forked from KhronosGroup/SPIRV-Cross
-
Notifications
You must be signed in to change notification settings - Fork 0
/
spirv_glsl.cpp
16491 lines (14401 loc) · 524 KB
/
spirv_glsl.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* Copyright 2015-2021 Arm Limited
* SPDX-License-Identifier: Apache-2.0 OR MIT
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* At your option, you may choose to accept this material under either:
* 1. The Apache License, Version 2.0, found at <http://www.apache.org/licenses/LICENSE-2.0>, or
* 2. The MIT License, found at <http://opensource.org/licenses/MIT>.
*/
#include "spirv_glsl.hpp"
#include "GLSL.std.450.h"
#include "spirv_common.hpp"
#include <algorithm>
#include <assert.h>
#include <cmath>
#include <limits>
#include <locale.h>
#include <utility>
#ifndef _WIN32
#include <langinfo.h>
#endif
#include <locale.h>
using namespace spv;
using namespace SPIRV_CROSS_NAMESPACE;
using namespace std;
enum ExtraSubExpressionType
{
// Create masks above any legal ID range to allow multiple address spaces into the extra_sub_expressions map.
EXTRA_SUB_EXPRESSION_TYPE_STREAM_OFFSET = 0x10000000,
EXTRA_SUB_EXPRESSION_TYPE_AUX = 0x20000000
};
static bool is_unsigned_opcode(Op op)
{
// Don't have to be exhaustive, only relevant for legacy target checking ...
switch (op)
{
case OpShiftRightLogical:
case OpUGreaterThan:
case OpUGreaterThanEqual:
case OpULessThan:
case OpULessThanEqual:
case OpUConvert:
case OpUDiv:
case OpUMod:
case OpUMulExtended:
case OpConvertUToF:
case OpConvertFToU:
return true;
default:
return false;
}
}
static bool is_unsigned_glsl_opcode(GLSLstd450 op)
{
// Don't have to be exhaustive, only relevant for legacy target checking ...
switch (op)
{
case GLSLstd450UClamp:
case GLSLstd450UMin:
case GLSLstd450UMax:
case GLSLstd450FindUMsb:
return true;
default:
return false;
}
}
static bool packing_is_vec4_padded(BufferPackingStandard packing)
{
switch (packing)
{
case BufferPackingHLSLCbuffer:
case BufferPackingHLSLCbufferPackOffset:
case BufferPackingStd140:
case BufferPackingStd140EnhancedLayout:
return true;
default:
return false;
}
}
static bool packing_is_hlsl(BufferPackingStandard packing)
{
switch (packing)
{
case BufferPackingHLSLCbuffer:
case BufferPackingHLSLCbufferPackOffset:
return true;
default:
return false;
}
}
static bool packing_has_flexible_offset(BufferPackingStandard packing)
{
switch (packing)
{
case BufferPackingStd140:
case BufferPackingStd430:
case BufferPackingScalar:
case BufferPackingHLSLCbuffer:
return false;
default:
return true;
}
}
static bool packing_is_scalar(BufferPackingStandard packing)
{
switch (packing)
{
case BufferPackingScalar:
case BufferPackingScalarEnhancedLayout:
return true;
default:
return false;
}
}
static BufferPackingStandard packing_to_substruct_packing(BufferPackingStandard packing)
{
switch (packing)
{
case BufferPackingStd140EnhancedLayout:
return BufferPackingStd140;
case BufferPackingStd430EnhancedLayout:
return BufferPackingStd430;
case BufferPackingHLSLCbufferPackOffset:
return BufferPackingHLSLCbuffer;
case BufferPackingScalarEnhancedLayout:
return BufferPackingScalar;
default:
return packing;
}
}
void CompilerGLSL::init()
{
if (ir.source.known)
{
options.es = ir.source.es;
options.version = ir.source.version;
}
// Query the locale to see what the decimal point is.
// We'll rely on fixing it up ourselves in the rare case we have a comma-as-decimal locale
// rather than setting locales ourselves. Settings locales in a safe and isolated way is rather
// tricky.
