external/vulkancts/modules/vulkan/subgroups/vktSubgroupsShapeTests.cpp

/*------------------------------------------------------------------------
 * Vulkan Conformance Tests
 * ------------------------
 *
 * Copyright (c) 2019 The Khronos Group Inc.
 * Copyright (c) 2019 Google Inc.
 * Copyright (c) 2017 Codeplay Software 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 Subgroups Tests
 */ /*--------------------------------------------------------------------*/

#include "vktSubgroupsShapeTests.hpp"
#include "vktSubgroupsTestsUtils.hpp"

#include <string>
#include <vector>

using namespace tcu;
using namespace std;
using namespace vk;
using namespace vkt;

namespace
{
enum OpType
{
	OPTYPE_CLUSTERED = 0,
	OPTYPE_QUAD,
	OPTYPE_LAST
};

struct CaseDefinition
{
	OpType				opType;
	VkShaderStageFlags	shaderStage;
	de::SharedPtr<bool>	geometryPointSizeSupported;
	deBool				requiredSubgroupSize;
};

static bool checkVertexPipelineStages (const void*			internalData,
									   vector<const void*>	datas,
									   deUint32				width,
									   deUint32)
{
	DE_UNREF(internalData);

	return subgroups::check(datas, width, 1);
}

static bool checkCompute (const void*			internalData,
						  vector<const void*>	datas,
						  const deUint32		numWorkgroups[3],
						  const deUint32		localSize[3],
						  deUint32)
{
	DE_UNREF(internalData);

	return subgroups::checkCompute(datas, numWorkgroups, localSize, 1);
}

string getOpTypeName (const OpType opType)
{
	switch (opType)
	{
		case OPTYPE_CLUSTERED:	return "clustered";
		case OPTYPE_QUAD:		return "quad";
		default:				TCU_THROW(InternalError, "Unsupported op type");
	}
}

string getExtHeader (const CaseDefinition& caseDef)
{
	const string	testExtensions	= (OPTYPE_CLUSTERED == caseDef.opType)
									? "#extension GL_KHR_shader_subgroup_clustered: enable\n"
									: "#extension GL_KHR_shader_subgroup_quad: enable\n";
	const string	extensions		= testExtensions
									+ "#extension GL_KHR_shader_subgroup_ballot: enable\n";

	return extensions;
}

string getBodySource (const CaseDefinition& caseDef)
{
	ostringstream	bdy;

	bdy << "  uint tempResult = 0x1;\n"
		<< "  uvec4 mask = subgroupBallot(true);\n";

	if (OPTYPE_CLUSTERED == caseDef.opType)
	{
		for (deUint32 i = 1; i <= subgroups::maxSupportedSubgroupSize(); i *= 2)
		{
			bdy << "  if (gl_SubgroupSize >= " << i << ")\n"
				<< "  {\n"
				<< "    uvec4 contribution = uvec4(0);\n"
				<< "    const uint modID = gl_SubgroupInvocationID % 32;\n"
				<< "    switch (gl_SubgroupInvocationID / 32)\n"
				<< "    {\n"
				<< "    case 0: contribution.x = 1 << modID; break;\n"
				<< "    case 1: contribution.y = 1 << modID; break;\n"
				<< "    case 2: contribution.z = 1 << modID; break;\n"
				<< "    case 3: contribution.w = 1 << modID; break;\n"
				<< "    }\n"
				<< "    uvec4 result = subgroupClusteredOr(contribution, " << i << ");\n"
				<< "    uint rootID = gl_SubgroupInvocationID & ~(" << i - 1 << ");\n"
				<< "    for (uint i = 0; i < " << i << "; i++)\n"
				<< "    {\n"
				<< "      uint nextID = rootID + i;\n"
				<< "      if (subgroupBallotBitExtract(mask, nextID) ^^ subgroupBallotBitExtract(result, nextID))\n"
				<< "      {\n"
				<< "        tempResult = 0;\n"
				<< "      }\n"
				<< "    }\n"
				<< "  }\n";
		}
	}
	else
	{
		bdy << "  uint cluster[4] =\n"
			<< "  {\n"
			<< "    subgroupQuadBroadcast(gl_SubgroupInvocationID, 0),\n"
			<< "    subgroupQuadBroadcast(gl_SubgroupInvocationID, 1),\n"
			<< "    subgroupQuadBroadcast(gl_SubgroupInvocationID, 2),\n"
			<< "    subgroupQuadBroadcast(gl_SubgroupInvocationID, 3)\n"
			<< "  };\n"
			<< "  uint rootID = gl_SubgroupInvocationID & ~0x3;\n"
			<< "  for (uint i = 0; i < 4; i++)\n"
			<< "  {\n"
			<< "    uint nextID = rootID + i;\n"
			<< "    if (subgroupBallotBitExtract(mask, nextID) && (cluster[i] != nextID))\n"
			<< "    {\n"
			<< "      tempResult = mask.x;\n"
			<< "    }\n"
			<< "  }\n";
	}

	bdy << "  tempRes = tempResult;\n";

	return bdy.str();
}

vector<string> getFramebufferPerStageHeadDeclarations (const CaseDefinition& caseDef)
{
	vector<string>	result;

