/
mojoshader.c
7514 lines (6338 loc) · 242 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: this file really needs to be split up.
12
13
// !!! FIXME: I keep changing coding styles for symbols and typedefs.
14
15
16
// Shader bytecode format is described at MSDN:
// http://msdn2.microsoft.com/en-us/library/ms800307.aspx
17
#include <stdio.h>
18
#include <string.h>
19
#include <stdlib.h>
20
#include <stdarg.h>
21
#include <assert.h>
22
23
#include "mojoshader.h"
24
25
26
27
28
29
30
// This is the highest shader version we currently support.
#define MAX_SHADER_MAJOR 3
#define MAX_SHADER_MINOR 0
31
32
33
// If SUPPORT_PROFILE_* isn't defined, we assume an implicit desire to support.
// You get all the profiles unless you go out of your way to disable them.
34
35
36
37
38
#ifndef SUPPORT_PROFILE_D3D
#define SUPPORT_PROFILE_D3D 1
#endif
39
40
41
42
#ifndef SUPPORT_PROFILE_PASSTHROUGH
#define SUPPORT_PROFILE_PASSTHROUGH 1
#endif
43
44
45
46
#ifndef SUPPORT_PROFILE_GLSL
#define SUPPORT_PROFILE_GLSL 1
#endif
47
48
49
50
#ifndef SUPPORT_PROFILE_ARB1
#define SUPPORT_PROFILE_ARB1 1
#endif
51
52
53
54
// Get basic wankery out of the way here...
typedef unsigned int uint; // this is a printf() helper. don't use for code.
55
56
#ifdef _MSC_VER
57
#include <malloc.h>
58
59
#define snprintf _snprintf
typedef unsigned __int8 uint8;
60
typedef unsigned __int16 uint16;
61
62
typedef unsigned __int32 uint32;
typedef unsigned __int32 int32;
63
64
65
// Warning Level 4 considered harmful. :)
#pragma warning(disable: 4100) // "unreferenced formal parameter"
#pragma warning(disable: 4389) // "signed/unsigned mismatch"
66
67
#else
#include <stdint.h>
68
typedef uint8_t uint8;
69
typedef uint16_t uint16;
70
typedef uint32_t uint32;
71
typedef int32_t int32;
72
#endif
73
74
75
76
77
78
79
#ifdef __GNUC__
#define ISPRINTF(x,y) __attribute__((format (printf, x, y)))
#else
#define ISPRINTF(x,y)
#endif
80
81
#define STATICARRAYLEN(x) ( (sizeof ((x))) / (sizeof ((x)[0])) )
82
83
84
85
86
87
#ifdef _WINDOWS // !!! FIXME: bleh
const char *endline_str = "\r\n";
#else
const char *endline_str = "\n";
#endif
88
89
90
// we need to reference this by explicit value occasionally.
#define OPCODE_RET 28
91
92
93
94
95
// Special-case return values from the parsing pipeline...
#define FAIL (-1)
#define NOFAIL (-2)
#define END_OF_STREAM (-3)
96
97
98
99
100
101
102
103
104
// 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
105
106
107
108
109
static inline uint16 SWAP16(uint16 x)
{
__asm__ __volatile__("lhbrx %0,0,%1" : "=r" (x) : "r" (&x));
return x;
} // SWAP16
110
111
112
113
114
115
#elif defined(__POWERPC__)
static inline uint32 SWAP32(uint32 x)
{
return ( (((x) >> 24) & 0x000000FF) | (((x) >> 8) & 0x0000FF00) |
(((x) << 8) & 0x00FF0000) | (((x) << 24) & 0xFF000000) );
} // SWAP32
116
117
118
119
static inline uint16 SWAP16(uint16 x)
{
return ( (((x) >> 8) & 0x00FF) | (((x) << 8) & 0xFF00) );
} // SWAP16
120
#else
121
# define SWAP16(x) (x)
122
123
124
# define SWAP32(x) (x)
#endif
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
typedef enum
{
REG_TYPE_TEMP = 0,
REG_TYPE_INPUT = 1,
REG_TYPE_CONST = 2,
REG_TYPE_ADDRESS = 3,
REG_TYPE_TEXTURE = 3, // ALSO 3!
REG_TYPE_RASTOUT = 4,
REG_TYPE_ATTROUT = 5,
REG_TYPE_TEXCRDOUT = 6,
REG_TYPE_OUTPUT = 6, // ALSO 6!
