dxvk/src/d3d11/d3d11_context_imm.cpp

#include "d3d11_cmdlist.h"
#include "d3d11_context_imm.h"
#include "d3d11_device.h"
#include "d3d11_fence.h"
#include "d3d11_texture.h"

#include "../util/util_win32_compat.h"

constexpr static uint32_t MinFlushIntervalUs = 750;
constexpr static uint32_t IncFlushIntervalUs = 250;
constexpr static uint32_t MaxPendingSubmits  = 6;

namespace dxvk {
  
  D3D11ImmediateContext::D3D11ImmediateContext(
          D3D11Device*    pParent,
    const Rc<DxvkDevice>& Device)
  : D3D11CommonContext<D3D11ImmediateContext>(pParent, Device, 0, DxvkCsChunkFlag::SingleUse),
    m_csThread(Device, Device->createContext(DxvkContextType::Primary)),
    m_maxImplicitDiscardSize(pParent->GetOptions()->maxImplicitDiscardSize),
    m_submissionFence(new sync::CallbackFence()),
    m_multithread(this, false, pParent->GetOptions()->enableContextLock),
    m_videoContext(this, Device) {
    EmitCs([
      cDevice                 = m_device,
      cRelaxedBarriers        = pParent->GetOptions()->relaxedBarriers,
      cIgnoreGraphicsBarriers = pParent->GetOptions()->ignoreGraphicsBarriers
    ] (DxvkContext* ctx) {
      ctx->beginRecording(cDevice->createCommandList());

      DxvkBarrierControlFlags barrierControl;

      if (cRelaxedBarriers)
        barrierControl.set(DxvkBarrierControl::IgnoreWriteAfterWrite);

      if (cIgnoreGraphicsBarriers)
        barrierControl.set(DxvkBarrierControl::IgnoreGraphicsBarriers);

      ctx->setBarrierControl(barrierControl);
    });
    
    ClearState();
  }
  
  
  D3D11ImmediateContext::~D3D11ImmediateContext() {
    // Avoids hanging when in this state, see comment
    // in DxvkDevice::~DxvkDevice.
    if (this_thread::isInModuleDetachment())
      return;

    ExecuteFlush(GpuFlushType::ExplicitFlush, nullptr);
    SynchronizeCsThread(DxvkCsThread::SynchronizeAll);
    SynchronizeDevice();
  }
  
  
  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::QueryInterface(REFIID riid, void** ppvObject) {
    if (riid == __uuidof(ID3D10Multithread)) {
      *ppvObject = ref(&m_multithread);
      return S_OK;
    }

    if (riid == __uuidof(ID3D11VideoContext)) {
      *ppvObject = ref(&m_videoContext);
      return S_OK;
    }

    return D3D11CommonContext<D3D11ImmediateContext>::QueryInterface(riid, ppvObject);
  }


  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::GetData(
          ID3D11Asynchronous*               pAsync,
          void*                             pData,
          UINT                              DataSize,
          UINT                              GetDataFlags) {
    if (!pAsync || (DataSize && !pData))
      return E_INVALIDARG;
    
    // Check whether the data size is actually correct
    if (DataSize && DataSize != pAsync->GetDataSize())
      return E_INVALIDARG;
    
    // Passing a non-null pData is actually allowed if
    // DataSize is 0, but we should ignore that pointer
    pData = DataSize ? pData : nullptr;

    // Get query status directly from the query object
    auto query = static_cast<D3D11Query*>(pAsync);
    HRESULT hr = query->GetData(pData, GetDataFlags);
    
    // If we're likely going to spin on the asynchronous object,
    // flush the context so that we're keeping the GPU busy.
    if (hr == S_FALSE) {
      // Don't mark the event query as stalling if the app does
      // not intend to spin on it. This reduces flushes on End.
      if (!(GetDataFlags & D3D11_ASYNC_GETDATA_DONOTFLUSH))
        query->NotifyStall();

      // Ignore the DONOTFLUSH flag here as some games will spin
      // on queries without ever flushing the context otherwise.
      D3D10DeviceLock lock = LockContext();
      ConsiderFlush(GpuFlushType::ImplicitSynchronization);
    }
    
    return hr;
  }
  
  
  void STDMETHODCALLTYPE D3D11ImmediateContext::Begin(ID3D11Asynchronous* pAsync) {
    D3D10DeviceLock lock = LockContext();

    if (unlikely(!pAsync))
      return;
    
    auto query = static_cast<D3D11Query*>(pAsync);

    if (unlikely(!query->DoBegin()))
      return;

