Add GPUQueryType::BinaryOcclusion for simpler visibility checks

Source/Engine/Graphics/Enums.h

@@ -410,6 +410,11 @@ enum class GPUQueryType
     /// </summary>
     Occlusion = 1,
 
+    /// <summary>
+    /// Tests object visibility by checking depth and stencil tests. Works like Occlusion query, except returns simply a binary 0/1 result: 0 indicates that no samples passed depth and stencil testing; 1 indicates that at least one sample passed.
+    /// </summary>
+    BinaryOcclusion = 2,
+
     MAX
 };
 

Source/Engine/GraphicsDevice/DirectX/DX11/GPUContextDX11.cpp

@@ -303,7 +303,7 @@ void GPUContextDX11::SetRenderTarget(GPUTextureView* depthBuffer, GPUTextureView
 
 void GPUContextDX11::SetRenderTarget(GPUTextureView* depthBuffer, const Span<GPUTextureView*>& rts)
 {
-    ASSERT(Math::IsInRange(rts.Length(), 1, GPU_MAX_RT_BINDED));
+    ASSERT(Math::IsInRange(rts.Length(), 0, GPU_MAX_RT_BINDED));
 
     auto depthBufferDX11 = static_cast<GPUTextureViewDX11*>(depthBuffer);
     ID3D11DepthStencilView* dsv = depthBufferDX11 ? depthBufferDX11->DSV() : nullptr;
@@ -590,7 +590,18 @@ uint64 GPUContextDX11::BeginQuery(GPUQueryType type)
         auto& query = _device->_queries.AddOne();
         query.Type = type;
         D3D11_QUERY_DESC queryDesc;
-        queryDesc.Query = type == GPUQueryType::Occlusion ? D3D11_QUERY_OCCLUSION : D3D11_QUERY_TIMESTAMP;
+        switch (type)
+        {
+        case GPUQueryType::Timer:
+            queryDesc.Query = D3D11_QUERY_TIMESTAMP;
+            break;
+        case GPUQueryType::Occlusion:
+            queryDesc.Query = D3D11_QUERY_OCCLUSION;
+            break;
+        case GPUQueryType::BinaryOcclusion:
+            queryDesc.Query = D3D11_QUERY_OCCLUSION_PREDICATE;
+            break;
+        }
         queryDesc.MiscFlags = 0;
         HRESULT hr = _device->GetDevice()->CreateQuery(&queryDesc, &query.Query);
         LOG_DIRECTX_RESULT_WITH_RETURN(hr, 0);

Source/Engine/GraphicsDevice/DirectX/DX11/GPUDeviceDX11.cpp

@@ -1060,9 +1060,15 @@ RETRY:
             }
         }
     }
+    else if (q.Type == (uint16)GPUQueryType::Occlusion)
+    {
+        hasData = context->GetData(query.Query, &result, sizeof(UINT64), 0) == S_OK;
+    }
     else
     {
-        hasData = context->GetData(query.Query, &result, sizeof(uint64), 0) == S_OK;
+        BOOL resultBool;
+        hasData = context->GetData(query.Query, &resultBool, sizeof(BOOL), 0) == S_OK;
+        result = resultBool ? 1 : 0;
     }
 
     if (!hasData && wait)

Source/Engine/GraphicsDevice/DirectX/DX11/GPUDeviceDX11.h

@@ -93,7 +93,7 @@ private:
     bool _allowTearing = false;
     GPUBuffer* _dummyVB = nullptr;
     Array<GPUQueryDataDX11> _queries;
-    Array<uint16> _readyQueries[2]; // Timer and Occlusion
+    Array<uint16> _readyQueries[3]; // Timer and Occlusion and BinaryOcclusion
 
     // Static Samplers
     ID3D11SamplerState* _samplerLinearClamp = nullptr;

