/
mojoshader.c
7645 lines (6455 loc) · 246 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
// !!! FIXME: this file really needs to be split up.
11
12
// !!! FIXME: I keep changing coding styles for symbols and typedefs.
13
14
#define __MOJOSHADER_INTERNAL__ 1
#include "mojoshader_internal.h"
15
16
17
// predeclare.
18
typedef struct Context Context;
19
struct ConstantsList;
20
21
// one emit function for each opcode in each profile.
22
typedef void (*emit_function)(Context *ctx);
23
24
// one emit function for starting output in each profile.
25
typedef void (*emit_start)(Context *ctx, const char *profilestr);
26
27
// one emit function for ending output in each profile.
28
typedef void (*emit_end)(Context *ctx);
29
30
31
32
// one emit function for phase opcode output in each profile.
typedef void (*emit_phase)(Context *ctx);
33
34
35
36
37
38
// one emit function for finalizing output in each profile.
typedef void (*emit_finalize)(Context *ctx);
// one emit function for global definitions in each profile.
typedef void (*emit_global)(Context *ctx, RegisterType regtype, int regnum);
39
// one emit function for relative uniform arrays in each profile.
40
typedef void (*emit_array)(Context *ctx, int base, int size);
41
42
43
44
45
46
// one emit function for relative constants arrays in each profile.
typedef void (*emit_const_array)(Context *ctx,
const struct ConstantsList *constslist,
int base, int size);
47
// one emit function for uniforms in each profile.
48
49
typedef void (*emit_uniform)(Context *ctx, RegisterType regtype, int regnum,
int arraybase, int arraysize);
50
51
52
53
// one emit function for samplers in each profile.
typedef void (*emit_sampler)(Context *ctx, int stage, TextureType ttype);
54
55
// one emit function for attributes in each profile.
typedef void (*emit_attribute)(Context *ctx, RegisterType regtype, int regnum,
56
57
MOJOSHADER_usage usage, int index, int wmask,
int flags);
58
59
60
61
// one args function for each possible sequence of opcode arguments.
typedef int (*args_function)(Context *ctx);
62
// one state function for each opcode where we have state machine updates.
63
typedef void (*state_function)(Context *ctx);
64
65
66
67
68
// one function for varnames in each profile.
typedef const char *(*varname_function)(Context *c, RegisterType t, int num);
// one function for const var array in each profile.
69
typedef const char *(*const_array_varname_function)(Context *c, int base, int size);
70
71
72
73
74
75
typedef struct
{
const char *name;
emit_start start_emitter;
emit_end end_emitter;
76
emit_phase phase_emitter;
77
emit_global global_emitter;
78
emit_array array_emitter;
79
emit_const_array const_array_emitter;
80
emit_uniform uniform_emitter;
81
emit_sampler sampler_emitter;
82
emit_attribute attribute_emitter;
83
emit_finalize finalize_emitter;
84
85
varname_function get_varname;
const_array_varname_function get_const_array_varname;
86
} Profile;
87
88
89
90
91
// 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.
92
typedef struct OutputListNode
93
94
{
char *str;
95
96
97
98
99
100
101
struct OutputListNode *next;
} OutputListNode;
typedef struct OutputList
{
OutputListNode head;
OutputListNode *tail;
102
103
} OutputList;
104
105
106
107
108
109
typedef struct ConstantsList
{
MOJOSHADER_constant constant;
struct ConstantsList *next;
} ConstantsList;
110
111
112
113
114
typedef struct VariableList
{
MOJOSHADER_uniformType type;
int index;
int count;
115
ConstantsList *constant;
116
int used;
117
118
struct VariableList *next;
} VariableList;
119
120
121
122
123
typedef struct RegisterList
{
RegisterType regtype;
int regnum;
124
125
126
MOJOSHADER_usage usage;
int index;
int writemask;
127
int misc;
128
const VariableList *array;
129
130
131
struct RegisterList *next;
} RegisterList;
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
typedef struct
{
const uint32 *token; // this is the unmolested token in the stream.
int regnum;
int swizzle; // xyzw (all four, not split out).
int swizzle_x;
int swizzle_y;
int swizzle_z;
int swizzle_w;
SourceMod src_mod;
RegisterType regtype;
int relative;
RegisterType relative_regtype;
int relative_regnum;
int relative_component;
const VariableList *relative_array;
} SourceArgInfo;
149
150
#define SCRATCH_BUFFER_SIZE 128
151
#define SCRATCH_BUFFERS 32
152
153
// !!! FIXME: the scratch buffers make Context pretty big.