    EmitCs([cQuery = Com<D3D11Query, false>(query)]
    (DxvkContext* ctx) {
      cQuery->Begin(ctx);
    });
  }


  void STDMETHODCALLTYPE D3D11ImmediateContext::End(ID3D11Asynchronous* pAsync) {
    D3D10DeviceLock lock = LockContext();

    if (unlikely(!pAsync))
      return;
    
    auto query = static_cast<D3D11Query*>(pAsync);

    if (unlikely(!query->DoEnd())) {
      EmitCs([cQuery = Com<D3D11Query, false>(query)]
      (DxvkContext* ctx) {
        cQuery->Begin(ctx);
      });
    }

    EmitCs([cQuery = Com<D3D11Query, false>(query)]
    (DxvkContext* ctx) {
      cQuery->End(ctx);
    });

    if (unlikely(query->TrackStalls())) {
      query->NotifyEnd();

      if (query->IsStalling())
        ExecuteFlush(GpuFlushType::ImplicitSynchronization, nullptr);
      else if (query->IsEvent())
        ConsiderFlush(GpuFlushType::ImplicitStrongHint);
    }
  }


  void STDMETHODCALLTYPE D3D11ImmediateContext::Flush() {
    D3D10DeviceLock lock = LockContext();

    ExecuteFlush(GpuFlushType::ExplicitFlush, nullptr);
  }


  void STDMETHODCALLTYPE D3D11ImmediateContext::Flush1(
          D3D11_CONTEXT_TYPE          ContextType,
          HANDLE                      hEvent) {
    D3D10DeviceLock lock = LockContext();

    ExecuteFlush(GpuFlushType::ExplicitFlush, hEvent);
  }
  
  
  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::Signal(
          ID3D11Fence*                pFence,
          UINT64                      Value) {
    D3D10DeviceLock lock = LockContext();
    auto fence = static_cast<D3D11Fence*>(pFence);

    if (!fence)
      return E_INVALIDARG;

    EmitCs([
      cFence = fence->GetFence(),
      cValue = Value
    ] (DxvkContext* ctx) {
      ctx->signalFence(cFence, cValue);
    });

    ExecuteFlush(GpuFlushType::ExplicitFlush, nullptr);
    return S_OK;
  }


  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::Wait(
          ID3D11Fence*                pFence,
          UINT64                      Value) {
    D3D10DeviceLock lock = LockContext();
    auto fence = static_cast<D3D11Fence*>(pFence);

    if (!fence)
      return E_INVALIDARG;

    ExecuteFlush(GpuFlushType::ExplicitFlush, nullptr);

    EmitCs([
      cFence = fence->GetFence(),
      cValue = Value
    ] (DxvkContext* ctx) {
      ctx->waitFence(cFence, cValue);
    });

    return S_OK;
  }


  void STDMETHODCALLTYPE D3D11ImmediateContext::ExecuteCommandList(
          ID3D11CommandList*  pCommandList,
          BOOL                RestoreContextState) {
    D3D10DeviceLock lock = LockContext();

    auto commandList = static_cast<D3D11CommandList*>(pCommandList);
    
    // Clear state so that the command list can't observe any
    // current context state. The command list itself will clean
    // up after execution to ensure that no state changes done
    // by the command list are visible to the immediate context.
    ResetCommandListState();

    // Flush any outstanding commands so that
    // we don't mess up the execution order
    FlushCsChunk();
    
    // As an optimization, flush everything if the
    // number of pending draw calls is high enough.
    ConsiderFlush(GpuFlushType::ImplicitWeakHint);

    // Dispatch command list to the CS thread
    commandList->EmitToCsThread([this] (DxvkCsChunkRef&& chunk, GpuFlushType flushType) {
      EmitCsChunk(std::move(chunk));

      // Return the sequence number from before the flush since
      // that is actually going to be needed for resource tracking
      uint64_t csSeqNum = m_csSeqNum;

      // Consider a flush after every chunk in case the app
      // submits a very large command list or the GPU is idle
      ConsiderFlush(flushType);
      return csSeqNum;
    });

