-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathContextVk.cpp
6345 lines (5412 loc) · 254 KB
/
ContextVk.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 2016 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// ContextVk.cpp:
// Implements the class methods for ContextVk.
//
#include "libANGLE/renderer/vulkan/ContextVk.h"
#include "common/bitset_utils.h"
#include "common/debug.h"
#include "common/utilities.h"
#include "libANGLE/Context.h"
#include "libANGLE/Display.h"
#include "libANGLE/Program.h"
#include "libANGLE/Semaphore.h"
#include "libANGLE/Surface.h"
#include "libANGLE/angletypes.h"
#include "libANGLE/renderer/renderer_utils.h"
#include "libANGLE/renderer/vulkan/BufferVk.h"
#include "libANGLE/renderer/vulkan/CompilerVk.h"
#include "libANGLE/renderer/vulkan/DisplayVk.h"
#include "libANGLE/renderer/vulkan/FenceNVVk.h"
#include "libANGLE/renderer/vulkan/FramebufferVk.h"
#include "libANGLE/renderer/vulkan/MemoryObjectVk.h"
#include "libANGLE/renderer/vulkan/OverlayVk.h"
#include "libANGLE/renderer/vulkan/ProgramPipelineVk.h"
#include "libANGLE/renderer/vulkan/ProgramVk.h"
#include "libANGLE/renderer/vulkan/QueryVk.h"
#include "libANGLE/renderer/vulkan/RenderbufferVk.h"
#include "libANGLE/renderer/vulkan/RendererVk.h"
#include "libANGLE/renderer/vulkan/SamplerVk.h"
#include "libANGLE/renderer/vulkan/SemaphoreVk.h"
#include "libANGLE/renderer/vulkan/ShaderVk.h"
#include "libANGLE/renderer/vulkan/SurfaceVk.h"
#include "libANGLE/renderer/vulkan/SyncVk.h"
#include "libANGLE/renderer/vulkan/TextureVk.h"
#include "libANGLE/renderer/vulkan/TransformFeedbackVk.h"
#include "libANGLE/renderer/vulkan/VertexArrayVk.h"
#include "libANGLE/trace.h"
#include <iostream>
namespace rx
{
namespace
{
// For DesciptorSetUpdates
constexpr size_t kDescriptorBufferInfosInitialSize = 8;
constexpr size_t kDescriptorImageInfosInitialSize = 4;
constexpr size_t kDescriptorWriteInfosInitialSize =
kDescriptorBufferInfosInitialSize + kDescriptorImageInfosInitialSize;
// For shader uniforms such as gl_DepthRange and the viewport size.
struct GraphicsDriverUniforms
{
std::array<float, 4> viewport;
// 32 bits for 32 clip planes
uint32_t enabledClipPlanes;
uint32_t xfbActiveUnpaused;
int32_t xfbVerticesPerInstance;
// Used to replace gl_NumSamples. Because gl_NumSamples cannot be recognized in SPIR-V.
int32_t numSamples;
std::array<int32_t, 4> xfbBufferOffsets;
// .xy contain packed 8-bit values for atomic counter buffer offsets. These offsets are
// within Vulkan's minStorageBufferOffsetAlignment limit and are used to support unaligned
// offsets allowed in GL.
//
// .zw are unused.
std::array<uint32_t, 4> acbBufferOffsets;
// We'll use x, y, z for near / far / diff respectively.
std::array<float, 4> depthRange;
};
static_assert(sizeof(GraphicsDriverUniforms) % (sizeof(uint32_t) * 4) == 0,
"GraphicsDriverUniforms should 16bytes aligned");
// TODO: http://issuetracker.google.com/173636783 Once the bug is fixed, we should remove this.
struct GraphicsDriverUniformsExtended
{
GraphicsDriverUniforms common;
// Used to flip gl_FragCoord (both .xy for Android pre-rotation; only .y for desktop)
std::array<float, 2> halfRenderArea;
std::array<float, 2> flipXY;
std::array<float, 2> negFlipXY;
std::array<int32_t, 2> padding;
// Used to pre-rotate gl_FragCoord for swapchain images on Android (a mat2, which is padded to
// the size of two vec4's).
std::array<float, 8> fragRotation;
};
struct ComputeDriverUniforms
{
// Atomic counter buffer offsets with the same layout as in GraphicsDriverUniforms.
std::array<uint32_t, 4> acbBufferOffsets;
};
GLenum DefaultGLErrorCode(VkResult result)
{
switch (result)
{
case VK_ERROR_OUT_OF_HOST_MEMORY:
case VK_ERROR_OUT_OF_DEVICE_MEMORY:
case VK_ERROR_TOO_MANY_OBJECTS:
return GL_OUT_OF_MEMORY;
default:
return GL_INVALID_OPERATION;
}
}
constexpr gl::ShaderMap<vk::ImageLayout> kShaderReadOnlyImageLayouts = {
{gl::ShaderType::Vertex, vk::ImageLayout::VertexShaderReadOnly},
{gl::ShaderType::TessControl, vk::ImageLayout::PreFragmentShadersReadOnly},
{gl::ShaderType::TessEvaluation, vk::ImageLayout::PreFragmentShadersReadOnly},
{gl::ShaderType::Geometry, vk::ImageLayout::PreFragmentShadersReadOnly},
{gl::ShaderType::Fragment, vk::ImageLayout::FragmentShaderReadOnly},
{gl::ShaderType::Compute, vk::ImageLayout::ComputeShaderReadOnly}};
constexpr gl::ShaderMap<vk::ImageLayout> kShaderWriteImageLayouts = {
{gl::ShaderType::Vertex, vk::ImageLayout::VertexShaderWrite},
{gl::ShaderType::TessControl, vk::ImageLayout::PreFragmentShadersWrite},
{gl::ShaderType::TessEvaluation, vk::ImageLayout::PreFragmentShadersWrite},
{gl::ShaderType::Geometry, vk::ImageLayout::PreFragmentShadersWrite},
{gl::ShaderType::Fragment, vk::ImageLayout::FragmentShaderWrite},
{gl::ShaderType::Compute, vk::ImageLayout::ComputeShaderWrite}};
constexpr VkBufferUsageFlags kVertexBufferUsage = VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
constexpr size_t kDefaultValueSize = sizeof(gl::VertexAttribCurrentValueData::Values);
constexpr size_t kDefaultBufferSize = kDefaultValueSize * 16;
constexpr size_t kDriverUniformsAllocatorPageSize = 4 * 1024;
uint32_t GetCoverageSampleCount(const gl::State &glState, FramebufferVk *drawFramebuffer)
{
if (!glState.isSampleCoverageEnabled())
{
return 0;
}
// Get a fraction of the samples based on the coverage parameters.