154
155
// !!! FIXME: might be worth having one set of static scratch buffers that
// !!! FIXME: are mutex protected?
156
157
// Context...this is state that changes as we parse through a shader...
158
struct Context
159
{
160
161
int isfail;
int out_of_memory;
162
163
MOJOSHADER_malloc malloc;
MOJOSHADER_free free;
164
void *malloc_data;
165
const uint32 *orig_tokens;
166
167
const uint32 *tokens;
uint32 tokencount;
168
MOJOSHADER_parsePhase parse_phase;
169
170
const MOJOSHADER_swizzle *swizzles;
unsigned int swizzles_count;
171
172
OutputList *output;
OutputList globals;
173
OutputList helpers;
174
OutputList subroutines;
175
OutputList mainline_intro;
176
OutputList mainline;
177
178
OutputList ignore;
OutputList *output_stack[2];
179
uint8 *output_bytes; // can be used instead of the OutputLists.
180
181
int indent_stack[2];
int output_stack_len;
182
int output_len; // total strlen; prevents walking the lists just to malloc.
183
int indent;
184
const char *shader_type_str;
185
186
const char *endline;
int endline_len;
187
char scratch[SCRATCH_BUFFERS][SCRATCH_BUFFER_SIZE];
188
189
int scratchidx; // current scratch buffer.
int profileid;
190
const Profile *profile;
191
MOJOSHADER_shaderType shader_type;
192
193
uint8 major_ver;
uint8 minor_ver;
194
DestArgInfo dest_arg;
195
SourceArgInfo source_args[5];
196
SourceArgInfo predicate_arg; // for predicated instructions.
197
uint32 dwords[4];
198
uint32 version_token;
199
int instruction_count;
200
uint32 instruction_controls;
201
202
uint32 previous_opcode;
int loops;
203
int reps;
204
int max_reps;
205
int cmps;
206
207
int scratch_registers;
int max_scratch_registers;
208
209
210
int branch_labels_stack_index;
int branch_labels_stack[32];
int assigned_branch_labels;
211
int assigned_vertex_attributes;
212
int last_address_reg_component;
213
214
RegisterList used_registers;
RegisterList defined_registers;
215
216
int error_count;
ErrorList *errors;
217
218
int constant_count;
ConstantsList *constants;
219
220
int uniform_count;
RegisterList uniforms;
221
222
int attribute_count;
RegisterList attributes;
223
224
int sampler_count;
RegisterList samplers;
225
VariableList *variables; // variables to register mapping.
226
227
int centroid_allowed;
int have_ctab;
228
int have_relative_input_registers;
229
230
231
232
233
234
235
int determined_constants_arrays;
int predicated;
int support_nv2;
int support_nv3;
int support_nv4;
int support_glsl120;
int glsl_generated_lit_opcode;
236
237
238
};
239
// Convenience functions for allocators...
240
#if !MOJOSHADER_FORCE_ALLOCATOR
241
242
void *MOJOSHADER_internal_malloc(int bytes, void *d) { return malloc(bytes); }
void MOJOSHADER_internal_free(void *ptr, void *d) { free(ptr); }
243
#endif
244
245
246
247
248
MOJOSHADER_error MOJOSHADER_out_of_mem_error = { "Out of memory", NULL, -1 };
MOJOSHADER_parseData MOJOSHADER_out_of_mem_data = {
1, &MOJOSHADER_out_of_mem_error, 0, 0, 0, 0,
MOJOSHADER_TYPE_UNKNOWN, 0, 0, 0, 0
249
250
};
251
static inline void out_of_memory(Context *ctx)
252
{
253
ctx->isfail = ctx->out_of_memory = 1;
254
255
} // out_of_memory
256
static inline void *Malloc(Context *ctx, const size_t len)
257
{
258
void *retval = ctx->malloc((int) len, ctx->malloc_data);
259
260
261
if (retval == NULL)
out_of_memory(ctx);
return retval;
262
263
} // Malloc
264
265
266
static inline char *StrDup(Context *ctx, const char *str)
{
char *retval = (char *) Malloc(ctx, strlen(str) + 1);
267
if (retval != NULL)
268
269
270
strcpy(retval, str);
return retval;
} // StrDup
271
272
static inline void Free(Context *ctx, void *ptr)
273
{
274
if (ptr != NULL) // check for NULL in case of dumb free() impl.
