external/vulkancts/modules/vulkan/api/vktApiCopiesAndBlittingTests.cpp

/*------------------------------------------------------------------------
 * Vulkan Conformance Tests
 * ------------------------
 *
 * Copyright (c) 2015-2016 The Khronos Group Inc.
 * Copyright (c) 2015-2016 Samsung Electronics Co., Ltd.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 *//*!
 * \file
 * \brief Vulkan Copies And Blitting Tests
 *//*--------------------------------------------------------------------*/

#include "vktApiCopiesAndBlittingTests.hpp"

#include "deStringUtil.hpp"
#include "deUniquePtr.hpp"

#include "tcuImageCompare.hpp"
#include "tcuTexture.hpp"
#include "tcuTextureUtil.hpp"
#include "tcuVectorType.hpp"
#include "tcuVectorUtil.hpp"
#include "tcuTestLog.hpp"
#include "tcuTexLookupVerifier.hpp"

#include "vkImageUtil.hpp"
#include "vkMemUtil.hpp"
#include "vkPrograms.hpp"
#include "vkQueryUtil.hpp"
#include "vkRefUtil.hpp"
#include "vktTestCase.hpp"
#include "vktTestCaseUtil.hpp"
#include "vktTestGroupUtil.hpp"
#include "vkTypeUtil.hpp"

#include <set>

namespace vkt
{

namespace api
{

namespace
{
enum MirrorMode
{
	MIRROR_MODE_NONE = 0,
	MIRROR_MODE_X = (1<<0),
	MIRROR_MODE_Y = (1<<1),
	MIRROR_MODE_XY = MIRROR_MODE_X | MIRROR_MODE_Y,

	MIRROR_MODE_LAST
};

enum AllocationKind
{
	ALLOCATION_KIND_SUBALLOCATED,
	ALLOCATION_KIND_DEDICATED,
};

template <typename Type>
class BinaryCompare
{
public:
	bool operator() (const Type& a, const Type& b) const
	{
		return deMemCmp(&a, &b, sizeof(Type)) < 0;
	}
};

typedef std::set<vk::VkFormat, BinaryCompare<vk::VkFormat> >	FormatSet;

FormatSet dedicatedAllocationImageToImageFormatsToTestSet;
FormatSet dedicatedAllocationBlittingFormatsToTestSet;

using namespace vk;

VkImageAspectFlags getAspectFlags (tcu::TextureFormat format)
{
	VkImageAspectFlags	aspectFlag	= 0;
	aspectFlag |= (tcu::hasDepthComponent(format.order)? VK_IMAGE_ASPECT_DEPTH_BIT : 0);
	aspectFlag |= (tcu::hasStencilComponent(format.order)? VK_IMAGE_ASPECT_STENCIL_BIT : 0);

	if (!aspectFlag)
		aspectFlag = VK_IMAGE_ASPECT_COLOR_BIT;

	return aspectFlag;
}

// This is effectively same as vk::isFloatFormat(mapTextureFormat(format))
// except that it supports some formats that are not mappable to VkFormat.
// When we are checking combined depth and stencil formats, each aspect is
// checked separately, and in some cases we construct PBA with a format that
// is not mappable to VkFormat.
bool isFloatFormat (tcu::TextureFormat format)
{
	return tcu::getTextureChannelClass(format.type) == tcu::TEXTURECHANNELCLASS_FLOATING_POINT;
}

union CopyRegion
{
	VkBufferCopy		bufferCopy;
	VkImageCopy			imageCopy;
	VkBufferImageCopy	bufferImageCopy;
	VkImageBlit			imageBlit;
	VkImageResolve		imageResolve;
};

struct ImageParms
{
	VkImageType		imageType;
	VkFormat		format;
	VkExtent3D		extent;
	VkImageLayout	operationLayout;
};

struct TestParams
{
	union Data
	{
		struct Buffer
		{
			VkDeviceSize	size;
		} buffer;

		ImageParms	image;
	} src, dst;

	std::vector<CopyRegion>	regions;

	union
	{
		VkFilter				filter;
		VkSampleCountFlagBits	samples;
	};

	AllocationKind allocationKind;
};

de::MovePtr<Allocation> allocateBuffer (const InstanceInterface&	vki,
										const DeviceInterface&		vkd,
										const VkPhysicalDevice&		physDevice,
										const VkDevice				device,
										const VkBuffer&				buffer,
										const MemoryRequirement		requirement,
										Allocator&					allocator,
										AllocationKind				allocationKind)
{
	switch (allocationKind)
	{
		case ALLOCATION_KIND_SUBALLOCATED:
		{
			const VkMemoryRequirements memoryRequirements = getBufferMemoryRequirements(vkd, device, buffer);

			return allocator.allocate(memoryRequirements, requirement);
		}

		case ALLOCATION_KIND_DEDICATED:
		{
			return allocateDedicated(vki, vkd, physDevice, device, buffer, requirement);
		}

		default:
		{
			TCU_THROW(InternalError, "Invalid allocation kind");
		}
	}
}

de::MovePtr<Allocation> allocateImage (const InstanceInterface&		vki,
									   const DeviceInterface&		vkd,
									   const VkPhysicalDevice&		physDevice,
									   const VkDevice				device,
									   const VkImage&				image,
									   const MemoryRequirement		requirement,
									   Allocator&					allocator,
									   AllocationKind				allocationKind)
{
	switch (allocationKind)
	{
		case ALLOCATION_KIND_SUBALLOCATED:
		{
			const VkMemoryRequirements memoryRequirements = getImageMemoryRequirements(vkd, device, image);

			return allocator.allocate(memoryRequirements, requirement);
		}

		case ALLOCATION_KIND_DEDICATED:
		{
			return allocateDedicated(vki, vkd, physDevice, device, image, requirement);
		}

		default:
		{
			TCU_THROW(InternalError, "Invalid allocation kind");
		}
	}
}


inline deUint32 getArraySize(const ImageParms& parms)
{
	return (parms.imageType == VK_IMAGE_TYPE_2D) ? parms.extent.depth : 1u;
}

inline VkExtent3D getExtent3D(const ImageParms& parms)
{
	const VkExtent3D		extent					=
	{
		parms.extent.width,
		parms.extent.height,
		(parms.imageType == VK_IMAGE_TYPE_2D) ? 1u : parms.extent.depth
	};
	return extent;
}

const tcu::TextureFormat mapCombinedToDepthTransferFormat (const tcu::TextureFormat& combinedFormat)
{
	tcu::TextureFormat format;
	switch (combinedFormat.type)
	{
		case tcu::TextureFormat::UNSIGNED_INT_16_8_8:
			format = tcu::TextureFormat(tcu::TextureFormat::D, tcu::TextureFormat::UNORM_INT16);
			break;
		case tcu::TextureFormat::UNSIGNED_INT_24_8_REV:
			format = tcu::TextureFormat(tcu::TextureFormat::D, tcu::TextureFormat::UNSIGNED_INT_24_8_REV);
			break;
		case tcu::TextureFormat::FLOAT_UNSIGNED_INT_24_8_REV:
			format = tcu::TextureFormat(tcu::TextureFormat::D, tcu::TextureFormat::FLOAT);
			break;
		default:
			DE_ASSERT(false);
			break;
	}
	return format;
}

class CopiesAndBlittingTestInstance : public vkt::TestInstance
{
public:
										CopiesAndBlittingTestInstance		(Context&	context,
																			 TestParams	testParams);
	virtual tcu::TestStatus				iterate								(void) = 0;

	enum FillMode
	{
		FILL_MODE_GRADIENT = 0,
		FILL_MODE_WHITE,
		FILL_MODE_RED,
		FILL_MODE_MULTISAMPLE,

		FILL_MODE_LAST
	};

protected:
	const TestParams					m_params;

	Move<VkCommandPool>					m_cmdPool;
	Move<VkCommandBuffer>				m_cmdBuffer;
	Move<VkFence>						m_fence;
	de::MovePtr<tcu::TextureLevel>		m_sourceTextureLevel;
	de::MovePtr<tcu::TextureLevel>		m_destinationTextureLevel;
	de::MovePtr<tcu::TextureLevel>		m_expectedTextureLevel;

	VkCommandBufferBeginInfo			m_cmdBufferBeginInfo;

	void								generateBuffer						(tcu::PixelBufferAccess buffer, int width, int height, int depth = 1, FillMode = FILL_MODE_GRADIENT);
	virtual void						generateExpectedResult				(void);
	void								uploadBuffer						(tcu::ConstPixelBufferAccess bufferAccess, const Allocation& bufferAlloc);
	void								uploadImage							(const tcu::ConstPixelBufferAccess& src, VkImage dst, const ImageParms& parms);
	virtual tcu::TestStatus				checkTestResult						(tcu::ConstPixelBufferAccess result);
	virtual void						copyRegionToTextureLevel			(tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region) = 0;
	deUint32							calculateSize						(tcu::ConstPixelBufferAccess src) const
										{
											return src.getWidth() * src.getHeight() * src.getDepth() * tcu::getPixelSize(src.getFormat());
										}

	de::MovePtr<tcu::TextureLevel>		readImage							(vk::VkImage				image,
																			 const ImageParms&			imageParms);
	void								submitCommandsAndWait				(const DeviceInterface&		vk,
																			const VkDevice				device,
																			const VkQueue				queue,
																			const VkCommandBuffer&		cmdBuffer);

private:
	void								uploadImageAspect					(const tcu::ConstPixelBufferAccess&	src,
																			 const VkImage&						dst,
																			 const ImageParms&					parms);
	void								readImageAspect						(vk::VkImage						src,
																			 const tcu::PixelBufferAccess&		dst,
																			 const ImageParms&					parms);
};

CopiesAndBlittingTestInstance::CopiesAndBlittingTestInstance (Context& context, TestParams testParams)
	: vkt::TestInstance	(context)
	, m_params			(testParams)
{
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();

	if (m_params.allocationKind == ALLOCATION_KIND_DEDICATED)
	{
		const std::string extensionName("VK_KHR_dedicated_allocation");

		if (!de::contains(context.getDeviceExtensions().begin(), context.getDeviceExtensions().end(), extensionName))
			TCU_THROW(NotSupportedError, std::string(extensionName + " is not supported").c_str());
	}

	// Create command pool
	m_cmdPool = createCommandPool(vk, vkDevice, VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT, queueFamilyIndex);

	// Create command buffer
	m_cmdBuffer = allocateCommandBuffer(vk, vkDevice, *m_cmdPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY);

	// Create fence
	m_fence = createFence(vk, vkDevice);
}

void CopiesAndBlittingTestInstance::generateBuffer (tcu::PixelBufferAccess buffer, int width, int height, int depth, FillMode mode)
{
	const tcu::TextureChannelClass	channelClass	= tcu::getTextureChannelClass(buffer.getFormat().type);
	tcu::Vec4						maxValue		(1.0f);

	if (buffer.getFormat().order == tcu::TextureFormat::S)
	{
		// Stencil-only is stored in the first component. Stencil is always 8 bits.
		maxValue.x() = 1 << 8;
	}
	else if (buffer.getFormat().order == tcu::TextureFormat::DS)
	{
		// In a combined format, fillWithComponentGradients expects stencil in the fourth component.
		maxValue.w() = 1 << 8;
	}
	else if (channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER || channelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER)
	{
		// The tcu::Vectors we use as pixels are 32-bit, so clamp to that.
		const tcu::IVec4	bits	= tcu::min(tcu::getTextureFormatBitDepth(buffer.getFormat()), tcu::IVec4(32));
		const int			signBit	= (channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER ? 1 : 0);

		for (int i = 0; i < 4; ++i)
		{
			if (bits[i] != 0)
				maxValue[i] = static_cast<float>((1 << (bits[i] - signBit)) - 1);
		}
	}

	if (mode == FILL_MODE_GRADIENT)
	{
		tcu::fillWithComponentGradients(buffer, tcu::Vec4(0.0f, 0.0f, 0.0f, 0.0f), maxValue);
		return;
	}

	const tcu::Vec4		redColor	(maxValue.x(),	0.0,			0.0,			maxValue.w());
	const tcu::Vec4		greenColor	(0.0,			maxValue.y(),	0.0,			maxValue.w());
	const tcu::Vec4		blueColor	(0.0,			0.0,			maxValue.z(),	maxValue.w());
	const tcu::Vec4		whiteColor	(maxValue.x(),	maxValue.y(),	maxValue.z(),	maxValue.w());

	for (int z = 0; z < depth;  ++z)
	for (int y = 0; y < height; ++y)
	for (int x = 0; x < width;  ++x)
	{
		switch (mode)
		{
			case FILL_MODE_WHITE:
				if (tcu::isCombinedDepthStencilType(buffer.getFormat().type))
				{
					buffer.setPixDepth(1.0f, x, y, z);
					if (tcu::hasStencilComponent(buffer.getFormat().order))
						buffer.setPixStencil(255, x, y, z);
				}
				else
					buffer.setPixel(whiteColor, x, y, z);
				break;

			case FILL_MODE_RED:
				if (tcu::isCombinedDepthStencilType(buffer.getFormat().type))
				{
					buffer.setPixDepth(redColor[x % 4], x, y, z);
					if (tcu::hasStencilComponent(buffer.getFormat().order))
						buffer.setPixStencil(255 * (int)redColor[y % 4], x, y, z);
				}
				else
					buffer.setPixel(redColor, x, y, z);
				break;

			case FILL_MODE_MULTISAMPLE:
				buffer.setPixel((x == y) ? tcu::Vec4(0.0, 0.5, 0.5, 1.0) : ((x > y) ? greenColor : blueColor), x, y, z);
				break;

			default:
				break;
		}
	}
}

void CopiesAndBlittingTestInstance::uploadBuffer (tcu::ConstPixelBufferAccess bufferAccess, const Allocation& bufferAlloc)
{
	const DeviceInterface&		vk			= m_context.getDeviceInterface();
	const VkDevice				vkDevice	= m_context.getDevice();
	const deUint32				bufferSize	= calculateSize(bufferAccess);

