projects / mesa.git / blob
? search:


59d7c0f826295cf6de3fff7a32d1371de80234b8
[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
32 #include "s_nvfragprog.h"
33 #include "s_span.h"
34 #include "s_texture.h"
35
36
37 /* if 1, print some debugging info */
38 #define DEBUG_FRAG 0
39
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 /**
96 * Fetch a 4-element float vector from the given source register.
97 * Apply swizzling and negating as needed.
98 */
99 static void
100 fetch_vector4( const struct fp_src_register *source,
101 const struct fp_machine *machine,
102 const struct fragment_program *program,
103 GLfloat result[4] )
104 {
105 const GLfloat *src;
106
107 if (source->IsParameter) {
108 src = program->Parameters[source->Register].Values;
109 }
110 else {
111 src = machine->Registers[source->Register];
112 }
113
114 result[0] = src[source->Swizzle[0]];
115 result[1] = src[source->Swizzle[1]];
116 result[2] = src[source->Swizzle[2]];
117 result[3] = src[source->Swizzle[3]];
118
119 if (source->NegateBase) {
120 result[0] = -result[0];
121 result[1] = -result[1];
122 result[2] = -result[2];
123 result[3] = -result[3];
124 }
125 if (source->Abs) {
126 result[0] = FABSF(result[0]);
127 result[1] = FABSF(result[1]);
128 result[2] = FABSF(result[2]);
129 result[3] = FABSF(result[3]);
130 }
131 if (source->NegateAbs) {
132 result[0] = -result[0];
133 result[1] = -result[1];
134 result[2] = -result[2];
135 result[3] = -result[3];
136 }
137 }
138
139
140 /**
141 * Fetch the derivative with respect to X for the given register.
142 * \return GL_TRUE if it was easily computed or GL_FALSE if we
143 * need to execute another instance of the program (ugh)!
144 */
145 static GLboolean
146 fetch_vector4_deriv( const struct fp_src_register *source,
147 const struct sw_span *span,
148 char xOrY, GLfloat result[4] )
149 {
150 GLfloat src[4];
151
152 ASSERT(xOrY == 'X' || xOrY == 'Y');
153
154 switch (source->Register) {
155 case FRAG_ATTRIB_WPOS:
156 if (xOrY == 'X') {
157 src[0] = 1.0;
158 src[1] = 0.0;
159 src[2] = span->dzdx;
160 src[3] = span->dwdx;
161 }
162 else {
163 src[0] = 0.0;
164 src[1] = 1.0;
165 src[2] = span->dzdy;
166 src[3] = span->dwdy;
167 }
168 break;
169 case FRAG_ATTRIB_COL0:
170 if (xOrY == 'X') {
171 src[0] = span->drdx * (1.0F / CHAN_MAXF);
172 src[1] = span->dgdx * (1.0F / CHAN_MAXF);
173 src[2] = span->dbdx * (1.0F / CHAN_MAXF);
174 src[3] = span->dadx * (1.0F / CHAN_MAXF);
175 }
176 else {
177 src[0] = span->drdy * (1.0F / CHAN_MAXF);
178 src[1] = span->dgdy * (1.0F / CHAN_MAXF);
179 src[2] = span->dbdy * (1.0F / CHAN_MAXF);
180 src[3] = span->dady * (1.0F / CHAN_MAXF);
181 }
182 break;
183 case FRAG_ATTRIB_COL1:
184 if (xOrY == 'X') {
185 src[0] = span->dsrdx * (1.0F / CHAN_MAXF);
186 src[1] = span->dsgdx * (1.0F / CHAN_MAXF);
187 src[2] = span->dsbdx * (1.0F / CHAN_MAXF);
188 src[3] = 0.0; /* XXX need this */
189 }
190 else {
191 src[0] = span->dsrdy * (1.0F / CHAN_MAXF);
192 src[1] = span->dsgdy * (1.0F / CHAN_MAXF);
193 src[2] = span->dsbdy * (1.0F / CHAN_MAXF);
194 src[3] = 0.0; /* XXX need this */
195 }
196 break;
197 case FRAG_ATTRIB_FOGC:
198 if (xOrY == 'X') {
199 src[0] = span->dfogdx;
200 src[1] = 0.0;
201 src[2] = 0.0;
202 src[3] = 0.0;
203 }
204 else {
205 src[0] = span->dfogdy;
206 src[1] = 0.0;
207 src[2] = 0.0;
208 src[3] = 0.0;
209 }
210 break;
211 case FRAG_ATTRIB_TEX0:
212 case FRAG_ATTRIB_TEX1:
213 case FRAG_ATTRIB_TEX2:
214 case FRAG_ATTRIB_TEX3:
215 case FRAG_ATTRIB_TEX4:
216 case FRAG_ATTRIB_TEX5:
217 case FRAG_ATTRIB_TEX6:
218 case FRAG_ATTRIB_TEX7:
219 if (xOrY == 'X') {
220 const GLuint u = source->Register - FRAG_ATTRIB_TEX0;
221 src[0] = span->texStepX[u][0] * (1.0F / CHAN_MAXF);
222 src[1] = span->texStepX[u][1] * (1.0F / CHAN_MAXF);
223 src[2] = span->texStepX[u][2] * (1.0F / CHAN_MAXF);
224 src[3] = span->texStepX[u][3] * (1.0F / CHAN_MAXF);
225 }
226 else {
227 const GLuint u = source->Register - FRAG_ATTRIB_TEX0;
228 src[0] = span->texStepY[u][0] * (1.0F / CHAN_MAXF);
229 src[1] = span->texStepY[u][1] * (1.0F / CHAN_MAXF);
230 src[2] = span->texStepY[u][2] * (1.0F / CHAN_MAXF);
231 src[3] = span->texStepY[u][3] * (1.0F / CHAN_MAXF);
232 }
233 break;
234 default:
235 return GL_FALSE;
236 }
237
238 result[0] = src[source->Swizzle[0]];
239 result[1] = src[source->Swizzle[1]];
240 result[2] = src[source->Swizzle[2]];
241 result[3] = src[source->Swizzle[3]];
242
243 if (source->NegateBase) {
244 result[0] = -result[0];
245 result[1] = -result[1];
246 result[2] = -result[2];
247 result[3] = -result[3];
248 }
249 if (source->Abs) {
250 result[0] = FABSF(result[0]);
251 result[1] = FABSF(result[1]);
252 result[2] = FABSF(result[2]);
253 result[3] = FABSF(result[3]);
254 }
255 if (source->NegateAbs) {
256 result[0] = -result[0];
257 result[1] = -result[1];
258 result[2] = -result[2];
259 result[3] = -result[3];
260 }
261 return GL_TRUE;
262 }
263
264
265 /**
266 * As above, but only return result[0] element.