275
ctx->free(ptr, ctx->malloc_data);
276
277
278
} // Free
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
// jump between output sections in the context...
static inline void push_output(Context *ctx, OutputList *section)
{
assert(ctx->output_stack_len < STATICARRAYLEN(ctx->output_stack));
ctx->output_stack[ctx->output_stack_len] = ctx->output;
ctx->indent_stack[ctx->output_stack_len] = ctx->indent;
ctx->output_stack_len++;
ctx->output = section;
ctx->indent = 0;
} // push_output
static inline void pop_output(Context *ctx)
{
assert(ctx->output_stack_len > 0);
ctx->output_stack_len--;
ctx->output = ctx->output_stack[ctx->output_stack_len];
ctx->indent = ctx->indent_stack[ctx->output_stack_len];
} // pop_output
// Shader model version magic...
303
304
static inline uint32 ver_ui32(const uint8 major, const uint8 minor)
{
305
return ( (((uint32) major) << 16) | (((minor) == 0xFF) ? 1 : (minor)) );
306
307
} // version_ui32
308
static inline int shader_version_supported(const uint8 maj, const uint8 min)
309
310
311
312
{
return (ver_ui32(maj,min) <= ver_ui32(MAX_SHADER_MAJOR, MAX_SHADER_MINOR));
} // shader_version_supported
313
314
static inline int shader_version_atleast(const Context *ctx, const uint8 maj,
const uint8 min)
315
316
317
318
{
return (ver_ui32(ctx->major_ver, ctx->minor_ver) >= ver_ui32(maj, min));
} // shader_version_atleast
319
320
321
322
323
324
static inline int shader_version_exactly(const Context *ctx, const uint8 maj,
const uint8 min)
{
return ((ctx->major_ver == maj) && (ctx->minor_ver == min));
} // shader_version_exactly
325
326
327
328
329
330
331
332
333
334
static inline int shader_is_pixel(const Context *ctx)
{
return (ctx->shader_type == MOJOSHADER_TYPE_PIXEL);
} // shader_is_pixel
static inline int shader_is_vertex(const Context *ctx)
{
return (ctx->shader_type == MOJOSHADER_TYPE_VERTEX);
} // shader_is_vertex
335
336
337
static inline int isfail(const Context *ctx)
{
338
return ctx->isfail;
339
340
} // isfail
341
342
static inline char *get_scratch_buffer(Context *ctx)
343
{
344
assert(ctx->scratchidx < SCRATCH_BUFFERS);
345
ctx->scratchidx = (ctx->scratchidx + 1) % SCRATCH_BUFFERS;
346
return ctx->scratch[ctx->scratchidx];
347
348
} // get_scratch_buffer
349
350
351
static void failf(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static void failf(Context *ctx, const char *fmt, ...)
352
{
353
354
ctx->isfail = 1;
if (ctx->out_of_memory)
355
return;
356
357
358
int error_position = 0;
switch (ctx->parse_phase)
359
{
360
361
362
363
364
365
366
367
368
369
370
case MOJOSHADER_PARSEPHASE_NOTSTARTED:
error_position = -2;
break;
case MOJOSHADER_PARSEPHASE_WORKING:
error_position = (ctx->tokens - ctx->orig_tokens) * sizeof (uint32);
break;
case MOJOSHADER_PARSEPHASE_DONE:
error_position = -1;
break;
default:
assert(0 && "Unexpected value");
371
return;
372
373
374
375
} // switch
ErrorList *error = (ErrorList *) Malloc(ctx, sizeof (ErrorList));
if (error == NULL)
376
return;
377
378
379
380
381
382
char *scratch = get_scratch_buffer(ctx);
va_list ap;
va_start(ap, fmt);
const int len = vsnprintf(scratch, SCRATCH_BUFFER_SIZE, fmt, ap);
va_end(ap);
383
384
385
386
387
388
389
390
391
392
char *failstr = (char *) Malloc(ctx, len + 1);
if (failstr == NULL)
Free(ctx, error);
else
{
// see comments about scratch buffer overflow in output_line().
if (len < SCRATCH_BUFFER_SIZE)
strcpy(failstr, scratch); // copy it over.