	// Write buffer data
	deMemcpy(bufferAlloc.getHostPtr(), bufferAccess.getDataPtr(), bufferSize);
	flushMappedMemoryRange(vk, vkDevice, bufferAlloc.getMemory(), bufferAlloc.getOffset(), bufferSize);
}

void CopiesAndBlittingTestInstance::uploadImageAspect (const tcu::ConstPixelBufferAccess& imageAccess, const VkImage& image, const ImageParms& parms)
{
	const InstanceInterface&	vki					= m_context.getInstanceInterface();
	const DeviceInterface&		vk					= m_context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= m_context.getPhysicalDevice();
	const VkDevice				vkDevice			= m_context.getDevice();
	const VkQueue				queue				= m_context.getUniversalQueue();
	const deUint32				queueFamilyIndex	= m_context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= m_context.getDefaultAllocator();
	Move<VkBuffer>				buffer;
	const deUint32				bufferSize			= calculateSize(imageAccess);
	de::MovePtr<Allocation>		bufferAlloc;
	const deUint32				arraySize			= getArraySize(parms);
	const VkExtent3D			imageExtent			= getExtent3D(parms);

	// Create source buffer
	{
		const VkBufferCreateInfo			bufferParams			=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			bufferSize,									// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_SRC_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		buffer		= createBuffer(vk, vkDevice, &bufferParams);
		bufferAlloc = allocateBuffer(vki, vk, vkPhysDevice, vkDevice, *buffer, MemoryRequirement::HostVisible, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(vkDevice, *buffer, bufferAlloc->getMemory(), bufferAlloc->getOffset()));
	}

	// Barriers for copying buffer to image
	const VkBufferMemoryBarrier				preBufferBarrier		=
	{
		VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,		// VkStructureType	sType;
		DE_NULL,										// const void*		pNext;
		VK_ACCESS_HOST_WRITE_BIT,						// VkAccessFlags	srcAccessMask;
		VK_ACCESS_TRANSFER_READ_BIT,					// VkAccessFlags	dstAccessMask;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32			srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32			dstQueueFamilyIndex;
		*buffer,										// VkBuffer			buffer;
		0u,												// VkDeviceSize		offset;
		bufferSize										// VkDeviceSize		size;
	};

	const VkImageAspectFlags				formatAspect			= getAspectFlags(mapVkFormat(parms.format));
	const bool								skipPreImageBarrier		= formatAspect == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT) &&
																	  getAspectFlags(imageAccess.getFormat()) == VK_IMAGE_ASPECT_STENCIL_BIT;
	const VkImageMemoryBarrier				preImageBarrier			=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,			// VkStructureType			sType;
		DE_NULL,										// const void*				pNext;
		0u,												// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_WRITE_BIT,					// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_UNDEFINED,						// VkImageLayout			oldLayout;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,			// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32					dstQueueFamilyIndex;
		image,											// VkImage					image;
		{												// VkImageSubresourceRange	subresourceRange;
			formatAspect,	// VkImageAspectFlags	aspect;
			0u,				// deUint32				baseMipLevel;
			1u,				// deUint32				mipLevels;
			0u,				// deUint32				baseArraySlice;
			arraySize,		// deUint32				arraySize;
		}
	};

	const VkImageMemoryBarrier				postImageBarrier		=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,			// VkStructureType			sType;
		DE_NULL,										// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,					// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_WRITE_BIT,					// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,			// VkImageLayout			oldLayout;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,			// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,						// deUint32					dstQueueFamilyIndex;
		image,											// VkImage					image;
		{												// VkImageSubresourceRange	subresourceRange;
			formatAspect,				// VkImageAspectFlags	aspect;
			0u,							// deUint32				baseMipLevel;
			1u,							// deUint32				mipLevels;
			0u,							// deUint32				baseArraySlice;
			arraySize,					// deUint32				arraySize;
		}
	};

	const VkBufferImageCopy		copyRegion		=
	{
		0u,												// VkDeviceSize				bufferOffset;
		(deUint32)imageAccess.getWidth(),				// deUint32					bufferRowLength;
		(deUint32)imageAccess.getHeight(),				// deUint32					bufferImageHeight;
		{
			getAspectFlags(imageAccess.getFormat()),		// VkImageAspectFlags	aspect;
			0u,												// deUint32				mipLevel;
			0u,												// deUint32				baseArrayLayer;
			arraySize,										// deUint32				layerCount;
		},												// VkImageSubresourceLayers	imageSubresource;
		{ 0, 0, 0 },									// VkOffset3D				imageOffset;
		imageExtent										// VkExtent3D				imageExtent;
	};

	// Write buffer data
	deMemcpy(bufferAlloc->getHostPtr(), imageAccess.getDataPtr(), bufferSize);
	flushMappedMemoryRange(vk, vkDevice, bufferAlloc->getMemory(), bufferAlloc->getOffset(), bufferSize);

	// Copy buffer to image
	const VkCommandBufferBeginInfo			cmdBufferBeginInfo		=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL,
						  1, &preBufferBarrier, (skipPreImageBarrier ? 0 : 1), (skipPreImageBarrier ? DE_NULL : &preImageBarrier));
	vk.cmdCopyBufferToImage(*m_cmdBuffer, *buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1u, &copyRegion);
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &postImageBarrier);
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));

	submitCommandsAndWait(vk, vkDevice, queue, *m_cmdBuffer);
}

void CopiesAndBlittingTestInstance::uploadImage (const tcu::ConstPixelBufferAccess& src, VkImage dst, const ImageParms& parms)
{
	if (tcu::isCombinedDepthStencilType(src.getFormat().type))
	{
		if (tcu::hasDepthComponent(src.getFormat().order))
		{
			tcu::TextureLevel	depthTexture	(mapCombinedToDepthTransferFormat(src.getFormat()), src.getWidth(), src.getHeight(), src.getDepth());
			tcu::copy(depthTexture.getAccess(), tcu::getEffectiveDepthStencilAccess(src, tcu::Sampler::MODE_DEPTH));
			uploadImageAspect(depthTexture.getAccess(), dst, parms);
		}

		if (tcu::hasStencilComponent(src.getFormat().order))
		{
			tcu::TextureLevel	stencilTexture	(tcu::getEffectiveDepthStencilTextureFormat(src.getFormat(), tcu::Sampler::MODE_STENCIL), src.getWidth(), src.getHeight(), src.getDepth());
			tcu::copy(stencilTexture.getAccess(), tcu::getEffectiveDepthStencilAccess(src, tcu::Sampler::MODE_STENCIL));
			uploadImageAspect(stencilTexture.getAccess(), dst, parms);
		}
	}
	else
		uploadImageAspect(src, dst, parms);
}

tcu::TestStatus CopiesAndBlittingTestInstance::checkTestResult (tcu::ConstPixelBufferAccess result)
{
	const tcu::ConstPixelBufferAccess	expected	= m_expectedTextureLevel->getAccess();

	if (isFloatFormat(result.getFormat()))
	{
		const tcu::Vec4	threshold (0.0f);
		if (!tcu::floatThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expected, result, threshold, tcu::COMPARE_LOG_RESULT))
			return tcu::TestStatus::fail("CopiesAndBlitting test");
	}
	else
	{
		const tcu::UVec4 threshold (0u);
		if (!tcu::intThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expected, result, threshold, tcu::COMPARE_LOG_RESULT))
			return tcu::TestStatus::fail("CopiesAndBlitting test");
	}

	return tcu::TestStatus::pass("CopiesAndBlitting test");
}

void CopiesAndBlittingTestInstance::generateExpectedResult (void)
{
	const tcu::ConstPixelBufferAccess	src	= m_sourceTextureLevel->getAccess();
	const tcu::ConstPixelBufferAccess	dst	= m_destinationTextureLevel->getAccess();

	m_expectedTextureLevel	= de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(dst.getFormat(), dst.getWidth(), dst.getHeight(), dst.getDepth()));
	tcu::copy(m_expectedTextureLevel->getAccess(), dst);

	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		copyRegionToTextureLevel(src, m_expectedTextureLevel->getAccess(), m_params.regions[i]);
}

class CopiesAndBlittingTestCase : public vkt::TestCase
{
public:
							CopiesAndBlittingTestCase	(tcu::TestContext&			testCtx,
														 const std::string&			name,
														 const std::string&			description)
								: vkt::TestCase	(testCtx, name, description)
							{}

	virtual TestInstance*	createInstance				(Context&					context) const = 0;
};

void CopiesAndBlittingTestInstance::readImageAspect (vk::VkImage					image,
													 const tcu::PixelBufferAccess&	dst,
													 const ImageParms&				imageParms)
{
	const InstanceInterface&	vki					= m_context.getInstanceInterface();
	const DeviceInterface&		vk					= m_context.getDeviceInterface();
	const VkPhysicalDevice		physDevice			= m_context.getPhysicalDevice();
	const VkDevice				device				= m_context.getDevice();
	const VkQueue				queue				= m_context.getUniversalQueue();
	Allocator&					allocator			= m_context.getDefaultAllocator();

	Move<VkBuffer>				buffer;
	de::MovePtr<Allocation>		bufferAlloc;
	const deUint32				queueFamilyIndex	= m_context.getUniversalQueueFamilyIndex();
	const VkDeviceSize			pixelDataSize		= calculateSize(dst);
	const VkExtent3D			imageExtent			= getExtent3D(imageParms);

	// Create destination buffer
	{
		const VkBufferCreateInfo			bufferParams			=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			pixelDataSize,								// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_DST_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		buffer		= createBuffer(vk, device, &bufferParams);
		bufferAlloc = allocateBuffer(vki, vk, physDevice, device, *buffer, MemoryRequirement::HostVisible, allocator, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(device, *buffer, bufferAlloc->getMemory(), bufferAlloc->getOffset()));

		deMemset(bufferAlloc->getHostPtr(), 0, static_cast<size_t>(pixelDataSize));
		flushMappedMemoryRange(vk, device, bufferAlloc->getMemory(), bufferAlloc->getOffset(), pixelDataSize);
	}

	// Barriers for copying image to buffer
	const VkImageAspectFlags				formatAspect			= getAspectFlags(mapVkFormat(imageParms.format));
	const VkImageMemoryBarrier				imageBarrier			=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags			dstAccessMask;
		imageParms.operationLayout,					// VkImageLayout			oldLayout;
		VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,		// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		image,										// VkImage					image;
		{											// VkImageSubresourceRange	subresourceRange;
			formatAspect,			// VkImageAspectFlags	aspectMask;
			0u,						// deUint32				baseMipLevel;
			1u,						// deUint32				mipLevels;
			0u,						// deUint32				baseArraySlice;
			getArraySize(imageParms)// deUint32				arraySize;
		}
	};

	const VkBufferMemoryBarrier				bufferBarrier			=
	{
		VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,	// VkStructureType	sType;
		DE_NULL,									// const void*		pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags	srcAccessMask;
		VK_ACCESS_HOST_READ_BIT,					// VkAccessFlags	dstAccessMask;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			dstQueueFamilyIndex;
		*buffer,									// VkBuffer			buffer;
		0u,											// VkDeviceSize		offset;
		pixelDataSize								// VkDeviceSize		size;
	};

	const VkImageMemoryBarrier				postImageBarrier		=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,		// VkImageLayout			oldLayout;
		imageParms.operationLayout,					// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		image,										// VkImage					image;
		{
			formatAspect,								// VkImageAspectFlags	aspectMask;
			0u,											// deUint32				baseMipLevel;
			1u,											// deUint32				mipLevels;
			0u,											// deUint32				baseArraySlice;
			getArraySize(imageParms)					// deUint32				arraySize;
		}											// VkImageSubresourceRange	subresourceRange;
	};