#ifdef _WIN32
// On Windows, localeconv uses thread-local storage, so it should be fine.
const struct lconv *conv = localeconv();
if (conv && conv->decimal_point)
current_locale_radix_character = *conv->decimal_point;
#elif defined(__ANDROID__) && __ANDROID_API__ < 26
// nl_langinfo is not supported on this platform, fall back to the worse alternative.
const struct lconv *conv = localeconv();
if (conv && conv->decimal_point)
current_locale_radix_character = *conv->decimal_point;
#else
// localeconv, the portable function is not MT safe ...
const char *decimal_point = nl_langinfo(RADIXCHAR);
if (decimal_point && *decimal_point != '\0')
current_locale_radix_character = *decimal_point;
#endif
}
static const char *to_pls_layout(PlsFormat format)
{
switch (format)
{
case PlsR11FG11FB10F:
return "layout(r11f_g11f_b10f) ";
case PlsR32F:
return "layout(r32f) ";
case PlsRG16F:
return "layout(rg16f) ";
case PlsRGB10A2:
return "layout(rgb10_a2) ";
case PlsRGBA8:
return "layout(rgba8) ";
case PlsRG16:
return "layout(rg16) ";
case PlsRGBA8I:
return "layout(rgba8i)";
case PlsRG16I:
return "layout(rg16i) ";
case PlsRGB10A2UI:
return "layout(rgb10_a2ui) ";
case PlsRGBA8UI:
return "layout(rgba8ui) ";
case PlsRG16UI:
return "layout(rg16ui) ";
case PlsR32UI:
return "layout(r32ui) ";
default:
return "";
}
}
static SPIRType::BaseType pls_format_to_basetype(PlsFormat format)
{
switch (format)
{
default:
case PlsR11FG11FB10F:
case PlsR32F:
case PlsRG16F:
case PlsRGB10A2:
case PlsRGBA8:
case PlsRG16:
return SPIRType::Float;
case PlsRGBA8I:
case PlsRG16I:
return SPIRType::Int;
case PlsRGB10A2UI:
case PlsRGBA8UI:
case PlsRG16UI:
case PlsR32UI:
return SPIRType::UInt;
}
}
static uint32_t pls_format_to_components(PlsFormat format)
{
switch (format)
{
default:
case PlsR32F:
case PlsR32UI:
return 1;
case PlsRG16F:
case PlsRG16:
case PlsRG16UI:
case PlsRG16I:
return 2;
case PlsR11FG11FB10F:
return 3;
case PlsRGB10A2:
case PlsRGBA8:
case PlsRGBA8I:
case PlsRGB10A2UI:
case PlsRGBA8UI:
return 4;
}
}
const char *CompilerGLSL::vector_swizzle(int vecsize, int index)
{
static const char *const swizzle[4][4] = {
{ ".x", ".y", ".z", ".w" },
{ ".xy", ".yz", ".zw", nullptr },
{ ".xyz", ".yzw", nullptr, nullptr },
#if defined(__GNUC__) && (__GNUC__ == 9)
// This works around a GCC 9 bug, see details in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90947.
// This array ends up being compiled as all nullptrs, tripping the assertions below.
{ "", nullptr, nullptr, "$" },
#else
{ "", nullptr, nullptr, nullptr },
#endif
};
assert(vecsize >= 1 && vecsize <= 4);
assert(index >= 0 && index < 4);
assert(swizzle[vecsize - 1][index]);
return swizzle[vecsize - 1][index];
}
void CompilerGLSL::reset(uint32_t iteration_count)
{
// Sanity check the iteration count to be robust against a certain class of bugs where
// we keep forcing recompilations without making clear forward progress.
// In buggy situations we will loop forever, or loop for an unbounded number of iterations.
// Certain types of recompilations are considered to make forward progress,
// but in almost all situations, we'll never see more than 3 iterations.
// It is highly context-sensitive when we need to force recompilation,
// and it is not practical with the current architecture
// to resolve everything up front.
if (iteration_count >= options.force_recompile_max_debug_iterations && !is_force_recompile_forward_progress)
SPIRV_CROSS_THROW("Maximum compilation loops detected and no forward progress was made. Must be a SPIRV-Cross bug!");
// We do some speculative optimizations which should pretty much always work out,
// but just in case the SPIR-V is rather weird, recompile until it's happy.
