projects / mesa.git / blob
? search:


94458238b517f727f6f31a79e4e288cf67872abf
[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 * Fetch a texel.
43 */
44 static void
45 fetch_texel( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda,
46 GLuint unit, GLfloat color[4] )
47 {
48 GLchan rgba[4];
49 SWcontext *swrast = SWRAST_CONTEXT(ctx);
50
51 swrast->TextureSample[unit](ctx, unit, ctx->Texture.Unit[unit]._Current,
52 1, (const GLfloat (*)[4]) texcoord,
53 &lambda, &rgba);
54 color[0] = CHAN_TO_FLOAT(rgba[0]);
55 color[1] = CHAN_TO_FLOAT(rgba[1]);
56 color[2] = CHAN_TO_FLOAT(rgba[2]);
57 color[3] = CHAN_TO_FLOAT(rgba[3]);
58 }
59
60
61 /**
62 * Fetch a texel with the given partial derivatives to compute a level
63 * of detail in the mipmap.
64 */
65 static void
66 fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4],
67 const GLfloat texdx[4], const GLfloat texdy[4],
68 GLuint unit, GLfloat color[4] )
69 {
70 SWcontext *swrast = SWRAST_CONTEXT(ctx);
71 const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
72 const struct gl_texture_image *texImg = texObj->Image[texObj->BaseLevel];
73 const GLfloat texW = (GLfloat) texImg->WidthScale;
74 const GLfloat texH = (GLfloat) texImg->HeightScale;
75 GLchan rgba[4];
76
77 GLfloat lambda = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */
78 texdx[1], texdy[1], /* dt/dx, dt/dy */
79 texdx[3], texdy[2], /* dq/dx, dq/dy */
80 texW, texH,
81 texcoord[0], texcoord[1], texcoord[3],
82 1.0F / texcoord[3]);
83
84 swrast->TextureSample[unit](ctx, unit, ctx->Texture.Unit[unit]._Current,
85 1, (const GLfloat (*)[4]) texcoord,
86 &lambda, &rgba);
87 color[0] = CHAN_TO_FLOAT(rgba[0]);
88 color[1] = CHAN_TO_FLOAT(rgba[1]);
89 color[2] = CHAN_TO_FLOAT(rgba[2]);
90 color[3] = CHAN_TO_FLOAT(rgba[3]);
91 }
92
93
94 /**
95 * Return a pointer to the 4-element float vector specified by the given
96 * source register.
97 */
98 static INLINE const GLfloat *
99 get_register_pointer( GLcontext *ctx,
100 const struct fp_src_register *source,
101 const struct fp_machine *machine,
102 const struct fragment_program *program )
103 {
104 const GLfloat *src;
105 switch (source->File) {
106 case PROGRAM_TEMPORARY:
107 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_TEMPS);
108 src = machine->Temporaries[source->Index];
109 break;
110 case PROGRAM_INPUT:
111 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
112 src = machine->Inputs[source->Index];
113 break;
114 case PROGRAM_LOCAL_PARAM:
115 ASSERT(source->Index < MAX_PROGRAM_LOCAL_PARAMS);
116 src = program->Base.LocalParams[source->Index];
117 break;
118 case PROGRAM_ENV_PARAM:
119 ASSERT(source->Index < MAX_NV_FRAGMENT_PROGRAM_PARAMS);
120 src = ctx->FragmentProgram.Parameters[source->Index];
121 break;
122
123 case PROGRAM_STATE_VAR:
124 /* Fallthrough */
125
126 case PROGRAM_NAMED_PARAM:
127 ASSERT(source->Index < (GLint) program->Parameters->NumParameters);
128 src = program->Parameters->Parameters[source->Index].Values;
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 /* XXX: This should go someplace else, but it is safe here (and slow!)