	// Copy image to buffer
	const VkImageAspectFlags	aspect			= getAspectFlags(dst.getFormat());
	const VkBufferImageCopy		copyRegion		=
	{
		0u,									// VkDeviceSize				bufferOffset;
		(deUint32)dst.getWidth(),			// deUint32					bufferRowLength;
		(deUint32)dst.getHeight(),			// deUint32					bufferImageHeight;
		{
			aspect,								// VkImageAspectFlags		aspect;
			0u,									// deUint32					mipLevel;
			0u,									// deUint32					baseArrayLayer;
			getArraySize(imageParms),			// deUint32					layerCount;
		},									// VkImageSubresourceLayers	imageSubresource;
		{ 0, 0, 0 },						// VkOffset3D				imageOffset;
		imageExtent							// VkExtent3D				imageExtent;
	};

	const VkCommandBufferBeginInfo			cmdBufferBeginInfo		=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &imageBarrier);
	vk.cmdCopyImageToBuffer(*m_cmdBuffer, image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, *buffer, 1u, &copyRegion);
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_HOST_BIT|VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 1, &bufferBarrier, 1, &postImageBarrier);
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));

	submitCommandsAndWait(vk, device, queue, *m_cmdBuffer);

	// Read buffer data
	invalidateMappedMemoryRange(vk, device, bufferAlloc->getMemory(), bufferAlloc->getOffset(), pixelDataSize);
	tcu::copy(dst, tcu::ConstPixelBufferAccess(dst.getFormat(), dst.getSize(), bufferAlloc->getHostPtr()));
}

void CopiesAndBlittingTestInstance::submitCommandsAndWait (const DeviceInterface& vk, const VkDevice device, const VkQueue queue, const VkCommandBuffer& cmdBuffer)
{
	const VkSubmitInfo						submitInfo				=
	{
		VK_STRUCTURE_TYPE_SUBMIT_INFO,	// VkStructureType			sType;
		DE_NULL,						// const void*				pNext;
		0u,								// deUint32					waitSemaphoreCount;
		DE_NULL,						// const VkSemaphore*		pWaitSemaphores;
		(const VkPipelineStageFlags*)DE_NULL,
		1u,								// deUint32					commandBufferCount;
		&cmdBuffer,						// const VkCommandBuffer*	pCommandBuffers;
		0u,								// deUint32					signalSemaphoreCount;
		DE_NULL							// const VkSemaphore*		pSignalSemaphores;
	};

	VK_CHECK(vk.resetFences(device, 1, &m_fence.get()));
	VK_CHECK(vk.queueSubmit(queue, 1, &submitInfo, *m_fence));
	VK_CHECK(vk.waitForFences(device, 1, &m_fence.get(), true, ~(0ull) /* infinity */));
}

de::MovePtr<tcu::TextureLevel> CopiesAndBlittingTestInstance::readImage	(vk::VkImage		image,
																		 const ImageParms&	parms)
{
	const tcu::TextureFormat		imageFormat	= mapVkFormat(parms.format);
	de::MovePtr<tcu::TextureLevel>	resultLevel	(new tcu::TextureLevel(imageFormat, parms.extent.width, parms.extent.height, parms.extent.depth));

	if (tcu::isCombinedDepthStencilType(imageFormat.type))
	{
		if (tcu::hasDepthComponent(imageFormat.order))
		{
			tcu::TextureLevel	depthTexture	(mapCombinedToDepthTransferFormat(imageFormat), parms.extent.width, parms.extent.height, parms.extent.depth);
			readImageAspect(image, depthTexture.getAccess(), parms);
			tcu::copy(tcu::getEffectiveDepthStencilAccess(resultLevel->getAccess(), tcu::Sampler::MODE_DEPTH), depthTexture.getAccess());
		}

		if (tcu::hasStencilComponent(imageFormat.order))
		{
			tcu::TextureLevel	stencilTexture	(tcu::getEffectiveDepthStencilTextureFormat(imageFormat, tcu::Sampler::MODE_STENCIL), parms.extent.width, parms.extent.height, parms.extent.depth);
			readImageAspect(image, stencilTexture.getAccess(), parms);
			tcu::copy(tcu::getEffectiveDepthStencilAccess(resultLevel->getAccess(), tcu::Sampler::MODE_STENCIL), stencilTexture.getAccess());
		}
	}
	else
		readImageAspect(image, resultLevel->getAccess(), parms);

	return resultLevel;
}

// Copy from image to image.

class CopyImageToImage : public CopiesAndBlittingTestInstance
{
public:
										CopyImageToImage			(Context&	context,
																	 TestParams params);
	virtual tcu::TestStatus				iterate						(void);

protected:
	virtual tcu::TestStatus				checkTestResult				(tcu::ConstPixelBufferAccess result);

private:
	Move<VkImage>						m_source;
	de::MovePtr<Allocation>				m_sourceImageAlloc;
	Move<VkImage>						m_destination;
	de::MovePtr<Allocation>				m_destinationImageAlloc;

	virtual void						copyRegionToTextureLevel	(tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region);
};

CopyImageToImage::CopyImageToImage (Context& context, TestParams params)
	: CopiesAndBlittingTestInstance(context, params)
{
	const InstanceInterface&	vki					= context.getInstanceInterface();
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= context.getPhysicalDevice();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= context.getDefaultAllocator();

	if ((m_params.dst.image.imageType == VK_IMAGE_TYPE_3D && m_params.src.image.imageType == VK_IMAGE_TYPE_2D) ||
		(m_params.dst.image.imageType == VK_IMAGE_TYPE_2D && m_params.src.image.imageType == VK_IMAGE_TYPE_3D))
	{
		if (std::find(context.getDeviceExtensions().begin(), context.getDeviceExtensions().end(), "VK_KHR_maintenance1") == context.getDeviceExtensions().end())
			TCU_THROW(NotSupportedError, "Extension VK_KHR_maintenance1 not supported");
	}

	VkImageFormatProperties properties;
	if ((context.getInstanceInterface().getPhysicalDeviceImageFormatProperties (context.getPhysicalDevice(),
																				m_params.src.image.format,
																				m_params.src.image.imageType,
																				VK_IMAGE_TILING_OPTIMAL,
																				VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
																				0,
																				&properties) == VK_ERROR_FORMAT_NOT_SUPPORTED) ||
		(context.getInstanceInterface().getPhysicalDeviceImageFormatProperties (context.getPhysicalDevice(),
																				m_params.dst.image.format,
																				m_params.dst.image.imageType,
																				VK_IMAGE_TILING_OPTIMAL,
																				VK_IMAGE_USAGE_TRANSFER_DST_BIT,
																				0,
																				&properties) == VK_ERROR_FORMAT_NOT_SUPPORTED))
	{
		TCU_THROW(NotSupportedError, "Format not supported");
	}

	// Create source image
	{
		const VkImageCreateInfo	sourceImageParams		=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.src.image.imageType,			// VkImageType			imageType;
			m_params.src.image.format,				// VkFormat				format;
			getExtent3D(m_params.src.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.src.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_source				= createImage(vk, vkDevice, &sourceImageParams);
		m_sourceImageAlloc		= allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_source, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_source, m_sourceImageAlloc->getMemory(), m_sourceImageAlloc->getOffset()));
	}

	// Create destination image
	{
		const VkImageCreateInfo	destinationImageParams	=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.dst.image.imageType,			// VkImageType			imageType;
			m_params.dst.image.format,				// VkFormat				format;
			getExtent3D(m_params.dst.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.dst.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_destination			= createImage(vk, vkDevice, &destinationImageParams);
		m_destinationImageAlloc	= allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_destination, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_destination, m_destinationImageAlloc->getMemory(), m_destinationImageAlloc->getOffset()));
	}
}

tcu::TestStatus CopyImageToImage::iterate (void)
{
	const tcu::TextureFormat	srcTcuFormat		= mapVkFormat(m_params.src.image.format);
	const tcu::TextureFormat	dstTcuFormat		= mapVkFormat(m_params.dst.image.format);
	m_sourceTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(srcTcuFormat,
																				(int)m_params.src.image.extent.width,
																				(int)m_params.src.image.extent.height,
																				(int)m_params.src.image.extent.depth));
	generateBuffer(m_sourceTextureLevel->getAccess(), m_params.src.image.extent.width, m_params.src.image.extent.height, m_params.src.image.extent.depth, FILL_MODE_RED);
	m_destinationTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(dstTcuFormat,
																				(int)m_params.dst.image.extent.width,
																				(int)m_params.dst.image.extent.height,
																				(int)m_params.dst.image.extent.depth));
	generateBuffer(m_destinationTextureLevel->getAccess(), m_params.dst.image.extent.width, m_params.dst.image.extent.height, m_params.dst.image.extent.depth, FILL_MODE_GRADIENT);
	generateExpectedResult();

	uploadImage(m_sourceTextureLevel->getAccess(), m_source.get(), m_params.src.image);
	uploadImage(m_destinationTextureLevel->getAccess(), m_destination.get(), m_params.dst.image);

	const DeviceInterface&		vk					= m_context.getDeviceInterface();
	const VkDevice				vkDevice			= m_context.getDevice();
	const VkQueue				queue				= m_context.getUniversalQueue();

	std::vector<VkImageCopy>	imageCopies;
	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		imageCopies.push_back(m_params.regions[i].imageCopy);

	const VkImageMemoryBarrier	imageBarriers[]		=
	{
		// source image
		{
			VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
			DE_NULL,									// const void*				pNext;
			VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
			VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags			dstAccessMask;
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
			m_params.src.image.operationLayout,			// VkImageLayout			newLayout;
			VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
			VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
			m_source.get(),								// VkImage					image;
			{											// VkImageSubresourceRange	subresourceRange;
				getAspectFlags(srcTcuFormat),	// VkImageAspectFlags	aspectMask;
				0u,								// deUint32				baseMipLevel;
				1u,								// deUint32				mipLevels;
				0u,								// deUint32				baseArraySlice;
				getArraySize(m_params.src.image)// deUint32				arraySize;
			}
		},
		// destination image
		{
			VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
			DE_NULL,									// const void*				pNext;
			VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
			VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			dstAccessMask;
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
			m_params.dst.image.operationLayout,			// VkImageLayout			newLayout;
			VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
			VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
			m_destination.get(),						// VkImage					image;
			{											// VkImageSubresourceRange	subresourceRange;
				getAspectFlags(dstTcuFormat),	// VkImageAspectFlags	aspectMask;
				0u,								// deUint32				baseMipLevel;
				1u,								// deUint32				mipLevels;
				0u,								// deUint32				baseArraySlice;
				getArraySize(m_params.dst.image)// deUint32				arraySize;
			}
		},
	};

	const VkCommandBufferBeginInfo	cmdBufferBeginInfo	=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, DE_LENGTH_OF_ARRAY(imageBarriers), imageBarriers);
	vk.cmdCopyImage(*m_cmdBuffer, m_source.get(), m_params.src.image.operationLayout, m_destination.get(), m_params.dst.image.operationLayout, (deUint32)m_params.regions.size(), imageCopies.data());
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));

	submitCommandsAndWait (vk, vkDevice, queue, *m_cmdBuffer);

	de::MovePtr<tcu::TextureLevel>	resultTextureLevel	= readImage(*m_destination, m_params.dst.image);

	return checkTestResult(resultTextureLevel->getAccess());
}

tcu::TestStatus CopyImageToImage::checkTestResult (tcu::ConstPixelBufferAccess result)
{
	const tcu::Vec4	fThreshold (0.0f);
	const tcu::UVec4 uThreshold (0u);

	if (tcu::isCombinedDepthStencilType(result.getFormat().type))
	{
		if (tcu::hasDepthComponent(result.getFormat().order))
		{
			const tcu::Sampler::DepthStencilMode	mode				= tcu::Sampler::MODE_DEPTH;
			const tcu::ConstPixelBufferAccess		depthResult			= tcu::getEffectiveDepthStencilAccess(result, mode);
			const tcu::ConstPixelBufferAccess		expectedResult		= tcu::getEffectiveDepthStencilAccess(m_expectedTextureLevel->getAccess(), mode);

			if (isFloatFormat(result.getFormat()))
			{
				if (!tcu::floatThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expectedResult, depthResult, fThreshold, tcu::COMPARE_LOG_RESULT))
					return tcu::TestStatus::fail("CopiesAndBlitting test");
			}
			else
			{
				if (!tcu::intThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expectedResult, depthResult, uThreshold, tcu::COMPARE_LOG_RESULT))
					return tcu::TestStatus::fail("CopiesAndBlitting test");
			}
		}

		if (tcu::hasStencilComponent(result.getFormat().order))
		{
			const tcu::Sampler::DepthStencilMode	mode				= tcu::Sampler::MODE_STENCIL;
			const tcu::ConstPixelBufferAccess		stencilResult		= tcu::getEffectiveDepthStencilAccess(result, mode);
			const tcu::ConstPixelBufferAccess		expectedResult		= tcu::getEffectiveDepthStencilAccess(m_expectedTextureLevel->getAccess(), mode);

			if (isFloatFormat(result.getFormat()))
			{
				if (!tcu::floatThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expectedResult, stencilResult, fThreshold, tcu::COMPARE_LOG_RESULT))
					return tcu::TestStatus::fail("CopiesAndBlitting test");
			}
			else
			{
				if (!tcu::intThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", expectedResult, stencilResult, uThreshold, tcu::COMPARE_LOG_RESULT))
					return tcu::TestStatus::fail("CopiesAndBlitting test");
			}
		}
	}
	else
	{
		if (isFloatFormat(result.getFormat()))
		{
			if (!tcu::floatThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", m_expectedTextureLevel->getAccess(), result, fThreshold, tcu::COMPARE_LOG_RESULT))
				return tcu::TestStatus::fail("CopiesAndBlitting test");
		}
		else
		{
			if (!tcu::intThresholdCompare(m_context.getTestContext().getLog(), "Compare", "Result comparison", m_expectedTextureLevel->getAccess(), result, uThreshold, tcu::COMPARE_LOG_RESULT))
				return tcu::TestStatus::fail("CopiesAndBlitting test");
		}
	}

	return tcu::TestStatus::pass("CopiesAndBlitting test");
}

void CopyImageToImage::copyRegionToTextureLevel (tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region)
{
	VkOffset3D	srcOffset	= region.imageCopy.srcOffset;
	VkOffset3D	dstOffset	= region.imageCopy.dstOffset;
	VkExtent3D	extent		= region.imageCopy.extent;

	if (m_params.src.image.imageType == VK_IMAGE_TYPE_3D && m_params.dst.image.imageType == VK_IMAGE_TYPE_2D)
		dstOffset.z = srcOffset.z;
	if (m_params.src.image.imageType == VK_IMAGE_TYPE_2D && m_params.dst.image.imageType == VK_IMAGE_TYPE_3D)
	{
		srcOffset.z = dstOffset.z;
		extent.depth = std::max(region.imageCopy.extent.depth, region.imageCopy.srcSubresource.layerCount);
	}


	if (tcu::isCombinedDepthStencilType(src.getFormat().type))
	{
		DE_ASSERT(src.getFormat() == dst.getFormat());

