BansheeEngine/BansheeD3D11RenderAPI/Source/BsD3D11RenderAPI.cpp

//********************************** Banshee Engine (www.banshee3d.com) **************************************************//
//**************** Copyright (c) 2016 Marko Pintera ([email protected]). All rights reserved. **********************//
#include "BsD3D11RenderAPI.h"
#include "BsD3D11DriverList.h"
#include "BsD3D11Driver.h"
#include "BsD3D11Device.h"
#include "BsD3D11TextureManager.h"
#include "BsD3D11Texture.h"
#include "BsD3D11HardwareBufferManager.h"
#include "BsD3D11RenderWindowManager.h"
#include "BsD3D11HLSLProgramFactory.h"
#include "BsD3D11BlendState.h"
#include "BsD3D11RasterizerState.h"
#include "BsD3D11DepthStencilState.h"
#include "BsD3D11SamplerState.h"
#include "BsD3D11GpuProgram.h"
#include "BsD3D11Mappings.h"
#include "BsD3D11VertexBuffer.h"
#include "BsD3D11IndexBuffer.h"
#include "BsD3D11RenderStateManager.h"
#include "BsD3D11GpuParamBlockBuffer.h"
#include "BsD3D11InputLayoutManager.h"
#include "BsD3D11TextureView.h"
#include "BsD3D11RenderUtility.h"
#include "BsGpuParams.h"
#include "BsCoreThread.h"
#include "BsD3D11QueryManager.h"
#include "BsDebug.h"
#include "BsException.h"
#include "BsRenderStats.h"
#include "BsGpuParamDesc.h"

namespace BansheeEngine
{
	D3D11RenderAPI::D3D11RenderAPI()
		: mDXGIFactory(nullptr), mDevice(nullptr), mDriverList(nullptr)
		, mActiveD3DDriver(nullptr), mFeatureLevel(D3D_FEATURE_LEVEL_11_0)
		, mHLSLFactory(nullptr), mIAManager(nullptr)
		, mStencilRef(0), mActiveDrawOp(DOT_TRIANGLE_LIST)
		, mViewportNorm(0.0f, 0.0f, 1.0f, 1.0f)
	{
		mClipPlanesDirty = false; // DX11 handles clip planes through shaders
	}

	D3D11RenderAPI::~D3D11RenderAPI()
	{

	}

	const StringID& D3D11RenderAPI::getName() const
	{
		static StringID strName("D3D11RenderAPI");
		return strName;
	}

	const String& D3D11RenderAPI::getShadingLanguageName() const
	{
		static String strName("hlsl");
		return strName;
	}

	void D3D11RenderAPI::initializePrepare()
	{
		THROW_IF_NOT_CORE_THREAD;

		HRESULT hr = CreateDXGIFactory(__uuidof(IDXGIFactory), (void**)&mDXGIFactory);
		if(FAILED(hr))
			BS_EXCEPT(RenderingAPIException, "Failed to create Direct3D11 DXGIFactory");

		mDriverList = bs_new<D3D11DriverList>(mDXGIFactory);
		mActiveD3DDriver = mDriverList->item(0); // TODO: Always get first driver, for now
		mVideoModeInfo = mActiveD3DDriver->getVideoModeInfo();

		IDXGIAdapter* selectedAdapter = mActiveD3DDriver->getDeviceAdapter();

		D3D_FEATURE_LEVEL requestedLevels[] = {
			D3D_FEATURE_LEVEL_11_0,
			D3D_FEATURE_LEVEL_10_1,
			D3D_FEATURE_LEVEL_10_0,
			D3D_FEATURE_LEVEL_9_3,
			D3D_FEATURE_LEVEL_9_2,
			D3D_FEATURE_LEVEL_9_1
		};

		UINT32 numRequestedLevel = sizeof(requestedLevels) / sizeof(requestedLevels[0]);

		UINT32 deviceFlags = 0;

#if BS_DEBUG_MODE
		deviceFlags |= D3D11_CREATE_DEVICE_DEBUG;
#endif

		ID3D11Device* device;
		hr = D3D11CreateDevice(selectedAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr, deviceFlags, 
			requestedLevels, numRequestedLevel, D3D11_SDK_VERSION, &device, &mFeatureLevel, nullptr);

		if(FAILED(hr))         
			BS_EXCEPT(RenderingAPIException, "Failed to create Direct3D11 object. D3D11CreateDeviceN returned this error code: " + toString(hr));

		mDevice = bs_new<D3D11Device>(device);
		
		// This must query for DirectX 10 interface as this is unsupported for DX11
		LARGE_INTEGER driverVersion; 
		if(SUCCEEDED(selectedAdapter->CheckInterfaceSupport(IID_ID3D10Device, &driverVersion)))
		{
			mDriverVersion.major =  HIWORD(driverVersion.HighPart);
			mDriverVersion.minor = LOWORD(driverVersion.HighPart);
			mDriverVersion.release = HIWORD(driverVersion.LowPart);
			mDriverVersion.build = LOWORD(driverVersion.LowPart);
		}

		// Create the texture manager for use by others		
		TextureManager::startUp<D3D11TextureManager>();
		TextureCoreManager::startUp<D3D11TextureCoreManager>();

		// Create hardware buffer manager		
		HardwareBufferManager::startUp();
		HardwareBufferCoreManager::startUp<D3D11HardwareBufferCoreManager>(std::ref(*mDevice));

		// Create render window manager
		RenderWindowManager::startUp<D3D11RenderWindowManager>(this);
		RenderWindowCoreManager::startUp<D3D11RenderWindowCoreManager>(this);

		// Create & register HLSL factory		
		mHLSLFactory = bs_new<D3D11HLSLProgramFactory>();

		// Create render state manager
		RenderStateCoreManager::startUp<D3D11RenderStateCoreManager>();

		mCurrentCapabilities = createRenderSystemCapabilities();

		mCurrentCapabilities->addShaderProfile("hlsl");
		GpuProgramCoreManager::instance().addFactory(mHLSLFactory);

		mIAManager = bs_new<D3D11InputLayoutManager>();

