projects / mesa.git / blob
? search:


c728aa3d00611951e387cc2ed9f9adaa2bafa5e0
[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 /* init condition codes */
499 dMachine->CondCodes[0] = COND_EQ;
500 dMachine->CondCodes[1] = COND_EQ;
501 dMachine->CondCodes[2] = COND_EQ;
502 dMachine->CondCodes[3] = COND_EQ;
503 }
504
505
506 /**
507 * Execute the given vertex program.
508 * NOTE: we do everything in single-precision floating point; we don't
509 * currently observe the single/half/fixed-precision qualifiers.
510 * \param ctx - rendering context
511 * \param program - the fragment program to execute
512 * \param machine - machine state (register file)
513 * \param maxInst - max number of instructions to execute
514 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
515 */
516 static GLboolean
517 execute_program( GLcontext *ctx,
518 const struct fragment_program *program, GLuint maxInst,
519 struct fp_machine *machine, const struct sw_span *span,
520 GLuint column )
521 {
522 GLuint pc;
523
524 #if DEBUG_FRAG
525 printf("execute fragment program --------------------\n");
526 #endif
527
528 for (pc = 0; pc < maxInst; pc++) {
529 const struct fp_instruction *inst = program->Instructions + pc;
530 switch (inst->Opcode) {
531 case FP_OPCODE_ADD:
532 {
533 GLfloat a[4], b[4], result[4];
534 fetch_vector4( &inst->SrcReg[0], machine, program, a );
535 fetch_vector4( &inst->SrcReg[1], machine, program, b );
536 result[0] = a[0] + b[0];
537 result[1] = a[1] + b[1];
538 result[2] = a[2] + b[2];
539 result[3] = a[3] + b[3];
540 store_vector4( inst, machine, result );
541 }
542 break;
543 case FP_OPCODE_COS:
544 {
545 GLfloat a[4], result[4];
546 fetch_vector1( &inst->SrcReg[0], machine, program, a );
547 result[0] = result[1] = result[2] = result[3] = _mesa_cos(a[0]);
548 store_vector4( inst, machine, result );
549 }
550 break;
551 case FP_OPCODE_DDX: /* Partial derivative with respect to X */
552 {
553 GLfloat a[4], aNext[4], result[4];
554 struct fp_machine dMachine;
555 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', result)) {
556 /* This is tricky. Make a copy of the current machine state,
557 * increment the input registers by the dx or dy partial
558 * derivatives, then re-execute the program up to the
559 * preceeding instruction, then fetch the source register.
560 * Finally, find the difference in the register values for
561 * the original and derivative runs.
562 */
563 fetch_vector4( &inst->SrcReg[0], machine, program, a);
564 init_machine_deriv(ctx, machine, program, span,
565 'X', &dMachine);
566 execute_program(ctx, program, pc, &dMachine, span, column);
567 fetch_vector4( &inst->SrcReg[0], &dMachine, program, aNext );
568 result[0] = aNext[0] - a[0];
569 result[1] = aNext[1] - a[1];
570 result[2] = aNext[2] - a[2];
571 result[3] = aNext[3] - a[3];
572 }
573 store_vector4( inst, machine, result );
574 }
575 break;
576 case FP_OPCODE_DDY: /* Partial derivative with respect to Y */
577 {
578 GLfloat a[4], aNext[4], result[4];
579 struct fp_machine dMachine;
580 if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', result)) {
581 init_machine_deriv(ctx, machine, program, span,
582 'Y', &dMachine);
583 fetch_vector4( &inst->SrcReg[0], machine, program, a);
584 execute_program(ctx, program, pc, &dMachine, span, column);
585 fetch_vector4( &inst->SrcReg[0], &dMachine, program, aNext );
586 result[0] = aNext[0] - a[0];
587 result[1] = aNext[1] - a[1];
588 result[2] = aNext[2] - a[2];
589 result[3] = aNext[3] - a[3];
590 }
591 store_vector4( inst, machine, result );
592 }
593 break;
594 case FP_OPCODE_DP3:
595 {
596 GLfloat a[4], b[4], result[4];
