/
mojoshader.c
7643 lines (6453 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
241
void *MOJOSHADER_internal_malloc(int bytes, void *d) { return malloc(bytes); }
void MOJOSHADER_internal_free(void *ptr, void *d) { free(ptr); }
242
243
244
245
246
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
247
248
};
249
static inline void out_of_memory(Context *ctx)
250
{
251
ctx->isfail = ctx->out_of_memory = 1;
252
253
} // out_of_memory
254
static inline void *Malloc(Context *ctx, const size_t len)
255
{
256
void *retval = ctx->malloc((int) len, ctx->malloc_data);
257
258
259
if (retval == NULL)
out_of_memory(ctx);
return retval;
260
261
} // Malloc
262
263
264
static inline char *StrDup(Context *ctx, const char *str)
{
char *retval = (char *) Malloc(ctx, strlen(str) + 1);
265
if (retval != NULL)
266
267
268
strcpy(retval, str);
return retval;
} // StrDup
269
270
static inline void Free(Context *ctx, void *ptr)
271
{
272
if (ptr != NULL) // check for NULL in case of dumb free() impl.
273
ctx->free(ptr, ctx->malloc_data);
274
275
276
} // Free
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
// 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...
301
302
static inline uint32 ver_ui32(const uint8 major, const uint8 minor)
{
303
return ( (((uint32) major) << 16) | (((minor) == 0xFF) ? 1 : (minor)) );
304
305
} // version_ui32
306
static inline int shader_version_supported(const uint8 maj, const uint8 min)
307
308
309
310
{
return (ver_ui32(maj,min) <= ver_ui32(MAX_SHADER_MAJOR, MAX_SHADER_MINOR));
} // shader_version_supported
311
312
static inline int shader_version_atleast(const Context *ctx, const uint8 maj,
const uint8 min)
313
314
315
316
{
return (ver_ui32(ctx->major_ver, ctx->minor_ver) >= ver_ui32(maj, min));
} // shader_version_atleast
317
318
319
320
321
322
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
323
324
325
326
327
328
329
330
331
332
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
333
334
335
static inline int isfail(const Context *ctx)
{
336
return ctx->isfail;
337
338
} // isfail
339
340
static inline char *get_scratch_buffer(Context *ctx)
341
{
342
assert(ctx->scratchidx < SCRATCH_BUFFERS);
343
ctx->scratchidx = (ctx->scratchidx + 1) % SCRATCH_BUFFERS;
344
return ctx->scratch[ctx->scratchidx];
345
346
} // get_scratch_buffer
347
348
349
static void failf(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static void failf(Context *ctx, const char *fmt, ...)
350
{
351
352
ctx->isfail = 1;
if (ctx->out_of_memory)
353
return;
354
355
356
int error_position = 0;
switch (ctx->parse_phase)
357
{
358
359
360
361
362
363
364
365
366
367
368
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");
369
return;
370
371
372
373
} // switch
ErrorList *error = (ErrorList *) Malloc(ctx, sizeof (ErrorList));
if (error == NULL)
374
return;
375
376
377
378
379
380
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);
381
382
383
384
385
386
387
388
389
390
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
391
{
392
393
394
395
396
397
398
399
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;
400
error->next = NULL;
401
402
ErrorList *prev = NULL;
403
404
ErrorList *item = ctx->errors;
while (item != NULL)
405
{
406
prev = item;
407
item = item->next;
408
409
} // while
410
if (prev == NULL)
411
ctx->errors = error;
412
413
else
prev->next = error;
414
415
416
ctx->error_count++;
} // else
417
418
419
} // failf
420
static inline void fail(Context *ctx, const char *reason)
421
{
422
failf(ctx, "%s", reason);
423
424
425
} // fail
426
427
static void output_line(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static void output_line(Context *ctx, const char *fmt, ...)
428
{
429
430
OutputListNode *item = NULL;
431
432
if (isfail(ctx) || ctx->out_of_memory)
return; // we failed previously, don't go on...