		// Copy depth.
		if (tcu::hasDepthComponent(src.getFormat().order))
		{
			const tcu::ConstPixelBufferAccess	srcSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcOffset.z, extent.width, extent.height, extent.depth), tcu::Sampler::MODE_DEPTH);
			const tcu::PixelBufferAccess		dstSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(dst, dstOffset.x, dstOffset.y, dstOffset.z, extent.width, extent.height, extent.depth), tcu::Sampler::MODE_DEPTH);
			tcu::copy(dstSubRegion, srcSubRegion);
		}

		// Copy stencil.
		if (tcu::hasStencilComponent(src.getFormat().order))
		{
			const tcu::ConstPixelBufferAccess	srcSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcOffset.z, extent.width, extent.height, extent.depth), tcu::Sampler::MODE_STENCIL);
			const tcu::PixelBufferAccess		dstSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(dst, dstOffset.x, dstOffset.y, dstOffset.z, extent.width, extent.height, extent.depth), tcu::Sampler::MODE_STENCIL);
			tcu::copy(dstSubRegion, srcSubRegion);
		}
	}
	else
	{
		const tcu::ConstPixelBufferAccess	srcSubRegion		= tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcOffset.z, extent.width, extent.height, extent.depth);
		// CopyImage acts like a memcpy. Replace the destination format with the srcformat to use a memcpy.
		const tcu::PixelBufferAccess		dstWithSrcFormat	(srcSubRegion.getFormat(), dst.getSize(), dst.getDataPtr());
		const tcu::PixelBufferAccess		dstSubRegion		= tcu::getSubregion(dstWithSrcFormat, dstOffset.x, dstOffset.y, dstOffset.z, extent.width, extent.height, extent.depth);

		tcu::copy(dstSubRegion, srcSubRegion);
	}
}

class CopyImageToImageTestCase : public vkt::TestCase
{
public:
							CopyImageToImageTestCase	(tcu::TestContext&				testCtx,
														 const std::string&				name,
														 const std::string&				description,
														 const TestParams				params)
								: vkt::TestCase	(testCtx, name, description)
								, m_params		(params)
							{}

	virtual TestInstance*	createInstance				(Context&						context) const
							{
								return new CopyImageToImage(context, m_params);
							}
private:
	TestParams				m_params;
};

// Copy from buffer to buffer.

class CopyBufferToBuffer : public CopiesAndBlittingTestInstance
{
public:
								CopyBufferToBuffer			(Context& context, TestParams params);
	virtual tcu::TestStatus		iterate						(void);
private:
	virtual void				copyRegionToTextureLevel	(tcu::ConstPixelBufferAccess, tcu::PixelBufferAccess, CopyRegion);
	Move<VkBuffer>				m_source;
	de::MovePtr<Allocation>		m_sourceBufferAlloc;
	Move<VkBuffer>				m_destination;
	de::MovePtr<Allocation>		m_destinationBufferAlloc;
};

CopyBufferToBuffer::CopyBufferToBuffer (Context& context, TestParams params)
	: CopiesAndBlittingTestInstance	(context, params)
{
	const InstanceInterface&	vki					= context.getInstanceInterface();
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= context.getPhysicalDevice();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= context.getDefaultAllocator();

	// Create source buffer
	{
		const VkBufferCreateInfo	sourceBufferParams		=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			m_params.src.buffer.size,					// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_SRC_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		m_source				= createBuffer(vk, vkDevice, &sourceBufferParams);
		m_sourceBufferAlloc		= allocateBuffer(vki, vk, vkPhysDevice, vkDevice, *m_source, MemoryRequirement::HostVisible, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(vkDevice, *m_source, m_sourceBufferAlloc->getMemory(), m_sourceBufferAlloc->getOffset()));
	}

	// Create destination buffer
	{
		const VkBufferCreateInfo	destinationBufferParams	=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			m_params.dst.buffer.size,					// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_DST_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		m_destination				= createBuffer(vk, vkDevice, &destinationBufferParams);
		m_destinationBufferAlloc	= allocateBuffer(vki, vk, vkPhysDevice, vkDevice, *m_destination, MemoryRequirement::HostVisible, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(vkDevice, *m_destination, m_destinationBufferAlloc->getMemory(), m_destinationBufferAlloc->getOffset()));
	}
}

tcu::TestStatus CopyBufferToBuffer::iterate (void)
{
	const int srcLevelWidth		= (int)(m_params.src.buffer.size/4); // Here the format is VK_FORMAT_R32_UINT, we need to divide the buffer size by 4
	m_sourceTextureLevel		= de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(mapVkFormat(VK_FORMAT_R32_UINT), srcLevelWidth, 1));
	generateBuffer(m_sourceTextureLevel->getAccess(), srcLevelWidth, 1, 1, FILL_MODE_RED);

	const int dstLevelWidth		= (int)(m_params.dst.buffer.size/4);
	m_destinationTextureLevel	= de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(mapVkFormat(VK_FORMAT_R32_UINT), dstLevelWidth, 1));
	generateBuffer(m_destinationTextureLevel->getAccess(), dstLevelWidth, 1, 1, FILL_MODE_WHITE);

	generateExpectedResult();

	uploadBuffer(m_sourceTextureLevel->getAccess(), *m_sourceBufferAlloc);
	uploadBuffer(m_destinationTextureLevel->getAccess(), *m_destinationBufferAlloc);

	const DeviceInterface&		vk			= m_context.getDeviceInterface();
	const VkDevice				vkDevice	= m_context.getDevice();
	const VkQueue				queue		= m_context.getUniversalQueue();

	const VkBufferMemoryBarrier		srcBufferBarrier	=
	{
		VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,	// VkStructureType	sType;
		DE_NULL,									// const void*		pNext;
		VK_ACCESS_HOST_WRITE_BIT,					// VkAccessFlags	srcAccessMask;
		VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags	dstAccessMask;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			dstQueueFamilyIndex;
		*m_source,									// VkBuffer			buffer;
		0u,											// VkDeviceSize		offset;
		m_params.src.buffer.size					// VkDeviceSize		size;
	};

	const VkBufferMemoryBarrier		dstBufferBarrier	=
	{
		VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,	// VkStructureType	sType;
		DE_NULL,									// const void*		pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags	srcAccessMask;
		VK_ACCESS_HOST_READ_BIT,					// VkAccessFlags	dstAccessMask;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			dstQueueFamilyIndex;
		*m_destination,								// VkBuffer			buffer;
		0u,											// VkDeviceSize		offset;
		m_params.dst.buffer.size					// VkDeviceSize		size;
	};

	std::vector<VkBufferCopy>		bufferCopies;
	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		bufferCopies.push_back(m_params.regions[i].bufferCopy);

	const VkCommandBufferBeginInfo	cmdBufferBeginInfo	=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 1, &srcBufferBarrier, 0, (const VkImageMemoryBarrier*)DE_NULL);
	vk.cmdCopyBuffer(*m_cmdBuffer, m_source.get(), m_destination.get(), (deUint32)m_params.regions.size(), &bufferCopies[0]);
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 1, &dstBufferBarrier, 0, (const VkImageMemoryBarrier*)DE_NULL);
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));
	submitCommandsAndWait(vk, vkDevice, queue, *m_cmdBuffer);



	// Read buffer data
	de::MovePtr<tcu::TextureLevel>	resultLevel		(new tcu::TextureLevel(mapVkFormat(VK_FORMAT_R32_UINT), dstLevelWidth, 1));
	invalidateMappedMemoryRange(vk, vkDevice, m_destinationBufferAlloc->getMemory(), m_destinationBufferAlloc->getOffset(), m_params.dst.buffer.size);
	tcu::copy(*resultLevel, tcu::ConstPixelBufferAccess(resultLevel->getFormat(), resultLevel->getSize(), m_destinationBufferAlloc->getHostPtr()));

	return checkTestResult(resultLevel->getAccess());
}

void CopyBufferToBuffer::copyRegionToTextureLevel (tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region)
{
	deMemcpy((deUint8*) dst.getDataPtr() + region.bufferCopy.dstOffset,
			 (deUint8*) src.getDataPtr() + region.bufferCopy.srcOffset,
			 (size_t)region.bufferCopy.size);
}

class BufferToBufferTestCase : public vkt::TestCase
{
public:
							BufferToBufferTestCase	(tcu::TestContext&	testCtx,
													 const std::string&	name,
													 const std::string&	description,
													 const TestParams	params)
								: vkt::TestCase	(testCtx, name, description)
								, m_params		(params)
							{}

	virtual TestInstance*	createInstance			(Context& context) const
							{
								return new CopyBufferToBuffer(context, m_params);
							}
private:
	TestParams				m_params;
};

// Copy from image to buffer.

class CopyImageToBuffer : public CopiesAndBlittingTestInstance
{
public:
								CopyImageToBuffer			(Context&	context,
															 TestParams	testParams);
	virtual tcu::TestStatus		iterate						(void);
private:
	virtual void				copyRegionToTextureLevel	(tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region);

	tcu::TextureFormat			m_textureFormat;
	VkDeviceSize				m_bufferSize;

	Move<VkImage>				m_source;
	de::MovePtr<Allocation>		m_sourceImageAlloc;
	Move<VkBuffer>				m_destination;
	de::MovePtr<Allocation>		m_destinationBufferAlloc;
};

CopyImageToBuffer::CopyImageToBuffer (Context& context, TestParams testParams)
	: CopiesAndBlittingTestInstance(context, testParams)
	, m_textureFormat(mapVkFormat(testParams.src.image.format))
	, m_bufferSize(m_params.dst.buffer.size * tcu::getPixelSize(m_textureFormat))
{
	const InstanceInterface&	vki					= context.getInstanceInterface();
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= context.getPhysicalDevice();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= context.getDefaultAllocator();

	// Create source image
	{
		const VkImageCreateInfo		sourceImageParams		=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.src.image.imageType,			// VkImageType			imageType;
			m_params.src.image.format,				// VkFormat				format;
			getExtent3D(m_params.src.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.src.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_source			= createImage(vk, vkDevice, &sourceImageParams);
		m_sourceImageAlloc	= allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_source, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_source, m_sourceImageAlloc->getMemory(), m_sourceImageAlloc->getOffset()));
	}

	// Create destination buffer
	{
		const VkBufferCreateInfo	destinationBufferParams	=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			m_bufferSize,								// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_DST_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		m_destination				= createBuffer(vk, vkDevice, &destinationBufferParams);
		m_destinationBufferAlloc	= allocateBuffer(vki, vk, vkPhysDevice, vkDevice, *m_destination, MemoryRequirement::HostVisible, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(vkDevice, *m_destination, m_destinationBufferAlloc->getMemory(), m_destinationBufferAlloc->getOffset()));
	}
}

tcu::TestStatus CopyImageToBuffer::iterate (void)
{
	m_sourceTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(m_textureFormat,
																				m_params.src.image.extent.width,
																				m_params.src.image.extent.height,
																				m_params.src.image.extent.depth));
	generateBuffer(m_sourceTextureLevel->getAccess(), m_params.src.image.extent.width, m_params.src.image.extent.height, m_params.src.image.extent.depth);
	m_destinationTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(m_textureFormat, (int)m_params.dst.buffer.size, 1));
	generateBuffer(m_destinationTextureLevel->getAccess(), (int)m_params.dst.buffer.size, 1, 1);

	generateExpectedResult();

	uploadImage(m_sourceTextureLevel->getAccess(), *m_source, m_params.src.image);
	uploadBuffer(m_destinationTextureLevel->getAccess(), *m_destinationBufferAlloc);

	const DeviceInterface&		vk			= m_context.getDeviceInterface();
	const VkDevice				vkDevice	= m_context.getDevice();
	const VkQueue				queue		= m_context.getUniversalQueue();

