projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
r5xx: Dump shader constants when dumping program assembly.
[mesa.git] / src / mesa / drivers / dri / r300 / r500_fragprog.c
1 /*
2 * Copyright (C) 2005 Ben Skeggs.
3 *
4 * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
5 * Adaptation and modification for ATI/AMD Radeon R500 GPU chipsets.
6 *
7 * All Rights Reserved.
8 *
9 * Permission is hereby granted, free of charge, to any person obtaining
10 * a copy of this software and associated documentation files (the
11 * "Software"), to deal in the Software without restriction, including
12 * without limitation the rights to use, copy, modify, merge, publish,
13 * distribute, sublicense, and/or sell copies of the Software, and to
14 * permit persons to whom the Software is furnished to do so, subject to
15 * the following conditions:
16 *
17 * The above copyright notice and this permission notice (including the
18 * next paragraph) shall be included in all copies or substantial
19 * portions of the Software.
20 *
21 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
22 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
23 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
24 * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
25 * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
26 * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
27 * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28 *
29 */
30
31 /**
32 * \file
33 *
34 * \author Ben Skeggs <darktama@iinet.net.au>
35 *
36 * \author Jerome Glisse <j.glisse@gmail.com>
37 *
38 * \author Corbin Simpson <MostAwesomeDude@gmail.com>
39 *
40 * \todo Depth write, WPOS/FOGC inputs
41 *
42 * \todo FogOption
43 *
44 * \todo Verify results of opcodes for accuracy, I've only checked them in
45 * specific cases.
46 */
47
48 #include "glheader.h"
49 #include "macros.h"
50 #include "enums.h"
51 #include "shader/prog_instruction.h"
52 #include "shader/prog_parameter.h"
53 #include "shader/prog_print.h"
54
55 #include "r300_context.h"
56 #include "r500_fragprog.h"
57 #include "r300_reg.h"
58 #include "r300_state.h"
59
60 /*
61 * Useful macros and values
62 */
63 #define ERROR(fmt, args...) do { \
64 fprintf(stderr, "%s::%s(): " fmt "\n", \
65 __FILE__, __FUNCTION__, ##args); \
66 fp->error = GL_TRUE; \
67 } while(0)
68
69 #define COMPILE_STATE struct r300_pfs_compile_state *cs = fp->cs
70
71 #define R500_US_NUM_TEMP_REGS 128
72 #define R500_US_NUM_CONST_REGS 256
73
74 /* "Register" flags */
75 #define REG_CONSTANT (1 << 8)
76 #define REG_SRC_REL (1 << 9)
77 #define REG_DEST_REL (1 << 7)
78
79 /* Swizzle tools */
80 #define R500_SWIZZLE_ZERO 4
81 #define R500_SWIZZLE_HALF 5
82 #define R500_SWIZZLE_ONE 6
83 #define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6))
84 #define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6))
85 #define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6))
86 #define R500_SWIZ_MOD_NEG 1
87 #define R500_SWIZ_MOD_ABS 2
88 #define R500_SWIZ_MOD_NEG_ABS 3
89 /* Swizzles for inst2 */
90 #define MAKE_SWIZ_TEX_STRQ(x) (x << 8)
91 #define MAKE_SWIZ_TEX_RGBA(x) (x << 24)
92 /* Swizzles for inst3 */
93 #define MAKE_SWIZ_RGB_A(x) (x << 2)
94 #define MAKE_SWIZ_RGB_B(x) (x << 15)
95 /* Swizzles for inst4 */
96 #define MAKE_SWIZ_ALPHA_A(x) (x << 14)
97 #define MAKE_SWIZ_ALPHA_B(x) (x << 21)
98 /* Swizzle for inst5 */
99 #define MAKE_SWIZ_RGBA_C(x) (x << 14)
100 #define MAKE_SWIZ_ALPHA_C(x) (x << 27)
101
102 /* Writemasks */
103 #define R500_WRITEMASK_G 0x2
104 #define R500_WRITEMASK_RGB 0x7
105 #define R500_WRITEMASK_A 0x8
106 #define R500_WRITEMASK_AR 0x9
107 #define R500_WRITEMASK_AG 0xA
108 #define R500_WRITEMASK_ARG 0xB
109 #define R500_WRITEMASK_AB 0xC
110 #define R500_WRITEMASK_ARGB 0xF
111
112 /* 1/(2pi), needed for quick modulus in trig insts
113 * Thanks to glisse for pointing out how to do it! */
114 static const GLfloat RCP_2PI[] = {0.15915494309189535,
115 0.15915494309189535,
116 0.15915494309189535,
117 0.15915494309189535};
118
119 static const GLfloat LIT[] = {127.999999,
120 127.999999,
121 127.999999,
122 -127.999999};
123
124 static void dump_program(struct r500_fragment_program *fp);
125
126 static inline GLuint make_rgb_swizzle(struct prog_src_register src) {
127 GLuint swiz = 0x0;
128 GLuint temp;
129 /* This could be optimized, but it should be plenty fast already. */
130 int i;
131 for (i = 0; i < 3; i++) {
132 temp = GET_SWZ(src.Swizzle, i);
133 /* Fix SWIZZLE_ONE */
134 if (temp == 5) temp++;
135 swiz |= temp << i*3;
136 }
137 if (src.NegateBase)
138 swiz |= (R500_SWIZ_MOD_NEG << 9);
139 return swiz;
140 }
141
142 static inline GLuint make_rgba_swizzle(GLuint src) {
143 GLuint swiz = 0x0;
144 GLuint temp;
145 int i;
146 for (i = 0; i < 4; i++) {
147 temp = GET_SWZ(src, i);
148 /* Fix SWIZZLE_ONE */
149 if (temp == 5) temp++;
150 swiz |= temp << i*3;
151 }
152 return swiz;
153 }
154
155 static inline GLuint make_alpha_swizzle(struct prog_src_register src) {
156 GLuint swiz = GET_SWZ(src.Swizzle, 3);
157
158 if (swiz == 5) swiz++;
159
160 if (src.NegateBase)
161 swiz |= (R500_SWIZ_MOD_NEG << 3);
162
163 return swiz;
164 }
165
166 static inline GLuint make_sop_swizzle(struct prog_src_register src) {
167 GLuint swiz = GET_SWZ(src.Swizzle, 0);
168
169 if (swiz == 5) swiz++;
170 return swiz;
171 }
172
173 static inline GLuint make_strq_swizzle(struct prog_src_register src) {
174 GLuint swiz = 0x0, temp = 0x0;
175 int i;
176 for (i = 0; i < 4; i++) {
177 temp = GET_SWZ(src.Swizzle, i) & 0x3;
178 swiz |= temp << i*2;
179 }
180 return swiz;
181 }
182
183 static int get_temp(struct r500_fragment_program *fp, int slot) {
184
185 COMPILE_STATE;
186
187 int r = fp->temp_reg_offset + cs->temp_in_use + slot;
188
189 if (r > R500_US_NUM_TEMP_REGS) {
190 ERROR("Too many temporary registers requested, can't compile!\n");
191 }
192
193 return r;
194 }
195
196 /* Borrowed verbatim from r300_fragprog since it hasn't changed. */
197 static GLuint emit_const4fv(struct r500_fragment_program *fp,
198 const GLfloat * cp)
199 {
200 GLuint reg = 0x0;
201 int index;
202
203 for (index = 0; index < fp->const_nr; ++index) {
204 if (fp->constant[index] == cp)
205 break;
206 }
207
208 if (index >= fp->const_nr) {
209 if (index >= R500_US_NUM_CONST_REGS) {
210 ERROR("Out of hw constants!\n");
211 return reg;
212 }
213
214 fp->const_nr++;
215 fp->constant[index] = cp;
216 }
217
218 reg = index | REG_CONSTANT;
219 return reg;
220 }
221
222 static GLuint make_src(struct r500_fragment_program *fp, struct prog_src_register src) {
223 COMPILE_STATE;
224 GLuint reg;
225 switch (src.File) {
226 case PROGRAM_TEMPORARY:
227 reg = src.Index + fp->temp_reg_offset;
228 break;
229 case PROGRAM_INPUT:
230 reg = cs->inputs[src.Index].reg;
231 break;
232 case PROGRAM_LOCAL_PARAM:
233 reg = emit_const4fv(fp,
234 fp->mesa_program.Base.LocalParams[src.
235 Index]);
236 break;
237 case PROGRAM_ENV_PARAM:
238 reg = emit_const4fv(fp,
239 fp->ctx->FragmentProgram.Parameters[src.