// This typically only means one extra pass.
clear_force_recompile();
// Clear invalid expression tracking.
invalid_expressions.clear();
current_function = nullptr;
// Clear temporary usage tracking.
expression_usage_counts.clear();
forwarded_temporaries.clear();
suppressed_usage_tracking.clear();
// Ensure that we declare phi-variable copies even if the original declaration isn't deferred
flushed_phi_variables.clear();
reset_name_caches();
ir.for_each_typed_id<SPIRFunction>([&](uint32_t, SPIRFunction &func) {
func.active = false;
func.flush_undeclared = true;
});
ir.for_each_typed_id<SPIRVariable>([&](uint32_t, SPIRVariable &var) { var.dependees.clear(); });
ir.reset_all_of_type<SPIRExpression>();
ir.reset_all_of_type<SPIRAccessChain>();
statement_count = 0;
indent = 0;
current_loop_level = 0;
}
void CompilerGLSL::remap_pls_variables()
{
for (auto &input : pls_inputs)
{
auto &var = get<SPIRVariable>(input.id);
bool input_is_target = false;
if (var.storage == StorageClassUniformConstant)
{
auto &type = get<SPIRType>(var.basetype);
input_is_target = type.image.dim == DimSubpassData;
}
if (var.storage != StorageClassInput && !input_is_target)
SPIRV_CROSS_THROW("Can only use in and target variables for PLS inputs.");
var.remapped_variable = true;
}
for (auto &output : pls_outputs)
{
auto &var = get<SPIRVariable>(output.id);
if (var.storage != StorageClassOutput)
SPIRV_CROSS_THROW("Can only use out variables for PLS outputs.");
var.remapped_variable = true;
}
}
void CompilerGLSL::remap_ext_framebuffer_fetch(uint32_t input_attachment_index, uint32_t color_location, bool coherent)
{
subpass_to_framebuffer_fetch_attachment.push_back({ input_attachment_index, color_location });
inout_color_attachments.push_back({ color_location, coherent });
}
bool CompilerGLSL::location_is_framebuffer_fetch(uint32_t location) const
{
return std::find_if(begin(inout_color_attachments), end(inout_color_attachments),
[&](const std::pair<uint32_t, bool> &elem) {
return elem.first == location;
}) != end(inout_color_attachments);
}
bool CompilerGLSL::location_is_non_coherent_framebuffer_fetch(uint32_t location) const
{
return std::find_if(begin(inout_color_attachments), end(inout_color_attachments),
[&](const std::pair<uint32_t, bool> &elem) {
return elem.first == location && !elem.second;
}) != end(inout_color_attachments);
}
void CompilerGLSL::find_static_extensions()
{
ir.for_each_typed_id<SPIRType>([&](uint32_t, const SPIRType &type) {
if (type.basetype == SPIRType::Double)
{
if (options.es)
SPIRV_CROSS_THROW("FP64 not supported in ES profile.");
if (!options.es && options.version < 400)
require_extension_internal("GL_ARB_gpu_shader_fp64");
}
else if (type.basetype == SPIRType::Int64 || type.basetype == SPIRType::UInt64)
{
if (options.es)
SPIRV_CROSS_THROW("64-bit integers not supported in ES profile.");
if (!options.es)
require_extension_internal("GL_ARB_gpu_shader_int64");
}
else if (type.basetype == SPIRType::Half)
{
require_extension_internal("GL_EXT_shader_explicit_arithmetic_types_float16");
if (options.vulkan_semantics)
require_extension_internal("GL_EXT_shader_16bit_storage");
}
else if (type.basetype == SPIRType::SByte || type.basetype == SPIRType::UByte)
{
require_extension_internal("GL_EXT_shader_explicit_arithmetic_types_int8");
if (options.vulkan_semantics)
require_extension_internal("GL_EXT_shader_8bit_storage");
}
else if (type.basetype == SPIRType::Short || type.basetype == SPIRType::UShort)
{
require_extension_internal("GL_EXT_shader_explicit_arithmetic_types_int16");
if (options.vulkan_semantics)
require_extension_internal("GL_EXT_shader_16bit_storage");
}
});
auto &execution = get_entry_point();
switch (execution.model)
{
case ExecutionModelGLCompute:
if (!options.es && options.version < 430)
require_extension_internal("GL_ARB_compute_shader");
if (options.es && options.version < 310)
SPIRV_CROSS_THROW("At least ESSL 3.10 required for compute shaders.");
break;
case ExecutionModelGeometry:
if (options.es && options.version < 320)
require_extension_internal("GL_EXT_geometry_shader");
if (!options.es && options.version < 150)
require_extension_internal("GL_ARB_geometry_shader4");
if (execution.flags.get(ExecutionModeInvocations) && execution.invocations != 1)
{
// Instanced GS is part of 400 core or this extension.
