/
mojoshader.c
2614 lines (2161 loc) · 76.3 KB
1
/**
2
3
* MojoShader; generate shader programs from bytecode of compiled
* Direct3D shaders.
4
5
6
7
8
9
*
* Please see the file LICENSE.txt in the source's root directory.
*
* This file written by Ryan C. Gordon.
*/
10
11
// !!! FIXME: I keep changing coding styles for symbols and typedefs.
12
13
14
// !!! FIXME: do DEF* and DCL_* opcodes have to come before instructions?
// !!! FIXME: my reading of the msdn spec suggests no.
15
16
17
// Shader bytecode format is described at MSDN:
// http://msdn2.microsoft.com/en-us/library/ms800307.aspx
18
#include <stdio.h>
19
#include <string.h>
20
#include <stdlib.h>
21
#include <stdint.h>
22
#include <stdarg.h>
23
#include <assert.h>
24
25
#include "mojoshader.h"
26
27
28
29
30
31
// This is the highest shader version we currently support.
#define MAX_SHADER_MAJOR 3
#define MAX_SHADER_MINOR 0
32
33
34
// You get all the profiles unless you go out of your way to disable them.
35
36
37
38
39
40
41
42
#ifndef SUPPORT_PROFILE_D3D
#define SUPPORT_PROFILE_D3D 1
#endif
#ifndef SUPPORT_PROFILE_GLSL
#define SUPPORT_PROFILE_GLSL 1
#endif
43
44
45
46
47
48
// Get basic wankery out of the way here...
typedef unsigned int uint; // this is a printf() helper. don't use for code.
typedef uint8_t uint8;
typedef uint32_t uint32;
49
typedef int32_t int32;
50
51
52
53
54
55
56
#ifdef __GNUC__
#define ISPRINTF(x,y) __attribute__((format (printf, x, y)))
#else
#define ISPRINTF(x,y)
#endif
57
58
#define STATICARRAYLEN(x) ( (sizeof ((x))) / (sizeof ((x)[0])) )
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
// Byteswap magic...
#if ((defined __GNUC__) && (defined __POWERPC__))
static inline uint32 SWAP32(uint32 x)
{
__asm__ __volatile__("lwbrx %0,0,%1" : "=r" (x) : "r" (&x));
return x;
} // SWAP32
#elif defined(__POWERPC__)
static inline uint32 SWAP32(uint32 x)
{
return ( (((x) >> 24) & 0x000000FF) | (((x) >> 8) & 0x0000FF00) |
(((x) << 8) & 0x00FF0000) | (((x) << 24) & 0xFF000000) );
} // SWAP32
#else
# define SWAP32(x) (x)
#endif
78
79
// predeclare.
80
typedef struct Context Context;
81
82
// one emit function for each opcode in each profile.
83
typedef void (*emit_function)(Context *ctx);
84
85
// one emit function for comments in each profile.
86
typedef void (*emit_comment)(Context *ctx, const char *str);
87
88
// one emit function for starting output in each profile.
89
typedef void (*emit_start)(Context *ctx);
90
91
// one emit function for ending output in each profile.
92
typedef void (*emit_end)(Context *ctx);
93
94
95
96
// one args function for each possible sequence of opcode arguments.
typedef int (*args_function)(Context *ctx);
97
// one state function for each opcode where we have state machine updates.
98
typedef void (*state_function)(Context *ctx);
99
100
101
102
103
104
105
typedef struct
{
const char *name;
emit_start start_emitter;
emit_end end_emitter;
emit_comment comment_emitter;
106
} Profile;
107
108
109
110
// These are enum values, but they also can be used in bitmasks, so we can
// test if an opcode is acceptable: if (op->shader_types & ourtype) {} ...
111
112
typedef enum
{
113
114
115
116
117
SHADER_TYPE_UNKNOWN = 0,
SHADER_TYPE_PIXEL = (1 << 0),
SHADER_TYPE_VERTEX = (1 << 1),
SHADER_TYPE_GEOMETRY = (1 << 2),
SHADER_TYPE_ANY = 0xFFFFFFFF
118
} ShaderType;
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
typedef enum
{
REGISTER_TYPE_TEMP = 0,
REGISTER_TYPE_INPUT = 1,
REGISTER_TYPE_CONST = 2,
REGISTER_TYPE_ADDR = 3,
REGISTER_TYPE_TEXTURE = 3, // ALSO 3!
REGISTER_TYPE_RASTOUT = 4,
REGISTER_TYPE_ATTROUT = 5,
REGISTER_TYPE_TEXCRDOUT = 6,
REGISTER_TYPE_OUTPUT = 6, // ALSO 6!
