summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/nvfx/nvfx_fragprog.c
blob: ee41f03b9b8fd7064f64ff75efeb63f266c427c5 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
#include "pipe/p_context.h"
#include "pipe/p_defines.h"
#include "pipe/p_state.h"
#include "util/u_inlines.h"

#include "pipe/p_shader_tokens.h"
#include "tgsi/tgsi_parse.h"
#include "tgsi/tgsi_util.h"

#include "nvfx_context.h"
#include "nvfx_shader.h"

#define MAX_CONSTS 128
#define MAX_IMM 32
struct nvfx_fpc {
	struct nvfx_fragment_program *fp;

	uint attrib_map[PIPE_MAX_SHADER_INPUTS];

	unsigned r_temps;
	unsigned r_temps_discard;
	struct nvfx_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
	struct nvfx_sreg *r_temp;

	int num_regs;

	unsigned inst_offset;
	unsigned have_const;

	struct {
		int pipe;
		float vals[4];
	} consts[MAX_CONSTS];
	int nr_consts;

	struct nvfx_sreg imm[MAX_IMM];
	unsigned nr_imm;
};

static INLINE struct nvfx_sreg
temp(struct nvfx_fpc *fpc)
{
	int idx = ffs(~fpc->r_temps) - 1;

	if (idx < 0) {
		NOUVEAU_ERR("out of temps!!\n");
		assert(0);
		return nvfx_sr(NVFXSR_TEMP, 0);
	}

	fpc->r_temps |= (1 << idx);
	fpc->r_temps_discard |= (1 << idx);
	return nvfx_sr(NVFXSR_TEMP, idx);
}

static INLINE void
release_temps(struct nvfx_fpc *fpc)
{
	fpc->r_temps &= ~fpc->r_temps_discard;
	fpc->r_temps_discard = 0;
}

static INLINE struct nvfx_sreg
constant(struct nvfx_fpc *fpc, int pipe, float vals[4])
{
	int idx;

	if (fpc->nr_consts == MAX_CONSTS)
		assert(0);
	idx = fpc->nr_consts++;

	fpc->consts[idx].pipe = pipe;
	if (pipe == -1)
		memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float));
	return nvfx_sr(NVFXSR_CONST, idx);
}

#define arith(cc,s,o,d,m,s0,s1,s2) \
	nvfx_fp_arith((cc), (s), NVFX_FP_OP_OPCODE_##o, \
			(d), (m), (s0), (s1), (s2))
#define tex(cc,s,o,u,d,m,s0,s1,s2) \
	nvfx_fp_tex((cc), (s), NVFX_FP_OP_OPCODE_##o, (u), \
		    (d), (m), (s0), none, none)

static void
grow_insns(struct nvfx_fpc *fpc, int size)
{
	struct nvfx_fragment_program *fp = fpc->fp;

	fp->insn_len += size;
	fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len);
}

static void
emit_src(struct nvfx_fpc *fpc, int pos, struct nvfx_sreg src)
{
	struct nvfx_fragment_program *fp = fpc->fp;
	uint32_t *hw = &fp->insn[fpc->inst_offset];
	uint32_t sr = 0;

	switch (src.type) {
	case NVFXSR_INPUT:
		sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT);
		hw[0] |= (src.index << NVFX_FP_OP_INPUT_SRC_SHIFT);
		break;
	case NVFXSR_OUTPUT:
		sr |= NVFX_FP_REG_SRC_HALF;
		/* fall-through */
	case NVFXSR_TEMP:
		sr |= (NVFX_FP_REG_TYPE_TEMP << NVFX_FP_REG_TYPE_SHIFT);
		sr |= (src.index << NVFX_FP_REG_SRC_SHIFT);
		break;
	case NVFXSR_CONST:
		if (!fpc->have_const) {
			grow_insns(fpc, 4);
			fpc->have_const = 1;
		}

		hw = &fp->insn[fpc->inst_offset];
		if (fpc->consts[src.index].pipe >= 0) {
			struct nvfx_fragment_program_data *fpd;

			fp->consts = realloc(fp->consts, ++fp->nr_consts *
					     sizeof(*fpd));
			fpd = &fp->consts[fp->nr_consts - 1];
			fpd->offset = fpc->inst_offset + 4;
			fpd->index = fpc->consts[src.index].pipe;
			memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4);
		} else {
			memcpy(&fp->insn[fpc->inst_offset + 4],
				fpc->consts[src.index].vals,
				sizeof(uint32_t) * 4);
		}