	DE_UNREF(caseDef);

	result.push_back("layout(location = 0) out float result;\n");
	result.push_back("layout(location = 0) out float out_color;\n");
	result.push_back("layout(location = 0) out float out_color[];\n");
	result.push_back("layout(location = 0) out float out_color;\n");

	return result;
}

void initFrameBufferPrograms (SourceCollections& programCollection, CaseDefinition caseDef)
{
	const ShaderBuildOptions	buildOptions		(programCollection.usedVulkanVersion, SPIRV_VERSION_1_3, 0u);
	const string				extHeader			= getExtHeader(caseDef);
	const string				testSrc				= getBodySource(caseDef);
	const vector<string>		headDeclarations	= getFramebufferPerStageHeadDeclarations(caseDef);
	const bool					pointSizeSupported	= *caseDef.geometryPointSizeSupported;

	subgroups::initStdFrameBufferPrograms(programCollection, buildOptions, caseDef.shaderStage, VK_FORMAT_R32_UINT, pointSizeSupported, extHeader, testSrc, "", headDeclarations);
}

vector<string> getPerStageHeadDeclarations (const CaseDefinition& caseDef)
{
	const deUint32	stageCount	= subgroups::getStagesCount(caseDef.shaderStage);
	const bool		fragment	= (caseDef.shaderStage & VK_SHADER_STAGE_FRAGMENT_BIT) != 0;
	vector<string>	result		(stageCount, string());

	if (fragment)
		result.reserve(result.size() + 1);

	for (size_t i = 0; i < result.size(); ++i)
	{
		result[i] =
			"layout(set = 0, binding = " + de::toString(i) + ", std430) buffer Buffer1\n"
			"{\n"
			"  uint result[];\n"
			"};\n";
	}

	if (fragment)
	{
		const string	fragPart	=
			"layout(location = 0) out uint result;\n";

		result.push_back(fragPart);
	}

	return result;
}

void initPrograms (SourceCollections& programCollection, CaseDefinition caseDef)
{
#ifndef CTS_USES_VULKANSC
	const bool					spirv14required		= isAllRayTracingStages(caseDef.shaderStage);
#else
	const bool					spirv14required		= false;
#endif // CTS_USES_VULKANSC
	const SpirvVersion			spirvVersion		= spirv14required ? SPIRV_VERSION_1_4 : SPIRV_VERSION_1_3;
	const ShaderBuildOptions	buildOptions		(programCollection.usedVulkanVersion, spirvVersion, 0u);
	const string				extHeader			= getExtHeader(caseDef);
	const string				testSrc				= getBodySource(caseDef);
	const vector<string>		headDeclarations	= getPerStageHeadDeclarations(caseDef);
	const bool					pointSizeSupport	= *caseDef.geometryPointSizeSupported;

	subgroups::initStdPrograms(programCollection, buildOptions, caseDef.shaderStage, VK_FORMAT_R32_UINT, pointSizeSupport, extHeader, testSrc, "", headDeclarations);
}

void supportedCheck (Context& context, CaseDefinition caseDef)
{
	if (!subgroups::isSubgroupSupported(context))
		TCU_THROW(NotSupportedError, "Subgroup operations are not supported");

	if (!subgroups::isSubgroupFeatureSupportedForDevice(context, VK_SUBGROUP_FEATURE_BALLOT_BIT))
	{
		TCU_THROW(NotSupportedError, "Device does not support subgroup ballot operations");
	}

	if (OPTYPE_CLUSTERED == caseDef.opType)
	{
		if (!subgroups::isSubgroupFeatureSupportedForDevice(context, VK_SUBGROUP_FEATURE_CLUSTERED_BIT))
		{
			TCU_THROW(NotSupportedError, "Subgroup shape tests require that clustered operations are supported!");
		}
	}

	if (OPTYPE_QUAD == caseDef.opType)
	{
		if (!subgroups::isSubgroupFeatureSupportedForDevice(context, VK_SUBGROUP_FEATURE_QUAD_BIT))
		{
			TCU_THROW(NotSupportedError, "Subgroup shape tests require that quad operations are supported!");
		}
	}