if (!options.es && options.version < 400)
require_extension_internal("GL_ARB_gpu_shader5");
}
break;
case ExecutionModelTessellationEvaluation:
case ExecutionModelTessellationControl:
if (options.es && options.version < 320)
require_extension_internal("GL_EXT_tessellation_shader");
if (!options.es && options.version < 400)
require_extension_internal("GL_ARB_tessellation_shader");
break;
case ExecutionModelRayGenerationKHR:
case ExecutionModelIntersectionKHR:
case ExecutionModelAnyHitKHR:
case ExecutionModelClosestHitKHR:
case ExecutionModelMissKHR:
case ExecutionModelCallableKHR:
// NV enums are aliases.
if (options.es || options.version < 460)
SPIRV_CROSS_THROW("Ray tracing shaders require non-es profile with version 460 or above.");
if (!options.vulkan_semantics)
SPIRV_CROSS_THROW("Ray tracing requires Vulkan semantics.");
// Need to figure out if we should target KHR or NV extension based on capabilities.
for (auto &cap : ir.declared_capabilities)
{
if (cap == CapabilityRayTracingKHR || cap == CapabilityRayQueryKHR ||
cap == CapabilityRayTraversalPrimitiveCullingKHR)
{
ray_tracing_is_khr = true;
break;
}
}
if (ray_tracing_is_khr)
{
// In KHR ray tracing we pass payloads by pointer instead of location,
// so make sure we assign locations properly.
ray_tracing_khr_fixup_locations();
require_extension_internal("GL_EXT_ray_tracing");
}
else
require_extension_internal("GL_NV_ray_tracing");
break;
default:
break;
}
if (!pls_inputs.empty() || !pls_outputs.empty())
{
if (execution.model != ExecutionModelFragment)
SPIRV_CROSS_THROW("Can only use GL_EXT_shader_pixel_local_storage in fragment shaders.");
require_extension_internal("GL_EXT_shader_pixel_local_storage");
}
if (!inout_color_attachments.empty())
{
if (execution.model != ExecutionModelFragment)
SPIRV_CROSS_THROW("Can only use GL_EXT_shader_framebuffer_fetch in fragment shaders.");
if (options.vulkan_semantics)
SPIRV_CROSS_THROW("Cannot use EXT_shader_framebuffer_fetch in Vulkan GLSL.");
bool has_coherent = false;
bool has_incoherent = false;
for (auto &att : inout_color_attachments)
{
if (att.second)
has_coherent = true;
else
has_incoherent = true;
}
if (has_coherent)
require_extension_internal("GL_EXT_shader_framebuffer_fetch");
if (has_incoherent)
require_extension_internal("GL_EXT_shader_framebuffer_fetch_non_coherent");
}
if (options.separate_shader_objects && !options.es && options.version < 410)
require_extension_internal("GL_ARB_separate_shader_objects");
if (ir.addressing_model == AddressingModelPhysicalStorageBuffer64EXT)
{
if (!options.vulkan_semantics)
SPIRV_CROSS_THROW("GL_EXT_buffer_reference is only supported in Vulkan GLSL.");
if (options.es && options.version < 320)
SPIRV_CROSS_THROW("GL_EXT_buffer_reference requires ESSL 320.");
else if (!options.es && options.version < 450)
SPIRV_CROSS_THROW("GL_EXT_buffer_reference requires GLSL 450.");
require_extension_internal("GL_EXT_buffer_reference");
}
else if (ir.addressing_model != AddressingModelLogical)
{
SPIRV_CROSS_THROW("Only Logical and PhysicalStorageBuffer64EXT addressing models are supported.");
}
// Check for nonuniform qualifier and passthrough.
