projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Assorted casts to silence g++ warnings.
[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 * Return a pointer to the 4-element float vector specified by the given
97 * source register.
98 */
99 static INLINE const GLfloat *
100 get_register_pointer( GLcontext *ctx,
101 const struct fp_src_register *source,
102 const struct fp_machine *machine,
103 const struct fragment_program *program )
104 {
105 const GLfloat *src;
106 switch (source->File) {
107 case PROGRAM_TEMPORARY:
108 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_TEMPS);
109 src = machine->Temporaries[source->Index];
110 break;
111 case PROGRAM_INPUT:
112 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
113 src = machine->Inputs[source->Index];
114 break;
115 case PROGRAM_LOCAL_PARAM:
116 ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
117 src = program->Base.LocalParams[source->Index];
118 break;
119 case PROGRAM_ENV_PARAM:
120 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_PARAMS);
121 src = ctx->FragmentProgram.Parameters[source->Index];
122 break;
123 case PROGRAM_NAMED_PARAM:
124 ASSERT(source->Index < (GLint) program->Parameters->NumParameters);
125 src = program->Parameters->Parameters[source->Index].Values;
126 break;
127 case PROGRAM_STATE_VAR:
128 src = NULL;
129 break;
130 default:
131 _mesa_problem(ctx, "Invalid input register file in fetch_vector4");
132 src = NULL;
133 }
134 return src;
135 }
136
137
138 /**
139 * Fetch a 4-element float vector from the given source register.
140 * Apply swizzling and negating as needed.
141 */
142 static void
143 fetch_vector4( GLcontext *ctx,
144 const struct fp_src_register *source,
145 const struct fp_machine *machine,
146 const struct fragment_program *program,
147 GLfloat result[4] )
148 {
149 const GLfloat *src = get_register_pointer(ctx, source, machine, program);
150 ASSERT(src);
151
152 result[0] = src[source->Swizzle[0]];
153 result[1] = src[source->Swizzle[1]];
154 result[2] = src[source->Swizzle[2]];
155 result[3] = src[source->Swizzle[3]];
156
157 if (source->NegateBase) {
158 result[0] = -result[0];
159 result[1] = -result[1];
160 result[2] = -result[2];
161 result[3] = -result[3];
162 }
163 if (source->Abs) {
164 result[0] = FABSF(result[0]);
165 result[1] = FABSF(result[1]);
166 result[2] = FABSF(result[2]);
167 result[3] = FABSF(result[3]);
168 }
169 if (source->NegateAbs) {
170 result[0] = -result[0];
171 result[1] = -result[1];
172 result[2] = -result[2];
173 result[3] = -result[3];
174 }
175 }
176
177
178 /**
179 * Fetch the derivative with respect to X for the given register.
180 * \return GL_TRUE if it was easily computed or GL_FALSE if we
181 * need to execute another instance of the program (ugh)!
182 */
183 static GLboolean
184 fetch_vector4_deriv( const struct fp_src_register *source,
185 const struct sw_span *span,
186 char xOrY, GLfloat result[4] )
187 {
188 GLfloat src[4];
189
190 ASSERT(xOrY == 'X' || xOrY == 'Y');
191
192 assert(source->File == PROGRAM_INPUT);
193
194 switch (source->Index) {
195 case FRAG_ATTRIB_WPOS:
196 if (xOrY == 'X') {
197 src[0] = 1.0;
198 src[1] = 0.0;
199 src[2] = span->dzdx;
200 src[3] = span->dwdx;
201 }
202 else {
203 src[0] = 0.0;
204 src[1] = 1.0;
205 src[2] = span->dzdy;
206 src[3] = span->dwdy;
207 }
208 break;
209 case FRAG_ATTRIB_COL0:
210 if (xOrY == 'X') {
211 src[0] = span->drdx * (1.0F / CHAN_MAXF);
212 src[1] = span->dgdx * (1.0F / CHAN_MAXF);
213 src[2] = span->dbdx * (1.0F / CHAN_MAXF);
214 src[3] = span->dadx * (1.0F / CHAN_MAXF);
215 }
216 else {
217 src[0] = span->drdy * (1.0F / CHAN_MAXF);
218 src[1] = span->dgdy * (1.0F / CHAN_MAXF);
219 src[2] = span->dbdy * (1.0F / CHAN_MAXF);
220 src[3] = span->dady * (1.0F / CHAN_MAXF);
221 }
222 break;
223 case FRAG_ATTRIB_COL1:
224 if (xOrY == 'X') {
225 src[0] = span->dsrdx * (1.0F / CHAN_MAXF);
226 src[1] = span->dsgdx * (1.0F / CHAN_MAXF);
227 src[2] = span->dsbdx * (1.0F / CHAN_MAXF);
228 src[3] = 0.0; /* XXX need this */
229 }
230 else {
231 src[0] = span->dsrdy * (1.0F / CHAN_MAXF);
232 src[1] = span->dsgdy * (1.0F / CHAN_MAXF);
233 src[2] = span->dsbdy * (1.0F / CHAN_MAXF);
234 src[3] = 0.0; /* XXX need this */
235 }
236 break;
237 case FRAG_ATTRIB_FOGC:
238 if (xOrY == 'X') {
239 src[0] = span->dfogdx;
240 src[1] = 0.0;
241 src[2] = 0.0;
242 src[3] = 0.0;
243 }
244 else {
245 src[0] = span->dfogdy;
246 src[1] = 0.0;
247 src[2] = 0.0;
248 src[3] = 0.0;
249 }
250 break;
251 case FRAG_ATTRIB_TEX0:
252 case FRAG_ATTRIB_TEX1:
253 case FRAG_ATTRIB_TEX2:
254 case FRAG_ATTRIB_TEX3:
255 case FRAG_ATTRIB_TEX4:
256 case FRAG_ATTRIB_TEX5:
257 case FRAG_ATTRIB_TEX6:
258 case FRAG_ATTRIB_TEX7:
259 if (xOrY == 'X') {
260 const GLuint u = source->Index - FRAG_ATTRIB_TEX0;