	// Barriers for copying image to buffer
	const VkImageMemoryBarrier		imageBarrier		=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
		VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,		// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		*m_source,									// VkImage					image;
		{											// VkImageSubresourceRange	subresourceRange;
			getAspectFlags(m_textureFormat),	// VkImageAspectFlags	aspectMask;
			0u,								// deUint32				baseMipLevel;
			1u,								// deUint32				mipLevels;
			0u,								// deUint32				baseArraySlice;
			1u								// deUint32				arraySize;
		}
	};

	const VkBufferMemoryBarrier		bufferBarrier		=
	{
		VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,	// VkStructureType	sType;
		DE_NULL,									// const void*		pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags	srcAccessMask;
		VK_ACCESS_HOST_READ_BIT,					// VkAccessFlags	dstAccessMask;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32			dstQueueFamilyIndex;
		*m_destination,								// VkBuffer			buffer;
		0u,											// VkDeviceSize		offset;
		m_bufferSize								// VkDeviceSize		size;
	};

	// Copy from image to buffer
	std::vector<VkBufferImageCopy>	bufferImageCopies;
	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		bufferImageCopies.push_back(m_params.regions[i].bufferImageCopy);

	const VkCommandBufferBeginInfo	cmdBufferBeginInfo	=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &imageBarrier);
	vk.cmdCopyImageToBuffer(*m_cmdBuffer, m_source.get(), VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, m_destination.get(), (deUint32)m_params.regions.size(), &bufferImageCopies[0]);
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 1, &bufferBarrier, 0, (const VkImageMemoryBarrier*)DE_NULL);
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));

	submitCommandsAndWait (vk, vkDevice, queue, *m_cmdBuffer);

	// Read buffer data
	de::MovePtr<tcu::TextureLevel>	resultLevel		(new tcu::TextureLevel(m_textureFormat, (int)m_params.dst.buffer.size, 1));
	invalidateMappedMemoryRange(vk, vkDevice, m_destinationBufferAlloc->getMemory(), m_destinationBufferAlloc->getOffset(), m_bufferSize);
	tcu::copy(*resultLevel, tcu::ConstPixelBufferAccess(resultLevel->getFormat(), resultLevel->getSize(), m_destinationBufferAlloc->getHostPtr()));

	return checkTestResult(resultLevel->getAccess());
}

class CopyImageToBufferTestCase : public vkt::TestCase
{
public:
							CopyImageToBufferTestCase	(tcu::TestContext&		testCtx,
														 const std::string&		name,
														 const std::string&		description,
														 const TestParams		params)
								: vkt::TestCase	(testCtx, name, description)
								, m_params		(params)
							{}

	virtual TestInstance*	createInstance				(Context&				context) const
							{
								return new CopyImageToBuffer(context, m_params);
							}
private:
	TestParams				m_params;
};

void CopyImageToBuffer::copyRegionToTextureLevel (tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region)
{
	deUint32			rowLength	= region.bufferImageCopy.bufferRowLength;
	if (!rowLength)
		rowLength = region.bufferImageCopy.imageExtent.width;

	deUint32			imageHeight	= region.bufferImageCopy.bufferImageHeight;
	if (!imageHeight)
		imageHeight = region.bufferImageCopy.imageExtent.height;

	const int			texelSize	= src.getFormat().getPixelSize();
	const VkExtent3D	extent		= region.bufferImageCopy.imageExtent;
	const VkOffset3D	srcOffset	= region.bufferImageCopy.imageOffset;
	const int			texelOffset	= (int) region.bufferImageCopy.bufferOffset / texelSize;

	for (deUint32 z = 0; z < extent.depth; z++)
	{
		for (deUint32 y = 0; y < extent.height; y++)
		{
			int									texelIndex		= texelOffset + (z * imageHeight + y) *	rowLength;
			const tcu::ConstPixelBufferAccess	srcSubRegion	= tcu::getSubregion(src, srcOffset.x, srcOffset.y + y, srcOffset.z + z,
																					region.bufferImageCopy.imageExtent.width, 1, 1);
			const tcu::PixelBufferAccess		dstSubRegion	= tcu::getSubregion(dst, texelIndex, 0, region.bufferImageCopy.imageExtent.width, 1);
			tcu::copy(dstSubRegion, srcSubRegion);
		}
	}
}

// Copy from buffer to image.

class CopyBufferToImage : public CopiesAndBlittingTestInstance
{
public:
								CopyBufferToImage			(Context&	context,
															 TestParams	testParams);
	virtual tcu::TestStatus		iterate						(void);
private:
	virtual void				copyRegionToTextureLevel	(tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region);

	tcu::TextureFormat			m_textureFormat;
	VkDeviceSize				m_bufferSize;

	Move<VkBuffer>				m_source;
	de::MovePtr<Allocation>		m_sourceBufferAlloc;
	Move<VkImage>				m_destination;
	de::MovePtr<Allocation>		m_destinationImageAlloc;
};

CopyBufferToImage::CopyBufferToImage (Context& context, TestParams testParams)
	: CopiesAndBlittingTestInstance(context, testParams)
	, m_textureFormat(mapVkFormat(testParams.dst.image.format))
	, m_bufferSize(m_params.src.buffer.size * tcu::getPixelSize(m_textureFormat))
{
	const InstanceInterface&	vki					= context.getInstanceInterface();
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= context.getPhysicalDevice();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= context.getDefaultAllocator();

	// Create source buffer
	{
		const VkBufferCreateInfo	sourceBufferParams		=
		{
			VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,		// VkStructureType		sType;
			DE_NULL,									// const void*			pNext;
			0u,											// VkBufferCreateFlags	flags;
			m_bufferSize,								// VkDeviceSize			size;
			VK_BUFFER_USAGE_TRANSFER_SRC_BIT,			// VkBufferUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,					// VkSharingMode		sharingMode;
			1u,											// deUint32				queueFamilyIndexCount;
			&queueFamilyIndex,							// const deUint32*		pQueueFamilyIndices;
		};

		m_source				= createBuffer(vk, vkDevice, &sourceBufferParams);
		m_sourceBufferAlloc		= allocateBuffer(vki, vk, vkPhysDevice, vkDevice, *m_source, MemoryRequirement::HostVisible, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindBufferMemory(vkDevice, *m_source, m_sourceBufferAlloc->getMemory(), m_sourceBufferAlloc->getOffset()));
	}

	// Create destination image
	{
		const VkImageCreateInfo		destinationImageParams	=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.dst.image.imageType,			// VkImageType			imageType;
			m_params.dst.image.format,				// VkFormat				format;
			getExtent3D(m_params.dst.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.dst.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_destination			= createImage(vk, vkDevice, &destinationImageParams);
		m_destinationImageAlloc	= allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_destination, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_destination, m_destinationImageAlloc->getMemory(), m_destinationImageAlloc->getOffset()));
	}
}

tcu::TestStatus CopyBufferToImage::iterate (void)
{
	m_sourceTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(m_textureFormat, (int)m_params.src.buffer.size, 1));
	generateBuffer(m_sourceTextureLevel->getAccess(), (int)m_params.src.buffer.size, 1, 1);
	m_destinationTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(m_textureFormat,
																					m_params.dst.image.extent.width,
																					m_params.dst.image.extent.height,
																					m_params.dst.image.extent.depth));

	generateBuffer(m_destinationTextureLevel->getAccess(), m_params.dst.image.extent.width, m_params.dst.image.extent.height, m_params.dst.image.extent.depth);

	generateExpectedResult();

	uploadBuffer(m_sourceTextureLevel->getAccess(), *m_sourceBufferAlloc);
	uploadImage(m_destinationTextureLevel->getAccess(), *m_destination, m_params.dst.image);

	const DeviceInterface&		vk			= m_context.getDeviceInterface();
	const VkDevice				vkDevice	= m_context.getDevice();
	const VkQueue				queue		= m_context.getUniversalQueue();

	const VkImageMemoryBarrier	imageBarrier	=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		*m_destination,								// VkImage					image;
		{											// VkImageSubresourceRange	subresourceRange;
			getAspectFlags(m_textureFormat),	// VkImageAspectFlags	aspectMask;
			0u,								// deUint32				baseMipLevel;
			1u,								// deUint32				mipLevels;
			0u,								// deUint32				baseArraySlice;
			1u								// deUint32				arraySize;
		}
	};

	// Copy from buffer to image
	std::vector<VkBufferImageCopy>		bufferImageCopies;
	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		bufferImageCopies.push_back(m_params.regions[i].bufferImageCopy);

	const VkCommandBufferBeginInfo	cmdBufferBeginInfo	=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &imageBarrier);
	vk.cmdCopyBufferToImage(*m_cmdBuffer, m_source.get(), m_destination.get(), VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, (deUint32)m_params.regions.size(), bufferImageCopies.data());
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));

	submitCommandsAndWait (vk, vkDevice, queue, *m_cmdBuffer);

	de::MovePtr<tcu::TextureLevel>	resultLevel	= readImage(*m_destination, m_params.dst.image);

	return checkTestResult(resultLevel->getAccess());
}

class CopyBufferToImageTestCase : public vkt::TestCase
{
public:
							CopyBufferToImageTestCase	(tcu::TestContext&		testCtx,
														 const std::string&		name,
														 const std::string&		description,
														 const TestParams		params)
								: vkt::TestCase	(testCtx, name, description)
								, m_params		(params)
							{}

	virtual					~CopyBufferToImageTestCase	(void) {}

	virtual TestInstance*	createInstance				(Context&				context) const
							{
								return new CopyBufferToImage(context, m_params);
							}
private:
	TestParams				m_params;
};

void CopyBufferToImage::copyRegionToTextureLevel (tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region)
{
	deUint32			rowLength	= region.bufferImageCopy.bufferRowLength;
	if (!rowLength)
		rowLength = region.bufferImageCopy.imageExtent.width;

	deUint32			imageHeight	= region.bufferImageCopy.bufferImageHeight;
	if (!imageHeight)
		imageHeight = region.bufferImageCopy.imageExtent.height;

	const int			texelSize	= dst.getFormat().getPixelSize();
	const VkExtent3D	extent		= region.bufferImageCopy.imageExtent;
	const VkOffset3D	dstOffset	= region.bufferImageCopy.imageOffset;
	const int			texelOffset	= (int) region.bufferImageCopy.bufferOffset / texelSize;

	for (deUint32 z = 0; z < extent.depth; z++)
	{
		for (deUint32 y = 0; y < extent.height; y++)
		{
			int									texelIndex		= texelOffset + (z * imageHeight + y) *	rowLength;
			const tcu::ConstPixelBufferAccess	srcSubRegion	= tcu::getSubregion(src, texelIndex, 0, region.bufferImageCopy.imageExtent.width, 1);
			const tcu::PixelBufferAccess		dstSubRegion	= tcu::getSubregion(dst, dstOffset.x, dstOffset.y + y, dstOffset.z + z,
																					region.bufferImageCopy.imageExtent.width, 1, 1);
			tcu::copy(dstSubRegion, srcSubRegion);
		}
	}
}

// Copy from image to image with scaling.

class BlittingImages : public CopiesAndBlittingTestInstance
{
public:
										BlittingImages					(Context&	context,
																		 TestParams params);
	virtual tcu::TestStatus				iterate							(void);
protected:
	virtual tcu::TestStatus				checkTestResult					(tcu::ConstPixelBufferAccess result);
	virtual void						copyRegionToTextureLevel		(tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region);
	virtual void						generateExpectedResult			(void);
private:
	bool								checkLinearFilteredResult		(const tcu::ConstPixelBufferAccess&	result,
																		 const tcu::ConstPixelBufferAccess&	clampedReference,
																		 const tcu::ConstPixelBufferAccess&	unclampedReference,
																		 const tcu::TextureFormat&			sourceFormat);
	bool								checkNearestFilteredResult		(const tcu::ConstPixelBufferAccess&	result,
																		 const tcu::ConstPixelBufferAccess& source);

	Move<VkImage>						m_source;
	de::MovePtr<Allocation>				m_sourceImageAlloc;
	Move<VkImage>						m_destination;
	de::MovePtr<Allocation>				m_destinationImageAlloc;

	de::MovePtr<tcu::TextureLevel>		m_unclampedExpectedTextureLevel;
};

BlittingImages::BlittingImages (Context& context, TestParams params)
	: CopiesAndBlittingTestInstance(context, params)
{
	const InstanceInterface&	vki					= context.getInstanceInterface();
	const DeviceInterface&		vk					= context.getDeviceInterface();
	const VkPhysicalDevice		vkPhysDevice		= context.getPhysicalDevice();
	const VkDevice				vkDevice			= context.getDevice();
	const deUint32				queueFamilyIndex	= context.getUniversalQueueFamilyIndex();
	Allocator&					memAlloc			= context.getDefaultAllocator();