// Instead of looping over all decorations to find this, just look at capabilities.
for (auto &cap : ir.declared_capabilities)
{
switch (cap)
{
case CapabilityShaderNonUniformEXT:
if (!options.vulkan_semantics)
require_extension_internal("GL_NV_gpu_shader5");
else
require_extension_internal("GL_EXT_nonuniform_qualifier");
break;
case CapabilityRuntimeDescriptorArrayEXT:
if (!options.vulkan_semantics)
SPIRV_CROSS_THROW("GL_EXT_nonuniform_qualifier is only supported in Vulkan GLSL.");
require_extension_internal("GL_EXT_nonuniform_qualifier");
break;
case CapabilityGeometryShaderPassthroughNV:
if (execution.model == ExecutionModelGeometry)
{
require_extension_internal("GL_NV_geometry_shader_passthrough");
execution.geometry_passthrough = true;
}
break;
case CapabilityVariablePointers:
case CapabilityVariablePointersStorageBuffer:
SPIRV_CROSS_THROW("VariablePointers capability is not supported in GLSL.");
case CapabilityMultiView:
if (options.vulkan_semantics)
require_extension_internal("GL_EXT_multiview");
else
{
require_extension_internal("GL_OVR_multiview2");
if (options.ovr_multiview_view_count == 0)
SPIRV_CROSS_THROW("ovr_multiview_view_count must be non-zero when using GL_OVR_multiview2.");
if (get_execution_model() != ExecutionModelVertex)
SPIRV_CROSS_THROW("OVR_multiview2 can only be used with Vertex shaders.");
}
break;
case CapabilityRayQueryKHR:
if (options.es || options.version < 460 || !options.vulkan_semantics)
SPIRV_CROSS_THROW("RayQuery requires Vulkan GLSL 460.");
require_extension_internal("GL_EXT_ray_query");
ray_tracing_is_khr = true;
break;
case CapabilityRayTraversalPrimitiveCullingKHR:
if (options.es || options.version < 460 || !options.vulkan_semantics)
SPIRV_CROSS_THROW("RayQuery requires Vulkan GLSL 460.");
require_extension_internal("GL_EXT_ray_flags_primitive_culling");
ray_tracing_is_khr = true;
break;
default:
break;
}
}
if (options.ovr_multiview_view_count)
{
if (options.vulkan_semantics)
SPIRV_CROSS_THROW("OVR_multiview2 cannot be used with Vulkan semantics.");
if (get_execution_model() != ExecutionModelVertex)
SPIRV_CROSS_THROW("OVR_multiview2 can only be used with Vertex shaders.");
require_extension_internal("GL_OVR_multiview2");
}
}
void CompilerGLSL::ray_tracing_khr_fixup_locations()
{
uint32_t location = 0;
ir.for_each_typed_id<SPIRVariable>([&](uint32_t, SPIRVariable &var) {
// Incoming payload storage can also be used for tracing.
if (var.storage != StorageClassRayPayloadKHR && var.storage != StorageClassCallableDataKHR &&
var.storage != StorageClassIncomingRayPayloadKHR && var.storage != StorageClassIncomingCallableDataKHR)
return;
if (is_hidden_variable(var))
return;
set_decoration(var.self, DecorationLocation, location++);
});
}
string CompilerGLSL::compile()
{
ir.fixup_reserved_names();
if (!options.vulkan_semantics)
{
// only NV_gpu_shader5 supports divergent indexing on OpenGL, and it does so without extra qualifiers
backend.nonuniform_qualifier = "";
backend.needs_row_major_load_workaround = true;
}
backend.allow_precision_qualifiers = options.vulkan_semantics || options.es;
backend.force_gl_in_out_block = true;
backend.supports_extensions = true;
backend.use_array_constructor = true;
backend.workgroup_size_is_hidden = true;
backend.support_precise_qualifier = (!options.es && options.version >= 400) || (options.es && options.version >= 320);
if (is_legacy_es())
backend.support_case_fallthrough = false;
// Scan the SPIR-V to find trivial uses of extensions.
fixup_anonymous_struct_names();
fixup_type_alias();
reorder_type_alias();
build_function_control_flow_graphs_and_analyze();
find_static_extensions();
fixup_image_load_store_access();
update_active_builtins();
analyze_image_and_sampler_usage();
analyze_interlocked_resource_usage();
if (!inout_color_attachments.empty())
emit_inout_fragment_outputs_copy_to_subpass_inputs();
// Shaders might cast unrelated data to pointers of non-block types.