		sr |= (NVFX_FP_REG_TYPE_CONST << NVFX_FP_REG_TYPE_SHIFT);
		break;
	case NVFXSR_NONE:
		sr |= (NVFX_FP_REG_TYPE_INPUT << NVFX_FP_REG_TYPE_SHIFT);
		break;
	default:
		assert(0);
	}

	if (src.negate)
		sr |= NVFX_FP_REG_NEGATE;

	if (src.abs)
		hw[1] |= (1 << (29 + pos));

	sr |= ((src.swz[0] << NVFX_FP_REG_SWZ_X_SHIFT) |
	       (src.swz[1] << NVFX_FP_REG_SWZ_Y_SHIFT) |
	       (src.swz[2] << NVFX_FP_REG_SWZ_Z_SHIFT) |
	       (src.swz[3] << NVFX_FP_REG_SWZ_W_SHIFT));

	hw[pos + 1] |= sr;
}

static void
emit_dst(struct nvfx_fpc *fpc, struct nvfx_sreg dst)
{
	struct nvfx_fragment_program *fp = fpc->fp;
	uint32_t *hw = &fp->insn[fpc->inst_offset];

	switch (dst.type) {
	case NVFXSR_TEMP:
		if (fpc->num_regs < (dst.index + 1))
			fpc->num_regs = dst.index + 1;
		break;
	case NVFXSR_OUTPUT:
		if (dst.index == 1) {
			fp->fp_control |= 0xe;
		} else {
			hw[0] |= NVFX_FP_OP_OUT_REG_HALF;
		}
		break;
	case NVFXSR_NONE:
		hw[0] |= (1 << 30);
		break;
	default:
		assert(0);
	}

	hw[0] |= (dst.index << NVFX_FP_OP_OUT_REG_SHIFT);
}

static void
nvfx_fp_arith(struct nvfx_fpc *fpc, int sat, int op,
	      struct nvfx_sreg dst, int mask,
	      struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2)
{
	struct nvfx_fragment_program *fp = fpc->fp;
	uint32_t *hw;

	fpc->inst_offset = fp->insn_len;
	fpc->have_const = 0;
	grow_insns(fpc, 4);
	hw = &fp->insn[fpc->inst_offset];
	memset(hw, 0, sizeof(uint32_t) * 4);

	if (op == NVFX_FP_OP_OPCODE_KIL)
		fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL;
	hw[0] |= (op << NVFX_FP_OP_OPCODE_SHIFT);
	hw[0] |= (mask << NVFX_FP_OP_OUTMASK_SHIFT);
	hw[2] |= (dst.dst_scale << NVFX_FP_OP_DST_SCALE_SHIFT);

	if (sat)
		hw[0] |= NVFX_FP_OP_OUT_SAT;

	if (dst.cc_update)
		hw[0] |= NVFX_FP_OP_COND_WRITE_ENABLE;
	hw[1] |= (dst.cc_test << NVFX_FP_OP_COND_SHIFT);
	hw[1] |= ((dst.cc_swz[0] << NVFX_FP_OP_COND_SWZ_X_SHIFT) |
		  (dst.cc_swz[1] << NVFX_FP_OP_COND_SWZ_Y_SHIFT) |
		  (dst.cc_swz[2] << NVFX_FP_OP_COND_SWZ_Z_SHIFT) |
		  (dst.cc_swz[3] << NVFX_FP_OP_COND_SWZ_W_SHIFT));

	emit_dst(fpc, dst);
	emit_src(fpc, 0, s0);
	emit_src(fpc, 1, s1);
	emit_src(fpc, 2, s2);
}

static void
nvfx_fp_tex(struct nvfx_fpc *fpc, int sat, int op, int unit,
	    struct nvfx_sreg dst, int mask,
	    struct nvfx_sreg s0, struct nvfx_sreg s1, struct nvfx_sreg s2)
{
	struct nvfx_fragment_program *fp = fpc->fp;

	nvfx_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);

	fp->insn[fpc->inst_offset] |= (unit << NVFX_FP_OP_TEX_UNIT_SHIFT);
	fp->samplers |= (1 << unit);
}

static INLINE struct nvfx_sreg
tgsi_src(struct nvfx_fpc *fpc, const struct tgsi_full_src_register *fsrc)
{
	struct nvfx_sreg src = { 0 };