267 */
268 static void
269 fetch_vector1( const struct fp_src_register *source,
270 const struct fp_machine *machine,
271 const struct fragment_program *program,
272 GLfloat result[4] )
273 {
274 const GLfloat *src;
275
276 if (source->IsParameter) {
277 src = program->Parameters[source->Register].Values;
278 }
279 else {
280 src = machine->Registers[source->Register];
281 }
282
283 result[0] = src[source->Swizzle[0]];
284
285 if (source->NegateBase) {
286 result[0] = -result[0];
287 }
288 if (source->Abs) {
289 result[0] = FABSF(result[0]);
290 }
291 if (source->NegateAbs) {
292 result[0] = -result[0];
293 }
294 }
295
296
297 /*
298 * Test value against zero and return GT, LT, EQ or UN if NaN.
299 */
300 static INLINE GLuint
301 generate_cc( float value )
302 {
303 if (value != value)
304 return COND_UN; /* NaN */
305 if (value > 0.0F)
306 return COND_GT;
307 if (value < 0.0F)
308 return COND_LT;
309 return COND_EQ;
310 }
311
312 /*
313 * Test if the ccMaskRule is satisfied by the given condition code.
314 * Used to mask destination writes according to the current condition codee.
315 */
316 static INLINE GLboolean
317 test_cc(GLuint condCode, GLuint ccMaskRule)
318 {
319 switch (ccMaskRule) {
320 case COND_EQ: return (condCode == COND_EQ);
321 case COND_NE: return (condCode != COND_EQ);
322 case COND_LT: return (condCode == COND_LT);
323 case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
324 case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
325 case COND_GT: return (condCode == COND_GT);
326 case COND_TR: return GL_TRUE;
327 case COND_FL: return GL_FALSE;
328 default: return GL_TRUE;
329 }
330 }
331
332
333 /**
334 * Store 4 floats into a register. Observe the instructions saturate and
335 * set-condition-code flags.
336 */
337 static void
338 store_vector4( const struct fp_instruction *inst,
339 struct fp_machine *machine,
340 const GLfloat value[4] )
341 {
342 const struct fp_dst_register *dest = &(inst->DstReg);
343 const GLboolean clamp = inst->Saturate;
344 const GLboolean updateCC = inst->UpdateCondRegister;
345 GLfloat *dstReg = machine->Registers[dest->Register];
346 GLfloat clampedValue[4];
347 const GLboolean *writeMask = dest->WriteMask;
348 GLboolean condWriteMask[4];
349
350 #if DEBUG_FRAG
351 if (value[0] > 1.0e10 ||
352 IS_INF_OR_NAN(value[0]) ||
353 IS_INF_OR_NAN(value[1]) ||
354 IS_INF_OR_NAN(value[2]) ||
355 IS_INF_OR_NAN(value[3]) )
356 printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
357 #endif
358
359 if (clamp) {
360 clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
361 clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
362 clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
363 clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
364 value = clampedValue;
365 }
366
367 if (dest->CondMask != COND_TR) {
368 condWriteMask[0] = writeMask[0]
369 && test_cc(machine->CondCodes[dest->CondSwizzle[0]], dest->CondMask);
370 condWriteMask[1] = writeMask[1]
371 && test_cc(machine->CondCodes[dest->CondSwizzle[1]], dest->CondMask);
372 condWriteMask[2] = writeMask[2]
373 && test_cc(machine->CondCodes[dest->CondSwizzle[2]], dest->CondMask);
374 condWriteMask[3] = writeMask[3]
375 && test_cc(machine->CondCodes[dest->CondSwizzle[3]], dest->CondMask);
376 writeMask = condWriteMask;
377 }
378
379 if (writeMask[0]) {
380 dstReg[0] = value[0];
381 if (updateCC)
382 machine->CondCodes[0] = generate_cc(value[0]);
383 }
384 if (writeMask[1]) {
385 dstReg[1] = value[1];
386 if (updateCC)
387 machine->CondCodes[1] = generate_cc(value[1]);
388 }
389 if (writeMask[2]) {
390 dstReg[2] = value[2];
391 if (updateCC)
392 machine->CondCodes[2] = generate_cc(value[2]);
393 }
394 if (writeMask[3]) {
395 dstReg[3] = value[3];
396 if (updateCC)
397 machine->CondCodes[3] = generate_cc(value[3]);
398 }
399 }
400
401
402 /**
403 * Initialize a new machine state instance from an existing one, adding
404 * the partial derivatives onto the input registers.