REGISTER_TYPE_CONSTINT = 7,
REGISTER_TYPE_COLOROUT = 8,
REGISTER_TYPE_DEPTHOUT = 9,
REGISTER_TYPE_SAMPLER = 10,
REGISTER_TYPE_CONST2 = 11,
REGISTER_TYPE_CONST3 = 12,
REGISTER_TYPE_CONST4 = 13,
REGISTER_TYPE_CONSTBOOL = 14,
REGISTER_TYPE_LOOP = 15,
REGISTER_TYPE_TEMPFLOAT16 = 16,
REGISTER_TYPE_MISCTYPE = 17,
REGISTER_TYPE_LABEL = 18,
REGISTER_TYPE_PREDICATE = 19,
REGISTER_TYPE_MAX = 19
145
} RegisterType;
146
147
148
149
150
151
152
typedef enum
{
RASTOUT_TYPE_POSITION = 0,
RASTOUT_TYPE_FOG = 1,
RASTOUT_TYPE_POINT_SIZE = 2,
RASTOUT_TYPE_MAX = 2
153
} RastOutType;
154
155
156
157
158
159
typedef enum
{
MISCTYPE_TYPE_POSITION = 0,
MISCTYPE_TYPE_FACE = 1,
MISCTYPE_TYPE_MAX = 1
160
} MiscTypeType;
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
typedef enum
{
DECLUSAGE_POSITION = 0,
DECLUSAGE_BLENDWEIGHT = 1,
DECLUSAGE_BLENDINDICES = 2,
DECLUSAGE_NORMAL = 3,
DECLUSAGE_PSIZE = 4,
DECLUSAGE_TEXCOORD = 5,
DECLUSAGE_TANGENT = 6,
DECLUSAGE_BINORMAL = 7,
DECLUSAGE_TESSFACTOR = 8,
DECLUSAGE_POSITIONT = 9,
DECLUSAGE_COLOR = 10,
DECLUSAGE_FOG = 11,
DECLUSAGE_DEPTH = 12,
DECLUSAGE_SAMPLE = 13
178
} DeclUsageType;
179
180
181
182
183
184
typedef enum
{
TEXTURE_TYPE_2D = 2,
TEXTURE_TYPE_CUBE = 3,
TEXTURE_TYPE_VOLUME = 4,
185
} TextureType;
186
187
188
189
190
191
192
193
194
195
196
// A simple linked list of strings, so we can build the final output without
// realloc()ing for each new line, and easily insert lines into the middle
// of the output without much trouble.
typedef struct OutputList
{
char *str;
struct OutputList *next;
} OutputList;
197
198
199
200
201
202
// result modifiers.
#define MOD_SATURATE 0x01
#define MOD_PP 0x02
#define MOD_CENTROID 0x04
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
// source modifiers.
typedef enum
{
SRCMOD_NONE,
SRCMOD_NEGATE,
SRCMOD_BIAS,
SRCMOD_BIASNEGATE,
SRCMOD_SIGN,
SRCMOD_SIGNNEGATE,
SRCMOD_COMPLEMENT,
SRCMOD_X2,
SRCMOD_X2NEGATE,
SRCMOD_DZ,
SRCMOD_DW,
SRCMOD_ABS,
SRCMOD_ABSNEGATE,
SRCMOD_NOT,
SRCMOD_TOTAL
221
} SourceMod;
222
223
224
225
typedef struct
{
226
const uint32 *token; // this is the unmolested token in the stream.
227
228
int regnum;
int relative;
229
230
231
232
233
int writemask; // xyzw or rgba (all four, not split out).
int writemask0; // x or red
int writemask1; // y or green
int writemask2; // z or blue
int writemask3; // w or alpha
234
235
236
237
238
239
240
int result_mod;
int result_shift;
int regtype;
} DestArgInfo;
typedef struct
{
241
const uint32 *token; // this is the unmolested token in the stream.
242
243
int regnum;
int relative;
244
int swizzle; // xyzw (all four, not split out).
245
246
247
248
249
250
251
252
253
int swizzle_x;
int swizzle_y;
int swizzle_z;
int swizzle_w;
int src_mod;
int regtype;
} SourceArgInfo;
254
255
#define SCRATCH_BUFFER_SIZE 256
#define SCRATCH_BUFFERS 10
256
257
// Context...this is state that changes as we parse through a shader...
258
struct Context
259
{
260
261
MOJOSHADER_malloc malloc;
MOJOSHADER_free free;
262
263
const uint32 *tokens;
uint32 tokencount;
264
265
266
OutputList output;
OutputList *output_tail;
int output_len; // total strlen; prevents walking the list just to malloc.
