Allow distinct K0/K1 values for A/B block descriptor (#98)

* add gitignore

* host tensor: allow generating sequentially increasing value in a given dimension

* gridwise gemm v3r1: allow distinct K0/K1 values for A/B block descriptor

- remove dangling header include
- modify example gemm_xdl accordingly
- infer KPack value from M/NPerXdl
- device conv2d fwd: update parameters accordingly for the underlying gridwise gemm v3r1
(API for conv2d fwd stays the same for now until we decide to expose individual K0s for activation and weight)

* add LDS data dump utility

* profiler: reflect API change for distinct K0/K1 for A/B matrices

* profiler: add conflict-free LDS write FP16 kernel instances

* fix accidental perf regression

* address feedback; cosmetic changes

* clang-format for new files

* format
Co-authored-by: Chao Liu <chao.liu2@amd.com>

.gitignore

+# Compiled Object files
+*.slo
+*.lo
+*.o
+*.obj
+
+# Precompiled Headers
+*.gch
+*.pch
+*.ipch
+
+# Compiled Dynamic libraries
+*.so
+*.dylib
+*.dll
+
+# Fortran module files
+*.mod
+
+# Compiled Static libraries
+*.lai
+*.la
+*.a
+*.lib
+
+# Executables
+*.exe
+*.out
+*.app
+
+# vim tags
+tags
+.tags
+.*.swp
+
+# Editors
+.vscode
+
+# build-in-source directory
+build*
+
+# emacs temporary/backup files
+.\#*
+\#*\#
+*~
+
+# GDB temporary files
+.gdb_history
\ No newline at end of file

composable_kernel/include/tensor_operation/blockwise_gemm_xdlops.hpp

@@ -17,7 +17,7 @@ template <index_t BlockSize,
           index_t NPerXDL,
           index_t MRepeat,
           index_t NRepeat,
-          index_t K1>
+          index_t KPack>
 struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 {
     static constexpr auto I0 = Number<0>{};
@@ -29,10 +29,15 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 
     static constexpr index_t MPerBlock = AK0MK1BlockDesc{}.GetLength(I1);
     static constexpr index_t NPerBlock = BK0NK1BlockDesc{}.GetLength(I1);
+    static constexpr index_t KPerBlock =
+        BK0NK1BlockDesc{}.GetLength(I0) * BK0NK1BlockDesc{}.GetLength(I2);
 
-    static constexpr index_t K0 = BK0NK1BlockDesc{}.GetLength(I0);
+    static constexpr index_t A_K0 = AK0MK1BlockDesc{}.GetLength(I0);
+    static constexpr index_t B_K0 = BK0NK1BlockDesc{}.GetLength(I0);
+    static constexpr index_t A_K1 = AK0MK1BlockDesc{}.GetLength(I2);
+    static constexpr index_t B_K1 = BK0NK1BlockDesc{}.GetLength(I2);
 
-    static constexpr auto xdlops_gemm = XdlopsGemm<FloatAB, MPerXDL, NPerXDL, K1>{};
+    static constexpr auto xdlops_gemm = XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack>{};
 
     static constexpr index_t MWaves = MPerBlock / (MRepeat * MPerXDL);
     static constexpr index_t NWaves = NPerBlock / (NRepeat * NPerXDL);
@@ -66,7 +71,7 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 
         const auto xdlops_a_idx = xdlops_gemm.CalculateAThreadOriginDataIndex();
 
-        return make_tuple(xdlops_a_idx[I0], 0, waveId_m, xdlops_a_idx[I1], 0);
+        return make_tuple(0, waveId_m, xdlops_a_idx[I1], Number<KPack>{} * xdlops_a_idx[I0]);
     }
 
     __device__ static auto CalculateBThreadOriginDataIndex()
@@ -77,7 +82,7 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 
         const auto xdlops_b_idx = xdlops_gemm.CalculateBThreadOriginDataIndex();
 