		RenderAPICore::initializePrepare();
	}

	void D3D11RenderAPI::initializeFinalize(const SPtr<RenderWindowCore>& primaryWindow)
	{
		D3D11RenderUtility::startUp(mDevice);

		QueryManager::startUp<D3D11QueryManager>();

		RenderAPICore::initializeFinalize(primaryWindow);
	}

    void D3D11RenderAPI::destroyCore()
	{
		THROW_IF_NOT_CORE_THREAD;

		for (auto& boundUAV : mBoundUAVs)
		{
			if (boundUAV.second != nullptr)
				boundUAV.first->releaseView(boundUAV.second);
		}

		QueryManager::shutDown();
		D3D11RenderUtility::shutDown();

		if(mIAManager != nullptr)
		{
			bs_delete(mIAManager);
			mIAManager = nullptr;
		}

		if(mHLSLFactory != nullptr)
		{
			bs_delete(mHLSLFactory);
			mHLSLFactory = nullptr;
		}

		mActiveVertexDeclaration = nullptr;
		mActiveVertexShader = nullptr;
		mActiveRenderTarget = nullptr;

		RenderStateCoreManager::shutDown();
		RenderWindowCoreManager::shutDown();
		RenderWindowManager::shutDown();
		HardwareBufferCoreManager::shutDown();
		HardwareBufferManager::shutDown();
		TextureCoreManager::shutDown();
		TextureManager::shutDown();

		SAFE_RELEASE(mDXGIFactory);

		if(mDevice != nullptr)
		{
			bs_delete(mDevice);
			mDevice = nullptr;
		}

		if(mDriverList != nullptr)
		{
			bs_delete(mDriverList);
			mDriverList = nullptr;
		}

		mActiveD3DDriver = nullptr;

		RenderAPICore::destroyCore();
	}

	void D3D11RenderAPI::setSamplerState(GpuProgramType gptype, UINT16 texUnit, const SPtr<SamplerStateCore>& samplerState)
	{
		THROW_IF_NOT_CORE_THREAD;

		// TODO - I'm setting up views one by one, it might be more efficient to hold them in an array
		//  and then set them all up at once before rendering? Needs testing

		ID3D11SamplerState* samplerArray[1];
		D3D11SamplerStateCore* d3d11SamplerState = static_cast<D3D11SamplerStateCore*>(const_cast<SamplerStateCore*>(samplerState.get()));
		samplerArray[0] = d3d11SamplerState->getInternal();

		switch(gptype)
		{
		case GPT_VERTEX_PROGRAM:
			mDevice->getImmediateContext()->VSSetSamplers(texUnit, 1, samplerArray);
			break;
		case GPT_FRAGMENT_PROGRAM:
			mDevice->getImmediateContext()->PSSetSamplers(texUnit, 1, samplerArray);
			break;
		case GPT_GEOMETRY_PROGRAM:
			mDevice->getImmediateContext()->GSSetSamplers(texUnit, 1, samplerArray);
			break;
		case GPT_DOMAIN_PROGRAM:
			mDevice->getImmediateContext()->DSSetSamplers(texUnit, 1, samplerArray);
			break;
		case GPT_HULL_PROGRAM:
			mDevice->getImmediateContext()->HSSetSamplers(texUnit, 1, samplerArray);
			break;
		case GPT_COMPUTE_PROGRAM:
			mDevice->getImmediateContext()->CSSetSamplers(texUnit, 1, samplerArray);
			break;
		default:
			BS_EXCEPT(InvalidParametersException, "Unsupported gpu program type: " + toString(gptype));
		}

		BS_INC_RENDER_STAT(NumSamplerBinds);
	}

	void D3D11RenderAPI::setBlendState(const SPtr<BlendStateCore>& blendState)
	{
		THROW_IF_NOT_CORE_THREAD;

		D3D11BlendStateCore* d3d11BlendState = static_cast<D3D11BlendStateCore*>(const_cast<BlendStateCore*>(blendState.get()));
		mDevice->getImmediateContext()->OMSetBlendState(d3d11BlendState->getInternal(), nullptr, 0xFFFFFFFF);

		BS_INC_RENDER_STAT(NumBlendStateChanges);
	}

	void D3D11RenderAPI::setRasterizerState(const SPtr<RasterizerStateCore>& rasterizerState)
	{
		THROW_IF_NOT_CORE_THREAD;

		D3D11RasterizerStateCore* d3d11RasterizerState = static_cast<D3D11RasterizerStateCore*>(const_cast<RasterizerStateCore*>(rasterizerState.get()));
		mDevice->getImmediateContext()->RSSetState(d3d11RasterizerState->getInternal());

		BS_INC_RENDER_STAT(NumRasterizerStateChanges);
	}

	void D3D11RenderAPI::setDepthStencilState(const SPtr<DepthStencilStateCore>& depthStencilState, UINT32 stencilRefValue)
	{
		THROW_IF_NOT_CORE_THREAD;

		D3D11DepthStencilStateCore* d3d11RasterizerState = static_cast<D3D11DepthStencilStateCore*>(const_cast<DepthStencilStateCore*>(depthStencilState.get()));
		mDevice->getImmediateContext()->OMSetDepthStencilState(d3d11RasterizerState->getInternal(), stencilRefValue);

		BS_INC_RENDER_STAT(NumDepthStencilStateChanges);
	}

	void D3D11RenderAPI::setTexture(GpuProgramType gptype, UINT16 unit, bool enabled, const SPtr<TextureCore>& texPtr)
	{
		THROW_IF_NOT_CORE_THREAD;

		// TODO - I'm setting up views one by one, it might be more efficient to hold them in an array
		//  and then set them all up at once before rendering? Needs testing