    // Restore the immediate context's state
    if (RestoreContextState)
      RestoreCommandListState();
    else
      ResetContextState();
  }
  
  
  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::FinishCommandList(
          BOOL                RestoreDeferredContextState,
          ID3D11CommandList   **ppCommandList) {
    InitReturnPtr(ppCommandList);
    
    Logger::err("D3D11: FinishCommandList called on immediate context");
    return DXGI_ERROR_INVALID_CALL;
  }
  
  
  HRESULT STDMETHODCALLTYPE D3D11ImmediateContext::Map(
          ID3D11Resource*             pResource,
          UINT                        Subresource,
          D3D11_MAP                   MapType,
          UINT                        MapFlags,
          D3D11_MAPPED_SUBRESOURCE*   pMappedResource) {
    D3D10DeviceLock lock = LockContext();

    if (unlikely(!pResource))
      return E_INVALIDARG;
    
    D3D11_RESOURCE_DIMENSION resourceDim = D3D11_RESOURCE_DIMENSION_UNKNOWN;
    pResource->GetType(&resourceDim);

    HRESULT hr;
    
    if (likely(resourceDim == D3D11_RESOURCE_DIMENSION_BUFFER)) {
      hr = MapBuffer(
        static_cast<D3D11Buffer*>(pResource),
        MapType, MapFlags, pMappedResource);
    } else {
      hr = MapImage(
        GetCommonTexture(pResource),
        Subresource, MapType, MapFlags,
        pMappedResource);
    }

    if (unlikely(FAILED(hr)))
      *pMappedResource = D3D11_MAPPED_SUBRESOURCE();

    return hr;
  }
  
  
  void STDMETHODCALLTYPE D3D11ImmediateContext::Unmap(
          ID3D11Resource*             pResource,
          UINT                        Subresource) {
    // Since it is very uncommon for images to be mapped compared
    // to buffers, we count the currently mapped images in order
    // to avoid a virtual method call in the common case.
    if (unlikely(m_mappedImageCount > 0)) {
      D3D11_RESOURCE_DIMENSION resourceDim = D3D11_RESOURCE_DIMENSION_UNKNOWN;
      pResource->GetType(&resourceDim);

      if (resourceDim != D3D11_RESOURCE_DIMENSION_BUFFER) {
        D3D10DeviceLock lock = LockContext();
        UnmapImage(GetCommonTexture(pResource), Subresource);
      }
    }
  }


  HRESULT D3D11ImmediateContext::MapBuffer(
          D3D11Buffer*                pResource,
          D3D11_MAP                   MapType,
          UINT                        MapFlags,
          D3D11_MAPPED_SUBRESOURCE*   pMappedResource) {
    if (unlikely(!pMappedResource))
      return E_INVALIDARG;

    if (unlikely(pResource->GetMapMode() == D3D11_COMMON_BUFFER_MAP_MODE_NONE)) {
      Logger::err("D3D11: Cannot map a device-local buffer");
      return E_INVALIDARG;
    }

    VkDeviceSize bufferSize = pResource->Desc()->ByteWidth;

    if (likely(MapType == D3D11_MAP_WRITE_DISCARD)) {
      // Allocate a new backing slice for the buffer and set
      // it as the 'new' mapped slice. This assumes that the
      // only way to invalidate a buffer is by mapping it.
      auto physSlice = pResource->DiscardSlice();
      pMappedResource->pData      = physSlice.mapPtr;
      pMappedResource->RowPitch   = bufferSize;
      pMappedResource->DepthPitch = bufferSize;
      
      EmitCs([
        cBuffer      = pResource->GetBuffer(),
        cBufferSlice = physSlice
      ] (DxvkContext* ctx) {
        ctx->invalidateBuffer(cBuffer, cBufferSlice);
      });

      return S_OK;
    } else if (likely(MapType == D3D11_MAP_WRITE_NO_OVERWRITE)) {
      // Put this on a fast path without any extra checks since it's
      // a somewhat desired method to partially update large buffers
      DxvkBufferSliceHandle physSlice = pResource->GetMappedSlice();
      pMappedResource->pData      = physSlice.mapPtr;
      pMappedResource->RowPitch   = bufferSize;
      pMappedResource->DepthPitch = bufferSize;
      return S_OK;
    } else {
      // Quantum Break likes using MAP_WRITE on resources which would force
      // us to synchronize with the GPU multiple times per frame. In those
      // situations, if there are no pending GPU writes to the resource, we
      // can promote it to MAP_WRITE_DISCARD, but preserve the data by doing
      // a CPU copy from the previous buffer slice, to avoid the sync point.
      bool doInvalidatePreserve = false;