240 Index]);
241 break;
242 case PROGRAM_STATE_VAR:
243 case PROGRAM_NAMED_PARAM:
244 case PROGRAM_CONSTANT:
245 reg = emit_const4fv(fp, fp->mesa_program.Base.Parameters->
246 ParameterValues[src.Index]);
247 break;
248 default:
249 ERROR("Can't handle src.File %x\n", src.File);
250 reg = 0x0;
251 break;
252 }
253 return reg;
254 }
255
256 static GLuint make_dest(struct r500_fragment_program *fp, struct prog_dst_register dest) {
257 GLuint reg;
258 switch (dest.File) {
259 case PROGRAM_TEMPORARY:
260 reg = dest.Index + fp->temp_reg_offset;
261 break;
262 case PROGRAM_OUTPUT:
263 /* Eventually we may need to handle multiple
264 * rendering targets... */
265 reg = dest.Index;
266 break;
267 default:
268 ERROR("Can't handle dest.File %x\n", dest.File);
269 reg = 0x0;
270 break;
271 }
272 return reg;
273 }
274
275 static void emit_tex(struct r500_fragment_program *fp,
276 struct prog_instruction *fpi, int dest, int counter)
277 {
278 int hwsrc, hwdest;
279 GLuint mask;
280
281 mask = fpi->DstReg.WriteMask << 11;
282 hwsrc = make_src(fp, fpi->SrcReg[0]);
283
284 if (fpi->DstReg.File == PROGRAM_OUTPUT) {
285 hwdest = get_temp(fp, 0);
286 } else {
287 hwdest = dest;
288 }
289
290 fp->inst[counter].inst0 = R500_INST_TYPE_TEX | mask
291 | R500_INST_TEX_SEM_WAIT;
292
293 fp->inst[counter].inst1 = R500_TEX_ID(fpi->TexSrcUnit)
294 | R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED;
295
296 if (fpi->TexSrcTarget == TEXTURE_RECT_INDEX)
297 fp->inst[counter].inst1 |= R500_TEX_UNSCALED;
298
299 switch (fpi->Opcode) {
300 case OPCODE_KIL:
301 fp->inst[counter].inst1 |= R500_TEX_INST_TEXKILL;
302 break;
303 case OPCODE_TEX:
304 fp->inst[counter].inst1 |= R500_TEX_INST_LD;
305 break;
306 case OPCODE_TXB:
307 fp->inst[counter].inst1 |= R500_TEX_INST_LODBIAS;
308 break;
309 case OPCODE_TXP:
310 fp->inst[counter].inst1 |= R500_TEX_INST_PROJ;
311 break;
312 default:
313 ERROR("emit_tex can't handle opcode %x\n", fpi->Opcode);
314 }
315
316 fp->inst[counter].inst2 = R500_TEX_SRC_ADDR(hwsrc)
317 | MAKE_SWIZ_TEX_STRQ(make_strq_swizzle(fpi->SrcReg[0]))
318 /* | R500_TEX_SRC_S_SWIZ_R | R500_TEX_SRC_T_SWIZ_G
319 | R500_TEX_SRC_R_SWIZ_B | R500_TEX_SRC_Q_SWIZ_A */
320 | R500_TEX_DST_ADDR(hwdest)
321 | R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G
322 | R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A;
323
324 fp->inst[counter].inst3 = 0x0;
325 fp->inst[counter].inst4 = 0x0;
326 fp->inst[counter].inst5 = 0x0;
327
328 if (fpi->DstReg.File == PROGRAM_OUTPUT) {
329 counter++;
330 fp->inst[counter].inst0 = R500_INST_TYPE_OUT
331 | R500_INST_TEX_SEM_WAIT | (mask << 4);
332 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
333 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
334 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
335 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB)
336 | R500_ALU_RGB_SEL_B_SRC0
337 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB)
338 | R500_ALU_RGB_OMOD_DISABLE;
339 fp->inst[counter].inst4 = R500_ALPHA_OP_CMP
340 | R500_ALPHA_ADDRD(dest)
341 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_ALPHA_SWIZ_A_A)
342 | R500_ALPHA_SEL_B_SRC0 | MAKE_SWIZ_ALPHA_B(R500_ALPHA_SWIZ_A_A)
343 | R500_ALPHA_OMOD_DISABLE;
344 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
345 | R500_ALU_RGBA_ADDRD(dest)
346 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
347 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
348 }
349 }
350
351 static void emit_alu(struct r500_fragment_program *fp, int counter, struct prog_instruction *fpi) {
352 if (fpi->DstReg.File == PROGRAM_OUTPUT) {
353 fp->inst[counter].inst0 = R500_INST_TYPE_OUT;
354
355 if (fpi->DstReg.Index == FRAG_RESULT_COLR)
356 fp->inst[counter].inst0 |= (fpi->DstReg.WriteMask << 15);
357
358 if (fpi->DstReg.Index == FRAG_RESULT_DEPR)
359 fp->inst[counter].inst4 = R500_ALPHA_W_OMASK;
360 } else {
361 fp->inst[counter].inst0 = R500_INST_TYPE_ALU
362 /* pixel_mask */
363 | (fpi->DstReg.WriteMask << 11);
364 }
365
366 fp->inst[counter].inst0 |= R500_INST_TEX_SEM_WAIT;
367
368 /* Ideally, we shouldn't have to explicitly clear memory here! */
369 fp->inst[counter].inst1 = 0x0;
370 fp->inst[counter].inst2 = 0x0;
371 fp->inst[counter].inst3 = 0x0;
372 fp->inst[counter].inst5 = 0x0;
373 }
374
375 static void emit_mov(struct r500_fragment_program *fp, int counter, struct prog_instruction *fpi, GLuint src_reg, GLuint swizzle, GLuint dest) {
376 /* The r3xx shader uses MAD to implement MOV. We are using CMP, since
377 * it is technically more accurate and recommended by ATI/AMD. */
378 emit_alu(fp, counter, fpi);
379 fp->inst[counter].inst1 = R500_RGB_ADDR0(src_reg);
380 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src_reg);
381 /* (De)mangle the swizzle from Mesa to R500. */
382 swizzle = make_rgba_swizzle(swizzle);
383 /* 0x1FF is 9 bits, size of an RGB swizzle. */
384 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
385 | MAKE_SWIZ_RGB_A((swizzle & 0x1ff))
386 | R500_ALU_RGB_SEL_B_SRC0
387 | MAKE_SWIZ_RGB_B((swizzle & 0x1ff))
388 | R500_ALU_RGB_OMOD_DISABLE;
389 fp->inst[counter].inst4 |= R500_ALPHA_OP_CMP
390 | R500_ALPHA_ADDRD(dest)
391 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(GET_SWZ(swizzle, 3))
392 | R500_ALPHA_SEL_B_SRC0 | MAKE_SWIZ_ALPHA_B(GET_SWZ(swizzle, 3))
393 | R500_ALPHA_OMOD_DISABLE;
394 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
395 | R500_ALU_RGBA_ADDRD(dest)
396 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
397 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
398 }
399
400 static void emit_mad(struct r500_fragment_program *fp, int counter, struct prog_instruction *fpi, int one, int two, int three) {
401 /* Note: This code was all Corbin's. Corbin is a rather hackish coder.
