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

v4.3 update. (#2709)

Browse Source 
* v4.3 update.

* Update the cute_dsl_api changelog's doc link

* Update version to 4.3.0

* Update the example link

* Update doc to encourage user to install DSL from requirements.txt

---------

Co-authored-by: Larry Wu <larwu@nvidia.com>
This commit is contained in:
Junkai-Wu
2025-10-21 14:26:30 -04:00
committed by GitHub
co-authored by Larry Wu
parent e6e2cc29f5
commit b1d6e2c9b3
244 changed files with 59272 additions and 10455 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/library/src/gemm_operation_3x.hpp
+18 -15
View File
@@ -185,23 +185,11 @@ public:
/// Constructor
GemmUniversal3xOperation(char const *name = "unknown_gemm"):
GemmOperation3xBase<Operator_>(name, GemmKind::kUniversal) {
if constexpr (Operator::ArchTag::kMinComputeCapability == 90) {
dim3 cluster_dims(
cute::size<0>(typename Operator::GemmKernel::ClusterShape{}),
cute::size<1>(typename Operator::GemmKernel::ClusterShape{}),
cute::size<2>(typename Operator::GemmKernel::ClusterShape{}));
uint32_t threads_per_block = Operator::GemmKernel::MaxThreadsPerBlock;
void const* kernel_ptr = (void*)(device_kernel<typename Operator::GemmKernel>);
max_active_clusters = cutlass::KernelHardwareInfo::query_device_max_active_clusters(
cluster_dims,
threads_per_block,
kernel_ptr);
}
}
GemmOperation3xBase<Operator_>(name, GemmKind::kUniversal) {}
private:
int max_active_clusters{};
// mutable because it needs to be set in initialize (see comment in initialize)
mutable int max_active_clusters{};
protected:
@@ -683,6 +671,21 @@ public:
void *host_workspace,
void *device_workspace,
cudaStream_t stream = nullptr) const override {
// this would ideally go in the constructor, but
// the constructor is called at profiler startup for EVERY kernel,
// REGARDLESS of whether the kernel is actually supported on the device
if constexpr (Operator::ArchTag::kMinComputeCapability == 90) {
dim3 cluster_dims(
cute::size<0>(typename Operator::GemmKernel::ClusterShape{}),
cute::size<1>(typename Operator::GemmKernel::ClusterShape{}),
cute::size<2>(typename Operator::GemmKernel::ClusterShape{}));
uint32_t threads_per_block = Operator::GemmKernel::MaxThreadsPerBlock;
void const* kernel_ptr = (void*)(device_kernel<typename Operator::GemmKernel>);
max_active_clusters = cutlass::KernelHardwareInfo::query_device_max_active_clusters(
cluster_dims,
threads_per_block,
kernel_ptr);
}
Operator *op = new (host_workspace) Operator;
return Status::kSuccess;
}