else
393
{
394
395
396
397
398
399
400
401
va_start(ap, fmt);
vsnprintf(failstr, len + 1, fmt, ap); // rebuild it.
va_end(ap);
} // else
error->error.error = failstr;
error->error.filename = NULL; // no filename at this level.
error->error.error_position = error_position;
402
error->next = NULL;
403
404
ErrorList *prev = NULL;
405
406
ErrorList *item = ctx->errors;
while (item != NULL)
407
{
408
prev = item;
409
item = item->next;
410
411
} // while
412
if (prev == NULL)
413
ctx->errors = error;
414
415
else
prev->next = error;
416
417
418
ctx->error_count++;
} // else
419
420
421
} // failf
422
static inline void fail(Context *ctx, const char *reason)
423
{
424
failf(ctx, "%s", reason);
425
426
427
} // fail
428
429
static void output_line(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static void output_line(Context *ctx, const char *fmt, ...)
430
{
431
432
OutputListNode *item = NULL;
433
434
if (isfail(ctx) || ctx->out_of_memory)
return; // we failed previously, don't go on...
435
436
char *scratch = get_scratch_buffer(ctx);
437
438
const int indent = ctx->indent;
439
440
if (indent > 0)
memset(scratch, '\t', indent);
441
442
443
va_list ap;
va_start(ap, fmt);
444
const int len = vsnprintf(scratch+indent, SCRATCH_BUFFER_SIZE-indent, fmt, ap) + indent;
445
446
va_end(ap);
447
item = (OutputListNode *) Malloc(ctx, sizeof (OutputListNode));
448
if (item == NULL)
449
return;
450
451
item->str = (char *) Malloc(ctx, len + 1);
452
if (item->str == NULL)
453
{
454
Free(ctx, item);
455
return;
456
457
} // if
458
459
460
// 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().
461
if (len < SCRATCH_BUFFER_SIZE)
462
463
464
strcpy(item->str, scratch); // copy it over.
else
{
465
466
if (indent > 0)
memset(item->str, '\t', indent);
467
va_start(ap, fmt);
468
vsnprintf(item->str+indent, len + 1, fmt, ap); // rebuild it.
469
470
va_end(ap);
} // else
471
472
item->next = NULL;
473
474
475
ctx->output->tail->next = item;
ctx->output->tail = item;
476
ctx->output_len += len + ctx->endline_len;
477
478
479
} // output_line
480
// this is just to stop gcc whining.
481
static inline void output_blank_line(Context *ctx)
482
{
483
output_line(ctx, "%s", "");
484
485
486
} // output_blank_line
487
// !!! FIXME: this is sort of nasty.
488
489
static void floatstr(Context *ctx, char *buf, size_t bufsize, float f,
int leavedecimal)
490
491
{
const size_t len = snprintf(buf, bufsize, "%f", f);
492
if ((len+2) >= bufsize)
493
494
495
496
497
498
fail(ctx, "BUG: internal buffer is too small");
else
{
char *end = buf + len;
char *ptr = strchr(buf, '.');
if (ptr == NULL)
499
500
501
{
if (leavedecimal)
strcat(buf, ".0");
502
return; // done.
503
} // if
504
505
506
507
508
509
510
511
512
while (--end != ptr)
{
if (*end != '0')
{
end++;
break;
} // if
} // while
513
514
if ((leavedecimal) && (end == ptr))
end += 2;
515
516
517
518
*end = '\0'; // chop extra '0' or all decimal places off.
} // else
} // floatstr
519
520
521
// Deal with register lists... !!! FIXME: I sort of hate this.
522
static void free_reglist(MOJOSHADER_free f, void *d, RegisterList *item)
523
524
525
526
{
while (item != NULL)
{
RegisterList *next = item->next;
527
f(item, d);
528
529
530
531
532
533
534
535
536
item = next;
} // while
} // free_reglist
static inline uint32 reg_to_ui32(const RegisterType regtype, const int regnum)
{
return ( ((uint32) regtype) | (((uint32) regnum) << 16) );
} // reg_to_uint32
537
538
539
static RegisterList *reglist_insert(Context *ctx, RegisterList *prev,
const RegisterType regtype,
const int regnum)
540
541
542
543
544
545
546
{
const uint32 newval = reg_to_ui32(regtype, regnum);
RegisterList *item = prev->next;
while (item != NULL)
{
const uint32 val = reg_to_ui32(item->regtype, item->regnum);
if (newval == val)
547
return item; // already set, so we're done.
