leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 24 Packages Projects Releases Wiki Activity

Updates and Bug fixes to CUTLASS 3.3 (#1232)

Browse Source
This commit is contained in:
Pradeep Ramani
2023-12-05 09:50:49 -05:00
committed by GitHub
parent 4a1709e17e
commit e9e30c2304
31 changed files with 534 additions and 227 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cute/swizzle_layout.hpp
+24
View File
@@ -449,6 +449,30 @@ recast_layout(Swizzle<B,M,S> const& swizzle)
// Other operations
//
template <int B, int M, int S, class Offset, class LayoutB, class Shape, class Stride>
CUTE_HOST_DEVICE constexpr
auto
max_common_layout(ComposedLayout<Swizzle<B,M,S>,Offset,LayoutB> const& a,
Layout<Shape,Stride> const& b)
{
auto common = max_common_layout(a.layout_b(), b);
auto base = Int<(1 << M)>{};
if constexpr (base < size(common)) {
return common.compose(base); // Truncate common to size base
} else {
return common;
}
}
template <class Shape, class Stride, int B, int M, int S, class Offset, class LayoutB>
CUTE_HOST_DEVICE constexpr
auto
max_common_layout(Layout<Shape,Stride> const& a,
ComposedLayout<Swizzle<B,M,S>,Offset,LayoutB> const& b)
{
return max_common_layout(b, a);
}
template <int B, int M, int S, class Offset, class LayoutB, class Shape, class Stride>
CUTE_HOST_DEVICE constexpr
auto