433
434
char *scratch = get_scratch_buffer(ctx);
435
436
const int indent = ctx->indent;
437
438
if (indent > 0)
memset(scratch, '\t', indent);
439
440
441
va_list ap;
va_start(ap, fmt);
442
const int len = vsnprintf(scratch+indent, SCRATCH_BUFFER_SIZE-indent, fmt, ap) + indent;
443
444
va_end(ap);
445
item = (OutputListNode *) Malloc(ctx, sizeof (OutputListNode));
446
if (item == NULL)
447
return;
448
449
item->str = (char *) Malloc(ctx, len + 1);
450
if (item->str == NULL)
451
{
452
Free(ctx, item);
453
return;
454
455
} // if
456
457
458
// 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().
459
if (len < SCRATCH_BUFFER_SIZE)
460
461
462
strcpy(item->str, scratch); // copy it over.
else
{
463
464
if (indent > 0)
memset(item->str, '\t', indent);
465
va_start(ap, fmt);
466
vsnprintf(item->str+indent, len + 1, fmt, ap); // rebuild it.
467
468
va_end(ap);
} // else
469
470
item->next = NULL;
471
472
473
ctx->output->tail->next = item;
ctx->output->tail = item;
474
ctx->output_len += len + ctx->endline_len;
475
476
477
} // output_line
478
// this is just to stop gcc whining.
479
static inline void output_blank_line(Context *ctx)
480
{
481
output_line(ctx, "%s", "");
482
483
484
} // output_blank_line
485
// !!! FIXME: this is sort of nasty.
486
487
static void floatstr(Context *ctx, char *buf, size_t bufsize, float f,
int leavedecimal)
488
489
{
const size_t len = snprintf(buf, bufsize, "%f", f);
490
if ((len+2) >= bufsize)
491
492
493
494
495
496
fail(ctx, "BUG: internal buffer is too small");
else
{
char *end = buf + len;
char *ptr = strchr(buf, '.');
if (ptr == NULL)
497
498
499
{
if (leavedecimal)
strcat(buf, ".0");
500
return; // done.
501
} // if
502
503
504
505
506
507
508
509
510
while (--end != ptr)
{
if (*end != '0')
{
end++;
break;
} // if
} // while
511
512
if ((leavedecimal) && (end == ptr))
end += 2;
513
514
515
516
*end = '\0'; // chop extra '0' or all decimal places off.
} // else
} // floatstr
517
518
519
// Deal with register lists... !!! FIXME: I sort of hate this.
520
static void free_reglist(MOJOSHADER_free f, void *d, RegisterList *item)
521
522
523
524
{
while (item != NULL)
{
RegisterList *next = item->next;
525
f(item, d);
526
527
528
529
530
531
532
533
534
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
535
536
537
static RegisterList *reglist_insert(Context *ctx, RegisterList *prev,
const RegisterType regtype,
const int regnum)
538
539
540
541
542
543
544
{
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)
545
return item; // already set, so we're done.
546
547
548
549
550
551
552
553
554
555
556
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).
557
item = (RegisterList *) Malloc(ctx, sizeof (RegisterList));
558
if (item != NULL)
559
560
561
{
item->regtype = regtype;
item->regnum = regnum;
562
item->usage = MOJOSHADER_USAGE_UNKNOWN;
563
564
item->index = 0;
item->writemask = 0;
565
item->misc = 0;
566
item->array = NULL;
567
568
item->next = prev->next;
prev->next = item;
569
} // if
570
571
return item;
572
573
} // reglist_insert
574
575
static RegisterList *reglist_find(const RegisterList *prev,
const RegisterType rtype, const int regnum)
576
{
577
578
const uint32 newval = reg_to_ui32(rtype, regnum);
RegisterList *item = prev->next;
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
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.