267
int indent;
268
269
270
const char *endline;
int endline_len;
const char *failstr;
271
char scratch[SCRATCH_BUFFERS][SCRATCH_BUFFER_SIZE];
272
273
int scratchidx; // current scratch buffer.
int profileid;
274
275
const Profile *profile;
ShaderType shader_type;
276
277
uint32 major_ver;
uint32 minor_ver;
278
279
DestArgInfo dest_args[1];
SourceArgInfo source_args[4];
280
uint32 dwords[4];
281
uint32 instruction_count;
282
uint32 instruction_controls;
283
284
285
};
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
// Shader model version magic.
static inline uint32 ver_ui32(const uint8 major, const uint8 minor)
{
return ( (((uint32) major) << 16) | (((minor) == 0xFF) ? 0 : (minor)) );
} // version_ui32
static int shader_version_supported(uint8 maj, uint8 min)
{
return (ver_ui32(maj,min) <= ver_ui32(MAX_SHADER_MAJOR, MAX_SHADER_MINOR));
} // shader_version_supported
static int shader_version_atleast(const Context *ctx, uint8 maj, uint8 min)
{
return (ver_ui32(ctx->major_ver, ctx->minor_ver) >= ver_ui32(maj, min));
} // shader_version_atleast
304
static inline char *get_scratch_buffer(Context *ctx)
305
{
306
ctx->scratchidx = (ctx->scratchidx + 1) % SCRATCH_BUFFERS;
307
return ctx->scratch[ctx->scratchidx];
308
309
} // get_scratch_buffer
310
311
312
// Special-case return values from the parsing pipeline...
#define FAIL (-1)
313
314
#define NOFAIL (-2)
#define END_OF_STREAM (-3)
315
316
317
318
319
320
321
static inline isfail(const Context *ctx)
{
return (ctx->failstr != NULL);
} // isfail
322
323
324
325
326
327
328
329
330
331
332
static const char *out_of_mem_string = "Out of memory";
static inline int out_of_memory(Context *ctx)
{
if (ctx->failstr == NULL)
ctx->failstr = out_of_mem_string; // fail() would call malloc().
return FAIL;
} // out_of_memory
static int failf(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static int failf(Context *ctx, const char *fmt, ...)
333
{
334
335
if (ctx->failstr == NULL) // don't change existing error.
{
336
char *scratch = get_scratch_buffer(ctx);
337
338
va_list ap;
va_start(ap, fmt);
339
const int len = vsnprintf(scratch,SCRATCH_BUFFER_SIZE,fmt,ap);
340
341
va_end(ap);
342
343
344
345
346
347
char *failstr = (char *) ctx->malloc(len + 1);
if (failstr == NULL)
out_of_memory(ctx);
else
{
// see comments about scratch buffer overflow in output_line().
348
if (len < SCRATCH_BUFFER_SIZE)
349
350
351
352
353
354
355
356
357
strcpy(failstr, scratch); // copy it over.
else
{
va_start(ap, fmt);
vsnprintf(failstr, len + 1, fmt, ap); // rebuild it.
va_end(ap);
} // else
ctx->failstr = failstr;
} // else
358
359
360
361
362
363
} // if
return FAIL;
} // failf
364
static inline int fail(Context *ctx, const char *reason)
365
366
367
368
369
{
return failf(ctx, "%s", reason);
} // fail
370
371
static int output_line(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static int output_line(Context *ctx, const char *fmt, ...)
372
{
373
if (isfail(ctx))
374
375
return FAIL; // we failed previously, don't go on...
376
377
378
379
380
OutputList *item = (OutputList *) ctx->malloc(sizeof (OutputList));
if (item == NULL)
return out_of_memory(ctx);
char *scratch = get_scratch_buffer(ctx);
381
382
const int indent = ctx->indent;
383
384
if (indent > 0)
memset(scratch, '\t', indent);
385
386
va_list ap;
va_start(ap, fmt);
387
const int len = vsnprintf(scratch+indent, SCRATCH_BUFFER_SIZE-indent, fmt, ap) + indent;
388
389
va_end(ap);
390
391
item->str = (char *) ctx->malloc(len + 1);
if (item->str == NULL)
392
{
393
394
free(item);
return out_of_memory(ctx);
395
396
} // if
397
398
399
// If we overflowed our scratch buffer, that's okay. We were going to
// allocate anyhow...the scratch buffer just lets us avoid a second
// run of vsnprintf().
400
if (len < SCRATCH_BUFFER_SIZE)
401
402
403
strcpy(item->str, scratch); // copy it over.
else
{
404
405
if (indent > 0)
memset(item->str, '\t', indent);
406
va_start(ap, fmt);
407
vsnprintf(item->str+indent, len + 1, fmt, ap); // rebuild it.
