Merge pull request #896 from billhollings/master
Support vertex attribute offsets larger than the vertex buffer stride.
diff --git a/Docs/Whats_New.md b/Docs/Whats_New.md
index 9afd2df..c1ce731 100644
--- a/Docs/Whats_New.md
+++ b/Docs/Whats_New.md
@@ -20,22 +20,54 @@
Released 2020/06/01
- Add support for extensions:
+ - `VK_GOOGLE_display_timing`
- `VK_KHR_external_memory` (non-functional groundwork for future
Metal-resource Vulkan extension).
- `VK_KHR_external_memory_capabilities` (non-functional groundwork
for future Metal-resource Vulkan extension).
-- Memory consumption improvements.
+- Memory consumption improvements in command handling and vector pre-allocation optimizations.
+- `vkQueuePresentKHR()` returns a `VkResult` for each swapchain.
+- `MVKPipeline` disable fragment shader outputs for unused attachments.
+- `MVKBuffer` support texel buffers in host-coherent memory on Mac.
+- `MVKDescriptor` pass buffers to shaders that do atomic image accesses.
+- Support vertex attribute offsets larger than the vertex buffer stride.
+- Fix crash when more than two GPUs.
+- Fix issue where `vkGetPhysicalDeviceFormatProperties()` incorrectly returned
+ properties for unsupported formats.
+- Fix stack overflow in when logging and reporting very long messages.
+- Fix situation where compute pipeline state not retained across copy and renderpass operations.
+- Fix buffer offset calculation.
+- Fixes to maximum FPS calculations.
+- Enables format atomic capabilities only when format supports it.
+- Add `MVKSmallVector` as a more memory efficient substitute of `MVKVector`.
- Reinstate `VulkanSamples API-Samples` demo apps and add
`input_attachment` and `push_descriptors` demos.
-- `vkQueuePresentKHR()` returns a `VkResult` for each swapchain.
+- Add `MVK_CONFIG_AUTO_GPU_CAPTURE_OUTPUT_FILE` environment variable
+ to support capturing GPU traces to a file.
- Consolidate frame and non-frame performance reporting.
- Remove `vkGetSwapchainPerformanceMVK()` from API.
- Swapchain performance can be retrieved with other activity performance
through `vkGetPerformanceStatisticsMVK()`.
- Add `MVK_CONFIG_PERFORMANCE_LOGGING_INLINE` env var to enable/disable
logging of performance of each activity when it happens.
-- Fix crash when more than two GPUs.
+ - Reduce thread locking on performance statistics collection.
+- Numerous documentation typo corrections.
- Support Xcode 11.5.
+- Update to latest SPIRV-Cross version:
+ - MSL: mark `BuiltInFragCoord` as implicitly used for subpass reads.
+ - MSL: Deal with cases where builtin is implicitly needed, declared, but unused.
+ - MSL: Do not use base expression with PhysicalTypeID `OpCompositeExtract`.
+ - MSL: Add options to control emission of fragment outputs.
+ - MSL: Force disabled fragment builtins to have the right name.
+ - MSL: Allow removing clip distance user varyings.
+ - MSL: Support edge case with DX layout in scalar block layout.
+ - MSL: Deal correctly with initializers on Private variables.
+ - MSL: Fix case where `subpassInput` is passed to leaf functions.
+ - MSL: Redirect member indices when buffer has been sorted by Offset.
+ - MSL: If the packed type is scalar, don't emit "pack_" prefix.
+ - MSL: Avoid packed arrays in more cases.
+ - Do not add NonWritable/NonReadable decorations for regular images.
+ - Expose a query if samplers or images are comparison resources.