576 * - karl
577 */
578 _mesa_load_state_parameters(ctx, program->Parameters);
579
580
581 for (pc = 0; pc < maxInst; pc++) {
582 const struct fp_instruction *inst = program->Instructions + pc;
583
584 if (ctx->FragmentProgram.CallbackEnabled &&
585 ctx->FragmentProgram.Callback) {
586 ctx->FragmentProgram.CurrentPosition = inst->StringPos;
587 ctx->FragmentProgram.Callback(program->Base.Target,
588 ctx->FragmentProgram.CallbackData);
589 }
590
591 switch (inst->Opcode) {
592 case FP_OPCODE_ABS:
593 {
594 GLfloat a[4], result[4];
595 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
596 result[0] = FABSF(a[0]);
597 result[1] = FABSF(a[1]);
598 result[2] = FABSF(a[2]);
599 result[3] = FABSF(a[3]);
600 store_vector4( inst, machine, result );
601 }
602 break;
603 case FP_OPCODE_ADD:
604 {
605 GLfloat a[4], b[4], result[4];
606 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
607 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
608 result[0] = a[0] + b[0];
609 result[1] = a[1] + b[1];
610 result[2] = a[2] + b[2];
611 result[3] = a[3] + b[3];
612 store_vector4( inst, machine, result );
613 }
614 break;
615 case FP_OPCODE_CMP:
616 {
617 GLfloat a[4], b[4], c[4], result[4];
618 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
619 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
620 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
621 result[0] = a[0] < 0.0F ? b[0] : c[0];
622 result[1] = a[1] < 0.0F ? b[1] : c[1];
623 result[2] = a[2] < 0.0F ? b[2] : c[2];
624 result[3] = a[3] < 0.0F ? b[3] : c[3];
625 store_vector4( inst, machine, result );
626 }
627 break;
628 case FP_OPCODE_COS:
629 {
630 GLfloat a[4], result[4];
631 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
632 result[0] = result[1] = result[2] = result[3] = (GLfloat)_mesa_cos(a[0]);
633 store_vector4( inst, machine, result );
634 }
635 break;
636 case FP_OPCODE_DDX: /* Partial derivative with respect to X */
637 {
638 GLfloat a[4], aNext[4], result[4];
639 struct fp_machine dMachine;
640 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', result)) {
641 /* This is tricky. Make a copy of the current machine state,
642 * increment the input registers by the dx or dy partial
643 * derivatives, then re-execute the program up to the
644 * preceeding instruction, then fetch the source register.
645 * Finally, find the difference in the register values for
646 * the original and derivative runs.
647 */
648 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
649 init_machine_deriv(ctx, machine, program, span,
650 'X', &dMachine);
651 execute_program(ctx, program, pc, &dMachine, span, column);
652 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
653 result[0] = aNext[0] - a[0];
654 result[1] = aNext[1] - a[1];
655 result[2] = aNext[2] - a[2];
656 result[3] = aNext[3] - a[3];
657 }
658 store_vector4( inst, machine, result );
659 }
660 break;
661 case FP_OPCODE_DDY: /* Partial derivative with respect to Y */
662 {
663 GLfloat a[4], aNext[4], result[4];
664 struct fp_machine dMachine;
665 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', result)) {
666 init_machine_deriv(ctx, machine, program, span,
667 'Y', &dMachine);
668 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
669 execute_program(ctx, program, pc, &dMachine, span, column);
670 fetch_vector4( ctx, &inst->SrcReg[0], &dMachine, program, aNext );
671 result[0] = aNext[0] - a[0];
672 result[1] = aNext[1] - a[1];
673 result[2] = aNext[2] - a[2];
674 result[3] = aNext[3] - a[3];
675 }
676 store_vector4( inst, machine, result );
677 }
678 break;
679 case FP_OPCODE_DP3:
680 {
681 GLfloat a[4], b[4], result[4];
682 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
683 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
684 result[0] = result[1] = result[2] = result[3] =
685 a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
686 store_vector4( inst, machine, result );
687 #if DEBUG_FRAG
688 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
689 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
690 #endif
691 }
692 break;
693 case FP_OPCODE_DP4:
694 {
695 GLfloat a[4], b[4], result[4];
696 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
697 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
698 result[0] = result[1] = result[2] = result[3] =
699 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
700 store_vector4( inst, machine, result );
701 }
702 break;
703 case FP_OPCODE_DPH:
704 {
705 GLfloat a[4], b[4], result[4];