REG_TYPE_CONSTINT = 7,
REG_TYPE_COLOROUT = 8,
REG_TYPE_DEPTHOUT = 9,
REG_TYPE_SAMPLER = 10,
REG_TYPE_CONST2 = 11,
REG_TYPE_CONST3 = 12,
REG_TYPE_CONST4 = 13,
REG_TYPE_CONSTBOOL = 14,
REG_TYPE_LOOP = 15,
REG_TYPE_TEMPFLOAT16 = 16,
REG_TYPE_MISCTYPE = 17,
REG_TYPE_LABEL = 18,
REG_TYPE_PREDICATE = 19,
REG_TYPE_MAX = 19
} RegisterType;
151
152
153
154
155
156
157
158
typedef enum
{
TEXTURE_TYPE_2D = 2,
TEXTURE_TYPE_CUBE = 3,
TEXTURE_TYPE_VOLUME = 4,
} TextureType;
159
// predeclare.
160
typedef struct Context Context;
161
struct ConstantsList;
162
163
// one emit function for each opcode in each profile.
164
typedef void (*emit_function)(Context *ctx);
165
166
// one emit function for starting output in each profile.
167
typedef void (*emit_start)(Context *ctx, const char *profilestr);
168
169
// one emit function for ending output in each profile.
170
typedef void (*emit_end)(Context *ctx);
171
172
173
174
// one emit function for phase opcode output in each profile.
typedef void (*emit_phase)(Context *ctx);
175
176
177
178
179
180
// 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);
181
// one emit function for relative uniform arrays in each profile.
182
typedef void (*emit_array)(Context *ctx, int base, int size);
183
184
185
186
187
188
// 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);
189
// one emit function for uniforms in each profile.
190
191
typedef void (*emit_uniform)(Context *ctx, RegisterType regtype, int regnum,
int arraybase, int arraysize);
192
193
194
195
// one emit function for samplers in each profile.
typedef void (*emit_sampler)(Context *ctx, int stage, TextureType ttype);
196
197
198
199
// one emit function for attributes in each profile.
typedef void (*emit_attribute)(Context *ctx, RegisterType regtype, int regnum,
MOJOSHADER_usage usage, int index, int wmask);
200
201
202
// one args function for each possible sequence of opcode arguments.
typedef int (*args_function)(Context *ctx);
203
// one state function for each opcode where we have state machine updates.
204
typedef void (*state_function)(Context *ctx);
205
206
207
208
209
// 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.
210
typedef const char *(*const_array_varname_function)(Context *c, int base, int size);
211
212
213
214
215
216
typedef struct
{
const char *name;
emit_start start_emitter;
emit_end end_emitter;
217
emit_phase phase_emitter;
218
emit_global global_emitter;
219
emit_array array_emitter;
220
emit_const_array const_array_emitter;
221
emit_uniform uniform_emitter;
222
emit_sampler sampler_emitter;
223
emit_attribute attribute_emitter;
224
emit_finalize finalize_emitter;
225
226
varname_function get_varname;
const_array_varname_function get_const_array_varname;
227
} Profile;
228
229
230
231
232
233
234
typedef enum
{
RASTOUT_TYPE_POSITION = 0,
RASTOUT_TYPE_FOG = 1,
RASTOUT_TYPE_POINT_SIZE = 2,
RASTOUT_TYPE_MAX = 2
235
} RastOutType;
236
237
238
239
240
241
typedef enum
{
MISCTYPE_TYPE_POSITION = 0,
MISCTYPE_TYPE_FACE = 1,
MISCTYPE_TYPE_MAX = 1
242
} MiscTypeType;
243
244
245
246
247
// 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.
248
typedef struct OutputListNode
249
250
{
char *str;
251
252
253
254
255
256
257
struct OutputListNode *next;
} OutputListNode;
typedef struct OutputList
{
OutputListNode head;
OutputListNode *tail;
258
259
} OutputList;
260
261
262
263
264
265
typedef struct ConstantsList
{
MOJOSHADER_constant constant;
struct ConstantsList *next;
} ConstantsList;
266
267
268
269
270
typedef struct VariableList
{
MOJOSHADER_uniformType type;
int index;
int count;
271
ConstantsList *constant;
272
int used;
273
274
struct VariableList *next;
} VariableList;
275
276
277
278
279
typedef struct RegisterList
{
RegisterType regtype;
int regnum;
280
281
282
MOJOSHADER_usage usage;
int index;
int writemask;
283
int misc;
284
const VariableList *array;
285
286
287
struct RegisterList *next;
} RegisterList;
288
289
290
291
292
// result modifiers.