597 fetch_vector4( &inst->SrcReg[0], machine, program, a );
598 fetch_vector4( &inst->SrcReg[1], machine, program, b );
599 result[0] = result[1] = result[2] = result[3] =
600 a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
601 store_vector4( inst, machine, result );
602 #if DEBUG_FRAG
603 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
604 result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
605 #endif
606 }
607 break;
608 case FP_OPCODE_DP4:
609 {
610 GLfloat a[4], b[4], result[4];
611 fetch_vector4( &inst->SrcReg[0], machine, program, a );
612 fetch_vector4( &inst->SrcReg[1], machine, program, b );
613 result[0] = result[1] = result[2] = result[3] =
614 a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
615 store_vector4( inst, machine, result );
616 }
617 break;
618 case FP_OPCODE_DST: /* Distance vector */
619 {
620 GLfloat a[4], b[4], result[4];
621 fetch_vector4( &inst->SrcReg[0], machine, program, a );
622 fetch_vector4( &inst->SrcReg[1], machine, program, b );
623 result[0] = 1.0F;
624 result[1] = a[1] * b[1];
625 result[2] = a[2];
626 result[3] = b[3];
627 store_vector4( inst, machine, result );
628 }
629 break;
630 case FP_OPCODE_EX2: /* Exponential base 2 */
631 {
632 GLfloat a[4], result[4];
633 fetch_vector1( &inst->SrcReg[0], machine, program, a );
634 result[0] = result[1] = result[2] = result[3] =
635 (GLfloat) _mesa_pow(2.0, a[0]);
636 store_vector4( inst, machine, result );
637 }
638 break;
639 case FP_OPCODE_FLR:
640 {
641 GLfloat a[4], result[4];
642 fetch_vector4( &inst->SrcReg[0], machine, program, a );
643 result[0] = FLOORF(a[0]);
644 result[1] = FLOORF(a[1]);
645 result[2] = FLOORF(a[2]);
646 result[3] = FLOORF(a[3]);
647 store_vector4( inst, machine, result );
648 }
649 break;
650 case FP_OPCODE_FRC:
651 {
652 GLfloat a[4], result[4];
653 fetch_vector4( &inst->SrcReg[0], machine, program, a );
654 result[0] = a[0] - FLOORF(a[0]);
655 result[1] = a[1] - FLOORF(a[1]);
656 result[2] = a[2] - FLOORF(a[2]);
657 result[3] = a[3] - FLOORF(a[3]);
658 store_vector4( inst, machine, result );
659 }
660 break;
661 case FP_OPCODE_KIL:
662 {
663 const GLuint *swizzle = inst->DstReg.CondSwizzle;
664 const GLuint condMask = inst->DstReg.CondMask;
665 if (test_cc(machine->CondCodes[swizzle[0]], condMask) ||
666 test_cc(machine->CondCodes[swizzle[1]], condMask) ||
667 test_cc(machine->CondCodes[swizzle[2]], condMask) ||
668 test_cc(machine->CondCodes[swizzle[3]], condMask)) {
669 return GL_FALSE;
670 }
671 }
672 break;
673 case FP_OPCODE_LG2: /* log base 2 */
674 {
675 GLfloat a[4], result[4];
676 fetch_vector1( &inst->SrcReg[0], machine, program, a );
677 result[0] = result[1] = result[2] = result[3]
678 = LOG2(a[0]);
679 store_vector4( inst, machine, result );
680 }
681 break;
682 case FP_OPCODE_LIT:
683 {
684 GLfloat a[4], result[4];
685 fetch_vector4( &inst->SrcReg[0], machine, program, a );
686 if (a[0] < 0.0F)
687 a[0] = 0.0F;
688 if (a[1] < 0.0F)
689 a[1] = 0.0F;
690 result[0] = 1.0F;
691 result[1] = a[0];
692 result[2] = (a[0] > 0.0) ? _mesa_pow(2.0, a[3]) : 0.0F;
693 result[3] = 1.0F;
694 store_vector4( inst, machine, result );
695 }
696 break;
697 case FP_OPCODE_LRP:
698 {
699 GLfloat a[4], b[4], c[4], result[4];
700 fetch_vector4( &inst->SrcReg[0], machine, program, a );
701 fetch_vector4( &inst->SrcReg[1], machine, program, b );
702 fetch_vector4( &inst->SrcReg[2], machine, program, c );
703 result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
704 result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
705 result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
706 result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
