projects / mesa.git / blob
? search:


c230ff38de82ad55029948979c4a106ac2e6dee8
[mesa.git] / src / mesa / swrast / s_nvfragprog.c
1 /*
2 * Mesa 3-D graphics library
3 * Version: 5.1
4 *
5 * Copyright (C) 1999-2003 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 #include "glheader.h"
27 #include "colormac.h"
28 #include "context.h"
29 #include "nvfragprog.h"
30 #include "macros.h"
31 #include "program.h"
32
33 #include "s_nvfragprog.h"
34 #include "s_span.h"
35 #include "s_texture.h"
36
37
38 /* if 1, print some debugging info */
39 #define DEBUG_FRAG 0
40
41
42 /**
43 * Fetch a texel.
44 */
45 static void
46 fetch_texel( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda,
47 GLuint unit, GLfloat color[4] )
48 {
49 GLchan rgba[4];
50 SWcontext *swrast = SWRAST_CONTEXT(ctx);
51
52 swrast->TextureSample[unit](ctx, unit, ctx->Texture.Unit[unit]._Current,
53 1, (const GLfloat (*)[4]) texcoord,
54 &lambda, &rgba);
55 color[0] = CHAN_TO_FLOAT(rgba[0]);
56 color[1] = CHAN_TO_FLOAT(rgba[1]);
57 color[2] = CHAN_TO_FLOAT(rgba[2]);
58 color[3] = CHAN_TO_FLOAT(rgba[3]);
59 }
60
61
62 /**
63 * Fetch a texel with the given partial derivatives to compute a level
64 * of detail in the mipmap.
65 */
66 static void
67 fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4],
68 const GLfloat texdx[4], const GLfloat texdy[4],
69 GLuint unit, GLfloat color[4] )
70 {
71 SWcontext *swrast = SWRAST_CONTEXT(ctx);
72 const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
73 const struct gl_texture_image *texImg = texObj->Image[texObj->BaseLevel];
74 const GLfloat texW = (GLfloat) texImg->WidthScale;
75 const GLfloat texH = (GLfloat) texImg->HeightScale;
76 GLchan rgba[4];
77
78 GLfloat lambda = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */
79 texdx[1], texdy[1], /* dt/dx, dt/dy */
80 texdx[3], texdy[2], /* dq/dx, dq/dy */
81 texW, texH,
82 texcoord[0], texcoord[1], texcoord[3],
83 1.0F / texcoord[3]);
84
85 swrast->TextureSample[unit](ctx, unit, ctx->Texture.Unit[unit]._Current,
86 1, (const GLfloat (*)[4]) texcoord,
87 &lambda, &rgba);
88 color[0] = CHAN_TO_FLOAT(rgba[0]);
89 color[1] = CHAN_TO_FLOAT(rgba[1]);
90 color[2] = CHAN_TO_FLOAT(rgba[2]);
91 color[3] = CHAN_TO_FLOAT(rgba[3]);
92 }
93
94
95
96 /**
97 * Fetch a 4-element float vector from the given source register.
98 * Apply swizzling and negating as needed.
99 */
100 static void
101 fetch_vector4( GLcontext *ctx,
102 const struct fp_src_register *source,
103 struct fp_machine *machine,
104 const struct fragment_program *program,
105 GLfloat result[4] )
106 {
107 const GLfloat *src;
108
109 switch (source->File) {
110 case PROGRAM_TEMPORARY:
111 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_TEMPS);
112 src = machine->Temporaries[source->Index];
113 break;
114 case PROGRAM_INPUT:
115 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
116 src = machine->Inputs[source->Index];
117 break;
118 case PROGRAM_LOCAL_PARAM:
119 ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
120 src = program->Base.LocalParams[source->Index];
121 break;
122 case PROGRAM_ENV_PARAM:
123 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_PARAMS);
124 src = ctx->FragmentProgram.Parameters[source->Index];
125 break;
126 case PROGRAM_NAMED_PARAM:
127 ASSERT(source->Index < program->Parameters->NumParameters);
128 src = program->Parameters->Parameters[source->Index].Values;
129 break;
130 case PROGRAM_STATE_VAR:
131 abort();
132 default:
133 _mesa_problem(ctx, "Invalid input register file in fetch_vector4");
134 return;
135 }
136
137 result[0] = src[source->Swizzle[0]];
138 result[1] = src[source->Swizzle[1]];
139 result[2] = src[source->Swizzle[2]];
140 result[3] = src[source->Swizzle[3]];
141
142 if (source->NegateBase) {
143 result[0] = -result[0];
144 result[1] = -result[1];
145 result[2] = -result[2];
146 result[3] = -result[3];
147 }
148 if (source->Abs) {
149 result[0] = FABSF(result[0]);
150 result[1] = FABSF(result[1]);
151 result[2] = FABSF(result[2]);
152 result[3] = FABSF(result[3]);
153 }
154 if (source->NegateAbs) {
155 result[0] = -result[0];
156 result[1] = -result[1];
157 result[2] = -result[2];
158 result[3] = -result[3];
159 }
160 }
161
162
163 /**
164 * Fetch the derivative with respect to X for the given register.
165 * \return GL_TRUE if it was easily computed or GL_FALSE if we
166 * need to execute another instance of the program (ugh)!