// There are multiple ways to obtain an integer value from a float -
// truncation, ceil and round
//
// round() provides a more even distribution of values but doesn't seem to play well
// with all vendors (AMD). A way to work around this is to increase the comparison threshold
// of deqp tests. Though this takes care of deqp tests other apps would still have issues.
//
// Truncation provides an uneven distribution near the edges of the interval but seems to
// play well with all vendors.
//
// We are going with truncation for expediency.
return static_cast<uint32_t>(glState.getSampleCoverageValue() * drawFramebuffer->getSamples());
}
void ApplySampleCoverage(const gl::State &glState,
uint32_t coverageSampleCount,
uint32_t maskNumber,
uint32_t *maskOut)
{
if (!glState.isSampleCoverageEnabled())
{
return;
}
uint32_t maskBitOffset = maskNumber * 32;
uint32_t coverageMask = coverageSampleCount >= (maskBitOffset + 32)
? std::numeric_limits<uint32_t>::max()
: (1u << (coverageSampleCount - maskBitOffset)) - 1;
if (glState.getSampleCoverageInvert())
{
coverageMask = ~coverageMask;
}
*maskOut &= coverageMask;
}
bool IsRenderPassStartedAndUsesImage(const vk::CommandBufferHelper &renderPassCommands,
const vk::ImageHelper &image)
{
return renderPassCommands.started() && renderPassCommands.usesImageInRenderPass(image);
}
// When an Android surface is rotated differently than the device's native orientation, ANGLE must
// rotate gl_Position in the last pre-rasterization shader and gl_FragCoord in the fragment shader.
// Rotation of gl_Position is done in SPIR-V. The following are the rotation matrices for the
// fragment shader.
//
// Note: these are mat2's that are appropriately padded (4 floats per row).
using PreRotationMatrixValues = std::array<float, 8>;
constexpr angle::PackedEnumMap<rx::SurfaceRotation,
PreRotationMatrixValues,
angle::EnumSize<rx::SurfaceRotation>()>
kFragRotationMatrices = {
{{rx::SurfaceRotation::Identity, {{1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::Rotated90Degrees,
{{0.0f, 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::Rotated180Degrees,
{{1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::Rotated270Degrees,
{{0.0f, 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::FlippedIdentity, {{1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::FlippedRotated90Degrees,
{{0.0f, 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::FlippedRotated180Degrees,
{{1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f}}},
{rx::SurfaceRotation::FlippedRotated270Degrees,
{{0.0f, 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f}}}}};
bool IsRotatedAspectRatio(SurfaceRotation rotation)
{
return ((rotation == SurfaceRotation::Rotated90Degrees) ||
(rotation == SurfaceRotation::Rotated270Degrees) ||
(rotation == SurfaceRotation::FlippedRotated90Degrees) ||
(rotation == SurfaceRotation::FlippedRotated270Degrees));
}
SurfaceRotation DetermineSurfaceRotation(gl::Framebuffer *framebuffer,
WindowSurfaceVk *windowSurface)
{
if (windowSurface && framebuffer->isDefault())
{
switch (windowSurface->getPreTransform())
{
case VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR:
// Do not rotate gl_Position (surface matches the device's orientation):
return SurfaceRotation::Identity;
case VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR:
// Rotate gl_Position 90 degrees:
return SurfaceRotation::Rotated90Degrees;
case VK_SURFACE_TRANSFORM_ROTATE_180_BIT_KHR:
// Rotate gl_Position 180 degrees:
return SurfaceRotation::Rotated180Degrees;
case VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR:
// Rotate gl_Position 270 degrees:
return SurfaceRotation::Rotated270Degrees;
default:
UNREACHABLE();
return SurfaceRotation::Identity;
}
}
else
{
// Do not rotate gl_Position (offscreen framebuffer):
return SurfaceRotation::Identity;
}
}
// Should not generate a copy with modern C++.