diff --git a/ExternalDependencies.xcodeproj/project.pbxproj b/ExternalDependencies.xcodeproj/project.pbxproj
index 1b0f51c..ec56c28 100644
--- a/ExternalDependencies.xcodeproj/project.pbxproj
+++ b/ExternalDependencies.xcodeproj/project.pbxproj
@@ -2980,7 +2980,7 @@
A9F55D25198BE6A7004EC31B /* Project object */ = {
isa = PBXProject;
attributes = {
- LastUpgradeCheck = 1140;
+ LastUpgradeCheck = 1150;
ORGANIZATIONNAME = "The Brenwill Workshop Ltd.";
TargetAttributes = {
A972A7E421CEC72F0013AB25 = {
diff --git "a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies \050Debug\051.xcscheme" "b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies \050Debug\051.xcscheme"
index 18a6d08..9677c5a 100644
--- "a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies \050Debug\051.xcscheme"
+++ "b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies \050Debug\051.xcscheme"
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-iOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-iOS.xcscheme
index fab5380..f5912bb 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-iOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-iOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-macOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-macOS.xcscheme
index 2313055..72e1129 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-macOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies-macOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies.xcscheme
index 2ebf29f..7d3952c 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/ExternalDependencies.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-iOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-iOS.xcscheme
index 944ddf4..50db954 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-iOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-iOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-macOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-macOS.xcscheme
index 9f89d19..a4ceebf 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-macOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Cross-macOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-iOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-iOS.xcscheme
index 03437e5..915c7e2 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-iOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-iOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-macOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-macOS.xcscheme
index cbfbdd3..2f53c1a 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-macOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/SPIRV-Tools-macOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-iOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-iOS.xcscheme
index dc8c0d1..62d3081 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-iOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-iOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-macOS.xcscheme b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-macOS.xcscheme
index 4a85559..4b203b9 100644
--- a/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-macOS.xcscheme
+++ b/ExternalDependencies.xcodeproj/xcshareddata/xcschemes/glslang-macOS.xcscheme
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
<Scheme
- LastUpgradeVersion = "1140"
+ LastUpgradeVersion = "1150"
version = "2.0">
<BuildAction
parallelizeBuildables = "YES"
diff --git a/ExternalRevisions/SPIRV-Cross_repo_revision b/ExternalRevisions/SPIRV-Cross_repo_revision
index e20a35a..9fa21a3 100644
--- a/ExternalRevisions/SPIRV-Cross_repo_revision
+++ b/ExternalRevisions/SPIRV-Cross_repo_revision
@@ -1 +1 @@
-fbc560782c42bab4581dbc7705da5013bf864956
+d385bf096f5dabbc4cdaeb6872b0f64be1a63ad0
diff --git a/MoltenVK/MoltenVK/Commands/MVKCommandEncoderState.mm b/MoltenVK/MoltenVK/Commands/MVKCommandEncoderState.mm
index d63ba37..c06b32a 100644
--- a/MoltenVK/MoltenVK/Commands/MVKCommandEncoderState.mm
+++ b/MoltenVK/MoltenVK/Commands/MVKCommandEncoderState.mm
@@ -598,22 +598,33 @@
void MVKGraphicsResourcesCommandEncoderState::encodeImpl(uint32_t stage) {
- MVKPipeline* pipeline = _cmdEncoder->_graphicsPipelineState.getPipeline();
+ MVKGraphicsPipeline* pipeline = (MVKGraphicsPipeline*)_cmdEncoder->_graphicsPipelineState.getPipeline();
bool fullImageViewSwizzle = pipeline->fullImageViewSwizzle() || _cmdEncoder->getDevice()->_pMetalFeatures->nativeTextureSwizzle;
- bool forTessellation = ((MVKGraphicsPipeline*)pipeline)->isTessellationPipeline();