706 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
707 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
708 result[0] = result[1] = result[2] = result[3] =
709 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + b[3];
710 store_vector4( inst, machine, result );
711 }
712 break;
713 case FP_OPCODE_DST: /* Distance vector */
714 {
715 GLfloat a[4], b[4], result[4];
716 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
717 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
718 result[0] = 1.0F;
719 result[1] = a[1] * b[1];
720 result[2] = a[2];
721 result[3] = b[3];
722 store_vector4( inst, machine, result );
723 }
724 break;
725 case FP_OPCODE_EX2: /* Exponential base 2 */
726 {
727 GLfloat a[4], result[4];
728 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
729 result[0] = result[1] = result[2] = result[3] =
730 (GLfloat) _mesa_pow(2.0, a[0]);
731 store_vector4( inst, machine, result );
732 }
733 break;
734 case FP_OPCODE_FLR:
735 {
736 GLfloat a[4], result[4];
737 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
738 result[0] = FLOORF(a[0]);
739 result[1] = FLOORF(a[1]);
740 result[2] = FLOORF(a[2]);
741 result[3] = FLOORF(a[3]);
742 store_vector4( inst, machine, result );
743 }
744 break;
745 case FP_OPCODE_FRC:
746 {
747 GLfloat a[4], result[4];
748 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
749 result[0] = a[0] - FLOORF(a[0]);
750 result[1] = a[1] - FLOORF(a[1]);
751 result[2] = a[2] - FLOORF(a[2]);
752 result[3] = a[3] - FLOORF(a[3]);
753 store_vector4( inst, machine, result );
754 }
755 break;
756 case FP_OPCODE_KIL:
757 {
758 const GLuint *swizzle = inst->DstReg.CondSwizzle;
759 const GLuint condMask = inst->DstReg.CondMask;
760 if (test_cc(machine->CondCodes[swizzle[0]], condMask) ||
761 test_cc(machine->CondCodes[swizzle[1]], condMask) ||
762 test_cc(machine->CondCodes[swizzle[2]], condMask) ||
763 test_cc(machine->CondCodes[swizzle[3]], condMask)) {
764 return GL_FALSE;
765 }
766 }
767 break;
768 case FP_OPCODE_LG2: /* log base 2 */
769 {
770 GLfloat a[4], result[4];
771 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
772 result[0] = result[1] = result[2] = result[3]
773 = LOG2(a[0]);
774 store_vector4( inst, machine, result );
775 }
776 break;
777 case FP_OPCODE_LIT:
778 {
779 GLfloat a[4], result[4];
780 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
781 if (a[0] < 0.0F)
782 a[0] = 0.0F;
783 if (a[1] < 0.0F)
784 a[1] = 0.0F;
785 result[0] = 1.0F;
786 result[1] = a[0];
787 result[2] = (a[0] > 0.0F) ? (GLfloat)_mesa_pow(2.0, a[3]) : 0.0F;
788 result[3] = 1.0F;
789 store_vector4( inst, machine, result );
790 }
791 break;
792 case FP_OPCODE_LRP:
793 {
794 GLfloat a[4], b[4], c[4], result[4];
795 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
796 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
797 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
798 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
799 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
800 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
801 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
802 store_vector4( inst, machine, result );
803 }
804 break;
805 case FP_OPCODE_MAD:
806 {
807 GLfloat a[4], b[4], c[4], result[4];
808 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
809 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
810 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
811 result[0] = a[0] * b[0] + c[0];
812 result[1] = a[1] * b[1] + c[1];
813 result[2] = a[2] * b[2] + c[2];
814 result[3] = a[3] * b[3] + c[3];
815 store_vector4( inst, machine, result );
816 }
817 break;
818 case FP_OPCODE_MAX:
819 {
820 GLfloat a[4], b[4], result[4];
821 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
822 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
823 result[0] = MAX2(a[0], b[0]);
824 result[1] = MAX2(a[1], b[1]);
825 result[2] = MAX2(a[2], b[2]);
826 result[3] = MAX2(a[3], b[3]);
827 store_vector4( inst, machine, result );
828 }
829 break;
830 case FP_OPCODE_MIN:
831 {
832 GLfloat a[4], b[4], result[4];
833 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
834 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
835 result[0] = MIN2(a[0], b[0]);
836 result[1] = MIN2(a[1], b[1]);
837 result[2] = MIN2(a[2], b[2]);
838 result[3] = MIN2(a[3], b[3]);
839 store_vector4( inst, machine, result );
840 }
841 break;
842 case FP_OPCODE_MOV:
843 {
844 GLfloat result[4];
845 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, result );