EventName GetTraceEventName(const char *title, uint32_t counter)
{
EventName buf;
snprintf(buf.data(), kMaxGpuEventNameLen - 1, "%s %u", title, counter);
return buf;
}
vk::ResourceAccess GetDepthAccess(const gl::DepthStencilState &dsState)
{
if (!dsState.depthTest)
{
return vk::ResourceAccess::Unused;
}
return dsState.isDepthMaskedOut() ? vk::ResourceAccess::ReadOnly : vk::ResourceAccess::Write;
}
vk::ResourceAccess GetStencilAccess(const gl::DepthStencilState &dsState)
{
if (!dsState.stencilTest)
{
return vk::ResourceAccess::Unused;
}
return dsState.isStencilNoOp() && dsState.isStencilBackNoOp() ? vk::ResourceAccess::ReadOnly
: vk::ResourceAccess::Write;
}
egl::ContextPriority GetContextPriority(const gl::State &state)
{
return egl::FromEGLenum<egl::ContextPriority>(state.getContextPriority());
}
template <typename MaskT>
void AppendBufferVectorToDesc(vk::ShaderBuffersDescriptorDesc *desc,
const gl::BufferVector &buffers,
const MaskT &buffersMask,
bool appendOffset)
{
if (buffersMask.any())
{
typename MaskT::param_type lastBufferIndex = buffersMask.last();
for (typename MaskT::param_type bufferIndex = 0; bufferIndex <= lastBufferIndex;
++bufferIndex)
{
const gl::OffsetBindingPointer<gl::Buffer> &binding = buffers[bufferIndex];
const gl::Buffer *bufferGL = binding.get();
if (!bufferGL)
{
desc->append32BitValue(0);
continue;
}
BufferVk *bufferVk = vk::GetImpl(bufferGL);
if (!bufferVk->isBufferValid())
{
desc->append32BitValue(0);
continue;
}
VkDeviceSize bufferOffset = 0;
vk::BufferSerial bufferSerial =
bufferVk->getBufferAndOffset(&bufferOffset).getBufferSerial();
desc->appendBufferSerial(bufferSerial);
ASSERT(static_cast<uint64_t>(binding.getSize()) <=
static_cast<uint64_t>(std::numeric_limits<uint32_t>::max()));
desc->append32BitValue(static_cast<uint32_t>(binding.getSize()));
if (appendOffset)
{
ASSERT(static_cast<uint64_t>(binding.getOffset()) <
static_cast<uint64_t>(std::numeric_limits<uint32_t>::max()));
desc->append32BitValue(static_cast<uint32_t>(bufferOffset + binding.getOffset()));
}
}
}
desc->append32BitValue(std::numeric_limits<uint32_t>::max());
}
} // anonymous namespace
// Not necessary once upgraded to C++17.
constexpr ContextVk::DirtyBits ContextVk::kIndexAndVertexDirtyBits;
constexpr ContextVk::DirtyBits ContextVk::kPipelineDescAndBindingDirtyBits;
constexpr ContextVk::DirtyBits ContextVk::kTexturesAndDescSetDirtyBits;
constexpr ContextVk::DirtyBits ContextVk::kResourcesAndDescSetDirtyBits;
constexpr ContextVk::DirtyBits ContextVk::kXfbBuffersAndDescSetDirtyBits;
constexpr ContextVk::DirtyBits ContextVk::kDriverUniformsAndBindingDirtyBits;
ANGLE_INLINE void ContextVk::flushDescriptorSetUpdates()
{
if (mWriteDescriptorSets.empty())
{
ASSERT(mDescriptorBufferInfos.empty());
ASSERT(mDescriptorImageInfos.empty());
return;
}
vkUpdateDescriptorSets(getDevice(), static_cast<uint32_t>(mWriteDescriptorSets.size()),
mWriteDescriptorSets.data(), 0, nullptr);
mWriteDescriptorSets.clear();
mDescriptorBufferInfos.clear();
mDescriptorImageInfos.clear();
}
ANGLE_INLINE void ContextVk::onRenderPassFinished()
{
pauseRenderPassQueriesIfActive();
mRenderPassCommandBuffer = nullptr;
mGraphicsDirtyBits.set(DIRTY_BIT_RENDER_PASS);
}
// ContextVk::ScopedDescriptorSetUpdates implementation.
class ContextVk::ScopedDescriptorSetUpdates final : angle::NonCopyable
{
public:
ANGLE_INLINE ScopedDescriptorSetUpdates(ContextVk *contextVk) : mContextVk(contextVk) {}
ANGLE_INLINE ~ScopedDescriptorSetUpdates() { mContextVk->flushDescriptorSetUpdates(); }
private:
ContextVk *mContextVk;
};
ContextVk::DriverUniformsDescriptorSet::DriverUniformsDescriptorSet()
: descriptorSet(VK_NULL_HANDLE), dynamicOffset(0)
{}
ContextVk::DriverUniformsDescriptorSet::~DriverUniformsDescriptorSet() = default;
void ContextVk::DriverUniformsDescriptorSet::init(RendererVk *rendererVk)
{
size_t minAlignment = static_cast<size_t>(
rendererVk->getPhysicalDeviceProperties().limits.minUniformBufferOffsetAlignment);
dynamicBuffer.init(rendererVk, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, minAlignment,
kDriverUniformsAllocatorPageSize, true,
vk::DynamicBufferPolicy::FrequentSmallAllocations);
descriptorSetCache.clear();
}
void ContextVk::DriverUniformsDescriptorSet::destroy(RendererVk *renderer)
{
descriptorSetLayout.reset();
descriptorPoolBinding.reset();
dynamicBuffer.destroy(renderer);
descriptorSetCache.clear();
descriptorSetCache.destroy(renderer);
}
// ContextVk implementation.