		ID3D11ShaderResourceView* viewArray[1];
		if(texPtr != nullptr && enabled)
		{
			D3D11TextureCore* d3d11Texture = static_cast<D3D11TextureCore*>(texPtr.get());
			viewArray[0] = d3d11Texture->getSRV();
		}
		else
			viewArray[0] = nullptr;

		switch(gptype)
		{
		case GPT_VERTEX_PROGRAM:
			mDevice->getImmediateContext()->VSSetShaderResources(unit, 1, viewArray);
			break;
		case GPT_FRAGMENT_PROGRAM:
			mDevice->getImmediateContext()->PSSetShaderResources(unit, 1, viewArray);
			break;
		case GPT_GEOMETRY_PROGRAM:
			mDevice->getImmediateContext()->GSSetShaderResources(unit, 1, viewArray);
			break;
		case GPT_DOMAIN_PROGRAM:
			mDevice->getImmediateContext()->DSSetShaderResources(unit, 1, viewArray);
			break;
		case GPT_HULL_PROGRAM:
			mDevice->getImmediateContext()->HSSetShaderResources(unit, 1, viewArray);
			break;
		case GPT_COMPUTE_PROGRAM:
			mDevice->getImmediateContext()->CSSetShaderResources(unit, 1, viewArray);
			break;
		default:
			BS_EXCEPT(InvalidParametersException, "Unsupported gpu program type: " + toString(gptype));
		}

		BS_INC_RENDER_STAT(NumTextureBinds);
	}

	void D3D11RenderAPI::setLoadStoreTexture(GpuProgramType gptype, UINT16 unit, bool enabled, const SPtr<TextureCore>& texPtr,
		const TextureSurface& surface)
	{
		THROW_IF_NOT_CORE_THREAD;

		// TODO - This hasn't bee tested and might be incorrect. I might need to set UAVs together with render targets,
		// especially considering DX11 expects number of UAVs to match number of render targets.

		ID3D11UnorderedAccessView* viewArray[1];
		if (texPtr != nullptr && enabled)
		{
			D3D11TextureCore* d3d11Texture = static_cast<D3D11TextureCore*>(texPtr.get());
			TextureViewPtr texView = TextureCore::requestView(texPtr, surface.mipLevel, 1, 
				surface.arraySlice, surface.numArraySlices, GVU_RANDOMWRITE);

			D3D11TextureView* d3d11texView = static_cast<D3D11TextureView*>(texView.get());
			viewArray[0] = d3d11texView->getUAV();

			if (mBoundUAVs[unit].second != nullptr)
				mBoundUAVs[unit].first->releaseView(mBoundUAVs[unit].second);

			mBoundUAVs[unit] = std::make_pair(texPtr, texView);
		}
		else
		{
			viewArray[0] = nullptr;

			if (mBoundUAVs[unit].second != nullptr)
				mBoundUAVs[unit].first->releaseView(mBoundUAVs[unit].second);

			mBoundUAVs[unit] = std::pair<SPtr<TextureCore>, TextureViewPtr>();
		}

		if (gptype == GPT_FRAGMENT_PROGRAM)
		{
			mDevice->getImmediateContext()->OMSetRenderTargetsAndUnorderedAccessViews(
				D3D11_KEEP_RENDER_TARGETS_AND_DEPTH_STENCIL, nullptr, nullptr, unit, 1, viewArray, nullptr);
		}
		else if (gptype == GPT_COMPUTE_PROGRAM)
		{
			mDevice->getImmediateContext()->CSSetUnorderedAccessViews(unit, 1, viewArray, nullptr);
		}
		else
			BS_EXCEPT(InvalidParametersException, "Unsupported gpu program type: " + toString(gptype));

		BS_INC_RENDER_STAT(NumTextureBinds);
	}

	void D3D11RenderAPI::disableTextureUnit(GpuProgramType gptype, UINT16 texUnit)
	{
		THROW_IF_NOT_CORE_THREAD;

		setTexture(gptype, texUnit, false, nullptr);
	}

	void D3D11RenderAPI::beginFrame()
	{
		// Not used
	}

	void D3D11RenderAPI::endFrame()
	{
		// Not used
	}

	void D3D11RenderAPI::setViewport(const Rect2& vp)
	{
		THROW_IF_NOT_CORE_THREAD;

		mViewportNorm = vp;
		applyViewport();
	}

	void D3D11RenderAPI::applyViewport()
	{
		if (mActiveRenderTarget == nullptr)
			return;

		const RenderTargetProperties& rtProps = mActiveRenderTarget->getProperties();

		// Set viewport dimensions
		mViewport.TopLeftX = (FLOAT)(rtProps.getWidth() * mViewportNorm.x);
		mViewport.TopLeftY = (FLOAT)(rtProps.getHeight() * mViewportNorm.y);
		mViewport.Width = (FLOAT)(rtProps.getWidth() * mViewportNorm.width);
		mViewport.Height = (FLOAT)(rtProps.getHeight() * mViewportNorm.height);

		if (rtProps.requiresTextureFlipping())
		{
			// Convert "top-left" to "bottom-left"
			mViewport.TopLeftY = rtProps.getHeight() - mViewport.Height - mViewport.TopLeftY;
		}

		mViewport.MinDepth = 0.0f;
		mViewport.MaxDepth = 1.0f;

		mDevice->getImmediateContext()->RSSetViewports(1, &mViewport);
	}

	void D3D11RenderAPI::setVertexBuffers(UINT32 index, SPtr<VertexBufferCore>* buffers, UINT32 numBuffers)
	{
		THROW_IF_NOT_CORE_THREAD;

		UINT32 maxBoundVertexBuffers = mCurrentCapabilities->getMaxBoundVertexBuffers();
		if(index < 0 || (index + numBuffers) >= maxBoundVertexBuffers)
			BS_EXCEPT(InvalidParametersException, "Invalid vertex index: " + toString(index) + ". Valid range is 0 .. " + toString(maxBoundVertexBuffers - 1));