	switch (fsrc->Register.File) {
	case TGSI_FILE_INPUT:
		src = nvfx_sr(NVFXSR_INPUT,
			      fpc->attrib_map[fsrc->Register.Index]);
		break;
	case TGSI_FILE_CONSTANT:
		src = constant(fpc, fsrc->Register.Index, NULL);
		break;
	case TGSI_FILE_IMMEDIATE:
		assert(fsrc->Register.Index < fpc->nr_imm);
		src = fpc->imm[fsrc->Register.Index];
		break;
	case TGSI_FILE_TEMPORARY:
		src = fpc->r_temp[fsrc->Register.Index];
		break;
	/* NV40 fragprog result regs are just temps, so this is simple */
	case TGSI_FILE_OUTPUT:
		src = fpc->r_result[fsrc->Register.Index];
		break;
	default:
		NOUVEAU_ERR("bad src file\n");
		break;
	}

	src.abs = fsrc->Register.Absolute;
	src.negate = fsrc->Register.Negate;
	src.swz[0] = fsrc->Register.SwizzleX;
	src.swz[1] = fsrc->Register.SwizzleY;
	src.swz[2] = fsrc->Register.SwizzleZ;
	src.swz[3] = fsrc->Register.SwizzleW;
	return src;
}

static INLINE struct nvfx_sreg
tgsi_dst(struct nvfx_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
	switch (fdst->Register.File) {
	case TGSI_FILE_OUTPUT:
		return fpc->r_result[fdst->Register.Index];
	case TGSI_FILE_TEMPORARY:
		return fpc->r_temp[fdst->Register.Index];
	case TGSI_FILE_NULL:
		return nvfx_sr(NVFXSR_NONE, 0);
	default:
		NOUVEAU_ERR("bad dst file %d\n", fdst->Register.File);
		return nvfx_sr(NVFXSR_NONE, 0);
	}
}

static INLINE int
tgsi_mask(uint tgsi)
{
	int mask = 0;

	if (tgsi & TGSI_WRITEMASK_X) mask |= NVFX_FP_MASK_X;
	if (tgsi & TGSI_WRITEMASK_Y) mask |= NVFX_FP_MASK_Y;
	if (tgsi & TGSI_WRITEMASK_Z) mask |= NVFX_FP_MASK_Z;
	if (tgsi & TGSI_WRITEMASK_W) mask |= NVFX_FP_MASK_W;
	return mask;
}

static boolean
nvfx_fragprog_parse_instruction(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
				const struct tgsi_full_instruction *finst)
{
	const struct nvfx_sreg none = nvfx_sr(NVFXSR_NONE, 0);
	struct nvfx_sreg src[3], dst, tmp;
	int mask, sat, unit = 0;
	int ai = -1, ci = -1, ii = -1;
	int i;

	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
		return TRUE;

	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
		const struct tgsi_full_src_register *fsrc;

		fsrc = &finst->Src[i];
		if (fsrc->Register.File == TGSI_FILE_TEMPORARY) {
			src[i] = tgsi_src(fpc, fsrc);
		}
	}

	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
		const struct tgsi_full_src_register *fsrc;

		fsrc = &finst->Src[i];

		switch (fsrc->Register.File) {
		case TGSI_FILE_INPUT:
			if (ai == -1 || ai == fsrc->Register.Index) {
				ai = fsrc->Register.Index;
				src[i] = tgsi_src(fpc, fsrc);
			} else {
				src[i] = temp(fpc);
				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
				      tgsi_src(fpc, fsrc), none, none);
			}
			break;
		case TGSI_FILE_CONSTANT:
			if ((ci == -1 && ii == -1) ||
			    ci == fsrc->Register.Index) {
				ci = fsrc->Register.Index;
				src[i] = tgsi_src(fpc, fsrc);
			} else {
				src[i] = temp(fpc);
				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
				      tgsi_src(fpc, fsrc), none, none);
			}
			break;
		case TGSI_FILE_IMMEDIATE:
			if ((ci == -1 && ii == -1) ||
			    ii == fsrc->Register.Index) {
				ii = fsrc->Register.Index;
				src[i] = tgsi_src(fpc, fsrc);
			} else {
				src[i] = temp(fpc);
				arith(fpc, 0, MOV, src[i], NVFX_FP_MASK_ALL,
				      tgsi_src(fpc, fsrc), none, none);
			}
			break;
		case TGSI_FILE_TEMPORARY:
			/* handled above */
			break;
		case TGSI_FILE_SAMPLER:
			unit = fsrc->Register.Index;
			break;
		case TGSI_FILE_OUTPUT:
			break;
		default:
			NOUVEAU_ERR("bad src file\n");
			return FALSE;
		}
	}