846 store_vector4( inst, machine, result );
847 }
848 break;
849 case FP_OPCODE_MUL:
850 {
851 GLfloat a[4], b[4], result[4];
852 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
853 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
854 result[0] = a[0] * b[0];
855 result[1] = a[1] * b[1];
856 result[2] = a[2] * b[2];
857 result[3] = a[3] * b[3];
858 store_vector4( inst, machine, result );
859 #if DEBUG_FRAG
860 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
861 result[0], result[1], result[2], result[3],
862 a[0], a[1], a[2], a[3],
863 b[0], b[1], b[2], b[3]);
864 #endif
865 }
866 break;
867 case FP_OPCODE_PK2H: /* pack two 16-bit floats */
868 /* XXX this is probably wrong */
869 {
870 GLfloat a[4], result[4];
871 const GLuint *rawBits = (const GLuint *) a;
872 GLuint *rawResult = (GLuint *) result;
873 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
874 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
875 = rawBits[0] | (rawBits[1] << 16);
876 store_vector4( inst, machine, result );
877 }
878 break;
879 case FP_OPCODE_PK2US: /* pack two GLushorts */
880 {
881 GLfloat a[4], result[4];
882 GLuint usx, usy, *rawResult = (GLuint *) result;
883 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
884 a[0] = CLAMP(a[0], 0.0F, 1.0F);
885 a[1] = CLAMP(a[0], 0.0F, 1.0F);
886 usx = IROUND(a[0] * 65535.0F);
887 usy = IROUND(a[1] * 65535.0F);
888 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
889 = usx | (usy << 16);
890 store_vector4( inst, machine, result );
891 }
892 break;
893 case FP_OPCODE_PK4B: /* pack four GLbytes */
894 {
895 GLfloat a[4], result[4];
896 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
897 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
898 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
899 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
900 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
901 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
902 ubx = IROUND(127.0F * a[0] + 128.0F);
903 uby = IROUND(127.0F * a[1] + 128.0F);
904 ubz = IROUND(127.0F * a[2] + 128.0F);
905 ubw = IROUND(127.0F * a[3] + 128.0F);
906 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
907 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
908 store_vector4( inst, machine, result );
909 }
910 break;
911 case FP_OPCODE_PK4UB: /* pack four GLubytes */
912 {
913 GLfloat a[4], result[4];
914 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
915 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
916 a[0] = CLAMP(a[0], 0.0F, 1.0F);
917 a[1] = CLAMP(a[1], 0.0F, 1.0F);
918 a[2] = CLAMP(a[2], 0.0F, 1.0F);
919 a[3] = CLAMP(a[3], 0.0F, 1.0F);
920 ubx = IROUND(255.0F * a[0]);
921 uby = IROUND(255.0F * a[1]);
922 ubz = IROUND(255.0F * a[2]);
923 ubw = IROUND(255.0F * a[3]);
924 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
925 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
926 store_vector4( inst, machine, result );
927 }
928 break;
929 case FP_OPCODE_POW:
930 {
931 GLfloat a[4], b[4], result[4];
932 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
933 fetch_vector1( ctx, &inst->SrcReg[1], machine, program, b );
934 result[0] = result[1] = result[2] = result[3]
935 = (GLfloat)_mesa_pow(a[0], b[0]);
936 store_vector4( inst, machine, result );
937 }
938 break;
939 case FP_OPCODE_RCP:
940 {
941 GLfloat a[4], result[4];
942 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
943 #if DEBUG_FRAG
944 if (a[0] == 0)
945 printf("RCP(0)\n");
946 else if (IS_INF_OR_NAN(a[0]))
947 printf("RCP(inf)\n");
948 #endif
949 result[0] = result[1] = result[2] = result[3]
950 = 1.0F / a[0];
951 store_vector4( inst, machine, result );
952 }
953 break;
954 case FP_OPCODE_RFL:
955 {
956 GLfloat axis[4], dir[4], result[4], tmp[4];
957 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, axis );
958 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dir );
959 tmp[3] = axis[0] * axis[0]
960 + axis[1] * axis[1]
961 + axis[2] * axis[2];
962 tmp[0] = (2.0F * (axis[0] * dir[0] +
963 axis[1] * dir[1] +
964 axis[2] * dir[2])) / tmp[3];
965 result[0] = tmp[0] * axis[0] - dir[0];
966 result[1] = tmp[0] * axis[1] - dir[1];
967 result[2] = tmp[0] * axis[2] - dir[2];
968 /* result[3] is never written! XXX enforce in parser! */
969 store_vector4( inst, machine, result );
970 }
971 break;
972 case FP_OPCODE_RSQ: /* 1 / sqrt() */
973 {
974 GLfloat a[4], result[4];
975 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
976 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