405 * Used to implement DDX and DDY instructions in non-trivial cases.
406 */
407 static void
408 init_machine_deriv( GLcontext *ctx,
409 const struct fp_machine *machine,
410 const struct fragment_program *program,
411 const struct sw_span *span, char xOrY,
412 struct fp_machine *dMachine )
413 {
414 GLuint u;
415
416 ASSERT(xOrY == 'X' || xOrY == 'Y');
417
418 /* copy existing machine */
419 _mesa_memcpy(dMachine, machine, sizeof(struct fp_machine));
420
421 /* Clear temporary registers */
422 _mesa_bzero((GLfloat*) (machine->Registers + FP_TEMP_REG_START) ,
423 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
424
425 /* Add derivatives */
426 if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
427 GLfloat *wpos = (GLfloat*) machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_WPOS];
428 if (xOrY == 'X') {
429 wpos[0] += 1.0F;
430 wpos[1] += 0.0F;
431 wpos[2] += span->dzdx;
432 wpos[3] += span->dwdx;
433 }
434 else {
435 wpos[0] += 0.0F;
436 wpos[1] += 1.0F;
437 wpos[2] += span->dzdy;
438 wpos[3] += span->dwdy;
439 }
440 }
441 if (program->InputsRead & (1 << FRAG_ATTRIB_COL0)) {
442 GLfloat *col0 = (GLfloat*) machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL0];
443 if (xOrY == 'X') {
444 col0[0] += span->drdx * (1.0F / CHAN_MAXF);
445 col0[1] += span->dgdx * (1.0F / CHAN_MAXF);
446 col0[2] += span->dbdx * (1.0F / CHAN_MAXF);
447 col0[3] += span->dadx * (1.0F / CHAN_MAXF);
448 }
449 else {
450 col0[0] += span->drdy * (1.0F / CHAN_MAXF);
451 col0[1] += span->dgdy * (1.0F / CHAN_MAXF);
452 col0[2] += span->dbdy * (1.0F / CHAN_MAXF);
453 col0[3] += span->dady * (1.0F / CHAN_MAXF);
454 }
455 }
456 if (program->InputsRead & (1 << FRAG_ATTRIB_COL1)) {
457 GLfloat *col1 = (GLfloat*) machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL1];
458 if (xOrY == 'X') {
459 col1[0] += span->dsrdx * (1.0F / CHAN_MAXF);
460 col1[1] += span->dsgdx * (1.0F / CHAN_MAXF);
461 col1[2] += span->dsbdx * (1.0F / CHAN_MAXF);
462 col1[3] += 0.0; /*XXX fix */
463 }
464 else {
465 col1[0] += span->dsrdy * (1.0F / CHAN_MAXF);
466 col1[1] += span->dsgdy * (1.0F / CHAN_MAXF);
467 col1[2] += span->dsbdy * (1.0F / CHAN_MAXF);
468 col1[3] += 0.0; /*XXX fix */
469 }
470 }
471 if (program->InputsRead & (1 << FRAG_ATTRIB_FOGC)) {
472 GLfloat *fogc = (GLfloat*) machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_FOGC];
473 if (xOrY == 'X') {
474 fogc[0] += span->dfogdx;
475 }
476 else {
477 fogc[0] += span->dfogdy;
478 }
479 }
480 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
481 if (program->InputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
482 GLfloat *tex = (GLfloat*) machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_TEX0+u];
483 if (xOrY == 'X') {
484 tex[0] += span->texStepX[u][0];
485 tex[1] += span->texStepX[u][1];
486 tex[2] += span->texStepX[u][2];
487 tex[3] += span->texStepX[u][3];
488 }
489 else {
490 tex[0] += span->texStepY[u][0];
491 tex[1] += span->texStepY[u][1];
492 tex[2] += span->texStepY[u][2];
493 tex[3] += span->texStepY[u][3];
494 }
495 }
496 }
497 }
498
499
500 /**
501 * Execute the given vertex program.
502 * NOTE: we do everything in single-precision floating point; we don't
503 * currently observe the single/half/fixed-precision qualifiers.
