add small K1 test

example/11_conv2d_bwd_weight/conv2d_bwd_weight_xdl.cpp

@@ -46,26 +46,26 @@ using DeviceConvBwdWeightInstance = ck::tensor_operation::device::
         256,                              // BlockSize
         128,                              // MPerBlock
         128,                              // NPerBlock
-        4,                                // K0PerBlock
-        8,                                // K1
+        16,                                // K0PerBlock
+        2,                                // K1
         32,                               // MPerXdl
         32,                               // NPerXdl
         2,                                // MXdlPerWave
         2,                                // NXdlPerWave
-        S<1, 4, 16, 4>,                   // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<1, 4, 32, 1>,                   // ABlockTransferThreadClusterLengths_K0_M_K1
         S<0, 3, 1, 2>,                    // ABlockTransferThreadClusterArrangeOrder
         S<0, 2, 1, 3>,                    // ABlockTransferSrcAccessOrder
         2,                                // ABlockTransferSrcVectorDim
-        8,                                // ABlockTransferSrcScalarPerVector
+        4,                                // ABlockTransferSrcScalarPerVector
         2,                                // ABlockTransferDstScalarPerVector_K1
-        true,                             // ABlockLdsAddExtraM
-        S<1, 4, 16, 4>,                   // BBlockTransferThreadClusterLengths_K0_N_K1
+        false,                             // ABlockLdsAddExtraM
+        S<1, 4, 32, 1>,                   // BBlockTransferThreadClusterLengths_K0_N_K1
         S<0, 3, 1, 2>,                    // BBlockTransferThreadClusterArrangeOrder
         S<0, 2, 1, 3>,                    // BBlockTransferSrcAccessOrder
         2,                                // BBlockTransferSrcVectorDim
-        8,                                // BBlockTransferSrcScalarPerVector
+        4,                                // BBlockTransferSrcScalarPerVector
         2,                                // BBlockTransferDstScalarPerVector_K1
-        true,                             // BBlockLdsAddExtraN
+        false,                             // BBlockLdsAddExtraN
         1,                                // CShuffleMXdlPerWavePerShuffle
         1,                                // CShuffleNXdlPerWavePerShuffle
         S<1, 32, 1, 4>,                   // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r4r2.hpp

@@ -557,6 +557,9 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
         //       register
         // sanity check
 
+        constexpr index_t KPack = math::max(
+            K1, MfmaSelector<FloatAB, MPerXDL, NPerXDL>::selected_mfma.k_per_blk);
+
         auto blockwise_gemm =
             BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
                                                                 FloatAB,
@@ -567,7 +570,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
                                                                 NPerXDL,
                                                                 MRepeat,
                                                                 NRepeat,
-                                                                K1>{};
+                                                                KPack>{};
 
         auto c_thread_buf = blockwise_gemm.GetCThreadBuffer();