+ bool forTessellation = pipeline->isTessellationPipeline();
if (stage == (forTessellation ? kMVKGraphicsStageVertex : kMVKGraphicsStageRasterization)) {
encodeBindings(kMVKShaderStageVertex, "vertex", fullImageViewSwizzle,
- [](MVKCommandEncoder* cmdEncoder, MVKMTLBufferBinding& b)->void {
- if (b.isInline)
+ [pipeline](MVKCommandEncoder* cmdEncoder, MVKMTLBufferBinding& b)->void {
+ if (b.isInline) {
cmdEncoder->setVertexBytes(cmdEncoder->_mtlRenderEncoder,
b.mtlBytes,
b.size,
b.index);
- else
+ } else {
[cmdEncoder->_mtlRenderEncoder setVertexBuffer: b.mtlBuffer
offset: b.offset
atIndex: b.index];
+
+ // Add any translated vertex bindings for this binding
+ auto xltdVtxBindings = pipeline->getTranslatedVertexBindings();
+ for (auto& xltdBind : xltdVtxBindings) {
+ if (b.index == pipeline->getMetalBufferIndexForVertexAttributeBinding(xltdBind.binding)) {
+ [cmdEncoder->_mtlRenderEncoder setVertexBuffer: b.mtlBuffer
+ offset: b.offset + xltdBind.translationOffset
+ atIndex: pipeline->getMetalBufferIndexForVertexAttributeBinding(xltdBind.translationBinding)];
+ }
+ }
+ }
},
[](MVKCommandEncoder* cmdEncoder, MVKMTLBufferBinding& b, const MVKArrayRef<uint32_t>& s)->void {
cmdEncoder->setVertexBytes(cmdEncoder->_mtlRenderEncoder,
diff --git a/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.h b/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.h
index 9503ba9..5f01d27 100644
--- a/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.h
+++ b/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.h
@@ -184,6 +184,13 @@
#pragma mark -
#pragma mark MVKGraphicsPipeline
+/** Describes a buffer binding to accommodate vertex attributes with offsets greater than the stride. */
+struct MVKTranslatedVertexBinding {
+ uint16_t binding;
+ uint16_t translationBinding;
+ uint32_t translationOffset;
+};
+
typedef MVKSmallVector<MVKGraphicsStage, 4> MVKPiplineStages;
/** The number of dynamic states possible in Vulkan. */
@@ -230,6 +237,12 @@
/** Returns true if the tessellation control shader needs a buffer to store its per-patch output. */
bool needsTessCtlPatchOutputBuffer() { return _needsTessCtlPatchOutputBuffer; }
+ /** Returns the Metal vertex buffer index to use for the specified vertex attribute binding number. */
+ uint32_t getMetalBufferIndexForVertexAttributeBinding(uint32_t binding) { return _device->getMetalBufferIndexForVertexAttributeBinding(binding); }
+
+ /** Returns the collection of translated vertex bindings. */
+ MVKArrayRef<MVKTranslatedVertexBinding> getTranslatedVertexBindings() { return _translatedVertexBindings.contents(); }
+
/** Constructs an instance for the device and parent (which may be NULL). */
MVKGraphicsPipeline(MVKDevice* device,
MVKPipelineCache* pipelineCache,
@@ -260,6 +273,7 @@
void addFragmentOutputToPipeline(MTLRenderPipelineDescriptor* plDesc, const SPIRVTessReflectionData& reflectData, const VkGraphicsPipelineCreateInfo* pCreateInfo, bool isTessellationVertexPipeline = false);
bool isRenderingPoints(const VkGraphicsPipelineCreateInfo* pCreateInfo, const SPIRVTessReflectionData& reflectData);
bool verifyImplicitBuffer(bool needsBuffer, MVKShaderImplicitRezBinding& index, MVKShaderStage stage, const char* name, uint32_t reservedBuffers);
+ uint32_t getTranslatedVertexBinding(uint32_t binding, uint32_t translationOffset, uint32_t maxBinding);
const VkPipelineShaderStageCreateInfo* _pVertexSS = nullptr;
const VkPipelineShaderStageCreateInfo* _pTessCtlSS = nullptr;
@@ -272,6 +286,7 @@
MVKSmallVector<VkViewport, kMVKCachedViewportScissorCount> _viewports;
MVKSmallVector<VkRect2D, kMVKCachedViewportScissorCount> _scissors;
+ MVKSmallVector<MVKTranslatedVertexBinding> _translatedVertexBindings;
MTLComputePipelineDescriptor* _mtlTessControlStageDesc = nil;
diff --git a/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.mm b/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.mm
index 781681a..e2d61ec 100644
--- a/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.mm
+++ b/MoltenVK/MoltenVK/GPUObjects/MVKPipeline.mm
@@ -953,40 +953,12 @@
}
}
- // Vertex attributes
- uint32_t vaCnt = pVI->vertexAttributeDescriptionCount;
- uint32_t vbCnt = pVI->vertexBindingDescriptionCount;
- for (uint32_t i = 0; i < vaCnt; i++) {
- const VkVertexInputAttributeDescription* pVKVA = &pVI->pVertexAttributeDescriptions[i];
- if (shaderContext.isVertexAttributeLocationUsed(pVKVA->location)) {
-
- // Vulkan allows offsets to exceed the buffer stride, but Metal doesn't.
- // Only check non-zero offsets, as it's common for both to be zero when step rate is instance.