504 * \param ctx - rendering context
505 * \param program - the fragment program to execute
506 * \param machine - machine state (register file)
507 * \param maxInst - max number of instructions to execute
508 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
509 */
510 static GLboolean
511 execute_program( GLcontext *ctx,
512 const struct fragment_program *program, GLuint maxInst,
513 struct fp_machine *machine, const struct sw_span *span,
514 GLuint column )
515 {
516 GLuint pc;
517
518 #if DEBUG_FRAG
519 printf("execute fragment program --------------------\n");
520 #endif
521
522 for (pc = 0; pc < maxInst; pc++) {
523 const struct fp_instruction *inst = program->Instructions + pc;
524 switch (inst->Opcode) {
525 case FP_OPCODE_ADD:
526 {
527 GLfloat a[4], b[4], result[4];
528 fetch_vector4( &inst->SrcReg[0], machine, program, a );
529 fetch_vector4( &inst->SrcReg[1], machine, program, b );
530 result[0] = a[0] + b[0];
531 result[1] = a[1] + b[1];
532 result[2] = a[2] + b[2];
533 result[3] = a[3] + b[3];
534 store_vector4( inst, machine, result );
535 }
536 break;
537 case FP_OPCODE_COS:
538 {
539 GLfloat a[4], result[4];
540 fetch_vector1( &inst->SrcReg[0], machine, program, a );
541 result[0] = result[1] = result[2] = result[3] = _mesa_cos(a[0]);
542 store_vector4( inst, machine, result );
543 }
544 break;
545 case FP_OPCODE_DDX: /* Partial derivative with respect to X */
546 {
547 GLfloat a[4], aNext[4], result[4];
548 struct fp_machine dMachine;
549 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', result)) {
550 /* This is tricky. Make a copy of the current machine state,
551 * increment the input registers by the dx or dy partial
552 * derivatives, then re-execute the program up to the
553 * preceeding instruction, then fetch the source register.
554 * Finally, find the difference in the register values for
555 * the original and derivative runs.
556 */
557 fetch_vector4( &inst->SrcReg[0], machine, program, a);
558 init_machine_deriv(ctx, machine, program, span,
559 'X', &dMachine);
560 execute_program(ctx, program, pc, &dMachine, span, column);
561 fetch_vector4( &inst->SrcReg[0], &dMachine, program, aNext );
562 result[0] = aNext[0] - a[0];
563 result[1] = aNext[1] - a[1];
564 result[2] = aNext[2] - a[2];
565 result[3] = aNext[3] - a[3];
566 }
567 store_vector4( inst, machine, result );
568 }
569 break;
570 case FP_OPCODE_DDY: /* Partial derivative with respect to Y */
571 {
572 GLfloat a[4], aNext[4], result[4];
573 struct fp_machine dMachine;
574 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', result)) {
575 init_machine_deriv(ctx, machine, program, span,
576 'Y', &dMachine);
577 fetch_vector4( &inst->SrcReg[0], machine, program, a);
578 execute_program(ctx, program, pc, &dMachine, span, column);
579 fetch_vector4( &inst->SrcReg[0], &dMachine, program, aNext );
580 result[0] = aNext[0] - a[0];
581 result[1] = aNext[1] - a[1];
582 result[2] = aNext[2] - a[2];
583 result[3] = aNext[3] - a[3];
584 }
585 store_vector4( inst, machine, result );
586 }
587 break;
588 case FP_OPCODE_DP3:
589 {
590 GLfloat a[4], b[4], result[4];
591 fetch_vector4( &inst->SrcReg[0], machine, program, a );
592 fetch_vector4( &inst->SrcReg[1], machine, program, b );
593 result[0] = result[1] = result[2] = result[3] =
594 a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
595 store_vector4( inst, machine, result );
596 #if DEBUG_FRAG
597 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
598 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
599 #endif
600 }
601 break;
602 case FP_OPCODE_DP4:
603 {
604 GLfloat a[4], b[4], result[4];
605 fetch_vector4( &inst->SrcReg[0], machine, program, a );
606 fetch_vector4( &inst->SrcReg[1], machine, program, b );
607 result[0] = result[1] = result[2] = result[3] =
608 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
609 store_vector4( inst, machine, result );
610 }
611 break;
612 case FP_OPCODE_DST: /* Distance vector */
613 {
614 GLfloat a[4], b[4], result[4];
615 fetch_vector4( &inst->SrcReg[0], machine, program, a );
616 fetch_vector4( &inst->SrcReg[1], machine, program, b );
617 result[0] = 1.0F;
618 result[1] = a[1] * b[1];
619 result[2] = a[2];
620 result[3] = b[3];
621 store_vector4( inst, machine, result );
622 }
623 break;
624 case FP_OPCODE_EX2: /* Exponential base 2 */
625 {
626 GLfloat a[4], result[4];
627 fetch_vector1( &inst->SrcReg[0], machine, program, a );
628 result[0] = result[1] = result[2] = result[3] =
629 (GLfloat) _mesa_pow(2.0, a[0]);
630 store_vector4( inst, machine, result );
631 }
632 break;
633 case FP_OPCODE_FLR:
634 {
635 GLfloat a[4], result[4];
636 fetch_vector4( &inst->SrcReg[0], machine, program, a );
637 result[0] = FLOORF(a[0]);
638 result[1] = FLOORF(a[1]);
639 result[2] = FLOORF(a[2]);
640 result[3] = FLOORF(a[3]);
641 store_vector4( inst, machine, result );
642 }
643 break;