	VkImageFormatProperties properties;
	if ((context.getInstanceInterface().getPhysicalDeviceImageFormatProperties (context.getPhysicalDevice(),
																				m_params.src.image.format,
																				VK_IMAGE_TYPE_2D,
																				VK_IMAGE_TILING_OPTIMAL,
																				VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
																				0,
																				&properties) == VK_ERROR_FORMAT_NOT_SUPPORTED) ||
		(context.getInstanceInterface().getPhysicalDeviceImageFormatProperties (context.getPhysicalDevice(),
																				m_params.dst.image.format,
																				VK_IMAGE_TYPE_2D,
																				VK_IMAGE_TILING_OPTIMAL,
																				VK_IMAGE_USAGE_TRANSFER_DST_BIT,
																				0,
																				&properties) == VK_ERROR_FORMAT_NOT_SUPPORTED))
	{
		TCU_THROW(NotSupportedError, "Format not supported");
	}

	VkFormatProperties srcFormatProperties;
	context.getInstanceInterface().getPhysicalDeviceFormatProperties(context.getPhysicalDevice(), m_params.src.image.format, &srcFormatProperties);
	if (!(srcFormatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_BLIT_SRC_BIT))
	{
		TCU_THROW(NotSupportedError, "Format feature blit source not supported");
	}

	VkFormatProperties dstFormatProperties;
	context.getInstanceInterface().getPhysicalDeviceFormatProperties(context.getPhysicalDevice(), m_params.dst.image.format, &dstFormatProperties);
	if (!(dstFormatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_BLIT_DST_BIT))
	{
		TCU_THROW(NotSupportedError, "Format feature blit destination not supported");
	}

	if (m_params.filter == VK_FILTER_LINEAR)
	{
		if (!(srcFormatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT))
			TCU_THROW(NotSupportedError, "Source format feature sampled image filter linear not supported");
		if (!(dstFormatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT))
			TCU_THROW(NotSupportedError, "Destination format feature sampled image filter linear not supported");
	}

	// Create source image
	{
		const VkImageCreateInfo		sourceImageParams		=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.src.image.imageType,			// VkImageType			imageType;
			m_params.src.image.format,				// VkFormat				format;
			getExtent3D(m_params.src.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.src.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_source = createImage(vk, vkDevice, &sourceImageParams);
		m_sourceImageAlloc = allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_source, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_source, m_sourceImageAlloc->getMemory(), m_sourceImageAlloc->getOffset()));
	}

	// Create destination image
	{
		const VkImageCreateInfo		destinationImageParams	=
		{
			VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,	// VkStructureType		sType;
			DE_NULL,								// const void*			pNext;
			0u,										// VkImageCreateFlags	flags;
			m_params.dst.image.imageType,			// VkImageType			imageType;
			m_params.dst.image.format,				// VkFormat				format;
			getExtent3D(m_params.dst.image),		// VkExtent3D			extent;
			1u,										// deUint32				mipLevels;
			getArraySize(m_params.dst.image),		// deUint32				arraySize;
			VK_SAMPLE_COUNT_1_BIT,					// deUint32				samples;
			VK_IMAGE_TILING_OPTIMAL,				// VkImageTiling		tiling;
			VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
				VK_IMAGE_USAGE_TRANSFER_DST_BIT,	// VkImageUsageFlags	usage;
			VK_SHARING_MODE_EXCLUSIVE,				// VkSharingMode		sharingMode;
			1u,										// deUint32				queueFamilyCount;
			&queueFamilyIndex,						// const deUint32*		pQueueFamilyIndices;
			VK_IMAGE_LAYOUT_UNDEFINED,				// VkImageLayout		initialLayout;
		};

		m_destination = createImage(vk, vkDevice, &destinationImageParams);
		m_destinationImageAlloc = allocateImage(vki, vk, vkPhysDevice, vkDevice, *m_destination, MemoryRequirement::Any, memAlloc, m_params.allocationKind);
		VK_CHECK(vk.bindImageMemory(vkDevice, *m_destination, m_destinationImageAlloc->getMemory(), m_destinationImageAlloc->getOffset()));
	}
}

tcu::TestStatus BlittingImages::iterate (void)
{
	const tcu::TextureFormat	srcTcuFormat		= mapVkFormat(m_params.src.image.format);
	const tcu::TextureFormat	dstTcuFormat		= mapVkFormat(m_params.dst.image.format);
	m_sourceTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(srcTcuFormat,
																				m_params.src.image.extent.width,
																				m_params.src.image.extent.height,
																				m_params.src.image.extent.depth));
	generateBuffer(m_sourceTextureLevel->getAccess(), m_params.src.image.extent.width, m_params.src.image.extent.height, m_params.src.image.extent.depth, FILL_MODE_GRADIENT);
	m_destinationTextureLevel = de::MovePtr<tcu::TextureLevel>(new tcu::TextureLevel(dstTcuFormat,
																					 (int)m_params.dst.image.extent.width,
																					 (int)m_params.dst.image.extent.height,
																					 (int)m_params.dst.image.extent.depth));
	generateBuffer(m_destinationTextureLevel->getAccess(), m_params.dst.image.extent.width, m_params.dst.image.extent.height, m_params.dst.image.extent.depth, FILL_MODE_WHITE);
	generateExpectedResult();

	uploadImage(m_sourceTextureLevel->getAccess(), m_source.get(), m_params.src.image);
	uploadImage(m_destinationTextureLevel->getAccess(), m_destination.get(), m_params.dst.image);

	const DeviceInterface&		vk					= m_context.getDeviceInterface();
	const VkDevice				vkDevice			= m_context.getDevice();
	const VkQueue				queue				= m_context.getUniversalQueue();

	std::vector<VkImageBlit>	regions;
	for (deUint32 i = 0; i < m_params.regions.size(); i++)
		regions.push_back(m_params.regions[i].imageBlit);

	// Barriers for copying image to buffer
	const VkImageMemoryBarrier		srcImageBarrier		=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_READ_BIT,				// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
		m_params.src.image.operationLayout,			// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		m_source.get(),								// VkImage					image;
		{											// VkImageSubresourceRange	subresourceRange;
			getAspectFlags(srcTcuFormat),	// VkImageAspectFlags	aspectMask;
			0u,								// deUint32				baseMipLevel;
			1u,								// deUint32				mipLevels;
			0u,								// deUint32				baseArraySlice;
			1u								// deUint32				arraySize;
		}
	};

	const VkImageMemoryBarrier		dstImageBarrier		=
	{
		VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,		// VkStructureType			sType;
		DE_NULL,									// const void*				pNext;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			srcAccessMask;
		VK_ACCESS_TRANSFER_WRITE_BIT,				// VkAccessFlags			dstAccessMask;
		VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,		// VkImageLayout			oldLayout;
		m_params.dst.image.operationLayout,			// VkImageLayout			newLayout;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					srcQueueFamilyIndex;
		VK_QUEUE_FAMILY_IGNORED,					// deUint32					dstQueueFamilyIndex;
		m_destination.get(),						// VkImage					image;
		{											// VkImageSubresourceRange	subresourceRange;
			getAspectFlags(dstTcuFormat),	// VkImageAspectFlags	aspectMask;
			0u,								// deUint32				baseMipLevel;
			1u,								// deUint32				mipLevels;
			0u,								// deUint32				baseArraySlice;
			1u								// deUint32				arraySize;
		}
	};

	const VkCommandBufferBeginInfo	cmdBufferBeginInfo	=
	{
		VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,			// VkStructureType					sType;
		DE_NULL,												// const void*						pNext;
		VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,			// VkCommandBufferUsageFlags		flags;
		(const VkCommandBufferInheritanceInfo*)DE_NULL,
	};

	VK_CHECK(vk.beginCommandBuffer(*m_cmdBuffer, &cmdBufferBeginInfo));
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &srcImageBarrier);
	vk.cmdBlitImage(*m_cmdBuffer, m_source.get(), m_params.src.image.operationLayout, m_destination.get(), m_params.dst.image.operationLayout, (deUint32)m_params.regions.size(), &regions[0], m_params.filter);
	vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, (VkDependencyFlags)0, 0, (const VkMemoryBarrier*)DE_NULL, 0, (const VkBufferMemoryBarrier*)DE_NULL, 1, &dstImageBarrier);
	VK_CHECK(vk.endCommandBuffer(*m_cmdBuffer));
	submitCommandsAndWait(vk, vkDevice, queue, *m_cmdBuffer);

	de::MovePtr<tcu::TextureLevel> resultTextureLevel = readImage(*m_destination, m_params.dst.image);

	return checkTestResult(resultTextureLevel->getAccess());
}

static float calculateFloatConversionError (int srcBits)
{
	if (srcBits > 0)
	{
		const int	clampedBits	= de::clamp<int>(srcBits, 0, 32);
		const float	srcMaxValue	= de::max((float)(1ULL<<clampedBits) - 1.0f, 1.0f);
		const float	error		= 1.0f / srcMaxValue;

		return de::clamp<float>(error, 0.0f, 1.0f);
	}
	else
		return 1.0f;
}

tcu::Vec4 getFormatThreshold (const tcu::TextureFormat& format)
{
	tcu::Vec4 threshold(0.01f);

	switch (format.type)
	{
	case tcu::TextureFormat::HALF_FLOAT:
		threshold = tcu::Vec4(0.005f);
		break;

	case tcu::TextureFormat::FLOAT:
	case tcu::TextureFormat::FLOAT64:
		threshold = tcu::Vec4(0.001f);
		break;

	case tcu::TextureFormat::UNSIGNED_INT_11F_11F_10F_REV:
		threshold = tcu::Vec4(0.02f, 0.02f, 0.0625f, 1.0f);
		break;

	case tcu::TextureFormat::UNSIGNED_INT_999_E5_REV:
		threshold = tcu::Vec4(0.05f, 0.05f, 0.05f, 1.0f);
		break;

	default:
		const tcu::IVec4 bits = tcu::getTextureFormatMantissaBitDepth(format);
		threshold = tcu::Vec4(calculateFloatConversionError(bits.x()),
				      calculateFloatConversionError(bits.y()),
				      calculateFloatConversionError(bits.z()),
				      calculateFloatConversionError(bits.w()));
	}

	// Return value matching the channel order specified by the format
	if (format.order == tcu::TextureFormat::BGR || format.order == tcu::TextureFormat::BGRA)
		return threshold.swizzle(2, 1, 0, 3);
	else
		return threshold;
}

bool BlittingImages::checkLinearFilteredResult (const tcu::ConstPixelBufferAccess&	result,
												const tcu::ConstPixelBufferAccess&	clampedExpected,
												const tcu::ConstPixelBufferAccess&	unclampedExpected,
												const tcu::TextureFormat&			srcFormat)
{
	tcu::TestLog&				log			(m_context.getTestContext().getLog());
	const tcu::TextureFormat	dstFormat	= result.getFormat();
	bool						isOk		= false;

	log << tcu::TestLog::Section("ClampedSourceImage", "Region with clamped edges on source image.");

	if (isFloatFormat(dstFormat))
	{
		const bool		srcIsSRGB	= tcu::isSRGB(srcFormat);
		const tcu::Vec4	srcMaxDiff	= getFormatThreshold(srcFormat) * tcu::Vec4(srcIsSRGB ? 2.0f : 1.0f);
		const tcu::Vec4	dstMaxDiff	= getFormatThreshold(dstFormat);
		const tcu::Vec4	threshold	= tcu::max(srcMaxDiff, dstMaxDiff);

		isOk = tcu::floatThresholdCompare(log, "Compare", "Result comparsion", clampedExpected, result, threshold, tcu::COMPARE_LOG_RESULT);
		log << tcu::TestLog::EndSection;

		if (!isOk)
		{
			log << tcu::TestLog::Section("NonClampedSourceImage", "Region with non-clamped edges on source image.");
			isOk = tcu::floatThresholdCompare(log, "Compare", "Result comparsion", unclampedExpected, result, threshold, tcu::COMPARE_LOG_RESULT);
			log << tcu::TestLog::EndSection;
		}
	}
	else
	{
		tcu::UVec4	threshold;
		// Calculate threshold depending on channel width of destination format.
		const tcu::IVec4	bitDepth	= tcu::getTextureFormatBitDepth(dstFormat);
		for (deUint32 i = 0; i < 4; ++i)
			threshold[i] = de::max( (0x1 << bitDepth[i]) / 256, 1);

		isOk = tcu::intThresholdCompare(log, "Compare", "Result comparsion", clampedExpected, result, threshold, tcu::COMPARE_LOG_RESULT);
		log << tcu::TestLog::EndSection;

		if (!isOk)
		{
			log << tcu::TestLog::Section("NonClampedSourceImage", "Region with non-clamped edges on source image.");
			isOk = tcu::intThresholdCompare(log, "Compare", "Result comparsion", unclampedExpected, result, threshold, tcu::COMPARE_LOG_RESULT);
			log << tcu::TestLog::EndSection;
		}
	}