261 src[0] = span->texStepX[u][0] * (1.0F / CHAN_MAXF);
262 src[1] = span->texStepX[u][1] * (1.0F / CHAN_MAXF);
263 src[2] = span->texStepX[u][2] * (1.0F / CHAN_MAXF);
264 src[3] = span->texStepX[u][3] * (1.0F / CHAN_MAXF);
265 }
266 else {
267 const GLuint u = source->Index - FRAG_ATTRIB_TEX0;
268 src[0] = span->texStepY[u][0] * (1.0F / CHAN_MAXF);
269 src[1] = span->texStepY[u][1] * (1.0F / CHAN_MAXF);
270 src[2] = span->texStepY[u][2] * (1.0F / CHAN_MAXF);
271 src[3] = span->texStepY[u][3] * (1.0F / CHAN_MAXF);
272 }
273 break;
274 default:
275 return GL_FALSE;
276 }
277
278 result[0] = src[source->Swizzle[0]];
279 result[1] = src[source->Swizzle[1]];
280 result[2] = src[source->Swizzle[2]];
281 result[3] = src[source->Swizzle[3]];
282
283 if (source->NegateBase) {
284 result[0] = -result[0];
285 result[1] = -result[1];
286 result[2] = -result[2];
287 result[3] = -result[3];
288 }
289 if (source->Abs) {
290 result[0] = FABSF(result[0]);
291 result[1] = FABSF(result[1]);
292 result[2] = FABSF(result[2]);
293 result[3] = FABSF(result[3]);
294 }
295 if (source->NegateAbs) {
296 result[0] = -result[0];
297 result[1] = -result[1];
298 result[2] = -result[2];
299 result[3] = -result[3];
300 }
301 return GL_TRUE;
302 }
303
304
305 /**
306 * As above, but only return result[0] element.
307 */
308 static void
309 fetch_vector1( GLcontext *ctx,
310 const struct fp_src_register *source,
311 const struct fp_machine *machine,
312 const struct fragment_program *program,
313 GLfloat result[4] )
314 {
315 const GLfloat *src = get_register_pointer(ctx, source, machine, program);
316 ASSERT(src);
317
318 result[0] = src[source->Swizzle[0]];
319
320 if (source->NegateBase) {
321 result[0] = -result[0];
322 }
323 if (source->Abs) {
324 result[0] = FABSF(result[0]);
325 }
326 if (source->NegateAbs) {
327 result[0] = -result[0];
328 }
329 }
330
331
332 /*
333 * Test value against zero and return GT, LT, EQ or UN if NaN.
334 */
335 static INLINE GLuint
336 generate_cc( float value )
337 {
338 if (value != value)
339 return COND_UN; /* NaN */
340 if (value > 0.0F)
341 return COND_GT;
342 if (value < 0.0F)
343 return COND_LT;
344 return COND_EQ;
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 codee.
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 * Store 4 floats into a register. Observe the instructions saturate and
370 * set-condition-code flags.
371 */
372 static void
373 store_vector4( const struct fp_instruction *inst,
374 struct fp_machine *machine,
375 const GLfloat value[4] )
376 {
377 const struct fp_dst_register *dest = &(inst->DstReg);
378 const GLboolean clamp = inst->Saturate;
379 const GLboolean updateCC = inst->UpdateCondRegister;
380 GLfloat *dstReg;
381 GLfloat clampedValue[4];
382 const GLboolean *writeMask = dest->WriteMask;
383 GLboolean condWriteMask[4];
384
385 switch (dest->File) {
386 case PROGRAM_OUTPUT:
387 dstReg = machine->Outputs[dest->Index];
388 break;
389 case PROGRAM_TEMPORARY:
390 dstReg = machine->Temporaries[dest->Index];
391 break;
392 default:
393 _mesa_problem(NULL, "bad register file in store_vector4(fp)");
394 return;
395 }
396
397 #if DEBUG_FRAG
398 if (value[0] > 1.0e10 ||
399 IS_INF_OR_NAN(value[0]) ||
400 IS_INF_OR_NAN(value[1]) ||
401 IS_INF_OR_NAN(value[2]) ||
402 IS_INF_OR_NAN(value[3]) )
403 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
404 #endif
405
406 if (clamp) {
407 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
408 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
409 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
410 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
411 value = clampedValue;
412 }
413
414 if (dest->CondMask != COND_TR) {
415 condWriteMask[0] = writeMask[0]
416 && test_cc(machine->CondCodes[dest->CondSwizzle[0]], dest->CondMask);
417 condWriteMask[1] = writeMask[1]
418 && test_cc(machine->CondCodes[dest->CondSwizzle[1]], dest->CondMask);
419 condWriteMask[2] = writeMask[2]
420 && test_cc(machine->CondCodes[dest->CondSwizzle[2]], dest->CondMask);
421 condWriteMask[3] = writeMask[3]
422 && test_cc(machine->CondCodes[dest->CondSwizzle[3]], dest->CondMask);
423 writeMask = condWriteMask;
424 }
425
426 if (writeMask[0]) {
427 dstReg[0] = value[0];
428 if (updateCC)
429 machine->CondCodes[0] = generate_cc(value[0]);
430 }
431 if (writeMask[1]) {
432 dstReg[1] = value[1];
433 if (updateCC)
434 machine->CondCodes[1] = generate_cc(value[1]);
435 }
436 if (writeMask[2]) {
437 dstReg[2] = value[2];
438 if (updateCC)
439 machine->CondCodes[2] = generate_cc(value[2]);
440 }
441 if (writeMask[3]) {
442 dstReg[3] = value[3];
443 if (updateCC)
444 machine->CondCodes[3] = generate_cc(value[3]);
445 }
446 }
447
448
449 /**
450 * Initialize a new machine state instance from an existing one, adding
451 * the partial derivatives onto the input registers.