644 case FP_OPCODE_FRC:
645 {
646 GLfloat a[4], result[4];
647 fetch_vector4( &inst->SrcReg[0], machine, program, a );
648 result[0] = a[0] - FLOORF(a[0]);
649 result[1] = a[1] - FLOORF(a[1]);
650 result[2] = a[2] - FLOORF(a[2]);
651 result[3] = a[3] - FLOORF(a[3]);
652 store_vector4( inst, machine, result );
653 }
654 break;
655 case FP_OPCODE_KIL:
656 {
657 const GLuint *swizzle = inst->DstReg.CondSwizzle;
658 const GLuint condMask = inst->DstReg.CondMask;
659 if (test_cc(machine->CondCodes[swizzle[0]], condMask) ||
660 test_cc(machine->CondCodes[swizzle[1]], condMask) ||
661 test_cc(machine->CondCodes[swizzle[2]], condMask) ||
662 test_cc(machine->CondCodes[swizzle[3]], condMask))
663 return GL_FALSE;
664 }
665 break;
666 case FP_OPCODE_LG2: /* log base 2 */
667 {
668 GLfloat a[4], result[4];
669 fetch_vector1( &inst->SrcReg[0], machine, program, a );
670 result[0] = result[1] = result[2] = result[3]
671 = LOG2(a[0]);
672 store_vector4( inst, machine, result );
673 }
674 break;
675 case FP_OPCODE_LIT:
676 {
677 GLfloat a[4], result[4];
678 fetch_vector4( &inst->SrcReg[0], machine, program, a );
679 if (a[0] < 0.0F)
680 a[0] = 0.0F;
681 if (a[1] < 0.0F)
682 a[1] = 0.0F;
683 result[0] = 1.0F;
684 result[1] = a[0];
685 result[2] = (a[0] > 0.0) ? _mesa_pow(2.0, a[3]) : 0.0F;
686 result[3] = 1.0F;
687 store_vector4( inst, machine, result );
688 }
689 break;
690 case FP_OPCODE_LRP:
691 {
692 GLfloat a[4], b[4], c[4], result[4];
693 fetch_vector4( &inst->SrcReg[0], machine, program, a );
694 fetch_vector4( &inst->SrcReg[1], machine, program, b );
695 fetch_vector4( &inst->SrcReg[2], machine, program, c );
696 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
697 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
698 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
699 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
700 store_vector4( inst, machine, result );
701 }
702 break;
703 case FP_OPCODE_MAD:
704 {
705 GLfloat a[4], b[4], c[4], result[4];
706 fetch_vector4( &inst->SrcReg[0], machine, program, a );
707 fetch_vector4( &inst->SrcReg[1], machine, program, b );
708 fetch_vector4( &inst->SrcReg[2], machine, program, c );
709 result[0] = a[0] * b[0] + c[0];
710 result[1] = a[1] * b[1] + c[1];
711 result[2] = a[2] * b[2] + c[2];
712 result[3] = a[3] * b[3] + c[3];
713 store_vector4( inst, machine, result );
714 }
715 break;
716 case FP_OPCODE_MAX:
717 {
718 GLfloat a[4], b[4], result[4];
719 fetch_vector4( &inst->SrcReg[0], machine, program, a );
720 fetch_vector4( &inst->SrcReg[1], machine, program, b );
721 result[0] = MAX2(a[0], b[0]);
722 result[1] = MAX2(a[1], b[1]);
723 result[2] = MAX2(a[2], b[2]);
724 result[3] = MAX2(a[3], b[3]);
725 store_vector4( inst, machine, result );
726 }
727 break;
728 case FP_OPCODE_MIN:
729 {
730 GLfloat a[4], b[4], result[4];
731 fetch_vector4( &inst->SrcReg[0], machine, program, a );
732 fetch_vector4( &inst->SrcReg[1], machine, program, b );
733 result[0] = MIN2(a[0], b[0]);
734 result[1] = MIN2(a[1], b[1]);
735 result[2] = MIN2(a[2], b[2]);
736 result[3] = MIN2(a[3], b[3]);
737 store_vector4( inst, machine, result );
738 }
739 break;
740 case FP_OPCODE_MOV:
741 {
742 GLfloat result[4];
743 fetch_vector4( &inst->SrcReg[0], machine, program, result );
744 store_vector4( inst, machine, result );
745 }
746 break;
747 case FP_OPCODE_MUL:
748 {
749 GLfloat a[4], b[4], result[4];
750 fetch_vector4( &inst->SrcReg[0], machine, program, a );
751 fetch_vector4( &inst->SrcReg[1], machine, program, b );
752 result[0] = a[0] * b[0];
753 result[1] = a[1] * b[1];
754 result[2] = a[2] * b[2];
755 result[3] = a[3] * b[3];
756 store_vector4( inst, machine, result );
757 #if DEBUG_FRAG
758 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
759 result[0], result[1], result[2], result[3],
760 a[0], a[1], a[2], a[3],
761 b[0], b[1], b[2], b[3]);
762 #endif
763 }
764 break;
765 case FP_OPCODE_PK2H: /* pack two 16-bit floats */
766 /* XXX this is probably wrong */
767 {
768 GLfloat a[4], result[4];
769 const GLuint *rawBits = (const GLuint *) a;
770 GLuint *rawResult = (GLuint *) result;
771 fetch_vector4( &inst->SrcReg[0], machine, program, a );
772 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
773 = rawBits[0] | (rawBits[1] << 16);
774 store_vector4( inst, machine, result );
775 }
776 break;
777 case FP_OPCODE_PK2US: /* pack two GLushorts */
778 {
779 GLfloat a[4], result[4];
780 GLuint usx, usy, *rawResult = (GLuint *) result;
781 fetch_vector4( &inst->SrcReg[0], machine, program, a );
782 a[0] = CLAMP(a[0], 0.0F, 1.0F);
783 a[1] = CLAMP(a[0], 0.0F, 1.0F);
784 usx = IROUND(a[0] * 65535.0F);
785 usy = IROUND(a[1] * 65535.0F);
786 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
787 = usx | (usy << 16);
788 store_vector4( inst, machine, result );
789 }
790 break;
791 case FP_OPCODE_PK4B: /* pack four GLbytes */
792 {
793 GLfloat a[4], result[4];
