a);
1039 result[0] =
1040 result[1] =
1041 result[2] =
1042 result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
1043 store_vector4(inst, machine, result);
1044 }
1045 break;
1046 case OPCODE_NOP:
1047 break;
1048 case OPCODE_POW:
1049 {
1050 GLfloat a[4], b[4], result[4];
1051 fetch_vector1(&inst->SrcReg[0], machine, a);
1052 fetch_vector1(&inst->SrcReg[1], machine, b);
1053 result[0] = result[1] = result[2] = result[3]
1054 = (GLfloat) pow(a[0], b[0]);
1055 store_vector4(inst, machine, result);
1056 }
1057 break;
1058
1059 case OPCODE_RCP:
1060 {
1061 GLfloat a[4], result[4];
1062 fetch_vector1(&inst->SrcReg[0], machine, a);
1063 if (DEBUG_PROG) {
1064 if (a[0] == 0)
1065 printf("RCP(0)\n");
1066 else if (IS_INF_OR_NAN(a[0]))
1067 printf("RCP(inf)\n");
1068 }
1069 result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
1070 store_vector4(inst, machine, result);
1071 }
1072 break;
1073 case OPCODE_RET: /* return from subroutine (conditional) */
1074 if (eval_condition(machine, inst)) {
1075 if (machine->StackDepth == 0) {
1076 return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
1077 }
1078 /* subtract one because of pc++ in the for loop */
1079 pc = machine->CallStack[--machine->StackDepth] - 1;
1080 }
1081 break;
1082 case OPCODE_RSQ: /* 1 / sqrt() */
1083 {
1084 GLfloat a[4], result[4];
1085 fetch_vector1(&inst->SrcReg[0], machine, a);
1086 a[0] = FABSF(a[0]);
1087 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
1088 store_vector4(inst, machine, result);
1089 if (DEBUG_PROG) {
1090 printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
1091 }
1092 }
1093 break;
1094 case OPCODE_SCS: /* sine and cos */
1095 {
1096 GLfloat a[4], result[4];
1097 fetch_vector1(&inst->SrcReg[0], machine, a);
1098 result[0] = (GLfloat) cos(a[0]);
1099 result[1] = (GLfloat) sin(a[0]);
1100 result[2] = 0.0; /* undefined! */
1101 result[3] = 0.0; /* undefined! */
1102 store_vector4(inst, machine, result);
1103 }
1104 break;
1105 case OPCODE_SEQ: /* set on equal */
1106 {
1107 GLfloat a[4], b[4], result[4];
1108 fetch_vector4(&inst->SrcReg[0], machine, a);
1109 fetch_vector4(&inst->SrcReg[1], machine, b);
1110 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1111 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1112 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1113 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1114 store_vector4(inst, machine, result);
1115 if (DEBUG_PROG) {
1116 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1117 result[0], result[1], result[2], result[3],
1118 a[0], a[1], a[2], a[3],
1119 b[0], b[1], b[2], b[3]);
1120 }
1121 }
1122 break;
1123 case OPCODE_SGE: /* set on greater or equal */
1124 {
1125 GLfloat a[4], b[4], result[4];
1126 fetch_vector4(&inst->SrcReg[0], machine, a);
1127 fetch_vector4(&inst->SrcReg[1], machine, b);
1128 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1129 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1130 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1131 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1132 store_vector4(inst, machine, result);
1133 if (DEBUG_PROG) {
1134 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1135 result[0], result[1], result[2], result[3],
1136 a[0], a[1], a[2], a[3],
1137 b[0], b[1], b[2], b[3]);
1138 }
1139 }
1140 break;
1141 case OPCODE_SGT: /* set on greater */
1142 {
1143 GLfloat a[4], b[4], result[4];
1144 fetch_vector4(&inst->SrcReg[0], machine, a);
1145 fetch_vector4(&inst->SrcReg[1], machine, b);
1146 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1147 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1148 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1149 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1150 store_vector4(inst, machine, result);
1151 if (DEBUG_PROG) {
1152 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1153 result[0], result[1], result[2], result[3],
1154 a[0], a[1], a[2], a[3],
1155 b[0], b[1], b[2], b[3]);
1156 }
1157 }
1158 break;
1159 case OPCODE_SIN:
1160 {
1161 GLfloat a[4], result[4];
1162 fetch_vector1(&inst->SrcReg[0], machine, a);