Source/Engine/GraphicsDevice/DirectX/DX12/GPUContextDX12.cpp

@@ -948,7 +948,7 @@ void GPUContextDX12::SetRenderTarget(GPUTextureView* depthBuffer, GPUTextureView
 
 void GPUContextDX12::SetRenderTarget(GPUTextureView* depthBuffer, const Span<GPUTextureView*>& rts)
 {
-    ASSERT(Math::IsInRange(rts.Length(), 1, GPU_MAX_RT_BINDED));
+    ASSERT(Math::IsInRange(rts.Length(), 0, GPU_MAX_RT_BINDED));
 
     const auto depthBufferDX12 = static_cast<GPUTextureViewDX12*>(depthBuffer);
 

Source/Engine/GraphicsDevice/DirectX/DX12/GPUDeviceDX12.cpp

@@ -1061,7 +1061,7 @@ GPUQueryDX12 GPUDeviceDX12::AllocQuery(GPUQueryType type)
         // Allocate a new query heap
         PROFILE_MEM(GraphicsCommands);
         auto heap = New<QueryHeapDX12>();
-        int32 size = type == GPUQueryType::Occlusion ? 4096 : 1024;
+        int32 size = type == GPUQueryType::Timer ? 1024 : 4096;
         if (heap->Init(this, type, size))
         {
             Delete(heap);

Source/Engine/GraphicsDevice/DirectX/DX12/QueryHeapDX12.cpp

@@ -28,6 +28,11 @@ bool QueryHeapDX12::Init(GPUDeviceDX12* device, GPUQueryType type, uint32 size)
         QueryType = D3D12_QUERY_TYPE_OCCLUSION;
         heapDesc.Type = D3D12_QUERY_HEAP_TYPE_OCCLUSION;
         break;
+    case GPUQueryType::BinaryOcclusion:
+        _resultSize = sizeof(uint64);
+        QueryType = D3D12_QUERY_TYPE_BINARY_OCCLUSION;
+        heapDesc.Type = D3D12_QUERY_HEAP_TYPE_OCCLUSION;
+        break;
     case GPUQueryType::MAX:
         return true;
     }
@@ -100,11 +105,6 @@ void QueryHeapDX12::EndQueryBatchAndResolveQueryData(GPUContextDX12* context)
     StartQueryBatch();
 }
 
-bool QueryHeapDX12::CanAlloc(int32 count) const
-{
-    return _currentBatch.Open && _currentIndex + count <= GetQueryHeapCount();
-}
-
 void QueryHeapDX12::Alloc(ElementHandle& handle)
 {
     ASSERT(_currentBatch.Open);

Source/Engine/GraphicsDevice/DirectX/DX12/QueryHeapDX12.h

@@ -153,7 +153,10 @@ public:
     /// </summary>
     /// <param name="count">How many elements to allocate?</param>
     /// <returns>True if can alloc new query within the same batch.</returns>
-    bool CanAlloc(int32 count = 1) const;
+    bool CanAlloc(int32 count = 1) const
+    {
+        return _currentBatch.Open && _currentIndex + count <= _queryHeapCount;
+    }
 
     /// <summary>
     /// Allocates the query heap element.

Source/Engine/GraphicsDevice/Vulkan/GPUContextVulkan.cpp

@@ -1466,6 +1466,9 @@ uint64 GPUContextVulkan::BeginQuery(GPUQueryType type)
     case GPUQueryType::Occlusion:
         vkCmdBeginQuery(cmdBuffer->GetHandle(), pool->GetHandle(), query.QueryIndex, 0);
         break;
+    case GPUQueryType::BinaryOcclusion:
+        vkCmdBeginQuery(cmdBuffer->GetHandle(), pool->GetHandle(), query.QueryIndex, VK_QUERY_CONTROL_PRECISE_BIT);
+        break;
     }
     pool->MarkQueryAsStarted(query.QueryIndex);
 