707 store_vector4( inst, machine, result );
708 }
709 break;
710 case FP_OPCODE_MAD:
711 {
712 GLfloat a[4], b[4], c[4], result[4];
713 fetch_vector4( &inst->SrcReg[0], machine, program, a );
714 fetch_vector4( &inst->SrcReg[1], machine, program, b );
715 fetch_vector4( &inst->SrcReg[2], machine, program, c );
716 result[0] = a[0] * b[0] + c[0];
717 result[1] = a[1] * b[1] + c[1];
718 result[2] = a[2] * b[2] + c[2];
719 result[3] = a[3] * b[3] + c[3];
720 store_vector4( inst, machine, result );
721 }
722 break;
723 case FP_OPCODE_MAX:
724 {
725 GLfloat a[4], b[4], result[4];
726 fetch_vector4( &inst->SrcReg[0], machine, program, a );
727 fetch_vector4( &inst->SrcReg[1], machine, program, b );
728 result[0] = MAX2(a[0], b[0]);
729 result[1] = MAX2(a[1], b[1]);
730 result[2] = MAX2(a[2], b[2]);
731 result[3] = MAX2(a[3], b[3]);
732 store_vector4( inst, machine, result );
733 }
734 break;
735 case FP_OPCODE_MIN:
736 {
737 GLfloat a[4], b[4], result[4];
738 fetch_vector4( &inst->SrcReg[0], machine, program, a );
739 fetch_vector4( &inst->SrcReg[1], machine, program, b );
740 result[0] = MIN2(a[0], b[0]);
741 result[1] = MIN2(a[1], b[1]);
742 result[2] = MIN2(a[2], b[2]);
743 result[3] = MIN2(a[3], b[3]);
744 store_vector4( inst, machine, result );
745 }
746 break;
747 case FP_OPCODE_MOV:
748 {
749 GLfloat result[4];
750 fetch_vector4( &inst->SrcReg[0], machine, program, result );
751 store_vector4( inst, machine, result );
752 }
753 break;
754 case FP_OPCODE_MUL:
755 {
756 GLfloat a[4], b[4], result[4];
757 fetch_vector4( &inst->SrcReg[0], machine, program, a );
758 fetch_vector4( &inst->SrcReg[1], machine, program, b );
759 result[0] = a[0] * b[0];
760 result[1] = a[1] * b[1];
761 result[2] = a[2] * b[2];
762 result[3] = a[3] * b[3];
763 store_vector4( inst, machine, result );
764 #if DEBUG_FRAG
765 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
766 result[0], result[1], result[2], result[3],
767 a[0], a[1], a[2], a[3],
768 b[0], b[1], b[2], b[3]);
769 #endif
770 }
771 break;
772 case FP_OPCODE_PK2H: /* pack two 16-bit floats */
773 /* XXX this is probably wrong */
774 {
775 GLfloat a[4], result[4];
776 const GLuint *rawBits = (const GLuint *) a;
777 GLuint *rawResult = (GLuint *) result;
778 fetch_vector4( &inst->SrcReg[0], machine, program, a );
779 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
780 = rawBits[0] | (rawBits[1] << 16);
781 store_vector4( inst, machine, result );
782 }
783 break;
784 case FP_OPCODE_PK2US: /* pack two GLushorts */
785 {
786 GLfloat a[4], result[4];
787 GLuint usx, usy, *rawResult = (GLuint *) result;
788 fetch_vector4( &inst->SrcReg[0], machine, program, a );
789 a[0] = CLAMP(a[0], 0.0F, 1.0F);
790 a[1] = CLAMP(a[0], 0.0F, 1.0F);
791 usx = IROUND(a[0] * 65535.0F);
792 usy = IROUND(a[1] * 65535.0F);
793 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
794 = usx | (usy << 16);
795 store_vector4( inst, machine, result );
796 }
797 break;
798 case FP_OPCODE_PK4B: /* pack four GLbytes */
799 {
800 GLfloat a[4], result[4];
801 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
802 fetch_vector4( &inst->SrcReg[0], machine, program, a );
803 a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
804 a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
805 a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
806 a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
807 ubx = IROUND(127.0F * a[0] + 128.0F);
808 uby = IROUND(127.0F * a[1] + 128.0F);
809 ubz = IROUND(127.0F * a[2] + 128.0F);
810 ubw = IROUND(127.0F * a[3] + 128.0F);
811 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