977 store_vector4( inst, machine, result );
978 #if DEBUG_FRAG
979 printf("RSQ %g = 1/sqrt(%g)\n", result[0], a[0]);
980 #endif
981 }
982 break;
983 case FP_OPCODE_SCS: /* sine and cos */
984 {
985 GLfloat a[4], result[4];
986 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
987 result[0] = (GLfloat)cos(a[0]);
988 result[1] = (GLfloat)sin(a[0]);
989 result[2] = 0.0; /* undefined! */
990 result[3] = 0.0; /* undefined! */
991 store_vector4( inst, machine, result );
992 }
993 break;
994 case FP_OPCODE_SEQ: /* set on equal */
995 {
996 GLfloat a[4], b[4], result[4];
997 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
998 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
999 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
1000 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
1001 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
1002 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
1003 store_vector4( inst, machine, result );
1004 }
1005 break;
1006 case FP_OPCODE_SFL: /* set false, operands ignored */
1007 {
1008 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
1009 store_vector4( inst, machine, result );
1010 }
1011 break;
1012 case FP_OPCODE_SGE: /* set on greater or equal */
1013 {
1014 GLfloat a[4], b[4], result[4];
1015 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1016 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1017 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
1018 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
1019 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
1020 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
1021 store_vector4( inst, machine, result );
1022 }
1023 break;
1024 case FP_OPCODE_SGT: /* set on greater */
1025 {
1026 GLfloat a[4], b[4], result[4];
1027 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1028 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1029 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
1030 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
1031 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
1032 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
1033 store_vector4( inst, machine, result );
1034 }
1035 break;
1036 case FP_OPCODE_SIN:
1037 {
1038 GLfloat a[4], result[4];
1039 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1040 result[0] = result[1] = result[2] =
1041 result[3] = (GLfloat)_mesa_sin(a[0]);
1042 store_vector4( inst, machine, result );
1043 }
1044 break;
1045 case FP_OPCODE_SLE: /* set on less or equal */
1046 {
1047 GLfloat a[4], b[4], result[4];
1048 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1049 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1050 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
1051 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
1052 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
1053 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
1054 store_vector4( inst, machine, result );
1055 }
1056 break;
1057 case FP_OPCODE_SLT: /* set on less */
1058 {
1059 GLfloat a[4], b[4], result[4];
1060 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1061 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1062 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
1063 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
1064 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
1065 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
1066 store_vector4( inst, machine, result );
1067 }
1068 break;
1069 case FP_OPCODE_SNE: /* set on not equal */
1070 {
1071 GLfloat a[4], b[4], result[4];
1072 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1073 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1074 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
1075 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
1076 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
1077 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
1078 store_vector4( inst, machine, result );
1079 }
1080 break;
1081 case FP_OPCODE_STR: /* set true, operands ignored */
1082 {
1083 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
1084 store_vector4( inst, machine, result );
1085 }
1086 break;
1087 case FP_OPCODE_SUB:
1088 {
1089 GLfloat a[4], b[4], result[4];
1090 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1091 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1092 result[0] = a[0] - b[0];
1093 result[1] = a[1] - b[1];
1094 result[2] = a[2] - b[2];