167 */
168 static GLboolean
169 fetch_vector4_deriv( const struct fp_src_register *source,
170 const struct sw_span *span,
171 char xOrY, GLfloat result[4] )
172 {
173 GLfloat src[4];
174
175 ASSERT(xOrY == 'X' || xOrY == 'Y');
176
177 assert(source->File == PROGRAM_INPUT);
178
179 switch (source->Index) {
180 case FRAG_ATTRIB_WPOS:
181 if (xOrY == 'X') {
182 src[0] = 1.0;
183 src[1] = 0.0;
184 src[2] = span->dzdx;
185 src[3] = span->dwdx;
186 }
187 else {
188 src[0] = 0.0;
189 src[1] = 1.0;
190 src[2] = span->dzdy;
191 src[3] = span->dwdy;
192 }
193 break;
194 case FRAG_ATTRIB_COL0:
195 if (xOrY == 'X') {
196 src[0] = span->drdx * (1.0F / CHAN_MAXF);
197 src[1] = span->dgdx * (1.0F / CHAN_MAXF);
198 src[2] = span->dbdx * (1.0F / CHAN_MAXF);
199 src[3] = span->dadx * (1.0F / CHAN_MAXF);
200 }
201 else {
202 src[0] = span->drdy * (1.0F / CHAN_MAXF);
203 src[1] = span->dgdy * (1.0F / CHAN_MAXF);
204 src[2] = span->dbdy * (1.0F / CHAN_MAXF);
205 src[3] = span->dady * (1.0F / CHAN_MAXF);
206 }
207 break;
208 case FRAG_ATTRIB_COL1:
209 if (xOrY == 'X') {
210 src[0] = span->dsrdx * (1.0F / CHAN_MAXF);
211 src[1] = span->dsgdx * (1.0F / CHAN_MAXF);
212 src[2] = span->dsbdx * (1.0F / CHAN_MAXF);
213 src[3] = 0.0; /* XXX need this */
214 }
215 else {
216 src[0] = span->dsrdy * (1.0F / CHAN_MAXF);
217 src[1] = span->dsgdy * (1.0F / CHAN_MAXF);
218 src[2] = span->dsbdy * (1.0F / CHAN_MAXF);
219 src[3] = 0.0; /* XXX need this */
220 }
221 break;
222 case FRAG_ATTRIB_FOGC:
223 if (xOrY == 'X') {
224 src[0] = span->dfogdx;
225 src[1] = 0.0;
226 src[2] = 0.0;
227 src[3] = 0.0;
228 }
229 else {
230 src[0] = span->dfogdy;
231 src[1] = 0.0;
232 src[2] = 0.0;
233 src[3] = 0.0;
234 }
235 break;
236 case FRAG_ATTRIB_TEX0:
237 case FRAG_ATTRIB_TEX1:
238 case FRAG_ATTRIB_TEX2:
239 case FRAG_ATTRIB_TEX3:
240 case FRAG_ATTRIB_TEX4:
241 case FRAG_ATTRIB_TEX5:
242 case FRAG_ATTRIB_TEX6:
243 case FRAG_ATTRIB_TEX7:
244 if (xOrY == 'X') {
245 const GLuint u = source->Index - FRAG_ATTRIB_TEX0;
246 src[0] = span->texStepX[u][0] * (1.0F / CHAN_MAXF);
247 src[1] = span->texStepX[u][1] * (1.0F / CHAN_MAXF);
248 src[2] = span->texStepX[u][2] * (1.0F / CHAN_MAXF);
249 src[3] = span->texStepX[u][3] * (1.0F / CHAN_MAXF);
250 }
251 else {
252 const GLuint u = source->Index - FRAG_ATTRIB_TEX0;
253 src[0] = span->texStepY[u][0] * (1.0F / CHAN_MAXF);
254 src[1] = span->texStepY[u][1] * (1.0F / CHAN_MAXF);
255 src[2] = span->texStepY[u][2] * (1.0F / CHAN_MAXF);
256 src[3] = span->texStepY[u][3] * (1.0F / CHAN_MAXF);
257 }
258 break;
259 default:
260 return GL_FALSE;
261 }
262
263 result[0] = src[source->Swizzle[0]];
264 result[1] = src[source->Swizzle[1]];
265 result[2] = src[source->Swizzle[2]];
266 result[3] = src[source->Swizzle[3]];
267
268 if (source->NegateBase) {
269 result[0] = -result[0];
270 result[1] = -result[1];
271 result[2] = -result[2];
272 result[3] = -result[3];
273 }
274 if (source->Abs) {
275 result[0] = FABSF(result[0]);
276 result[1] = FABSF(result[1]);
277 result[2] = FABSF(result[2]);
278 result[3] = FABSF(result[3]);
279 }
280 if (source->NegateAbs) {
281 result[0] = -result[0];
282 result[1] = -result[1];
283 result[2] = -result[2];
284 result[3] = -result[3];
285 }
286 return GL_TRUE;
287 }
288
289
290 /**
291 * As above, but only return result[0] element.
292 */
293 static void
294 fetch_vector1( GLcontext *ctx,
295 const struct fp_src_register *source,
296 const struct fp_machine *machine,
297 const struct fragment_program *program,
298 GLfloat result[4] )
299 {
300 const GLfloat *src;
301
302 switch (source->File) {
303 case PROGRAM_TEMPORARY:
304 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_TEMPS);
305 src = machine->Temporaries[source->Index];
306 break;
307 case PROGRAM_INPUT:
308 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
309 src = machine->Inputs[source->Index];
310 break;
311 case PROGRAM_LOCAL_PARAM:
312 ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
313 src = program->Base.LocalParams[source->Index];
314 break;
315 case PROGRAM_ENV_PARAM:
316 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_PARAMS);
317 src = ctx->FragmentProgram.Parameters[source->Index];
318 break;
319 case PROGRAM_NAMED_PARAM:
320 ASSERT(source->Index < program->Parameters->NumParameters);
321 src = program->Parameters->Parameters[source->Index].Values;
322 break;
323 case PROGRAM_STATE_VAR:
324 abort();
325 default:
326 _mesa_problem(ctx, "Invalid input register file in fetch_vector1");
327 return;
328 }
329
330 result[0] = src[source->Swizzle[0]];
331
332 if (source->NegateBase) {
333 result[0] = -result[0];
334 }
335 if (source->Abs) {
336 result[0] = FABSF(result[0]);
337 }
338 if (source->NegateAbs) {
339 result[0] = -result[0];
340 }
341 }
342
343
344 /*
345 * Test value against zero and return GT, LT, EQ or UN if NaN.