408
409
va_end(ap);
} // else
410
411
412
413
414
item->next = NULL;
ctx->output_tail->next = item;
ctx->output_tail = item;
ctx->output_len += len + ctx->endline_len;
415
416
417
418
return 0;
} // output_line
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
// !!! FIXME: this is sort of nasty.
static void floatstr(Context *ctx, char *buf, size_t bufsize, float f)
{
const size_t len = snprintf(buf, bufsize, "%f", f);
if (len >= bufsize)
fail(ctx, "BUG: internal buffer is too small");
else
{
char *end = buf + len;
char *ptr = strchr(buf, '.');
if (ptr == NULL)
return; // done.
while (--end != ptr)
{
if (*end != '0')
{
end++;
break;
} // if
} // while
*end = '\0'; // chop extra '0' or all decimal places off.
} // else
} // floatstr
444
445
// if SUPPORT_PROFILE_* isn't defined, we assume an implicit desire to support.
446
447
448
449
450
451
452
453
454
#define AT_LEAST_ONE_PROFILE 0
#if !SUPPORT_PROFILE_D3D
#define PROFILE_EMITTER_D3D(op)
#else
#undef AT_LEAST_ONE_PROFILE
#define AT_LEAST_ONE_PROFILE 1
#define PROFILE_EMITTER_D3D(op) emit_D3D_##op,
455
static const char *get_D3D_register_string(Context *ctx,
456
RegisterType regtype,
457
458
int regnum, char *regnum_str,
size_t regnum_size)
459
{
460
const char *retval = NULL;
461
int has_number = 1;
462
463
switch (regtype)
464
465
{
case REGISTER_TYPE_TEMP:
466
retval = "r";
467
468
469
break;
case REGISTER_TYPE_INPUT:
470
retval = "v";
471
472
473
break;
case REGISTER_TYPE_CONST:
474
475
476
retval = "c";
break;
477
case REGISTER_TYPE_CONST2:
478
479
480
481
retval = "c";
regnum += 2048;
break;
482
case REGISTER_TYPE_CONST3:
483
484
485
486
retval = "c";
regnum += 4096;
break;
487
case REGISTER_TYPE_CONST4:
488
retval = "c";
489
regnum += 6144;
490
491
492
break;
case REGISTER_TYPE_ADDR: // (or REGISTER_TYPE_TEXTURE, same value.)
493
retval = (ctx->shader_type == SHADER_TYPE_VERTEX) ? "a" : "t";
494
495
496
break;
case REGISTER_TYPE_RASTOUT:
497
switch ((RastOutType) regnum)
498
{
499
500
501
case RASTOUT_TYPE_POSITION: retval = "oPos"; break;
case RASTOUT_TYPE_FOG: retval = "oFog"; break;
case RASTOUT_TYPE_POINT_SIZE: retval = "oPts"; break;
502
} // switch
503
has_number = 0;
504
505
506
break;
case REGISTER_TYPE_ATTROUT:
507
retval = "oD";
508
509
510
511
break;
case REGISTER_TYPE_TEXCRDOUT: // (or REGISTER_TYPE_OUTPUT, same value.)
if ((ctx->shader_type==SHADER_TYPE_VERTEX) && (ctx->major_ver>=3))
512
retval = "o";
513
else
514
retval = "oT";
515
516
517
break;
case REGISTER_TYPE_CONSTINT:
518
retval = "i";
519
520
521
break;
case REGISTER_TYPE_COLOROUT:
522
retval = "oC";
523
524
525
break;
case REGISTER_TYPE_DEPTHOUT:
526
retval = "oDepth";
527
has_number = 0;
528
529
530
break;
case REGISTER_TYPE_SAMPLER:
531
retval = "s";
532
533
534
break;
case REGISTER_TYPE_CONSTBOOL:
535
retval = "b";
536
537
538
break;
case REGISTER_TYPE_LOOP:
539
retval = "aL";
540
has_number = 0;
541
542
543
544
545
546
break;
// !!! FIXME: don't know what the asm string is for this..
