MVKCommandBuffer track rendering viewMask from commands.

MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h

@@ -299,12 +299,13 @@ typedef struct {
 	uint32_t maxActiveMetalCommandBuffersPerQueue;
 
 	/**
-	 * Metal allows only 8192 occlusion queries per MTLBuffer. If enabled, MoltenVK
-	 * allocates a MTLBuffer for each query pool, allowing each query pool to support
-	 * 8192 queries, which may slow performance or cause unexpected behaviour if the query
-	 * pool is not established prior to a Metal renderpass, or if the query pool is changed
-	 * within a renderpass. If disabled, one MTLBuffer will be shared by all query pools,
-	 * which improves performance, but limits the total device queries to 8192.
+	 * Depending on the GPU, Metal allows 8192 or 32768 occlusion queries per MTLBuffer.
+	 * If enabled, MoltenVK allocates a MTLBuffer for each query pool, allowing each query
+	 * pool to support that permitted number of queries. This may slow performance or cause
+	 * unexpected behaviour if the query pool is not established prior to a Metal renderpass,
+	 * or if the query pool is changed within a renderpass. If disabled, one MTLBuffer will
+	 * be shared by all query pools, which improves performance, but limits the total device
+	 * queries to the permitted number.
 	 *
 	 * The value of this parameter may be changed at any time during application runtime,
 	 * and the changed value will immediately effect subsequent MoltenVK behaviour.

MoltenVK/MoltenVK/Commands/MVKCmdRenderPass.mm

@@ -61,6 +61,8 @@ VkResult MVKCmdBeginRenderPassBase::setContent(MVKCommandBuffer* cmdBuff,
 		}
 	}
 
+	cmdBuff->_currentSubpassInfo.beginRenderpass(_renderPass);
+
 	return VK_SUCCESS;
 }
 
@@ -126,6 +128,8 @@ VkResult MVKCmdNextSubpass::setContent(MVKCommandBuffer* cmdBuff,
 									   VkSubpassContents contents) {
 	_contents = contents;
 
+	cmdBuff->_currentSubpassInfo.nextSubpass();
+
 	return VK_SUCCESS;
 }
 
@@ -144,12 +148,13 @@ void MVKCmdNextSubpass::encode(MVKCommandEncoder* cmdEncoder) {
 #pragma mark MVKCmdEndRenderPass
 
 VkResult MVKCmdEndRenderPass::setContent(MVKCommandBuffer* cmdBuff) {
+	cmdBuff->_currentSubpassInfo = {};
 	return VK_SUCCESS;
 }
 
 VkResult MVKCmdEndRenderPass::setContent(MVKCommandBuffer* cmdBuff,
 										 const VkSubpassEndInfo* pEndSubpassInfo) {
-	return VK_SUCCESS;
+	return setContent(cmdBuff);
 }
 
 void MVKCmdEndRenderPass::encode(MVKCommandEncoder* cmdEncoder) {
@@ -178,6 +183,8 @@ VkResult MVKCmdBeginRendering<N>::setContent(MVKCommandBuffer* cmdBuff,
 		_renderingInfo.pStencilAttachment = &_stencilAttachment;
 	}
 
+	cmdBuff->_currentSubpassInfo.beginRendering(pRenderingInfo->viewMask);
+
 	return VK_SUCCESS;
 }
 
@@ -196,6 +203,7 @@ template class MVKCmdBeginRendering<8>;
 #pragma mark MVKCmdEndRendering
 
 VkResult MVKCmdEndRendering::setContent(MVKCommandBuffer* cmdBuff) {
+	cmdBuff->_currentSubpassInfo = {};
 	return VK_SUCCESS;
 }
 

MoltenVK/MoltenVK/Commands/MVKCommandBuffer.h

@@ -65,6 +65,24 @@ private:
 } MVKCommandEncodingContext;
 
 
+#pragma mark -
+#pragma mark MVKCurrentSubpassInfo
+
+/** Tracks current render subpass information. */
+typedef struct MVKCurrentSubpassInfo {
+	MVKRenderPass* renderpass;
+	uint32_t subpassIndex;
+	uint32_t subpassViewMask;
+
+	void beginRenderpass(MVKRenderPass* rp);
+	void nextSubpass();
+	void beginRendering(uint32_t viewMask);
+
+private:
+	void updateViewMask();
+} MVKCurrentSubpassInfo;
+
+
 #pragma mark -
 #pragma mark MVKCommandBuffer
 
@@ -113,6 +131,9 @@ public:
 	 */
 	uint32_t getViewCount() const;
 