794 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
795 fetch_vector4( &inst->SrcReg[0], machine, program, a );
796 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
797 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
798 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
799 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
800 ubx = IROUND(127.0F * a[0] + 128.0F);
801 uby = IROUND(127.0F * a[1] + 128.0F);
802 ubz = IROUND(127.0F * a[2] + 128.0F);
803 ubw = IROUND(127.0F * a[3] + 128.0F);
804 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
805 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
806 store_vector4( inst, machine, result );
807 }
808 break;
809 case FP_OPCODE_PK4UB: /* pack four GLubytes */
810 {
811 GLfloat a[4], result[4];
812 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
813 fetch_vector4( &inst->SrcReg[0], machine, program, a );
814 a[0] = CLAMP(a[0], 0.0F, 1.0F);
815 a[1] = CLAMP(a[1], 0.0F, 1.0F);
816 a[2] = CLAMP(a[2], 0.0F, 1.0F);
817 a[3] = CLAMP(a[3], 0.0F, 1.0F);
818 ubx = IROUND(255.0F * a[0]);
819 uby = IROUND(255.0F * a[1]);
820 ubz = IROUND(255.0F * a[2]);
821 ubw = IROUND(255.0F * a[3]);
822 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
823 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
824 store_vector4( inst, machine, result );
825 }
826 break;
827 case FP_OPCODE_POW:
828 {
829 GLfloat a[4], b[4], result[4];
830 fetch_vector1( &inst->SrcReg[0], machine, program, a );
831 fetch_vector1( &inst->SrcReg[1], machine, program, b );
832 result[0] = result[1] = result[2] = result[3]
833 = _mesa_pow(a[0], b[0]);
834 store_vector4( inst, machine, result );
835 }
836 break;
837 case FP_OPCODE_RCP:
838 {
839 GLfloat a[4], result[4];
840 fetch_vector1( &inst->SrcReg[0], machine, program, a );
841 #if DEBUG_FRAG
842 if (a[0] == 0)
843 printf("RCP(0)\n");
844 else if (IS_INF_OR_NAN(a[0]))
845 printf("RCP(inf)\n");
846 #endif
847 result[0] = result[1] = result[2] = result[3]
848 = 1.0F / a[0];
849 store_vector4( inst, machine, result );
850 }
851 break;
852 case FP_OPCODE_RFL:
853 {
854 GLfloat axis[4], dir[4], result[4], tmp[4];
855 fetch_vector4( &inst->SrcReg[0], machine, program, axis );
856 fetch_vector4( &inst->SrcReg[1], machine, program, dir );
857 tmp[3] = axis[0] * axis[0]
858 + axis[1] * axis[1]
859 + axis[2] * axis[2];
860 tmp[0] = (2.0F * (axis[0] * dir[0] +
861 axis[1] * dir[1] +
862 axis[2] * dir[2])) / tmp[3];
863 result[0] = tmp[0] * axis[0] - dir[0];
864 result[1] = tmp[0] * axis[1] - dir[1];
865 result[2] = tmp[0] * axis[2] - dir[2];
866 /* result[3] is never written! XXX enforce in parser! */
867 store_vector4( inst, machine, result );
868 }
869 break;
870 case FP_OPCODE_RSQ: /* 1 / sqrt() */
871 {
872 GLfloat a[4], result[4];
873 fetch_vector1( &inst->SrcReg[0], machine, program, a );
874 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
875 store_vector4( inst, machine, result );
876 #if DEBUG_FRAG
877 printf("RSQ %g = 1/sqrt(%g)\n", result[0], a[0]);
878 #endif
879 }
880 break;
881 case FP_OPCODE_SEQ: /* set on equal */
882 {
883 GLfloat a[4], b[4], result[4];
884 fetch_vector4( &inst->SrcReg[0], machine, program, a );
885 fetch_vector4( &inst->SrcReg[1], machine, program, b );
886 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
887 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
888 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
889 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
890 store_vector4( inst, machine, result );
891 }
892 break;
893 case FP_OPCODE_SFL: /* set false, operands ignored */
894 {
895 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
896 store_vector4( inst, machine, result );
897 }
898 break;
899 case FP_OPCODE_SGE: /* set on greater or equal */
900 {
901 GLfloat a[4], b[4], result[4];
902 fetch_vector4( &inst->SrcReg[0], machine, program, a );
903 fetch_vector4( &inst->SrcReg[1], machine, program, b );
904 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
905 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
906 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
907 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
908 store_vector4( inst, machine, result );
909 }
910 break;
911 case FP_OPCODE_SGT: /* set on greater */
912 {
913 GLfloat a[4], b[4], result[4];
914 fetch_vector4( &inst->SrcReg[0], machine, program, a );
915 fetch_vector4( &inst->SrcReg[1], machine, program, b );
916 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
917 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
918 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
919 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
920 store_vector4( inst, machine, result );
921 }
922 break;
923 case FP_OPCODE_SIN:
924 {
925 GLfloat a[4], result[4];
926 fetch_vector1( &inst->SrcReg[0], machine, program, a );