	return isOk;
}

//! Utility to encapsulate coordinate computation and loops.
struct CompareEachPixelInEachRegion
{
	virtual		 ~CompareEachPixelInEachRegion  (void) {}
	virtual bool compare						(const void* pUserData, const int x, const int y, const tcu::Vec2& srcNormCoord) const = 0;

	bool forEach (const void*						pUserData,
				  const std::vector<CopyRegion>&	regions,
				  const int							sourceWidth,
				  const int							sourceHeight,
				  const tcu::PixelBufferAccess&		errorMask) const
	{
		bool compareOk = true;

		for (std::vector<CopyRegion>::const_iterator regionIter = regions.begin(); regionIter != regions.end(); ++regionIter)
		{
			const VkImageBlit& blit = regionIter->imageBlit;

			const int	dx		= deSign32(blit.dstOffsets[1].x - blit.dstOffsets[0].x);
			const int	dy		= deSign32(blit.dstOffsets[1].y - blit.dstOffsets[0].y);
			const float	xScale	= static_cast<float>(blit.srcOffsets[1].x - blit.srcOffsets[0].x) / static_cast<float>(blit.dstOffsets[1].x - blit.dstOffsets[0].x);
			const float	yScale	= static_cast<float>(blit.srcOffsets[1].y - blit.srcOffsets[0].y) / static_cast<float>(blit.dstOffsets[1].y - blit.dstOffsets[0].y);
			const float srcInvW	= 1.0f / static_cast<float>(sourceWidth);
			const float srcInvH	= 1.0f / static_cast<float>(sourceHeight);

			for (int y = blit.dstOffsets[0].y; y < blit.dstOffsets[1].y; y += dy)
			for (int x = blit.dstOffsets[0].x; x < blit.dstOffsets[1].x; x += dx)
			{
				const tcu::Vec2 srcNormCoord
				(
					(xScale * (static_cast<float>(x - blit.dstOffsets[0].x) + 0.5f) + static_cast<float>(blit.srcOffsets[0].x)) * srcInvW,
					(yScale * (static_cast<float>(y - blit.dstOffsets[0].y) + 0.5f) + static_cast<float>(blit.srcOffsets[0].y)) * srcInvH
				);

				if (!compare(pUserData, x, y, srcNormCoord))
				{
					errorMask.setPixel(tcu::Vec4(1.0f, 0.0f, 0.0f, 1.0f), x, y);
					compareOk = false;
				}
			}
		}
		return compareOk;
	}
};

tcu::Vec4 getFloatOrFixedPointFormatThreshold (const tcu::TextureFormat& format)
{
	const tcu::TextureChannelClass	channelClass	= tcu::getTextureChannelClass(format.type);
	const tcu::IVec4				bitDepth		= tcu::getTextureFormatBitDepth(format);

	if (channelClass == tcu::TEXTURECHANNELCLASS_FLOATING_POINT)
	{
		return getFormatThreshold(format);
	}
	else if (channelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT ||
			 channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT)
	{
		const bool	isSigned	= (channelClass == tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT);
		const float	range		= isSigned ? 1.0f - (-1.0f)
										   : 1.0f -   0.0f;

		tcu::Vec4 v;
		for (int i = 0; i < 4; ++i)
		{
			if (bitDepth[i] == 0)
				v[i] = 1.0f;
			else
				v[i] = range / static_cast<float>((1 << bitDepth[i]) - 1);
		}
		return v;
	}
	else
	{
		DE_ASSERT(0);
		return tcu::Vec4();
	}
}

bool floatNearestBlitCompare (const tcu::ConstPixelBufferAccess&	source,
							  const tcu::ConstPixelBufferAccess&	result,
							  const tcu::PixelBufferAccess&			errorMask,
							  const std::vector<CopyRegion>&		regions)
{
	const tcu::Sampler		sampler		(tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::NEAREST, tcu::Sampler::NEAREST);
	tcu::LookupPrecision	precision;

	{
		const tcu::IVec4	dstBitDepth	= tcu::getTextureFormatBitDepth(result.getFormat());
		const tcu::Vec4		srcMaxDiff	= getFloatOrFixedPointFormatThreshold(source.getFormat());
		const tcu::Vec4		dstMaxDiff	= getFloatOrFixedPointFormatThreshold(result.getFormat());

		precision.colorMask		 = tcu::notEqual(dstBitDepth, tcu::IVec4(0));
		precision.colorThreshold = tcu::max(srcMaxDiff, dstMaxDiff);
	}

	const struct Capture
	{
		const tcu::ConstPixelBufferAccess&	source;
		const tcu::ConstPixelBufferAccess&	result;
		const tcu::Sampler&					sampler;
		const tcu::LookupPrecision&			precision;
		const bool							isSRGB;
	} capture =
	{
		source, result, sampler, precision, tcu::isSRGB(result.getFormat())
	};

	const struct Loop : CompareEachPixelInEachRegion
	{
		Loop (void) {}

		bool compare (const void* pUserData, const int x, const int y, const tcu::Vec2& srcNormCoord) const
		{
			const Capture&					c					= *static_cast<const Capture*>(pUserData);
			const tcu::TexLookupScaleMode	lookupScaleDontCare	= tcu::TEX_LOOKUP_SCALE_MINIFY;
			tcu::Vec4						dstColor			= c.result.getPixel(x, y);

			// TexLookupVerifier performs a conversion to linear space, so we have to as well
			if (c.isSRGB)
				dstColor = tcu::sRGBToLinear(dstColor);

			return tcu::isLevel2DLookupResultValid(c.source, c.sampler, lookupScaleDontCare, c.precision, srcNormCoord, 0, dstColor);
		}
	} loop;

	return loop.forEach(&capture, regions, source.getWidth(), source.getHeight(), errorMask);
}

bool intNearestBlitCompare (const tcu::ConstPixelBufferAccess&	source,
							const tcu::ConstPixelBufferAccess&	result,
							const tcu::PixelBufferAccess&		errorMask,
							const std::vector<CopyRegion>&		regions)
{
	const tcu::Sampler		sampler		(tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::NEAREST, tcu::Sampler::NEAREST);
	tcu::IntLookupPrecision	precision;

	{
		const tcu::IVec4	srcBitDepth	= tcu::getTextureFormatBitDepth(source.getFormat());
		const tcu::IVec4	dstBitDepth	= tcu::getTextureFormatBitDepth(result.getFormat());

		for (deUint32 i = 0; i < 4; ++i) {
			precision.colorThreshold[i]	= de::max(de::max(srcBitDepth[i] / 8, dstBitDepth[i] / 8), 1);
			precision.colorMask[i]		= dstBitDepth[i] != 0;
		}
	}

	// Prepare a source image with a matching (converted) pixel format. Ideally, we would've used a wrapper that
	// does the conversion on the fly without wasting memory, but this approach is more straightforward.
	tcu::TextureLevel				convertedSourceTexture	(result.getFormat(), source.getWidth(), source.getHeight());
	const tcu::PixelBufferAccess	convertedSource			= convertedSourceTexture.getAccess();

	for (int y = 0; y < source.getHeight(); ++y)
	for (int x = 0; x < source.getWidth();  ++x)
		convertedSource.setPixel(source.getPixelInt(x, y), x, y);	// will be clamped to max. representable value

	const struct Capture
	{
		const tcu::ConstPixelBufferAccess&	source;
		const tcu::ConstPixelBufferAccess&	result;
		const tcu::Sampler&					sampler;
		const tcu::IntLookupPrecision&		precision;
	} capture =
	{
		convertedSource, result, sampler, precision
	};

	const struct Loop : CompareEachPixelInEachRegion
	{
		Loop (void) {}

		bool compare (const void* pUserData, const int x, const int y, const tcu::Vec2& srcNormCoord) const
		{
			const Capture&					c					= *static_cast<const Capture*>(pUserData);
			const tcu::TexLookupScaleMode	lookupScaleDontCare	= tcu::TEX_LOOKUP_SCALE_MINIFY;
			const tcu::IVec4				dstColor			= c.result.getPixelInt(x, y);

			return tcu::isLevel2DLookupResultValid(c.source, c.sampler, lookupScaleDontCare, c.precision, srcNormCoord, 0, dstColor);
		}
	} loop;

	return loop.forEach(&capture, regions, source.getWidth(), source.getHeight(), errorMask);
}

bool BlittingImages::checkNearestFilteredResult (const tcu::ConstPixelBufferAccess&	result,
												 const tcu::ConstPixelBufferAccess& source)
{
	tcu::TestLog&					log				(m_context.getTestContext().getLog());
	const tcu::TextureFormat		dstFormat		= result.getFormat();
	const tcu::TextureChannelClass	dstChannelClass = tcu::getTextureChannelClass(dstFormat.type);

	tcu::TextureLevel		errorMaskStorage	(tcu::TextureFormat(tcu::TextureFormat::RGB, tcu::TextureFormat::UNORM_INT8), result.getWidth(), result.getHeight());
	tcu::PixelBufferAccess	errorMask			= errorMaskStorage.getAccess();
	tcu::Vec4				pixelBias			(0.0f, 0.0f, 0.0f, 0.0f);
	tcu::Vec4				pixelScale			(1.0f, 1.0f, 1.0f, 1.0f);
	bool					ok					= false;

	tcu::clear(errorMask, tcu::Vec4(0.0f, 1.0f, 0.0f, 1.0));

	if (dstChannelClass == tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER ||
		dstChannelClass == tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER)
	{
		ok = intNearestBlitCompare(source, result, errorMask, m_params.regions);
	}
	else
		ok = floatNearestBlitCompare(source, result, errorMask, m_params.regions);

	if (result.getFormat() != tcu::TextureFormat(tcu::TextureFormat::RGBA, tcu::TextureFormat::UNORM_INT8))
		tcu::computePixelScaleBias(result, pixelScale, pixelBias);

	if (!ok)
	{
		log << tcu::TestLog::ImageSet("Compare", "Result comparsion")
			<< tcu::TestLog::Image("Result", "Result", result, pixelScale, pixelBias)
			<< tcu::TestLog::Image("ErrorMask",	"Error mask", errorMask)
			<< tcu::TestLog::EndImageSet;
	}
	else
	{
		log << tcu::TestLog::ImageSet("Compare", "Result comparsion")
			<< tcu::TestLog::Image("Result", "Result", result, pixelScale, pixelBias)
			<< tcu::TestLog::EndImageSet;
	}

	return ok;
}

tcu::TestStatus BlittingImages::checkTestResult (tcu::ConstPixelBufferAccess result)
{
	DE_ASSERT(m_params.filter == VK_FILTER_NEAREST || m_params.filter == VK_FILTER_LINEAR);
	const std::string failMessage("Result image is incorrect");

	if (m_params.filter == VK_FILTER_LINEAR)
	{
		if (tcu::isCombinedDepthStencilType(result.getFormat().type))
		{
			if (tcu::hasDepthComponent(result.getFormat().order))
			{
				const tcu::Sampler::DepthStencilMode	mode				= tcu::Sampler::MODE_DEPTH;
				const tcu::ConstPixelBufferAccess		depthResult			= tcu::getEffectiveDepthStencilAccess(result, mode);
				const tcu::ConstPixelBufferAccess		clampedExpected		= tcu::getEffectiveDepthStencilAccess(m_expectedTextureLevel->getAccess(), mode);
				const tcu::ConstPixelBufferAccess		unclampedExpected	= tcu::getEffectiveDepthStencilAccess(m_unclampedExpectedTextureLevel->getAccess(), mode);
				const tcu::TextureFormat				sourceFormat		= tcu::getEffectiveDepthStencilTextureFormat(mapVkFormat(m_params.src.image.format), mode);

				if (!checkLinearFilteredResult(depthResult, clampedExpected, unclampedExpected, sourceFormat))
					return tcu::TestStatus::fail(failMessage);
			}

			if (tcu::hasStencilComponent(result.getFormat().order))
			{
				const tcu::Sampler::DepthStencilMode	mode				= tcu::Sampler::MODE_STENCIL;
				const tcu::ConstPixelBufferAccess		stencilResult		= tcu::getEffectiveDepthStencilAccess(result, mode);
				const tcu::ConstPixelBufferAccess		clampedExpected		= tcu::getEffectiveDepthStencilAccess(m_expectedTextureLevel->getAccess(), mode);
				const tcu::ConstPixelBufferAccess		unclampedExpected	= tcu::getEffectiveDepthStencilAccess(m_unclampedExpectedTextureLevel->getAccess(), mode);
				const tcu::TextureFormat				sourceFormat		= tcu::getEffectiveDepthStencilTextureFormat(mapVkFormat(m_params.src.image.format), mode);

				if (!checkLinearFilteredResult(stencilResult, clampedExpected, unclampedExpected, sourceFormat))
					return tcu::TestStatus::fail(failMessage);
			}
		}
		else
		{
			const tcu::TextureFormat	sourceFormat	= mapVkFormat(m_params.src.image.format);

