anki-3d-engine/AnKi/Gr/BackendCommon/Functions.cpp

// Copyright (C) 2009-2023, Panagiotis Christopoulos Charitos and contributors.
// All rights reserved.
// Code licensed under the BSD License.
// http://www.anki3d.org/LICENSE

#include <AnKi/Gr/BackendCommon/Functions.h>

namespace anki {

template<typename T>
static void writeShaderBlockMemorySanityChecks(const ShaderVariableBlockInfo& varBlkInfo, [[maybe_unused]] const void* elements,
											   [[maybe_unused]] U32 elementsCount, [[maybe_unused]] void* buffBegin,
											   [[maybe_unused]] const void* buffEnd)
{
	// Check args
	ANKI_ASSERT(elements != nullptr);
	ANKI_ASSERT(elementsCount > 0);
	ANKI_ASSERT(buffBegin != nullptr);
	ANKI_ASSERT(buffEnd != nullptr);
	ANKI_ASSERT(buffBegin < buffEnd);

	// Check varBlkInfo
	ANKI_ASSERT(varBlkInfo.m_offset != -1);
	ANKI_ASSERT(varBlkInfo.m_arraySize > 0);

	if(varBlkInfo.m_arraySize > 1)
	{
		ANKI_ASSERT(varBlkInfo.m_arrayStride > 0);
	}

	// Check array size
	ANKI_ASSERT(I16(elementsCount) <= varBlkInfo.m_arraySize);
}

template<typename T>
static void writeShaderBlockMemorySimple(const ShaderVariableBlockInfo& varBlkInfo, const void* elements, U32 elementsCount, void* buffBegin,
										 const void* buffEnd)
{
	writeShaderBlockMemorySanityChecks<T>(varBlkInfo, elements, elementsCount, buffBegin, buffEnd);

	U8* outBuff = static_cast<U8*>(buffBegin) + varBlkInfo.m_offset;
	const U8* inBuff = static_cast<const U8*>(elements);
	for(U i = 0; i < elementsCount; i++)
	{
		ANKI_ASSERT(outBuff + sizeof(T) <= static_cast<const U8*>(buffEnd));

		// Memcpy because Vec might have SIMD alignment but not the output buffer
		memcpy(outBuff, inBuff + i * sizeof(T), sizeof(T));

		outBuff += varBlkInfo.m_arrayStride;
	}
}

template<typename T, typename Vec>
static void writeShaderBlockMemoryMatrix(const ShaderVariableBlockInfo& varBlkInfo, const void* elements, U32 elementsCount, void* buffBegin,
										 const void* buffEnd)
{
	writeShaderBlockMemorySanityChecks<T>(varBlkInfo, elements, elementsCount, buffBegin, buffEnd);
	ANKI_ASSERT(varBlkInfo.m_matrixStride > 0);
	ANKI_ASSERT(varBlkInfo.m_matrixStride >= static_cast<I16>(sizeof(Vec)));

	U8* buff = static_cast<U8*>(buffBegin) + varBlkInfo.m_offset;
	for(U i = 0; i < elementsCount; i++)
	{
		U8* subbuff = buff;
		const T& matrix = static_cast<const T*>(elements)[i];
		for(U j = 0; j < sizeof(T) / sizeof(Vec); j++)
		{
			ANKI_ASSERT((subbuff + sizeof(Vec)) <= static_cast<const U8*>(buffEnd));

			const Vec in = matrix.getRow(j);
			memcpy(subbuff, &in, sizeof(Vec)); // Memcpy because Vec might have SIMD alignment but not the output buffer
			subbuff += varBlkInfo.m_matrixStride;
		}

		buff += varBlkInfo.m_arrayStride;
	}
}

// This is some trickery to select calling between writeShaderBlockMemoryMatrix and writeShaderBlockMemorySimple
namespace {

template<typename T>
class IsShaderVarDataTypeAMatrix
{
public:
	static constexpr Bool kValue = false;
};

#define ANKI_SVDT_MACRO(type, baseType, rowCount, columnCount, isIntagralType) \
	template<> \
	class IsShaderVarDataTypeAMatrix<type> \
	{ \
	public: \
		static constexpr Bool kValue = rowCount * columnCount > 4; \
	};
#include <AnKi/Gr/ShaderVariableDataType.def.h>
#undef ANKI_SVDT_MACRO

template<typename T, Bool isMatrix = IsShaderVarDataTypeAMatrix<T>::kValue>
class WriteShaderBlockMemory
{
public:
	void operator()(const ShaderVariableBlockInfo& varBlkInfo, const void* elements, U32 elementsCount, void* buffBegin, const void* buffEnd)
	{
		using RowVec = typename T::RowVec;
		writeShaderBlockMemoryMatrix<T, RowVec>(varBlkInfo, elements, elementsCount, buffBegin, buffEnd);
	}
};

template<typename T>
class WriteShaderBlockMemory<T, false>
{
public:
	void operator()(const ShaderVariableBlockInfo& varBlkInfo, const void* elements, U32 elementsCount, void* buffBegin, const void* buffEnd)
	{
		writeShaderBlockMemorySimple<T>(varBlkInfo, elements, elementsCount, buffBegin, buffEnd);
	}
};