927 result[0] = result[1] = result[2] = result[3] = _mesa_sin(a[0]);
928 store_vector4( inst, machine, result );
929 }
930 break;
931 case FP_OPCODE_SLE: /* set on less or equal */
932 {
933 GLfloat a[4], b[4], result[4];
934 fetch_vector4( &inst->SrcReg[0], machine, program, a );
935 fetch_vector4( &inst->SrcReg[1], machine, program, b );
936 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
937 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
938 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
939 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
940 store_vector4( inst, machine, result );
941 }
942 break;
943 case FP_OPCODE_SLT: /* set on less */
944 {
945 GLfloat a[4], b[4], result[4];
946 fetch_vector4( &inst->SrcReg[0], machine, program, a );
947 fetch_vector4( &inst->SrcReg[1], machine, program, b );
948 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
949 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
950 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
951 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
952 store_vector4( inst, machine, result );
953 }
954 break;
955 case FP_OPCODE_SNE: /* set on not equal */
956 {
957 GLfloat a[4], b[4], result[4];
958 fetch_vector4( &inst->SrcReg[0], machine, program, a );
959 fetch_vector4( &inst->SrcReg[1], machine, program, b );
960 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
961 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
962 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
963 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
964 store_vector4( inst, machine, result );
965 }
966 break;
967 case FP_OPCODE_STR: /* set true, operands ignored */
968 {
969 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
970 store_vector4( inst, machine, result );
971 }
972 break;
973 case FP_OPCODE_SUB:
974 {
975 GLfloat a[4], b[4], result[4];
976 fetch_vector4( &inst->SrcReg[0], machine, program, a );
977 fetch_vector4( &inst->SrcReg[1], machine, program, b );
978 result[0] = a[0] - b[0];
979 result[1] = a[1] - b[1];
980 result[2] = a[2] - b[2];
981 result[3] = a[3] - b[3];
982 store_vector4( inst, machine, result );
983 }
984 break;
985 case FP_OPCODE_TEX:
986 /* Texel lookup */
987 {
988 GLfloat texcoord[4], color[4];
989 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
990 /* XXX: Undo perspective divide from interpolate_texcoords() */
991 fetch_texel( ctx, texcoord,
992 span->array->lambda[inst->TexSrcUnit][column],
993 inst->TexSrcUnit, color );
994 store_vector4( inst, machine, color );
995 }
996 break;
997 case FP_OPCODE_TXD:
998 /* Texture lookup w/ partial derivatives for LOD */
999 {
1000 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1001 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
1002 fetch_vector4( &inst->SrcReg[1], machine, program, dtdx );
1003 fetch_vector4( &inst->SrcReg[2], machine, program, dtdy );
1004 fetch_texel_deriv( ctx, texcoord, dtdx, dtdy, inst->TexSrcUnit,
1005 color );
1006 store_vector4( inst, machine, color );
1007 }
1008 break;
1009 case FP_OPCODE_TXP:
1010 /* Texture lookup w/ perspective divide */
1011 {
1012 GLfloat texcoord[4], color[4];
1013 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
1014 /* Already did perspective divide in interpolate_texcoords() */
1015 fetch_texel( ctx, texcoord,
1016 span->array->lambda[inst->TexSrcUnit][column],
1017 inst->TexSrcUnit, color );
1018 store_vector4( inst, machine, color );
1019 }
1020 break;
1021 case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
1022 /* XXX this is probably wrong */
1023 {
1024 GLfloat a[4], result[4];
1025 const GLuint *rawBits = (const GLuint *) a;
1026 GLuint *rawResult = (GLuint *) result;
1027 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1028 rawResult[0] = rawBits[0] & 0xffff;
1029 rawResult[1] = (rawBits[0] >> 16) & 0xffff;
1030 rawResult[2] = rawBits[0] & 0xffff;
1031 rawResult[3] = (rawBits[0] >> 16) & 0xffff;
1032 store_vector4( inst, machine, result );
1033 }
1034 break;
1035 case FP_OPCODE_UP2US: /* unpack two GLushorts */
1036 {
1037 GLfloat a[4], result[4];
1038 const GLuint *rawBits = (const GLuint *) a;
1039 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1040 result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
1041 result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
1042 result[2] = result[0];
1043 result[3] = result[1];
1044 store_vector4( inst, machine, result );
1045 }
1046 break;
1047 case FP_OPCODE_UP4B: /* unpack four GLbytes */
1048 {
1049 GLfloat a[4], result[4];
1050 const GLuint *rawBits = (const GLuint *) a;
1051 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1052 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1053 result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1054 result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1055 result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1056 store_vector4( inst, machine, result );