	dst  = tgsi_dst(fpc, &finst->Dst[0]);
	mask = tgsi_mask(finst->Dst[0].Register.WriteMask);
	sat  = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE);

	switch (finst->Instruction.Opcode) {
	case TGSI_OPCODE_ABS:
		arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none);
		break;
	case TGSI_OPCODE_ADD:
		arith(fpc, sat, ADD, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_CMP:
		tmp = nvfx_sr(NVFXSR_NONE, 0);
		tmp.cc_update = 1;
		arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none);
		dst.cc_test = NVFX_COND_GE;
		arith(fpc, sat, MOV, dst, mask, src[2], none, none);
		dst.cc_test = NVFX_COND_LT;
		arith(fpc, sat, MOV, dst, mask, src[1], none, none);
		break;
	case TGSI_OPCODE_COS:
		arith(fpc, sat, COS, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_DDX:
		if (mask & (NVFX_FP_MASK_Z | NVFX_FP_MASK_W)) {
			tmp = temp(fpc);
			arith(fpc, sat, DDX, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y,
			      swz(src[0], Z, W, Z, W), none, none);
			arith(fpc, 0, MOV, tmp, NVFX_FP_MASK_Z | NVFX_FP_MASK_W,
			      swz(tmp, X, Y, X, Y), none, none);
			arith(fpc, sat, DDX, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y, src[0],
			      none, none);
			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
		} else {
			arith(fpc, sat, DDX, dst, mask, src[0], none, none);
		}
		break;
	case TGSI_OPCODE_DDY:
		if (mask & (NVFX_FP_MASK_Z | NVFX_FP_MASK_W)) {
			tmp = temp(fpc);
			arith(fpc, sat, DDY, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y,
			      swz(src[0], Z, W, Z, W), none, none);
			arith(fpc, 0, MOV, tmp, NVFX_FP_MASK_Z | NVFX_FP_MASK_W,
			      swz(tmp, X, Y, X, Y), none, none);
			arith(fpc, sat, DDY, tmp, NVFX_FP_MASK_X | NVFX_FP_MASK_Y, src[0],
			      none, none);
			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
		} else {
			arith(fpc, sat, DDY, dst, mask, src[0], none, none);
		}
		break;
	case TGSI_OPCODE_DP3:
		arith(fpc, sat, DP3, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_DP4:
		arith(fpc, sat, DP4, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_DPH:
		tmp = temp(fpc);
		arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_X, src[0], src[1], none);
		arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X),
		      swz(src[1], W, W, W, W), none);
		break;
	case TGSI_OPCODE_DST:
		arith(fpc, sat, DST, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_EX2:
		arith(fpc, sat, EX2, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_FLR:
		arith(fpc, sat, FLR, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_FRC:
		arith(fpc, sat, FRC, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_KILP:
		arith(fpc, 0, KIL, none, 0, none, none, none);
		break;
	case TGSI_OPCODE_KIL:
		dst = nvfx_sr(NVFXSR_NONE, 0);
		dst.cc_update = 1;
		arith(fpc, 0, MOV, dst, NVFX_FP_MASK_ALL, src[0], none, none);
		dst.cc_update = 0; dst.cc_test = NVFX_COND_LT;
		arith(fpc, 0, KIL, dst, 0, none, none, none);
		break;
	case TGSI_OPCODE_LG2:
		arith(fpc, sat, LG2, dst, mask, src[0], none, none);
		break;
//	case TGSI_OPCODE_LIT:
	case TGSI_OPCODE_LRP:
		if(!nvfx->is_nv4x)
			arith(fpc, sat, LRP_NV30, dst, mask, src[0], src[1], src[2]);
		else {
			tmp = temp(fpc);
			arith(fpc, 0, MAD, tmp, mask, neg(src[0]), src[2], src[2]);
			arith(fpc, sat, MAD, dst, mask, src[0], src[1], tmp);
		}
		break;
	case TGSI_OPCODE_MAD:
		arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]);
		break;
	case TGSI_OPCODE_MAX:
		arith(fpc, sat, MAX, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_MIN:
		arith(fpc, sat, MIN, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_MOV:
		arith(fpc, sat, MOV, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_MUL:
		arith(fpc, sat, MUL, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_POW:
		if(!nvfx->is_nv4x)
			arith(fpc, sat, POW_NV30, dst, mask, src[0], src[1], none);
		else {
			tmp = temp(fpc);
			arith(fpc, 0, LG2, tmp, NVFX_FP_MASK_X,
			      swz(src[0], X, X, X, X), none, none);
			arith(fpc, 0, MUL, tmp, NVFX_FP_MASK_X, swz(tmp, X, X, X, X),
			      swz(src[1], X, X, X, X), none);
			arith(fpc, sat, EX2, dst, mask,
			      swz(tmp, X, X, X, X), none, none);
		}
		break;
	case TGSI_OPCODE_RCP:
		arith(fpc, sat, RCP, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_RET:
		assert(0);
		break;
	case TGSI_OPCODE_RFL:
		if(!nvfx->is_nv4x)
			arith(fpc, 0, RFL_NV30, dst, mask, src[0], src[1], none);
		else {
			tmp = temp(fpc);
			arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_X, src[0], src[0], none);
			arith(fpc, 0, DP3, tmp, NVFX_FP_MASK_Y, src[0], src[1], none);
			arith(fpc, 0, DIV, scale(tmp, 2X), NVFX_FP_MASK_Z,
			      swz(tmp, Y, Y, Y, Y), swz(tmp, X, X, X, X), none);
			arith(fpc, sat, MAD, dst, mask,
			      swz(tmp, Z, Z, Z, Z), src[0], neg(src[1]));
		}
		break;
	case TGSI_OPCODE_RSQ:
		if(!nvfx->is_nv4x)
			arith(fpc, sat, RSQ_NV30, dst, mask, abs(swz(src[0], X, X, X, X)), none, none);
		else {
			tmp = temp(fpc);
			arith(fpc, 0, LG2, scale(tmp, INV_2X), NVFX_FP_MASK_X,
			      abs(swz(src[0], X, X, X, X)), none, none);
			arith(fpc, sat, EX2, dst, mask,
			      neg(swz(tmp, X, X, X, X)), none, none);
		}
		break;
	case TGSI_OPCODE_SCS:
		/* avoid overwriting the source */
		if(src[0].swz[NVFX_SWZ_X] != NVFX_SWZ_X)
		{
			if (mask & NVFX_FP_MASK_X) {
				arith(fpc, sat, COS, dst, NVFX_FP_MASK_X,
				      swz(src[0], X, X, X, X), none, none);
			}
			if (mask & NVFX_FP_MASK_Y) {
				arith(fpc, sat, SIN, dst, NVFX_FP_MASK_Y,
				      swz(src[0], X, X, X, X), none, none);
			}
		}
		else
		{
			if (mask & NVFX_FP_MASK_Y) {
				arith(fpc, sat, SIN, dst, NVFX_FP_MASK_Y,
				      swz(src[0], X, X, X, X), none, none);
			}
			if (mask & NVFX_FP_MASK_X) {
				arith(fpc, sat, COS, dst, NVFX_FP_MASK_X,
				      swz(src[0], X, X, X, X), none, none);
			}
		}
		break;
	case TGSI_OPCODE_SEQ:
		arith(fpc, sat, SEQ, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SFL:
		arith(fpc, sat, SFL, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SGE:
		arith(fpc, sat, SGE, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SGT:
		arith(fpc, sat, SGT, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SIN:
		arith(fpc, sat, SIN, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_SLE:
		arith(fpc, sat, SLE, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SLT:
		arith(fpc, sat, SLT, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SNE:
		arith(fpc, sat, SNE, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_STR:
		arith(fpc, sat, STR, dst, mask, src[0], src[1], none);
		break;
	case TGSI_OPCODE_SUB:
		arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none);
		break;
	case TGSI_OPCODE_TEX:
		tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_TXB:
		tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_TXP:
		tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none);
		break;
	case TGSI_OPCODE_XPD:
		tmp = temp(fpc);
		arith(fpc, 0, MUL, tmp, mask,
		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
		arith(fpc, sat, MAD, dst, (mask & ~NVFX_FP_MASK_W),
		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
		      neg(tmp));
		break;
	default:
		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
		return FALSE;
	}