#define MOD_SATURATE 0x01
#define MOD_PP 0x02
#define MOD_CENTROID 0x04
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
// 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
311
} SourceMod;
312
313
314
315
typedef struct
{
316
const uint32 *token; // this is the unmolested token in the stream.
317
318
int regnum;
int relative;
319
320
321
322
323
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
324
int orig_writemask; // writemask before mojoshader tweaks it.
325
326
int result_mod;
int result_shift;
327
RegisterType regtype;
328
329
330
331
} DestArgInfo;
typedef struct
{
332
const uint32 *token; // this is the unmolested token in the stream.
333
int regnum;
334
int swizzle; // xyzw (all four, not split out).
335
336
337
338
int swizzle_x;
int swizzle_y;
int swizzle_z;
int swizzle_w;
339
SourceMod src_mod;
340
RegisterType regtype;
341
342
343
344
int relative;
RegisterType relative_regtype;
int relative_regnum;
int relative_component;
345
const VariableList *relative_array;
346
347
348
} SourceArgInfo;
349
350
#define SCRATCH_BUFFER_SIZE 128
351
#define SCRATCH_BUFFERS 32
352
353
// !!! FIXME: the scratch buffers make Context pretty big.
354
355
// !!! FIXME: might be worth having one set of static scratch buffers that
// !!! FIXME: are mutex protected?
356
357
// Context...this is state that changes as we parse through a shader...
358
struct Context
359
{
360
361
MOJOSHADER_malloc malloc;
MOJOSHADER_free free;
362
void *malloc_data;
363
364
const uint32 *tokens;
uint32 tokencount;
365
366
OutputList *output;
OutputList globals;
367
OutputList helpers;
368
OutputList subroutines;
369
OutputList mainline_intro;
370
OutputList mainline;
371
372
OutputList ignore;
OutputList *output_stack[2];
373
uint8 *output_bytes; // can be used instead of the OutputLists.
374
375
int indent_stack[2];
int output_stack_len;
376
int output_len; // total strlen; prevents walking the lists just to malloc.
377
int indent;
378
const char *shader_type_str;
379
380
381
const char *endline;
int endline_len;
const char *failstr;
382
char scratch[SCRATCH_BUFFERS][SCRATCH_BUFFER_SIZE];
383
384
int scratchidx; // current scratch buffer.
int profileid;
385
const Profile *profile;
386
MOJOSHADER_shaderType shader_type;
387
388
uint8 major_ver;
uint8 minor_ver;
389
DestArgInfo dest_arg;
390
SourceArgInfo source_args[5];
391
SourceArgInfo predicate_arg; // for predicated instructions.
392
uint32 dwords[4];
393
uint32 version_token;
394
int instruction_count;
395
uint32 instruction_controls;
396
397
uint32 previous_opcode;
int loops;
398
int reps;
399
int max_reps;
400
int cmps;
401
402
int scratch_registers;
int max_scratch_registers;
403
404
405
int branch_labels_stack_index;
int branch_labels_stack[32];
int assigned_branch_labels;
406
int assigned_vertex_attributes;
407
int last_address_reg_component;
408
409
RegisterList used_registers;
RegisterList defined_registers;
410
411
int constant_count;
ConstantsList *constants;
412
413
int uniform_count;
RegisterList uniforms;
414
415
int attribute_count;
RegisterList attributes;
416
417
int sampler_count;
RegisterList samplers;
418
419
VariableList *variables; // variables to register mapping.
int have_ctab:1;
420
int determined_constants_arrays:1;
421
422
int predicated:1;
int support_nv2:1;
423
int support_nv3:1;
424
int support_glsl120:1;
425
int glsl_generated_lit_opcode:1;
426
427
428
};
429
430
// Convenience functions for allocators...