452 * Used to implement DDX and DDY instructions in non-trivial cases.
453 */
454 static void
455 init_machine_deriv( GLcontext *ctx,
456 const struct fp_machine *machine,
457 const struct fragment_program *program,
458 const struct sw_span *span, char xOrY,
459 struct fp_machine *dMachine )
460 {
461 GLuint u;
462
463 ASSERT(xOrY == 'X' || xOrY == 'Y');
464
465 /* copy existing machine */
466 _mesa_memcpy(dMachine, machine, sizeof(struct fp_machine));
467
468 /* Clear temporary registers */
469 _mesa_bzero( (void*) machine->Temporaries,
470 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
471
472 /* Add derivatives */
473 if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
474 GLfloat *wpos = (GLfloat*) machine->Inputs[FRAG_ATTRIB_WPOS];
475 if (xOrY == 'X') {
476 wpos[0] += 1.0F;
477 wpos[1] += 0.0F;
478 wpos[2] += span->dzdx;
479 wpos[3] += span->dwdx;
480 }
481 else {
482 wpos[0] += 0.0F;
483 wpos[1] += 1.0F;
484 wpos[2] += span->dzdy;
485 wpos[3] += span->dwdy;
486 }
487 }
488 if (program->InputsRead & (1 << FRAG_ATTRIB_COL0)) {
489 GLfloat *col0 = (GLfloat*) machine->Inputs[FRAG_ATTRIB_COL0];
490 if (xOrY == 'X') {
491 col0[0] += span->drdx * (1.0F / CHAN_MAXF);
492 col0[1] += span->dgdx * (1.0F / CHAN_MAXF);
493 col0[2] += span->dbdx * (1.0F / CHAN_MAXF);
494 col0[3] += span->dadx * (1.0F / CHAN_MAXF);
495 }
496 else {
497 col0[0] += span->drdy * (1.0F / CHAN_MAXF);
498 col0[1] += span->dgdy * (1.0F / CHAN_MAXF);
499 col0[2] += span->dbdy * (1.0F / CHAN_MAXF);
500 col0[3] += span->dady * (1.0F / CHAN_MAXF);
501 }
502 }
503 if (program->InputsRead & (1 << FRAG_ATTRIB_COL1)) {
504 GLfloat *col1 = (GLfloat*) machine->Inputs[FRAG_ATTRIB_COL1];
505 if (xOrY == 'X') {
506 col1[0] += span->dsrdx * (1.0F / CHAN_MAXF);
507 col1[1] += span->dsgdx * (1.0F / CHAN_MAXF);
508 col1[2] += span->dsbdx * (1.0F / CHAN_MAXF);
509 col1[3] += 0.0; /*XXX fix */
510 }
511 else {
512 col1[0] += span->dsrdy * (1.0F / CHAN_MAXF);
513 col1[1] += span->dsgdy * (1.0F / CHAN_MAXF);
514 col1[2] += span->dsbdy * (1.0F / CHAN_MAXF);
515 col1[3] += 0.0; /*XXX fix */
516 }
517 }
518 if (program->InputsRead & (1 << FRAG_ATTRIB_FOGC)) {
519 GLfloat *fogc = (GLfloat*) machine->Inputs[FRAG_ATTRIB_FOGC];
520 if (xOrY == 'X') {
521 fogc[0] += span->dfogdx;
522 }
523 else {
524 fogc[0] += span->dfogdy;
525 }
526 }
527 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
528 if (program->InputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
529 GLfloat *tex = (GLfloat*) machine->Inputs[FRAG_ATTRIB_TEX0 + u];
530 if (xOrY == 'X') {
531 tex[0] += span->texStepX[u][0];
532 tex[1] += span->texStepX[u][1];
533 tex[2] += span->texStepX[u][2];
534 tex[3] += span->texStepX[u][3];
535 }
536 else {
537 tex[0] += span->texStepY[u][0];
538 tex[1] += span->texStepY[u][1];
539 tex[2] += span->texStepY[u][2];
540 tex[3] += span->texStepY[u][3];
541 }
542 }
543 }
544
545 /* init condition codes */
546 dMachine->CondCodes[0] = COND_EQ;
547 dMachine->CondCodes[1] = COND_EQ;
548 dMachine->CondCodes[2] = COND_EQ;
549 dMachine->CondCodes[3] = COND_EQ;
550 }
551
552
553 /**
554 * Execute the given vertex program.