// case REGISTER_TYPE_TEMPFLOAT16:
case REGISTER_TYPE_MISCTYPE:
547
switch ((MiscTypeType) regnum)
548
{
549
550
case MISCTYPE_TYPE_POSITION: retval = "vPos"; break;
case MISCTYPE_TYPE_FACE: retval = "vFace"; break;
551
} // switch
552
has_number = 0;
553
554
555
break;
case REGISTER_TYPE_LABEL:
556
retval = "l";
557
558
559
break;
case REGISTER_TYPE_PREDICATE:
560
retval = "p";
561
562
563
break;
} // switch
564
565
566
567
568
if (has_number)
snprintf(regnum_str, regnum_size, "%u", (uint) regnum);
else
regnum_str[0] = '\0';
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
return retval;
} // get_D3D_register_string
static char *make_D3D_destarg_string(Context *ctx, const int idx)
{
if (idx >= STATICARRAYLEN(ctx->dest_args))
{
fail(ctx, "Too many destination args");
return "";
} // if
const DestArgInfo *arg = &ctx->dest_args[idx];
const char *result_shift_str = "";
switch (arg->result_shift)
{
case 0x1: result_shift_str = "_x2"; break;
case 0x2: result_shift_str = "_x4"; break;
case 0x3: result_shift_str = "_x8"; break;
case 0xD: result_shift_str = "_d8"; break;
case 0xE: result_shift_str = "_d4"; break;
case 0xF: result_shift_str = "_d2"; break;
} // switch
const char *sat_str = (arg->result_mod & MOD_SATURATE) ? "_sat" : "";
const char *pp_str = (arg->result_mod & MOD_PP) ? "_pp" : "";
const char *cent_str = (arg->result_mod & MOD_CENTROID) ? "_centroid" : "";
char regnum_str[16];
const char *regtype_str = get_D3D_register_string(ctx,
600
(RegisterType) arg->regtype,
601
602
arg->regnum, regnum_str,
sizeof (regnum_str));
603
604
605
606
607
608
if (regtype_str == NULL)
{
fail(ctx, "Unknown destination register type.");
return "";
} // if
609
610
611
612
613
614
615
616
617
618
619
620
char writemask_str[6];
int i = 0;
if (arg->writemask != 0xF) // 0xF == 1111. No explicit mask.
{
writemask_str[i++] = '.';
if (arg->writemask0) writemask_str[i++] = 'x';
if (arg->writemask1) writemask_str[i++] = 'y';
if (arg->writemask2) writemask_str[i++] = 'z';
if (arg->writemask3) writemask_str[i++] = 'w';
} // if
writemask_str[i] = '\0';
assert(i < sizeof (writemask_str));
621
622
623
// may turn out something like "_x2_sat_pp_centroid r0.xyzw" ...
char *retval = get_scratch_buffer(ctx);
624
snprintf(retval, SCRATCH_BUFFER_SIZE, "%s%s%s%s %s%s%s",
625
result_shift_str, sat_str, pp_str, cent_str,
626
regtype_str, regnum_str, writemask_str);
627
return retval;
628
} // make_D3D_destarg_string
629
630
631
static char *make_D3D_sourcearg_string(Context *ctx, const int idx)
632
{
633
634
635
if (idx >= STATICARRAYLEN(ctx->source_args))
{
fail(ctx, "Too many source args");
636
return "";
637
} // if
638
639
640
641
642
const SourceArgInfo *arg = &ctx->source_args[idx];
const char *premod_str = "";
const char *postmod_str = "";
643
switch ((SourceMod) arg->src_mod)
644
645
646
647
648
649
650
651
652
653
654
{
case SRCMOD_NEGATE:
premod_str = "-";
break;
case SRCMOD_BIASNEGATE:
premod_str = "-";
// fall through.
case SRCMOD_BIAS:
postmod_str = "_bias";
break;
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
case SRCMOD_SIGNNEGATE:
premod_str = "-";
// fall through.
case SRCMOD_SIGN:
postmod_str = "_bx2";
break;
case SRCMOD_COMPLEMENT:
premod_str = "1-";
break;
case SRCMOD_X2NEGATE:
premod_str = "-";
// fall through.
case SRCMOD_X2:
postmod_str = "_x2";
break;
case SRCMOD_DZ:
postmod_str = "_dz";
break;
case SRCMOD_DW:
postmod_str = "_dw";
break;
case SRCMOD_ABSNEGATE:
premod_str = "-";
// fall through.
case SRCMOD_ABS:
postmod_str = "_abs";
break;
case SRCMOD_NOT:
premod_str = "!";
break;
} // switch
char regnum_str[16];
const char *regtype_str = get_D3D_register_string(ctx,
697
(RegisterType) arg->regtype,
698
699
700
701
702
703
704
705
706
707
708
arg->regnum, regnum_str,
sizeof (regnum_str));
if (regtype_str == NULL)
{
fail(ctx, "Unknown source register type.");
return "";
} // if
char swizzle_str[6];
int i = 0;
709
if (arg->swizzle != 0xE4) // 0xE4 == 11100100 ... 3 2 1 0. No swizzle.