431
432
433
434
435
436
static MOJOSHADER_parseData out_of_mem_data = {
"Out of memory", 0, 0, 0, 0, MOJOSHADER_TYPE_UNKNOWN, 0, 0, 0, 0
};
static const char *out_of_mem_str = "Out of memory";
static inline int out_of_memory(Context *ctx)
437
{
438
439
440
441
442
if (ctx->failstr == NULL)
ctx->failstr = out_of_mem_str; // fail() would call malloc().
return FAIL;
} // out_of_memory
443
static inline void *Malloc(Context *ctx, const size_t len)
444
{
445
void *retval = ctx->malloc((int) len, ctx->malloc_data);
446
447
448
if (retval == NULL)
out_of_memory(ctx);
return retval;
449
450
451
} // Malloc
452
static inline void Free(Context *ctx, void *ptr)
453
{
454
if (ptr != NULL) // check for NULL in case of dumb free() impl.
455
ctx->free(ptr, ctx->malloc_data);
456
457
458
} // Free
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
// 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...
483
484
485
486
487
static inline uint32 ver_ui32(const uint8 major, const uint8 minor)
{
return ( (((uint32) major) << 16) | (((minor) == 0xFF) ? 0 : (minor)) );
} // version_ui32
488
static inline int shader_version_supported(const uint8 maj, const uint8 min)
489
490
491
492
{
return (ver_ui32(maj,min) <= ver_ui32(MAX_SHADER_MAJOR, MAX_SHADER_MINOR));
} // shader_version_supported
493
494
static inline int shader_version_atleast(const Context *ctx, const uint8 maj,
const uint8 min)
495
496
497
498
{
return (ver_ui32(ctx->major_ver, ctx->minor_ver) >= ver_ui32(maj, min));
} // shader_version_atleast
499
500
501
502
503
504
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
505
506
507
508
509
510
511
512
513
514
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
515
516
517
518
519
520
static inline int isfail(const Context *ctx)
{
return (ctx->failstr != NULL);
} // isfail
521
522
static inline char *get_scratch_buffer(Context *ctx)
523
{
524
525
526
527
528
529
530
531
532
533
534
535
if ((ctx->scratchidx >= SCRATCH_BUFFERS) && !isfail(ctx))
{
// can't call fail() here, since it calls back into here.
const char *errstr = "BUG: overflowed scratch buffers";
char *failstr = (char *) Malloc(ctx, strlen(errstr) + 1);
if (failstr != NULL)
{
strcpy(failstr, errstr);
ctx->failstr = failstr;
} // if
} // if
536
ctx->scratchidx = (ctx->scratchidx + 1) % SCRATCH_BUFFERS;
537
return ctx->scratch[ctx->scratchidx];
538
539
} // get_scratch_buffer
540
541
542
static int failf(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static int failf(Context *ctx, const char *fmt, ...)
543
{
544
545
if (ctx->failstr == NULL) // don't change existing error.
{
546
char *scratch = get_scratch_buffer(ctx);
547
548
va_list ap;
va_start(ap, fmt);
549
const int len = vsnprintf(scratch, SCRATCH_BUFFER_SIZE, fmt, ap);
550
551
va_end(ap);
552
char *failstr = (char *) Malloc(ctx, len + 1);
553
if (failstr != NULL)
554
555
{
// see comments about scratch buffer overflow in output_line().
556
if (len < SCRATCH_BUFFER_SIZE)
557
558
559
560
561
562
563
564
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;
565
} // if
566
567
568
569
570
571
} // if
return FAIL;
} // failf
572
static inline int fail(Context *ctx, const char *reason)
573
574
575
576
577
{
return failf(ctx, "%s", reason);
} // fail
578
579
static int output_line(Context *ctx, const char *fmt, ...) ISPRINTF(2,3);
static int output_line(Context *ctx, const char *fmt, ...)
580
{
581
582
OutputListNode *item = NULL;
583
if (isfail(ctx))
584
585
return FAIL; // we failed previously, don't go on...
586
char *scratch = get_scratch_buffer(ctx);
587
588
const int indent = ctx->indent;
589
590
if (indent > 0)
memset(scratch, '\t', indent);
591
592
593
va_list ap;
va_start(ap, fmt);
594
const int len = vsnprintf(scratch+indent, SCRATCH_BUFFER_SIZE-indent, fmt, ap) + indent;
595
596
va_end(ap);
597
item = (OutputListNode *) Malloc(ctx, sizeof (OutputListNode));
598
if (item == NULL)
599
return FAIL;
600
601
item->str = (char *) Malloc(ctx, len + 1);
602
if (item->str == NULL)
603
{
604
Free(ctx, item);
605
return FAIL;
606
607
} // if
608
609
610
// 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().
