fix splitkv compiler issue since lse is used to select kernel instances

This commit is contained in:
fsx950223
2025-04-17 08:15:10 +00:00
parent 0d4f145078
commit aa9847e42d

View File

@@ -277,16 +277,10 @@ FMHA_FWD_SPLITKV_API_INNER_DISPATCH=""" {F_if}((t.is_group_mode == {F
static_assert({F_bm0} % kM0 == 0);
static_assert({F_bn1} % 32 == 0);
if (t.has_lse) {{
if constexpr (std::is_same_v<{F_dtype}, FmhaFwdFp8>) {{
return -1;
}} else {{
using traits2_ = fmha_fwd_splitkv_combine_traits_<{F_hdim}, {F_dtype}, {F_mode}, /*F_bn1=*/32, true, {F_squant}, {F_spad}, {F_dvpad}>;
return fmha_fwd_splitkv_<traits_, traits2_>(s, a);
}}
if (t.has_lse && std::is_same_v<{F_dtype}, FmhaFwdFp8>) {{
return -1;
}} else {{
using traits2_ = fmha_fwd_splitkv_combine_traits_<{F_hdim}, {F_dtype}, {F_mode}, /*F_bn1=*/32, false, {F_squant}, {F_spad}, {F_dvpad}>;
using traits2_ = fmha_fwd_splitkv_combine_traits_<{F_hdim}, {F_dtype}, {F_mode}, /*F_bn1=*/32, {F_lse}, {F_squant}, {F_spad}, {F_dvpad}>;
return fmha_fwd_splitkv_<traits_, traits2_>(s, a);
}}