Move CUTLASS pin to v3.9.0

alexsamardzic · alexsamardzic · commit daddb71229ec · 2025-04-28T13:35:27.000Z
diff --git a/third_party/cutlass b/third_party/cutlass
@@ -1 +1 @@
-Subproject commit afa1772203677c5118fcd82537a9c8fefbcc7008
+Subproject commit e94e888df3551224738bfa505787b515eae8352f
diff --git a/torchao/csrc/cuda/mx_kernels/mx_fp_cutlass_kernels.cu b/torchao/csrc/cuda/mx_kernels/mx_fp_cutlass_kernels.cu
@@ -107,7 +107,7 @@ void run_gemm(at::Tensor& a, at::Tensor& b, at::Tensor& a_scale,
   using StrideD   = typename Gemm::GemmKernel::StrideD;
   using LayoutSFA = typename Gemm::GemmKernel::CollectiveMainloop::LayoutSFA;
   using LayoutSFB = typename Gemm::GemmKernel::CollectiveMainloop::LayoutSFB;
-  using Sm100BlkScaledConfig = typename Gemm::GemmKernel::CollectiveMainloop::Sm100BlkScaledConfig;
+  using Sm100BlkScaledConfig = typename Gemm::GemmKernel::CollectiveMainloop::Sm1xxBlkScaledConfig;
 
   // Initialize strides using packed stride configuration
   auto stride_A = cutlass::make_cute_packed_stride(StrideA{}, make_shape(M, K, 1));