611
if (len < SCRATCH_BUFFER_SIZE)
612
613
614
strcpy(item->str, scratch); // copy it over.
else
{
615
616
if (indent > 0)
memset(item->str, '\t', indent);
617
va_start(ap, fmt);
618
vsnprintf(item->str+indent, len + 1, fmt, ap); // rebuild it.
619
620
va_end(ap);
} // else
621
622
item->next = NULL;
623
624
625
ctx->output->tail->next = item;
ctx->output->tail = item;
626
ctx->output_len += len + ctx->endline_len;
627
628
629
630
631
return 0;
} // output_line
632
633
634
635
636
637
638
// this is just to stop gcc whining.
static inline int output_blank_line(Context *ctx)
{
return output_line(ctx, "%s", "");
} // output_blank_line
639
// !!! FIXME: this is sort of nasty.
640
641
static void floatstr(Context *ctx, char *buf, size_t bufsize, float f,
int leavedecimal)
642
643
{
const size_t len = snprintf(buf, bufsize, "%f", f);
644
if ((len+2) >= bufsize)
645
646
647
648
649
650
fail(ctx, "BUG: internal buffer is too small");
else
{
char *end = buf + len;
char *ptr = strchr(buf, '.');
if (ptr == NULL)
651
652
653
{
if (leavedecimal)
strcat(buf, ".0");
654
return; // done.
655
} // if
656
657
658
659
660
661
662
663
664
while (--end != ptr)
{
if (*end != '0')
{
end++;
break;
} // if
} // while
665
666
if ((leavedecimal) && (end == ptr))
end += 2;
667
668
669
670
*end = '\0'; // chop extra '0' or all decimal places off.
} // else
} // floatstr
671
672
673
// Deal with register lists... !!! FIXME: I sort of hate this.
674
static void free_reglist(MOJOSHADER_free f, void *d, RegisterList *item)
675
676
677
678
{
while (item != NULL)
{
RegisterList *next = item->next;
679
f(item, d);
680
681
682
683
684
685
686
687
688
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
689
690
691
static RegisterList *reglist_insert(Context *ctx, RegisterList *prev,
const RegisterType regtype,
const int regnum)
692
693
694
695
696
697
698
{
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)
699
return item; // already set, so we're done.
700
701
702
703
704
705
706
707
708
709
710
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).
711
item = (RegisterList *) Malloc(ctx, sizeof (RegisterList));
712
if (item != NULL)
713
714
715
{
item->regtype = regtype;
item->regnum = regnum;
716
item->usage = MOJOSHADER_USAGE_UNKNOWN;
717
718
item->index = 0;
item->writemask = 0;
719
item->misc = 0;
720
item->array = NULL;
721
722
item->next = prev->next;
prev->next = item;
723
} // if
724
725
return item;
726
727
} // reglist_insert
728
729
static RegisterList *reglist_find(RegisterList *prev, const RegisterType rtype,
const int regnum)
730
{
731
732
const uint32 newval = reg_to_ui32(rtype, regnum);
RegisterList *item = prev->next;
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
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.
749
750
751
752
753
754
755
} // reglist_find
static inline const RegisterList *reglist_exists(RegisterList *prev,
const RegisterType regtype,
const int regnum)
{
return (reglist_find(prev, regtype, regnum));
756
757
758
759
760
761
762
763
764
765
766
767
} // 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);
768
} // get_used_register
769
770
771
772
773
774
775
776
777
778
779
780
781
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
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
static const RegisterList *declared_attribute(Context *ctx,
const MOJOSHADER_usage usage,
const int index)
{
const RegisterList *item = ctx->attributes.next;
while (item != NULL)
{
if ((item->usage == usage) && (item->index == index))
return item;
item = item->next;
} // while
return NULL;
} // declared_attribute
797
798
799
800
801
802
803
804
805
806
static void add_attribute_register(Context *ctx, const RegisterType rtype,
const int regnum, const MOJOSHADER_usage usage,
const int index, const int writemask)
{
RegisterList *item = reglist_insert(ctx, &ctx->attributes, rtype, regnum);
item->usage = usage;
item->index = index;
item->writemask = writemask;
} // add_attribute_register
807
808
809
static inline void add_sampler(Context *ctx, const RegisterType rtype,
const int regnum, const TextureType ttype)
{
810
// !!! FIXME: make sure it doesn't exist?