1095 result[3] = a[3] - b[3];
1096 store_vector4( inst, machine, result );
1097 }
1098 break;
1099 case FP_OPCODE_SWZ:
1100 {
1101 const struct fp_src_register *source = &inst->SrcReg[0];
1102 const GLfloat *src = get_register_pointer(ctx, source,
1103 machine, program);
1104 GLfloat result[4];
1105 GLuint i;
1106
1107 /* do extended swizzling here */
1108 for (i = 0; i < 3; i++) {
1109 if (source->Swizzle[i] == SWIZZLE_ZERO)
1110 result[i] = 0.0;
1111 else if (source->Swizzle[i] == SWIZZLE_ONE)
1112 result[i] = -1.0;
1113 else
1114 result[i] = -src[source->Swizzle[i]];
1115 if (source->NegateBase)
1116 result[i] = -result[i];
1117 }
1118 store_vector4( inst, machine, result );
1119 }
1120 break;
1121 case FP_OPCODE_TEX:
1122 /* Texel lookup */
1123 {
1124 GLfloat texcoord[4], color[4];
1125 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1126 /* XXX: Undo perspective divide from interpolate_texcoords() */
1127 fetch_texel( ctx, texcoord,
1128 span->array->lambda[inst->TexSrcUnit][column],
1129 inst->TexSrcUnit, color );
1130 store_vector4( inst, machine, color );
1131 }
1132 break;
1133 case FP_OPCODE_TXB:
1134 /* Texel lookup with LOD bias */
1135 {
1136 GLfloat texcoord[4], color[4], bias, lambda;
1137
1138 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1139 /* texcoord[3] is the bias to add to lambda */
1140 bias = ctx->Texture.Unit[inst->TexSrcUnit].LodBias
1141 + ctx->Texture.Unit[inst->TexSrcUnit]._Current->LodBias
1142 + texcoord[3];
1143 lambda = span->array->lambda[inst->TexSrcUnit][column] + bias;
1144 fetch_texel( ctx, texcoord, lambda,
1145 inst->TexSrcUnit, color );
1146 store_vector4( inst, machine, color );
1147 }
1148 break;
1149 case FP_OPCODE_TXD:
1150 /* Texture lookup w/ partial derivatives for LOD */
1151 {
1152 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1153 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1154 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, dtdx );
1155 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, dtdy );
1156 fetch_texel_deriv( ctx, texcoord, dtdx, dtdy, inst->TexSrcUnit,
1157 color );
1158 store_vector4( inst, machine, color );
1159 }
1160 break;
1161 case FP_OPCODE_TXP:
1162 /* Texture lookup w/ perspective divide */
1163 {
1164 GLfloat texcoord[4], color[4];
1165 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
1166 /* Already did perspective divide in interpolate_texcoords() */
1167 fetch_texel( ctx, texcoord,
1168 span->array->lambda[inst->TexSrcUnit][column],
1169 inst->TexSrcUnit, color );
1170 store_vector4( inst, machine, color );
1171 }
1172 break;
1173 case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
1174 /* XXX this is probably wrong */
1175 {
1176 GLfloat a[4], result[4];
1177 const GLuint *rawBits = (const GLuint *) a;
1178 GLuint *rawResult = (GLuint *) result;
1179 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1180 rawResult[0] = rawBits[0] & 0xffff;
1181 rawResult[1] = (rawBits[0] >> 16) & 0xffff;
1182 rawResult[2] = rawBits[0] & 0xffff;
1183 rawResult[3] = (rawBits[0] >> 16) & 0xffff;
1184 store_vector4( inst, machine, result );
1185 }
1186 break;
1187 case FP_OPCODE_UP2US: /* unpack two GLushorts */
1188 {
1189 GLfloat a[4], result[4];
1190 const GLuint *rawBits = (const GLuint *) a;
1191 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1192 result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
1193 result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
1194 result[2] = result[0];
1195 result[3] = result[1];
1196 store_vector4( inst, machine, result );
1197 }
1198 break;
1199 case FP_OPCODE_UP4B: /* unpack four GLbytes */
1200 {
1201 GLfloat a[4], result[4];
1202 const GLuint *rawBits = (const GLuint *) a;
1203 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1204 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1205 result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1206 result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1207 result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1208 store_vector4( inst, machine, result );
1209 }
1210 break;
1211 case FP_OPCODE_UP4UB: /* unpack four GLubytes */
1212 {
1213 GLfloat a[4], result[4];
1214 const GLuint *rawBits = (const GLuint *) a;
1215 fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
1216 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1217 result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1218 result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1219 result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1220 store_vector4( inst, machine, result );