402 * If you can make it pretty or fast, please do so! */
403 emit_alu(fp, counter, fpi);
404 /* Common MAD stuff */
405 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAD
406 | R500_ALPHA_ADDRD(make_dest(fp, fpi->DstReg));
407 fp->inst[counter].inst5 |= R500_ALU_RGBA_OP_MAD
408 | R500_ALU_RGBA_ADDRD(make_dest(fp, fpi->DstReg));
409 switch (one) {
410 case 0:
411 case 1:
412 case 2:
413 fp->inst[counter].inst1 |= R500_RGB_ADDR0(make_src(fp, fpi->SrcReg[one]));
414 fp->inst[counter].inst2 |= R500_ALPHA_ADDR0(make_src(fp, fpi->SrcReg[one]));
415 fp->inst[counter].inst3 |= R500_ALU_RGB_SEL_A_SRC0
416 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[one]));
417 fp->inst[counter].inst4 |= R500_ALPHA_SEL_A_SRC0
418 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[one]));
419 break;
420 case R500_SWIZZLE_ZERO:
421 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ZERO);
422 fp->inst[counter].inst4 |= MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ZERO);
423 break;
424 case R500_SWIZZLE_ONE:
425 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ONE);
426 fp->inst[counter].inst4 |= MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE);
427 break;
428 default:
429 ERROR("Bad src index in emit_mad: %d\n", one);
430 break;
431 }
432 switch (two) {
433 case 0:
434 case 1:
435 case 2:
436 fp->inst[counter].inst1 |= R500_RGB_ADDR1(make_src(fp, fpi->SrcReg[two]));
437 fp->inst[counter].inst2 |= R500_ALPHA_ADDR1(make_src(fp, fpi->SrcReg[two]));
438 fp->inst[counter].inst3 |= R500_ALU_RGB_SEL_B_SRC1
439 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[two]));
440 fp->inst[counter].inst4 |= R500_ALPHA_SEL_B_SRC1
441 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[two]));
442 break;
443 case R500_SWIZZLE_ZERO:
444 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ZERO);
445 fp->inst[counter].inst4 |= MAKE_SWIZ_ALPHA_B(R500_SWIZZLE_ZERO);
446 break;
447 case R500_SWIZZLE_ONE:
448 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ONE);
449 fp->inst[counter].inst4 |= MAKE_SWIZ_ALPHA_B(R500_SWIZZLE_ONE);
450 break;
451 default:
452 ERROR("Bad src index in emit_mad: %d\n", two);
453 break;
454 }
455 switch (three) {
456 case 0:
457 case 1:
458 case 2:
459 fp->inst[counter].inst1 |= R500_RGB_ADDR2(make_src(fp, fpi->SrcReg[three]));
460 fp->inst[counter].inst2 |= R500_ALPHA_ADDR2(make_src(fp, fpi->SrcReg[three]));
461 fp->inst[counter].inst5 |= R500_ALU_RGBA_SEL_C_SRC2
462 | MAKE_SWIZ_RGBA_C(make_rgb_swizzle(fpi->SrcReg[three]))
463 | R500_ALU_RGBA_ALPHA_SEL_C_SRC2
464 | MAKE_SWIZ_ALPHA_C(make_alpha_swizzle(fpi->SrcReg[three]));
465 break;
466 case R500_SWIZZLE_ZERO:
467 fp->inst[counter].inst5 |= MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
468 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
469 break;
470 case R500_SWIZZLE_ONE:
471 fp->inst[counter].inst5 |= MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ONE)
472 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ONE);
473 break;
474 default:
475 ERROR("Bad src index in emit_mad: %d\n", three);
476 break;
477 }
478 }
479
480 static GLboolean parse_program(struct r500_fragment_program *fp)
481 {
482 struct gl_fragment_program *mp = &fp->mesa_program;
483 const struct prog_instruction *inst = mp->Base.Instructions;
484 struct prog_instruction *fpi;
485 GLuint src[3], dest = 0;
486 int temp_swiz, counter = 0;
487
488 if (!inst || inst[0].Opcode == OPCODE_END) {
489 ERROR("The program is empty!\n");
490 return GL_FALSE;
491 }
492
493 for (fpi = mp->Base.Instructions; fpi->Opcode != OPCODE_END; fpi++) {
494
495 if (fpi->Opcode != OPCODE_KIL) {
496 dest = make_dest(fp, fpi->DstReg);
497 }
498
499 switch (fpi->Opcode) {
500 case OPCODE_ABS:
501 emit_mov(fp, counter, fpi, make_src(fp, fpi->SrcReg[0]), fpi->SrcReg[0].Swizzle, dest);
502 fp->inst[counter].inst3 |= R500_ALU_RGB_MOD_A_ABS
503 | R500_ALU_RGB_MOD_B_ABS;
504 fp->inst[counter].inst4 |= R500_ALPHA_MOD_A_ABS
505 | R500_ALPHA_MOD_B_ABS;
506 break;
507 case OPCODE_ADD:
508 /* Variation on MAD: 1*src0+src1 */
509 emit_mad(fp, counter, fpi, R500_SWIZZLE_ONE, 0, 1);
510 break;
511 case OPCODE_CMP:
512 /* This inst's selects need to be swapped as follows:
513 * 0 -> C ; 1 -> B ; 2 -> A */
514 src[0] = make_src(fp, fpi->SrcReg[0]);
515 src[1] = make_src(fp, fpi->SrcReg[1]);
516 src[2] = make_src(fp, fpi->SrcReg[2]);
517 emit_alu(fp, counter, fpi);
518 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[2])
519 | R500_RGB_ADDR1(src[1]) | R500_RGB_ADDR2(src[0]);
520 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[2])
521 | R500_ALPHA_ADDR1(src[1]) | R500_ALPHA_ADDR2(src[0]);
522 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
523 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[2]))
524 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
525 fp->inst[counter].inst4 |= R500_ALPHA_OP_CMP
526 | R500_ALPHA_ADDRD(dest)
527 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[2]))
528 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
529 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
530 | R500_ALU_RGBA_ADDRD(dest)
531 | R500_ALU_RGBA_SEL_C_SRC2
532 | MAKE_SWIZ_RGBA_C(make_rgb_swizzle(fpi->SrcReg[0]))
533 | R500_ALU_RGBA_ALPHA_SEL_C_SRC2
534 | MAKE_SWIZ_ALPHA_C(make_alpha_swizzle(fpi->SrcReg[0]));
535 break;
536 case OPCODE_COS:
537 src[0] = make_src(fp, fpi->SrcReg[0]);
538 src[1] = emit_const4fv(fp, RCP_2PI);
539 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
540 | (R500_WRITEMASK_ARGB << 11);
541 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
542 | R500_RGB_ADDR1(src[1]);
543 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
544 | R500_ALPHA_ADDR1(src[1]);
545 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
546 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB)
547 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB);
548 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
549 | R500_ALPHA_ADDRD(get_temp(fp, 0))
550 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A
551 | R500_ALPHA_SEL_B_SRC1 | R500_ALPHA_SWIZ_B_A;
552 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
553 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
554 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
555 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
556 counter++;
557 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_ARGB << 11);
558 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
559 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
560 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
561 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB);
562 fp->inst[counter].inst4 = R500_ALPHA_OP_FRC
563 | R500_ALPHA_ADDRD(get_temp(fp, 1))
564 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A;
565 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_FRC
566 | R500_ALU_RGBA_ADDRD(get_temp(fp, 1));
567 counter++;
568 emit_alu(fp, counter, fpi);
569 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 1));
570 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 1));
571 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0;
572 fp->inst[counter].inst4 |= R500_ALPHA_OP_COS
573 | R500_ALPHA_ADDRD(dest)
574 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
575 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
576 | R500_ALU_RGBA_ADDRD(dest);
577 break;
578 case OPCODE_DP3:
579 src[0] = make_src(fp, fpi->SrcReg[0]);
580 src[1] = make_src(fp, fpi->SrcReg[1]);
581 emit_alu(fp, counter, fpi);
582 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
583 | R500_RGB_ADDR1(src[1]);
584 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
585 | R500_ALPHA_ADDR1(src[1]);
586 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
587 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
588 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
589 fp->inst[counter].inst4 |= R500_ALPHA_OP_DP
590 | R500_ALPHA_ADDRD(dest)
591 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
592 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
593 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_DP3
594 | R500_ALU_RGBA_ADDRD(dest);
595 break;
596 case OPCODE_DP4:
597 src[0] = make_src(fp, fpi->SrcReg[0]);
598 src[1] = make_src(fp, fpi->SrcReg[1]);
599 /* Based on DP3 */
600 emit_alu(fp, counter, fpi);
601 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
602 | R500_RGB_ADDR1(src[1]);
603 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
604 | R500_ALPHA_ADDR1(src[1]);
605 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
606 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
607 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
608 fp->inst[counter].inst4 |= R500_ALPHA_OP_DP
609 | R500_ALPHA_ADDRD(dest)
610 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
611 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
612 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_DP4
613 | R500_ALU_RGBA_ADDRD(dest);
614 break;
615 case OPCODE_DPH:
616 src[0] = make_src(fp, fpi->SrcReg[0]);
617 src[1] = make_src(fp, fpi->SrcReg[1]);
618 /* Based on DP3 */
619 emit_alu(fp, counter, fpi);
620 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
621 | R500_RGB_ADDR1(src[1]);
622 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
623 | R500_ALPHA_ADDR1(src[1]);
624 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
625 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
626 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
627 fp->inst[counter].inst4 |= R500_ALPHA_OP_DP
628 | R500_ALPHA_ADDRD(dest)
629 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE)
630 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
631 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_DP4
632 | R500_ALU_RGBA_ADDRD(dest);
633 break;
634 case OPCODE_DST:
635 src[0] = make_src(fp, fpi->SrcReg[0]);
636 src[1] = make_src(fp, fpi->SrcReg[1]);
637 /* [1, src0.y*src1.y, src0.z, src1.w]
638 * So basically MUL with lotsa swizzling. */
639 emit_alu(fp, counter, fpi);
640 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
641 | R500_RGB_ADDR1(src[1]);
642 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
643 | R500_ALPHA_ADDR1(src[1]);
644 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
645 | R500_ALU_RGB_SEL_B_SRC1;
646 /* Select [1, y, z, 1] */
647 temp_swiz = (make_rgb_swizzle(fpi->SrcReg[0]) & ~0x7) | R500_SWIZZLE_ONE;
648 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_A(temp_swiz);
649 /* Select [1, y, 1, w] */
650 temp_swiz = (make_rgb_swizzle(fpi->SrcReg[0]) & ~0x1c7) | R500_SWIZZLE_ONE | (R500_SWIZZLE_ONE << 6);
651 fp->inst[counter].inst3 |= MAKE_SWIZ_RGB_B(temp_swiz);
652 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAD
653 | R500_ALPHA_ADDRD(dest)
654 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE)
655 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
656 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
657 | R500_ALU_RGBA_ADDRD(dest)
658 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
659 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
660 break;
661 case OPCODE_EX2:
662 src[0] = make_src(fp, fpi->SrcReg[0]);
663 emit_alu(fp, counter, fpi);
664 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
665 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
666 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
667 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
668 fp->inst[counter].inst4 |= R500_ALPHA_OP_EX2
669 | R500_ALPHA_ADDRD(dest)
670 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
671 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
672 | R500_ALU_RGBA_ADDRD(dest);
673 break;
674 case OPCODE_FRC:
675 src[0] = make_src(fp, fpi->SrcReg[0]);
676 emit_alu(fp, counter, fpi);
677 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
678 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
679 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
680 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
681 fp->inst[counter].inst4 |= R500_ALPHA_OP_FRC
682 | R500_ALPHA_ADDRD(dest)
683 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]));
684 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_FRC
685 | R500_ALU_RGBA_ADDRD(dest);
686 break;
687 case OPCODE_LG2:
688 src[0] = make_src(fp, fpi->SrcReg[0]);
689 emit_alu(fp, counter, fpi);
690 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
691 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
692 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
693 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
694 fp->inst[counter].inst4 |= R500_ALPHA_OP_LN2
695 | R500_ALPHA_ADDRD(dest)
696 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
697 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
698 | R500_ALU_RGBA_ADDRD(dest);
699 break;
700 case OPCODE_LIT:
701 /* To be honest, I have no idea how I came up with the following.
