leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 30 Packages Projects Releases Wiki Activity

Collection of changes to fix clang build. (#1200)

Browse Source 
* Remove unused variables

* Qualify calls to make_fragment_? from templated base class.

Fixes clang build error.

* Add missing `#include <cstdio>`

* Various changes to fix clang compile errors.

* More changes to fix clang build.

Remaining issues:

- `params` initializer of `CollectiveEpilogue`.
- `ops` initializer of `Sm90VisitorImplBase`.
- `__usAtomicCAS` needs to be added to clang upstream.

* Fix remaining clang build issues.

* Qualify `cute::rank()` calls.

* Qualify some more calls that are otherwise ambiguous between `cute` and `std` namespace.

* Double-escape special registers in inline asm.

* small change

---------

Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Christian Sigg
2023-12-08 14:42:12 -05:00
committed by GitHub
co-authored by Haicheng Wu
parent f4a0216601
commit e1483d5fa0
46 changed files with 308 additions and 273 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/53_hopper_gemm_permute/53_hopper_gemm_permute.cu
+3 -3
View File
@@ -197,14 +197,14 @@ template<class ... Shapes>
auto
select_mode_shape(Shapes const & ... shapes) {
auto permuted_shapes = filter_tuple(cute::make_tuple(shapes...), [](auto shape) {
if constexpr (rank(shape) > 1) {
if constexpr (cute::rank(shape) > 1) {
return cute::make_tuple(shape);
}
else {
return cute::make_tuple();
}
});
if constexpr (rank(permuted_shapes) == 0) {
if constexpr (cute::rank(permuted_shapes) == 0) {
return get<0>(cute::make_tuple(shapes...));
}
else {
@@ -251,7 +251,7 @@ auto
select_tile_shape(TileSize size, Shape const& shape)
{
static_assert(is_static<TileSize>::value, "Tile size must be static");
if constexpr (rank(Shape{}) == 0) {
if constexpr (cute::rank(Shape{}) == 0) {
return cute::make_tuple(size);
}
else {
examples/53_hopper_gemm_permute/permute_traits.hpp
+2 -2
View File
@@ -78,7 +78,7 @@ reshape(Shape const& shape, TargetShape const& target_shape)
template<class Permute, bool Transpose, class Shape, class Stride>
constexpr auto
make_permute_layout(Layout<Shape,Stride> const& layout) {
static_assert(rank(Shape{}) == 3, "Only rank-3 layouts are supported");
static_assert(cute::rank(Shape{}) == 3, "Only rank-3 layouts are supported");
if constexpr (Transpose) {
// Deal with tensor B by transposing appropriately before and after computing the permute layout.
// Its CuTe-canonical mode order is [N,K,L], while permute operations expect [row,col,batch].
@@ -135,7 +135,7 @@ using inverse_t = decltype(inverse(T{}));
template<class Permute, bool Transpose, class Shape, class Stride>
constexpr auto
make_original_layout(Layout<Shape,Stride> const& layout) {
static_assert(rank(Shape{}) == 3, "Only rank-3 layouts are supported");
static_assert(cute::rank(Shape{}) == 3, "Only rank-3 layouts are supported");
if constexpr (Transpose) {
// Deal with tensor B by transposing appropriately before and after computing the permute layout.
// Its CuTe-canonical mode order is [N,K,L], while permute operations expect [row,col,batch].