	if (caseDef.requiredSubgroupSize)
	{
		context.requireDeviceFunctionality("VK_EXT_subgroup_size_control");

		const VkPhysicalDeviceSubgroupSizeControlFeaturesEXT&	subgroupSizeControlFeatures		= context.getSubgroupSizeControlFeaturesEXT();
		const VkPhysicalDeviceSubgroupSizeControlPropertiesEXT&	subgroupSizeControlProperties	= context.getSubgroupSizeControlPropertiesEXT();

		if (subgroupSizeControlFeatures.subgroupSizeControl == DE_FALSE)
			TCU_THROW(NotSupportedError, "Device does not support varying subgroup sizes nor required subgroup size");

		if (subgroupSizeControlFeatures.computeFullSubgroups == DE_FALSE)
			TCU_THROW(NotSupportedError, "Device does not support full subgroups in compute shaders");

		if ((subgroupSizeControlProperties.requiredSubgroupSizeStages & caseDef.shaderStage) != caseDef.shaderStage)
			TCU_THROW(NotSupportedError, "Required subgroup size is not supported for shader stage");
	}

	*caseDef.geometryPointSizeSupported = subgroups::isTessellationAndGeometryPointSizeSupported(context);

#ifndef CTS_USES_VULKANSC
	if (isAllRayTracingStages(caseDef.shaderStage))
	{
		context.requireDeviceFunctionality("VK_KHR_ray_tracing_pipeline");
	}
#endif // CTS_USES_VULKANSC

	subgroups::supportedCheckShader(context, caseDef.shaderStage);
}

TestStatus noSSBOtest (Context& context, const CaseDefinition caseDef)
{
	switch (caseDef.shaderStage)
	{
		case VK_SHADER_STAGE_VERTEX_BIT:					return subgroups::makeVertexFrameBufferTest(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages);
		case VK_SHADER_STAGE_GEOMETRY_BIT:					return subgroups::makeGeometryFrameBufferTest(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages);
		case VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT:		return subgroups::makeTessellationEvaluationFrameBufferTest(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages, caseDef.shaderStage);
		case VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT:	return subgroups::makeTessellationEvaluationFrameBufferTest(context,  VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages, caseDef.shaderStage);
		default:											TCU_THROW(InternalError, "Unhandled shader stage");
	}
}

TestStatus test (Context& context, const CaseDefinition caseDef)
{
	if (isAllComputeStages(caseDef.shaderStage))
	{
		const VkPhysicalDeviceSubgroupSizeControlPropertiesEXT&	subgroupSizeControlProperties	= context.getSubgroupSizeControlPropertiesEXT();
		TestLog&												log								= context.getTestContext().getLog();

		if (caseDef.requiredSubgroupSize == DE_FALSE)
			return subgroups::makeComputeTest(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkCompute);

		log << TestLog::Message << "Testing required subgroup size range [" <<  subgroupSizeControlProperties.minSubgroupSize << ", "
			<< subgroupSizeControlProperties.maxSubgroupSize << "]" << TestLog::EndMessage;

		// According to the spec, requiredSubgroupSize must be a power-of-two integer.
		for (deUint32 size = subgroupSizeControlProperties.minSubgroupSize; size <= subgroupSizeControlProperties.maxSubgroupSize; size *= 2)
		{
			TestStatus result = subgroups::makeComputeTest(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkCompute,
																size, VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT_EXT);
			if (result.getCode() != QP_TEST_RESULT_PASS)
			{
				log << TestLog::Message << "subgroupSize " << size << " failed" << TestLog::EndMessage;
				return result;
			}
		}

		return TestStatus::pass("OK");
	}
	else if (isAllGraphicsStages(caseDef.shaderStage))
	{
		const VkShaderStageFlags	stages	= subgroups::getPossibleGraphicsSubgroupStages(context, caseDef.shaderStage);

		return subgroups::allStages(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages, stages);
	}
#ifndef CTS_USES_VULKANSC
	else if (isAllRayTracingStages(caseDef.shaderStage))
	{
		const VkShaderStageFlags	stages	= subgroups::getPossibleRayTracingSubgroupStages(context, caseDef.shaderStage);

		return subgroups::allRayTracingStages(context, VK_FORMAT_R32_UINT, DE_NULL, 0, DE_NULL, checkVertexPipelineStages, stages);
	}
#endif // CTS_USES_VULKANSC
	else
		TCU_THROW(InternalError, "Unknown stage or invalid stage set");
}
}

