From 91c829504aed2d522b9982a56d708e7ab3c94d2d Mon Sep 17 00:00:00 2001
From: ZheWang <35656954+eeezio@users.noreply.github.com>
Date: Tue, 13 Jan 2026 10:26:45 +0800
Subject: [PATCH] fix mxfp8-gemm example failure (#3531)

Co-authored-by: ZheWang <zhewan@amd.com>

[ROCm/composable_kernel commit: a575acb245847d96d54c1e6d198748bda3e57952]
---
 include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp b/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp
index 97f936fde9..4f636b5962 100644
--- a/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp
+++ b/include/ck_tile/ops/epilogue/cshuffle_epilogue.hpp
@@ -331,10 +331,15 @@ struct CShuffleEpilogue
             }
             else
             {
+#if defined(__gfx950__)
+                constexpr auto is_950 = true;
+#else
+                constexpr auto is_950 = false;
+#endif
                 constexpr int RakedXDLN_PerWarp = NumNXdlPerWavePerShuffle / BlockedXDLN_PerWarp;
                 // BlockedLayout
                 // this branch is for original a16w4
-                if constexpr(is_any_of<ADataType, pk_int4_t, pk_fp4_t>::value ||
+                if constexpr(is_950 || is_any_of<ADataType, pk_int4_t, pk_fp4_t>::value ||
                              is_any_of<BDataType, pk_int4_t, pk_fp4_t>::value)
                 {
                     return tile_distribution_encoding<