ContextVk::ContextVk(const gl::State &state, gl::ErrorSet *errorSet, RendererVk *renderer)
: ContextImpl(state, errorSet),
vk::Context(renderer),
mGraphicsDirtyBitHandlers{},
mComputeDirtyBitHandlers{},
mRenderPassCommandBuffer(nullptr),
mCurrentGraphicsPipeline(nullptr),
mCurrentComputePipeline(nullptr),
mCurrentDrawMode(gl::PrimitiveMode::InvalidEnum),
mCurrentWindowSurface(nullptr),
mCurrentRotationDrawFramebuffer(SurfaceRotation::Identity),
mCurrentRotationReadFramebuffer(SurfaceRotation::Identity),
mActiveRenderPassQueries{},
mVertexArray(nullptr),
mDrawFramebuffer(nullptr),
mProgram(nullptr),
mExecutable(nullptr),
mLastIndexBufferOffset(nullptr),
mCurrentIndexBufferOffset(0),
mCurrentDrawElementsType(gl::DrawElementsType::InvalidEnum),
mXfbBaseVertex(0),
mXfbVertexCountPerInstance(0),
mClearColorValue{},
mClearDepthStencilValue{},
mClearColorMasks(0),
mFlipYForCurrentSurface(false),
mFlipViewportForDrawFramebuffer(false),
mFlipViewportForReadFramebuffer(false),
mIsAnyHostVisibleBufferWritten(false),
mEmulateSeamfulCubeMapSampling(false),
mOutsideRenderPassCommands(nullptr),
mRenderPassCommands(nullptr),
mQueryEventType(GraphicsEventCmdBuf::NotInQueryCmd),
mGpuEventsEnabled(false),
mEGLSyncObjectPendingFlush(false),
mHasDeferredFlush(false),
mLastProgramUsesFramebufferFetch(false),
mGpuClockSync{std::numeric_limits<double>::max(), std::numeric_limits<double>::max()},
mGpuEventTimestampOrigin(0),
mPerfCounters{},
mContextPerfCounters{},
mCumulativeContextPerfCounters{},
mContextPriority(renderer->getDriverPriority(GetContextPriority(state))),
mShareGroupVk(vk::GetImpl(state.getShareGroup()))
{
ANGLE_TRACE_EVENT0("gpu.angle", "ContextVk::ContextVk");
memset(&mClearColorValue, 0, sizeof(mClearColorValue));
memset(&mClearDepthStencilValue, 0, sizeof(mClearDepthStencilValue));
memset(&mViewport, 0, sizeof(mViewport));
memset(&mScissor, 0, sizeof(mScissor));
// Ensure viewport is within Vulkan requirements
vk::ClampViewport(&mViewport);
mNonIndexedDirtyBitsMask.set();
mNonIndexedDirtyBitsMask.reset(DIRTY_BIT_INDEX_BUFFER);
mIndexedDirtyBitsMask.set();
// Once a command buffer is ended, all bindings (through |vkCmdBind*| calls) are lost per Vulkan
// spec. Once a new command buffer is allocated, we must make sure every previously bound
// resource is bound again.
//
// Note that currently these dirty bits are set every time a new render pass command buffer is
// begun. However, using ANGLE's SecondaryCommandBuffer, the Vulkan command buffer (which is
// the primary command buffer) is not ended, so technically we don't need to rebind these.
mNewGraphicsCommandBufferDirtyBits =
DirtyBits{DIRTY_BIT_RENDER_PASS, DIRTY_BIT_PIPELINE_BINDING,
DIRTY_BIT_TEXTURES, DIRTY_BIT_VERTEX_BUFFERS,
DIRTY_BIT_INDEX_BUFFER, DIRTY_BIT_SHADER_RESOURCES,
DIRTY_BIT_DESCRIPTOR_SETS, DIRTY_BIT_DRIVER_UNIFORMS_BINDING,
DIRTY_BIT_VIEWPORT, DIRTY_BIT_SCISSOR};
if (getFeatures().supportsTransformFeedbackExtension.enabled)
{
mNewGraphicsCommandBufferDirtyBits.set(DIRTY_BIT_TRANSFORM_FEEDBACK_BUFFERS);
}
mNewComputeCommandBufferDirtyBits =
DirtyBits{DIRTY_BIT_PIPELINE_BINDING, DIRTY_BIT_TEXTURES, DIRTY_BIT_SHADER_RESOURCES,
DIRTY_BIT_DESCRIPTOR_SETS, DIRTY_BIT_DRIVER_UNIFORMS_BINDING};
mGraphicsDirtyBitHandlers[DIRTY_BIT_MEMORY_BARRIER] =
&ContextVk::handleDirtyGraphicsMemoryBarrier;
mGraphicsDirtyBitHandlers[DIRTY_BIT_EVENT_LOG] = &ContextVk::handleDirtyGraphicsEventLog;
mGraphicsDirtyBitHandlers[DIRTY_BIT_DEFAULT_ATTRIBS] =
&ContextVk::handleDirtyGraphicsDefaultAttribs;
mGraphicsDirtyBitHandlers[DIRTY_BIT_PIPELINE_DESC] =
&ContextVk::handleDirtyGraphicsPipelineDesc;
mGraphicsDirtyBitHandlers[DIRTY_BIT_RENDER_PASS] = &ContextVk::handleDirtyGraphicsRenderPass;
mGraphicsDirtyBitHandlers[DIRTY_BIT_PIPELINE_BINDING] =
&ContextVk::handleDirtyGraphicsPipelineBinding;
mGraphicsDirtyBitHandlers[DIRTY_BIT_TEXTURES] = &ContextVk::handleDirtyGraphicsTextures;
mGraphicsDirtyBitHandlers[DIRTY_BIT_VERTEX_BUFFERS] =
&ContextVk::handleDirtyGraphicsVertexBuffers;
mGraphicsDirtyBitHandlers[DIRTY_BIT_INDEX_BUFFER] = &ContextVk::handleDirtyGraphicsIndexBuffer;
mGraphicsDirtyBitHandlers[DIRTY_BIT_DRIVER_UNIFORMS] =
&ContextVk::handleDirtyGraphicsDriverUniforms;
mGraphicsDirtyBitHandlers[DIRTY_BIT_DRIVER_UNIFORMS_BINDING] =
&ContextVk::handleDirtyGraphicsDriverUniformsBinding;
mGraphicsDirtyBitHandlers[DIRTY_BIT_SHADER_RESOURCES] =