702 * All I know is that it's based on the r3xx stuff, and was
703 * concieved with the help of NyQuil. Mmm, MyQuil. */
704
705 /* First instruction */
706 src[0] = make_src(fp, fpi->SrcReg[0]);
707 src[1] = emit_const4fv(fp, LIT);
708 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
709 | (R500_WRITEMASK_ARG << 11);
710 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]) | R500_RGB_ADDR1(src[1]);
711 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]) | R500_ALPHA_ADDR1(src[1]);
712 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
713 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
714 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ZERO);
715 fp->inst[counter].inst4 = R500_ALPHA_OP_MAX
716 | R500_ALPHA_ADDRD(get_temp(fp, 0))
717 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
718 | R500_ALPHA_SEL_B_SRC1 | R500_ALPHA_SWIZ_B_A;
719 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAX
720 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0));
721 counter++;
722 /* Second instruction */
723 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_AB << 11);
724 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0)) | R500_RGB_ADDR1(src[1]);
725 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
726 /* Select [z, z, z, y] */
727 temp_swiz = 2 | (2 << 3) | (2 << 6);
728 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
729 | MAKE_SWIZ_RGB_A(temp_swiz)
730 | R500_ALU_RGB_SEL_B_SRC0
731 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB);
732 fp->inst[counter].inst4 = R500_ALPHA_OP_LN2
733 | R500_ALPHA_ADDRD(get_temp(fp, 0))
734 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_G;
735 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MIN
736 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0));
737 counter++;
738 /* Third instruction */
739 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_AG << 11);
740 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
741 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
742 /* Select [x, x, x, z] */
743 temp_swiz = 0;
744 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
745 | MAKE_SWIZ_RGB_A(temp_swiz)
746 | R500_ALU_RGB_SEL_B_SRC0
747 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ONE);
748 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
749 | R500_ALPHA_ADDRD(get_temp(fp, 1))
750 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A
751 | R500_ALPHA_SEL_B_SRC0 | R500_ALPHA_SWIZ_B_B;
752 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
753 | R500_ALU_RGBA_ADDRD(get_temp(fp, 1))
754 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
755 | R500_ALU_RGBA_A_SWIZ_0;
756 counter++;
757 /* Fourth instruction */
758 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_AR << 11);
759 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
760 fp->inst[counter].inst3 = MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ONE)
761 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ONE);
762 fp->inst[counter].inst4 = R500_ALPHA_OP_EX2
763 | R500_ALPHA_ADDRD(get_temp(fp, 0))
764 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A;
765 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
766 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
767 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
768 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
769 counter++;
770 /* Fifth instruction */
771 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_AB << 11);
772 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
773 /* Select [w, w, w] */
774 temp_swiz = 3 | (3 << 3) | (3 << 6);
775 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
776 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ZERO)
777 | R500_ALU_RGB_SEL_B_SRC0
778 | MAKE_SWIZ_RGB_B(temp_swiz);
779 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAD
780 | R500_ALPHA_ADDRD(get_temp(fp, 0))
781 | R500_ALPHA_SWIZ_A_1
782 | R500_ALPHA_SWIZ_B_1;
783 /* Select [-y, -y, -y] */
784 temp_swiz = 1 | (1 << 3) | (1 << 6);
785 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
786 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
787 | MAKE_SWIZ_RGBA_C(temp_swiz)
788 | R500_ALU_RGBA_MOD_C_NEG
789 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
790 counter++;
791 /* Final instruction */
792 emit_mov(fp, counter, fpi, get_temp(fp, 0), SWIZZLE_NOOP, dest);
793 break;
794 case OPCODE_LRP:
795 /* src0 * src1 + INV(src0) * src2
796 * 1) MUL src0, src1, temp
797 * 2) PRE 1-src0; MAD srcp, src2, temp */
798 src[0] = make_src(fp, fpi->SrcReg[0]);
799 src[1] = make_src(fp, fpi->SrcReg[1]);
800 src[2] = make_src(fp, fpi->SrcReg[2]);
801 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
802 | R500_INST_NOP | (R500_WRITEMASK_ARGB << 11);
803 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
804 | R500_RGB_ADDR1(src[1]);
805 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
806 | R500_ALPHA_ADDR1(src[1]);
807 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
808 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
809 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
810 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
811 | R500_ALPHA_ADDRD(get_temp(fp, 0))
812 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
813 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
814 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
815 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
816 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
817 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
818 counter++;
819 emit_alu(fp, counter, fpi);
820 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
821 | R500_RGB_ADDR1(src[2])
822 | R500_RGB_ADDR2(get_temp(fp, 0))
823 | R500_RGB_SRCP_OP_1_MINUS_RGB0;
824 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
825 | R500_ALPHA_ADDR1(src[2])
826 | R500_ALPHA_ADDR2(get_temp(fp, 0))
827 | R500_ALPHA_SRCP_OP_1_MINUS_A0;
828 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRCP
829 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
830 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB);
831 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAD
832 | R500_ALPHA_ADDRD(dest)
833 | R500_ALPHA_SEL_A_SRCP | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
834 | R500_ALPHA_SEL_B_SRC1 | R500_ALPHA_SWIZ_B_A;
835 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
836 | R500_ALU_RGBA_ADDRD(dest)
837 | R500_ALU_RGBA_SEL_C_SRC2 | MAKE_SWIZ_RGBA_C(make_rgb_swizzle(fpi->SrcReg[2]))
838 | R500_ALU_RGBA_ALPHA_SEL_C_SRC2
839 | MAKE_SWIZ_ALPHA_C(make_alpha_swizzle(fpi->SrcReg[2]));
840 break;
841 case OPCODE_MAD:
842 emit_mad(fp, counter, fpi, 0, 1, 2);
843 break;
844 case OPCODE_MAX:
845 src[0] = make_src(fp, fpi->SrcReg[0]);
846 src[1] = make_src(fp, fpi->SrcReg[1]);
847 emit_alu(fp, counter, fpi);
848 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]) | R500_RGB_ADDR1(src[1]);
849 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]) | R500_ALPHA_ADDR1(src[1]);
850 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
851 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
852 | R500_ALU_RGB_SEL_B_SRC1
853 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
854 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAX
855 | R500_ALPHA_ADDRD(dest)
856 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
857 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
858 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAX
859 | R500_ALU_RGBA_ADDRD(dest);
860 break;
861 case OPCODE_MIN:
862 src[0] = make_src(fp, fpi->SrcReg[0]);
863 src[1] = make_src(fp, fpi->SrcReg[1]);
864 emit_alu(fp, counter, fpi);
865 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]) | R500_RGB_ADDR1(src[1]);
866 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]) | R500_ALPHA_ADDR1(src[1]);
867 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
868 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
869 | R500_ALU_RGB_SEL_B_SRC1
870 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
871 fp->inst[counter].inst4 |= R500_ALPHA_OP_MIN
872 | R500_ALPHA_ADDRD(dest)
873 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
874 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
875 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MIN
876 | R500_ALU_RGBA_ADDRD(dest);
877 break;
878 case OPCODE_MOV:
879 emit_mov(fp, counter, fpi, make_src(fp, fpi->SrcReg[0]), fpi->SrcReg[0].Swizzle, dest);
880 break;
881 case OPCODE_MUL:
882 /* Variation on MAD: src0*src1+0 */
883 emit_mad(fp, counter, fpi, 0, 1, R500_SWIZZLE_ZERO);
884 break;
885 case OPCODE_POW:
886 /* POW(a,b) = EX2(LN2(a)*b) */
887 src[0] = make_src(fp, fpi->SrcReg[0]);
888 src[1] = make_src(fp, fpi->SrcReg[1]);
889 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
890 | (R500_WRITEMASK_ARGB << 11);
891 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