		ID3D11Buffer* dx11buffers[MAX_BOUND_VERTEX_BUFFERS];
		UINT32 strides[MAX_BOUND_VERTEX_BUFFERS];
		UINT32 offsets[MAX_BOUND_VERTEX_BUFFERS];

		for(UINT32 i = 0; i < numBuffers; i++)
		{
			SPtr<D3D11VertexBufferCore> vertexBuffer = std::static_pointer_cast<D3D11VertexBufferCore>(buffers[i]);
			const VertexBufferProperties& vbProps = vertexBuffer->getProperties();

			dx11buffers[i] = vertexBuffer->getD3DVertexBuffer();

			strides[i] = vbProps.getVertexSize();
			offsets[i] = 0;
		}

		mDevice->getImmediateContext()->IASetVertexBuffers(index, numBuffers, dx11buffers, strides, offsets);

		BS_INC_RENDER_STAT(NumVertexBufferBinds);
	}

	void D3D11RenderAPI::setIndexBuffer(const SPtr<IndexBufferCore>& buffer)
	{
		THROW_IF_NOT_CORE_THREAD;

		SPtr<D3D11IndexBufferCore> indexBuffer = std::static_pointer_cast<D3D11IndexBufferCore>(buffer);

		DXGI_FORMAT indexFormat = DXGI_FORMAT_R16_UINT;
		if(indexBuffer->getProperties().getType() == IT_16BIT)
			indexFormat = DXGI_FORMAT_R16_UINT;
		else if (indexBuffer->getProperties().getType() == IT_32BIT)
			indexFormat = DXGI_FORMAT_R32_UINT;
		else
			BS_EXCEPT(InternalErrorException, "Unsupported index format: " + toString(indexBuffer->getProperties().getType()));

		mDevice->getImmediateContext()->IASetIndexBuffer(indexBuffer->getD3DIndexBuffer(), indexFormat, 0);

		BS_INC_RENDER_STAT(NumIndexBufferBinds);
	}

	void D3D11RenderAPI::setVertexDeclaration(const SPtr<VertexDeclarationCore>& vertexDeclaration)
	{
		THROW_IF_NOT_CORE_THREAD;

		mActiveVertexDeclaration = vertexDeclaration;
	}

	void D3D11RenderAPI::setDrawOperation(DrawOperationType op)
	{
		THROW_IF_NOT_CORE_THREAD;

		mDevice->getImmediateContext()->IASetPrimitiveTopology(D3D11Mappings::getPrimitiveType(op));
	}

	void D3D11RenderAPI::bindGpuProgram(const SPtr<GpuProgramCore>& prg)
	{
		THROW_IF_NOT_CORE_THREAD;

		switch(prg->getProperties().getType())
		{
		case GPT_VERTEX_PROGRAM:
			{
				D3D11GpuVertexProgramCore* d3d11GpuProgram = static_cast<D3D11GpuVertexProgramCore*>(prg.get());
				mDevice->getImmediateContext()->VSSetShader(d3d11GpuProgram->getVertexShader(), nullptr, 0);
				mActiveVertexShader = std::static_pointer_cast<D3D11GpuProgramCore>(prg);
				break;
			}
		case GPT_FRAGMENT_PROGRAM:
			{
				D3D11GpuFragmentProgramCore* d3d11GpuProgram = static_cast<D3D11GpuFragmentProgramCore*>(prg.get());
				mDevice->getImmediateContext()->PSSetShader(d3d11GpuProgram->getPixelShader(), nullptr, 0);
				break;
			}
		case GPT_GEOMETRY_PROGRAM:
			{
				D3D11GpuGeometryProgramCore* d3d11GpuProgram = static_cast<D3D11GpuGeometryProgramCore*>(prg.get());
				mDevice->getImmediateContext()->GSSetShader(d3d11GpuProgram->getGeometryShader(), nullptr, 0);
				break;
			}
		case GPT_DOMAIN_PROGRAM:
			{
				D3D11GpuDomainProgramCore* d3d11GpuProgram = static_cast<D3D11GpuDomainProgramCore*>(prg.get());
				mDevice->getImmediateContext()->DSSetShader(d3d11GpuProgram->getDomainShader(), nullptr, 0);
				break;
			}
		case GPT_HULL_PROGRAM:
			{
				D3D11GpuHullProgramCore* d3d11GpuProgram = static_cast<D3D11GpuHullProgramCore*>(prg.get());
				mDevice->getImmediateContext()->HSSetShader(d3d11GpuProgram->getHullShader(), nullptr, 0);
				break;
			}
		case GPT_COMPUTE_PROGRAM:
			{
				D3D11GpuComputeProgramCore* d3d11GpuProgram = static_cast<D3D11GpuComputeProgramCore*>(prg.get());
				mDevice->getImmediateContext()->CSSetShader(d3d11GpuProgram->getComputeShader(), nullptr, 0);
				break;
			}
		default:
			BS_EXCEPT(InvalidParametersException, "Unsupported gpu program type: " + toString(prg->getProperties().getType()));
		}

		if (mDevice->hasError())
			BS_EXCEPT(RenderingAPIException, "Failed to bindGpuProgram : " + mDevice->getErrorDescription());

		BS_INC_RENDER_STAT(NumGpuProgramBinds);
	}

	void D3D11RenderAPI::unbindGpuProgram(GpuProgramType gptype)
	{
		THROW_IF_NOT_CORE_THREAD;

		switch(gptype)
		{
		case GPT_VERTEX_PROGRAM:
			mDevice->getImmediateContext()->VSSetShader(nullptr, nullptr, 0);
			mActiveVertexShader = nullptr;
			break;
		case GPT_FRAGMENT_PROGRAM:
			mDevice->getImmediateContext()->PSSetShader(nullptr, nullptr, 0);
			break;
		case GPT_GEOMETRY_PROGRAM:
			mDevice->getImmediateContext()->GSSetShader(nullptr, nullptr, 0);
			break;
		case GPT_DOMAIN_PROGRAM:
			mDevice->getImmediateContext()->DSSetShader(nullptr, nullptr, 0);
			break;
		case GPT_HULL_PROGRAM:
			mDevice->getImmediateContext()->HSSetShader(nullptr, nullptr, 0);
			break;
		case GPT_COMPUTE_PROGRAM:
			mDevice->getImmediateContext()->CSSetShader(nullptr, nullptr, 0);
			break;
		default:
			BS_EXCEPT(InvalidParametersException, "Unsupported gpu program type: " + toString(gptype));
		}

