leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

CUTLASS 3.8 Release (#2059)

Browse Source 
* CUTLASS 3.8 Release

* update

* Update README.md

* Revert "Update README.md"

This reverts commit b353e36fe83e0815f99b44e46c0c95494c44726b.

* update

* update

---------

Co-authored-by: Haicheng Wu <57973641+hwu36@users.noreply.github.com>
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
mihir-awatramani
2025-01-24 23:44:06 -08:00
committed by GitHub
parent 9eb01fa0b0
commit 389e493055
290 changed files with 91223 additions and 292 deletions
Download Patch File Download Diff File Expand all files Collapse all files

132
include/cutlass/array.h
View File

@@ -2567,7 +2567,6 @@ struct bit_not<Array<uint1b_t, N>> {
}
};
/// bit_xor
template <int N>
struct bit_xor<Array<uint1b_t, N>> {
@@ -2590,6 +2589,137 @@ struct bit_xor<Array<uint1b_t, N>> {
}
};
/// Fused and-popc-add
template <typename T, int N>
struct and_popc_add<Array<T, N>, Array<T, N>, Array<T, N>> {
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
and_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], b[i], c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, T const &scalar, Array<T, N> const &c) const {
Array<T, N> result;
and_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], scalar, c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(T const &scalar, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
and_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(scalar, b[i], c[i]);
}
return result;
}
};
/// Fused or-popc-add
template <typename T, int N>
struct or_popc_add<Array<T, N>, Array<T, N>, Array<T, N>> {
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
or_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], b[i], c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, T const &scalar, Array<T, N> const &c) const {
Array<T, N> result;
or_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], scalar, c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(T const &scalar, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
or_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(scalar, b[i], c[i]);
}
return result;
}
};
/// Fused xor-popc-add
template <typename T, int N>
struct xor_popc_add<Array<T, N>, Array<T, N>, Array<T, N>> {
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
xor_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], b[i], c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(Array<T, N> const &a, T const &scalar, Array<T, N> const &c) const {
Array<T, N> result;
xor_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(a[i], scalar, c[i]);
}
return result;
}
CUTLASS_HOST_DEVICE
Array<T, N> operator()(T const &scalar, Array<T, N> const &b, Array<T, N> const &c) const {
Array<T, N> result;
xor_popc_add<T> scalar_op;
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < N; ++i) {
result[i] = scalar_op(scalar, b[i], c[i]);
}
return result;
}
};
/////////////////////////////////////////////////////////////////////////////////////////////////
// Operator overloads
/////////////////////////////////////////////////////////////////////////////////////////////////