892 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
893 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
894 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
895 fp->inst[counter].inst4 = R500_ALPHA_OP_LN2
896 | R500_ALPHA_ADDRD(get_temp(fp, 0))
897 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
898 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
899 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0));
900 counter++;
901 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_ARGB << 11);
902 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0))
903 | R500_RGB_ADDR1(src[1]);
904 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0))
905 | R500_ALPHA_ADDR1(src[1]);
906 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
907 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]))
908 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[1]));
909 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
910 | R500_ALPHA_ADDRD(get_temp(fp, 1))
911 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
912 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
913 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
914 | R500_ALU_RGBA_ADDRD(get_temp(fp, 1))
915 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
916 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
917 counter++;
918 emit_alu(fp, counter, fpi);
919 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 1));
920 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 1));
921 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
922 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
923 fp->inst[counter].inst4 |= R500_ALPHA_OP_EX2
924 | R500_ALPHA_ADDRD(dest)
925 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
926 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
927 | R500_ALU_RGBA_ADDRD(dest);
928 break;
929 case OPCODE_RCP:
930 src[0] = make_src(fp, fpi->SrcReg[0]);
931 emit_alu(fp, counter, fpi);
932 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
933 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
934 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
935 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
936 fp->inst[counter].inst4 |= R500_ALPHA_OP_RCP
937 | R500_ALPHA_ADDRD(dest)
938 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
939 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
940 | R500_ALU_RGBA_ADDRD(dest);
941 break;
942 case OPCODE_RSQ:
943 src[0] = make_src(fp, fpi->SrcReg[0]);
944 emit_alu(fp, counter, fpi);
945 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0]);
946 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0]);
947 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
948 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
949 fp->inst[counter].inst4 |= R500_ALPHA_OP_RSQ
950 | R500_ALPHA_ADDRD(dest)
951 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
952 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
953 | R500_ALU_RGBA_ADDRD(dest);
954 break;
955 case OPCODE_SCS:
956 src[0] = make_src(fp, fpi->SrcReg[0]);
957 src[1] = emit_const4fv(fp, RCP_2PI);
958 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
959 | (R500_WRITEMASK_ARGB << 11);
960 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
961 | R500_RGB_ADDR1(src[1]);
962 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
963 | R500_ALPHA_ADDR1(src[1]);
964 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
965 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB)
966 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB);
967 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
968 | R500_ALPHA_ADDRD(get_temp(fp, 0))
969 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A
970 | R500_ALPHA_SEL_B_SRC1 | R500_ALPHA_SWIZ_B_A;
971 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
972 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
973 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
974 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
975 counter++;
976 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_ARGB << 11);
977 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
978 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
979 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
980 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB);
981 fp->inst[counter].inst4 = R500_ALPHA_OP_FRC
982 | R500_ALPHA_ADDRD(get_temp(fp, 1))
983 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A;
984 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_FRC
985 | R500_ALU_RGBA_ADDRD(get_temp(fp, 1));
986 counter++;
987 /* Do a cosine, then a sine, masking out the channels we want to protect. */
988 /* Cosine only goes in R (x) channel. */
989 fpi->DstReg.WriteMask = 0x1;
990 emit_alu(fp, counter, fpi);
991 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 1));
992 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 1));
993 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
994 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
995 fp->inst[counter].inst4 |= R500_ALPHA_OP_COS
996 | R500_ALPHA_ADDRD(dest)
997 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
998 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
999 | R500_ALU_RGBA_ADDRD(dest);
1000 counter++;
1001 /* Sine only goes in G (y) channel. */
1002 fpi->DstReg.WriteMask = 0x2;
1003 emit_alu(fp, counter, fpi);
1004 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 1));
1005 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 1));
1006 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1007 | MAKE_SWIZ_RGB_A(make_rgb_swizzle(fpi->SrcReg[0]));
1008 fp->inst[counter].inst4 |= R500_ALPHA_OP_SIN
1009 | R500_ALPHA_ADDRD(dest)
1010 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
1011 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
1012 | R500_ALU_RGBA_ADDRD(dest);
1013 break;
1014 case OPCODE_SGE:
1015 src[0] = make_src(fp, fpi->SrcReg[0]);
1016 src[1] = make_src(fp, fpi->SrcReg[1]);
1017 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
1018 | (R500_WRITEMASK_ARGB << 11);
1019 fp->inst[counter].inst1 = R500_RGB_ADDR1(src[0])
1020 | R500_RGB_ADDR2(src[1]);
1021 fp->inst[counter].inst2 = R500_ALPHA_ADDR1(src[0])
1022 | R500_ALPHA_ADDR2(src[1]);
1023 fp->inst[counter].inst3 = /* 1 */
1024 MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ONE)
1025 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[0]));
1026 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
1027 | R500_ALPHA_ADDRD(get_temp(fp, 0))
1028 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE)
1029 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[0]));
1030 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
1031 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
1032 | R500_ALU_RGBA_SEL_C_SRC2
1033 | MAKE_SWIZ_RGBA_C(make_rgb_swizzle(fpi->SrcReg[1]))
1034 | R500_ALU_RGBA_MOD_C_NEG
1035 | R500_ALU_RGBA_ALPHA_SEL_C_SRC2
1036 | MAKE_SWIZ_ALPHA_C(make_alpha_swizzle(fpi->SrcReg[1]))
1037 | R500_ALU_RGBA_ALPHA_MOD_C_NEG;
1038 counter++;
1039 /* This inst's selects need to be swapped as follows:
1040 * 0 -> C ; 1 -> B ; 2 -> A */
1041 emit_alu(fp, counter, fpi);
1042 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
1043 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
1044 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1045 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ONE)
1046 | R500_ALU_RGB_SEL_B_SRC0
1047 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ZERO);
1048 fp->inst[counter].inst4 |= R500_ALPHA_OP_CMP
1049 | R500_ALPHA_ADDRD(dest)
1050 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE)
1051 | R500_ALPHA_SEL_B_SRC0 | MAKE_SWIZ_ALPHA_B(R500_SWIZZLE_ZERO);
1052 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
1053 | R500_ALU_RGBA_ADDRD(dest)
1054 | R500_ALU_RGBA_SEL_C_SRC0
1055 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_RGB)
1056 | R500_ALU_RGBA_ALPHA_SEL_C_SRC0
1057 | R500_ALU_RGBA_A_SWIZ_A;
1058 break;
1059 case OPCODE_SIN:
1060 src[0] = make_src(fp, fpi->SrcReg[0]);
1061 src[1] = emit_const4fv(fp, RCP_2PI);
1062 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
1063 | (R500_WRITEMASK_ARGB << 11);
1064 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
1065 | R500_RGB_ADDR1(src[1]);
1066 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
1067 | R500_ALPHA_ADDR1(src[1]);
1068 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1069 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB)
1070 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_RGB);
1071 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
1072 | R500_ALPHA_ADDRD(get_temp(fp, 0))
1073 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A
1074 | R500_ALPHA_SEL_B_SRC1 | R500_ALPHA_SWIZ_B_A;
1075 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