+	/** Updated as renderpass commands are added. */
+	MVKCurrentSubpassInfo _currentSubpassInfo;
+
     /**
      * Metal requires that a visibility buffer is established when a render pass is created, 
      * but Vulkan permits it to be set during a render pass. When the first occlusion query
@@ -176,7 +197,7 @@ protected:
 	MVKSmallVector<VkFormat, kMVKDefaultAttachmentCount> _colorAttachmentFormats;
 	MVKCommandPool* _commandPool;
 	VkCommandBufferInheritanceInfo _secondaryInheritanceInfo;
-	VkCommandBufferInheritanceRenderingInfo _secondaryInerhitanceRenderingInfo;
+	VkCommandBufferInheritanceRenderingInfo _secondaryInheritanceRenderingInfo;
 	id<MTLCommandBuffer> _prefilledMTLCmdBuffer = nil;
     MVKCommandEncodingContext* _immediateCmdEncodingContext = nullptr;
     MVKCommandEncoder* _immediateCmdEncoder = nullptr;

MoltenVK/MoltenVK/Commands/MVKCommandBuffer.mm

@@ -30,6 +30,7 @@
 
 using namespace std;
 
+
 #pragma mark -
 #pragma mark MVKCommandEncodingContext
 
@@ -57,6 +58,28 @@ MVKCommandEncodingContext::~MVKCommandEncodingContext() {
 }
 
 
+#pragma mark -
+#pragma mark MVKCurrentSubpassInfo
+
+void MVKCurrentSubpassInfo::beginRenderpass(MVKRenderPass* rp) {
+	renderpass = rp;
+	subpassIndex = 0;
+	updateViewMask();
+}
+void MVKCurrentSubpassInfo::nextSubpass() {
+	subpassIndex++;
+	updateViewMask();
+}
+void MVKCurrentSubpassInfo::beginRendering(uint32_t viewMask) {
+	renderpass = nullptr;
+	subpassIndex = 0;
+	subpassViewMask = viewMask;
+}
+void MVKCurrentSubpassInfo::updateViewMask() {
+	subpassViewMask = renderpass ? renderpass->getSubpass(subpassIndex)->getViewMask() : 0;
+}
+
+
 #pragma mark -
 #pragma mark MVKCommandBuffer
 
@@ -80,11 +103,11 @@ VkResult MVKCommandBuffer::begin(const VkCommandBufferBeginInfo* pBeginInfo) {
 		for (const auto* next = (VkBaseInStructure*)_secondaryInheritanceInfo.pNext; next; next = next->pNext) {
 			switch (next->sType) {
 				case VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_RENDERING_INFO: {
-					if (mvkSetOrClear(&_secondaryInerhitanceRenderingInfo, (VkCommandBufferInheritanceRenderingInfo*)next)) {
-						for (uint32_t caIdx = 0; caIdx < _secondaryInerhitanceRenderingInfo.colorAttachmentCount; caIdx++) {
-							_colorAttachmentFormats.push_back(_secondaryInerhitanceRenderingInfo.pColorAttachmentFormats[caIdx]);
+					if (mvkSetOrClear(&_secondaryInheritanceRenderingInfo, (VkCommandBufferInheritanceRenderingInfo*)next)) {
+						for (uint32_t caIdx = 0; caIdx < _secondaryInheritanceRenderingInfo.colorAttachmentCount; caIdx++) {
+							_colorAttachmentFormats.push_back(_secondaryInheritanceRenderingInfo.pColorAttachmentFormats[caIdx]);
 						}
-						_secondaryInerhitanceRenderingInfo.pColorAttachmentFormats = _colorAttachmentFormats.data();
+						_secondaryInheritanceRenderingInfo.pColorAttachmentFormats = _colorAttachmentFormats.data();
 					}
 					break;
 				}
@@ -147,6 +170,7 @@ VkResult MVKCommandBuffer::reset(VkCommandBufferResetFlags flags) {
 	_wasExecuted = false;
 	_isExecutingNonConcurrently.clear();
 	_commandCount = 0;
+	_currentSubpassInfo = {};
 	_needsVisibilityResultMTLBuffer = false;
 	_hasStageCounterTimestampCommand = false;
 	_lastTessellationPipeline = nullptr;
@@ -247,8 +271,10 @@ uint32_t MVKCommandBuffer::getViewCount() const {
 		if (inheritedRenderPass) {
 			viewMask = inheritedRenderPass->getSubpass(_secondaryInheritanceInfo.subpass)->getViewMask();
 		} else {
-			viewMask = _secondaryInerhitanceRenderingInfo.viewMask;
+			viewMask = _secondaryInheritanceRenderingInfo.viewMask;
 		}
+	} else {
+		viewMask = _currentSubpassInfo.subpassViewMask;
 	}
 	return max(__builtin_popcount(viewMask), 1);
 }

MoltenVK/MoltenVK/Utility/MVKEnvironment.h

@@ -118,7 +118,7 @@ void mvkSetConfig(const MVKConfiguration& mvkConfig);
 #   define MVK_CONFIG_MAX_ACTIVE_METAL_COMMAND_BUFFERS_PER_QUEUE    64
 #endif
 
-/** Support more than 8192 occlusion queries per buffer. Enabled by default. */
+/** Support more than 8192 or 32768 occlusion queries per device. Enabled by default. */
 #ifndef MVK_CONFIG_SUPPORT_LARGE_QUERY_POOLS
 #   define MVK_CONFIG_SUPPORT_LARGE_QUERY_POOLS    1
 #endif