- if (pVKVA->offset > 0) {
- const VkVertexInputBindingDescription* pVKVB = pVI->pVertexBindingDescriptions;
- for (uint32_t j = 0; j < vbCnt; j++, pVKVB++) {
- if (pVKVB->binding == pVKVA->binding) {
- if (pVKVA->offset >= pVKVB->stride) {
- setConfigurationResult(reportError(VK_ERROR_INITIALIZATION_FAILED, "Under Metal, vertex attribute offsets must not exceed the vertex buffer stride."));
- return false;
- }
- break;
- }
- }
- }
-
- MTLVertexAttributeDescriptor* vaDesc = plDesc.vertexDescriptor.attributes[pVKVA->location];
- vaDesc.format = getPixelFormats()->getMTLVertexFormat(pVKVA->format);
- vaDesc.bufferIndex = _device->getMetalBufferIndexForVertexAttributeBinding(pVKVA->binding);
- vaDesc.offset = pVKVA->offset;
- }
- }
-
// Vertex buffer bindings
+ uint32_t vbCnt = pVI->vertexBindingDescriptionCount;
+ uint32_t maxBinding = 0;
for (uint32_t i = 0; i < vbCnt; i++) {
const VkVertexInputBindingDescription* pVKVB = &pVI->pVertexBindingDescriptions[i];
- uint32_t vbIdx = _device->getMetalBufferIndexForVertexAttributeBinding(pVKVB->binding);
- if (shaderContext.isVertexBufferUsed(vbIdx)) {
+ if (shaderContext.isVertexBufferUsed(pVKVB->binding)) {
// Vulkan allows any stride, but Metal only allows multiples of 4.
// TODO: We should try to expand the buffer to the required alignment in that case.
@@ -995,6 +967,8 @@
return false;
}
+ maxBinding = max(pVKVB->binding, maxBinding);
+ uint32_t vbIdx = getMetalBufferIndexForVertexAttributeBinding(pVKVB->binding);
MTLVertexBufferLayoutDescriptor* vbDesc = plDesc.vertexDescriptor.layouts[vbIdx];
vbDesc.stride = (pVKVB->stride == 0) ? sizeof(simd::float4) : pVKVB->stride; // Vulkan allows zero stride but Metal doesn't. Default to float4
vbDesc.stepFunction = mvkMTLVertexStepFunctionFromVkVertexInputRate(pVKVB->inputRate);
@@ -1007,7 +981,7 @@
vbCnt = pVertexInputDivisorState->vertexBindingDivisorCount;
for (uint32_t i = 0; i < vbCnt; i++) {
const VkVertexInputBindingDivisorDescriptionEXT* pVKVB = &pVertexInputDivisorState->pVertexBindingDivisors[i];
- uint32_t vbIdx = _device->getMetalBufferIndexForVertexAttributeBinding(pVKVB->binding);
+ uint32_t vbIdx = getMetalBufferIndexForVertexAttributeBinding(pVKVB->binding);
if (shaderContext.isVertexBufferUsed(vbIdx)) {
MTLVertexBufferLayoutDescriptor* vbDesc = plDesc.vertexDescriptor.layouts[vbIdx];
if (vbDesc.stepFunction == MTLVertexStepFunctionPerInstance) {
@@ -1019,9 +993,93 @@
}
}
+ // Vertex attributes
+ uint32_t vaCnt = pVI->vertexAttributeDescriptionCount;
+ for (uint32_t i = 0; i < vaCnt; i++) {
+ const VkVertexInputAttributeDescription* pVKVA = &pVI->pVertexAttributeDescriptions[i];
+ if (shaderContext.isVertexAttributeLocationUsed(pVKVA->location)) {
+ uint32_t vaBinding = pVKVA->binding;
+ uint32_t vaOffset = pVKVA->offset;
+
+ // Vulkan allows offsets to exceed the buffer stride, but Metal doesn't.
+ // If this is the case, fetch an a translated artificial buffer binding, using the same MTLBuffer,
+ // but that is translated so that the reduced VA offset fits into the binding stride.
+ // Only check non-zero offsets, as it's common for both to be zero when step rate is instance.
+ if (vaOffset > 0) {
+ const VkVertexInputBindingDescription* pVKVB = pVI->pVertexBindingDescriptions;
+ for (uint32_t j = 0; j < vbCnt; j++, pVKVB++) {
+ if (pVKVB->binding == pVKVA->binding) {
+ if (vaOffset >= pVKVB->stride) {
+ // Move vertex attribute offset into the stride. This vertex attribute may be
+ // combined with other vertex attributes into the same translated buffer binding.
+ // But if the reduced offset combined with the vertex attribute size still won't
+ // fit into the buffer binding stride, force the vertex attribute offset to zero,
+ // effectively dedicating this vertex attribute to its own buffer binding.