1076 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
1077 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
1078 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
1079 counter++;
1080 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | (R500_WRITEMASK_ARGB << 11);
1081 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
1082 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
1083 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1084 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_RGB);
1085 fp->inst[counter].inst4 = R500_ALPHA_OP_FRC
1086 | R500_ALPHA_ADDRD(get_temp(fp, 1))
1087 | R500_ALPHA_SEL_A_SRC0 | R500_ALPHA_SWIZ_A_A;
1088 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_FRC
1089 | R500_ALU_RGBA_ADDRD(get_temp(fp, 1));
1090 counter++;
1091 emit_alu(fp, counter, fpi);
1092 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 1));
1093 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 1));
1094 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0;
1095 fp->inst[counter].inst4 |= R500_ALPHA_OP_SIN
1096 | R500_ALPHA_ADDRD(dest)
1097 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_sop_swizzle(fpi->SrcReg[0]));
1098 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_SOP
1099 | R500_ALU_RGBA_ADDRD(dest);
1100 break;
1101 case OPCODE_SLT:
1102 src[0] = make_src(fp, fpi->SrcReg[0]);
1103 src[1] = make_src(fp, fpi->SrcReg[1]);
1104 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
1105 | (R500_WRITEMASK_ARGB << 11);
1106 fp->inst[counter].inst1 = R500_RGB_ADDR1(src[0])
1107 | R500_RGB_ADDR2(src[1]);
1108 fp->inst[counter].inst2 = R500_ALPHA_ADDR1(src[0])
1109 | R500_ALPHA_ADDR2(src[1]);
1110 fp->inst[counter].inst3 = /* 1 */
1111 MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ONE)
1112 | R500_ALU_RGB_SEL_B_SRC1 | MAKE_SWIZ_RGB_B(make_rgb_swizzle(fpi->SrcReg[0]));
1113 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
1114 | R500_ALPHA_ADDRD(get_temp(fp, 0))
1115 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ONE)
1116 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[0]));
1117 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
1118 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
1119 | R500_ALU_RGBA_SEL_C_SRC2
1120 | MAKE_SWIZ_RGBA_C(make_rgb_swizzle(fpi->SrcReg[1]))
1121 | R500_ALU_RGBA_MOD_C_NEG
1122 | R500_ALU_RGBA_ALPHA_SEL_C_SRC2
1123 | MAKE_SWIZ_ALPHA_C(make_alpha_swizzle(fpi->SrcReg[1]))
1124 | R500_ALU_RGBA_ALPHA_MOD_C_NEG;
1125 counter++;
1126 /* This inst's selects need to be swapped as follows:
1127 * 0 -> C ; 1 -> B ; 2 -> A */
1128 emit_alu(fp, counter, fpi);
1129 fp->inst[counter].inst1 = R500_RGB_ADDR0(get_temp(fp, 0));
1130 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(get_temp(fp, 0));
1131 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1132 | MAKE_SWIZ_RGB_A(R500_SWIZ_RGB_ZERO)
1133 | R500_ALU_RGB_SEL_B_SRC0
1134 | MAKE_SWIZ_RGB_B(R500_SWIZ_RGB_ONE);
1135 fp->inst[counter].inst4 |= R500_ALPHA_OP_CMP
1136 | R500_ALPHA_ADDRD(dest)
1137 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(R500_SWIZZLE_ZERO)
1138 | R500_ALPHA_SEL_B_SRC0 | MAKE_SWIZ_ALPHA_B(R500_SWIZZLE_ONE);
1139 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_CMP
1140 | R500_ALU_RGBA_ADDRD(dest)
1141 | R500_ALU_RGBA_SEL_C_SRC0
1142 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_RGB)
1143 | R500_ALU_RGBA_ALPHA_SEL_C_SRC0
1144 | R500_ALU_RGBA_A_SWIZ_A;
1145 break;
1146 case OPCODE_SUB:
1147 /* Variation on MAD: 1*src0-src1 */
1148 fpi->SrcReg[1].NegateBase = 0xF; /* NEG_XYZW */
1149 emit_mad(fp, counter, fpi, R500_SWIZZLE_ONE, 0, 1);
1150 break;
1151 case OPCODE_SWZ:
1152 /* TODO: The rarer negation masks! */
1153 emit_mov(fp, counter, fpi, make_src(fp, fpi->SrcReg[0]), fpi->SrcReg[0].Swizzle, dest);
1154 break;
1155 case OPCODE_XPD:
1156 /* src0 * src1 - src1 * src0
1157 * 1) MUL temp.xyz, src0.yzx, src1.zxy
1158 * 2) MAD src0.zxy, src1.yzx, -temp.xyz */
1159 src[0] = make_src(fp, fpi->SrcReg[0]);
1160 src[1] = make_src(fp, fpi->SrcReg[1]);
1161 fp->inst[counter].inst0 = R500_INST_TYPE_ALU | R500_INST_TEX_SEM_WAIT
1162 | (R500_WRITEMASK_RGB << 11);
1163 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
1164 | R500_RGB_ADDR1(src[1]);
1165 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
1166 | R500_ALPHA_ADDR1(src[1]);
1167 /* Select [y, z, x] */
1168 temp_swiz = make_rgb_swizzle(fpi->SrcReg[0]);
1169 temp_swiz = (GET_SWZ(temp_swiz, 1) << 0) | (GET_SWZ(temp_swiz, 2) << 3) | (GET_SWZ(temp_swiz, 0) << 6);
1170 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1171 | MAKE_SWIZ_RGB_A(temp_swiz);
1172 /* Select [z, x, y] */
1173 temp_swiz = make_rgb_swizzle(fpi->SrcReg[1]);
1174 temp_swiz = (GET_SWZ(temp_swiz, 2) << 0) | (GET_SWZ(temp_swiz, 0) << 3) | (GET_SWZ(temp_swiz, 1) << 6);
1175 fp->inst[counter].inst3 |= R500_ALU_RGB_SEL_B_SRC1
1176 | MAKE_SWIZ_RGB_B(temp_swiz);
1177 fp->inst[counter].inst4 = R500_ALPHA_OP_MAD
1178 | R500_ALPHA_ADDRD(get_temp(fp, 0))
1179 | R500_ALPHA_SEL_A_SRC0 | MAKE_SWIZ_ALPHA_A(make_alpha_swizzle(fpi->SrcReg[0]))
1180 | R500_ALPHA_SEL_B_SRC1 | MAKE_SWIZ_ALPHA_B(make_alpha_swizzle(fpi->SrcReg[1]));
1181 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
1182 | R500_ALU_RGBA_ADDRD(get_temp(fp, 0))
1183 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_ZERO)
1184 | MAKE_SWIZ_ALPHA_C(R500_SWIZZLE_ZERO);
1185 counter++;
1186 emit_alu(fp, counter, fpi);
1187 fp->inst[counter].inst1 = R500_RGB_ADDR0(src[0])
1188 | R500_RGB_ADDR1(src[1])
1189 | R500_RGB_ADDR2(get_temp(fp, 0));
1190 fp->inst[counter].inst2 = R500_ALPHA_ADDR0(src[0])
1191 | R500_ALPHA_ADDR1(src[1])
1192 | R500_ALPHA_ADDR2(get_temp(fp, 0));
1193 /* Select [z, x, y] */
1194 temp_swiz = make_rgb_swizzle(fpi->SrcReg[0]);
1195 temp_swiz = (GET_SWZ(temp_swiz, 2) << 0) | (GET_SWZ(temp_swiz, 0) << 3) | (GET_SWZ(temp_swiz, 1) << 6);
1196 fp->inst[counter].inst3 = R500_ALU_RGB_SEL_A_SRC0
1197 | MAKE_SWIZ_RGB_A(temp_swiz);
1198 /* Select [y, z, x] */
1199 temp_swiz = make_rgb_swizzle(fpi->SrcReg[1]);
1200 temp_swiz = (GET_SWZ(temp_swiz, 1) << 0) | (GET_SWZ(temp_swiz, 2) << 3) | (GET_SWZ(temp_swiz, 0) << 6);
1201 fp->inst[counter].inst3 |= R500_ALU_RGB_SEL_B_SRC1
1202 | MAKE_SWIZ_RGB_B(temp_swiz);
1203 fp->inst[counter].inst4 |= R500_ALPHA_OP_MAD
1204 | R500_ALPHA_ADDRD(dest)
1205 | R500_ALPHA_SWIZ_A_1
1206 | R500_ALPHA_SWIZ_B_1;
1207 fp->inst[counter].inst5 = R500_ALU_RGBA_OP_MAD
1208 | R500_ALU_RGBA_ADDRD(dest)
1209 | R500_ALU_RGBA_SEL_C_SRC2
1210 | MAKE_SWIZ_RGBA_C(R500_SWIZ_RGB_RGB)
1211 | R500_ALU_RGBA_MOD_C_NEG
1212 | R500_ALU_RGBA_A_SWIZ_0;
1213 break;
1214 case OPCODE_KIL:
1215 case OPCODE_TEX:
1216 case OPCODE_TXB:
1217 case OPCODE_TXP:
1218 emit_tex(fp, fpi, dest, counter);
1219 if (fpi->DstReg.File == PROGRAM_OUTPUT)
1220 counter++;
1221 break;
1222 default:
1223 ERROR("unknown fpi->Opcode %s\n", _mesa_opcode_string(fpi->Opcode));
1224 break;
1225 }
1226
1227 /* Finishing touches */
1228 if (fpi->SaturateMode == SATURATE_ZERO_ONE) {
1229 fp->inst[counter].inst0 |= R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP;
1230 }
1231
1232 counter++;
1233
1234 if (fp->error)
1235 return GL_FALSE;
1236
1237 }
1238
1239 /* Finish him! (If it's an ALU/OUT instruction...) */
1240 if ((fp->inst[counter-1].inst0 & 0x3) == 1) {
1241 fp->inst[counter-1].inst0 |= R500_INST_LAST;
1242 } else {
1243 /* We still need to put an output inst, right? */
1244 WARN_ONCE("Final FP instruction is not an OUT.\n");
1245 }
1246
1247 fp->cs->nrslots = counter;
1248
1249 fp->max_temp_idx++;
1250
1251 return GL_TRUE;
1252 }
1253
1254 static void init_program(r300ContextPtr r300, struct r500_fragment_program *fp)
1255 {
1256 struct r300_pfs_compile_state *cs = NULL;
1257 struct gl_fragment_program *mp = &fp->mesa_program;
1258 struct prog_instruction *fpi;
1259 GLuint InputsRead = mp->Base.InputsRead;
1260 GLuint temps_used = 0;
1261 int i, j;
1262
1263 /* New compile, reset tracking data */
1264 fp->optimization =
1265 driQueryOptioni(&r300->radeon.optionCache, "fp_optimization");
1266 fp->translated = GL_FALSE;
1267 fp->error = GL_FALSE;
1268 fp->cs = cs = &(R300_CONTEXT(fp->ctx)->state.pfs_compile);
1269 fp->const_nr = 0;
1270 /* Size of pixel stack, plus 1. */
1271 fp->max_temp_idx = 1;
1272 /* Temp register offset. */
1273 fp->temp_reg_offset = 0;
1274
1275 _mesa_memset(cs, 0, sizeof(*fp->cs));
1276 for (i = 0; i < PFS_MAX_ALU_INST; i++) {
1277 for (j = 0; j < 3; j++) {
1278 cs->slot[i].vsrc[j] = SRC_CONST;
1279 cs->slot[i].ssrc[j] = SRC_CONST;
1280 }
1281 }
1282
1283 /* Work out what temps the Mesa inputs correspond to, this must match
1284 * what setup_rs_unit does, which shouldn't be a problem as rs_unit
1285 * configures itself based on the fragprog's InputsRead
1286 *
1287 * NOTE: this depends on get_hw_temp() allocating registers in order,
1288 * starting from register 0, so we're just going to do that instead.
