[rocm-libraries] ROCm/rocm-libraries#5729 (commit 516c974)

[CK_TILE] Changed cshuffle LDS descriptor to naive layout
 (#5729)

## Motivation
This PR changes gemm/convolution cshuffle layout into plain one. to
improve cshuffle operation performance.

## Technical Details
The purpose is that before this change the cshuffle layout was having
some descriptor transformations that were probably aimed at reducing LDS
bank conflicts, but the transformations itself were terribly slow, which
negatively impacted the performance.

## Test Plan
There is no need for additional tests, since current tests cover this
functionality.

Author: jakpiase

include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp

@@ -336,9 +336,6 @@ struct CShuffleEpilogue
             constexpr index_t BaseWords  = ToWords(BaseStrideElems);
             constexpr index_t PadWords   = ((BaseWords % 2) == 0) ? 1 : 0;
             constexpr auto PaddingAmount = PadWords * ElemsPer4B;
-#else
-            constexpr auto PaddingAmount = 0;
-#endif
 
             constexpr auto lds_block_desc_0 = make_naive_tensor_descriptor(
                 make_tuple(number<MPerIterationShuffle / MLdsLayer>{},
@@ -369,6 +366,18 @@ struct CShuffleEpilogue
                 make_tuple(sequence<0>{}, sequence<1>{}));
 
             return lds_block_desc;
+
+#else
+            constexpr auto PaddingAmount = 0;
+
+            constexpr auto lds_block_desc = make_naive_tensor_descriptor(
+                make_tuple(number<MPerIterationShuffle>{}, number<NPerIterationShuffle>{}),
+                make_tuple(number<NPerIterationShuffle + PaddingAmount>{}, number<1>{}),
+                number<VectorLen>{},
+                number<1>{});
+
+            return lds_block_desc;
+#endif
         }
         // M is contiguous dimension
         else if constexpr(std::is_same_v<ELayout, tensor_layout::gemm::ColumnMajor>)