-        return make_tuple(xdlops_b_idx[I0], 0, waveId_n, xdlops_b_idx[I1], 0);
+        return make_tuple(0, waveId_n, xdlops_b_idx[I1], Number<KPack>{} * xdlops_b_idx[I0]);
     }
 
     template <index_t m0, index_t n0, index_t xdlops_i, index_t blk_i>
@@ -115,12 +120,6 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
                           BK0NK1BlockDesc::IsKnownAtCompileTime(),
                       "wrong! Desc should be known at compile-time");
 
-        static_assert(AK0MK1BlockDesc{}.GetLength(I0) == BK0NK1BlockDesc{}.GetLength(I0),
-                      "wrong! K0 dimension not consistent");
-
-        static_assert(AK0MK1BlockDesc{}.GetLength(I2) == BK0NK1BlockDesc{}.GetLength(I2),
-                      "wrong! K1 dimension not consistent");
-
         static_assert(BlockSize == MWaves * NWaves * WaveSize,
                       "BlockSize != MWaves * NWaves * WaveSize\n");
 
@@ -219,32 +218,32 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
             c_grid_desc_g_m0_n0_m1_n1_m2_n2);
     }
 
-    __host__ __device__ static constexpr auto MakeABlockDescriptor_K0_M0_M1_M2_K1()
+    __host__ __device__ static constexpr auto MakeABlockDescriptor_M0_M1_M2_K()
     {
         return transform_tensor_descriptor(
             AK0MK1BlockDesc{},
-            make_tuple(make_pass_through_transform(Number<K0>{}),
-                       make_unmerge_transform(
-                           make_tuple(Number<MRepeat>{}, Number<MWaves>{}, Number<MPerXDL>{})),
-                       make_pass_through_transform(Number<K1>{})),
-            make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
-            make_tuple(Sequence<0>{}, Sequence<1, 2, 3>{}, Sequence<4>{}));
+            make_tuple(
+                make_merge_transform_v3_division_mod(make_tuple(Number<A_K0>{}, Number<A_K1>{})),
+                make_unmerge_transform(
+                    make_tuple(Number<MRepeat>{}, Number<MWaves>{}, Number<MPerXDL>{}))),
+            make_tuple(Sequence<0, 2>{}, Sequence<1>{}),
+            make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}));
     }
 
-    __host__ __device__ static constexpr auto MakeBBlockDescriptor_K0_N0_N1_N2_K1()
+    __host__ __device__ static constexpr auto MakeBBlockDescriptor_N0_N1_N2_K()
     {
         return transform_tensor_descriptor(
             BK0NK1BlockDesc{},
-            make_tuple(make_pass_through_transform(Number<K0>{}),
-                       make_unmerge_transform(
-                           make_tuple(Number<NRepeat>{}, Number<NWaves>{}, Number<NPerXDL>{})),
-                       make_pass_through_transform(Number<K1>{})),
-            make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
-            make_tuple(Sequence<0>{}, Sequence<1, 2, 3>{}, Sequence<4>{}));
+            make_tuple(
+                make_merge_transform_v3_division_mod(make_tuple(Number<B_K0>{}, Number<B_K1>{})),
+                make_unmerge_transform(
+                    make_tuple(Number<NRepeat>{}, Number<NWaves>{}, Number<NPerXDL>{}))),
+            make_tuple(Sequence<0, 2>{}, Sequence<1>{}),
+            make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}));
     }
 
-    static constexpr auto a_block_desc_k0_m0_m1_m2_k1 = MakeABlockDescriptor_K0_M0_M1_M2_K1();
-    static constexpr auto b_block_desc_k0_n0_n1_n2_k1 = MakeBBlockDescriptor_K0_N0_N1_N2_K1();
+    static constexpr auto a_block_desc_m0_m1_m2_k = MakeABlockDescriptor_M0_M1_M2_K();
+    static constexpr auto b_block_desc_n0_n1_n2_k = MakeBBlockDescriptor_N0_N1_N2_K();
 