&ContextVk::handleDirtyGraphicsShaderResources;
mGraphicsDirtyBitHandlers[DIRTY_BIT_FRAMEBUFFER_FETCH_BARRIER] =
&ContextVk::handleDirtyGraphicsFramebufferFetchBarrier;
if (getFeatures().supportsTransformFeedbackExtension.enabled)
{
mGraphicsDirtyBitHandlers[DIRTY_BIT_TRANSFORM_FEEDBACK_BUFFERS] =
&ContextVk::handleDirtyGraphicsTransformFeedbackBuffersExtension;
mGraphicsDirtyBitHandlers[DIRTY_BIT_TRANSFORM_FEEDBACK_RESUME] =
&ContextVk::handleDirtyGraphicsTransformFeedbackResume;
}
else if (getFeatures().emulateTransformFeedback.enabled)
{
mGraphicsDirtyBitHandlers[DIRTY_BIT_TRANSFORM_FEEDBACK_BUFFERS] =
&ContextVk::handleDirtyGraphicsTransformFeedbackBuffersEmulation;
}
mGraphicsDirtyBitHandlers[DIRTY_BIT_DESCRIPTOR_SETS] =
&ContextVk::handleDirtyGraphicsDescriptorSets;
mGraphicsDirtyBitHandlers[DIRTY_BIT_VIEWPORT] = &ContextVk::handleDirtyGraphicsViewport;
mGraphicsDirtyBitHandlers[DIRTY_BIT_SCISSOR] = &ContextVk::handleDirtyGraphicsScissor;
mComputeDirtyBitHandlers[DIRTY_BIT_MEMORY_BARRIER] =
&ContextVk::handleDirtyComputeMemoryBarrier;
mComputeDirtyBitHandlers[DIRTY_BIT_EVENT_LOG] = &ContextVk::handleDirtyComputeEventLog;
mComputeDirtyBitHandlers[DIRTY_BIT_PIPELINE_DESC] = &ContextVk::handleDirtyComputePipelineDesc;
mComputeDirtyBitHandlers[DIRTY_BIT_PIPELINE_BINDING] =
&ContextVk::handleDirtyComputePipelineBinding;
mComputeDirtyBitHandlers[DIRTY_BIT_TEXTURES] = &ContextVk::handleDirtyComputeTextures;
mComputeDirtyBitHandlers[DIRTY_BIT_DRIVER_UNIFORMS] =
&ContextVk::handleDirtyComputeDriverUniforms;
mComputeDirtyBitHandlers[DIRTY_BIT_DRIVER_UNIFORMS_BINDING] =
&ContextVk::handleDirtyComputeDriverUniformsBinding;
mComputeDirtyBitHandlers[DIRTY_BIT_SHADER_RESOURCES] =
&ContextVk::handleDirtyComputeShaderResources;
mComputeDirtyBitHandlers[DIRTY_BIT_DESCRIPTOR_SETS] =
&ContextVk::handleDirtyComputeDescriptorSets;
mGraphicsDirtyBits = mNewGraphicsCommandBufferDirtyBits;
mComputeDirtyBits = mNewComputeCommandBufferDirtyBits;
mActiveTextures.fill({nullptr, nullptr, true});
mActiveImages.fill(nullptr);
// The following dirty bits don't affect the program pipeline:
//
// - READ_FRAMEBUFFER_BINDING only affects operations that read from said framebuffer,
// - CLEAR_* only affect following clear calls,
// - PACK/UNPACK_STATE only affect texture data upload/download,
// - *_BINDING only affect descriptor sets.
//
mPipelineDirtyBitsMask.set();
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_READ_FRAMEBUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_CLEAR_COLOR);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_CLEAR_DEPTH);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_CLEAR_STENCIL);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_UNPACK_STATE);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_UNPACK_BUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_PACK_STATE);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_PACK_BUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_RENDERBUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_DRAW_INDIRECT_BUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_DISPATCH_INDIRECT_BUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_SAMPLER_BINDINGS);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_TEXTURE_BINDINGS);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_IMAGE_BINDINGS);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_TRANSFORM_FEEDBACK_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_UNIFORM_BUFFER_BINDINGS);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_SHADER_STORAGE_BUFFER_BINDING);
mPipelineDirtyBitsMask.reset(gl::State::DIRTY_BIT_ATOMIC_COUNTER_BUFFER_BINDING);
// Reserve reasonable amount of spaces so that for majority of apps we don't need to grow at all
mDescriptorBufferInfos.reserve(kDescriptorBufferInfosInitialSize);
mDescriptorImageInfos.reserve(kDescriptorImageInfosInitialSize);
mWriteDescriptorSets.reserve(kDescriptorWriteInfosInitialSize);
}
ContextVk::~ContextVk() = default;
void ContextVk::onDestroy(const gl::Context *context)
{
outputCumulativePerfCounters();
// Remove context from the share group
mShareGroupVk->getContexts()->erase(this);
// This will not destroy any resources. It will release them to be collected after finish.
mIncompleteTextures.onDestroy(context);
// Flush and complete current outstanding work before destruction.
