CUTLASS 3.8 Release (#2059)

* CUTLASS 3.8 Release

* update

* Update README.md

* Revert "Update README.md"

This reverts commit b353e36fe83e0815f99b44e46c0c95494c44726b.

* update

* update

---------

Co-authored-by: Haicheng Wu <57973641+hwu36@users.noreply.github.com>
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
mihir-awatramani
2025-01-25 02:44:06 -05:00
committed by GitHub
co-authored by Haicheng Wu Haicheng Wu
parent 9eb01fa0b0
commit 389e493055
290 changed files with 91222 additions and 291 deletions
@@ -388,6 +388,17 @@ public:
[[maybe_unused]] dim3 cluster(cute::size<0>(typename GemmKernel::DispatchPolicy::ClusterShape{}),
cute::size<1>(typename GemmKernel::DispatchPolicy::ClusterShape{}),
cute::size<2>(typename GemmKernel::DispatchPolicy::ClusterShape{}));
// Dynamic cluster support
[[maybe_unused]] dim3 fallback_cluster = dim3{0,0,0};
if constexpr (GemmKernel::ArchTag::kMinComputeCapability == 100
) {
if constexpr (!cute::is_static_v<typename GemmKernel::DispatchPolicy::ClusterShape>) {
fallback_cluster = params.hw_info.cluster_shape_fallback;
cluster = params.hw_info.cluster_shape;
}
}
[[maybe_unused]] void* kernel_params[] = {&params};
if constexpr (kEnableCudaHostAdapter) {
@@ -415,6 +426,7 @@ public:
else {
launch_result = cuda_adapter->launch(grid,
cluster,
fallback_cluster,
block,
smem_size,
stream,
@@ -430,8 +442,7 @@ public:
else {
CUTLASS_ASSERT(cuda_adapter == nullptr);
[[maybe_unused]] void const* kernel = (void const*) device_kernel<GemmKernel>;
static constexpr bool kClusterLaunch = GemmKernel::ArchTag::kMinComputeCapability == 90
;
static constexpr bool kClusterLaunch = GemmKernel::ArchTag::kMinComputeCapability == 90;
if constexpr (kClusterLaunch) {
if constexpr (is_static_1x1x1) {
#if (CUTLASS_DEBUG_TRACE_LEVEL > 1)
@@ -456,6 +467,42 @@ public:
grid, cluster, block, smem_size, stream, kernel, kernel_params, launch_with_pdl);
}
}
else {
if constexpr (GemmKernel::ArchTag::kMinComputeCapability == 100
) {
if constexpr (is_static_1x1x1) {
#if (CUTLASS_DEBUG_TRACE_LEVEL > 1)
CUTLASS_TRACE_HOST("GemmUniversal::run: Launching static 1x1x1 kernel");
#endif
launch_result = cutlass::kernel_launch<GemmKernel>(grid, block, smem_size, stream, params, launch_with_pdl);
if (launch_result != Status::kSuccess) {
CUTLASS_TRACE_HOST("GemmUniversal::run: cutlass::kernel_launch reports failure");
}
#if (CUTLASS_DEBUG_TRACE_LEVEL > 1)
else {
CUTLASS_TRACE_HOST("GemmUniversal::run: cutlass::kernel_launch reports success");
}
#endif
}
else {
#if (CUTLASS_DEBUG_TRACE_LEVEL > 1)
CUTLASS_TRACE_HOST("GemmUniversal::run: Launching kernel with fall-back cluster");
#endif
launch_result = ClusterLauncher::launch_with_fallback_cluster(
grid,
cluster,
fallback_cluster,
block,
smem_size,
stream,
kernel,
kernel_params,
launch_with_pdl);
}
}
}
}
}
else {