// Find all such instances and make sure we can cast the pointers to a synthesized block type.
if (ir.addressing_model == AddressingModelPhysicalStorageBuffer64EXT)
analyze_non_block_pointer_types();
uint32_t pass_count = 0;
do
{
reset(pass_count);
buffer.reset();
emit_header();
emit_resources();
emit_extension_workarounds(get_execution_model());
emit_function(get<SPIRFunction>(ir.default_entry_point), Bitset());
pass_count++;
} while (is_forcing_recompilation());
// Implement the interlocked wrapper function at the end.
// The body was implemented in lieu of main().
if (interlocked_is_complex)
{
statement("void main()");
begin_scope();
statement("// Interlocks were used in a way not compatible with GLSL, this is very slow.");
statement("SPIRV_Cross_beginInvocationInterlock();");
statement("spvMainInterlockedBody();");
statement("SPIRV_Cross_endInvocationInterlock();");
end_scope();
}
// Entry point in GLSL is always main().
get_entry_point().name = "main";
return buffer.str();
}
std::string CompilerGLSL::get_partial_source()
{
return buffer.str();
}
void CompilerGLSL::build_workgroup_size(SmallVector<string> &arguments, const SpecializationConstant &wg_x,
const SpecializationConstant &wg_y, const SpecializationConstant &wg_z)
{
auto &execution = get_entry_point();
bool builtin_workgroup = execution.workgroup_size.constant != 0;
bool use_local_size_id = !builtin_workgroup && execution.flags.get(ExecutionModeLocalSizeId);
if (wg_x.id)
{
if (options.vulkan_semantics)
arguments.push_back(join("local_size_x_id = ", wg_x.constant_id));
else
arguments.push_back(join("local_size_x = ", get<SPIRConstant>(wg_x.id).specialization_constant_macro_name));
}
else if (use_local_size_id && execution.workgroup_size.id_x)
arguments.push_back(join("local_size_x = ", get<SPIRConstant>(execution.workgroup_size.id_x).scalar()));
else
arguments.push_back(join("local_size_x = ", execution.workgroup_size.x));
if (wg_y.id)
{
if (options.vulkan_semantics)
arguments.push_back(join("local_size_y_id = ", wg_y.constant_id));
else
arguments.push_back(join("local_size_y = ", get<SPIRConstant>(wg_y.id).specialization_constant_macro_name));
}
else if (use_local_size_id && execution.workgroup_size.id_y)
arguments.push_back(join("local_size_y = ", get<SPIRConstant>(execution.workgroup_size.id_y).scalar()));
else
arguments.push_back(join("local_size_y = ", execution.workgroup_size.y));
if (wg_z.id)
{
if (options.vulkan_semantics)
arguments.push_back(join("local_size_z_id = ", wg_z.constant_id));
else
arguments.push_back(join("local_size_z = ", get<SPIRConstant>(wg_z.id).specialization_constant_macro_name));
}
else if (use_local_size_id && execution.workgroup_size.id_z)
arguments.push_back(join("local_size_z = ", get<SPIRConstant>(execution.workgroup_size.id_z).scalar()));
else
arguments.push_back(join("local_size_z = ", execution.workgroup_size.z));
}
void CompilerGLSL::request_subgroup_feature(ShaderSubgroupSupportHelper::Feature feature)
{
if (options.vulkan_semantics)
{
auto khr_extension = ShaderSubgroupSupportHelper::get_KHR_extension_for_feature(feature);
require_extension_internal(ShaderSubgroupSupportHelper::get_extension_name(khr_extension));
}
else
{
if (!shader_subgroup_supporter.is_feature_requested(feature))
force_recompile();
shader_subgroup_supporter.request_feature(feature);
}
}
void CompilerGLSL::emit_header()
{
auto &execution = get_entry_point();
statement("#version ", options.version, options.es && options.version > 100 ? " es" : "");
if (!options.es && options.version < 420)
{
// Needed for binding = # on UBOs, etc.