      auto buffer = pResource->GetBuffer();
      auto sequenceNumber = pResource->GetSequenceNumber();

      if (MapType != D3D11_MAP_READ && !MapFlags && bufferSize <= m_maxImplicitDiscardSize) {
        SynchronizeCsThread(sequenceNumber);

        bool hasWoAccess = buffer->isInUse(DxvkAccess::Write);
        bool hasRwAccess = buffer->isInUse(DxvkAccess::Read);

        if (hasRwAccess && !hasWoAccess) {
          // Uncached reads can be so slow that a GPU sync may actually be faster
          doInvalidatePreserve = buffer->memFlags() & VK_MEMORY_PROPERTY_HOST_CACHED_BIT;
        }
      }

      if (doInvalidatePreserve) {
        auto prevSlice = pResource->GetMappedSlice();
        auto physSlice = pResource->DiscardSlice();

        EmitCs([
          cBuffer      = std::move(buffer),
          cBufferSlice = physSlice
        ] (DxvkContext* ctx) {
          ctx->invalidateBuffer(cBuffer, cBufferSlice);
        });

        std::memcpy(physSlice.mapPtr, prevSlice.mapPtr, physSlice.length);
        pMappedResource->pData      = physSlice.mapPtr;
        pMappedResource->RowPitch   = bufferSize;
        pMappedResource->DepthPitch = bufferSize;
        return S_OK;
      } else {
        if (!WaitForResource(buffer, sequenceNumber, MapType, MapFlags))
          return DXGI_ERROR_WAS_STILL_DRAWING;

        DxvkBufferSliceHandle physSlice = pResource->GetMappedSlice();
        pMappedResource->pData      = physSlice.mapPtr;
        pMappedResource->RowPitch   = bufferSize;
        pMappedResource->DepthPitch = bufferSize;
        return S_OK;
      }
    }
  }
  
  
  HRESULT D3D11ImmediateContext::MapImage(
          D3D11CommonTexture*         pResource,
          UINT                        Subresource,
          D3D11_MAP                   MapType,
          UINT                        MapFlags,
          D3D11_MAPPED_SUBRESOURCE*   pMappedResource) {
    const Rc<DxvkImage>  mappedImage  = pResource->GetImage();
    const Rc<DxvkBuffer> mappedBuffer = pResource->GetMappedBuffer(Subresource);

    auto mapMode = pResource->GetMapMode();
    
    if (unlikely(mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_NONE)) {
      Logger::err("D3D11: Cannot map a device-local image");
      return E_INVALIDARG;
    }

    if (unlikely(Subresource >= pResource->CountSubresources()))
      return E_INVALIDARG;
    
    if (likely(pMappedResource != nullptr)) {
      // Resources with an unknown memory layout cannot return a pointer
      if (pResource->Desc()->Usage         == D3D11_USAGE_DEFAULT
       && pResource->Desc()->TextureLayout == D3D11_TEXTURE_LAYOUT_UNDEFINED)
        return E_INVALIDARG;
    } else {
      if (pResource->Desc()->Usage != D3D11_USAGE_DEFAULT)
        return E_INVALIDARG;
    }

    VkFormat packedFormat = m_parent->LookupPackedFormat(
      pResource->Desc()->Format, pResource->GetFormatMode()).Format;
    
    uint64_t sequenceNumber = pResource->GetSequenceNumber(Subresource);

    auto formatInfo = lookupFormatInfo(packedFormat);
    void* mapPtr;

    if (mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_DIRECT) {
      // Wait for the resource to become available. We do not
      // support image renaming, so stall on DISCARD instead.
      if (MapType == D3D11_MAP_WRITE_DISCARD)
        MapFlags &= ~D3D11_MAP_FLAG_DO_NOT_WAIT;

      if (MapType != D3D11_MAP_WRITE_NO_OVERWRITE) {
        if (!WaitForResource(mappedImage, sequenceNumber, MapType, MapFlags))
          return DXGI_ERROR_WAS_STILL_DRAWING;
      }
      