+ uint32_t origOffset = vaOffset;
+ vaOffset %= pVKVB->stride;
+ if (vaOffset + getPixelFormats()->getBytesPerBlock(pVKVA->format) > pVKVB->stride) {
+ vaOffset = 0;
+ }
+ vaBinding = getTranslatedVertexBinding(vaBinding, origOffset - vaOffset, maxBinding);
+ }
+ break;
+ }
+ }
+ }
+
+ MTLVertexAttributeDescriptor* vaDesc = plDesc.vertexDescriptor.attributes[pVKVA->location];
+ vaDesc.format = getPixelFormats()->getMTLVertexFormat(pVKVA->format);
+ vaDesc.bufferIndex = getMetalBufferIndexForVertexAttributeBinding(vaBinding);
+ vaDesc.offset = vaOffset;
+ }
+ }
+
+ // Run through the vertex bindings. Add a new Metal vertex layout for each translated binding,
+ // identical to the original layout. The translated binding will index into the same MTLBuffer,
+ // but at an offset that is one or more strides away from the original.
+ for (uint32_t i = 0; i < vbCnt; i++) {
+ const VkVertexInputBindingDescription* pVKVB = &pVI->pVertexBindingDescriptions[i];
+ uint32_t vbVACnt = shaderContext.countVertexAttributesAt(pVKVB->binding);
+ if (vbVACnt > 0) {
+ uint32_t vbIdx = getMetalBufferIndexForVertexAttributeBinding(pVKVB->binding);
+ MTLVertexBufferLayoutDescriptor* vbDesc = plDesc.vertexDescriptor.layouts[vbIdx];
+
+ uint32_t xldtVACnt = 0;
+ for (auto& xltdBind : _translatedVertexBindings) {
+ if (xltdBind.binding == pVKVB->binding) {
+ uint32_t vbXltdIdx = getMetalBufferIndexForVertexAttributeBinding(xltdBind.translationBinding);
+ MTLVertexBufferLayoutDescriptor* vbXltdDesc = plDesc.vertexDescriptor.layouts[vbXltdIdx];
+ vbXltdDesc.stride = vbDesc.stride;
+ vbXltdDesc.stepFunction = vbDesc.stepFunction;
+ vbXltdDesc.stepRate = vbDesc.stepRate;
+ xldtVACnt++;
+ }
+ }
+
+ // If all of the vertex attributes at this vertex buffer binding have been translated, remove it.
+ if (xldtVACnt == vbVACnt) { vbDesc.stride = 0; }
+ }
+ }
+
return true;
}
+// Returns a translated binding for the existing binding and translation offset, creating it if needed.
+uint32_t MVKGraphicsPipeline::getTranslatedVertexBinding(uint32_t binding, uint32_t translationOffset, uint32_t maxBinding) {
+ // See if a translated binding already exists (for example if more than one VA needs the same translation).
+ for (auto& xltdBind : _translatedVertexBindings) {
+ if (xltdBind.binding == binding && xltdBind.translationOffset == translationOffset) {
+ return xltdBind.translationBinding;
+ }
+ }
+
+ // Get next available binding point and add a translation binding description for it
+ uint16_t xltdBindPt = (uint16_t)(maxBinding + _translatedVertexBindings.size() + 1);
+ _translatedVertexBindings.push_back( {.binding = (uint16_t)binding, .translationBinding = xltdBindPt, .translationOffset = translationOffset} );
+
+ return xltdBindPt;
+}
+
void MVKGraphicsPipeline::addTessellationToPipeline(MTLRenderPipelineDescriptor* plDesc,
const SPIRVTessReflectionData& reflectData,
const VkPipelineTessellationStateCreateInfo* pTS) {
@@ -1186,8 +1244,7 @@
// Set binding and offset from Vulkan vertex attribute
MSLVertexAttribute va;
va.vertexAttribute.location = pVKVA->location;
- va.vertexAttribute.msl_buffer = _device->getMetalBufferIndexForVertexAttributeBinding(pVKVA->binding);
- va.vertexAttribute.msl_offset = pVKVA->offset;
+ va.binding = pVKVA->binding;
// Metal can't do signedness conversions on vertex buffers (rdar://45922847). If the shader
// and the vertex attribute have mismatched signedness, we have to fix the shader
@@ -1223,17 +1280,6 @@
}
- // Set stride and input rate of vertex attribute from corresponding Vulkan vertex bindings
- uint32_t vbCnt = pCreateInfo->pVertexInputState->vertexBindingDescriptionCount;
- for (uint32_t vbIdx = 0; vbIdx < vbCnt; vbIdx++) {
- const VkVertexInputBindingDescription* pVKVB = &pCreateInfo->pVertexInputState->pVertexBindingDescriptions[vbIdx];