(void)finishImpl();
VkDevice device = getDevice();
for (DriverUniformsDescriptorSet &driverUniforms : mDriverUniforms)
{
driverUniforms.destroy(mRenderer);
}
for (vk::DynamicDescriptorPool &dynamicDescriptorPool : mDriverUniformsDescriptorPools)
{
dynamicDescriptorPool.destroy(device);
}
mDefaultUniformStorage.release(mRenderer);
mEmptyBuffer.release(mRenderer);
mStagingBuffer.release(mRenderer);
for (vk::DynamicBuffer &defaultBuffer : mDefaultAttribBuffers)
{
defaultBuffer.destroy(mRenderer);
}
for (vk::DynamicQueryPool &queryPool : mQueryPools)
{
queryPool.destroy(device);
}
// Recycle current commands buffers.
mRenderer->recycleCommandBufferHelper(mOutsideRenderPassCommands);
mRenderer->recycleCommandBufferHelper(mRenderPassCommands);
mOutsideRenderPassCommands = nullptr;
mRenderPassCommands = nullptr;
mRenderer->releaseSharedResources(&mResourceUseList);
mUtils.destroy(mRenderer);
mRenderPassCache.destroy(mRenderer);
mShaderLibrary.destroy(device);
mGpuEventQueryPool.destroy(device);
mCommandPool.destroy(device);
ASSERT(mCurrentGarbage.empty());
ASSERT(mResourceUseList.empty());
}
angle::Result ContextVk::getIncompleteTexture(const gl::Context *context,
gl::TextureType type,
gl::SamplerFormat format,
gl::Texture **textureOut)
{
return mIncompleteTextures.getIncompleteTexture(context, type, format, this, textureOut);
}
angle::Result ContextVk::initialize()
{
ANGLE_TRACE_EVENT0("gpu.angle", "ContextVk::initialize");
ANGLE_TRY(mQueryPools[gl::QueryType::AnySamples].init(this, VK_QUERY_TYPE_OCCLUSION,
vk::kDefaultOcclusionQueryPoolSize));
ANGLE_TRY(mQueryPools[gl::QueryType::AnySamplesConservative].init(
this, VK_QUERY_TYPE_OCCLUSION, vk::kDefaultOcclusionQueryPoolSize));
// Only initialize the timestamp query pools if the extension is available.
if (mRenderer->getQueueFamilyProperties().timestampValidBits > 0)
{
ANGLE_TRY(mQueryPools[gl::QueryType::Timestamp].init(this, VK_QUERY_TYPE_TIMESTAMP,
vk::kDefaultTimestampQueryPoolSize));
ANGLE_TRY(mQueryPools[gl::QueryType::TimeElapsed].init(this, VK_QUERY_TYPE_TIMESTAMP,
vk::kDefaultTimestampQueryPoolSize));
}
if (getFeatures().supportsTransformFeedbackExtension.enabled)
{
ANGLE_TRY(mQueryPools[gl::QueryType::TransformFeedbackPrimitivesWritten].init(
this, VK_QUERY_TYPE_TRANSFORM_FEEDBACK_STREAM_EXT,
vk::kDefaultTransformFeedbackQueryPoolSize));
}
// The primitives generated query is provided through the Vulkan pipeline statistics query if
// supported. TODO: If VK_EXT_primitives_generated_query is supported, use that instead.
// http://anglebug.com/5430
if (getFeatures().supportsPipelineStatisticsQuery.enabled)
{
ANGLE_TRY(mQueryPools[gl::QueryType::PrimitivesGenerated].init(
this, VK_QUERY_TYPE_PIPELINE_STATISTICS, vk::kDefaultPrimitivesGeneratedQueryPoolSize));
}
// Init GLES to Vulkan index type map.
initIndexTypeMap();
// Init driver uniforms and get the descriptor set layouts.
constexpr angle::PackedEnumMap<PipelineType, VkShaderStageFlags> kPipelineStages = {
{PipelineType::Graphics, VK_SHADER_STAGE_ALL_GRAPHICS},
{PipelineType::Compute, VK_SHADER_STAGE_COMPUTE_BIT},
};
for (PipelineType pipeline : angle::AllEnums<PipelineType>())
{
mDriverUniforms[pipeline].init(mRenderer);
vk::DescriptorSetLayoutDesc desc =
getDriverUniformsDescriptorSetDesc(kPipelineStages[pipeline]);
ANGLE_TRY(getDescriptorSetLayoutCache().getDescriptorSetLayout(
this, desc, &mDriverUniforms[pipeline].descriptorSetLayout));
vk::DescriptorSetLayoutBindingVector bindingVector;
std::vector<VkSampler> immutableSamplers;
desc.unpackBindings(&bindingVector, &immutableSamplers);
std::vector<VkDescriptorPoolSize> descriptorPoolSizes;
for (const VkDescriptorSetLayoutBinding &binding : bindingVector)
{
if (binding.descriptorCount > 0)
{
VkDescriptorPoolSize poolSize = {};
poolSize.type = binding.descriptorType;
poolSize.descriptorCount = binding.descriptorCount;
descriptorPoolSizes.emplace_back(poolSize);
}
}
if (!descriptorPoolSizes.empty())
{
ANGLE_TRY(mDriverUniformsDescriptorPools[pipeline].init(
this, descriptorPoolSizes.data(), descriptorPoolSizes.size(),
mDriverUniforms[pipeline].descriptorSetLayout.get().getHandle()));
}
}
mGraphicsPipelineDesc.reset(new vk::GraphicsPipelineDesc());
mGraphicsPipelineDesc->initDefaults(this);
// Initialize current value/default attribute buffers.