1289 */
1290
1291 /* Texcoords come first */
1292 for (i = 0; i < fp->ctx->Const.MaxTextureUnits; i++) {
1293 if (InputsRead & (FRAG_BIT_TEX0 << i)) {
1294 cs->inputs[FRAG_ATTRIB_TEX0 + i].refcount = 0;
1295 cs->inputs[FRAG_ATTRIB_TEX0 + i].reg =
1296 fp->temp_reg_offset;
1297 fp->temp_reg_offset++;
1298 }
1299 }
1300 InputsRead &= ~FRAG_BITS_TEX_ANY;
1301
1302 /* fragment position treated as a texcoord */
1303 if (InputsRead & FRAG_BIT_WPOS) {
1304 cs->inputs[FRAG_ATTRIB_WPOS].refcount = 0;
1305 cs->inputs[FRAG_ATTRIB_WPOS].reg =
1306 fp->temp_reg_offset;
1307 fp->temp_reg_offset++;
1308 }
1309 InputsRead &= ~FRAG_BIT_WPOS;
1310
1311 /* Then primary colour */
1312 if (InputsRead & FRAG_BIT_COL0) {
1313 cs->inputs[FRAG_ATTRIB_COL0].refcount = 0;
1314 cs->inputs[FRAG_ATTRIB_COL0].reg =
1315 fp->temp_reg_offset;
1316 fp->temp_reg_offset++;
1317 }
1318 InputsRead &= ~FRAG_BIT_COL0;
1319
1320 /* Secondary color */
1321 if (InputsRead & FRAG_BIT_COL1) {
1322 cs->inputs[FRAG_ATTRIB_COL1].refcount = 0;
1323 cs->inputs[FRAG_ATTRIB_COL1].reg =
1324 fp->temp_reg_offset;
1325 fp->temp_reg_offset++;
1326 }
1327 InputsRead &= ~FRAG_BIT_COL1;
1328
1329 /* Anything else */
1330 if (InputsRead) {
1331 WARN_ONCE("Don't know how to handle inputs 0x%x\n", InputsRead);
1332 /* force read from hwreg 0 for now */
1333 for (i = 0; i < 32; i++)
1334 if (InputsRead & (1 << i))
1335 cs->inputs[i].reg = 0;
1336 }
1337
1338 if (!mp->Base.Instructions) {
1339 ERROR("No instructions found in program, going to go die now.\n");
1340 return;
1341 }
1342
1343 for (fpi = mp->Base.Instructions; fpi->Opcode != OPCODE_END; fpi++) {
1344 for (i = 0; i < 3; i++) {
1345 if (fpi->SrcReg[i].File == PROGRAM_TEMPORARY) {
1346 if (fpi->SrcReg[i].Index >= temps_used)
1347 temps_used = fpi->SrcReg[i].Index + 1;
1348 }
1349 }
1350 }
1351
1352 cs->temp_in_use = temps_used + 1;
1353
1354 fp->max_temp_idx = fp->temp_reg_offset + cs->temp_in_use;
1355
1356 if (RADEON_DEBUG & DEBUG_PIXEL)
1357 fprintf(stderr, "FP temp indices: fp->max_temp_idx: %d cs->temp_in_use: %d\n", fp->max_temp_idx, cs->temp_in_use);
1358 }
1359
1360 static void update_params(struct r500_fragment_program *fp)
1361 {
1362 struct gl_fragment_program *mp = &fp->mesa_program;
1363
1364 /* Ask Mesa nicely to fill in ParameterValues for us */
1365 if (mp->Base.Parameters)
1366 _mesa_load_state_parameters(fp->ctx, mp->Base.Parameters);
1367 }
1368
1369 static void dumb_shader(struct r500_fragment_program *fp)
1370 {
1371 fp->inst[0].inst0 = R500_INST_TYPE_TEX
1372 | R500_INST_TEX_SEM_WAIT
1373 | R500_INST_RGB_WMASK_R
1374 | R500_INST_RGB_WMASK_G
1375 | R500_INST_RGB_WMASK_B
1376 | R500_INST_ALPHA_WMASK
1377 | R500_INST_RGB_CLAMP
1378 | R500_INST_ALPHA_CLAMP;
1379 fp->inst[0].inst1 = R500_TEX_ID(0)
1380 | R500_TEX_INST_LD
1381 | R500_TEX_SEM_ACQUIRE
1382 | R500_TEX_IGNORE_UNCOVERED;
1383 fp->inst[0].inst2 = R500_TEX_SRC_ADDR(0)
1384 | R500_TEX_SRC_S_SWIZ_R
1385 | R500_TEX_SRC_T_SWIZ_G
1386 | R500_TEX_DST_ADDR(0)
1387 | R500_TEX_DST_R_SWIZ_R
1388 | R500_TEX_DST_G_SWIZ_G
1389 | R500_TEX_DST_B_SWIZ_B
1390 | R500_TEX_DST_A_SWIZ_A;
1391 fp->inst[0].inst3 = R500_DX_ADDR(0)
1392 | R500_DX_S_SWIZ_R
1393 | R500_DX_T_SWIZ_R
1394 | R500_DX_R_SWIZ_R
1395 | R500_DX_Q_SWIZ_R
1396 | R500_DY_ADDR(0)
1397 | R500_DY_S_SWIZ_R
1398 | R500_DY_T_SWIZ_R
1399 | R500_DY_R_SWIZ_R
1400 | R500_DY_Q_SWIZ_R;
1401 fp->inst[0].inst4 = 0x0;
1402 fp->inst[0].inst5 = 0x0;
1403
1404 fp->inst[1].inst0 = R500_INST_TYPE_OUT |
1405 R500_INST_TEX_SEM_WAIT |
1406 R500_INST_LAST |
1407 R500_INST_RGB_OMASK_R |
1408 R500_INST_RGB_OMASK_G |
1409 R500_INST_RGB_OMASK_B |
1410 R500_INST_ALPHA_OMASK;
1411 fp->inst[1].inst1 = R500_RGB_ADDR0(0) |
1412 R500_RGB_ADDR1(0) |
1413 R500_RGB_ADDR1_CONST |
1414 R500_RGB_ADDR2(0) |
1415 R500_RGB_ADDR2_CONST |
1416 R500_RGB_SRCP_OP_1_MINUS_2RGB0;
1417 fp->inst[1].inst2 = R500_ALPHA_ADDR0(0) |
1418 R500_ALPHA_ADDR1(0) |
1419 R500_ALPHA_ADDR1_CONST |
1420 R500_ALPHA_ADDR2(0) |
1421 R500_ALPHA_ADDR2_CONST |
1422 R500_ALPHA_SRCP_OP_1_MINUS_2A0;
1423 fp->inst[1].inst3 = R500_ALU_RGB_SEL_A_SRC0 |
1424 R500_ALU_RGB_R_SWIZ_A_R |
1425 R500_ALU_RGB_G_SWIZ_A_G |
1426 R500_ALU_RGB_B_SWIZ_A_B |
1427 R500_ALU_RGB_SEL_B_SRC0 |
1428 R500_ALU_RGB_R_SWIZ_B_1 |
1429 R500_ALU_RGB_B_SWIZ_B_1 |
1430 R500_ALU_RGB_G_SWIZ_B_1;
1431 fp->inst[1].inst4 = R500_ALPHA_OP_MAD |
1432 R500_ALPHA_SWIZ_A_A |
1433 R500_ALPHA_SWIZ_B_1;
1434 fp->inst[1].inst5 = R500_ALU_RGBA_OP_MAD |
1435 R500_ALU_RGBA_R_SWIZ_0 |
1436 R500_ALU_RGBA_G_SWIZ_0 |
1437 R500_ALU_RGBA_B_SWIZ_0 |
1438 R500_ALU_RGBA_A_SWIZ_0;
1439
1440 fp->cs->nrslots = 2;
1441 fp->translated = GL_TRUE;
1442 }
1443
1444 void r500TranslateFragmentShader(r300ContextPtr r300,
1445 struct r500_fragment_program *fp)
1446 {
1447
1448 struct r300_pfs_compile_state *cs = NULL;
1449
1450 if (!fp->translated) {
1451
1452 init_program(r300, fp);
1453 cs = fp->cs;
1454
1455 if (parse_program(fp) == GL_FALSE) {
1456 ERROR("Huh. Couldn't parse program. There should be additional errors explaining why.\nUsing dumb shader...\n");
1457 dumb_shader(fp);
1458 fp->inst_offset = 0;
1459 fp->inst_end = cs->nrslots - 1;
1460 return;
1461 }
1462 fp->inst_offset = 0;
1463 fp->inst_end = cs->nrslots - 1;
1464
1465 fp->translated = GL_TRUE;
1466 if (RADEON_DEBUG & DEBUG_PIXEL) {
1467 fprintf(stderr, "Mesa program:\n");
1468 fprintf(stderr, "-------------\n");
1469 _mesa_print_program(&fp->mesa_program.Base);
1470 fflush(stdout);
1471 dump_program(fp);
1472 }
1473
1474
1475 r300UpdateStateParameters(fp->ctx, _NEW_PROGRAM);
1476 }
1477
1478 update_params(fp);
1479
1480 }
1481
1482 static char *toswiz(int swiz_val) {
1483 switch(swiz_val) {
1484 case 0: return "R";
1485 case 1: return "G";
1486 case 2: return "B";
1487 case 3: return "A";
1488 case 4: return "0";
1489 case 5: return "1/2";
1490 case 6: return "1";
1491 case 7: return "U";
1492 }
1493 return NULL;
1494 }
1495
1496 static char *toop(int op_val)
1497 {
1498 char *str = NULL;
1499 switch (op_val) {
1500 case 0: str = "MAD"; break;
1501 case 1: str = "DP3"; break;
1502 case 2: str = "DP4"; break;
1503 case 3: str = "D2A"; break;
1504 case 4: str = "MIN"; break;
1505 case 5: str = "MAX"; break;
1506 case 6: str = "Reserved"; break;