346 */
347 static INLINE GLuint
348 generate_cc( float value )
349 {
350 if (value != value)
351 return COND_UN; /* NaN */
352 if (value > 0.0F)
353 return COND_GT;
354 if (value < 0.0F)
355 return COND_LT;
356 return COND_EQ;
357 }
358
359 /*
360 * Test if the ccMaskRule is satisfied by the given condition code.
361 * Used to mask destination writes according to the current condition codee.
362 */
363 static INLINE GLboolean
364 test_cc(GLuint condCode, GLuint ccMaskRule)
365 {
366 switch (ccMaskRule) {
367 case COND_EQ: return (condCode == COND_EQ);
368 case COND_NE: return (condCode != COND_EQ);
369 case COND_LT: return (condCode == COND_LT);
370 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
371 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
372 case COND_GT: return (condCode == COND_GT);
373 case COND_TR: return GL_TRUE;
374 case COND_FL: return GL_FALSE;
375 default: return GL_TRUE;
376 }
377 }
378
379
380 /**
381 * Store 4 floats into a register. Observe the instructions saturate and
382 * set-condition-code flags.
383 */
384 static void
385 store_vector4( const struct fp_instruction *inst,
386 struct fp_machine *machine,
387 const GLfloat value[4] )
388 {
389 const struct fp_dst_register *dest = &(inst->DstReg);
390 const GLboolean clamp = inst->Saturate;
391 const GLboolean updateCC = inst->UpdateCondRegister;
392 GLfloat *dstReg;
393 GLfloat clampedValue[4];
394 const GLboolean *writeMask = dest->WriteMask;
395 GLboolean condWriteMask[4];
396
397 switch (dest->File) {
398 case PROGRAM_OUTPUT:
399 dstReg = machine->Outputs[dest->Index];
400 break;
401 case PROGRAM_TEMPORARY:
402 dstReg = machine->Temporaries[dest->Index];
403 break;
404 default:
405 _mesa_problem(NULL, "bad register file in store_vector4(fp)");
406 return;
407 }
408
409 #if DEBUG_FRAG
410 if (value[0] > 1.0e10 ||
411 IS_INF_OR_NAN(value[0]) ||
412 IS_INF_OR_NAN(value[1]) ||
413 IS_INF_OR_NAN(value[2]) ||
414 IS_INF_OR_NAN(value[3]) )
415 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
416 #endif
417
418 if (clamp) {
419 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
420 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
421 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
422 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
423 value = clampedValue;
424 }
425
426 if (dest->CondMask != COND_TR) {
427 condWriteMask[0] = writeMask[0]
428 && test_cc(machine->CondCodes[dest->CondSwizzle[0]], dest->CondMask);
429 condWriteMask[1] = writeMask[1]
430 && test_cc(machine->CondCodes[dest->CondSwizzle[1]], dest->CondMask);
431 condWriteMask[2] = writeMask[2]
432 && test_cc(machine->CondCodes[dest->CondSwizzle[2]], dest->CondMask);
433 condWriteMask[3] = writeMask[3]
434 && test_cc(machine->CondCodes[dest->CondSwizzle[3]], dest->CondMask);
435 writeMask = condWriteMask;
436 }
437
438 if (writeMask[0]) {
439 dstReg[0] = value[0];
440 if (updateCC)
441 machine->CondCodes[0] = generate_cc(value[0]);
442 }
443 if (writeMask[1]) {
444 dstReg[1] = value[1];
445 if (updateCC)
446 machine->CondCodes[1] = generate_cc(value[1]);
447 }
448 if (writeMask[2]) {
449 dstReg[2] = value[2];
450 if (updateCC)
451 machine->CondCodes[2] = generate_cc(value[2]);
452 }
453 if (writeMask[3]) {
454 dstReg[3] = value[3];
455 if (updateCC)
456 machine->CondCodes[3] = generate_cc(value[3]);
457 }
458 }
459
460
461 /**
462 * Initialize a new machine state instance from an existing one, adding
463 * the partial derivatives onto the input registers.
464 * Used to implement DDX and DDY instructions in non-trivial cases.
