Function GetThreadGroupSize

shadercompiler/EffectCompilerMetal.cpp:1705–1752

}

void GetThreadGroupSize( StageInput& stage, ParserState& state, ASTNode* callNode )
{
	stage.threadGroupSize[0] = 1;
	stage.threadGroupSize[1] = 1;
	stage.threadGroupSize[2] = 1;

	if( stage.type != COMPUTE_STAGE )
	{
		return;
	}

	Symbol* entryPointSymbol = callNode->GetSymbol();
	if( !entryPointSymbol || !entryPointSymbol->definition )
	{
		return;
	}

	ASTNode* attribList = entryPointSymbol->definition->GetChildOrNull( 2 );
	if( attribList && attribList->GetNodeType() == NT_FUNCTION_ATTRIBUTE_LIST )
	{
		// Find [numthreads(X, Y, Z)] attribute.
		for( size_t i = 0, n = attribList->GetChildrenCount(); i < n; ++i )
		{
			ASTNode* attribute = attribList->GetChild( i );
			assert( attribute->GetNodeType() == NT_FUNCTION_ATTRIBUTE );

			if( attribute->GetToken()->stringValue != "numthreads" )
			{
				continue;
			}

			for( size_t j = 0; j < 3; ++j )
			{
				ASTNode* child = attribute->GetChildOrNull( j );

				InlineString valueString = ( child && child->GetToken() ) ? child->GetToken()->stringValue : InlineString{ nullptr, nullptr };
				unsigned long value = valueString ? std::strtoul( valueString.start, nullptr, 10 ) : 1;

				stage.threadGroupSize[j] = (uint32_t)value;
			}

			if( stage.threadGroupSize[0] * stage.threadGroupSize[1] * stage.threadGroupSize[2] > 512 )
			{
				state.ShowMessage( attribute->GetLocation(), EC_CUSTOM_WARNING, "number of threads per group exceeds Intel limit of 512; computer shader invoketion will fail on Intel macs" );
			}
		}
	}
}

bool AutoAssignRegistersForNode( ParserState& state, ASTNode* functionHeader, const FileLocation& messageLocation )
{