	release_temps(fpc);
	return TRUE;
}

static boolean
nvfx_fragprog_parse_decl_attrib(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
				const struct tgsi_full_declaration *fdec)
{
	int hw;

	switch (fdec->Semantic.Name) {
	case TGSI_SEMANTIC_POSITION:
		hw = NVFX_FP_OP_INPUT_SRC_POSITION;
		break;
	case TGSI_SEMANTIC_COLOR:
		if (fdec->Semantic.Index == 0) {
			hw = NVFX_FP_OP_INPUT_SRC_COL0;
		} else
		if (fdec->Semantic.Index == 1) {
			hw = NVFX_FP_OP_INPUT_SRC_COL1;
		} else {
			NOUVEAU_ERR("bad colour semantic index\n");
			return FALSE;
		}
		break;
	case TGSI_SEMANTIC_FOG:
		hw = NVFX_FP_OP_INPUT_SRC_FOGC;
		break;
	case TGSI_SEMANTIC_GENERIC:
		if (fdec->Semantic.Index <= 7) {
			hw = NVFX_FP_OP_INPUT_SRC_TC(fdec->Semantic.
						     Index);
		} else {
			NOUVEAU_ERR("bad generic semantic index\n");
			return FALSE;
		}
		break;
	default:
		NOUVEAU_ERR("bad input semantic\n");
		return FALSE;
	}

	fpc->attrib_map[fdec->Range.First] = hw;
	return TRUE;
}

static boolean
nvfx_fragprog_parse_decl_output(struct nvfx_context* nvfx, struct nvfx_fpc *fpc,
				const struct tgsi_full_declaration *fdec)
{
	unsigned idx = fdec->Range.First;
	unsigned hw;

	switch (fdec->Semantic.Name) {
	case TGSI_SEMANTIC_POSITION:
		hw = 1;
		break;
	case TGSI_SEMANTIC_COLOR:
		hw = ~0;
		switch (fdec->Semantic.Index) {
		case 0: hw = 0; break;
		case 1: hw = 2; break;
		case 2: hw = 3; break;
		case 3: hw = 4; break;
		}
		if(hw > ((nvfx->is_nv4x) ? 4 : 2)) {
			NOUVEAU_ERR("bad rcol index\n");
			return FALSE;
		}
		break;
	default:
		NOUVEAU_ERR("bad output semantic\n");
		return FALSE;
	}

	fpc->r_result[idx] = nvfx_sr(NVFXSR_OUTPUT, hw);
	fpc->r_temps |= (1 << hw);
	return TRUE;
}

static boolean
nvfx_fragprog_prepare(struct nvfx_context* nvfx, struct nvfx_fpc *fpc)
{
	struct tgsi_parse_context p;
	int high_temp = -1, i;

	tgsi_parse_init(&p, fpc->fp->pipe.tokens);
	while (!tgsi_parse_end_of_tokens(&p)) {
		const union tgsi_full_token *tok = &p.FullToken;

		tgsi_parse_token(&p);
		switch(tok->Token.Type) {
		case TGSI_TOKEN_TYPE_DECLARATION:
		{
			const struct tgsi_full_declaration *fdec;
			fdec = &p.FullToken.FullDeclaration;
			switch (fdec->Declaration.File) {
			case TGSI_FILE_INPUT:
				if (!nvfx_fragprog_parse_decl_attrib(nvfx, fpc, fdec))
					goto out_err;
				break;
			case TGSI_FILE_OUTPUT:
				if (!nvfx_fragprog_parse_decl_output(nvfx, fpc, fdec))
					goto out_err;
				break;
			case TGSI_FILE_TEMPORARY:
				if (fdec->Range.Last > high_temp) {
					high_temp =
						fdec->Range.Last;
				}
				break;
			default:
				break;
			}
		}
			break;
		case TGSI_TOKEN_TYPE_IMMEDIATE:
		{
			struct tgsi_full_immediate *imm;
			float vals[4];

			imm = &p.FullToken.FullImmediate;
			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
			assert(fpc->nr_imm < MAX_IMM);

			vals[0] = imm->u[0].Float;
			vals[1] = imm->u[1].Float;
			vals[2] = imm->u[2].Float;
			vals[3] = imm->u[3].Float;
			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
		}
			break;
		default:
			break;
		}
	}
	tgsi_parse_free(&p);

	if (++high_temp) {
		fpc->r_temp = CALLOC(high_temp, sizeof(struct nvfx_sreg));
		for (i = 0; i < high_temp; i++)
			fpc->r_temp[i] = temp(fpc);
		fpc->r_temps_discard = 0;
	}

	return TRUE;

out_err:
	if (fpc->r_temp)
		FREE(fpc->r_temp);
	tgsi_parse_free(&p);
	return FALSE;
}

static void
nvfx_fragprog_translate(struct nvfx_context *nvfx,
			struct nvfx_fragment_program *fp)
{
	struct tgsi_parse_context parse;
	struct nvfx_fpc *fpc = NULL;

