tempsave, trload+asyncload done

This commit is contained in:
aska-0096
2025-07-21 05:55:55 +00:00
parent afd96d8180
commit 1b468bac0b
6 changed files with 168 additions and 17 deletions

View File

@@ -788,7 +788,7 @@ struct FmhaFwdDecodeKernel
amd_buffer_coherence_enum::SYSTEM_NT1>(
data, // will update this pointer if using paged-kvcache
make_tuple(length, kargs.hdim_v),
make_tuple(kargs.stride_v, 1),
make_tuple(kargs.hdim_v, 1),
number<FmhaPipeline::kAlignmentV>{},
number<1>{});