if (options.enable_420pack_extension)
{
statement("#ifdef GL_ARB_shading_language_420pack");
statement("#extension GL_ARB_shading_language_420pack : require");
statement("#endif");
}
// Needed for: layout(early_fragment_tests) in;
if (execution.flags.get(ExecutionModeEarlyFragmentTests))
require_extension_internal("GL_ARB_shader_image_load_store");
}
// Needed for: layout(post_depth_coverage) in;
if (execution.flags.get(ExecutionModePostDepthCoverage))
require_extension_internal("GL_ARB_post_depth_coverage");
// Needed for: layout({pixel,sample}_interlock_[un]ordered) in;
bool interlock_used = execution.flags.get(ExecutionModePixelInterlockOrderedEXT) ||
execution.flags.get(ExecutionModePixelInterlockUnorderedEXT) ||
execution.flags.get(ExecutionModeSampleInterlockOrderedEXT) ||
execution.flags.get(ExecutionModeSampleInterlockUnorderedEXT);
if (interlock_used)
{
if (options.es)
{
if (options.version < 310)
SPIRV_CROSS_THROW("At least ESSL 3.10 required for fragment shader interlock.");
require_extension_internal("GL_NV_fragment_shader_interlock");
}
else
{
if (options.version < 420)
require_extension_internal("GL_ARB_shader_image_load_store");
require_extension_internal("GL_ARB_fragment_shader_interlock");
}
}
for (auto &ext : forced_extensions)
{
if (ext == "GL_EXT_shader_explicit_arithmetic_types_float16")
{
// Special case, this extension has a potential fallback to another vendor extension in normal GLSL.
// GL_AMD_gpu_shader_half_float is a superset, so try that first.
statement("#if defined(GL_AMD_gpu_shader_half_float)");
statement("#extension GL_AMD_gpu_shader_half_float : require");
if (!options.vulkan_semantics)
{
statement("#elif defined(GL_NV_gpu_shader5)");
statement("#extension GL_NV_gpu_shader5 : require");
}
else
{
statement("#elif defined(GL_EXT_shader_explicit_arithmetic_types_float16)");
statement("#extension GL_EXT_shader_explicit_arithmetic_types_float16 : require");
}
statement("#else");
statement("#error No extension available for FP16.");
statement("#endif");
}
else if (ext == "GL_EXT_shader_explicit_arithmetic_types_int16")
{
if (options.vulkan_semantics)
statement("#extension GL_EXT_shader_explicit_arithmetic_types_int16 : require");
else
{
statement("#if defined(GL_AMD_gpu_shader_int16)");
statement("#extension GL_AMD_gpu_shader_int16 : require");
statement("#elif defined(GL_NV_gpu_shader5)");
statement("#extension GL_NV_gpu_shader5 : require");
statement("#else");
statement("#error No extension available for Int16.");
statement("#endif");
}
}
else if (ext == "GL_ARB_post_depth_coverage")
{
if (options.es)
statement("#extension GL_EXT_post_depth_coverage : require");
else
{
statement("#if defined(GL_ARB_post_depth_coverge)");
statement("#extension GL_ARB_post_depth_coverage : require");
statement("#else");
statement("#extension GL_EXT_post_depth_coverage : require");
statement("#endif");
}
}
else if (!options.vulkan_semantics && ext == "GL_ARB_shader_draw_parameters")
{
// Soft-enable this extension on plain GLSL.
statement("#ifdef ", ext);
statement("#extension ", ext, " : enable");
statement("#endif");
}
else if (ext == "GL_EXT_control_flow_attributes")
{
// These are just hints so we can conditionally enable and fallback in the shader.