} // namespace

void writeShaderBlockMemory(ShaderVariableDataType type, const ShaderVariableBlockInfo& varBlkInfo, const void* elements, U32 elementsCount,
							void* buffBegin, const void* buffEnd)
{
	switch(type)
	{
#define ANKI_SVDT_MACRO(type, baseType, rowCount, columnCount, isIntagralType) \
	case ShaderVariableDataType::k##type: \
		WriteShaderBlockMemory<type>()(varBlkInfo, elements, elementsCount, buffBegin, buffEnd); \
		break;
#include <AnKi/Gr/ShaderVariableDataType.def.h>
#undef ANKI_SVDT_MACRO

	default:
		ANKI_ASSERT(0);
	}
}

Error linkShaderReflection(const ShaderReflection& a, const ShaderReflection& b, ShaderReflection& c_)
{
	ShaderReflection c;
	c.m_descriptorSetMask = a.m_descriptorSetMask | b.m_descriptorSetMask;

	for(U32 set = 0; set < kMaxDescriptorSets; ++set)
	{
		if(!c.m_descriptorSetMask.get(set))
		{
			continue;
		}

		for(U32 binding = 0; binding < kMaxBindingsPerDescriptorSet; ++binding)
		{
			const Bool bindingExists = a.m_descriptorArraySizes[set][binding] > 0 || b.m_descriptorArraySizes[set][binding] > 0;
			if(!bindingExists)
			{
				continue;
			}

			const Bool sizeCorrect = a.m_descriptorArraySizes[set][binding] == 0 || b.m_descriptorArraySizes[set][binding] == 0
									 || a.m_descriptorArraySizes[set][binding] == b.m_descriptorArraySizes[set][binding];

			if(!sizeCorrect)
			{
				ANKI_GR_LOGE("Can't link shader reflection because of different array size. Set %u binding %u", set, binding);
				return Error::kFunctionFailed;
			}

			const Bool typeCorrect = a.m_descriptorTypes[set][binding] == DescriptorType::kCount
									 || b.m_descriptorTypes[set][binding] == DescriptorType::kCount
									 || a.m_descriptorTypes[set][binding] == b.m_descriptorTypes[set][binding];

			if(!typeCorrect)
			{
				ANKI_GR_LOGE("Can't link shader reflection because of different array size. Set %u binding %u", set, binding);
				return Error::kFunctionFailed;
			}

			const Bool flagsCorrect = a.m_descriptorFlags[set][binding] == DescriptorFlag::kNone
									  || b.m_descriptorFlags[set][binding] == DescriptorFlag::kNone
									  || a.m_descriptorFlags[set][binding] == b.m_descriptorFlags[set][binding];
			if(!flagsCorrect)
			{
				ANKI_GR_LOGE("Can't link shader reflection because of different discriptor flags. Set %u binding %u", set, binding);
				return Error::kFunctionFailed;
			}

			c.m_descriptorArraySizes[set][binding] = max(a.m_descriptorArraySizes[set][binding], b.m_descriptorArraySizes[set][binding]);
			c.m_descriptorTypes[set][binding] = min(a.m_descriptorTypes[set][binding], b.m_descriptorTypes[set][binding]);
			c.m_descriptorFlags[set][binding] = max(a.m_descriptorFlags[set][binding], b.m_descriptorFlags[set][binding]);
		}
	}

	c.m_vertexAttributeLocations = (a.m_vertexAttributeMask.getAnySet()) ? a.m_vertexAttributeLocations : b.m_vertexAttributeLocations;
	c.m_vertexAttributeMask = a.m_vertexAttributeMask | b.m_vertexAttributeMask;

	c.m_colorAttachmentWritemask = a.m_colorAttachmentWritemask | b.m_colorAttachmentWritemask;

	const Bool pushConstantsCorrect = a.m_pushConstantsSize == 0 || b.m_pushConstantsSize == 0 || a.m_pushConstantsSize == b.m_pushConstantsSize;
	if(!pushConstantsCorrect)
	{
		ANKI_GR_LOGE("Can't link shader reflection because of different push constants size");
		return Error::kFunctionFailed;
	}

	c.m_pushConstantsSize = max(a.m_pushConstantsSize, b.m_pushConstantsSize);

	c.m_discards = a.m_discards || b.m_discards;

	c_ = c;
	return Error::kNone;
}

} // end namespace anki