		BS_INC_RENDER_STAT(NumGpuProgramBinds);
	}

	void D3D11RenderAPI::setConstantBuffers(GpuProgramType gptype, const SPtr<GpuParamsCore>& bindableParams)
	{
		THROW_IF_NOT_CORE_THREAD;

		bindableParams->updateHardwareBuffers();
		const GpuParamDesc& paramDesc = bindableParams->getParamDesc();

		// TODO - I assign constant buffers one by one but it might be more efficient to do them all at once?
		ID3D11Buffer* bufferArray[1];
		for(auto iter = paramDesc.paramBlocks.begin(); iter != paramDesc.paramBlocks.end(); ++iter)
		{
			SPtr<GpuParamBlockBufferCore> currentBlockBuffer = bindableParams->getParamBlockBuffer(iter->second.slot);

			if(currentBlockBuffer != nullptr)
			{
				const D3D11GpuParamBlockBufferCore* d3d11paramBlockBuffer = 
					static_cast<const D3D11GpuParamBlockBufferCore*>(currentBlockBuffer.get());
				bufferArray[0] = d3d11paramBlockBuffer->getD3D11Buffer();
			}
			else
				bufferArray[0] = nullptr;

			switch(gptype)
			{
			case GPT_VERTEX_PROGRAM:
				mDevice->getImmediateContext()->VSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			case GPT_FRAGMENT_PROGRAM:
				mDevice->getImmediateContext()->PSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			case GPT_GEOMETRY_PROGRAM:
				mDevice->getImmediateContext()->GSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			case GPT_HULL_PROGRAM:
				mDevice->getImmediateContext()->HSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			case GPT_DOMAIN_PROGRAM:
				mDevice->getImmediateContext()->DSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			case GPT_COMPUTE_PROGRAM:
				mDevice->getImmediateContext()->CSSetConstantBuffers(iter->second.slot, 1, bufferArray);
				break;
			};

			BS_INC_RENDER_STAT(NumGpuParamBufferBinds);
		}

		if (mDevice->hasError())
			BS_EXCEPT(RenderingAPIException, "Failed to setConstantBuffers : " + mDevice->getErrorDescription());
	}

	void D3D11RenderAPI::draw(UINT32 vertexOffset, UINT32 vertexCount)
	{
		THROW_IF_NOT_CORE_THREAD;

		applyInputLayout();

		mDevice->getImmediateContext()->Draw(vertexCount, vertexOffset);

#if BS_DEBUG_MODE
		if(mDevice->hasError())
			LOGWRN(mDevice->getErrorDescription());
#endif

		UINT32 primCount = vertexCountToPrimCount(mActiveDrawOp, vertexCount);

		BS_INC_RENDER_STAT(NumDrawCalls);
		BS_ADD_RENDER_STAT(NumVertices, vertexCount);
		BS_ADD_RENDER_STAT(NumPrimitives, primCount);
	}

	void D3D11RenderAPI::drawIndexed(UINT32 startIndex, UINT32 indexCount, UINT32 vertexOffset, UINT32 vertexCount)
	{
		THROW_IF_NOT_CORE_THREAD;

		applyInputLayout();

		mDevice->getImmediateContext()->DrawIndexed(indexCount, startIndex, vertexOffset);

#if BS_DEBUG_MODE
		if(mDevice->hasError())
			LOGWRN(mDevice->getErrorDescription());
#endif

		UINT32 primCount = vertexCountToPrimCount(mActiveDrawOp, vertexCount);

		BS_INC_RENDER_STAT(NumDrawCalls);
		BS_ADD_RENDER_STAT(NumVertices, vertexCount);
		BS_ADD_RENDER_STAT(NumPrimitives, primCount);
	}

	void D3D11RenderAPI::dispatchCompute(UINT32 numGroupsX, UINT32 numGroupsY, UINT32 numGroupsZ)
	{
		mDevice->getImmediateContext()->Dispatch(numGroupsX, numGroupsY, numGroupsZ);

#if BS_DEBUG_MODE
		if (mDevice->hasError())
			LOGWRN(mDevice->getErrorDescription());
#endif

		BS_INC_RENDER_STAT(NumComputeCalls);
	}

	void D3D11RenderAPI::setScissorRect(UINT32 left, UINT32 top, UINT32 right, UINT32 bottom)
	{
		THROW_IF_NOT_CORE_THREAD;

		mScissorRect.left = static_cast<LONG>(left);
		mScissorRect.top = static_cast<LONG>(top);
		mScissorRect.bottom = static_cast<LONG>(bottom);
		mScissorRect.right = static_cast<LONG>(right);

		mDevice->getImmediateContext()->RSSetScissorRects(1, &mScissorRect);
	}

	void D3D11RenderAPI::clearViewport(UINT32 buffers, const Color& color, float depth, UINT16 stencil, UINT8 targetMask)
	{
		THROW_IF_NOT_CORE_THREAD;

		if(mActiveRenderTarget == nullptr)
			return;

		const RenderTargetProperties& rtProps = mActiveRenderTarget->getProperties();

		Rect2I clearArea((int)mViewport.TopLeftX, (int)mViewport.TopLeftY, (int)mViewport.Width, (int)mViewport.Height);

		bool clearEntireTarget = clearArea.width == 0 || clearArea.height == 0;
		clearEntireTarget |= (clearArea.x == 0 && clearArea.y == 0 && clearArea.width == rtProps.getWidth() && clearArea.height == rtProps.getHeight());

		if (!clearEntireTarget)
		{
			// TODO - Ignoring targetMask here
			D3D11RenderUtility::instance().drawClearQuad(buffers, color, depth, stencil);
			BS_INC_RENDER_STAT(NumClears);
		}
		else
			clearRenderTarget(buffers, color, depth, stencil, targetMask);
	}

	void D3D11RenderAPI::clearRenderTarget(UINT32 buffers, const Color& color, float depth, UINT16 stencil, UINT8 targetMask)
	{
		THROW_IF_NOT_CORE_THREAD;

		if(mActiveRenderTarget == nullptr)
			return;

