leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 24 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
include/cutlass/gemm/kernel/sm70_gemm.hpp
+7 -7
View File
@@ -60,7 +60,7 @@ public:
//
using ProblemShape = ProblemShape_;
static_assert(rank(ProblemShape{}) == 3 or rank(ProblemShape{}) == 4,
static_assert(cute::rank(ProblemShape{}) == 3 or cute::rank(ProblemShape{}) == 4,
"ProblemShape{} should be <M,N,K> or <M,N,K,L>");
// Mainloop derived types
@@ -142,7 +142,7 @@ public:
static bool
can_implement(Arguments const& args) {
return args.mode == GemmUniversalMode::kGemm or
(args.mode == GemmUniversalMode::kBatched && rank(ProblemShape{}) == 4);
(args.mode == GemmUniversalMode::kBatched && cute::rank(ProblemShape{}) == 4);
}
static int
@@ -159,7 +159,7 @@ public:
static dim3
get_grid_shape(Params const& params) {
int batch_count = 1;
if constexpr (rank(ProblemShape{}) == 4) {
if constexpr (cute::rank(ProblemShape{}) == 4) {
batch_count = cute::size<3>(params.problem_shape);
}
@@ -193,10 +193,10 @@ public:
auto L = get<3>(problem_shape_MNKL);
// Preconditions
static_assert(rank(StrideA{}) == 3, "StrideA must be rank-3: [M, K, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(rank(StrideB{}) == 3, "StrideB must be rank-3: [N, K, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(rank(StrideC{}) == 3, "StrideC must be rank-3: [M, N, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(rank(StrideD{}) == 3, "StrideD must be rank-3: [M, N, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(cute::rank(StrideA{}) == 3, "StrideA must be rank-3: [M, K, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(cute::rank(StrideB{}) == 3, "StrideB must be rank-3: [N, K, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(cute::rank(StrideC{}) == 3, "StrideC must be rank-3: [M, N, L]. If batch mode is not needed, set L stride to Int<0>.");
static_assert(cute::rank(StrideD{}) == 3, "StrideD must be rank-3: [M, N, L]. If batch mode is not needed, set L stride to Int<0>.");
// Get the appropriate blocks for this thread block -- potential for thread block locality
int thread_idx = int(threadIdx.x);