     template <typename ABlockBuffer, typename BBlockBuffer, typename CThreadBuffer>
     __device__ void Run(const ABlockBuffer& a_block_buf,
@@ -258,31 +257,31 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 
         static_for<0, MRepeat, 1>{}([&](auto m0) {
             // read A
-            a_thread_copy_.Run(a_block_desc_k0_m0_m1_m2_k1,
-                               make_tuple(I0, m0, I0, I0, I0),
+            a_thread_copy_.Run(a_block_desc_m0_m1_m2_k,
+                               make_tuple(m0, I0, I0, I0),
                                a_block_buf,
                                a_thread_desc_,
-                               make_tuple(I0, I0, I0, I0, I0),
+                               make_tuple(I0, I0, I0, I0),
                                a_thread_buf);
 
             static_for<0, NRepeat, 1>{}([&](auto n0) {
                 // read B
-                b_thread_copy_.Run(b_block_desc_k0_n0_n1_n2_k1,
-                                   make_tuple(I0, n0, I0, I0, I0),
+                b_thread_copy_.Run(b_block_desc_n0_n1_n2_k,
+                                   make_tuple(n0, I0, I0, I0),
                                    b_block_buf,
                                    b_thread_desc_,
-                                   make_tuple(I0, I0, I0, I0, I0),
+                                   make_tuple(I0, I0, I0, I0),
                                    b_thread_buf);
 
-                static_for<0, K0, xdlops_gemm.K0PerXdlops>{}([&](auto k0) {
-                    vector_type<FloatAB, K1> a_thread_vec;
-                    vector_type<FloatAB, K1> b_thread_vec;
+                static_for<0, KPerBlock, KPack * xdlops_gemm.K0PerXdlops>{}([&](auto k) {
+                    vector_type<FloatAB, KPack> a_thread_vec;
+                    vector_type<FloatAB, KPack> b_thread_vec;
 
-                    static_for<0, K1, 1>{}([&](auto i) {
+                    static_for<0, KPack, 1>{}([&](auto i) {
                         a_thread_vec.template AsType<FloatAB>()(i) = a_thread_buf
-                            [Number<a_thread_desc_.CalculateOffset(make_tuple(k0, 0, 0, 0, i))>{}];
+                            [Number<a_thread_desc_.CalculateOffset(make_tuple(0, 0, 0, k + i))>{}];
                         b_thread_vec.template AsType<FloatAB>()(i) = b_thread_buf
-                            [Number<b_thread_desc_.CalculateOffset(make_tuple(k0, 0, 0, 0, i))>{}];
+                            [Number<b_thread_desc_.CalculateOffset(make_tuple(0, 0, 0, k + i))>{}];
                     });
 
                     using mfma_input_type =
@@ -301,13 +300,13 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
     }
 
     private:
-    // A[K0, M0, M1, M2, K1]
+    // A[M0, M1, M2, KPerBlock]
     static constexpr auto a_thread_desc_ =
-        make_naive_tensor_descriptor_packed(make_tuple(Number<K0>{}, I1, I1, I1, Number<K1>{}));
+        make_naive_tensor_descriptor_packed(make_tuple(I1, I1, I1, Number<KPerBlock>{}));
 
-    // B[K0, N0, N1, N2, K1]
+    // B[N0, N1, N2, KPerBlock]
     static constexpr auto b_thread_desc_ =
-        make_naive_tensor_descriptor_packed(make_tuple(Number<K0>{}, I1, I1, I1, Number<K1>{}));
+        make_naive_tensor_descriptor_packed(make_tuple(I1, I1, I1, Number<KPerBlock>{}));
 