710
711
712
{
static const char channel[] = { 'x', 'y', 'z', 'w' };
swizzle_str[i++] = '.';
713
714
715
716
717
718
719
720
swizzle_str[i++] = channel[arg->swizzle_x];
swizzle_str[i++] = channel[arg->swizzle_y];
swizzle_str[i++] = channel[arg->swizzle_z];
swizzle_str[i++] = channel[arg->swizzle_w];
// .xyzz is the same as .xyz, .z is the same as .zzzz, etc.
while (swizzle_str[i-1] == swizzle_str[i-2])
i--;
721
722
723
724
725
} // if
swizzle_str[i] = '\0';
assert(i < sizeof (swizzle_str));
char *retval = get_scratch_buffer(ctx);
726
727
snprintf(retval, SCRATCH_BUFFER_SIZE, "%s%s%s%s%s",
premod_str, regtype_str, regnum_str, postmod_str, swizzle_str);
728
return retval;
729
} // make_D3D_sourcearg_string
730
731
732
static void emit_D3D_start(Context *ctx)
733
734
735
{
const uint major = (uint) ctx->major_ver;
const uint minor = (uint) ctx->minor_ver;
736
737
738
739
740
741
742
743
744
const char *shadertype_str = NULL;
char minor_str[16];
if (minor == 0xFF)
strcpy(minor_str, "sw");
else if (minor == 0x1) // apparently this is "vs_2_x". Weird.
strcpy(minor_str, "x");
else
snprintf(minor_str, sizeof (minor_str), "%u", (uint) minor);
745
746
if (ctx->shader_type == SHADER_TYPE_PIXEL)
747
shadertype_str = "ps";
748
else if (ctx->shader_type == SHADER_TYPE_VERTEX)
749
shadertype_str = "vs";
750
else
751
752
753
{
failf(ctx, "Shader type %u unsupported in this profile.",
(uint) ctx->shader_type);
754
return;
755
} // else
756
757
output_line(ctx, "%s_%u_%s", shadertype_str, major, minor_str);
758
759
} // emit_D3D_start
760
761
static void emit_D3D_end(Context *ctx)
762
{
763
output_line(ctx, "end");
764
} // emit_D3D_end
765
766
767
static void emit_D3D_comment(Context *ctx, const char *str)
768
{
769
770
output_line(ctx, "; %s", str);
} // emit_D3D_comment
771
772
773
static void emit_D3D_RESERVED(Context *ctx)
774
{
775
776
777
// do nothing; fails in the state machine.
} // emit_D3D_RESERVED
778
779
780
781
// Generic D3D opcode emitters. A list of macros generate all the entry points
// that call into these...
782
783
784
785
786
787
788
789
790
791
792
793
static char *lowercase(char *dst, const char *src)
{
int i = 0;
do
{
const char ch = src[i];
dst[i] = (((ch >= 'A') && (ch <= 'Z')) ? (ch - ('A' - 'a')) : ch);
} while (src[i++]);
return dst;
} // lowercase
794
static void emit_D3D_opcode_d(Context *ctx, const char *opcode)
795
{
796
const char *dst0 = make_D3D_destarg_string(ctx, 0);
797
opcode = lowercase(get_scratch_buffer(ctx), opcode);
798
output_line(ctx, "%s%s", opcode, dst0);
799
800
801
} // emit_D3D_opcode_d
802
static void emit_D3D_opcode_s(Context *ctx, const char *opcode)
803
{
804
const char *src0 = make_D3D_destarg_string(ctx, 0);
805
opcode = lowercase(get_scratch_buffer(ctx), opcode);
806
output_line(ctx, "%s %s", opcode, src0);
807
808
809
} // emit_D3D_opcode_s
810
static void emit_D3D_opcode_ss(Context *ctx, const char *opcode)
811
{
812
813
const char *src0 = make_D3D_sourcearg_string(ctx, 0);
const char *src1 = make_D3D_sourcearg_string(ctx, 1);
814
opcode = lowercase(get_scratch_buffer(ctx), opcode);
815
output_line(ctx, "%s %s, %s", opcode, src0, src1);
816