465 */
466 static void
467 init_machine_deriv( GLcontext *ctx,
468 const struct fp_machine *machine,
469 const struct fragment_program *program,
470 const struct sw_span *span, char xOrY,
471 struct fp_machine *dMachine )
472 {
473 GLuint u;
474
475 ASSERT(xOrY == 'X' || xOrY == 'Y');
476
477 /* copy existing machine */
478 _mesa_memcpy(dMachine, machine, sizeof(struct fp_machine));
479
480 /* Clear temporary registers */
481 _mesa_bzero( (void*) machine->Temporaries,
482 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
483
484 /* Add derivatives */
485 if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
486 GLfloat *wpos = (GLfloat*) machine->Inputs[FRAG_ATTRIB_WPOS];
487 if (xOrY == 'X') {
488 wpos[0] += 1.0F;
489 wpos[1] += 0.0F;
490 wpos[2] += span->dzdx;
491 wpos[3] += span->dwdx;
492 }
493 else {
494 wpos[0] += 0.0F;
495 wpos[1] += 1.0F;
496 wpos[2] += span->dzdy;
497 wpos[3] += span->dwdy;
498 }
499 }
500 if (program->InputsRead & (1 << FRAG_ATTRIB_COL0)) {
501 GLfloat *col0 = (GLfloat*) machine->Inputs[FRAG_ATTRIB_COL0];
502 if (xOrY == 'X') {
503 col0[0] += span->drdx * (1.0F / CHAN_MAXF);
504 col0[1] += span->dgdx * (1.0F / CHAN_MAXF);
505 col0[2] += span->dbdx * (1.0F / CHAN_MAXF);
506 col0[3] += span->dadx * (1.0F / CHAN_MAXF);
507 }
508 else {
509 col0[0] += span->drdy * (1.0F / CHAN_MAXF);
510 col0[1] += span->dgdy * (1.0F / CHAN_MAXF);
511 col0[2] += span->dbdy * (1.0F / CHAN_MAXF);
512 col0[3] += span->dady * (1.0F / CHAN_MAXF);
513 }
514 }
515 if (program->InputsRead & (1 << FRAG_ATTRIB_COL1)) {
516 GLfloat *col1 = (GLfloat*) machine->Inputs[FRAG_ATTRIB_COL1];
517 if (xOrY == 'X') {
518 col1[0] += span->dsrdx * (1.0F / CHAN_MAXF);
519 col1[1] += span->dsgdx * (1.0F / CHAN_MAXF);
520 col1[2] += span->dsbdx * (1.0F / CHAN_MAXF);
521 col1[3] += 0.0; /*XXX fix */
522 }
523 else {
524 col1[0] += span->dsrdy * (1.0F / CHAN_MAXF);
525 col1[1] += span->dsgdy * (1.0F / CHAN_MAXF);
526 col1[2] += span->dsbdy * (1.0F / CHAN_MAXF);
527 col1[3] += 0.0; /*XXX fix */
528 }
529 }
530 if (program->InputsRead & (1 << FRAG_ATTRIB_FOGC)) {
531 GLfloat *fogc = (GLfloat*) machine->Inputs[FRAG_ATTRIB_FOGC];
532 if (xOrY == 'X') {
533 fogc[0] += span->dfogdx;
534 }
535 else {
536 fogc[0] += span->dfogdy;
537 }
538 }
539 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
540 if (program->InputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
541 GLfloat *tex = (GLfloat*) machine->Inputs[FRAG_ATTRIB_TEX0 + u];
542 if (xOrY == 'X') {
543 tex[0] += span->texStepX[u][0];
544 tex[1] += span->texStepX[u][1];
545 tex[2] += span->texStepX[u][2];
546 tex[3] += span->texStepX[u][3];
547 }
548 else {
549 tex[0] += span->texStepY[u][0];
550 tex[1] += span->texStepY[u][1];
551 tex[2] += span->texStepY[u][2];
552 tex[3] += span->texStepY[u][3];
553 }
554 }
555 }
556
557 /* init condition codes */
558 dMachine->CondCodes[0] = COND_EQ;
559 dMachine->CondCodes[1] = COND_EQ;
560 dMachine->CondCodes[2] = COND_EQ;
561 dMachine->CondCodes[3] = COND_EQ;
562 }
563
564
565 /**
566 * Execute the given vertex program.
567 * NOTE: we do everything in single-precision floating point; we don't
568 * currently observe the single/half/fixed-precision qualifiers.
569 * \param ctx - rendering context
570 * \param program - the fragment program to execute
571 * \param machine - machine state (register file)
572 * \param maxInst - max number of instructions to execute
573 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
574 */
575 static GLboolean
576 execute_program( GLcontext *ctx,
577 const struct fragment_program *program, GLuint maxInst,
578 struct fp_machine *machine, const struct sw_span *span,
579 GLuint column )
580 {
581 GLuint pc;
582
583 #if DEBUG_FRAG
584 printf("execute fragment program --------------------\n");
585 #endif
586
587 for (pc = 0; pc < maxInst; pc++) {
588 const struct fp_instruction *inst = program->Instructions + pc;
589
590 if (ctx->FragmentProgram.CallbackEnabled &&
591 ctx->FragmentProgram.Callback) {
592 ctx->FragmentProgram.CurrentPosition = inst->StringPos;
593 ctx->FragmentProgram.Callback(program->Base.Target,
594 ctx->FragmentProgram.CallbackData);
595 }
596
597 switch (inst->Opcode) {
598 case FP_OPCODE_ABS:
599 {
600 GLfloat a[4], result[4];
601 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
602 result[0] = FABSF(a[0]);
603 result[1] = FABSF(a[1]);
604 result[2] = FABSF(a[2]);
605 result[3] = FABSF(a[3]);
606 store_vector4( inst, machine, result );
607 }
608 break;
609 case FP_OPCODE_ADD:
610 {
611 GLfloat a[4], b[4], result[4];
612 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
613 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
614 result[0] = a[0] + b[0];
615 result[1] = a[1] + b[1];
616 result[2] = a[2] + b[2];
617 result[3] = a[3] + b[3];
618 store_vector4( inst, machine, result );
619 }
620 break;
621 case FP_OPCODE_CMP:
622 {
623 GLfloat a[4], b[4], c[4], result[4];
624 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
625 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
626 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
627 result[0] = a[0] < 0.0F ? b[0] : c[0];
628 result[1] = a[1] < 0.0F ? b[1] : c[1];
629 result[2] = a[2] < 0.0F ? b[2] : c[2];
630 result[3] = a[3] < 0.0F ? b[3] : c[3];
631 store_vector4( inst, machine, result );
632 }
633 break;
634 case FP_OPCODE_COS:
635 {
636 GLfloat a[4], result[4];
637 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
638 result[0] = result[1] = result[2] = result[3] = (GLfloat)_mesa_cos(a[0]);
639 store_vector4( inst, machine, result );
640 }
641 break;
642 case FP_OPCODE_DDX: /* Partial derivative with respect to X */
643 {
644 GLfloat a[4], aNext[4], result[4];
645 struct fp_machine dMachine;
646 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', result)) {
647 /* This is tricky. Make a copy of the current machine state,
648 * increment the input registers by the dx or dy partial
649 * derivatives, then re-execute the program up to the
650 * preceeding instruction, then fetch the source register.