812 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
813 store_vector4( inst, machine, result );
814 }
815 break;
816 case FP_OPCODE_PK4UB: /* pack four GLubytes */
817 {
818 GLfloat a[4], result[4];
819 GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
820 fetch_vector4( &inst->SrcReg[0], machine, program, a );
821 a[0] = CLAMP(a[0], 0.0F, 1.0F);
822 a[1] = CLAMP(a[1], 0.0F, 1.0F);
823 a[2] = CLAMP(a[2], 0.0F, 1.0F);
824 a[3] = CLAMP(a[3], 0.0F, 1.0F);
825 ubx = IROUND(255.0F * a[0]);
826 uby = IROUND(255.0F * a[1]);
827 ubz = IROUND(255.0F * a[2]);
828 ubw = IROUND(255.0F * a[3]);
829 rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
830 = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
831 store_vector4( inst, machine, result );
832 }
833 break;
834 case FP_OPCODE_POW:
835 {
836 GLfloat a[4], b[4], result[4];
837 fetch_vector1( &inst->SrcReg[0], machine, program, a );
838 fetch_vector1( &inst->SrcReg[1], machine, program, b );
839 result[0] = result[1] = result[2] = result[3]
840 = _mesa_pow(a[0], b[0]);
841 store_vector4( inst, machine, result );
842 }
843 break;
844 case FP_OPCODE_RCP:
845 {
846 GLfloat a[4], result[4];
847 fetch_vector1( &inst->SrcReg[0], machine, program, a );
848 #if DEBUG_FRAG
849 if (a[0] == 0)
850 printf("RCP(0)\n");
851 else if (IS_INF_OR_NAN(a[0]))
852 printf("RCP(inf)\n");
853 #endif
854 result[0] = result[1] = result[2] = result[3]
855 = 1.0F / a[0];
856 store_vector4( inst, machine, result );
857 }
858 break;
859 case FP_OPCODE_RFL:
860 {
861 GLfloat axis[4], dir[4], result[4], tmp[4];
862 fetch_vector4( &inst->SrcReg[0], machine, program, axis );
863 fetch_vector4( &inst->SrcReg[1], machine, program, dir );
864 tmp[3] = axis[0] * axis[0]
865 + axis[1] * axis[1]
866 + axis[2] * axis[2];
867 tmp[0] = (2.0F * (axis[0] * dir[0] +
868 axis[1] * dir[1] +
869 axis[2] * dir[2])) / tmp[3];
870 result[0] = tmp[0] * axis[0] - dir[0];
871 result[1] = tmp[0] * axis[1] - dir[1];
872 result[2] = tmp[0] * axis[2] - dir[2];
873 /* result[3] is never written! XXX enforce in parser! */
874 store_vector4( inst, machine, result );
875 }
876 break;
877 case FP_OPCODE_RSQ: /* 1 / sqrt() */
878 {
879 GLfloat a[4], result[4];
880 fetch_vector1( &inst->SrcReg[0], machine, program, a );
881 result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
882 store_vector4( inst, machine, result );
883 #if DEBUG_FRAG
884 printf("RSQ %g = 1/sqrt(%g)\n", result[0], a[0]);
885 #endif
886 }
887 break;
888 case FP_OPCODE_SEQ: /* set on equal */
889 {
890 GLfloat a[4], b[4], result[4];
891 fetch_vector4( &inst->SrcReg[0], machine, program, a );
892 fetch_vector4( &inst->SrcReg[1], machine, program, b );
893 result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
894 result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
895 result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
896 result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
897 store_vector4( inst, machine, result );
898 }
899 break;
900 case FP_OPCODE_SFL: /* set false, operands ignored */
901 {
902 static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
903 store_vector4( inst, machine, result );
904 }
905 break;
906 case FP_OPCODE_SGE: /* set on greater or equal */
907 {
908 GLfloat a[4], b[4], result[4];
909 fetch_vector4( &inst->SrcReg[0], machine, program, a );
910 fetch_vector4( &inst->SrcReg[1], machine, program, b );
911 result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
912 result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
913 result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
914 result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
