From adbfcad03b128ea6c415f8330c5d6bdfddc4615f Mon Sep 17 00:00:00 2001
From: "Graner, Johannes" <Johannes.Graner@amd.com>
Date: Thu, 18 Dec 2025 05:07:59 -0500
Subject: [PATCH] Re-enable two stage kernel

---
 ...multiple_d_splitk_xdl_cshuffle_two_stage.hpp | 17 -----------------
 1 file changed, 17 deletions(-)

diff --git a/include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_two_stage.hpp b/include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_two_stage.hpp
index 9b5aab5c85..08b4c87218 100644
--- a/include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_two_stage.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_two_stage.hpp
@@ -848,23 +848,6 @@ struct DeviceGroupedGemmMultipleDSplitKXdlCShuffleTwoStage
             return false;
         }
 
-        // TODO: Fix this.
-        // Error appears in `script/profiler_grouped_gemm.sh grouped_gemm 1 0 1 1 0 0`
-        if(std::is_same<ALayout, tensor_layout::gemm::RowMajor>::value &&
-           std::is_same<BLayout, tensor_layout::gemm::RowMajor>::value &&
-           std::is_same<ELayout, tensor_layout::gemm::RowMajor>::value &&
-           getGemmSpecializationString(GemmSpec) == "MNKPadding" && arg.K_BATCH > 2)
-        {
-            if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
-            {
-                std::cout
-                    << "All RowMajor layout with MNKPadding specialization and KBatch > 2 is not "
-                       "supported for all possible shapes!"
-                    << std::endl;
-            }
-            return false;
-        }
-
         bool supported = true;
         bool isWave64  = get_warp_size() == 64;
         for(std::size_t i = 0; i < arg.gemm_kernel_args_.size(); ++i)