651 * Finally, find the difference in the register values for
652 * the original and derivative runs.
653 */
654 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
655 init_machine_deriv(ctx, machine, program, span,
656 'X', &dMachine);
657 execute_program(ctx, program, pc, &dMachine, span, column);
658 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
659 result[0] = aNext[0] - a[0];
660 result[1] = aNext[1] - a[1];
661 result[2] = aNext[2] - a[2];
662 result[3] = aNext[3] - a[3];
663 }
664 store_vector4( inst, machine, result );
665 }
666 break;
667 case FP_OPCODE_DDY: /* Partial derivative with respect to Y */
668 {
669 GLfloat a[4], aNext[4], result[4];
670 struct fp_machine dMachine;
671 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', result)) {
672 init_machine_deriv(ctx, machine, program, span,
673 'Y', &dMachine);
674 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
675 execute_program(ctx, program, pc, &dMachine, span, column);
676 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
677 result[0] = aNext[0] - a[0];
678 result[1] = aNext[1] - a[1];
679 result[2] = aNext[2] - a[2];
680 result[3] = aNext[3] - a[3];
681 }
682 store_vector4( inst, machine, result );
683 }
684 break;
685 case FP_OPCODE_DP3:
686 {
687 GLfloat a[4], b[4], result[4];
688 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
689 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
690 result[0] = result[1] = result[2] = result[3] =
691 a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
692 store_vector4( inst, machine, result );
693 #if DEBUG_FRAG
694 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
695 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
696 #endif
697 }
698 break;
699 case FP_OPCODE_DP4:
700 {
701 GLfloat a[4], b[4], result[4];
702 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
703 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
704 result[0] = result[1] = result[2] = result[3] =
705 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
706 store_vector4( inst, machine, result );
707 }
708 break;
709 case FP_OPCODE_DPH:
710 {
711 GLfloat a[4], b[4], result[4];
712 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
713 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
714 result[0] = result[1] = result[2] = result[3] =
715 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + b[3];
716 store_vector4( inst, machine, result );
717 }
718 break;
719 case FP_OPCODE_DST: /* Distance vector */
720 {
721 GLfloat a[4], b[4], result[4];
722 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
723 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
724 result[0] = 1.0F;
725 result[1] = a[1] * b[1];
726 result[2] = a[2];
727 result[3] = b[3];
728 store_vector4( inst, machine, result );
729 }
730 break;
731 case FP_OPCODE_EX2: /* Exponential base 2 */
732 {
733 GLfloat a[4], result[4];
734 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
735 result[0] = result[1] = result[2] = result[3] =
736 (GLfloat) _mesa_pow(2.0, a[0]);
737 store_vector4( inst, machine, result );
738 }
739 break;
740 case FP_OPCODE_FLR:
741 {
742 GLfloat a[4], result[4];
743 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
744 result[0] = FLOORF(a[0]);
745 result[1] = FLOORF(a[1]);
746 result[2] = FLOORF(a[2]);
747 result[3] = FLOORF(a[3]);
748 store_vector4( inst, machine, result );
749 }
750 break;
751 case FP_OPCODE_FRC:
752 {
753 GLfloat a[4], result[4];
754 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
755 result[0] = a[0] - FLOORF(a[0]);
756 result[1] = a[1] - FLOORF(a[1]);
757 result[2] = a[2] - FLOORF(a[2]);
758 result[3] = a[3] - FLOORF(a[3]);
759 store_vector4( inst, machine, result );
760 }
761 break;
762 case FP_OPCODE_KIL:
763 {
764 const GLuint *swizzle = inst->DstReg.CondSwizzle;
765 const GLuint condMask = inst->DstReg.CondMask;
766 if (test_cc(machine->CondCodes[swizzle[0]], condMask) ||
767 test_cc(machine->CondCodes[swizzle[1]], condMask) ||
768 test_cc(machine->CondCodes[swizzle[2]], condMask) ||
769 test_cc(machine->CondCodes[swizzle[3]], condMask)) {
770 return GL_FALSE;
771 }
772 }
773 break;
774 case FP_OPCODE_LG2: /* log base 2 */
775 {
776 GLfloat a[4], result[4];
777 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
778 result[0] = result[1] = result[2] = result[3]
779 = LOG2(a[0]);
780 store_vector4( inst, machine, result );
781 }
782 break;
783 case FP_OPCODE_LIT:
784 {
785 GLfloat a[4], result[4];
786 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
787 if (a[0] < 0.0F)
788 a[0] = 0.0F;
789 if (a[1] < 0.0F)
790 a[1] = 0.0F;
791 result[0] = 1.0F;
792 result[1] = a[0];
793 result[2] = (a[0] > 0.0F) ? (GLfloat)_mesa_pow(2.0, a[3]) : 0.0F;
794 result[3] = 1.0F;
795 store_vector4( inst, machine, result );
796 }
797 break;
798 case FP_OPCODE_LRP:
799 {
800 GLfloat a[4], b[4], c[4], result[4];
801 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
802 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
803 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
804 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
805 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
806 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
807 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
808 store_vector4( inst, machine, result );
809 }
810 break;
811 case FP_OPCODE_MAD:
812 {
813 GLfloat a[4], b[4], c[4], result[4];
814 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
815 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
816 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
817 result[0] = a[0] * b[0] + c[0];
818 result[1] = a[1] * b[1] + c[1];
819 result[2] = a[2] * b[2] + c[2];
820 result[3] = a[3] * b[3] + c[3];
821 store_vector4( inst, machine, result );
822 }
823 break;
824 case FP_OPCODE_MAX:
825 {
826 GLfloat a[4], b[4], result[4];
827 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
828 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