statement("#if defined(GL_EXT_control_flow_attributes)");
statement("#extension GL_EXT_control_flow_attributes : require");
statement("#define SPIRV_CROSS_FLATTEN [[flatten]]");
statement("#define SPIRV_CROSS_BRANCH [[dont_flatten]]");
statement("#define SPIRV_CROSS_UNROLL [[unroll]]");
statement("#define SPIRV_CROSS_LOOP [[dont_unroll]]");
statement("#else");
statement("#define SPIRV_CROSS_FLATTEN");
statement("#define SPIRV_CROSS_BRANCH");
statement("#define SPIRV_CROSS_UNROLL");
statement("#define SPIRV_CROSS_LOOP");
statement("#endif");
}
else if (ext == "GL_NV_fragment_shader_interlock")
{
statement("#extension GL_NV_fragment_shader_interlock : require");
statement("#define SPIRV_Cross_beginInvocationInterlock() beginInvocationInterlockNV()");
statement("#define SPIRV_Cross_endInvocationInterlock() endInvocationInterlockNV()");
}
else if (ext == "GL_ARB_fragment_shader_interlock")
{
statement("#ifdef GL_ARB_fragment_shader_interlock");
statement("#extension GL_ARB_fragment_shader_interlock : enable");
statement("#define SPIRV_Cross_beginInvocationInterlock() beginInvocationInterlockARB()");
statement("#define SPIRV_Cross_endInvocationInterlock() endInvocationInterlockARB()");
statement("#elif defined(GL_INTEL_fragment_shader_ordering)");
statement("#extension GL_INTEL_fragment_shader_ordering : enable");
statement("#define SPIRV_Cross_beginInvocationInterlock() beginFragmentShaderOrderingINTEL()");
statement("#define SPIRV_Cross_endInvocationInterlock()");
statement("#endif");
}
else
statement("#extension ", ext, " : require");
}
if (!options.vulkan_semantics)
{
using Supp = ShaderSubgroupSupportHelper;
auto result = shader_subgroup_supporter.resolve();
for (uint32_t feature_index = 0; feature_index < Supp::FeatureCount; feature_index++)
{
auto feature = static_cast<Supp::Feature>(feature_index);
if (!shader_subgroup_supporter.is_feature_requested(feature))
continue;
auto exts = Supp::get_candidates_for_feature(feature, result);
if (exts.empty())
continue;
statement("");
for (auto &ext : exts)
{
const char *name = Supp::get_extension_name(ext);
const char *extra_predicate = Supp::get_extra_required_extension_predicate(ext);
auto extra_names = Supp::get_extra_required_extension_names(ext);
statement(&ext != &exts.front() ? "#elif" : "#if", " defined(", name, ")",
(*extra_predicate != '\0' ? " && " : ""), extra_predicate);
for (const auto &e : extra_names)
statement("#extension ", e, " : enable");
statement("#extension ", name, " : require");
}
if (!Supp::can_feature_be_implemented_without_extensions(feature))
{
statement("#else");
statement("#error No extensions available to emulate requested subgroup feature.");
}
statement("#endif");
}
}
for (auto &header : header_lines)
statement(header);
SmallVector<string> inputs;
SmallVector<string> outputs;
switch (execution.model)
{
case ExecutionModelVertex:
if (options.ovr_multiview_view_count)
inputs.push_back(join("num_views = ", options.ovr_multiview_view_count));
break;
case ExecutionModelGeometry:
if ((execution.flags.get(ExecutionModeInvocations)) && execution.invocations != 1)
inputs.push_back(join("invocations = ", execution.invocations));
if (execution.flags.get(ExecutionModeInputPoints))
inputs.push_back("points");
if (execution.flags.get(ExecutionModeInputLines))
inputs.push_back("lines");
if (execution.flags.get(ExecutionModeInputLinesAdjacency))
inputs.push_back("lines_adjacency");
if (execution.flags.get(ExecutionModeTriangles))
inputs.push_back("triangles");
if (execution.flags.get(ExecutionModeInputTrianglesAdjacency))
inputs.push_back("triangles_adjacency");
if (!execution.geometry_passthrough)
{
// For passthrough, these are implies and cannot be declared in shader.
outputs.push_back(join("max_vertices = ", execution.output_vertices));
if (execution.flags.get(ExecutionModeOutputTriangleStrip))
outputs.push_back("triangle_strip");
if (execution.flags.get(ExecutionModeOutputPoints))
outputs.push_back("points");
if (execution.flags.get(ExecutionModeOutputLineStrip))
outputs.push_back("line_strip");
}
break;
case ExecutionModelTessellationControl:
if (execution.flags.get(ExecutionModeOutputVertices))
outputs.push_back(join("vertices = ", execution.output_vertices));
break;
case ExecutionModelTessellationEvaluation:
if (execution.flags.get(ExecutionModeQuads))
inputs.push_back("quads");