811
812
813
814
RegisterList *item = reglist_insert(ctx, &ctx->samplers, rtype, regnum);
item->index = (int) ttype;
} // add_sampler
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
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
846
847
848
849
850
851
852
853
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
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
static inline int scalar_register(const RegisterType regtype, const int regnum)
{
switch (regtype)
{
case REG_TYPE_DEPTHOUT:
case REG_TYPE_CONSTBOOL:
case REG_TYPE_PREDICATE:
case REG_TYPE_LOOP:
return 1;
case REG_TYPE_MISCTYPE:
if ( ((const MiscTypeType) regnum) == MISCTYPE_TYPE_FACE )
return 1;
return 0;
default: break;
} // switch
return 0;
} // scalar_register
876
877
static inline int no_swizzle(const int swizzle)
{
878
return (swizzle == 0xE4); // 0xE4 == 11100100 ... 0 1 2 3. No swizzle.
879
880
881
} // no_swizzle
882
883
884
885
886
static inline int vecsize_from_writemask(const int m)
{
return (m & 1) + ((m >> 1) & 1) + ((m >> 2) & 1) + ((m >> 3) & 1);
} // vecsize_from_writemask
887
888
889
890
891
892
893
894
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
895
static int allocate_branch_label(Context *ctx)
896
{
897
898
return ctx->assigned_branch_labels++;
} // allocate_branch_label
899
900
901
902
// D3D stuff that's used in more than just the d3d profile...
903
904
905
static const char swizzle_channels[] = { 'x', 'y', 'z', 'w' };
906
907
908
909
910
static const char *usagestrs[] = {
"_position", "_blendweight", "_blendindices", "_normal", "_psize",
"_texcoord", "_tangent", "_binormal", "_tessfactor", "_positiont",
"_color", "_fog", "_depth", "_sample"
};
911
912
static const char *get_D3D_register_string(Context *ctx,
913
RegisterType regtype,
914
915
int regnum, char *regnum_str,
size_t regnum_size)
916
{
917
const char *retval = NULL;
918
int has_number = 1;
919
920
switch (regtype)
921
{
922
case REG_TYPE_TEMP:
923
retval = "r";
924
925
break;
926
case REG_TYPE_INPUT:
927
retval = "v";
928
929
break;
930
case REG_TYPE_CONST:
931
932
933
retval = "c";
break;
934
case REG_TYPE_ADDRESS: // (or REG_TYPE_TEXTURE, same value.)
935
retval = shader_is_vertex(ctx) ? "a" : "t";
936
937
break;
938
case REG_TYPE_RASTOUT:
939
switch ((RastOutType) regnum)
940
{
941
942
943
case RASTOUT_TYPE_POSITION: retval = "oPos"; break;
case RASTOUT_TYPE_FOG: retval = "oFog"; break;
case RASTOUT_TYPE_POINT_SIZE: retval = "oPts"; break;
944
} // switch
945
has_number = 0;
946
947
break;
948
case REG_TYPE_ATTROUT:
949
retval = "oD";
950
951
break;
952
case REG_TYPE_OUTPUT: // (or REG_TYPE_TEXCRDOUT, same value.)
953
if (shader_is_vertex(ctx) && shader_version_atleast(ctx, 3, 0))
954
retval = "o";
955
else
956
retval = "oT";
957
958
break;
959
case REG_TYPE_CONSTINT:
960
retval = "i";
961
962
break;
963
case REG_TYPE_COLOROUT:
964
retval = "oC";
965
966
break;
967
case REG_TYPE_DEPTHOUT:
968
retval = "oDepth";
969
has_number = 0;
970
971
break;
972
case REG_TYPE_SAMPLER:
973
retval = "s";
974
975
break;
976
case REG_TYPE_CONSTBOOL:
977
retval = "b";
978
979
break;
980
case REG_TYPE_LOOP:
981
retval = "aL";
982
has_number = 0;
983
984
break;
985
case REG_TYPE_MISCTYPE:
986
switch ((const MiscTypeType) regnum)
987
{
988
989
case MISCTYPE_TYPE_POSITION: retval = "vPos"; break;
case MISCTYPE_TYPE_FACE: retval = "vFace"; break;
990
} // switch
991
has_number = 0;
992
993
break;
994
case REG_TYPE_LABEL:
995
retval = "l";
996
997
break;
998
case REG_TYPE_PREDICATE:
999
retval = "p";
1000
break;