1163 result[0] = result[1] = result[2] = result[3]
1164 = (GLfloat) sin(a[0]);
1165 store_vector4(inst, machine, result);
1166 }
1167 break;
1168 case OPCODE_SLE: /* set on less or equal */
1169 {
1170 GLfloat a[4], b[4], result[4];
1171 fetch_vector4(&inst->SrcReg[0], machine, a);
1172 fetch_vector4(&inst->SrcReg[1], machine, b);
1173 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1174 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1175 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1176 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1177 store_vector4(inst, machine, result);
1178 if (DEBUG_PROG) {
1179 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1180 result[0], result[1], result[2], result[3],
1181 a[0], a[1], a[2], a[3],
1182 b[0], b[1], b[2], b[3]);
1183 }
1184 }
1185 break;
1186 case OPCODE_SLT: /* set on less */
1187 {
1188 GLfloat a[4], b[4], result[4];
1189 fetch_vector4(&inst->SrcReg[0], machine, a);
1190 fetch_vector4(&inst->SrcReg[1], machine, b);
1191 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1192 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1193 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1194 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1195 store_vector4(inst, machine, result);
1196 if (DEBUG_PROG) {
1197 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1198 result[0], result[1], result[2], result[3],
1199 a[0], a[1], a[2], a[3],
1200 b[0], b[1], b[2], b[3]);
1201 }
1202 }
1203 break;
1204 case OPCODE_SNE: /* set on not equal */
1205 {
1206 GLfloat a[4], b[4], result[4];
1207 fetch_vector4(&inst->SrcReg[0], machine, a);
1208 fetch_vector4(&inst->SrcReg[1], machine, b);
1209 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1210 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1211 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1212 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1213 store_vector4(inst, machine, result);
1214 if (DEBUG_PROG) {
1215 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1216 result[0], result[1], result[2], result[3],
1217 a[0], a[1], a[2], a[3],
1218 b[0], b[1], b[2], b[3]);
1219 }
1220 }
1221 break;
1222 case OPCODE_SSG: /* set sign (-1, 0 or +1) */
1223 {
1224 GLfloat a[4], result[4];
1225 fetch_vector4(&inst->SrcReg[0], machine, a);
1226 result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
1227 result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
1228 result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
1229 result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
1230 store_vector4(inst, machine, result);
1231 }
1232 break;
1233 case OPCODE_SUB:
1234 {
1235 GLfloat a[4], b[4], result[4];
1236 fetch_vector4(&inst->SrcReg[0], machine, a);
1237 fetch_vector4(&inst->SrcReg[1], machine, b);
1238 result[0] = a[0] - b[0];
1239 result[1] = a[1] - b[1];
1240 result[2] = a[2] - b[2];
1241 result[3] = a[3] - b[3];
1242 store_vector4(inst, machine, result);
1243 if (DEBUG_PROG) {
1244 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1245 result[0], result[1], result[2], result[3],
1246 a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
1247 }
1248 }
1249 break;
1250 case OPCODE_SWZ: /* extended swizzle */
1251 {
1252 const struct prog_src_register *source = &inst->SrcReg[0];
1253 const GLfloat *src = get_src_register_pointer(source, machine);
1254 GLfloat result[4];
1255 GLuint i;
1256 for (i = 0; i < 4; i++) {
1257 const GLuint swz = GET_SWZ(source->Swizzle, i);
1258 if (swz == SWIZZLE_ZERO)
1259 result[i] = 0.0;
1260 else if (swz == SWIZZLE_ONE)
1261 result[i] = 1.0;
1262 else {
1263 ASSERT(swz <= 3);
1264 result[i] = src[swz];
1265 }
1266 if (source->Negate & (1 << i))
1267 result[i] = -result[i];
1268 }
1269 store_vector4(inst, machine, result);
1270 }
1271 break;
1272 case OPCODE_TEX: /* Both ARB and NV frag prog */
1273 /* Simple texel lookup */
1274 {
1275 GLfloat texcoord[4], color[4];
1276 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1277
1278 /* For TEX, texcoord.Q should not be used and its value should not
1279 * matter (at most, we pass coord.xyz to texture3D() in GLSL).