} // emit_D3D_opcode_ss
817
818
819
static void emit_D3D_opcode_ds(Context *ctx, const char *opcode)
820
{
821
822
const char *dst0 = make_D3D_destarg_string(ctx, 0);
const char *src0 = make_D3D_sourcearg_string(ctx, 0);
823
opcode = lowercase(get_scratch_buffer(ctx), opcode);
824
output_line(ctx, "%s%s, %s", opcode, dst0, src0);
825
826
827
} // emit_D3D_opcode_ds
828
static void emit_D3D_opcode_dss(Context *ctx, const char *opcode)
829
{
830
831
832
const char *dst0 = make_D3D_destarg_string(ctx, 0);
const char *src0 = make_D3D_sourcearg_string(ctx, 0);
const char *src1 = make_D3D_sourcearg_string(ctx, 1);
833
opcode = lowercase(get_scratch_buffer(ctx), opcode);
834
output_line(ctx, "%s%s, %s, %s", opcode, dst0, src0, src1);
835
836
837
} // emit_D3D_opcode_dss
838
static void emit_D3D_opcode_dsss(Context *ctx, const char *opcode)
839
{
840
841
842
843
const char *dst0 = make_D3D_destarg_string(ctx, 0);
const char *src0 = make_D3D_sourcearg_string(ctx, 0);
const char *src1 = make_D3D_sourcearg_string(ctx, 1);
const char *src2 = make_D3D_sourcearg_string(ctx, 2);
844
opcode = lowercase(get_scratch_buffer(ctx), opcode);
845
output_line(ctx, "%s%s, %s, %s, %s", opcode, dst0, src0, src1, src2);
846
847
848
} // emit_D3D_opcode_dsss
849
static void emit_D3D_opcode_dssss(Context *ctx, const char *opcode)
850
{
851
852
853
854
855
const char *dst0 = make_D3D_destarg_string(ctx, 0);
const char *src0 = make_D3D_sourcearg_string(ctx, 0);
const char *src1 = make_D3D_sourcearg_string(ctx, 1);
const char *src2 = make_D3D_sourcearg_string(ctx, 2);
const char *src3 = make_D3D_sourcearg_string(ctx, 3);
856
opcode = lowercase(get_scratch_buffer(ctx), opcode);
857
output_line(ctx,"%s%s, %s, %s, %s, %s",opcode,dst0,src0,src1,src2,src3);
858
859
860
} // emit_D3D_opcode_dssss
861
862
863
864
865
866
867
static void emit_D3D_opcode(Context *ctx, const char *opcode)
{
opcode = lowercase(get_scratch_buffer(ctx), opcode);
output_line(ctx, "%s", opcode);
} // emit_D3D_opcode_dssss
868
#define EMIT_D3D_OPCODE_FUNC(op) \
869
static void emit_D3D_##op(Context *ctx) { \
870
emit_D3D_opcode(ctx, #op); \
871
872
}
#define EMIT_D3D_OPCODE_D_FUNC(op) \
873
static void emit_D3D_##op(Context *ctx) { \
874
875
876
emit_D3D_opcode_d(ctx, #op); \
}
#define EMIT_D3D_OPCODE_S_FUNC(op) \
877
static void emit_D3D_##op(Context *ctx) { \
878
879
880
emit_D3D_opcode_s(ctx, #op); \
}
#define EMIT_D3D_OPCODE_SS_FUNC(op) \
881
static void emit_D3D_##op(Context *ctx) { \
882
883
884
emit_D3D_opcode_ss(ctx, #op); \
}
#define EMIT_D3D_OPCODE_DS_FUNC(op) \
885
static void emit_D3D_##op(Context *ctx) { \
886
887
888
emit_D3D_opcode_ds(ctx, #op); \
}
#define EMIT_D3D_OPCODE_DSS_FUNC(op) \
889
static void emit_D3D_##op(Context *ctx) { \
890
891
892
emit_D3D_opcode_dss(ctx, #op); \
}
#define EMIT_D3D_OPCODE_DSSS_FUNC(op) \
893
static void emit_D3D_##op(Context *ctx) { \
894
895
896
emit_D3D_opcode_dsss(ctx, #op); \
}
#define EMIT_D3D_OPCODE_DSSSS_FUNC(op) \
897
static void emit_D3D_##op(Context *ctx) { \
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
emit_D3D_opcode_dssss(ctx, #op); \
}
EMIT_D3D_OPCODE_FUNC(NOP)
EMIT_D3D_OPCODE_DS_FUNC(MOV)