548
549
550
551
552
553
554
555
556
557
558
else if (newval < val) // insert it here.
break;
else // if (newval > val)
{
// keep going, we're not to the insertion point yet.
prev = item;
item = item->next;
} // else
} // while
// we need to insert an entry after (prev).
559
item = (RegisterList *) Malloc(ctx, sizeof (RegisterList));
560
if (item != NULL)
561
562
563
{
item->regtype = regtype;
item->regnum = regnum;
564
item->usage = MOJOSHADER_USAGE_UNKNOWN;
565
566
item->index = 0;
item->writemask = 0;
567
item->misc = 0;
568
item->array = NULL;
569
570
item->next = prev->next;
prev->next = item;
571
} // if
572
573
return item;
574
575
} // reglist_insert
576
577
static RegisterList *reglist_find(const RegisterList *prev,
const RegisterType rtype, const int regnum)
578
{
579
580
const uint32 newval = reg_to_ui32(rtype, regnum);
RegisterList *item = prev->next;
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
while (item != NULL)
{
const uint32 val = reg_to_ui32(item->regtype, item->regnum);
if (newval == val)
return item; // here it is.
else if (newval < val) // should have been here if it existed.
return NULL;
else // if (newval > val)
{
// keep going, we're not to the insertion point yet.
prev = item;
item = item->next;
} // else
} // while
return NULL; // wasn't in the list.
597
598
599
600
601
602
603
} // reglist_find
static inline const RegisterList *reglist_exists(RegisterList *prev,
const RegisterType regtype,
const int regnum)
{
return (reglist_find(prev, regtype, regnum));
604
605
606
607
608
609
610
611
612
613
614
615
} // reglist_exists
static inline void set_used_register(Context *ctx, const RegisterType regtype,
const int regnum)
{
reglist_insert(ctx, &ctx->used_registers, regtype, regnum);
} // set_used_register
static inline int get_used_register(Context *ctx, const RegisterType regtype,
const int regnum)
{
return (reglist_exists(&ctx->used_registers, regtype, regnum) != NULL);
616
} // get_used_register
617
618
619
620
621
622
623
624
625
626
627
628
629
static inline void set_defined_register(Context *ctx, const RegisterType rtype,
const int regnum)
{
reglist_insert(ctx, &ctx->defined_registers, rtype, regnum);
} // set_defined_register
static inline int get_defined_register(Context *ctx, const RegisterType rtype,
const int regnum)
{
return (reglist_exists(&ctx->defined_registers, rtype, regnum) != NULL);
} // get_defined_register
630
631
static void add_attribute_register(Context *ctx, const RegisterType rtype,
const int regnum, const MOJOSHADER_usage usage,
632
const int index, const int writemask, int flags)
633
634
635
636
637
{
RegisterList *item = reglist_insert(ctx, &ctx->attributes, rtype, regnum);
item->usage = usage;
item->index = index;
item->writemask = writemask;
638
item->misc = flags;
639
640
} // add_attribute_register
641
642
643
static inline void add_sampler(Context *ctx, const RegisterType rtype,
const int regnum, const TextureType ttype)
{
644
// !!! FIXME: make sure it doesn't exist?
645
646
647
648
RegisterList *item = reglist_insert(ctx, &ctx->samplers, rtype, regnum);
item->index = (int) ttype;
} // add_sampler
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
static inline int writemask_xyzw(const int writemask)
{
return (writemask == 0xF); // 0xF == 1111. No explicit mask (full!).
} // writemask_xyzw
static inline int writemask_xyz(const int writemask)
{
return (writemask == 0x7); // 0x7 == 0111. (that is: xyz)
} // writemask_xyz
static inline int writemask_xy(const int writemask)
{
return (writemask == 0x3); // 0x3 == 0011. (that is: xy)
} // writemask_xy
static inline int writemask_x(const int writemask)
{
return (writemask == 0x1); // 0x1 == 0001. (that is: x)
} // writemask_x
static inline int writemask_y(const int writemask)
{
return (writemask == 0x2); // 0x1 == 0010. (that is: y)
} // writemask_y
680
681
682
683
684
685
686
687
static inline int replicate_swizzle(const int swizzle)
{
return ( (((swizzle >> 0) & 0x3) == ((swizzle >> 2) & 0x3)) &&
(((swizzle >> 2) & 0x3) == ((swizzle >> 4) & 0x3)) &&
(((swizzle >> 4) & 0x3) == ((swizzle >> 6) & 0x3)) );
} // replicate_swizzle
688
689
static inline int no_swizzle(const int swizzle)
{
690
return (swizzle == 0xE4); // 0xE4 == 11100100 ... 0 1 2 3. No swizzle.