			if (!checkLinearFilteredResult(result, m_expectedTextureLevel->getAccess(), m_unclampedExpectedTextureLevel->getAccess(), sourceFormat))
				return tcu::TestStatus::fail(failMessage);
		}
	}
	else // NEAREST filtering
	{
		if (tcu::isCombinedDepthStencilType(result.getFormat().type))
		{
			if (tcu::hasDepthComponent(result.getFormat().order))
			{
				const tcu::Sampler::DepthStencilMode	mode			= tcu::Sampler::MODE_DEPTH;
				const tcu::ConstPixelBufferAccess		depthResult		= tcu::getEffectiveDepthStencilAccess(result, mode);
				const tcu::ConstPixelBufferAccess		depthSource		= tcu::getEffectiveDepthStencilAccess(m_sourceTextureLevel->getAccess(), mode);

				if (!checkNearestFilteredResult(depthResult, depthSource))
					return tcu::TestStatus::fail(failMessage);
			}

			if (tcu::hasStencilComponent(result.getFormat().order))
			{
				const tcu::Sampler::DepthStencilMode	mode			= tcu::Sampler::MODE_STENCIL;
				const tcu::ConstPixelBufferAccess		stencilResult	= tcu::getEffectiveDepthStencilAccess(result, mode);
				const tcu::ConstPixelBufferAccess		stencilSource	= tcu::getEffectiveDepthStencilAccess(m_sourceTextureLevel->getAccess(), mode);

				if (!checkNearestFilteredResult(stencilResult, stencilSource))
					return tcu::TestStatus::fail(failMessage);
			}
		}
		else
		{
			if (!checkNearestFilteredResult(result, m_sourceTextureLevel->getAccess()))
				return tcu::TestStatus::fail(failMessage);
		}
	}

	return tcu::TestStatus::pass("Pass");
}

tcu::Vec4 linearToSRGBIfNeeded (const tcu::TextureFormat& format, const tcu::Vec4& color)
{
	return isSRGB(format) ? linearToSRGB(color) : color;
}

void scaleFromWholeSrcBuffer (const tcu::PixelBufferAccess& dst, const tcu::ConstPixelBufferAccess& src, const VkOffset3D regionOffset, const VkOffset3D regionExtent, tcu::Sampler::FilterMode filter)
{
	DE_ASSERT(filter == tcu::Sampler::LINEAR);
	DE_ASSERT(dst.getDepth() == 1 && src.getDepth() == 1);

	tcu::Sampler sampler(tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE,
					filter, filter, 0.0f, false);

	float sX = (float)regionExtent.x / (float)dst.getWidth();
	float sY = (float)regionExtent.y / (float)dst.getHeight();

	for (int y = 0; y < dst.getHeight(); y++)
	for (int x = 0; x < dst.getWidth(); x++)
		dst.setPixel(linearToSRGBIfNeeded(dst.getFormat(), src.sample2D(sampler, filter, (float)regionOffset.x + ((float)x+0.5f)*sX, (float)regionOffset.y + ((float)y+0.5f)*sY, 0)), x, y);
}

void blit (const tcu::PixelBufferAccess& dst, const tcu::ConstPixelBufferAccess& src, const tcu::Sampler::FilterMode filter, const MirrorMode mirrorMode)
{
	DE_ASSERT(filter == tcu::Sampler::NEAREST || filter == tcu::Sampler::LINEAR);

	tcu::Sampler sampler(tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE, tcu::Sampler::CLAMP_TO_EDGE,
						 filter, filter, 0.0f, false);

	const float sX = (float)src.getWidth() / (float)dst.getWidth();
	const float sY = (float)src.getHeight() / (float)dst.getHeight();
	const float sZ = (float)src.getDepth() / (float)dst.getDepth();

	tcu::Mat2 rotMatrix;
	rotMatrix(0,0) = (mirrorMode & MIRROR_MODE_X) ? -1.0f : 1.0f;
	rotMatrix(0,1) = 0.0f;
	rotMatrix(1,0) = 0.0f;
	rotMatrix(1,1) = (mirrorMode & MIRROR_MODE_Y) ? -1.0f : 1.0f;

	const int xOffset = (mirrorMode & MIRROR_MODE_X) ? dst.getWidth() - 1 : 0;
	const int yOffset = (mirrorMode & MIRROR_MODE_Y) ? dst.getHeight() - 1 : 0;

	if (dst.getDepth() == 1 && src.getDepth() == 1)
	{
		for (int y = 0; y < dst.getHeight(); ++y)
		for (int x = 0; x < dst.getWidth(); ++x)
		{
			const tcu::Vec2 xy = rotMatrix * tcu::Vec2((float)x,(float)y);
			dst.setPixel(linearToSRGBIfNeeded(dst.getFormat(), src.sample2D(sampler, filter, ((float)x+0.5f)*sX, ((float)y+0.5f)*sY, 0)), (int)round(xy[0]) + xOffset, (int)round(xy[1]) + yOffset);
		}
	}
	else
	{
		for (int z = 0; z < dst.getDepth(); ++z)
		for (int y = 0; y < dst.getHeight(); ++y)
		for (int x = 0; x < dst.getWidth(); ++x)
		{
			const tcu::Vec2 xy = rotMatrix * tcu::Vec2((float)x,(float)y);
			dst.setPixel(linearToSRGBIfNeeded(dst.getFormat(), src.sample3D(sampler, filter, ((float)x+0.5f)*sX, ((float)y+0.5f)*sY, ((float)z+0.5f)*sZ)), (int)round(xy[0]) + xOffset, (int)round(xy[1]) + yOffset, z);
		}
	}
}

void flipCoordinates (CopyRegion& region, const MirrorMode mirrorMode)
{
	const VkOffset3D dstOffset0 = region.imageBlit.dstOffsets[0];
	const VkOffset3D dstOffset1 = region.imageBlit.dstOffsets[1];
	const VkOffset3D srcOffset0 = region.imageBlit.srcOffsets[0];
	const VkOffset3D srcOffset1 = region.imageBlit.srcOffsets[1];

	if (mirrorMode > MIRROR_MODE_NONE && mirrorMode < MIRROR_MODE_LAST)
	{
		//sourceRegion
		region.imageBlit.srcOffsets[0].x = std::min(srcOffset0.x, srcOffset1.x);
		region.imageBlit.srcOffsets[0].y = std::min(srcOffset0.y, srcOffset1.y);

		region.imageBlit.srcOffsets[1].x = std::max(srcOffset0.x, srcOffset1.x);
		region.imageBlit.srcOffsets[1].y = std::max(srcOffset0.y, srcOffset1.y);

		//destinationRegion
		region.imageBlit.dstOffsets[0].x = std::min(dstOffset0.x, dstOffset1.x);
		region.imageBlit.dstOffsets[0].y = std::min(dstOffset0.y, dstOffset1.y);

		region.imageBlit.dstOffsets[1].x = std::max(dstOffset0.x, dstOffset1.x);
		region.imageBlit.dstOffsets[1].y = std::max(dstOffset0.y, dstOffset1.y);
	}
}

MirrorMode getMirrorMode(const VkOffset3D x1, const VkOffset3D x2)
{
	if (x1.x >= x2.x && x1.y >= x2.y)
	{
		return MIRROR_MODE_XY;
	}
	else if (x1.x <= x2.x && x1.y <= x2.y)
	{
		return MIRROR_MODE_NONE;
	}
	else if (x1.x <= x2.x && x1.y >= x2.y)
	{
		return MIRROR_MODE_Y;
	}
	else if (x1.x >= x2.x && x1.y <= x2.y)
	{
		return MIRROR_MODE_X;
	}
	return MIRROR_MODE_LAST;
}

MirrorMode getMirrorMode(const VkOffset3D s1, const VkOffset3D s2, const VkOffset3D d1, const VkOffset3D d2)
{
	const MirrorMode source		 = getMirrorMode(s1, s2);
	const MirrorMode destination = getMirrorMode(d1, d2);

	if (source == destination)
	{
		return MIRROR_MODE_NONE;
	}
	else if ((source == MIRROR_MODE_XY && destination == MIRROR_MODE_X)	  || (destination == MIRROR_MODE_XY && source == MIRROR_MODE_X) ||
			 (source == MIRROR_MODE_Y && destination == MIRROR_MODE_NONE) || (destination == MIRROR_MODE_Y && source == MIRROR_MODE_NONE))
	{
		return MIRROR_MODE_Y;
	}
	else if ((source == MIRROR_MODE_XY && destination == MIRROR_MODE_Y)	  || (destination == MIRROR_MODE_XY && source == MIRROR_MODE_Y) ||
			 (source == MIRROR_MODE_X && destination == MIRROR_MODE_NONE) || (destination == MIRROR_MODE_X && source == MIRROR_MODE_NONE))
	{
		return MIRROR_MODE_X;
	}
	else if ((source == MIRROR_MODE_XY && destination == MIRROR_MODE_NONE) || (destination == MIRROR_MODE_XY && source == MIRROR_MODE_NONE))
	{
		return MIRROR_MODE_XY;
	}
	return MIRROR_MODE_LAST;
}

void BlittingImages::copyRegionToTextureLevel (tcu::ConstPixelBufferAccess src, tcu::PixelBufferAccess dst, CopyRegion region)
{
	const MirrorMode mirrorMode = getMirrorMode(region.imageBlit.srcOffsets[0],
												region.imageBlit.srcOffsets[1],
												region.imageBlit.dstOffsets[0],
												region.imageBlit.dstOffsets[1]);

	flipCoordinates(region, mirrorMode);

	const VkOffset3D					srcOffset		= region.imageBlit.srcOffsets[0];
	const VkOffset3D					srcExtent		=
	{
		region.imageBlit.srcOffsets[1].x - srcOffset.x,
		region.imageBlit.srcOffsets[1].y - srcOffset.y,
		region.imageBlit.srcOffsets[1].z - srcOffset.z
	};
	const VkOffset3D					dstOffset		= region.imageBlit.dstOffsets[0];
	const VkOffset3D					dstExtent		=
	{
		region.imageBlit.dstOffsets[1].x - dstOffset.x,
		region.imageBlit.dstOffsets[1].y - dstOffset.y,
		region.imageBlit.dstOffsets[1].z - dstOffset.z
	};
	const tcu::Sampler::FilterMode		filter			= (m_params.filter == VK_FILTER_LINEAR) ? tcu::Sampler::LINEAR : tcu::Sampler::NEAREST;

	if (tcu::isCombinedDepthStencilType(src.getFormat().type))
	{
		DE_ASSERT(src.getFormat() == dst.getFormat());
		// Scale depth.
		if (tcu::hasDepthComponent(src.getFormat().order))
		{
			const tcu::ConstPixelBufferAccess	srcSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcExtent.x, srcExtent.y), tcu::Sampler::MODE_DEPTH);
			const tcu::PixelBufferAccess		dstSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(dst, dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y), tcu::Sampler::MODE_DEPTH);
			tcu::scale(dstSubRegion, srcSubRegion, filter);

			if (filter == tcu::Sampler::LINEAR)
			{
				const tcu::ConstPixelBufferAccess	depthSrc			= getEffectiveDepthStencilAccess(src, tcu::Sampler::MODE_DEPTH);
				const tcu::PixelBufferAccess		unclampedSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(m_unclampedExpectedTextureLevel->getAccess(), dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y), tcu::Sampler::MODE_DEPTH);
				scaleFromWholeSrcBuffer(unclampedSubRegion, depthSrc, srcOffset, srcExtent, filter);
			}
		}

		// Scale stencil.
		if (tcu::hasStencilComponent(src.getFormat().order))
		{
			const tcu::ConstPixelBufferAccess	srcSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcExtent.x, srcExtent.y), tcu::Sampler::MODE_STENCIL);
			const tcu::PixelBufferAccess		dstSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(dst, dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y), tcu::Sampler::MODE_STENCIL);
			blit(dstSubRegion, srcSubRegion, filter, mirrorMode);

			if (filter == tcu::Sampler::LINEAR)
			{
				const tcu::ConstPixelBufferAccess	stencilSrc			= getEffectiveDepthStencilAccess(src, tcu::Sampler::MODE_STENCIL);
				const tcu::PixelBufferAccess		unclampedSubRegion	= getEffectiveDepthStencilAccess(tcu::getSubregion(m_unclampedExpectedTextureLevel->getAccess(), dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y), tcu::Sampler::MODE_STENCIL);
				scaleFromWholeSrcBuffer(unclampedSubRegion, stencilSrc, srcOffset, srcExtent, filter);
			}
		}
	}
	else
	{
		const tcu::ConstPixelBufferAccess	srcSubRegion	= tcu::getSubregion(src, srcOffset.x, srcOffset.y, srcExtent.x, srcExtent.y);
		const tcu::PixelBufferAccess		dstSubRegion	= tcu::getSubregion(dst, dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y);
		blit(dstSubRegion, srcSubRegion, filter, mirrorMode);

		if (filter == tcu::Sampler::LINEAR)
		{
			const tcu::PixelBufferAccess	unclampedSubRegion	= tcu::getSubregion(m_unclampedExpectedTextureLevel->getAccess(), dstOffset.x, dstOffset.y, dstExtent.x, dstExtent.y);
			scaleFromWholeSrcBuffer(unclampedSubRegion, src, srcOffset, srcExtent, filter);
		}
	}
}

void BlittingImages::generateExpectedResult (void)
{
	const tcu::