tweak

host/driver_offline/include/device_gemm_xdlops_km_kn_mn.hpp

@@ -33,7 +33,35 @@ void device_gemm_xdlops_km_kn_mn(const ADesc& a_k_m_grid_desc,
     b_k_n_device_buf.ToDevice(b_k_n.mData.data());
     c_m_n_device_buf.ToDevice(c_m_n.mData.data());
 
-#if 1
+#if 0
+    // [M, N, K0, K1] = [256, 128, 4, 4] for fp32
+    constexpr index_t BlockSize = 256;
+
+    constexpr index_t MPerBlock = 256;
+    constexpr index_t NPerBlock = 128;
+    constexpr index_t KPerBlock = 4;
+
+    constexpr index_t MPerXDL = 32;
+    constexpr index_t NPerXDL = 32;
+    constexpr index_t K1      = 4;
+
+    constexpr index_t MRepeat = 4;
+    constexpr index_t NRepeat = 2;
+
+    using ABlockTransferThreadSliceLengths_K0_M_K1   = Sequence<1, 4, 4>;
+    using ABlockTransferThreadClusterLengths_K0_M_K1 = Sequence<4, 64, 1>;
+
+    constexpr index_t ABlockTransferSrcScalarPerVector_M  = 4;
+    constexpr index_t ABlockTransferDstScalarPerVector_K1 = 4;
+
+    using BBlockTransferThreadSliceLengths_K0_N_K1   = Sequence<1, 2, 4>;
+    using BBlockTransferThreadClusterLengths_K0_N_K1 = Sequence<4, 64, 1>;
+
+    constexpr index_t BBlockTransferSrcScalarPerVector_N  = 2;
+    constexpr index_t BBlockTransferDstScalarPerVector_K1 = 4;
+
+    constexpr index_t CThreadTransferDstScalarPerVector = 1;
+#elif 1
     // [M, N, K0, K1] = [256, 128, 4, 8] for fp16
     constexpr index_t BlockSize = 256;
 

host/driver_offline/src/gemm_driver_offline.cpp

@@ -16,8 +16,8 @@
 #include "device_gemm_xdlops_mk_nk_mn.hpp"
 #include "device_gemm_xdlops_km_kn_mn.hpp"
 
-#define USE_GEMM_XDL_MK_KN_MN 0
-#define USE_GEMM_XDL_MK_NK_MN 0
+#define USE_GEMM_XDL_MK_KN_MN 1
+#define USE_GEMM_XDL_MK_NK_MN 1
 #define USE_GEMM_XDL_KM_KN_MN 1
 
 enum GemmAlgo