595
596
597
598
599
600
601
} // reglist_find
static inline const RegisterList *reglist_exists(RegisterList *prev,
const RegisterType regtype,
const int regnum)
{
return (reglist_find(prev, regtype, regnum));
602
603
604
605
606
607
608
609
610
611
612
613
} // 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);
614
} // get_used_register
615
616
617
618
619
620
621
622
623
624
625
626
627
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
628
629
static void add_attribute_register(Context *ctx, const RegisterType rtype,
const int regnum, const MOJOSHADER_usage usage,
630
const int index, const int writemask, int flags)
631
632
633
634
635
{
RegisterList *item = reglist_insert(ctx, &ctx->attributes, rtype, regnum);
item->usage = usage;
item->index = index;
item->writemask = writemask;
636
item->misc = flags;
637
638
} // add_attribute_register
639
640
641
static inline void add_sampler(Context *ctx, const RegisterType rtype,
const int regnum, const TextureType ttype)
{
642
// !!! FIXME: make sure it doesn't exist?
643
644
645
646
RegisterList *item = reglist_insert(ctx, &ctx->samplers, rtype, regnum);
item->index = (int) ttype;
} // add_sampler
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
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
678
679
680
681
682
683
684
685
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
686
687
static inline int no_swizzle(const int swizzle)
{
688
return (swizzle == 0xE4); // 0xE4 == 11100100 ... 0 1 2 3. No swizzle.
689
690
691
} // no_swizzle
692
693
694
695
696
static inline int vecsize_from_writemask(const int m)
{
return (m & 1) + ((m >> 1) & 1) + ((m >> 2) & 1) + ((m >> 3) & 1);
} // vecsize_from_writemask
697
698
699
700
701
702
703
704
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
705
static int allocate_branch_label(Context *ctx)
706
{
707
708
return ctx->assigned_branch_labels++;
} // allocate_branch_label
709
710
711
712
// D3D stuff that's used in more than just the d3d profile...
713
714
715
static const char swizzle_channels[] = { 'x', 'y', 'z', 'w' };
716
717
718
719
720
static const char *usagestrs[] = {
"_position", "_blendweight", "_blendindices", "_normal", "_psize",
"_texcoord", "_tangent", "_binormal", "_tessfactor", "_positiont",
"_color", "_fog", "_depth", "_sample"
};
721
722
static const char *get_D3D_register_string(Context *ctx,
723
RegisterType regtype,
724
725
int regnum, char *regnum_str,
size_t regnum_size)
726
{
727
const char *retval = NULL;
728
int has_number = 1;
729
730
switch (regtype)
731
{
732
case REG_TYPE_TEMP:
733
retval = "r";
734
735
break;
736
case REG_TYPE_INPUT:
737
retval = "v";
738
739
break;
740
case REG_TYPE_CONST:
741
742
743
retval = "c";
break;
744
case REG_TYPE_ADDRESS: // (or REG_TYPE_TEXTURE, same value.)
745
retval = shader_is_vertex(ctx) ? "a" : "t";
746
747
break;
748
case REG_TYPE_RASTOUT:
749
switch ((RastOutType) regnum)
750
{
751
752
753
case RASTOUT_TYPE_POSITION: retval = "oPos"; break;
case RASTOUT_TYPE_FOG: retval = "oFog"; break;
case RASTOUT_TYPE_POINT_SIZE: retval = "oPts"; break;
754
} // switch
755
has_number = 0;
756
757
break;
758
case REG_TYPE_ATTROUT:
759
retval = "oD";
760
761
break;
762
case REG_TYPE_OUTPUT: // (or REG_TYPE_TEXCRDOUT, same value.)