		// Clear render surfaces
		if (buffers & FBT_COLOR)
		{
			UINT32 maxRenderTargets = mCurrentCapabilities->getNumMultiRenderTargets();

			ID3D11RenderTargetView** views = bs_newN<ID3D11RenderTargetView*>(maxRenderTargets);
			memset(views, 0, sizeof(ID3D11RenderTargetView*) * maxRenderTargets);

			mActiveRenderTarget->getCustomAttribute("RTV", views);
			if (!views[0])
			{
				bs_deleteN(views, maxRenderTargets);
				return;
			}

			float clearColor[4];
			clearColor[0] = color.r;
			clearColor[1] = color.g;
			clearColor[2] = color.b;
			clearColor[3] = color.a;

			for(UINT32 i = 0; i < maxRenderTargets; i++)
			{
				if(views[i] != nullptr && ((1 << i) & targetMask) != 0)
					mDevice->getImmediateContext()->ClearRenderTargetView(views[i], clearColor);
			}

			bs_deleteN(views, maxRenderTargets);
		}

		// Clear depth stencil
		if((buffers & FBT_DEPTH) != 0 || (buffers & FBT_STENCIL) != 0)
		{
			ID3D11DepthStencilView* depthStencilView = nullptr;
			mActiveRenderTarget->getCustomAttribute("DSV", &depthStencilView);

			D3D11_CLEAR_FLAG clearFlag;

			if((buffers & FBT_DEPTH) != 0 && (buffers & FBT_STENCIL) != 0)
				clearFlag = (D3D11_CLEAR_FLAG)(D3D11_CLEAR_DEPTH | D3D11_CLEAR_STENCIL);
			else if((buffers & FBT_STENCIL) != 0)
				clearFlag = D3D11_CLEAR_STENCIL;
			else
				clearFlag = D3D11_CLEAR_DEPTH;

			if(depthStencilView != nullptr)
				mDevice->getImmediateContext()->ClearDepthStencilView(depthStencilView, clearFlag, depth, (UINT8)stencil);
		}

		BS_INC_RENDER_STAT(NumClears);
	}

	void D3D11RenderAPI::setRenderTarget(const SPtr<RenderTargetCore>& target, bool readOnlyDepthStencil)
	{
		THROW_IF_NOT_CORE_THREAD;

		mActiveRenderTarget = target;

		UINT32 maxRenderTargets = mCurrentCapabilities->getNumMultiRenderTargets();
		ID3D11RenderTargetView** views = bs_newN<ID3D11RenderTargetView*>(maxRenderTargets);
		memset(views, 0, sizeof(ID3D11RenderTargetView*) * maxRenderTargets);

		ID3D11DepthStencilView* depthStencilView = nullptr;

		if (target != nullptr)
		{
			target->getCustomAttribute("RTV", views);

			if(readOnlyDepthStencil)
				target->getCustomAttribute("RODSV", &depthStencilView);
			else
				target->getCustomAttribute("DSV", &depthStencilView);
		}

		// Bind render targets
		mDevice->getImmediateContext()->OMSetRenderTargets(maxRenderTargets, views, depthStencilView);
		if (mDevice->hasError())
			BS_EXCEPT(RenderingAPIException, "Failed to setRenderTarget : " + mDevice->getErrorDescription());

		bs_deleteN(views, maxRenderTargets);
		applyViewport();

		BS_INC_RENDER_STAT(NumRenderTargetChanges);
	}

	void D3D11RenderAPI::setClipPlanesImpl(const PlaneList& clipPlanes)
	{
		LOGWRN("This call will be ignored. DX11 uses shaders for setting clip planes.");
	}

	RenderAPICapabilities* D3D11RenderAPI::createRenderSystemCapabilities() const
	{
		THROW_IF_NOT_CORE_THREAD;

		RenderAPICapabilities* rsc = bs_new<RenderAPICapabilities>();

		rsc->setDriverVersion(mDriverVersion);
		rsc->setDeviceName(mActiveD3DDriver->getDriverDescription());
		rsc->setRenderAPIName(getName());

		rsc->setStencilBufferBitDepth(8);

		rsc->setCapability(RSC_ANISOTROPY);
		rsc->setCapability(RSC_AUTOMIPMAP);

		// Cube map
		rsc->setCapability(RSC_CUBEMAPPING);

		// We always support compression, D3DX will decompress if device does not support
		rsc->setCapability(RSC_TEXTURE_COMPRESSION);
		rsc->setCapability(RSC_TEXTURE_COMPRESSION_DXT);
		rsc->setCapability(RSC_TWO_SIDED_STENCIL);
		rsc->setCapability(RSC_STENCIL_WRAP);
		rsc->setCapability(RSC_HWOCCLUSION);
		rsc->setCapability(RSC_HWOCCLUSION_ASYNCHRONOUS);

		if(mFeatureLevel >= D3D_FEATURE_LEVEL_10_1)
			rsc->setMaxBoundVertexBuffers(32);
		else
			rsc->setMaxBoundVertexBuffers(16);

		if(mFeatureLevel >= D3D_FEATURE_LEVEL_10_0)
		{
			rsc->addShaderProfile("ps_4_0");
			rsc->addShaderProfile("vs_4_0");
			rsc->addShaderProfile("gs_4_0");

			rsc->addGpuProgramProfile(GPP_FS_4_0, "ps_4_0");
			rsc->addGpuProgramProfile(GPP_VS_4_0, "vs_4_0");
			rsc->addGpuProgramProfile(GPP_GS_4_0, "gs_4_0");

			rsc->setNumTextureUnits(GPT_FRAGMENT_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);
			rsc->setNumTextureUnits(GPT_VERTEX_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);
			rsc->setNumTextureUnits(GPT_GEOMETRY_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);

			rsc->setNumCombinedTextureUnits(rsc->getNumTextureUnits(GPT_FRAGMENT_PROGRAM)
				+ rsc->getNumTextureUnits(GPT_VERTEX_PROGRAM) + rsc->getNumTextureUnits(GPT_VERTEX_PROGRAM));