555 * NOTE: we do everything in single-precision floating point; we don't
556 * currently observe the single/half/fixed-precision qualifiers.
557 * \param ctx - rendering context
558 * \param program - the fragment program to execute
559 * \param machine - machine state (register file)
560 * \param maxInst - max number of instructions to execute
561 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
562 */
563 static GLboolean
564 execute_program( GLcontext *ctx,
565 const struct fragment_program *program, GLuint maxInst,
566 struct fp_machine *machine, const struct sw_span *span,
567 GLuint column )
568 {
569 GLuint pc;
570
571 #if DEBUG_FRAG
572 printf("execute fragment program --------------------\n");
573 #endif
574
575 for (pc = 0; pc < maxInst; pc++) {
576 const struct fp_instruction *inst = program->Instructions + pc;
577
578 if (ctx->FragmentProgram.CallbackEnabled &&
579 ctx->FragmentProgram.Callback) {
580 ctx->FragmentProgram.CurrentPosition = inst->StringPos;
581 ctx->FragmentProgram.Callback(program->Base.Target,
582 ctx->FragmentProgram.CallbackData);
583 }
584
585 switch (inst->Opcode) {
586 case FP_OPCODE_ABS:
587 {
588 GLfloat a[4], result[4];
589 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
590 result[0] = FABSF(a[0]);
591 result[1] = FABSF(a[1]);
592 result[2] = FABSF(a[2]);
593 result[3] = FABSF(a[3]);
594 store_vector4( inst, machine, result );
595 }
596 break;
597 case FP_OPCODE_ADD:
598 {
599 GLfloat a[4], b[4], result[4];
600 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
601 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
602 result[0] = a[0] + b[0];
603 result[1] = a[1] + b[1];
604 result[2] = a[2] + b[2];
605 result[3] = a[3] + b[3];
606 store_vector4( inst, machine, result );
607 }
608 break;
609 case FP_OPCODE_CMP:
610 {
611 GLfloat a[4], b[4], c[4], result[4];
612 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
613 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
614 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
615 result[0] = a[0] < 0.0F ? b[0] : c[0];
616 result[1] = a[1] < 0.0F ? b[1] : c[1];
617 result[2] = a[2] < 0.0F ? b[2] : c[2];
618 result[3] = a[3] < 0.0F ? b[3] : c[3];
619 store_vector4( inst, machine, result );
620 }
621 break;
622 case FP_OPCODE_COS:
623 {
624 GLfloat a[4], result[4];
625 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
626 result[0] = result[1] = result[2] = result[3] = (GLfloat)_mesa_cos(a[0]);
627 store_vector4( inst, machine, result );
628 }
629 break;
630 case FP_OPCODE_DDX: /* Partial derivative with respect to X */
631 {
632 GLfloat a[4], aNext[4], result[4];
633 struct fp_machine dMachine;
634 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', result)) {
635 /* This is tricky. Make a copy of the current machine state,
636 * increment the input registers by the dx or dy partial
637 * derivatives, then re-execute the program up to the
638 * preceeding instruction, then fetch the source register.
639 * Finally, find the difference in the register values for
640 * the original and derivative runs.
641 */
642 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
643 init_machine_deriv(ctx, machine, program, span,
644 'X', &dMachine);
645 execute_program(ctx, program, pc, &dMachine, span, column);
646 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
647 result[0] = aNext[0] - a[0];
648 result[1] = aNext[1] - a[1];
649 result[2] = aNext[2] - a[2];
650 result[3] = aNext[3] - a[3];
651 }
652 store_vector4( inst, machine, result );
653 }
654 break;
655 case FP_OPCODE_DDY: /* Partial derivative with respect to Y */
656 {
657 GLfloat a[4], aNext[4], result[4];
658 struct fp_machine dMachine;
659 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', result)) {
660 init_machine_deriv(ctx, machine, program, span,
661 'Y', &dMachine);
662 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
663 execute_program(ctx, program, pc, &dMachine, span, column);
664 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
665 result[0] = aNext[0] - a[0];
666 result[1] = aNext[1] - a[1];
667 result[2] = aNext[2] - a[2];
668 result[3] = aNext[3] - a[3];
669 }
670 store_vector4( inst, machine, result );
671 }
672 break;
673 case FP_OPCODE_DP3:
674 {
675 GLfloat a[4], b[4], result[4];
676 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
677 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
678 result[0] = result[1] = result[2] = result[3] =
679 a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
680 store_vector4( inst, machine, result );
681 #if DEBUG_FRAG
682 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
683 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
684 #endif
685 }
686 break;
687 case FP_OPCODE_DP4:
688 {
689 GLfloat a[4], b[4], result[4];
690 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
691 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
692 result[0] = result[1] = result[2] = result[3] =
693 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
694 store_vector4( inst, machine, result );
695 }
696 break;
697 case FP_OPCODE_DPH:
698 {
699 GLfloat a[4], b[4], result[4];
700 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
701 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