for (vk::DynamicBuffer &buffer : mDefaultAttribBuffers)
{
buffer.init(mRenderer, kVertexBufferUsage, 1, kDefaultBufferSize, true,
vk::DynamicBufferPolicy::FrequentSmallAllocations);
}
#if ANGLE_ENABLE_VULKAN_GPU_TRACE_EVENTS
angle::PlatformMethods *platform = ANGLEPlatformCurrent();
ASSERT(platform);
// GPU tracing workaround for anglebug.com/2927. The renderer should not emit gpu events
// during platform discovery.
const unsigned char *gpuEventsEnabled =
platform->getTraceCategoryEnabledFlag(platform, "gpu.angle.gpu");
mGpuEventsEnabled = gpuEventsEnabled && *gpuEventsEnabled;
#endif
mEmulateSeamfulCubeMapSampling = shouldEmulateSeamfulCubeMapSampling();
// Assign initial command buffers from queue
mOutsideRenderPassCommands = mRenderer->getCommandBufferHelper(false);
mRenderPassCommands = mRenderer->getCommandBufferHelper(true);
if (mGpuEventsEnabled)
{
// GPU events should only be available if timestamp queries are available.
ASSERT(mRenderer->getQueueFamilyProperties().timestampValidBits > 0);
// Calculate the difference between CPU and GPU clocks for GPU event reporting.
ANGLE_TRY(mGpuEventQueryPool.init(this, VK_QUERY_TYPE_TIMESTAMP,
vk::kDefaultTimestampQueryPoolSize));
ANGLE_TRY(synchronizeCpuGpuTime());
mPerfCounters.primaryBuffers++;
EventName eventName = GetTraceEventName("Primary", mPerfCounters.primaryBuffers);
ANGLE_TRY(traceGpuEvent(&mOutsideRenderPassCommands->getCommandBuffer(),
TRACE_EVENT_PHASE_BEGIN, eventName));
}
size_t minAlignment = static_cast<size_t>(
mRenderer->getPhysicalDeviceProperties().limits.minUniformBufferOffsetAlignment);
mDefaultUniformStorage.init(mRenderer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, minAlignment,
mRenderer->getDefaultUniformBufferSize(), true,
vk::DynamicBufferPolicy::FrequentSmallAllocations);
// Initialize an "empty" buffer for use with default uniform blocks where there are no uniforms,
// or atomic counter buffer array indices that are unused.
constexpr VkBufferUsageFlags kEmptyBufferUsage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
VkBufferCreateInfo emptyBufferInfo = {};
emptyBufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
emptyBufferInfo.flags = 0;
emptyBufferInfo.size = 16;
emptyBufferInfo.usage = kEmptyBufferUsage;
emptyBufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
emptyBufferInfo.queueFamilyIndexCount = 0;
emptyBufferInfo.pQueueFamilyIndices = nullptr;
constexpr VkMemoryPropertyFlags kMemoryType = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
ANGLE_TRY(mEmptyBuffer.init(this, emptyBufferInfo, kMemoryType));
constexpr VkImageUsageFlags kStagingBufferUsageFlags =
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
size_t stagingBufferAlignment =
static_cast<size_t>(mRenderer->getPhysicalDeviceProperties().limits.minMemoryMapAlignment);
constexpr size_t kStagingBufferSize = 1024u * 1024u; // 1M
mStagingBuffer.init(mRenderer, kStagingBufferUsageFlags, stagingBufferAlignment,
kStagingBufferSize, true, vk::DynamicBufferPolicy::SporadicTextureUpload);
// Add context into the share group
mShareGroupVk->getContexts()->insert(this);
return angle::Result::Continue;
}
angle::Result ContextVk::flush(const gl::Context *context)
{
// If a sync object has been used or this is a shared context, then we need to flush the
// commands and end the render pass to make sure the sync object (and any preceding commands)
// lands in the correct place within the command stream.
// EGL sync objects can span across context share groups, so don't defer flushes if there's one
// pending a flush.
if (getShareGroupVk()->isSyncObjectPendingFlush() && context->isShared() &&
!mEGLSyncObjectPendingFlush)
{
// Flush the commands to create a sync point in the command stream.
ANGLE_TRY(flushCommandsAndEndRenderPass());
// Move the resources to the share group, so they are released during the next vkQueueSubmit
// performed by any context in the share group. Note that this relies heavily on the global
// mutex to guarantee that no two contexts are modifying the lists at the same time.
getShareGroupVk()->acquireResourceUseList(std::move(mResourceUseList));
mHasDeferredFlush = true;
return angle::Result::Continue;
}
// EGL sync objects can span across context share groups, so don't defer flushes if there's one
// pending a flush.
if (!mEGLSyncObjectPendingFlush &&
mRenderer->getFeatures().deferFlushUntilEndRenderPass.enabled && hasStartedRenderPass())
{
mHasDeferredFlush = true;
return angle::Result::Continue;
}
return flushImpl(nullptr);
}
angle::Result ContextVk::finish(const gl::Context *context)
{
return finishImpl();
}
angle::Result ContextVk::setupDraw(const gl::Context *context,
gl::PrimitiveMode mode,
GLint firstVertexOrInvalid,
GLsizei vertexOrIndexCount,
GLsizei instanceCount,
gl::DrawElementsType indexTypeOrInvalid,
const void *indices,
DirtyBits dirtyBitMask)
{
// Set any dirty bits that depend on draw call parameters or other objects.
if (mode != mCurrentDrawMode)
{
invalidateCurrentGraphicsPipeline();
mCurrentDrawMode = mode;
mGraphicsPipelineDesc->updateTopology(&mGraphicsPipelineTransition, mCurrentDrawMode);
}
// Must be called before the command buffer is started. Can call finish.