			rsc->setNumGpuParamBlockBuffers(GPT_FRAGMENT_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
			rsc->setNumGpuParamBlockBuffers(GPT_VERTEX_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
			rsc->setNumGpuParamBlockBuffers(GPT_GEOMETRY_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);

			rsc->setNumCombinedGpuParamBlockBuffers(rsc->getNumGpuParamBlockBuffers(GPT_FRAGMENT_PROGRAM)
				+ rsc->getNumGpuParamBlockBuffers(GPT_VERTEX_PROGRAM) + rsc->getNumGpuParamBlockBuffers(GPT_VERTEX_PROGRAM));
		}

		if(mFeatureLevel >= D3D_FEATURE_LEVEL_10_1)
		{
			rsc->addShaderProfile("ps_4_1");
			rsc->addShaderProfile("vs_4_1");
			rsc->addShaderProfile("gs_4_1");

			rsc->addGpuProgramProfile(GPP_FS_4_1, "ps_4_1");
			rsc->addGpuProgramProfile(GPP_VS_4_1, "vs_4_1");
			rsc->addGpuProgramProfile(GPP_GS_4_1, "gs_4_1");
		}

		if(mFeatureLevel >= D3D_FEATURE_LEVEL_11_0)
		{
			rsc->addShaderProfile("ps_5_0");
			rsc->addShaderProfile("vs_5_0");
			rsc->addShaderProfile("gs_5_0");
			rsc->addShaderProfile("cs_5_0");
			rsc->addShaderProfile("hs_5_0");
			rsc->addShaderProfile("ds_5_0");

			rsc->addGpuProgramProfile(GPP_FS_5_0, "ps_5_0");
			rsc->addGpuProgramProfile(GPP_VS_5_0, "vs_5_0");
			rsc->addGpuProgramProfile(GPP_GS_5_0, "gs_5_0");
			rsc->addGpuProgramProfile(GPP_CS_5_0, "cs_5_0");
			rsc->addGpuProgramProfile(GPP_HS_5_0, "hs_5_0");
			rsc->addGpuProgramProfile(GPP_DS_5_0, "ds_5_0");

			rsc->setNumTextureUnits(GPT_HULL_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);
			rsc->setNumTextureUnits(GPT_DOMAIN_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);
			rsc->setNumTextureUnits(GPT_COMPUTE_PROGRAM, D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT);

			rsc->setNumCombinedTextureUnits(rsc->getNumTextureUnits(GPT_FRAGMENT_PROGRAM)
				+ rsc->getNumTextureUnits(GPT_VERTEX_PROGRAM) + rsc->getNumTextureUnits(GPT_VERTEX_PROGRAM)
				+ rsc->getNumTextureUnits(GPT_HULL_PROGRAM) + rsc->getNumTextureUnits(GPT_DOMAIN_PROGRAM)
				+ rsc->getNumTextureUnits(GPT_COMPUTE_PROGRAM));

			rsc->setNumGpuParamBlockBuffers(GPT_HULL_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
			rsc->setNumGpuParamBlockBuffers(GPT_DOMAIN_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
			rsc->setNumGpuParamBlockBuffers(GPT_COMPUTE_PROGRAM, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);

			rsc->setNumCombinedGpuParamBlockBuffers(rsc->getNumGpuParamBlockBuffers(GPT_FRAGMENT_PROGRAM)
				+ rsc->getNumGpuParamBlockBuffers(GPT_VERTEX_PROGRAM) + rsc->getNumGpuParamBlockBuffers(GPT_VERTEX_PROGRAM)
				+ rsc->getNumGpuParamBlockBuffers(GPT_HULL_PROGRAM) + rsc->getNumGpuParamBlockBuffers(GPT_DOMAIN_PROGRAM)
				+ rsc->getNumGpuParamBlockBuffers(GPT_COMPUTE_PROGRAM));

			rsc->setCapability(RSC_SHADER_SUBROUTINE);
		}

		rsc->setCapability(RSC_USER_CLIP_PLANES);
		rsc->setCapability(RSC_VERTEX_FORMAT_UBYTE4);

		// Adapter details
		const DXGI_ADAPTER_DESC& adapterID = mActiveD3DDriver->getAdapterIdentifier();

		// Determine vendor
		switch(adapterID.VendorId)
		{
		case 0x10DE:
			rsc->setVendor(GPU_NVIDIA);
			break;
		case 0x1002:
			rsc->setVendor(GPU_AMD);
			break;
		case 0x163C:
		case 0x8086:
			rsc->setVendor(GPU_INTEL);
			break;
		default:
			rsc->setVendor(GPU_UNKNOWN);
			break;
		};

		rsc->setCapability(RSC_INFINITE_FAR_PLANE);

		rsc->setCapability(RSC_TEXTURE_3D);
		rsc->setCapability(RSC_NON_POWER_OF_2_TEXTURES);
		rsc->setCapability(RSC_HWRENDER_TO_TEXTURE);
		rsc->setCapability(RSC_TEXTURE_FLOAT);

		rsc->setNumMultiRenderTargets(D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT);
		rsc->setCapability(RSC_MRT_DIFFERENT_BIT_DEPTHS);

		rsc->setCapability(RSC_POINT_SPRITES);
		rsc->setCapability(RSC_POINT_EXTENDED_PARAMETERS);
		rsc->setMaxPointSize(256);

		rsc->setCapability(RSC_VERTEX_TEXTURE_FETCH);

		rsc->setCapability(RSC_MIPMAP_LOD_BIAS);

		rsc->setCapability(RSC_PERSTAGECONSTANT);

		return rsc;
	}

	void D3D11RenderAPI::determineMultisampleSettings(UINT32 multisampleCount, DXGI_FORMAT format, DXGI_SAMPLE_DESC* outputSampleDesc)
	{
		bool tryCSAA = false; // Note: Disabled for now, but leaving the code for later so it might be useful
		enum CSAAMode { CSAA_Normal, CSAA_Quality };
		CSAAMode csaaMode = CSAA_Normal;