702 result[0] = result[1] = result[2] = result[3] =
703 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + b[3];
704 store_vector4( inst, machine, result );
705 }
706 break;
707 case FP_OPCODE_DST: /* Distance vector */
708 {
709 GLfloat a[4], b[4], result[4];
710 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
711 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
712 result[0] = 1.0F;
713 result[1] = a[1] * b[1];
714 result[2] = a[2];
715 result[3] = b[3];
716 store_vector4( inst, machine, result );
717 }
718 break;
719 case FP_OPCODE_EX2: /* Exponential base 2 */
720 {
721 GLfloat a[4], result[4];
722 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
723 result[0] = result[1] = result[2] = result[3] =
724 (GLfloat) _mesa_pow(2.0, a[0]);
725 store_vector4( inst, machine, result );
726 }
727 break;
728 case FP_OPCODE_FLR:
729 {
730 GLfloat a[4], result[4];
731 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
732 result[0] = FLOORF(a[0]);
733 result[1] = FLOORF(a[1]);
734 result[2] = FLOORF(a[2]);
735 result[3] = FLOORF(a[3]);
736 store_vector4( inst, machine, result );
737 }
738 break;
739 case FP_OPCODE_FRC:
740 {
741 GLfloat a[4], result[4];
742 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
743 result[0] = a[0] - FLOORF(a[0]);
744 result[1] = a[1] - FLOORF(a[1]);
745 result[2] = a[2] - FLOORF(a[2]);
746 result[3] = a[3] - FLOORF(a[3]);
747 store_vector4( inst, machine, result );
748 }
749 break;
750 case FP_OPCODE_KIL:
751 {
752 const GLuint *swizzle = inst->DstReg.CondSwizzle;
753 const GLuint condMask = inst->DstReg.CondMask;
754 if (test_cc(machine->CondCodes[swizzle[0]], condMask) ||
755 test_cc(machine->CondCodes[swizzle[1]], condMask) ||
756 test_cc(machine->CondCodes[swizzle[2]], condMask) ||
757 test_cc(machine->CondCodes[swizzle[3]], condMask)) {
758 return GL_FALSE;
759 }
760 }
761 break;
762 case FP_OPCODE_LG2: /* log base 2 */
763 {
764 GLfloat a[4], result[4];
765 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
766 result[0] = result[1] = result[2] = result[3]
767 = LOG2(a[0]);
768 store_vector4( inst, machine, result );
769 }
770 break;
771 case FP_OPCODE_LIT:
772 {
773 GLfloat a[4], result[4];
774 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
775 if (a[0] < 0.0F)
776 a[0] = 0.0F;
777 if (a[1] < 0.0F)
778 a[1] = 0.0F;
779 result[0] = 1.0F;
780 result[1] = a[0];
781 result[2] = (a[0] > 0.0F) ? (GLfloat)_mesa_pow(2.0, a[3]) : 0.0F;
782 result[3] = 1.0F;
783 store_vector4( inst, machine, result );
784 }
785 break;
786 case FP_OPCODE_LRP:
787 {
788 GLfloat a[4], b[4], c[4], result[4];
789 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
790 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
791 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
792 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
793 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
794 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
795 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
796 store_vector4( inst, machine, result );
797 }
798 break;
799 case FP_OPCODE_MAD:
800 {
801 GLfloat a[4], b[4], c[4], result[4];
802 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
803 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
804 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
805 result[0] = a[0] * b[0] + c[0];
806 result[1] = a[1] * b[1] + c[1];
807 result[2] = a[2] * b[2] + c[2];
808 result[3] = a[3] * b[3] + c[3];
809 store_vector4( inst, machine, result );
810 }
811 break;
812 case FP_OPCODE_MAX:
813 {
814 GLfloat a[4], b[4], result[4];
815 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
816 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
817 result[0] = MAX2(a[0], b[0]);
818 result[1] = MAX2(a[1], b[1]);
819 result[2] = MAX2(a[2], b[2]);
820 result[3] = MAX2(a[3], b[3]);
821 store_vector4( inst, machine, result );
822 }
823 break;
824 case FP_OPCODE_MIN:
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] = MIN2(a[0], b[0]);
830 result[1] = MIN2(a[1], b[1]);
831 result[2] = MIN2(a[2], b[2]);
832 result[3] = MIN2(a[3], b[3]);
833 store_vector4( inst, machine, result );
834 }
835 break;
836 case FP_OPCODE_MOV:
837 {
838 GLfloat result[4];
839 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, result );
840 store_vector4( inst, machine, result );
841 }
842 break;
843 case FP_OPCODE_MUL:
844 {
845 GLfloat a[4], b[4], result[4];
846 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
847 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
848 result[0] = a[0] * b[0];
849 result[1] = a[1] * b[1];
850 result[2] = a[2] * b[2];
851 result[3] = a[3] * b[3];
852 store_vector4( inst, machine, result );
853 #if DEBUG_FRAG
854 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
855 result[0], result[1], result[2], result[3],
856 a[0], a[1], a[2], a[3],
857 b[0], b[1], b[2], b[3]);
858 #endif
859 }
860 break;
861 case FP_OPCODE_PK2H: /* pack two 16-bit floats */
862 /* XXX this is probably wrong */
863 {
864 GLfloat a[4], result[4];
865 const GLuint *rawBits = (const GLuint *) a;
866 GLuint *rawResult = (GLuint *) result;
867 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
868 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