691
692
693
} // no_swizzle
694
695
696
697
698
static inline int vecsize_from_writemask(const int m)
{
return (m & 1) + ((m >> 1) & 1) + ((m >> 2) & 1) + ((m >> 3) & 1);
} // vecsize_from_writemask
699
700
701
702
703
704
705
706
static int allocate_scratch_register(Context *ctx)
{
const int retval = ctx->scratch_registers++;
if (retval >= ctx->max_scratch_registers)
ctx->max_scratch_registers = retval + 1;
return retval;
} // allocate_scratch_register
707
static int allocate_branch_label(Context *ctx)
708
{
709
710
return ctx->assigned_branch_labels++;
} // allocate_branch_label
711
712
713
714
// D3D stuff that's used in more than just the d3d profile...
715
716
717
static const char swizzle_channels[] = { 'x', 'y', 'z', 'w' };
718
719
720
721
722
static const char *usagestrs[] = {
"_position", "_blendweight", "_blendindices", "_normal", "_psize",
"_texcoord", "_tangent", "_binormal", "_tessfactor", "_positiont",
"_color", "_fog", "_depth", "_sample"
};
723
724
static const char *get_D3D_register_string(Context *ctx,
725
RegisterType regtype,
726
727
int regnum, char *regnum_str,
size_t regnum_size)
728
{
729
const char *retval = NULL;
730
int has_number = 1;
731
732
switch (regtype)
733
{
734
case REG_TYPE_TEMP:
735
retval = "r";
736
737
break;
738
case REG_TYPE_INPUT:
739
retval = "v";
740
741
break;
742
case REG_TYPE_CONST:
743
744
745
retval = "c";
break;
746
case REG_TYPE_ADDRESS: // (or REG_TYPE_TEXTURE, same value.)
747
retval = shader_is_vertex(ctx) ? "a" : "t";
748
749
break;
750
case REG_TYPE_RASTOUT:
751
switch ((RastOutType) regnum)
752
{
753
754
755
case RASTOUT_TYPE_POSITION: retval = "oPos"; break;
case RASTOUT_TYPE_FOG: retval = "oFog"; break;
case RASTOUT_TYPE_POINT_SIZE: retval = "oPts"; break;
756
} // switch
757
has_number = 0;
758
759
break;
760
case REG_TYPE_ATTROUT:
761
retval = "oD";
762
763
break;
764
case REG_TYPE_OUTPUT: // (or REG_TYPE_TEXCRDOUT, same value.)
765
if (shader_is_vertex(ctx) && shader_version_atleast(ctx, 3, 0))
766
retval = "o";
767
else
768
retval = "oT";
769
770
break;
771
case REG_TYPE_CONSTINT:
772
retval = "i";
773
774
break;
775
case REG_TYPE_COLOROUT:
776
retval = "oC";
777
778
break;
779
case REG_TYPE_DEPTHOUT:
780
retval = "oDepth";
781
has_number = 0;
782
783
break;
784
case REG_TYPE_SAMPLER:
785
retval = "s";
786
787
break;
788
case REG_TYPE_CONSTBOOL:
789
retval = "b";
790
791
break;
792
case REG_TYPE_LOOP:
793
retval = "aL";
794
has_number = 0;
795
796
break;
797
case REG_TYPE_MISCTYPE:
798
switch ((const MiscTypeType) regnum)
799
{
800
801
case MISCTYPE_TYPE_POSITION: retval = "vPos"; break;
case MISCTYPE_TYPE_FACE: retval = "vFace"; break;
802
} // switch
803
has_number = 0;
804
805
break;
806
case REG_TYPE_LABEL:
807
retval = "l";
808
809
break;
810
case REG_TYPE_PREDICATE:
811
retval = "p";
812
break;
813
814
815
816
817
818
819
//case REG_TYPE_TEMPFLOAT16: // !!! FIXME: don't know this asm string
default:
fail(ctx, "unknown register type");
retval = "???";
has_number = 0;
break;
820
821
} // switch
822
823
824
825
826
if (has_number)
snprintf(regnum_str, regnum_size, "%u", (uint) regnum);
else
regnum_str[0] = '\0';
827
828
829
830
return retval;
} // get_D3D_register_string
831
832
833
834
835
836
837
838
839
#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,
840
static const char *make_D3D_srcarg_string_in_buf(Context *ctx,
841
842
const SourceArgInfo *arg,
char *buf, size_t buflen)
843
844
845
{
const char *premod_str = "";
const char *postmod_str = "";
846
switch (arg->src_mod)
847
848
849
850
851
852
853
854
855
856
857
{
case SRCMOD_NEGATE:
premod_str = "-";
break;
case SRCMOD_BIASNEGATE:
premod_str = "-";
// fall through.