1057 }
1058 break;
1059 case FP_OPCODE_UP4UB: /* unpack four GLubytes */
1060 {
1061 GLfloat a[4], result[4];
1062 const GLuint *rawBits = (const GLuint *) a;
1063 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1064 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1065 result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1066 result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1067 result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1068 store_vector4( inst, machine, result );
1069 }
1070 break;
1071 case FP_OPCODE_X2D: /* 2-D matrix transform */
1072 {
1073 GLfloat a[4], b[4], c[4], result[4];
1074 fetch_vector4( &inst->SrcReg[0], machine, program, a );
1075 fetch_vector4( &inst->SrcReg[1], machine, program, b );
1076 fetch_vector4( &inst->SrcReg[2], machine, program, c );
1077 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1078 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1079 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1080 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1081 store_vector4( inst, machine, result );
1082 }
1083 break;
1084 case FP_OPCODE_END:
1085 return GL_TRUE;
1086 default:
1087 _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program",
1088 inst->Opcode);
1089 return GL_TRUE; /* return value doesn't matter */
1090 }
1091 }
1092 return GL_TRUE;
1093 }
1094
1095
1096 static void
1097 init_machine( GLcontext *ctx, struct fp_machine *machine,
1098 const struct fragment_program *program,
1099 const struct sw_span *span, GLuint col )
1100 {
1101 GLuint j, u;
1102
1103 /* Clear temporary registers */
1104 _mesa_bzero(machine->Registers + FP_TEMP_REG_START,
1105 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
1106
1107 /* Load program local parameters */
1108 for (j = 0; j < MAX_NV_FRAGMENT_PROGRAM_PARAMS; j++) {
1109 COPY_4V(machine->Registers[FP_PROG_REG_START + j],
1110 program->Base.LocalParams[j]);
1111 }
1112
1113 /* Load input registers */
1114 if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
1115 GLfloat *wpos = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_WPOS];
1116 wpos[0] = span->x + col;
1117 wpos[1] = span->y;
1118 wpos[2] = (GLfloat) span->array->z[col] / ctx->DepthMaxF;
1119 wpos[3] = span->w + col * span->dwdx;
1120 }
1121 if (program->InputsRead & (1 << FRAG_ATTRIB_COL0)) {
1122 GLfloat *col0 = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL0];
1123 col0[0] = CHAN_TO_FLOAT(span->array->rgba[col][RCOMP]);
1124 col0[1] = CHAN_TO_FLOAT(span->array->rgba[col][GCOMP]);
1125 col0[2] = CHAN_TO_FLOAT(span->array->rgba[col][BCOMP]);
1126 col0[3] = CHAN_TO_FLOAT(span->array->rgba[col][ACOMP]);
1127 }
1128 if (program->InputsRead & (1 << FRAG_ATTRIB_COL1)) {
1129 GLfloat *col1 = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL1];
1130 col1[0] = CHAN_TO_FLOAT(span->array->spec[col][RCOMP]);
1131 col1[1] = CHAN_TO_FLOAT(span->array->spec[col][GCOMP]);
1132 col1[2] = CHAN_TO_FLOAT(span->array->spec[col][BCOMP]);
1133 col1[3] = CHAN_TO_FLOAT(span->array->spec[col][ACOMP]);
1134 }
1135 if (program->InputsRead & (1 << FRAG_ATTRIB_FOGC)) {
1136 GLfloat *fogc = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_FOGC];
1137 fogc[0] = span->array->fog[col];
1138 fogc[1] = 0.0F;
1139 fogc[2] = 0.0F;
1140 fogc[3] = 0.0F;
1141 }
1142 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
1143 if (program->InputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
1144 GLfloat *tex = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_TEX0+u];
1145 ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));
1146 COPY_4V(tex, span->array->texcoords[u][col]);
1147 ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);
1148 }
1149 }
1150 }
1151
1152
1153 void
1154 _swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
1155 {
1156 const struct fragment_program *program = ctx->FragmentProgram.Current;
1157 GLuint i;
1158
1159 for (i = 0; i < span->end; i++) {
1160 if (span->array->mask[i]) {
1161 init_machine(ctx, &ctx->FragmentProgram.Machine,
1162 ctx->FragmentProgram.Current, span, i);
1163
1164 if (!execute_program(ctx, program, ~0,
1165 &ctx->FragmentProgram.Machine, span, i))
1166 span->array->mask[i] = GL_FALSE; /* killed fragment */
1167
1168 /* Store output registers */
1169 {
1170 const GLfloat *colOut
1171 = ctx->FragmentProgram.Machine.Registers[FP_OUTPUT_REG_START];
1172 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
1173 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
1174 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
1175 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
1176 }
1177 /* depth value */
1178 if (ctx->FragmentProgram.Current->OutputsWritten & 2)
1179 span->array->z[i] = IROUND(ctx->FragmentProgram.Machine.Registers[FP_OUTPUT_REG_START + 2][0] * ctx->DepthMaxF);
1180 }
1181 }
1182 }
1183