763
if (shader_is_vertex(ctx) && shader_version_atleast(ctx, 3, 0))
764
retval = "o";
765
else
766
retval = "oT";
767
768
break;
769
case REG_TYPE_CONSTINT:
770
retval = "i";
771
772
break;
773
case REG_TYPE_COLOROUT:
774
retval = "oC";
775
776
break;
777
case REG_TYPE_DEPTHOUT:
778
retval = "oDepth";
779
has_number = 0;
780
781
break;
782
case REG_TYPE_SAMPLER:
783
retval = "s";
784
785
break;
786
case REG_TYPE_CONSTBOOL:
787
retval = "b";
788
789
break;
790
case REG_TYPE_LOOP:
791
retval = "aL";
792
has_number = 0;
793
794
break;
795
case REG_TYPE_MISCTYPE:
796
switch ((const MiscTypeType) regnum)
797
{
798
799
case MISCTYPE_TYPE_POSITION: retval = "vPos"; break;
case MISCTYPE_TYPE_FACE: retval = "vFace"; break;
800
} // switch
801
has_number = 0;
802
803
break;
804
case REG_TYPE_LABEL:
805
retval = "l";
806
807
break;
808
case REG_TYPE_PREDICATE:
809
retval = "p";
810
break;
811
812
813
814
815
816
817
//case REG_TYPE_TEMPFLOAT16: // !!! FIXME: don't know this asm string
default:
fail(ctx, "unknown register type");
retval = "???";
has_number = 0;
break;
818
819
} // switch
820
821
822
823
824
if (has_number)
snprintf(regnum_str, regnum_size, "%u", (uint) regnum);
else
regnum_str[0] = '\0';
825
826
827
828
return retval;
} // get_D3D_register_string
829
830
831
832
833
834
835
836
837
#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,
838
static const char *make_D3D_srcarg_string_in_buf(Context *ctx,
839
840
const SourceArgInfo *arg,
char *buf, size_t buflen)
841
842
843
{
const char *premod_str = "";
const char *postmod_str = "";
844
switch (arg->src_mod)
845
846
847
848
849
850
851
852
853
854
855
{
case SRCMOD_NEGATE:
premod_str = "-";
break;
case SRCMOD_BIASNEGATE:
premod_str = "-";
// fall through.
case SRCMOD_BIAS:
postmod_str = "_bias";
break;
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
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;
893
894
895
896
case SRCMOD_NONE:
case SRCMOD_TOTAL:
break; // stop compiler whining.
897
898
899
900
} // switch
char regnum_str[16];
901
902
903
const char *regtype_str = get_D3D_register_string(ctx, arg->regtype,
arg->regnum, regnum_str,
sizeof (regnum_str));
904
905
906
907
if (regtype_str == NULL)
{
fail(ctx, "Unknown source register type.");
908
return "";
909
910
} // if
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
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
935
936
char swizzle_str[6];
int i = 0;
937
const int scalar = scalar_register(ctx->shader_type, arg->regtype, arg->regnum);
938
if (!scalar && !no_swizzle(arg->swizzle))
939
940
{
swizzle_str[i++] = '.';
941
942
943
944
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];
945
946
947
948
// .xyzz is the same as .xyz, .z is the same as .zzzz, etc.
while (swizzle_str[i-1] == swizzle_str[i-2])
i--;
949
950
951
952
} // if
swizzle_str[i] = '\0';
assert(i < sizeof (swizzle_str));
953
// !!! FIXME: c12[a0.x] actually needs to be c[a0.x + 12]
954
955
956
957
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);
958
// !!! FIXME: make sure the scratch buffer was large enough.
959
return buf;
960
} // make_D3D_srcarg_string_in_buf
961
962
963
static const char *make_D3D_destarg_string(Context *ctx)
964
{
965
const DestArgInfo *arg = &ctx->dest_arg;
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
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;
994
const int scalar = scalar_register(ctx->shader_type, arg->regtype, arg->regnum);
995
if (!scalar && !writemask_xyzw(arg->writemask))
996
997
998
999
1000
{
writemask_str[i++] = '.';
if (arg->writemask0) writemask_str[i++] = 'x';
if (arg->writemask1) writemask_str[i++] = 'y';
if (arg->writemask2) writemask_str[i++] = 'z';