Add shared memory version of cuda kernels

hpcgarage / spatter

Benchmark for measuring the performance of sparse and irregular memory access.

Other

74 stars 15 forks source link

Add shared memory version of cuda kernels #35

Open plavin opened 5 years ago

plavin commented 5 years ago

Per this Nvidia article, shared memory pays no penalty for stride memory access. We should have a shared memory version to test this.

jyoung3131 commented 2 months ago

Updated in 2023 - we discussed the need for a shared memory version of the CUDA kernels but will continue to discuss how to best implement this feature.