869 = rawBits[0] | (rawBits[1] << 16);
870 store_vector4( inst, machine, result );
871 }
872 break;
873 case FP_OPCODE_PK2US: /* pack two GLushorts */
874 {
875 GLfloat a[4], result[4];
876 GLuint usx, usy, *rawResult = (GLuint *) result;
877 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
878 a[0] = CLAMP(a[0], 0.0F, 1.0F);
879 a[1] = CLAMP(a[0], 0.0F, 1.0F);
880 usx = IROUND(a[0] * 65535.0F);
881 usy = IROUND(a[1] * 65535.0F);
882 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
883 = usx | (usy << 16);
884 store_vector4( inst, machine, result );
885 }
886 break;
887 case FP_OPCODE_PK4B: /* pack four GLbytes */
888 {
889 GLfloat a[4], result[4];
890 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
891 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
892 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
893 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
894 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
895 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
896 ubx = IROUND(127.0F * a[0] + 128.0F);
897 uby = IROUND(127.0F * a[1] + 128.0F);
898 ubz = IROUND(127.0F * a[2] + 128.0F);
899 ubw = IROUND(127.0F * a[3] + 128.0F);
900 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
901 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
902 store_vector4( inst, machine, result );
903 }
904 break;
905 case FP_OPCODE_PK4UB: /* pack four GLubytes */
906 {
907 GLfloat a[4], result[4];
908 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
909 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
910 a[0] = CLAMP(a[0], 0.0F, 1.0F);
911 a[1] = CLAMP(a[1], 0.0F, 1.0F);
912 a[2] = CLAMP(a[2], 0.0F, 1.0F);
913 a[3] = CLAMP(a[3], 0.0F, 1.0F);
914 ubx = IROUND(255.0F * a[0]);
915 uby = IROUND(255.0F * a[1]);
916 ubz = IROUND(255.0F * a[2]);
917 ubw = IROUND(255.0F * a[3]);
918 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
919 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
920 store_vector4( inst, machine, result );
921 }
922 break;
923 case FP_OPCODE_POW:
924 {
925 GLfloat a[4], b[4], result[4];
926 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
927 fetch_vector1( ctx, &inst->SrcReg[1], machine, program, b );
928 result[0] = result[1] = result[2] = result[3]
929 = (GLfloat)_mesa_pow(a[0], b[0]);
930 store_vector4( inst, machine, result );
931 }
932 break;
933 case FP_OPCODE_RCP:
934 {
935 GLfloat a[4], result[4];
936 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
937 #if DEBUG_FRAG
938 if (a[0] == 0)
939 printf("RCP(0)\n");
940 else if (IS_INF_OR_NAN(a[0]))
941 printf("RCP(inf)\n");
942 #endif
943 result[0] = result[1] = result[2] = result[3]
944 = 1.0F / a[0];
945 store_vector4( inst, machine, result );
946 }
947 break;
948 case FP_OPCODE_RFL:
949 {
950 GLfloat axis[4], dir[4], result[4], tmp[4];
951 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, axis );
952 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dir );
953 tmp[3] = axis[0] * axis[0]
954 + axis[1] * axis[1]
955 + axis[2] * axis[2];
956 tmp[0] = (2.0F * (axis[0] * dir[0] +
957 axis[1] * dir[1] +
958 axis[2] * dir[2])) / tmp[3];
959 result[0] = tmp[0] * axis[0] - dir[0];
960 result[1] = tmp[0] * axis[1] - dir[1];
961 result[2] = tmp[0] * axis[2] - dir[2];
962 /* result[3] is never written! XXX enforce in parser! */
963 store_vector4( inst, machine, result );
964 }
965 break;
966 case FP_OPCODE_RSQ: /* 1 / sqrt() */
967 {
968 GLfloat a[4], result[4];
969 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
970 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
971 store_vector4( inst, machine, result );
972 #if DEBUG_FRAG
973 printf("RSQ %g = 1/sqrt(%g)\n", result[0], a[0]);
974 #endif
975 }
976 break;
977 case FP_OPCODE_SCS: /* sine and cos */
978 {
979 GLfloat a[4], result[4];
980 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
981 result[0] = (GLfloat)cos(a[0]);
982 result[1] = (GLfloat)sin(a[0]);
983 result[2] = 0.0; /* undefined! */
984 result[3] = 0.0; /* undefined! */
985 store_vector4( inst, machine, result );
986 }
987 break;
988 case FP_OPCODE_SEQ: /* set on equal */
989 {
990 GLfloat a[4], b[4], result[4];
991 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
992 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
993 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
994 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
995 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
996 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
997 store_vector4( inst, machine, result );
998 }
999 break;
1000 case FP_OPCODE_SFL: /* set false, operands ignored */
1001 {
1002 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1003 store_vector4( inst, machine, result );
1004 }
1005 break;
1006 case FP_OPCODE_SGE: /* set on greater or equal */
1007 {
1008 GLfloat a[4], b[4], result[4];
1009 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1010 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1011 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1012 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1013 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1014 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1015 store_vector4( inst, machine, result );
1016 }