      // Query the subresource's memory layout and hope that
      // the application respects the returned pitch values.
      mapPtr = mappedImage->mapPtr(0);
    } else {
      constexpr uint32_t DoInvalidate = (1u << 0);
      constexpr uint32_t DoPreserve   = (1u << 1);
      constexpr uint32_t DoWait       = (1u << 2);
      uint32_t doFlags;

      if (mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_BUFFER) {
        // If the image can be written by the GPU, we need to update the
        // mapped staging buffer to reflect the current image contents.
        if (pResource->Desc()->Usage == D3D11_USAGE_DEFAULT) {
          bool needsReadback = !pResource->NeedsDirtyRegionTracking();

          needsReadback |= MapType == D3D11_MAP_READ
                        || MapType == D3D11_MAP_READ_WRITE;

          if (needsReadback)
            ReadbackImageBuffer(pResource, Subresource);
        }
      }

      if (MapType == D3D11_MAP_READ) {
        // Reads will not change the image content, so we only need
        // to wait for the GPU to finish writing to the mapped buffer.
        doFlags = DoWait;
      } else if (MapType == D3D11_MAP_WRITE_DISCARD) {
        doFlags = DoInvalidate;

        // If we know for sure that the mapped buffer is currently not
        // in use by the GPU, we don't have to allocate a new slice.
        if (m_csThread.lastSequenceNumber() >= sequenceNumber && !mappedBuffer->isInUse(DxvkAccess::Read))
          doFlags = 0;
      } else if (mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_STAGING && (MapFlags & D3D11_MAP_FLAG_DO_NOT_WAIT)) {
        // Always respect DO_NOT_WAIT for mapped staging images
        doFlags = DoWait;
      } else if (MapType != D3D11_MAP_WRITE_NO_OVERWRITE || mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_BUFFER) {
        // Need to synchronize thread to determine pending GPU accesses
        SynchronizeCsThread(sequenceNumber);

        // Don't implicitly discard large buffers or buffers of images with
        // multiple subresources, as that is likely to cause memory issues.
        VkDeviceSize bufferSize = pResource->GetMappedSlice(Subresource).length;

        if (bufferSize >= m_maxImplicitDiscardSize || pResource->CountSubresources() > 1) {
          // Don't check access flags, WaitForResource will return
          // early anyway if the resource is currently in use
          doFlags = DoWait;
        } else if (mappedBuffer->isInUse(DxvkAccess::Write)) {
          // There are pending GPU writes, need to wait for those
          doFlags = DoWait;
        } else if (mappedBuffer->isInUse(DxvkAccess::Read)) {
          // All pending GPU accesses are reads, so the buffer data
          // is still current, and we can prevent GPU synchronization
          // by creating a new slice with an exact copy of the data.
          doFlags = DoInvalidate | DoPreserve;
        } else {
          // There are no pending accesses, so we don't need to wait
          doFlags = 0;
        }
      } else {
        // No need to synchronize staging resources with NO_OVERWRITE
        // since the buffer will be used directly.
        doFlags = 0;
      }

      if (doFlags & DoInvalidate) {
        DxvkBufferSliceHandle prevSlice = pResource->GetMappedSlice(Subresource);
        DxvkBufferSliceHandle physSlice = pResource->DiscardSlice(Subresource);

        EmitCs([
          cImageBuffer = mappedBuffer,
          cBufferSlice = physSlice
        ] (DxvkContext* ctx) {
          ctx->invalidateBuffer(cImageBuffer, cBufferSlice);
        });

        if (doFlags & DoPreserve)
          std::memcpy(physSlice.mapPtr, prevSlice.mapPtr, physSlice.length);

        mapPtr = physSlice.mapPtr;
      } else {
        if (doFlags & DoWait) {
          // We cannot respect DO_NOT_WAIT for buffer-mapped resources since
          // our internal copies need to be transparent to the application.
          if (mapMode == D3D11_COMMON_TEXTURE_MAP_MODE_BUFFER)
            MapFlags &= ~D3D11_MAP_FLAG_DO_NOT_WAIT;

