ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 01:10:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

tempsave. asynccopy+trload sanity checked

Browse Source
This commit is contained in:
aska-0096
2025-07-22 08:04:05 +00:00
parent 1b468bac0b
commit 14e0ab70c6
4 changed files with 66 additions and 152 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
example/ck_tile/01_fmha/codegen/ops/fmha_fwd_decode.py
View File

@@ -39,7 +39,7 @@ K0_MAX_SUBMAX_MAP = {
SEQLENQ_MAP = {
"16" : "16",
# "32" : "32",
"32" : "32",
# "64" : "64"
}
@@ -652,14 +652,14 @@ def get_fmha_fwd_tile_dict_from_dtype(dtype : str) -> Optional[dict]:
'64': {
# # Specialize for different SeqQ
'16': FmhaFwdTileSize(16, 32, 64, 64, 32, 64, 1, 1, 1, 1, 1, 1, 16, 16, 32, 16, 16, 32, -1),
# '32': FmhaFwdTileSize(32, 32, 64, 64, 32, 64, 1, 1, 1, 1, 1, 1, 32, 32, 16, 32, 32, 16, -1),
'32': FmhaFwdTileSize(32, 32, 64, 64, 32, 64, 1, 1, 1, 1, 1, 1, 32, 32, 16, 32, 32, 16, -1),
# '64': FmhaFwdTileSize(64, 64, 64, 64, 64, 64, 4, 1, 1, 4, 1, 1, 16, 16, 32, 16, 16, 32, -1),
},
# '128': {
# '16': FmhaFwdTileSize(16, 32, 64, 128, 32, 128, 1, 1, 1, 1, 1, 1, 16, 16, 32, 16, 16, 32, -1),
# # '32': FmhaFwdTileSize(32, 32, 64, 128, 32, 128, 1, 1, 1, 1, 1, 1, 32, 32, 16, 32, 32, 16, -1),
'128': {
'16': FmhaFwdTileSize(16, 32, 128, 128, 32, 128, 1, 1, 1, 1, 1, 1, 16, 16, 32, 16, 16, 32, -1),
'32': FmhaFwdTileSize(32, 32, 128, 128, 32, 128, 1, 1, 1, 1, 1, 1, 32, 32, 16, 32, 32, 16, -1),
# # '64': FmhaFwdTileSize(64, 64, 64, 128, 64, 128, 4, 1, 1, 4, 1, 1, 16, 16, 32, 16, 16, 32, -1),
# },
},
}
else:
return None

18
example/ck_tile/01_fmha/fmha_fwd.cpp
View File

@@ -685,6 +685,24 @@ bool run(const ck_tile::ArgParser& arg_parser)
ck_tile::FillUniformDistribution<VDataType>{1.f, 1.f, seed}(vnew_host);
ck_tile::FillUniformDistribution<BiasDataType>{0.f, 1.f, seed}(bias_host);
}
else if(init_method == "k1" || init_method == "96")
{
ck_tile::FillUniformDistribution<QDataType>{0.f, 1.f, seed}(q_host);
ck_tile::FillUniformDistribution<KDataType>{1.f, 1.f, seed}(k_host);
ck_tile::FillUniformDistribution<KDataType>{1.f, 1.f, seed}(knew_host);
ck_tile::FillUniformDistribution<VDataType>{0.f, 1.f, seed}(v_host);
ck_tile::FillUniformDistribution<VDataType>{0.f, 1.f, seed}(vnew_host);
ck_tile::FillUniformDistribution<BiasDataType>{0.f, 1.f, seed}(bias_host);
}
else if(init_method == "q1" || init_method == "95")
{
ck_tile::FillUniformDistribution<QDataType>{1.f, 1.f, seed}(q_host);
ck_tile::FillUniformDistribution<KDataType>{0.f, 1.f, seed}(k_host);
ck_tile::FillUniformDistribution<KDataType>{0.f, 1.f, seed}(knew_host);
ck_tile::FillUniformDistribution<VDataType>{0.f, 1.f, seed}(v_host);
ck_tile::FillUniformDistribution<VDataType>{0.f, 1.f, seed}(vnew_host);
ck_tile::FillUniformDistribution<BiasDataType>{0.f, 1.f, seed}(bias_host);
}
else if(init_method == "kv1" || init_method == "98")
{
ck_tile::FillUniformDistribution<QDataType>{0.f, 1.f, seed}(q_host);