829 result[0] = MAX2(a[0], b[0]);
830 result[1] = MAX2(a[1], b[1]);
831 result[2] = MAX2(a[2], b[2]);
832 result[3] = MAX2(a[3], b[3]);
833 store_vector4( inst, machine, result );
834 }
835 break;
836 case FP_OPCODE_MIN:
837 {
838 GLfloat a[4], b[4], result[4];
839 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
840 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
841 result[0] = MIN2(a[0], b[0]);
842 result[1] = MIN2(a[1], b[1]);
843 result[2] = MIN2(a[2], b[2]);
844 result[3] = MIN2(a[3], b[3]);
845 store_vector4( inst, machine, result );
846 }
847 break;
848 case FP_OPCODE_MOV:
849 {
850 GLfloat result[4];
851 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, result );
852 store_vector4( inst, machine, result );
853 }
854 break;
855 case FP_OPCODE_MUL:
856 {
857 GLfloat a[4], b[4], result[4];
858 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
859 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
860 result[0] = a[0] * b[0];
861 result[1] = a[1] * b[1];
862 result[2] = a[2] * b[2];
863 result[3] = a[3] * b[3];
864 store_vector4( inst, machine, result );
865 #if DEBUG_FRAG
866 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
867 result[0], result[1], result[2], result[3],
868 a[0], a[1], a[2], a[3],
869 b[0], b[1], b[2], b[3]);
870 #endif
871 }
872 break;
873 case FP_OPCODE_PK2H: /* pack two 16-bit floats */
874 /* XXX this is probably wrong */
875 {
876 GLfloat a[4], result[4];
877 const GLuint *rawBits = (const GLuint *) a;
878 GLuint *rawResult = (GLuint *) result;
879 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
880 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
881 = rawBits[0] | (rawBits[1] << 16);
882 store_vector4( inst, machine, result );
883 }
884 break;
885 case FP_OPCODE_PK2US: /* pack two GLushorts */
886 {
887 GLfloat a[4], result[4];
888 GLuint usx, usy, *rawResult = (GLuint *) result;
889 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
890 a[0] = CLAMP(a[0], 0.0F, 1.0F);
891 a[1] = CLAMP(a[0], 0.0F, 1.0F);
892 usx = IROUND(a[0] * 65535.0F);
893 usy = IROUND(a[1] * 65535.0F);
894 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
895 = usx | (usy << 16);
896 store_vector4( inst, machine, result );
897 }
898 break;
899 case FP_OPCODE_PK4B: /* pack four GLbytes */
900 {
901 GLfloat a[4], result[4];
902 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
903 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
904 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
905 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
906 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
907 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
908 ubx = IROUND(127.0F * a[0] + 128.0F);
909 uby = IROUND(127.0F * a[1] + 128.0F);
910 ubz = IROUND(127.0F * a[2] + 128.0F);
911 ubw = IROUND(127.0F * a[3] + 128.0F);
912 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
913 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
914 store_vector4( inst, machine, result );
915 }
916 break;
917 case FP_OPCODE_PK4UB: /* pack four GLubytes */
918 {
919 GLfloat a[4], result[4];
920 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
921 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
922 a[0] = CLAMP(a[0], 0.0F, 1.0F);
923 a[1] = CLAMP(a[1], 0.0F, 1.0F);
924 a[2] = CLAMP(a[2], 0.0F, 1.0F);
925 a[3] = CLAMP(a[3], 0.0F, 1.0F);
926 ubx = IROUND(255.0F * a[0]);
927 uby = IROUND(255.0F * a[1]);
928 ubz = IROUND(255.0F * a[2]);
929 ubw = IROUND(255.0F * a[3]);
930 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
931 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
932 store_vector4( inst, machine, result );
933 }
934 break;
935 case FP_OPCODE_POW:
936 {
937 GLfloat a[4], b[4], result[4];
938 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
939 fetch_vector1( ctx, &inst->SrcReg[1], machine, program, b );
940 result[0] = result[1] = result[2] = result[3]
941 = (GLfloat)_mesa_pow(a[0], b[0]);
942 store_vector4( inst, machine, result );
943 }
944 break;
945 case FP_OPCODE_RCP:
946 {
947 GLfloat a[4], result[4];
948 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
949 #if DEBUG_FRAG
950 if (a[0] == 0)
951 printf("RCP(0)\n");
952 else if (IS_INF_OR_NAN(a[0]))
953 printf("RCP(inf)\n");
954 #endif
955 result[0] = result[1] = result[2] = result[3]
956 = 1.0F / a[0];
957 store_vector4( inst, machine, result );
958 }
959 break;
960 case FP_OPCODE_RFL:
961 {
962 GLfloat axis[4], dir[4], result[4], tmp[4];
963 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, axis );
964 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dir );
965 tmp[3] = axis[0] * axis[0]
966 + axis[1] * axis[1]
967 + axis[2] * axis[2];
968 tmp[0] = (2.0F * (axis[0] * dir[0] +
969 axis[1] * dir[1] +
970 axis[2] * dir[2])) / tmp[3];
971 result[0] = tmp[0] * axis[0] - dir[0];
972 result[1] = tmp[0] * axis[1] - dir[1];
973 result[2] = tmp[0] * axis[2] - dir[2];
974 /* result[3] is never written! XXX enforce in parser! */
975 store_vector4( inst, machine, result );
976 }
977 break;
978 case FP_OPCODE_RSQ: /* 1 / sqrt() */
979 {
980 GLfloat a[4], result[4];
981 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
982 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
983 store_vector4( inst, machine, result );
984 #if DEBUG_FRAG
985 printf("RSQ %g = 1/sqrt(%g)\n", result[0], a[0]);
986 #endif
987 }
988 break;
989 case FP_OPCODE_SCS: /* sine and cos */
990 {
991 GLfloat a[4], result[4];
992 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
993 result[0] = cos(a[0]);
994 result[1] = sin(a[0]);
995 result[2] = 0.0; /* undefined! */
996 result[3] = 0.0; /* undefined! */
997 store_vector4( inst, machine, result );
998 }
999 break;
1000 case FP_OPCODE_SEQ: /* set on equal */
1001 {
1002 GLfloat a[4], b[4], result[4];
1003 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1004 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1005 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1006 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1007 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1008 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1009 store_vector4( inst, machine, result );