          // Wait for mapped buffer to become available
          if (!WaitForResource(mappedBuffer, sequenceNumber, MapType, MapFlags))
            return DXGI_ERROR_WAS_STILL_DRAWING;
        }

        mapPtr = pResource->GetMappedSlice(Subresource).mapPtr;
      }
    }

    // Mark the given subresource as mapped
    pResource->SetMapType(Subresource, MapType);

    if (pMappedResource) {
      auto layout = pResource->GetSubresourceLayout(formatInfo->aspectMask, Subresource);
      pMappedResource->pData      = reinterpret_cast<char*>(mapPtr) + layout.Offset;
      pMappedResource->RowPitch   = layout.RowPitch;
      pMappedResource->DepthPitch = layout.DepthPitch;
    }

    m_mappedImageCount += 1;
    return S_OK;
  }
  
  
  void D3D11ImmediateContext::UnmapImage(
          D3D11CommonTexture*         pResource,
          UINT                        Subresource) {
    D3D11_MAP mapType = pResource->GetMapType(Subresource);
    pResource->SetMapType(Subresource, D3D11_MAP(~0u));

    if (mapType == D3D11_MAP(~0u))
      return;

    // Decrement mapped image counter only after making sure
    // the given subresource is actually mapped right now
    m_mappedImageCount -= 1;

    if ((mapType != D3D11_MAP_READ) && (pResource->GetMapMode() == D3D11_COMMON_TEXTURE_MAP_MODE_BUFFER)) {
      if (pResource->NeedsDirtyRegionTracking()) {
        for (uint32_t i = 0; i < pResource->GetDirtyRegionCount(Subresource); i++) {
          D3D11_COMMON_TEXTURE_REGION region = pResource->GetDirtyRegion(Subresource, i);
          UpdateDirtyImageRegion(pResource, Subresource, &region);
        }

        pResource->ClearDirtyRegions(Subresource);
      } else {
        UpdateDirtyImageRegion(pResource, Subresource, nullptr);
      }
    }
  }
  
  
  void D3D11ImmediateContext::ReadbackImageBuffer(
          D3D11CommonTexture*         pResource,
          UINT                        Subresource) {
    VkImageAspectFlags aspectMask = lookupFormatInfo(pResource->GetPackedFormat())->aspectMask;
    VkImageSubresource subresource = pResource->GetSubresourceFromIndex(aspectMask, Subresource);

    EmitCs([
      cSrcImage           = pResource->GetImage(),
      cSrcSubresource     = vk::makeSubresourceLayers(subresource),
      cDstBuffer          = pResource->GetMappedBuffer(Subresource),
      cPackedFormat       = pResource->GetPackedFormat()
    ] (DxvkContext* ctx) {
      VkOffset3D offset = { 0, 0, 0 };
      VkExtent3D extent = cSrcImage->mipLevelExtent(cSrcSubresource.mipLevel);

      if (cSrcSubresource.aspectMask != (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
        ctx->copyImageToBuffer(cDstBuffer, 0, 0, 0,
          cSrcImage, cSrcSubresource, offset, extent);
      } else {
        ctx->copyDepthStencilImageToPackedBuffer(cDstBuffer, 0,
          VkOffset2D { 0, 0 },
          VkExtent2D { extent.width, extent.height },
          cSrcImage, cSrcSubresource,
          VkOffset2D { 0, 0 },
          VkExtent2D { extent.width, extent.height },
          cPackedFormat);
      }
    });

    if (pResource->HasSequenceNumber())
      TrackTextureSequenceNumber(pResource, Subresource);
  }


  void D3D11ImmediateContext::UpdateDirtyImageRegion(
          D3D11CommonTexture*         pResource,
          UINT                        Subresource,
    const D3D11_COMMON_TEXTURE_REGION* pRegion) {
    auto formatInfo = lookupFormatInfo(pResource->GetPackedFormat());
    auto subresource = vk::makeSubresourceLayers(
      pResource->GetSubresourceFromIndex(formatInfo->aspectMask, Subresource));

    // Update the entire image if no dirty region was specified
    D3D11_COMMON_TEXTURE_REGION region;

    if (pRegion) {
      region = *pRegion;
    } else {
      region.Offset = VkOffset3D { 0, 0, 0 };
      region.Extent = pResource->MipLevelExtent(subresource.mipLevel);
    }

    auto subresourceLayout = pResource->GetSubresourceLayout(formatInfo->aspectMask, Subresource);