1507 case 7: str = "CND"; break;
1508 case 8: str = "CMP"; break;
1509 case 9: str = "FRC"; break;
1510 case 10: str = "SOP"; break;
1511 case 11: str = "MDH"; break;
1512 case 12: str = "MDV"; break;
1513 }
1514 return str;
1515 }
1516
1517 static char *to_alpha_op(int op_val)
1518 {
1519 char *str = NULL;
1520 switch (op_val) {
1521 case 0: str = "MAD"; break;
1522 case 1: str = "DP"; break;
1523 case 2: str = "MIN"; break;
1524 case 3: str = "MAX"; break;
1525 case 4: str = "Reserved"; break;
1526 case 5: str = "CND"; break;
1527 case 6: str = "CMP"; break;
1528 case 7: str = "FRC"; break;
1529 case 8: str = "EX2"; break;
1530 case 9: str = "LN2"; break;
1531 case 10: str = "RCP"; break;
1532 case 11: str = "RSQ"; break;
1533 case 12: str = "SIN"; break;
1534 case 13: str = "COS"; break;
1535 case 14: str = "MDH"; break;
1536 case 15: str = "MDV"; break;
1537 }
1538 return str;
1539 }
1540
1541 static char *to_mask(int val)
1542 {
1543 char *str = NULL;
1544 switch(val) {
1545 case 0: str = "NONE"; break;
1546 case 1: str = "R"; break;
1547 case 2: str = "G"; break;
1548 case 3: str = "RG"; break;
1549 case 4: str = "B"; break;
1550 case 5: str = "RB"; break;
1551 case 6: str = "GB"; break;
1552 case 7: str = "RGB"; break;
1553 case 8: str = "A"; break;
1554 case 9: str = "AR"; break;
1555 case 10: str = "AG"; break;
1556 case 11: str = "ARG"; break;
1557 case 12: str = "AB"; break;
1558 case 13: str = "ARB"; break;
1559 case 14: str = "AGB"; break;
1560 case 15: str = "ARGB"; break;
1561 }
1562 return str;
1563 }
1564
1565 static char *to_texop(int val)
1566 {
1567 switch(val) {
1568 case 0: return "NOP";
1569 case 1: return "LD";
1570 case 2: return "TEXKILL";
1571 case 3: return "PROJ";
1572 case 4: return "LODBIAS";
1573 case 5: return "LOD";
1574 case 6: return "DXDY";
1575 }
1576 return NULL;
1577 }
1578
1579 static void dump_program(struct r500_fragment_program *fp)
1580 {
1581
1582 fprintf(stderr, "R500 Fragment Program:\n--------\n");
1583
1584 int n;
1585 uint32_t inst;
1586 uint32_t inst0;
1587 char *str = NULL;
1588
1589 if (fp->const_nr) {
1590 fprintf(stderr, "--------\nConstants:\n");
1591 for (n = 0; n < fp->const_nr; n++) {
1592 fprintf(stderr, "Constant %d: %f %f\n\t %f %f\n", n,
1593 fp->constant[n][0], fp->constant[n][1], fp->constant[n][2],
1594 fp->constant[n][3]);
1595 }
1596 fprintf(stderr, "--------\n");
1597 }
1598
1599 for (n = 0; n < fp->inst_end+1; n++) {
1600 inst0 = inst = fp->inst[n].inst0;
1601 fprintf(stderr,"%d\t0:CMN_INST 0x%08x:", n, inst);
1602 switch(inst & 0x3) {
1603 case R500_INST_TYPE_ALU: str = "ALU"; break;
1604 case R500_INST_TYPE_OUT: str = "OUT"; break;
1605 case R500_INST_TYPE_FC: str = "FC"; break;
1606 case R500_INST_TYPE_TEX: str = "TEX"; break;
1607 };
1608 fprintf(stderr,"%s %s %s %s %s ", str,
1609 inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "",
1610 inst & R500_INST_LAST ? "LAST" : "",
1611 inst & R500_INST_NOP ? "NOP" : "",
1612 inst & R500_INST_ALU_WAIT ? "ALU WAIT" : "");
1613 fprintf(stderr,"wmask: %s omask: %s\n", to_mask((inst >> 11) & 0xf),
1614 to_mask((inst >> 15) & 0xf));
1615
1616 switch(inst0 & 0x3) {
1617 case 0:
1618 case 1:
1619 fprintf(stderr,"\t1:RGB_ADDR 0x%08x:", fp->inst[n].inst1);
1620 inst = fp->inst[n].inst1;
1621
1622 fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
1623 inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
1624 (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
1625 (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
1626 (inst >> 30));
1627
1628 fprintf(stderr,"\t2:ALPHA_ADDR 0x%08x:", fp->inst[n].inst2);
1629 inst = fp->inst[n].inst2;
1630 fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
1631 inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
1632 (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
1633 (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
1634 (inst >> 30));
1635 fprintf(stderr,"\t3 RGB_INST: 0x%08x:", fp->inst[n].inst3);
1636 inst = fp->inst[n].inst3;
1637 fprintf(stderr,"rgb_A_src:%d %s/%s/%s %d rgb_B_src:%d %s/%s/%s %d\n",
1638 (inst) & 0x3, toswiz((inst >> 2) & 0x7), toswiz((inst >> 5) & 0x7), toswiz((inst >> 8) & 0x7),
1639 (inst >> 11) & 0x3,
1640 (inst >> 13) & 0x3, toswiz((inst >> 15) & 0x7), toswiz((inst >> 18) & 0x7), toswiz((inst >> 21) & 0x7),
1641 (inst >> 24) & 0x3);
1642
1643
1644 fprintf(stderr,"\t4 ALPHA_INST:0x%08x:", fp->inst[n].inst4);
1645 inst = fp->inst[n].inst4;
1646 fprintf(stderr,"%s dest:%d%s alp_A_src:%d %s %d alp_B_src:%d %s %d w:%d\n", to_alpha_op(inst & 0xf),
1647 (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
1648 (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), (inst >> 17) & 0x3,
1649 (inst >> 19) & 0x3, toswiz((inst >> 21) & 0x7), (inst >> 24) & 0x3,
1650 (inst >> 31) & 0x1);
1651
1652 fprintf(stderr,"\t5 RGBA_INST: 0x%08x:", fp->inst[n].inst5);
1653 inst = fp->inst[n].inst5;
1654 fprintf(stderr,"%s dest:%d%s rgb_C_src:%d %s/%s/%s %d alp_C_src:%d %s %d\n", toop(inst & 0xf),
1655 (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
1656 (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), toswiz((inst >> 17) & 0x7), toswiz((inst >> 20) & 0x7),
1657 (inst >> 23) & 0x3,
1658 (inst >> 25) & 0x3, toswiz((inst >> 27) & 0x7), (inst >> 30) & 0x3);
1659 break;
1660 case 2:
1661 break;
1662 case 3:
1663 inst = fp->inst[n].inst1;
1664 fprintf(stderr,"\t1:TEX_INST: 0x%08x: id: %d op:%s, %s, %s %s\n", inst, (inst >> 16) & 0xf,
1665 to_texop((inst >> 22) & 0x7), (inst & (1<<25)) ? "ACQ" : "",
1666 (inst & (1<<26)) ? "IGNUNC" : "", (inst & (1<<27)) ? "UNSCALED" : "SCALED");
1667 inst = fp->inst[n].inst2;
1668 fprintf(stderr,"\t2:TEX_ADDR: 0x%08x: src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n", inst,
1669 inst & 127, inst & (1<<7) ? "(rel)" : "",
1670 toswiz((inst >> 8) & 0x3), toswiz((inst >> 10) & 0x3),
1671 toswiz((inst >> 12) & 0x3), toswiz((inst >> 14) & 0x3),
1672 (inst >> 16) & 127, inst & (1<<23) ? "(rel)" : "",
1673 toswiz((inst >> 24) & 0x3), toswiz((inst >> 26) & 0x3),
1674 toswiz((inst >> 28) & 0x3), toswiz((inst >> 30) & 0x3));
1675
1676 fprintf(stderr,"\t3:TEX_DXDY: 0x%08x\n", fp->inst[n].inst3);
1677 break;
1678 }
1679 fprintf(stderr,"\n");
1680 }
1681
1682 }