namespace vkt
{
namespace subgroups
{
TestCaseGroup* createSubgroupsShapeTests (TestContext& testCtx)
{
	de::MovePtr<TestCaseGroup>	group				(new TestCaseGroup(testCtx, "shape", "Subgroup shape category tests"));
	de::MovePtr<TestCaseGroup>	graphicGroup		(new TestCaseGroup(testCtx, "graphics", "Subgroup shape category tests: graphics"));
	de::MovePtr<TestCaseGroup>	computeGroup		(new TestCaseGroup(testCtx, "compute", "Subgroup shape category tests: compute"));
	de::MovePtr<TestCaseGroup>	framebufferGroup	(new TestCaseGroup(testCtx, "framebuffer", "Subgroup shape category tests: framebuffer"));
#ifndef CTS_USES_VULKANSC
	de::MovePtr<TestCaseGroup>	raytracingGroup		(new TestCaseGroup(testCtx, "ray_tracing", "Subgroup shape category tests: ray tracing"));
#endif // CTS_USES_VULKANSC
	const VkShaderStageFlags	stages[]			=
	{
		VK_SHADER_STAGE_VERTEX_BIT,
		VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT,
		VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT,
		VK_SHADER_STAGE_GEOMETRY_BIT,
	};
	const deBool				boolValues[]		=
	{
		DE_FALSE,
		DE_TRUE
	};

	for (int opTypeIndex = 0; opTypeIndex < OPTYPE_LAST; ++opTypeIndex)
	{
		const OpType	opType	= static_cast<OpType>(opTypeIndex);
		const string	op		= de::toLower(getOpTypeName(opType));

		for (size_t groupSizeNdx = 0; groupSizeNdx < DE_LENGTH_OF_ARRAY(boolValues); ++groupSizeNdx)
		{
			const deBool			requiredSubgroupSize	= boolValues[groupSizeNdx];
			const string			testName				= op + (requiredSubgroupSize ? "_requiredsubgroupsize" : "");
			const CaseDefinition	caseDef					=
			{
				opType,							//  OpType				opType;
				VK_SHADER_STAGE_COMPUTE_BIT,	//  VkShaderStageFlags	shaderStage;
				de::SharedPtr<bool>(new bool),	//  de::SharedPtr<bool>	geometryPointSizeSupported;
				requiredSubgroupSize			//  deBool				requiredSubgroupSize;
			};

			addFunctionCaseWithPrograms(computeGroup.get(), testName, "", supportedCheck, initPrograms, test, caseDef);
		}

		{
			const CaseDefinition	caseDef		=
			{
				opType,							//  OpType				opType;
				VK_SHADER_STAGE_ALL_GRAPHICS,	//  VkShaderStageFlags	shaderStage;
				de::SharedPtr<bool>(new bool),	//  de::SharedPtr<bool>	geometryPointSizeSupported;
				DE_FALSE						//  deBool				requiredSubgroupSize;
			};

			addFunctionCaseWithPrograms(graphicGroup.get(), op, "", supportedCheck, initPrograms, test, caseDef);
		}

#ifndef CTS_USES_VULKANSC
		{
			const CaseDefinition	caseDef		=
			{
				opType,							//  OpType				opType;
				SHADER_STAGE_ALL_RAY_TRACING,	//  VkShaderStageFlags	shaderStage;
				de::SharedPtr<bool>(new bool),	//  de::SharedPtr<bool>	geometryPointSizeSupported;
				DE_FALSE						//  deBool				requiredSubgroupSize;
			};

			addFunctionCaseWithPrograms(raytracingGroup.get(), op, "", supportedCheck, initPrograms, test, caseDef);
		}
#endif // CTS_USES_VULKANSC

		for (int stageIndex = 0; stageIndex < DE_LENGTH_OF_ARRAY(stages); ++stageIndex)
		{
			const CaseDefinition	caseDef		=
			{
				opType,							//  OpType				opType;
				stages[stageIndex],				//  VkShaderStageFlags	shaderStage;
				de::SharedPtr<bool>(new bool),	//  de::SharedPtr<bool>	geometryPointSizeSupported;
				DE_FALSE						//  deBool				requiredSubgroupSize;
			};
			const string			testName	= op + "_" + getShaderStageName(caseDef.shaderStage);

			addFunctionCaseWithPrograms(framebufferGroup.get(), testName, "", supportedCheck, initFrameBufferPrograms, noSSBOtest, caseDef);
		}
	}

	group->addChild(graphicGroup.release());
	group->addChild(computeGroup.release());
	group->addChild(framebufferGroup.release());
#ifndef CTS_USES_VULKANSC
	group->addChild(raytracingGroup.release());
#endif // CTS_USES_VULKANSC

	return group.release();
}

} // subgroups
} // vkt