    // Update dirty region one aspect at a time, due to
    // how the data is laid out in the staging buffer.
    for (uint32_t i = 0; i < pResource->GetPlaneCount(); i++) {
      subresource.aspectMask = formatInfo->aspectMask;

      if (formatInfo->flags.test(DxvkFormatFlag::MultiPlane))
        subresource.aspectMask = vk::getPlaneAspect(i);

      EmitCs([
        cDstImage       = pResource->GetImage(),
        cDstSubresource = subresource,
        cDstOffset      = region.Offset,
        cDstExtent      = region.Extent,
        cSrcBuffer      = pResource->GetMappedBuffer(Subresource),
        cSrcOffset      = pResource->ComputeMappedOffset(Subresource, i, region.Offset),
        cSrcRowPitch    = subresourceLayout.RowPitch,
        cSrcDepthPitch  = subresourceLayout.DepthPitch,
        cPackedFormat   = pResource->GetPackedFormat()
      ] (DxvkContext* ctx) {
        if (cDstSubresource.aspectMask != (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
          ctx->copyBufferToImage(
            cDstImage, cDstSubresource, cDstOffset, cDstExtent,
            cSrcBuffer, cSrcOffset, cSrcRowPitch, cSrcDepthPitch);
        } else {
          ctx->copyPackedBufferToDepthStencilImage(
            cDstImage, cDstSubresource,
            VkOffset2D { cDstOffset.x, cDstOffset.y },
            VkExtent2D { cDstExtent.width, cDstExtent.height },
            cSrcBuffer, 0,
            VkOffset2D { cDstOffset.x, cDstOffset.y },
            VkExtent2D { cDstExtent.width, cDstExtent.height },
            cPackedFormat);
        }
      });
    }

    if (pResource->HasSequenceNumber())
      TrackTextureSequenceNumber(pResource, Subresource);
  }


  void D3D11ImmediateContext::UpdateMappedBuffer(
          D3D11Buffer*                  pDstBuffer,
          UINT                          Offset,
          UINT                          Length,
    const void*                         pSrcData,
          UINT                          CopyFlags) {
    DxvkBufferSliceHandle slice;

    if (likely(CopyFlags != D3D11_COPY_NO_OVERWRITE)) {
      slice = pDstBuffer->DiscardSlice();

      EmitCs([
        cBuffer      = pDstBuffer->GetBuffer(),
        cBufferSlice = slice
      ] (DxvkContext* ctx) {
        ctx->invalidateBuffer(cBuffer, cBufferSlice);
      });
    } else {
      slice = pDstBuffer->GetMappedSlice();
    }

    std::memcpy(reinterpret_cast<char*>(slice.mapPtr) + Offset, pSrcData, Length);
  }


  void STDMETHODCALLTYPE D3D11ImmediateContext::SwapDeviceContextState(
          ID3DDeviceContextState*           pState,
          ID3DDeviceContextState**          ppPreviousState) {
    InitReturnPtr(ppPreviousState);

    if (!pState)
      return;

    // Reset all state affected by the current context state
    ResetCommandListState();

    Com<D3D11DeviceContextState, false> oldState = std::move(m_stateObject);
    Com<D3D11DeviceContextState, false> newState = static_cast<D3D11DeviceContextState*>(pState);

    if (oldState == nullptr)
      oldState = new D3D11DeviceContextState(m_parent);
    
    if (ppPreviousState)
      *ppPreviousState = oldState.ref();
    
    m_stateObject = newState;

    oldState->SetState(m_state);
    newState->GetState(m_state);

    // Restore all state affected by the new context state
    RestoreCommandListState();
  }


  void D3D11ImmediateContext::SynchronizeCsThread(uint64_t SequenceNumber) {
    D3D10DeviceLock lock = LockContext();

    // Dispatch current chunk so that all commands
    // recorded prior to this function will be run
    if (SequenceNumber > m_csSeqNum)
      FlushCsChunk();
    
    m_csThread.synchronize(SequenceNumber);
  }
  
  
  void D3D11ImmediateContext::SynchronizeDevice() {
    m_device->waitForIdle();
  }
  
  
  void D3D11ImmediateContext::EndFrame() {
    D3D10DeviceLock lock = LockContext();

