MHYangAMD 9fcf21a4ec Fix fmha fwd precision issue on MI3XX series (#2285) ...

* Fix fmha fwd precision issue on MI3XX series

For fmha fwd fp16 cases, we found that using
impl::cast_tile_pk_fp16_fp32 for casting P would lead to precision
issues, since it uses __builtin_amdgcn_cvt_pkrtz, which is round to zero.

For examaple, fixing K,V to be all 1, and Q is random, which outputs are
expected to be all 1. But we found that it would have some incorrect
outputs 0.9995, which are smaller than the atol 0.001. (1 - 0.9995 =
0.0005 < 0.001) Thus, ck do not report this error.

* Add option to switch rtn/rtz for fmha fwd

2025-06-10 15:03:23 +08:00

..

block

[CK_TILE] fMHA batch_prefill block index & logits soft-capping optimizations (#2198)

2025-05-16 15:14:46 +08:00

kernel

[CK_TILE] For FMHA forward kernels, assign block indices reversely if using mask (#2209)

2025-05-27 10:58:58 +08:00

pipeline

Fix fmha fwd precision issue on MI3XX series (#2285)

2025-06-10 15:03:23 +08:00