Series · OpenCL · OpenMP
Parallel Computing
OpenCL and OpenMP from a first kernel to an LU decomposition, then the same decomposition with SSE intrinsics, each version timed.
01
OpenCL
A first OpenCL kernel that squares an array on the CPU or GPU, checked on the host.
October 02, 2019
02
LU Matrix Decomposition, OpenCL
LU decomposition as an OpenCL kernel, one work item per element.
October 10, 2019
03
OpenMP
OpenMP’s fork-join model and the one directive that starts a parallel section.
October 12, 2019
04
LU Matrix Decomposition
LU decomposition written plainly in C++: the lower and upper factors and how they rebuild the matrix.
October 18, 2019
05
LU Matrix Decomposition, Intrinsics
The same decomposition with SSE intrinsics for loads and stores, each beside the line it replaces.
October 22, 2019
06
LU Matrix Decomposition, Intrinsics Complete
Intrinsics for the arithmetic too, with the execution time of each version.
October 26, 2019
07
LU Matrix Decomposition, OpenMP
Finishing the LU decomposition in parallel with OpenMP.
October 30, 2019