if (mVertexArray->getStreamingVertexAttribsMask().any())
{
// All client attribs & any emulated buffered attribs will be updated
ANGLE_TRY(mVertexArray->updateStreamedAttribs(context, firstVertexOrInvalid,
vertexOrIndexCount, instanceCount,
indexTypeOrInvalid, indices));
mGraphicsDirtyBits.set(DIRTY_BIT_VERTEX_BUFFERS);
}
// Create a local object to ensure we flush the descriptor updates to device when we leave this
// function
ScopedDescriptorSetUpdates descriptorSetUpdates(this);
if (mProgram && mProgram->dirtyUniforms())
{
ANGLE_TRY(mProgram->updateUniforms(this));
mGraphicsDirtyBits.set(DIRTY_BIT_DESCRIPTOR_SETS);
}
else if (mProgramPipeline && mProgramPipeline->dirtyUniforms(getState()))
{
ANGLE_TRY(mProgramPipeline->updateUniforms(this));
mGraphicsDirtyBits.set(DIRTY_BIT_DESCRIPTOR_SETS);
}
// Update transform feedback offsets on every draw call when emulating transform feedback. This
// relies on the fact that no geometry/tessellation, indirect or indexed calls are supported in
// ES3.1 (and emulation is not done for ES3.2).
if (getFeatures().emulateTransformFeedback.enabled &&
mState.isTransformFeedbackActiveUnpaused())
{
ASSERT(firstVertexOrInvalid != -1);
mXfbBaseVertex = firstVertexOrInvalid;
mXfbVertexCountPerInstance = vertexOrIndexCount;
invalidateGraphicsDriverUniforms();
}
DirtyBits dirtyBits = mGraphicsDirtyBits & dirtyBitMask;
if (dirtyBits.none())
{
ASSERT(mRenderPassCommandBuffer);
return angle::Result::Continue;
}
// Flush any relevant dirty bits.
for (DirtyBits::Iterator dirtyBitIter = dirtyBits.begin(); dirtyBitIter != dirtyBits.end();
++dirtyBitIter)
{
ASSERT(mGraphicsDirtyBitHandlers[*dirtyBitIter]);
ANGLE_TRY((this->*mGraphicsDirtyBitHandlers[*dirtyBitIter])(&dirtyBitIter, dirtyBitMask));
}
mGraphicsDirtyBits &= ~dirtyBitMask;
// Render pass must be always available at this point.
ASSERT(mRenderPassCommandBuffer);
return angle::Result::Continue;
}
angle::Result ContextVk::setupIndexedDraw(const gl::Context *context,
gl::PrimitiveMode mode,
GLsizei indexCount,
GLsizei instanceCount,
gl::DrawElementsType indexType,
const void *indices)
{
ASSERT(mode != gl::PrimitiveMode::LineLoop);
if (indexType != mCurrentDrawElementsType)
{
mCurrentDrawElementsType = indexType;
ANGLE_TRY(onIndexBufferChange(nullptr));
}
const gl::Buffer *elementArrayBuffer = mVertexArray->getState().getElementArrayBuffer();
if (!elementArrayBuffer)
{
mGraphicsDirtyBits.set(DIRTY_BIT_INDEX_BUFFER);
ANGLE_TRY(mVertexArray->convertIndexBufferCPU(this, indexType, indexCount, indices));
mCurrentIndexBufferOffset = 0;
}
else
{
mCurrentIndexBufferOffset = reinterpret_cast<VkDeviceSize>(indices);
if (indices != mLastIndexBufferOffset)
{
mGraphicsDirtyBits.set(DIRTY_BIT_INDEX_BUFFER);
mLastIndexBufferOffset = indices;
}
if (shouldConvertUint8VkIndexType(indexType) && mGraphicsDirtyBits[DIRTY_BIT_INDEX_BUFFER])
{
ANGLE_PERF_WARNING(getDebug(), GL_DEBUG_SEVERITY_LOW,
"Potential inefficiency emulating uint8 vertex attributes due to "
"lack of hardware support");
BufferVk *bufferVk = vk::GetImpl(elementArrayBuffer);
VkDeviceSize bufferOffset = 0;
vk::BufferHelper &bufferHelper = bufferVk->getBufferAndOffset(&bufferOffset);
if (bufferHelper.isHostVisible() &&
!bufferHelper.isCurrentlyInUse(getLastCompletedQueueSerial()))
{
uint8_t *src = nullptr;
ANGLE_TRY(bufferVk->mapImpl(this, reinterpret_cast<void **>(&src)));
// Note: bufferOffset is not added here because mapImpl already adds it.
src += reinterpret_cast<uintptr_t>(indices);
const size_t byteCount = static_cast<size_t>(elementArrayBuffer->getSize()) -
reinterpret_cast<uintptr_t>(indices);
ANGLE_TRY(mVertexArray->convertIndexBufferCPU(this, indexType, byteCount, src));
ANGLE_TRY(bufferVk->unmapImpl(this));
}
else
{
ANGLE_TRY(mVertexArray->convertIndexBufferGPU(this, bufferVk, indices));
}
mCurrentIndexBufferOffset = 0;
}
}
return setupDraw(context, mode, 0, indexCount, instanceCount, indexType, indices,
mIndexedDirtyBitsMask);
}
angle::Result ContextVk::setupIndirectDraw(const gl::Context *context,
gl::PrimitiveMode mode,
DirtyBits dirtyBitMask,
vk::BufferHelper *indirectBuffer,
VkDeviceSize indirectBufferOffset)
{
GLint firstVertex = -1;
GLsizei vertexCount = 0;
GLsizei instanceCount = 1;