EMIT_D3D_OPCODE_DSS_FUNC(ADD)
EMIT_D3D_OPCODE_DSS_FUNC(SUB)
EMIT_D3D_OPCODE_DSSS_FUNC(MAD)
EMIT_D3D_OPCODE_DSS_FUNC(MUL)
EMIT_D3D_OPCODE_DS_FUNC(RCP)
EMIT_D3D_OPCODE_DS_FUNC(RSQ)
EMIT_D3D_OPCODE_DSS_FUNC(DP3)
EMIT_D3D_OPCODE_DSS_FUNC(DP4)
EMIT_D3D_OPCODE_DSS_FUNC(MIN)
EMIT_D3D_OPCODE_DSS_FUNC(MAX)
EMIT_D3D_OPCODE_DSS_FUNC(SLT)
EMIT_D3D_OPCODE_DSS_FUNC(SGE)
EMIT_D3D_OPCODE_DS_FUNC(EXP)
EMIT_D3D_OPCODE_DS_FUNC(LOG)
EMIT_D3D_OPCODE_DS_FUNC(LIT)
EMIT_D3D_OPCODE_DSS_FUNC(DST)
EMIT_D3D_OPCODE_DSSS_FUNC(LRP)
EMIT_D3D_OPCODE_DS_FUNC(FRC)
EMIT_D3D_OPCODE_DSS_FUNC(M4X4)
EMIT_D3D_OPCODE_DSS_FUNC(M4X3)
EMIT_D3D_OPCODE_DSS_FUNC(M3X4)
EMIT_D3D_OPCODE_DSS_FUNC(M3X3)
EMIT_D3D_OPCODE_DSS_FUNC(M3X2)
EMIT_D3D_OPCODE_S_FUNC(CALL)
EMIT_D3D_OPCODE_SS_FUNC(CALLNZ)
928
EMIT_D3D_OPCODE_SS_FUNC(LOOP)
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
EMIT_D3D_OPCODE_FUNC(RET)
EMIT_D3D_OPCODE_FUNC(ENDLOOP)
EMIT_D3D_OPCODE_S_FUNC(LABEL)
EMIT_D3D_OPCODE_DSS_FUNC(POW)
EMIT_D3D_OPCODE_DSS_FUNC(CRS)
EMIT_D3D_OPCODE_DSSS_FUNC(SGN)
EMIT_D3D_OPCODE_DS_FUNC(ABS)
EMIT_D3D_OPCODE_DS_FUNC(NRM)
EMIT_D3D_OPCODE_DS_FUNC(SINCOS)
EMIT_D3D_OPCODE_S_FUNC(REP)
EMIT_D3D_OPCODE_FUNC(ENDREP)
EMIT_D3D_OPCODE_S_FUNC(IF)
EMIT_D3D_OPCODE_FUNC(ELSE)
EMIT_D3D_OPCODE_FUNC(ENDIF)
EMIT_D3D_OPCODE_FUNC(BREAK)
EMIT_D3D_OPCODE_DS_FUNC(MOVA)
EMIT_D3D_OPCODE_D_FUNC(TEXKILL)
EMIT_D3D_OPCODE_DS_FUNC(TEXBEM)
EMIT_D3D_OPCODE_DS_FUNC(TEXBEML)
EMIT_D3D_OPCODE_DS_FUNC(TEXREG2AR)
EMIT_D3D_OPCODE_DS_FUNC(TEXREG2GB)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X2PAD)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X2TEX)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X3PAD)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X3TEX)
EMIT_D3D_OPCODE_DSS_FUNC(TEXM3X3SPEC)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X3VSPEC)
EMIT_D3D_OPCODE_DS_FUNC(EXPP)
EMIT_D3D_OPCODE_DS_FUNC(LOGP)
EMIT_D3D_OPCODE_DSSS_FUNC(CND)
EMIT_D3D_OPCODE_DS_FUNC(TEXREG2RGB)
EMIT_D3D_OPCODE_DS_FUNC(TEXDP3TEX)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X2DEPTH)
EMIT_D3D_OPCODE_DS_FUNC(TEXDP3)
EMIT_D3D_OPCODE_DS_FUNC(TEXM3X3)
EMIT_D3D_OPCODE_D_FUNC(TEXDEPTH)
EMIT_D3D_OPCODE_DSSS_FUNC(CMP)
EMIT_D3D_OPCODE_DSS_FUNC(BEM)
EMIT_D3D_OPCODE_DSSS_FUNC(DP2ADD)
EMIT_D3D_OPCODE_DS_FUNC(DSX)
EMIT_D3D_OPCODE_DS_FUNC(DSY)
EMIT_D3D_OPCODE_DSSSS_FUNC(TEXLDD)
EMIT_D3D_OPCODE_DSS_FUNC(TEXLDL)
972
EMIT_D3D_OPCODE_S_FUNC(BREAKP)
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
// special cases for comparison opcodes...
static const char *get_D3D_comparison_string(Context *ctx)
{
static const char *comps[] = {
"", "_gt", "_eq", "_ge", "_lt", "_ne", "_le"
};
if (ctx->instruction_controls >= STATICARRAYLEN(comps))
{
fail(ctx, "unknown comparison control");
return "";
} // if
return comps[ctx->instruction_controls];
} // get_D3D_comparison_string
static void emit_D3D_BREAKC(Context *ctx)
{
char op[16];
993
snprintf(op, sizeof (op), "break%s", get_D3D_comparison_string(ctx));
994
995
996
997
998
999
emit_D3D_opcode_ss(ctx, op);
} // emit_D3D_BREAKC
static void emit_D3D_IFC(Context *ctx)
{
char op[16];
1000
snprintf(op, sizeof (op), "if%s", get_D3D_comparison_string(ctx));