Avoided getting device properties CUDA threads setup - it is extremely slow. (692932d3) · Commits · TNL / tnl-dev

src/TNL/Devices/Cuda.cu

+4 −2

Original line number	Diff line number	Diff line
		@@ -31,10 +31,11 @@ void Cuda::setupThreads( const dim3& blockSize,
		* TODO: Fix the following:
		* I do not known how to get max grid size in kernels :(
		*
		/* Also, this is very slow.
		int currentDevice( 0 );
		cudaGetDevice( currentDevice );
		cudaDeviceProp properties;
		cudaGetDeviceProperties( &properties, currentDevice );
		cudaGetDeviceProperties( &properties, currentDevice );*/
		gridsCount.x = blocksCount.x / properties.maxGridSize[ 0 ] + ( blocksCount.x % properties.maxGridSize[ 0 ] != 0 );
		gridsCount.y = blocksCount.y / properties.maxGridSize[ 1 ] + ( blocksCount.y % properties.maxGridSize[ 1 ] != 0 );
		gridsCount.z = blocksCount.z / properties.maxGridSize[ 2 ] + ( blocksCount.z % properties.maxGridSize[ 2 ] != 0 );
		@@ -49,10 +50,11 @@ void Cuda::setupGrid( const dim3& blocksCount,
		const dim3& gridIdx,
		dim3& gridSize )
		{
		/* TODO: this is extremely slow!!!!
		int currentDevice( 0 );
		cudaGetDevice( &currentDevice );
		cudaDeviceProp properties;
		cudaGetDeviceProperties( &properties, currentDevice );
		cudaGetDeviceProperties( &properties, currentDevice );*/

		/****
		* TODO: fix the following