Skip to content

[Reorder] add optimized e8m0 -> f32 upconversion #544

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[Reorder] add optimized e8m0 -> f32 upconversion #544

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions include/cute/arch/reorder_xe.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1236,7 +1236,33 @@ struct Xe_Reorder<ReorderKind::VV, float_e2m1_t, bfloat16_t>
}
};

template <>
struct Xe_Reorder<ReorderKind::UU, float_ue8m0_t, float>
{
using SRegisters = intel::uchar4[1];
using DRegisters = intel::float4[1];

CUTE_HOST_DEVICE static void
reorder(intel::uchar4 const& src0, intel::float4& dst0)
{
#if defined(CUTE_ARCH_COPY_XE_ENABLED)
asm ( /* 2 cycles/output register */
"{\n"
".decl IN_UB v_type=G type=UB num_elts=64 alias=<%1,0>\n"
".decl OUT_UW v_type=G type=UW num_elts=128 alias=<%0,0>\n"
"shl (M1_NM, 32) OUT_UW(0,1)<2> IN_UB(0,0)<1;1,0> 7:uw\n"
"shl (M1_NM, 32) OUT_UW(2,1)<2> IN_UB(0,32)<1;1,0> 7:uw\n"
"add.sat (M1_NM, 32) OUT_UW(0,0)<2> IN_UB(0,0)<1;1,0> -254:w\n"
"add.sat (M1_NM, 32) OUT_UW(2,0)<2> IN_UB(0,32)<1;1,0> -254:w\n"
"}\n"
: "=rw"(dst0)
: "rw"(src0)
);
#else
CUTE_INVALID_CONTROL_PATH("Not Xe");
#endif
}
};


} // end namespace cute