1280 * Set Q=1 so that FetchTexelDeriv() doesn't get a garbage value
1281 * which is effectively what happens when the texcoord swizzle
1282 * is .xyzz
1283 */
1284 texcoord[3] = 1.0f;
1285
1286 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1287
1288 if (DEBUG_PROG) {
1289 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1290 color[0], color[1], color[2], color[3],
1291 inst->TexSrcUnit,
1292 texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
1293 }
1294 store_vector4(inst, machine, color);
1295 }
1296 break;
1297 case OPCODE_TXB: /* GL_ARB_fragment_program only */
1298 /* Texel lookup with LOD bias */
1299 {
1300 GLfloat texcoord[4], color[4], lodBias;
1301
1302 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1303
1304 /* texcoord[3] is the bias to add to lambda */
1305 lodBias = texcoord[3];
1306
1307 fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
1308
1309 if (DEBUG_PROG) {
1310 printf("TXB (%g, %g, %g, %g) = texture[%d][%g %g %g %g]"
1311 " bias %g\n",
1312 color[0], color[1], color[2], color[3],
1313 inst->TexSrcUnit,
1314 texcoord[0],
1315 texcoord[1],
1316 texcoord[2],
1317 texcoord[3],
1318 lodBias);
1319 }
1320
1321 store_vector4(inst, machine, color);
1322 }
1323 break;
1324 case OPCODE_TXD: /* GL_NV_fragment_program only */
1325 /* Texture lookup w/ partial derivatives for LOD */
1326 {
1327 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1328 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1329 fetch_vector4(&inst->SrcReg[1], machine, dtdx);
1330 fetch_vector4(&inst->SrcReg[2], machine, dtdy);
1331 machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
1332 0.0, /* lodBias */
1333 inst->TexSrcUnit, color);
1334 store_vector4(inst, machine, color);
1335 }
1336 break;
1337 case OPCODE_TXL:
1338 /* Texel lookup with explicit LOD */
1339 {
1340 GLfloat texcoord[4], color[4], lod;
1341
1342 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1343
1344 /* texcoord[3] is the LOD */
1345 lod = texcoord[3];
1346
1347 machine->FetchTexelLod(ctx, texcoord, lod,
1348 machine->Samplers[inst->TexSrcUnit], color);
1349
1350 store_vector4(inst, machine, color);
1351 }
1352 break;
1353 case OPCODE_TXP: /* GL_ARB_fragment_program only */
1354 /* Texture lookup w/ projective divide */
1355 {
1356 GLfloat texcoord[4], color[4];
1357
1358 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1359 /* Not so sure about this test - if texcoord[3] is
1360 * zero, we'd probably be fine except for an ASSERT in
1361 * IROUND_POS() which gets triggered by the inf values created.
1362 */
1363 if (texcoord[3] != 0.0) {
1364 texcoord[0] /= texcoord[3];
1365 texcoord[1] /= texcoord[3];
1366 texcoord[2] /= texcoord[3];
1367 }
1368
1369 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1370
1371 store_vector4(inst, machine, color);
1372 }
1373 break;
1374 case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
1375 /* Texture lookup w/ projective divide, as above, but do not
1376 * do the divide by w if sampling from a cube map.
1377 */
1378 {
1379 GLfloat texcoord[4], color[4];
1380
1381 fetch_vector4(&inst->SrcReg[0], machine, texcoord);
1382 if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
1383 texcoord[3] != 0.0) {
1384 texcoord[0] /= texcoord[3];
1385 texcoord[1] /= texcoord[3];
1386 texcoord[2] /= texcoord[3];
1387 }
1388
1389 fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
1390
1391 store_vector4(inst, machine, color);
1392 }
1393 break;
1394 case OPCODE_TRUNC: /* truncate toward zero */
1395 {
1396 GLfloat a[4], result[4];
1397 fetch_vector4(&inst->SrcReg[0], machine, a);
1398 result[0] = (GLfloat) (GLint) a[0];
1399 result[1] = (GLfloat) (GLint) a[1];
1400 result[2] = (GLfloat) (GLint) a[2];
1401 result[3] = (GLfloat) (GLint) a[3];
1402 store_vector4(inst, machine, result);
1403 }
1404 break;
1405 case OPCODE_XPD: /* cross product */
1406 {
1407 GLfloat a[4], b[4], result[4];
1408 fetch_vector4(&inst->SrcReg[0], machine, a);
1409 fetch_vector4(&inst->SrcReg[1], machine, b);
1410 result[0] = a[1] * b[2] - a[2] * b[1];
1411 result[1] = a[2] * b[0] - a[0] * b[2];
1412 result[2] = a[0] * b[1] - a[1] * b[0];
1413 result[3] = 1.0;
1414 store_vector4(inst, machine, result);
1415 if (DEBUG_PROG) {
1416 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1417 result[0], result[1], result[2], result[3],
1418 a[0], a[1], a[2], b[0], b[1], b[2]);
1419 }
1420 }
1421 break;
1422 case OPCODE_END:
1423 return GL_TRUE;
1424 default:
1425 _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
1426 inst->Opcode);
1427 return GL_TRUE; /* return value doesn't matter */
1428 }
1429
1430 numExec++;
1431 if (numExec > maxExec) {
1432 static GLboolean reported = GL_FALSE;
1433 if (!reported) {
1434 _mesa_problem(ctx, "Infinite loop detected in fragment program");
1435 reported = GL_TRUE;
1436 }
1437 return GL_TRUE;
1438 }
1439
1440 } /* for pc */
1441
1442 return GL_TRUE;
1443 }