Updates for CUTLASS 3.4.1 (#1346)

* Updates for CUTLASS 3.4.1

* minor epi change

include/cute/arch/copy_sm90_desc.hpp

@@ -265,7 +265,7 @@ tma_descriptor_replace_dims_strides_in_shared_mem(TmaDescriptor
     asm volatile (
       "tensormap.replace.tile.global_dim.shared::cta.b1024.b32 [%0], 2, %1;"
       :: "l"(smem_int64_desc), "r"(prob_shape[2]));
-    // Strides must be a multiple of 16. Also, stride for the intermost dimension is implicitly 1 
+    // Strides must be a multiple of 16. Also, stride for the intermost dimension is implicitly 1
     asm volatile (
       "tensormap.replace.tile.global_stride.shared::cta.b1024.b64 [%0], 0, %1;"
       :: "l"(smem_int64_desc), "l"(prob_stride[1] >> 4));

include/cute/atom/mma_atom.hpp

@@ -391,7 +391,7 @@ struct TiledMMA : MMA_Atom
     } else {
       return cute::max(core_size, perm_size);
     }
-
+  
     CUTE_GCC_UNREACHABLE;
   }
 

include/cute/util/type_traits.hpp

@@ -125,6 +125,9 @@ using CUTE_STL_NAMESPACE::invoke_result_t;
 using CUTE_STL_NAMESPACE::common_type;
 using CUTE_STL_NAMESPACE::common_type_t;
 
+using CUTE_STL_NAMESPACE::remove_pointer;
+using CUTE_STL_NAMESPACE::remove_pointer_t;
+
 // <utility>
 using CUTE_STL_NAMESPACE::declval;