    EmitCs<false>([] (DxvkContext* ctx) {
      ctx->endFrame();
    });
  }


  bool D3D11ImmediateContext::WaitForResource(
    const Rc<DxvkResource>&                 Resource,
          uint64_t                          SequenceNumber,
          D3D11_MAP                         MapType,
          UINT                              MapFlags) {
    // Determine access type to wait for based on map mode
    DxvkAccess access = MapType == D3D11_MAP_READ
      ? DxvkAccess::Write
      : DxvkAccess::Read;
    
    // Wait for any CS chunk using the resource to execute, since
    // otherwise we cannot accurately determine if the resource is
    // actually being used by the GPU right now.
    bool isInUse = Resource->isInUse(access);

    if (!isInUse) {
      SynchronizeCsThread(SequenceNumber);
      isInUse = Resource->isInUse(access);
    }

    if (MapFlags & D3D11_MAP_FLAG_DO_NOT_WAIT) {
      if (isInUse) {
        // We don't have to wait, but misbehaving games may
        // still try to spin on `Map` until the resource is
        // idle, so we should flush pending commands
        ConsiderFlush(GpuFlushType::ImplicitSynchronization);
        return false;
      }
    } else {
      if (isInUse) {
        // Make sure pending commands using the resource get
        // executed on the the GPU if we have to wait for it
        ExecuteFlush(GpuFlushType::ImplicitSynchronization, nullptr);
        SynchronizeCsThread(SequenceNumber);

        m_device->waitForResource(Resource, access);
      }
    }

    return true;
  }
  
  
  void D3D11ImmediateContext::EmitCsChunk(DxvkCsChunkRef&& chunk) {
    m_csSeqNum = m_csThread.dispatchChunk(std::move(chunk));
  }


  void D3D11ImmediateContext::TrackTextureSequenceNumber(
          D3D11CommonTexture*         pResource,
          UINT                        Subresource) {
    uint64_t sequenceNumber = GetCurrentSequenceNumber();
    pResource->TrackSequenceNumber(Subresource, sequenceNumber);

    ConsiderFlush(GpuFlushType::ImplicitStrongHint);
  }


  void D3D11ImmediateContext::TrackBufferSequenceNumber(
          D3D11Buffer*                pResource) {
    uint64_t sequenceNumber = GetCurrentSequenceNumber();
    pResource->TrackSequenceNumber(sequenceNumber);

    ConsiderFlush(GpuFlushType::ImplicitStrongHint);
  }


  uint64_t D3D11ImmediateContext::GetCurrentSequenceNumber() {
    // We do not flush empty chunks, so if we are tracking a resource
    // immediately after a flush, we need to use the sequence number
    // of the previously submitted chunk to prevent deadlocks.
    return m_csChunk->empty() ? m_csSeqNum : m_csSeqNum + 1;
  }


  uint64_t D3D11ImmediateContext::GetPendingCsChunks() {
    return GetCurrentSequenceNumber() - m_flushSeqNum;
  }


  void D3D11ImmediateContext::ConsiderFlush(
          GpuFlushType                FlushType) {
    uint64_t chunkId = GetCurrentSequenceNumber();
    uint64_t submissionId = m_submissionFence->value();

    if (m_flushTracker.considerFlush(FlushType, chunkId, submissionId))
      ExecuteFlush(FlushType, nullptr);
  }


  void D3D11ImmediateContext::ExecuteFlush(
          GpuFlushType                FlushType,
          HANDLE                      hEvent) {
    // Flush init context so that new resources are fully initialized
    // before the app can access them in any way. This has to happen
    // unconditionally since we may otherwise deadlock on Map.
    m_parent->FlushInitContext();

    // Exit early if there's nothing to do
    if (!GetPendingCsChunks() && !hEvent)
      return;

    // Signal the submission fence and flush the command list
    uint64_t submissionId = ++m_submissionId;

    if (hEvent) {
      m_submissionFence->setCallback(submissionId, [hEvent] {
        SetEvent(hEvent);
      });
    }

    EmitCs<false>([
      cSubmissionFence  = m_submissionFence,
      cSubmissionId     = submissionId
    ] (DxvkContext* ctx) {
      ctx->signal(cSubmissionFence, cSubmissionId);
      ctx->flushCommandList(nullptr);
    });

    FlushCsChunk();

    // Notify flush tracker about the flush
    m_flushSeqNum = m_csSeqNum;
    m_flushTracker.notifyFlush(m_flushSeqNum, submissionId);
  }
  
}