	fpc = CALLOC(1, sizeof(struct nvfx_fpc));
	if (!fpc)
		return;
	fpc->fp = fp;
	fpc->num_regs = 2;

	if (!nvfx_fragprog_prepare(nvfx, fpc)) {
		FREE(fpc);
		return;
	}

	tgsi_parse_init(&parse, fp->pipe.tokens);

	while (!tgsi_parse_end_of_tokens(&parse)) {
		tgsi_parse_token(&parse);

		switch (parse.FullToken.Token.Type) {
		case TGSI_TOKEN_TYPE_INSTRUCTION:
		{
			const struct tgsi_full_instruction *finst;

			finst = &parse.FullToken.FullInstruction;
			if (!nvfx_fragprog_parse_instruction(nvfx, fpc, finst))
				goto out_err;
		}
			break;
		default:
			break;
		}
	}

	if(!nvfx->is_nv4x)
		fp->fp_control |= (fpc->num_regs-1)/2;
	else
		fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT;

	/* Terminate final instruction */
	if(fp->insn)
                fp->insn[fpc->inst_offset] |= 0x00000001;

	/* Append NOP + END instruction, may or may not be necessary. */
	fpc->inst_offset = fp->insn_len;
	grow_insns(fpc, 4);
	fp->insn[fpc->inst_offset + 0] = 0x00000001;
	fp->insn[fpc->inst_offset + 1] = 0x00000000;
	fp->insn[fpc->inst_offset + 2] = 0x00000000;
	fp->insn[fpc->inst_offset + 3] = 0x00000000;

	fp->translated = TRUE;
out_err:
	tgsi_parse_free(&parse);
	if (fpc->r_temp)
		FREE(fpc->r_temp);
	FREE(fpc);
}

static inline void
nvfx_fp_memcpy(void* dst, const void* src, size_t len)
{
#ifndef WORDS_BIGENDIAN
	memcpy(dst, src, len);
#else
	size_t i;
	for(i = 0; i < len; i += 4) {
		uint32_t v = (uint32_t*)((char*)src + i);
		*(uint32_t*)((char*)dst + i) = (v >> 16) | (v << 16);
	}
#endif
}

void
nvfx_fragprog_validate(struct nvfx_context *nvfx)
{
	struct nouveau_channel* chan = nvfx->screen->base.channel;
	struct nvfx_fragment_program *fp = nvfx->fragprog;
	int update = 0;

	if (!fp->translated)
	{
		const int min_size = 4096;

		nvfx_fragprog_translate(nvfx, fp);
		if (!fp->translated) {
			static unsigned dummy[8] = {1, 0, 0, 0, 1, 0, 0, 0};
			static int warned = 0;
			if(!warned)
			{
				fprintf(stderr, "nvfx: failed to translate fragment program!\n");
				warned = 1;
			}

			/* use dummy program: we cannot fail here */
			fp->translated = TRUE;
			fp->insn = malloc(sizeof(dummy));
			memcpy(fp->insn, dummy, sizeof(dummy));
			fp->insn_len = sizeof(dummy) / sizeof(dummy[0]);
		}
		update = TRUE;

		fp->prog_size = (fp->insn_len * 4 + 63) & ~63;

		if(fp->prog_size >= min_size)
			fp->progs_per_bo = 1;
		else
			fp->progs_per_bo = min_size / fp->prog_size;
		fp->bo_prog_idx = fp->progs_per_bo - 1;
	}

	/* we must update constants even on "just" fragprog changes, because
	   we don't check whether the current constant buffer matches the latest
	   one bound to this fragment program */
	if (nvfx->dirty & (NVFX_NEW_FRAGCONST | NVFX_NEW_FRAGPROG))
		update = TRUE;

	if(update) {
		int offset;

		++fp->bo_prog_idx;
		if(fp->bo_prog_idx >= fp->progs_per_bo)
		{
			if(fp->fpbo && !nouveau_bo_busy(fp->fpbo->next->bo, NOUVEAU_BO_WR))
			{
				fp->fpbo = fp->fpbo->next;
			}
			else
			{
				struct nvfx_fragment_program_bo* fpbo = os_malloc_aligned(sizeof(struct nvfx_fragment_program) + fp->prog_size * fp->progs_per_bo, 16);
				char *map, *buf;
				int i;

