Harisankar Sadasivan

icon indicating a website link https://sites.google.com/umich.edu/hariss/home

Results 2 comments of Harisankar Sadasivan

redundant s_waitcnt lgkmcnt(0) with barrier

Seems to happen with rocm6.1

threadwise_tensor_slice_transfer_v5r1 issue

Hello @joye , Thank You for reaching out to the Composable Kernel team. We looked into this internally. We think logically the implementation is correct although initializing tmp to zero...