1017 break;
1018 case FP_OPCODE_SGT: /* set on greater */
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_SIN:
1031 {
1032 GLfloat a[4], result[4];
1033 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1034 result[0] = result[1] = result[2] =
1035 result[3] = (GLfloat)_mesa_sin(a[0]);
1036 store_vector4( inst, machine, result );
1037 }
1038 break;
1039 case FP_OPCODE_SLE: /* set on less or equal */
1040 {
1041 GLfloat a[4], b[4], result[4];
1042 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1043 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1044 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1045 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1046 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1047 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1048 store_vector4( inst, machine, result );
1049 }
1050 break;
1051 case FP_OPCODE_SLT: /* set on less */
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_SNE: /* set on not equal */
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_STR: /* set true, operands ignored */
1076 {
1077 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1078 store_vector4( inst, machine, result );
1079 }
1080 break;
1081 case FP_OPCODE_SUB:
1082 {
1083 GLfloat a[4], b[4], result[4];
1084 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1085 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1086 result[0] = a[0] - b[0];
1087 result[1] = a[1] - b[1];
1088 result[2] = a[2] - b[2];
1089 result[3] = a[3] - b[3];
1090 store_vector4( inst, machine, result );
1091 }
1092 break;
1093 case FP_OPCODE_SWZ:
1094 {
1095 const struct fp_src_register *source = &inst->SrcReg[0];
1096 const GLfloat *src = get_register_pointer(ctx, source,
1097 machine, program);
1098 GLfloat result[4];
1099 GLuint i;
1100
1101 /* do extended swizzling here */
1102 for (i = 0; i < 3; i++) {
1103 if (source->Swizzle[i] == SWIZZLE_ZERO)
1104 result[i] = 0.0;
1105 else if (source->Swizzle[i] == SWIZZLE_ONE)
1106 result[i] = -1.0;
1107 else
1108 result[i] = -src[source->Swizzle[i]];
1109 if (source->NegateBase)
1110 result[i] = -result[i];
1111 }
1112 store_vector4( inst, machine, result );
1113 }
1114 break;
1115 case FP_OPCODE_TEX:
1116 /* Texel lookup */
1117 {
1118 GLfloat texcoord[4], color[4];
1119 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1120 /* XXX: Undo perspective divide from interpolate_texcoords() */
1121 fetch_texel( ctx, texcoord,
1122 span->array->lambda[inst->TexSrcUnit][column],
1123 inst->TexSrcUnit, color );
1124 store_vector4( inst, machine, color );
1125 }
1126 break;
1127 case FP_OPCODE_TXB:
1128 /* Texel lookup with LOD bias */
1129 {
1130 GLfloat texcoord[4], color[4], bias, lambda;
1131
1132 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1133 /* texcoord[3] is the bias to add to lambda */
1134 bias = ctx->Texture.Unit[inst->TexSrcUnit].LodBias
1135 + ctx->Texture.Unit[inst->TexSrcUnit]._Current->LodBias
1136 + texcoord[3];
1137 lambda = span->array->lambda[inst->TexSrcUnit][column] + bias;
1138 fetch_texel( ctx, texcoord, lambda,
1139 inst->TexSrcUnit, color );
1140 store_vector4( inst, machine, color );
1141 }
1142 break;
1143 case FP_OPCODE_TXD:
1144 /* Texture lookup w/ partial derivatives for LOD */
1145 {
1146 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1147 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1148 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dtdx );
1149 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, dtdy );
1150 fetch_texel_deriv( ctx, texcoord, dtdx, dtdy, inst->TexSrcUnit,
1151 color );
1152 store_vector4( inst, machine, color );
1153 }
1154 break;
1155 case FP_OPCODE_TXP:
1156 /* Texture lookup w/ perspective divide */
1157 {
1158 GLfloat texcoord[4], color[4];
1159 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1160 /* Already did perspective divide in interpolate_texcoords() */
1161 fetch_texel( ctx, texcoord,
1162 span->array->lambda[inst->TexSrcUnit][column],
1163 inst->TexSrcUnit, color );
1164 store_vector4( inst, machine, color );
1165 }
1166 break;
1167 case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
1168 /* XXX this is probably wrong */
1169 {
1170 GLfloat a[4], result[4];
1171 const GLuint *rawBits = (const GLuint *) a;
1172 GLuint *rawResult = (GLuint *) result;
1173 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1174 rawResult[0] = rawBits[0] & 0xffff;
1175 rawResult[1] = (rawBits[0] >> 16) & 0xffff;
1176 rawResult[2] = rawBits[0] & 0xffff;
1177 rawResult[3] = (rawBits[0] >> 16) & 0xffff;
1178 store_vector4( inst, machine, result );
1179 }
1180 break;
1181 case FP_OPCODE_UP2US: /* unpack two GLushorts */
1182 {
1183 GLfloat a[4], result[4];
1184 const GLuint *rawBits = (const GLuint *) a;
1185 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1186 result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
1187 result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
1188 result[2] = result[0];
1189 result[3] = result[1];
1190 store_vector4( inst, machine, result );
1191 }
1192 break;
1193 case FP_OPCODE_UP4B: /* unpack four GLbytes */
1194 {
1195 GLfloat a[4], result[4];
1196 const GLuint *rawBits = (const GLuint *) a;