				if(fp->fpbo)
				{
					fpbo->next = fp->fpbo->next;
					fp->fpbo->next = fpbo;
				}
				else
					fpbo->next = fpbo;
				fp->fpbo = fpbo;
				fpbo->bo = 0;
				nouveau_bo_new(nvfx->screen->base.device, NOUVEAU_BO_VRAM | NOUVEAU_BO_MAP, 64, fp->prog_size * fp->progs_per_bo, &fpbo->bo);
				nouveau_bo_map(fpbo->bo, NOUVEAU_BO_NOSYNC);

				map = fpbo->bo->map;
				buf = fpbo->insn;
				for(i = 0; i < fp->progs_per_bo; ++i)
				{
					memcpy(buf, fp->insn, fp->insn_len * 4);
					nvfx_fp_memcpy(map, fp->insn, fp->insn_len * 4);
					map += fp->prog_size;
					buf += fp->prog_size;
				}
			}
			fp->bo_prog_idx = 0;
		}

		offset = fp->bo_prog_idx * fp->prog_size;

		if(nvfx->constbuf[PIPE_SHADER_FRAGMENT]) {
			struct pipe_resource* constbuf = nvfx->constbuf[PIPE_SHADER_FRAGMENT];
			// TODO: avoid using transfers, just directly the buffer
			struct pipe_transfer* transfer;
			// TODO: does this check make any sense, or should we do this unconditionally?
			uint32_t* map = pipe_buffer_map(&nvfx->pipe, constbuf, PIPE_TRANSFER_READ, &transfer);
			uint32_t* fpmap = (uint32_t*)((char*)fp->fpbo->bo->map + offset);
			uint32_t* buf = (uint32_t*)((char*)fp->fpbo->insn + offset);
			int i;
			for (i = 0; i < fp->nr_consts; ++i) {
				unsigned off = fp->consts[i].offset;
				unsigned idx = fp->consts[i].index * 4;

				/* TODO: is checking a good idea? */
				if(memcmp(&buf[off], &map[idx], 4 * sizeof(uint32_t))) {
					memcpy(&buf[off], &map[idx], 4 * sizeof(uint32_t));
					nvfx_fp_memcpy(&fpmap[off], &map[idx], 4 * sizeof(uint32_t));
				}
			}
			pipe_buffer_unmap(&nvfx->pipe, constbuf, transfer);
		}
	}

	if(update || (nvfx->dirty & NVFX_NEW_FRAGPROG)) {
		int offset = fp->bo_prog_idx * fp->prog_size;
		MARK_RING(chan, 8, 1);
		OUT_RING(chan, RING_3D(NV34TCL_FP_ACTIVE_PROGRAM, 1));
		OUT_RELOC(chan, fp->fpbo->bo, offset, NOUVEAU_BO_VRAM |
			      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
			      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
			      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
		OUT_RING(chan, RING_3D(NV34TCL_FP_CONTROL, 1));
		OUT_RING(chan, fp->fp_control);
		if(!nvfx->is_nv4x) {
			OUT_RING(chan, RING_3D(NV34TCL_FP_REG_CONTROL, 1));
			OUT_RING(chan, (1<<16)|0x4);
			OUT_RING(chan, RING_3D(NV34TCL_TX_UNITS_ENABLE, 1));
			OUT_RING(chan, fp->samplers);
		}
	}
}

void
nvfx_fragprog_relocate(struct nvfx_context *nvfx)
{
	struct nouveau_channel* chan = nvfx->screen->base.channel;
	struct nvfx_fragment_program *fp = nvfx->fragprog;
	struct nouveau_bo* bo = fp->fpbo->bo;
	int offset = fp->bo_prog_idx * fp->prog_size;
	unsigned fp_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD; // TODO: GART?
	fp_flags |= NOUVEAU_BO_DUMMY;
	MARK_RING(chan, 2, 2);
	OUT_RELOC(chan, bo, RING_3D(NV34TCL_FP_ACTIVE_PROGRAM, 1), fp_flags, 0, 0);
	OUT_RELOC(chan, bo, offset, fp_flags | NOUVEAU_BO_LOW |
		      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
		      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
}

void
nvfx_fragprog_destroy(struct nvfx_context *nvfx,
		      struct nvfx_fragment_program *fp)
{
	struct nvfx_fragment_program_bo* fpbo = fp->fpbo;
	if(fpbo)
	{
		do
		{
			struct nvfx_fragment_program_bo* next = fpbo->next;
			nouveau_bo_unmap(fpbo->bo);
			nouveau_bo_ref(0, &fpbo->bo);
			free(fpbo);
			fpbo = next;
		}
		while(fpbo != fp->fpbo);
	}

	if (fp->insn_len)
		FREE(fp->insn);
}