Astha Rai 6e4ddf54ef cleaning up errors for different blockwise gemm pipelines: fp8 works for all pipelines, bf16 has validation errors for v2 atomic and memory issues v1 reduction

2025-06-11 22:27:00 +00:00

..

block

Add the instances for small sized GEMM in preshuffle and improve CMake Flag (#2212)

2025-05-20 15:05:08 -07:00

device

comment debug prints

2025-06-05 00:22:38 +00:00

element

Change relu to clamp for grouped conv fwd instances (#2249)

2025-05-29 00:51:25 +02:00

grid

cleaning up errors for different blockwise gemm pipelines: fp8 works for all pipelines, bf16 has validation errors for v2 atomic and memory issues v1 reduction

2025-06-11 22:27:00 +00:00

thread

Moe gemm activation (#2026)

2025-04-23 10:35:34 +08:00

warp

Use new mfma instructions for FP8 on gfx950 (#2202)

2025-05-19 17:29:51 -07:00