915 store_vector4( inst, machine, result );
916 }
917 break;
918 case FP_OPCODE_SGT: /* set on greater */
919 {
920 GLfloat a[4], b[4], result[4];
921 fetch_vector4( &inst->SrcReg[0], machine, program, a );
922 fetch_vector4( &inst->SrcReg[1], machine, program, b );
923 result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
924 result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
925 result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
926 result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
927 store_vector4( inst, machine, result );
928 }
929 break;
930 case FP_OPCODE_SIN:
931 {
932 GLfloat a[4], result[4];
933 fetch_vector1( &inst->SrcReg[0], machine, program, a );
934 result[0] = result[1] = result[2] = result[3] = _mesa_sin(a[0]);
935 store_vector4( inst, machine, result );
936 }
937 break;
938 case FP_OPCODE_SLE: /* set on less or equal */
939 {
940 GLfloat a[4], b[4], result[4];
941 fetch_vector4( &inst->SrcReg[0], machine, program, a );
942 fetch_vector4( &inst->SrcReg[1], machine, program, b );
943 result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
944 result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
945 result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
946 result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
947 store_vector4( inst, machine, result );
948 }
949 break;
950 case FP_OPCODE_SLT: /* set on less */
951 {
952 GLfloat a[4], b[4], result[4];
953 fetch_vector4( &inst->SrcReg[0], machine, program, a );
954 fetch_vector4( &inst->SrcReg[1], machine, program, b );
955 result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
956 result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
957 result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
958 result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
959 store_vector4( inst, machine, result );
960 }
961 break;
962 case FP_OPCODE_SNE: /* set on not equal */
963 {
964 GLfloat a[4], b[4], result[4];
965 fetch_vector4( &inst->SrcReg[0], machine, program, a );
966 fetch_vector4( &inst->SrcReg[1], machine, program, b );
967 result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
968 result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
969 result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
970 result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
971 store_vector4( inst, machine, result );
972 }
973 break;
974 case FP_OPCODE_STR: /* set true, operands ignored */
975 {
976 static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
977 store_vector4( inst, machine, result );
978 }
979 break;
980 case FP_OPCODE_SUB:
981 {
982 GLfloat a[4], b[4], result[4];
983 fetch_vector4( &inst->SrcReg[0], machine, program, a );
984 fetch_vector4( &inst->SrcReg[1], machine, program, b );
985 result[0] = a[0] - b[0];
986 result[1] = a[1] - b[1];
987 result[2] = a[2] - b[2];
988 result[3] = a[3] - b[3];
989 store_vector4( inst, machine, result );
990 }
991 break;
992 case FP_OPCODE_TEX:
993 /* Texel lookup */
994 {
995 GLfloat texcoord[4], color[4];
996 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
997 /* XXX: Undo perspective divide from interpolate_texcoords() */
998 fetch_texel( ctx, texcoord,
999 span->array->lambda[inst->TexSrcUnit][column],
1000 inst->TexSrcUnit, color );
1001 store_vector4( inst, machine, color );
1002 }
1003 break;
1004 case FP_OPCODE_TXD:
1005 /* Texture lookup w/ partial derivatives for LOD */
1006 {
1007 GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
1008 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
1009 fetch_vector4( &inst->SrcReg[1], machine, program, dtdx );