- if (pVKVB->binding == pVKVA->binding) {
- va.vertexAttribute.msl_stride = pVKVB->stride;
- va.vertexAttribute.per_instance = (pVKVB->inputRate == VK_VERTEX_INPUT_RATE_INSTANCE);
- break;
- }
- }
-
shaderContext.vertexAttributes.push_back(va);
}
}
@@ -1652,10 +1698,6 @@
template<class Archive>
void serialize(Archive & archive, MSLVertexAttr& va) {
archive(va.location,
- va.msl_buffer,
- va.msl_offset,
- va.msl_stride,
- va.per_instance,
va.format,
va.builtin);
}
@@ -1731,6 +1773,7 @@
template<class Archive>
void serialize(Archive & archive, MSLVertexAttribute& va) {
archive(va.vertexAttribute,
+ va.binding,
va.isUsedByShader);
}
diff --git a/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.cpp b/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.cpp
index 6ee2d97..19fc8f8 100644
--- a/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.cpp
+++ b/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.cpp
@@ -103,12 +103,9 @@
MVK_PUBLIC_SYMBOL bool MSLVertexAttribute::matches(const MSLVertexAttribute& other) const {
if (vertexAttribute.location != other.vertexAttribute.location) { return false; }
- if (vertexAttribute.msl_buffer != other.vertexAttribute.msl_buffer) { return false; }
- if (vertexAttribute.msl_offset != other.vertexAttribute.msl_offset) { return false; }
- if (vertexAttribute.msl_stride != other.vertexAttribute.msl_stride) { return false; }
if (vertexAttribute.format != other.vertexAttribute.format) { return false; }
if (vertexAttribute.builtin != other.vertexAttribute.builtin) { return false; }
- if (!!vertexAttribute.per_instance != !!other.vertexAttribute.per_instance) { return false; }
+ if (binding != other.binding) { return false; }
return true;
}
@@ -158,12 +155,12 @@
return false;
}
-// Check them all in case inactive VA's duplicate buffers used by active VA's.
-MVK_PUBLIC_SYMBOL bool SPIRVToMSLConversionConfiguration::isVertexBufferUsed(uint32_t mslBuffer) const {
- for (auto& va : vertexAttributes) {
- if ((va.vertexAttribute.msl_buffer == mslBuffer) && va.isUsedByShader) { return true; }
- }
- return false;
+MVK_PUBLIC_SYMBOL uint32_t SPIRVToMSLConversionConfiguration::countVertexAttributesAt(uint32_t binding) const {
+ uint32_t vaCnt = 0;
+ for (auto& va : vertexAttributes) {
+ if ((va.binding == binding) && va.isUsedByShader) { vaCnt++; }
+ }
+ return vaCnt;
}
MVK_PUBLIC_SYMBOL void SPIRVToMSLConversionConfiguration::markAllAttributesAndResourcesUsed() {
diff --git a/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.h b/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.h
index 6cb0cea..0d5ddb7 100644
--- a/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.h
+++ b/MoltenVKShaderConverter/MoltenVKSPIRVToMSLConverter/SPIRVToMSLConverter.h
@@ -74,6 +74,7 @@
typedef struct MSLVertexAttribute {
SPIRV_CROSS_NAMESPACE::MSLVertexAttr vertexAttribute;
+ uint32_t binding = 0;
bool isUsedByShader = false;
/**
@@ -135,8 +136,11 @@
/** Returns whether the vertex attribute at the specified location is used by the shader. */
bool isVertexAttributeLocationUsed(uint32_t location) const;
- /** Returns whether the vertex buffer at the specified Metal binding index is used by the shader. */
- bool isVertexBufferUsed(uint32_t mslBuffer) const;
+ /** Returns the number of vertex attributes bound to the specified Vulkan buffer binding, and used by the shader. */
+ uint32_t countVertexAttributesAt(uint32_t binding) const;
+
+ /** Returns whether the vertex buffer at the specified Vulkan binding is used by the shader. */
+ bool isVertexBufferUsed(uint32_t binding) const { return countVertexAttributesAt(binding) > 0; }
/** Marks all vertex attributes and resources as being used by the shader. */
void markAllAttributesAndResourcesUsed();