1197 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1198 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1199 result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1200 result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1201 result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1202 store_vector4( inst, machine, result );
1203 }
1204 break;
1205 case FP_OPCODE_UP4UB: /* unpack four GLubytes */
1206 {
1207 GLfloat a[4], result[4];
1208 const GLuint *rawBits = (const GLuint *) a;
1209 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1210 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1211 result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1212 result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1213 result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1214 store_vector4( inst, machine, result );
1215 }
1216 break;
1217 case FP_OPCODE_X2D: /* 2-D matrix transform */
1218 {
1219 GLfloat a[4], b[4], c[4], result[4];
1220 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1221 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1222 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
1223 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1224 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1225 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1226 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1227 store_vector4( inst, machine, result );
1228 }
1229 break;
1230 case FP_OPCODE_END:
1231 return GL_TRUE;
1232 default:
1233 _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program",
1234 inst->Opcode);
1235 return GL_TRUE; /* return value doesn't matter */
1236 }
1237 }
1238 return GL_TRUE;
1239 }
1240
1241
1242 static void
1243 init_machine( GLcontext *ctx, struct fp_machine *machine,
1244 const struct fragment_program *program,
1245 const struct sw_span *span, GLuint col )
1246 {
1247 GLuint inputsRead = program->InputsRead;
1248 GLuint u;
1249
1250 if (ctx->FragmentProgram.CallbackEnabled)
1251 inputsRead = ~0;
1252
1253 /* Clear temporary registers */
1254 _mesa_bzero(machine->Temporaries,
1255 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
1256
1257 /* Load input registers */
1258 if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) {
1259 GLfloat *wpos = machine->Inputs[FRAG_ATTRIB_WPOS];
1260 wpos[0] = (GLfloat) span->x + col;
1261 wpos[1] = (GLfloat) span->y;
1262 wpos[2] = (GLfloat) span->array->z[col] / ctx->DepthMaxF;
1263 wpos[3] = span->w + col * span->dwdx;
1264 }
1265 if (inputsRead & (1 << FRAG_ATTRIB_COL0)) {
1266 GLfloat *col0 = machine->Inputs[FRAG_ATTRIB_COL0];
1267 col0[0] = CHAN_TO_FLOAT(span->array->rgba[col][RCOMP]);
1268 col0[1] = CHAN_TO_FLOAT(span->array->rgba[col][GCOMP]);
1269 col0[2] = CHAN_TO_FLOAT(span->array->rgba[col][BCOMP]);
1270 col0[3] = CHAN_TO_FLOAT(span->array->rgba[col][ACOMP]);
1271 }
1272 if (inputsRead & (1 << FRAG_ATTRIB_COL1)) {
1273 GLfloat *col1 = machine->Inputs[FRAG_ATTRIB_COL1];
1274 col1[0] = CHAN_TO_FLOAT(span->array->spec[col][RCOMP]);
1275 col1[1] = CHAN_TO_FLOAT(span->array->spec[col][GCOMP]);
1276 col1[2] = CHAN_TO_FLOAT(span->array->spec[col][BCOMP]);
1277 col1[3] = CHAN_TO_FLOAT(span->array->spec[col][ACOMP]);
1278 }
1279 if (inputsRead & (1 << FRAG_ATTRIB_FOGC)) {
1280 GLfloat *fogc = machine->Inputs[FRAG_ATTRIB_FOGC];
1281 fogc[0] = span->array->fog[col];
1282 fogc[1] = 0.0F;
1283 fogc[2] = 0.0F;
1284 fogc[3] = 0.0F;
1285 }
1286 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
1287 if (inputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
1288 GLfloat *tex = machine->Inputs[FRAG_ATTRIB_TEX0 + u];
1289 /*ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));*/
1290 COPY_4V(tex, span->array->texcoords[u][col]);
1291 /*ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);*/
1292 }
1293 }
1294
1295 /* init condition codes */
1296 machine->CondCodes[0] = COND_EQ;
1297 machine->CondCodes[1] = COND_EQ;
1298 machine->CondCodes[2] = COND_EQ;
1299 machine->CondCodes[3] = COND_EQ;
1300 }
1301
1302
1303 void
1304 _swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
1305 {
1306 const struct fragment_program *program = ctx->FragmentProgram.Current;
1307 GLuint i;
1308
1309 ctx->_CurrentProgram = GL_FRAGMENT_PROGRAM_ARB; /* or NV, doesn't matter */
1310
1311 for (i = 0; i < span->end; i++) {
1312 if (span->array->mask[i]) {
1313 init_machine(ctx, &ctx->FragmentProgram.Machine,
1314 ctx->FragmentProgram.Current, span, i);
1315
1316 if (!execute_program(ctx, program, ~0,
1317 &ctx->FragmentProgram.Machine, span, i)) {
1318 span->array->mask[i] = GL_FALSE; /* killed fragment */
1319 }
1320
1321 /* Store output registers */
1322 {
1323 const GLfloat *colOut
1324 = ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_COLR];
1325 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
1326 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
1327 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
1328 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
1329 }
1330 /* depth value */
1331 if (program->OutputsWritten & (1 << FRAG_OUTPUT_DEPR))
1332 span->array->z[i] = IROUND(ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_DEPR][0] * ctx->DepthMaxF);
1333 }
1334 }
1335
1336 ctx->_CurrentProgram = 0;
1337 }
1338