1010 }
1011 break;
1012 case FP_OPCODE_SFL: /* set false, operands ignored */
1013 {
1014 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1015 store_vector4( inst, machine, result );
1016 }
1017 break;
1018 case FP_OPCODE_SGE: /* set on greater or equal */
1019 {
1020 GLfloat a[4], b[4], result[4];
1021 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1022 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1023 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1024 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1025 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1026 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1027 store_vector4( inst, machine, result );
1028 }
1029 break;
1030 case FP_OPCODE_SGT: /* set on greater */
1031 {
1032 GLfloat a[4], b[4], result[4];
1033 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1034 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1035 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1036 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1037 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1038 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1039 store_vector4( inst, machine, result );
1040 }
1041 break;
1042 case FP_OPCODE_SIN:
1043 {
1044 GLfloat a[4], result[4];
1045 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1046 result[0] = result[1] = result[2] =
1047 result[3] = (GLfloat)_mesa_sin(a[0]);
1048 store_vector4( inst, machine, result );
1049 }
1050 break;
1051 case FP_OPCODE_SLE: /* set on less or equal */
1052 {
1053 GLfloat a[4], b[4], result[4];
1054 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1055 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1056 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1057 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1058 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1059 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1060 store_vector4( inst, machine, result );
1061 }
1062 break;
1063 case FP_OPCODE_SLT: /* set on less */
1064 {
1065 GLfloat a[4], b[4], result[4];
1066 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1067 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1068 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1069 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1070 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1071 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1072 store_vector4( inst, machine, result );
1073 }
1074 break;
1075 case FP_OPCODE_SNE: /* set on not equal */
1076 {
1077 GLfloat a[4], b[4], result[4];
1078 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1079 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1080 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1081 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1082 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1083 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1084 store_vector4( inst, machine, result );
1085 }
1086 break;
1087 case FP_OPCODE_STR: /* set true, operands ignored */
1088 {
1089 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1090 store_vector4( inst, machine, result );
1091 }
1092 break;
1093 case FP_OPCODE_SUB:
1094 {
1095 GLfloat a[4], b[4], result[4];
1096 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1097 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1098 result[0] = a[0] - b[0];
1099 result[1] = a[1] - b[1];
1100 result[2] = a[2] - b[2];
1101 result[3] = a[3] - b[3];
1102 store_vector4( inst, machine, result );
1103 }
1104 break;
1105 case FP_OPCODE_SWZ:
1106 {
1107 /* XXX to do: extended swizzle */
1108 }
1109 break;
1110 case FP_OPCODE_TEX:
1111 /* Texel lookup */
1112 {
1113 GLfloat texcoord[4], color[4];
1114 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1115 /* XXX: Undo perspective divide from interpolate_texcoords() */
1116 fetch_texel( ctx, texcoord,
1117 span->array->lambda[inst->TexSrcUnit][column],
1118 inst->TexSrcUnit, color );
1119 store_vector4( inst, machine, color );
1120 }
1121 break;
1122 case FP_OPCODE_TXB:
1123 /* Texel lookup with LOD bias */
1124 {
1125 GLfloat texcoord[4], color[4];
1126 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1127 /* XXX: apply bias from texcoord[3]!!! */
1128 fetch_texel( ctx, texcoord,
1129 span->array->lambda[inst->TexSrcUnit][column],
1130 inst->TexSrcUnit, color );
1131 store_vector4( inst, machine, color );
1132 }
1133 break;
1134 case FP_OPCODE_TXD:
1135 /* Texture lookup w/ partial derivatives for LOD */
1136 {
1137 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1138 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1139 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dtdx );
1140 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, dtdy );
1141 fetch_texel_deriv( ctx, texcoord, dtdx, dtdy, inst->TexSrcUnit,
1142 color );
1143 store_vector4( inst, machine, color );
1144 }
1145 break;
1146 case FP_OPCODE_TXP:
1147 /* Texture lookup w/ perspective divide */
1148 {
1149 GLfloat texcoord[4], color[4];
1150 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1151 /* Already did perspective divide in interpolate_texcoords() */
1152 fetch_texel( ctx, texcoord,
1153 span->array->lambda[inst->TexSrcUnit][column],
1154 inst->TexSrcUnit, color );
1155 store_vector4( inst, machine, color );
1156 }
1157 break;
1158 case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
1159 /* XXX this is probably wrong */
1160 {
1161 GLfloat a[4], result[4];
1162 const GLuint *rawBits = (const GLuint *) a;
1163 GLuint *rawResult = (GLuint *) result;
1164 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1165 rawResult[0] = rawBits[0] & 0xffff;
1166 rawResult[1] = (rawBits[0] >> 16) & 0xffff;
1167 rawResult[2] = rawBits[0] & 0xffff;