case SRCMOD_BIAS:
postmod_str = "_bias";
break;
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
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;
895
896
897
898
case SRCMOD_NONE:
case SRCMOD_TOTAL:
break; // stop compiler whining.
899
900
901
902
} // switch
char regnum_str[16];
903
904
905
const char *regtype_str = get_D3D_register_string(ctx, arg->regtype,
arg->regnum, regnum_str,
sizeof (regnum_str));
906
907
908
909
if (regtype_str == NULL)
{
fail(ctx, "Unknown source register type.");
910
return "";
911
912
} // if
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
const char *rel_lbracket = "";
const char *rel_rbracket = "";
char rel_swizzle[4] = { '\0' };
char rel_regnum_str[16] = { '\0' };
const char *rel_regtype_str = "";
if (arg->relative)
{
rel_swizzle[0] = '.';
rel_swizzle[1] = swizzle_channels[arg->relative_component];
rel_swizzle[2] = '\0';
rel_lbracket = "[";
rel_rbracket = "]";
rel_regtype_str = get_D3D_register_string(ctx, arg->relative_regtype,
arg->relative_regnum,
rel_regnum_str,
sizeof (rel_regnum_str));
if (regtype_str == NULL)
{
fail(ctx, "Unknown relative source register type.");
return "";
} // if
} // if
937
938
char swizzle_str[6];
int i = 0;
939
const int scalar = scalar_register(ctx->shader_type, arg->regtype, arg->regnum);
940
if (!scalar && !no_swizzle(arg->swizzle))
941
942
{
swizzle_str[i++] = '.';
943
944
945
946
swizzle_str[i++] = swizzle_channels[arg->swizzle_x];
swizzle_str[i++] = swizzle_channels[arg->swizzle_y];
swizzle_str[i++] = swizzle_channels[arg->swizzle_z];
swizzle_str[i++] = swizzle_channels[arg->swizzle_w];
947
948
949
950
// .xyzz is the same as .xyz, .z is the same as .zzzz, etc.
while (swizzle_str[i-1] == swizzle_str[i-2])
i--;
951
952
953
954
} // if
swizzle_str[i] = '\0';
assert(i < sizeof (swizzle_str));
955
// !!! FIXME: c12[a0.x] actually needs to be c[a0.x + 12]
956
957
958
959
snprintf(buf, buflen, "%s%s%s%s%s%s%s%s%s%s",
premod_str, regtype_str, regnum_str, postmod_str,
rel_lbracket, rel_regtype_str, rel_regnum_str, rel_swizzle,
rel_rbracket, swizzle_str);
960
// !!! FIXME: make sure the scratch buffer was large enough.
961
return buf;
962
} // make_D3D_srcarg_string_in_buf
963
964
965
static const char *make_D3D_destarg_string(Context *ctx)
966
{
967
const DestArgInfo *arg = &ctx->dest_arg;
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
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, arg->regtype,
arg->regnum, regnum_str,
sizeof (regnum_str));
if (regtype_str == NULL)
{
fail(ctx, "Unknown destination register type.");
return "";
} // if
char writemask_str[6];
int i = 0;
996
const int scalar = scalar_register(ctx->shader_type, arg->regtype, arg->regnum);
997
if (!scalar && !writemask_xyzw(arg->writemask))
998
999
1000
{
writemask_str[i++] = '.';
if (arg->writemask0) writemask_str[i++] = 'x';