		bool foundValid = false;
		size_t origNumSamples = multisampleCount;
		while (!foundValid)
		{
			// Deal with special cases
			if (tryCSAA)
			{
				switch(multisampleCount)
				{
				case 8:
					if (csaaMode == CSAA_Quality)
					{
						outputSampleDesc->Count = 8;
						outputSampleDesc->Quality = 8;
					}
					else
					{
						outputSampleDesc->Count = 4;
						outputSampleDesc->Quality = 8;
					}
					break;
				case 16:
					if (csaaMode == CSAA_Quality)
					{
						outputSampleDesc->Count = 8;
						outputSampleDesc->Quality = 16;
					}
					else
					{
						outputSampleDesc->Count = 4;
						outputSampleDesc->Quality = 16;
					}
					break;
				}
			}
			else // !CSAA
			{
				outputSampleDesc->Count = multisampleCount == 0 ? 1 : multisampleCount;
				outputSampleDesc->Quality = D3D11_STANDARD_MULTISAMPLE_PATTERN;
			}


			HRESULT hr;
			UINT outQuality;
			hr = mDevice->getD3D11Device()->CheckMultisampleQualityLevels(format, outputSampleDesc->Count, &outQuality);

			if (SUCCEEDED(hr) && (!tryCSAA || outQuality > outputSampleDesc->Quality))
			{
				foundValid = true;
			}
			else
			{
				// Downgrade
				if (tryCSAA && multisampleCount == 8)
				{
					// For CSAA, we'll try downgrading with quality mode at all samples.
					// then try without quality, then drop CSAA
					if (csaaMode == CSAA_Quality)
					{
						// Drop quality first
						csaaMode = CSAA_Normal;
					}
					else
					{
						// Drop CSAA entirely 
						tryCSAA = false;
					}

					// Return to original requested samples
					multisampleCount = static_cast<UINT32>(origNumSamples);
				}
				else
				{
					// Drop samples
					multisampleCount--;

					if (multisampleCount == 1)
					{
						// Ran out of options, no multisampling
						multisampleCount = 0;
						foundValid = true;
					}
				}
			}
		} 
	}

	VertexElementType D3D11RenderAPI::getColorVertexElementType() const
	{
		return VET_COLOR_ABGR;
	}

	void D3D11RenderAPI::convertProjectionMatrix(const Matrix4& matrix, Matrix4& dest)
	{
		dest = matrix;

		// Convert depth range from [-1,+1] to [0,1]
		dest[2][0] = (dest[2][0] + dest[3][0]) / 2;
		dest[2][1] = (dest[2][1] + dest[3][1]) / 2;
		dest[2][2] = (dest[2][2] + dest[3][2]) / 2;
		dest[2][3] = (dest[2][3] + dest[3][3]) / 2;
	}

	float D3D11RenderAPI::getHorizontalTexelOffset()
	{
		return 0.0f;
	}

	float D3D11RenderAPI::getVerticalTexelOffset()
	{
		return 0.0f;
	}

	float D3D11RenderAPI::getMinimumDepthInputValue()
	{
		return 0.0f;
	}

	float D3D11RenderAPI::getMaximumDepthInputValue()
	{
		return 1.0f;
	}

	GpuParamBlockDesc D3D11RenderAPI::generateParamBlockDesc(const String& name, Vector<GpuParamDataDesc>& params)
	{
		GpuParamBlockDesc block;
		block.blockSize = 0;
		block.isShareable = true;
		block.name = name;
		block.slot = 0;

		for (auto& param : params)
		{
			const GpuParamDataTypeInfo& typeInfo = GpuParams::PARAM_SIZES.lookup[param.type];
			UINT32 size = typeInfo.size / 4;

			if (param.arraySize > 1)
			{
				// Arrays perform no packing and their elements are always padded and aligned to four component vectors
				UINT32 alignOffset = size % typeInfo.baseTypeSize;
				if (alignOffset != 0)
				{
					UINT32 padding = (typeInfo.baseTypeSize - alignOffset);
					size += padding;
				}

				alignOffset = block.blockSize % typeInfo.baseTypeSize;
				if (alignOffset != 0)
				{
					UINT32 padding = (typeInfo.baseTypeSize - alignOffset);
					block.blockSize += padding;
				}

				param.elementSize = size;
				param.arrayElementStride = size;
				param.cpuMemOffset = block.blockSize;
				param.gpuMemOffset = 0;

				block.blockSize += size * param.arraySize;
			}
			else
			{
				// Pack everything as tightly as possible as long as the data doesn't cross 16 byte boundary
				UINT32 alignOffset = block.blockSize % 4;
				if (alignOffset != 0 && size > (4 - alignOffset))
				{
					UINT32 padding = (4 - alignOffset);
					block.blockSize += padding;
				}

				param.elementSize = size;
				param.arrayElementStride = size;
				param.cpuMemOffset = block.blockSize;
				param.gpuMemOffset = 0;

				block.blockSize += size;
			}

			param.paramBlockSlot = 0;
		}

		// Constant buffer size must always be a multiple of 16
		if (block.blockSize % 4 != 0)
			block.blockSize += (4 - (block.blockSize % 4));

		return block;
	}

	/************************************************************************/
	/* 								PRIVATE		                     		*/
	/************************************************************************/

	void D3D11RenderAPI::applyInputLayout()
	{
		if(mActiveVertexDeclaration == nullptr)
		{
			LOGWRN("Cannot apply input layout without a vertex declaration. Set vertex declaration before calling this method.");
			return;
		}

		if(mActiveVertexShader == nullptr)
		{
			LOGWRN("Cannot apply input layout without a vertex shader. Set vertex shader before calling this method.");
			return;
		}

		ID3D11InputLayout* ia = mIAManager->retrieveInputLayout(mActiveVertexShader->getInputDeclaration(), mActiveVertexDeclaration, *mActiveVertexShader);

		mDevice->getImmediateContext()->IASetInputLayout(ia);
	}
}