1168 rawResult[3] = (rawBits[0] >> 16) & 0xffff;
1169 store_vector4( inst, machine, result );
1170 }
1171 break;
1172 case FP_OPCODE_UP2US: /* unpack two GLushorts */
1173 {
1174 GLfloat a[4], result[4];
1175 const GLuint *rawBits = (const GLuint *) a;
1176 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1177 result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
1178 result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
1179 result[2] = result[0];
1180 result[3] = result[1];
1181 store_vector4( inst, machine, result );
1182 }
1183 break;
1184 case FP_OPCODE_UP4B: /* unpack four GLbytes */
1185 {
1186 GLfloat a[4], result[4];
1187 const GLuint *rawBits = (const GLuint *) a;
1188 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1189 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1190 result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1191 result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1192 result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1193 store_vector4( inst, machine, result );
1194 }
1195 break;
1196 case FP_OPCODE_UP4UB: /* unpack four GLubytes */
1197 {
1198 GLfloat a[4], result[4];
1199 const GLuint *rawBits = (const GLuint *) a;
1200 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1201 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1202 result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1203 result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1204 result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1205 store_vector4( inst, machine, result );
1206 }
1207 break;
1208 case FP_OPCODE_X2D: /* 2-D matrix transform */
1209 {
1210 GLfloat a[4], b[4], c[4], result[4];
1211 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1212 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1213 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
1214 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1215 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1216 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1217 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1218 store_vector4( inst, machine, result );
1219 }
1220 break;
1221 case FP_OPCODE_END:
1222 return GL_TRUE;
1223 default:
1224 _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program",
1225 inst->Opcode);
1226 return GL_TRUE; /* return value doesn't matter */
1227 }
1228 }
1229 return GL_TRUE;
1230 }
1231
1232
1233 static void
1234 init_machine( GLcontext *ctx, struct fp_machine *machine,
1235 const struct fragment_program *program,
1236 const struct sw_span *span, GLuint col )
1237 {
1238 GLuint inputsRead = program->InputsRead;
1239 GLuint u;
1240
1241 if (ctx->FragmentProgram.CallbackEnabled)
1242 inputsRead = ~0;
1243
1244 /* Clear temporary registers */
1245 _mesa_bzero(machine->Temporaries,
1246 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
1247
1248 /* Load input registers */
1249 if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) {
1250 GLfloat *wpos = machine->Inputs[FRAG_ATTRIB_WPOS];
1251 wpos[0] = (GLfloat) span->x + col;
1252 wpos[1] = (GLfloat) span->y;
1253 wpos[2] = (GLfloat) span->array->z[col] / ctx->DepthMaxF;
1254 wpos[3] = span->w + col * span->dwdx;
1255 }
1256 if (inputsRead & (1 << FRAG_ATTRIB_COL0)) {
1257 GLfloat *col0 = machine->Inputs[FRAG_ATTRIB_COL0];
1258 col0[0] = CHAN_TO_FLOAT(span->array->rgba[col][RCOMP]);
1259 col0[1] = CHAN_TO_FLOAT(span->array->rgba[col][GCOMP]);
1260 col0[2] = CHAN_TO_FLOAT(span->array->rgba[col][BCOMP]);
1261 col0[3] = CHAN_TO_FLOAT(span->array->rgba[col][ACOMP]);
1262 }
1263 if (inputsRead & (1 << FRAG_ATTRIB_COL1)) {
1264 GLfloat *col1 = machine->Inputs[FRAG_ATTRIB_COL1];
1265 col1[0] = CHAN_TO_FLOAT(span->array->spec[col][RCOMP]);
1266 col1[1] = CHAN_TO_FLOAT(span->array->spec[col][GCOMP]);
1267 col1[2] = CHAN_TO_FLOAT(span->array->spec[col][BCOMP]);
1268 col1[3] = CHAN_TO_FLOAT(span->array->spec[col][ACOMP]);
1269 }
1270 if (inputsRead & (1 << FRAG_ATTRIB_FOGC)) {
1271 GLfloat *fogc = machine->Inputs[FRAG_ATTRIB_FOGC];
1272 fogc[0] = span->array->fog[col];
1273 fogc[1] = 0.0F;
1274 fogc[2] = 0.0F;
1275 fogc[3] = 0.0F;
1276 }
1277 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
1278 if (inputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
1279 GLfloat *tex = machine->Inputs[FRAG_ATTRIB_TEX0 + u];
1280 /*ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));*/
1281 COPY_4V(tex, span->array->texcoords[u][col]);
1282 /*ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);*/
1283 }
1284 }
1285
1286 /* init condition codes */
1287 machine->CondCodes[0] = COND_EQ;
1288 machine->CondCodes[1] = COND_EQ;
1289 machine->CondCodes[2] = COND_EQ;
1290 machine->CondCodes[3] = COND_EQ;
1291 }
1292
1293
1294 void
1295 _swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
1296 {
1297 const struct fragment_program *program = ctx->FragmentProgram.Current;
1298 GLuint i;
1299
1300 ctx->_CurrentProgram = GL_FRAGMENT_PROGRAM_ARB; /* or NV, doesn't matter */
1301
1302 for (i = 0; i < span->end; i++) {
1303 if (span->array->mask[i]) {
1304 init_machine(ctx, &ctx->FragmentProgram.Machine,
1305 ctx->FragmentProgram.Current, span, i);
1306
1307 if (!execute_program(ctx, program, ~0,
1308 &ctx->FragmentProgram.Machine, span, i)) {
1309 span->array->mask[i] = GL_FALSE; /* killed fragment */
1310 }
1311
1312 /* Store output registers */
1313 {
1314 const GLfloat *colOut
1315 = ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_COLR];
1316 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
1317 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
1318 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
1319 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
1320 }
1321 /* depth value */
1322 if (program->OutputsWritten & (1 << FRAG_OUTPUT_DEPR))
1323 span->array->z[i] = IROUND(ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_DEPR][0] * ctx->DepthMaxF);
1324 }
1325 }
1326
1327 ctx->_CurrentProgram = 0;
1328 }
1329