[CK_TILE] Simplify the codes in splitkv_combine pipeline (#1549)

* Simplify the codes in splitkv_combine pipeline

* Always set kPadSeqLenK=true for fmha splitkv kernels

* Change in Oacc Alignment and TileDistribution to be more adaptable to tile sizes

---------

Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>
This commit is contained in:
Qianfeng
2024-10-08 10:44:34 +08:00
committed by GitHub
parent 7733ae167b
commit 74d68e3b99
3 changed files with 67 additions and 50 deletions

View File

@@ -600,8 +600,8 @@ def get_fwd_splitkv_blobs(kernel_filter : Optional[str], receipt, mask_impl) ->
# TODO: use async pipeline when compiler is more stable
if hdim == 256 or hdim in [32, 64, 128]:
# if True:
pipelines.append(Pipeline('qr', 'row', 'f', 'f', 'f', 'f', bias, lse, squant, pagedkv, mask))
pipelines.append(Pipeline('qr', 'col', 'f', 'f', 'f', 'f', bias, lse, squant, pagedkv, mask))
pipelines.append(Pipeline('qr', 'row', 'f', 't', 'f', 'f', bias, lse, squant, pagedkv, mask))
pipelines.append(Pipeline('qr', 'col', 'f', 't', 'f', 'f', bias, lse, squant, pagedkv, mask))
pipelines.append(Pipeline('qr', 'row', 't', 't', 't', 't', bias, lse, squant, pagedkv, mask))
pipelines.append(Pipeline('qr', 'col', 't', 't', 't', 't', bias, lse, squant, pagedkv, mask))