Add a kernel to measure HBM bandwidth vs WGs

ROCm / triton

Development repository for the Triton language and compiler

MIT License

83 stars 27 forks source link

Add a kernel to measure HBM bandwidth vs WGs #430

Closed vgokhale closed 8 months ago

vgokhale commented 8 months ago

The kernel moves 2^32 bytes of data from one part of HBM to another. Num WGs are swept and there is a way to switch the grid size and block size being moved per WG.