1010 fetch_vector4( &inst->SrcReg[2], machine, program, dtdy );
1011 fetch_texel_deriv( ctx, texcoord, dtdx, dtdy, inst->TexSrcUnit,
1012 color );
1013 store_vector4( inst, machine, color );
1014 }
1015 break;
1016 case FP_OPCODE_TXP:
1017 /* Texture lookup w/ perspective divide */
1018 {
1019 GLfloat texcoord[4], color[4];
1020 fetch_vector4( &inst->SrcReg[0], machine, program, texcoord );
1021 /* Already did perspective divide in interpolate_texcoords() */
1022 fetch_texel( ctx, texcoord,
1023 span->array->lambda[inst->TexSrcUnit][column],
1024 inst->TexSrcUnit, color );
1025 store_vector4( inst, machine, color );
1026 }
1027 break;
1028 case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
1029 /* XXX this is probably wrong */
1030 {
1031 GLfloat a[4], result[4];
1032 const GLuint *rawBits = (const GLuint *) a;
1033 GLuint *rawResult = (GLuint *) result;
1034 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1035 rawResult[0] = rawBits[0] & 0xffff;
1036 rawResult[1] = (rawBits[0] >> 16) & 0xffff;
1037 rawResult[2] = rawBits[0] & 0xffff;
1038 rawResult[3] = (rawBits[0] >> 16) & 0xffff;
1039 store_vector4( inst, machine, result );
1040 }
1041 break;
1042 case FP_OPCODE_UP2US: /* unpack two GLushorts */
1043 {
1044 GLfloat a[4], result[4];
1045 const GLuint *rawBits = (const GLuint *) a;
1046 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1047 result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
1048 result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
1049 result[2] = result[0];
1050 result[3] = result[1];
1051 store_vector4( inst, machine, result );
1052 }
1053 break;
1054 case FP_OPCODE_UP4B: /* unpack four GLbytes */
1055 {
1056 GLfloat a[4], result[4];
1057 const GLuint *rawBits = (const GLuint *) a;
1058 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1059 result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
1060 result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
1061 result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
1062 result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
1063 store_vector4( inst, machine, result );
1064 }
1065 break;
1066 case FP_OPCODE_UP4UB: /* unpack four GLubytes */
1067 {
1068 GLfloat a[4], result[4];
1069 const GLuint *rawBits = (const GLuint *) a;
1070 fetch_vector1( &inst->SrcReg[0], machine, program, a );
1071 result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
1072 result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
1073 result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
1074 result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
1075 store_vector4( inst, machine, result );
1076 }
1077 break;
1078 case FP_OPCODE_X2D: /* 2-D matrix transform */
1079 {
1080 GLfloat a[4], b[4], c[4], result[4];
1081 fetch_vector4( &inst->SrcReg[0], machine, program, a );
1082 fetch_vector4( &inst->SrcReg[1], machine, program, b );
1083 fetch_vector4( &inst->SrcReg[2], machine, program, c );
1084 result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
1085 result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
1086 result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
1087 result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
1088 store_vector4( inst, machine, result );
1089 }
1090 break;
1091 case FP_OPCODE_END:
1092 return GL_TRUE;
1093 default:
1094 _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program",
1095 inst->Opcode);
1096 return GL_TRUE; /* return value doesn't matter */
1097 }
1098 }
1099 return GL_TRUE;
1100 }
1101
1102
1103 static void
1104 init_machine( GLcontext *ctx, struct fp_machine *machine,
