Commit e6f771cd authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Optimizing (slightly) SpMV of Adaptive Rg CSR matrix format.

parent b740f688

ChangeLog

+9 −0
Original line number Diff line number Diff line 
2011-06-29  oberhuber  <oberhuber@yvonne>



	* cmake/AddCompilerFlag.cmake: New file.

	* cmake/MacroEnsureVersion.cmake: New file.

	* cmake/OptimizeForArchitecture.cmake: New file.

	* src/matrix/.tnlRgCSRMatrix.h.swp: New file.

	* src/matrix/tnlAdaptiveRgCSRMatrix.h: 

	 - optimizing Adaptive Rg CSR format by adding break to SpMV loop.



2011-06-15  oberhuber  <oberhuber@yvonne>



	* src/core/tnlCudaSupport.cpp: Removed.

src/matrix/tnlAdaptiveRgCSRMatrix.h

+2 −0
Original line number Diff line number Diff line
@@ -1088,6 +1088,8 @@ __global__ void AdaptiveRgCSRMatrixVectorProductKernel( Real* target, 
			const Index column = columns[ offset ];

			if( column != -1 )

				sum += nonzeroElements[ offset ] * vect[ column ];

			else

			   break;

		}

		partialSums[ threadIdx. x ] = sum;

		__syncthreads();