1221 }
1222 break;
1223 case FP_OPCODE_X2D: /* 2-D matrix transform */
1224 {
1225 GLfloat a[4], b[4], c[4], result[4];
1226 fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
1227 fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
1228 fetch_vector4( ctx, &inst->SrcReg[2], machine, program, c );
1229 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1230 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1231 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1232 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1233 store_vector4( inst, machine, result );
1234 }
1235 break;
1236 case FP_OPCODE_END:
1237 return GL_TRUE;
1238 default:
1239 _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program",
1240 inst->Opcode);
1241 return GL_TRUE; /* return value doesn't matter */
1242 }
1243 }
1244 return GL_TRUE;
1245 }
1246
1247
1248 static void
1249 init_machine( GLcontext *ctx, struct fp_machine *machine,
1250 const struct fragment_program *program,
1251 const struct sw_span *span, GLuint col )
1252 {
1253 GLuint inputsRead = program->InputsRead;
1254 GLuint u;
1255
1256 if (ctx->FragmentProgram.CallbackEnabled)
1257 inputsRead = ~0;
1258
1259 /* Clear temporary registers */
1260 _mesa_bzero(machine->Temporaries,
1261 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
1262
1263 /* Load input registers */
1264 if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) {
1265 GLfloat *wpos = machine->Inputs[FRAG_ATTRIB_WPOS];
1266 wpos[0] = (GLfloat) span->x + col;
1267 wpos[1] = (GLfloat) span->y;
1268 wpos[2] = (GLfloat) span->array->z[col] / ctx->DepthMaxF;
1269 wpos[3] = span->w + col * span->dwdx;
1270 }
1271 if (inputsRead & (1 << FRAG_ATTRIB_COL0)) {
1272 GLfloat *col0 = machine->Inputs[FRAG_ATTRIB_COL0];
1273 col0[0] = CHAN_TO_FLOAT(span->array->rgba[col][RCOMP]);
1274 col0[1] = CHAN_TO_FLOAT(span->array->rgba[col][GCOMP]);
1275 col0[2] = CHAN_TO_FLOAT(span->array->rgba[col][BCOMP]);
1276 col0[3] = CHAN_TO_FLOAT(span->array->rgba[col][ACOMP]);
1277 }
1278 if (inputsRead & (1 << FRAG_ATTRIB_COL1)) {
1279 GLfloat *col1 = machine->Inputs[FRAG_ATTRIB_COL1];
1280 col1[0] = CHAN_TO_FLOAT(span->array->spec[col][RCOMP]);
1281 col1[1] = CHAN_TO_FLOAT(span->array->spec[col][GCOMP]);
1282 col1[2] = CHAN_TO_FLOAT(span->array->spec[col][BCOMP]);
1283 col1[3] = CHAN_TO_FLOAT(span->array->spec[col][ACOMP]);
1284 }
1285 if (inputsRead & (1 << FRAG_ATTRIB_FOGC)) {
1286 GLfloat *fogc = machine->Inputs[FRAG_ATTRIB_FOGC];
1287 fogc[0] = span->array->fog[col];
1288 fogc[1] = 0.0F;
1289 fogc[2] = 0.0F;
1290 fogc[3] = 0.0F;
1291 }
1292 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
1293 if (inputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
1294 GLfloat *tex = machine->Inputs[FRAG_ATTRIB_TEX0 + u];
1295 /*ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));*/
1296 COPY_4V(tex, span->array->texcoords[u][col]);
1297 /*ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);*/
1298 }
1299 }
1300
1301 /* init condition codes */
1302 machine->CondCodes[0] = COND_EQ;
1303 machine->CondCodes[1] = COND_EQ;
1304 machine->CondCodes[2] = COND_EQ;
1305 machine->CondCodes[3] = COND_EQ;
1306 }
1307
1308
1309 void
1310 _swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
1311 {
1312 const struct fragment_program *program = ctx->FragmentProgram.Current;
1313 GLuint i;
1314
1315 ctx->_CurrentProgram = GL_FRAGMENT_PROGRAM_ARB; /* or NV, doesn't matter */
1316
1317 for (i = 0; i < span->end; i++) {
1318 if (span->array->mask[i]) {
1319 init_machine(ctx, &ctx->FragmentProgram.Machine,
1320 ctx->FragmentProgram.Current, span, i);
1321
1322 if (!execute_program(ctx, program, ~0,
1323 &ctx->FragmentProgram.Machine, span, i)) {
1324 span->array->mask[i] = GL_FALSE; /* killed fragment */
1325 }
1326
1327 /* Store output registers */
1328 {
1329 const GLfloat *colOut
1330 = ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_COLR];
1331 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
1332 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
1333 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
1334 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
1335 }
1336 /* depth value */
1337 if (program->OutputsWritten & (1 << FRAG_OUTPUT_DEPR))
1338 span->array->z[i] = IROUND(ctx->FragmentProgram.Machine.Outputs[FRAG_OUTPUT_DEPR][0] * ctx->DepthMaxF);
1339 }
1340 }
1341
1342 ctx->_CurrentProgram = 0;
1343 }
1344