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

cutlass 3.9 update (#2255)

Browse Source 
* cutlass 3.9 update

* rebase

* fixes out of shared memory for blockwise Blackwell

* doc format

* fix issue 2253

* disable host ref by default

* fix sm120 smem capacity

---------

Co-authored-by: yuzhai <yuzhai@nvidia.com>
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Yujia Zhai
2025-04-24 15:42:40 -04:00
committed by GitHub
co-authored by yuzhai Haicheng Wu
parent 8e345c5c5b
commit 331a1f5b3f
143 changed files with 18089 additions and 5935 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cute/layout.hpp
+3 -3
View File
@@ -834,7 +834,7 @@ coalesce_x(Layout<Shape,Stride> const& layout)
} else {
return detail::bw_coalesce<R-2>(flat_shape, flat_stride, get<R-1>(flat_shape), get<R-1>(flat_stride));
}
CUTE_GCC_UNREACHABLE;
}
@@ -1030,7 +1030,7 @@ template <class LShape, class LStride,
class RShape, class RStride>
CUTE_HOST_DEVICE constexpr
auto
composition_impl(LShape const& lhs_shape, LStride const& lhs_stride,
composition_impl(LShape const& lhs_shape, [[maybe_unused]] LStride const& lhs_stride,
RShape const& rhs_shape, RStride const& rhs_stride)
{
if constexpr (is_tuple<RShape>::value) { // Right-distributivity of Layout composition for RHS tuple
@@ -1067,7 +1067,7 @@ composition_impl(LShape const& lhs_shape, LStride const& lhs_stride,
auto rest_stride = get<3>(init);
auto curr_shape = get<curr_i>(lhs_shape);
auto curr_stride = get<curr_i>(lhs_stride);
[[maybe_unused]] auto curr_stride = get<curr_i>(lhs_stride);
// Strong divisibility condition -- requires composition to be statically verifiable.
//CUTE_STATIC_ASSERT_V(((rest_stride % curr_shape) == Int<0>{}) or (rest_stride < curr_shape), "Stride Divisibility Condition");