1105 const struct fragment_program *program,
1106 const struct sw_span *span, GLuint col )
1107 {
1108 GLuint j, u;
1109
1110 /* Clear temporary registers */
1111 _mesa_bzero(machine->Registers + FP_TEMP_REG_START,
1112 MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
1113
1114 /* Load program local parameters */
1115 for (j = 0; j < MAX_NV_FRAGMENT_PROGRAM_PARAMS; j++) {
1116 COPY_4V(machine->Registers[FP_PROG_REG_START + j],
1117 program->Base.LocalParams[j]);
1118 }
1119
1120 /* Load input registers */
1121 if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
1122 GLfloat *wpos = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_WPOS];
1123 wpos[0] = span->x + col;
1124 wpos[1] = span->y;
1125 wpos[2] = (GLfloat) span->array->z[col] / ctx->DepthMaxF;
1126 wpos[3] = span->w + col * span->dwdx;
1127 }
1128 if (program->InputsRead & (1 << FRAG_ATTRIB_COL0)) {
1129 GLfloat *col0 = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL0];
1130 col0[0] = CHAN_TO_FLOAT(span->array->rgba[col][RCOMP]);
1131 col0[1] = CHAN_TO_FLOAT(span->array->rgba[col][GCOMP]);
1132 col0[2] = CHAN_TO_FLOAT(span->array->rgba[col][BCOMP]);
1133 col0[3] = CHAN_TO_FLOAT(span->array->rgba[col][ACOMP]);
1134 }
1135 if (program->InputsRead & (1 << FRAG_ATTRIB_COL1)) {
1136 GLfloat *col1 = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_COL1];
1137 col1[0] = CHAN_TO_FLOAT(span->array->spec[col][RCOMP]);
1138 col1[1] = CHAN_TO_FLOAT(span->array->spec[col][GCOMP]);
1139 col1[2] = CHAN_TO_FLOAT(span->array->spec[col][BCOMP]);
1140 col1[3] = CHAN_TO_FLOAT(span->array->spec[col][ACOMP]);
1141 }
1142 if (program->InputsRead & (1 << FRAG_ATTRIB_FOGC)) {
1143 GLfloat *fogc = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_FOGC];
1144 fogc[0] = span->array->fog[col];
1145 fogc[1] = 0.0F;
1146 fogc[2] = 0.0F;
1147 fogc[3] = 0.0F;
1148 }
1149 for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
1150 if (program->InputsRead & (1 << (FRAG_ATTRIB_TEX0 + u))) {
1151 GLfloat *tex = machine->Registers[FP_INPUT_REG_START+FRAG_ATTRIB_TEX0+u];
1152 ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));
1153 COPY_4V(tex, span->array->texcoords[u][col]);
1154 ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);
1155 }
1156 }
1157
1158 /* init condition codes */
1159 machine->CondCodes[0] = COND_EQ;
1160 machine->CondCodes[1] = COND_EQ;
1161 machine->CondCodes[2] = COND_EQ;
1162 machine->CondCodes[3] = COND_EQ;
1163 }
1164
1165
1166 void
1167 _swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
1168 {
1169 const struct fragment_program *program = ctx->FragmentProgram.Current;
1170 GLuint i;
1171
1172 for (i = 0; i < span->end; i++) {
1173 if (span->array->mask[i]) {
1174 init_machine(ctx, &ctx->FragmentProgram.Machine,
1175 ctx->FragmentProgram.Current, span, i);
1176
1177 if (!execute_program(ctx, program, ~0,
1178 &ctx->FragmentProgram.Machine, span, i)) {
1179 span->array->mask[i] = GL_FALSE; /* killed fragment */
1180 }
1181
1182 /* Store output registers */
1183 {
1184 const GLfloat *colOut
1185 = ctx->FragmentProgram.Machine.Registers[FP_OUTPUT_REG_START];
1186 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][RCOMP], colOut[0]);
1187 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][GCOMP], colOut[1]);
1188 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][BCOMP], colOut[2]);
1189 UNCLAMPED_FLOAT_TO_CHAN(span->array->rgba[i][ACOMP], colOut[3]);
1190 }
1191 /* depth value */
1192 if (program->OutputsWritten & (1 << FRAG_OUTPUT_DEPR))
1193 span->array->z[i] = IROUND(ctx->FragmentProgram.Machine.Registers[FP_OUTPUT_REG_START + 2][0] * ctx->DepthMaxF);
1194 }
1195 }
1196 }
1197