@@ -1492,6 +1495,7 @@ void GPUContextVulkan::EndQuery(uint64 queryID)
 #endif
         break;
     case GPUQueryType::Occlusion:
+    case GPUQueryType::BinaryOcclusion:
         vkCmdEndQuery(cmdBuffer->GetHandle(), pool->GetHandle(), query.QueryIndex);
         break;
     }

Source/Engine/GraphicsDevice/Vulkan/GPUDeviceVulkan.cpp

@@ -642,7 +642,7 @@ QueryPoolVulkan::QueryPoolVulkan(GPUDeviceVulkan* device, int32 capacity, GPUQue
 {
     VkQueryPoolCreateInfo createInfo;
     RenderToolsVulkan::ZeroStruct(createInfo, VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO);
-    createInfo.queryType = type == GPUQueryType::Occlusion ? VK_QUERY_TYPE_OCCLUSION : VK_QUERY_TYPE_TIMESTAMP;
+    createInfo.queryType = type != GPUQueryType::Timer ? VK_QUERY_TYPE_TIMESTAMP : VK_QUERY_TYPE_OCCLUSION;
     createInfo.queryCount = capacity;
     VALIDATE_VULKAN_RESULT(vkCreateQueryPool(device->Device, &createInfo, nullptr, &_handle));
 
@@ -1282,7 +1282,7 @@ int32 GPUDeviceVulkan::GetOrCreateQueryPool(GPUQueryType type)
 
     PROFILE_CPU_NAMED("Create Create Pool");
     PROFILE_MEM(GraphicsCommands);
-    auto pool = New<BufferedQueryPoolVulkan>(this, type == GPUQueryType::Occlusion ? 4096 : 1024, type);
+    auto pool = New<BufferedQueryPoolVulkan>(this, type == GPUQueryType::Timer ? 1024 : 4096, type);
     QueryPools.Add(pool);
     return QueryPools.Count() - 1;
 }
@@ -2217,13 +2217,17 @@ RETRY:
         }
         break;
     case GPUQueryType::Occlusion:
+    case GPUQueryType::BinaryOcclusion:
         hasData = pool->GetResults(query.QueryIndex, result);
+        if (hasData && pool->Type == GPUQueryType::BinaryOcclusion && result > 1)
+            result = 1; // Clamp binary result
         break;
     }
 
     if (!hasData && wait)
     {
         // Wait until data is ready
+        // TODO: use VK_QUERY_RESULT_WAIT_BIT maybe?
         Platform::Yield();
         goto RETRY;
     }

Source/Engine/GraphicsDevice/WebGPU/GPUContextWebGPU.cpp

@@ -229,7 +229,7 @@ void GPUContextWebGPU::SetRenderTarget(GPUTextureView* depthBuffer, GPUTextureVi
 
 void GPUContextWebGPU::SetRenderTarget(GPUTextureView* depthBuffer, const Span<GPUTextureView*>& rts)
 {
-    ASSERT(Math::IsInRange(rts.Length(), 1, GPU_MAX_RT_BINDED));
+    ASSERT(Math::IsInRange(rts.Length(), 0, GPU_MAX_RT_BINDED));
     auto depthBufferGPU = (GPUTextureViewWebGPU*)depthBuffer;
     if (_renderTargetCount != rts.Length() || _depthStencil != depthBufferGPU || Platform::MemoryCompare(_renderTargets, rts.Get(), rts.Length() * sizeof(void*)) != 0)
     {

Source/Engine/GraphicsDevice/WebGPU/GPUDeviceWebGPU.cpp

@@ -860,7 +860,7 @@ GPUQueryWebGPU GPUDeviceWebGPU::AllocateQuery(GPUQueryType type)
 
         // Allocate a new query heap
         PROFILE_MEM(GraphicsCommands);
-        uint32 size = type == GPUQueryType::Occlusion ? 4096 : 1024;
+        uint32 size = type == GPUQueryType::Timer ? 1024 : 4096;
         auto set = New<GPUQuerySetWebGPU>(Device, type, size);
         QuerySets[QuerySetsCount++] = set;
     }