     // C[M, N, NumRegXdlops]
     static constexpr auto c_thread_desc_ = make_naive_tensor_descriptor_packed(
@@ -315,23 +314,23 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
 
     using AThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
                                                          FloatAB,
-                                                         decltype(a_block_desc_k0_m0_m1_m2_k1),
+                                                         decltype(a_block_desc_m0_m1_m2_k),
                                                          decltype(a_thread_desc_),
-                                                         Sequence<K0, 1, 1, 1, K1>,
-                                                         Sequence<0, 1, 2, 3, 4>,
-                                                         4,
-                                                         K1,
-                                                         K1>;
+                                                         Sequence<1, 1, 1, KPerBlock>,
+                                                         Sequence<0, 1, 2, 3>,
+                                                         3,
+                                                         A_K1,
+                                                         A_K1>;
 
     using BThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
                                                          FloatAB,
-                                                         decltype(b_block_desc_k0_n0_n1_n2_k1),
+                                                         decltype(b_block_desc_n0_n1_n2_k),
                                                          decltype(b_thread_desc_),
-                                                         Sequence<K0, 1, 1, 1, K1>,
-                                                         Sequence<0, 1, 2, 3, 4>,
-                                                         4,
-                                                         K1,
-                                                         K1>;
+                                                         Sequence<1, 1, 1, KPerBlock>,
+                                                         Sequence<0, 1, 2, 3>,
+                                                         3,
+                                                         B_K1,
+                                                         B_K1>;
 
     AThreadCopy a_thread_copy_{CalculateAThreadOriginDataIndex()};
     BThreadCopy b_thread_copy_{CalculateBThreadOriginDataIndex()};

composable_kernel/include/utility/common_header.hpp

@@ -35,6 +35,7 @@
 #include "transpose_vectors.hpp"
 #include "inner_product.hpp"
 #include "element_wise_operation.hpp"
+#include "debug.hpp"
 
 // TODO: remove this
 #if CK_USE_AMD_INLINE_ASM

composable_kernel/include/utility/debug.hpp

+#ifndef UTILITY_DEBUG_HPP
+#define UTILITY_DEBUG_HPP
+
+namespace ck {
+namespace debug {
+
+namespace detail {
+template <typename T, typename Enable = void>
+struct PrintAsType;
+
+template <typename T>
+struct PrintAsType<T, typename std::enable_if<std::is_floating_point<T>::value>::value>
+{
+    using type = float;
+};
+
+template <>
+struct PrintAsType<ck::half_t, void>
+{
+    using type = float;
+};
+
+template <typename T>
+struct PrintAsType<T, typename std::enable_if<std::is_integral<T>::value>::value>
+{
+    using type = int;
+};
+} // namespace detail
+
+// Print at runtime the data in shared memory in 128 bytes per row format given shared mem pointer
+// and the number of elements. Can optionally specify strides between elements and how many bytes'
+// worth of data per row.
+//
+// Usage example:
+//
+//   debug::print_shared(a_block_buf.p_data_, index_t(a_block_desc_k0_m_k1.GetElementSpaceSize()));
+//
+template <typename T, index_t element_stride = 1, index_t row_bytes = 128>
+__device__ void print_shared(T const* p_shared, index_t num_elements)
+{
+    using PrintType                = typename detail::PrintAsType<T>::type;
+    constexpr index_t row_elements = row_bytes / sizeof(T);
+    static_assert((element_stride >= 1 && element_stride <= row_elements),
+                  "element_stride should between [1, row_elements]");
+
+    index_t wgid = blockIdx.x + blockIdx.y * gridDim.x + gridDim.x * gridDim.y * blockIdx.z;
+    index_t tid =
+        (threadIdx.z * (blockDim.x * blockDim.y)) + (threadIdx.y * blockDim.x) + threadIdx.x;
+
+    __syncthreads();
+
+    if(tid == 0)
+    {
+        printf("\nWorkgroup id %d, bytes per row %d, element stride %d\n\n",
+               wgid,
+               row_bytes,
+               element_stride);
+        for(index_t i = 0; i < num_elements; i += row_elements)
+        {
+            printf("elem %5d: ", i);
+            for(index_t j = 0; j < row_elements; j += element_stride)
+            {
+                printf("%.0f ", static_cast<PrintType>(p_shared[i + j]));
+            }
+
+            printf("\n");
+        }
+        printf("\n");
+    }
+
+    __syncthreads();
+}
+
+} // namespace debug
+} // namespace ck
+
+#endif // UTILITY_DEBUG_HPP

device_operation/include/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp

@@ -432,10 +432,11 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
         OutElementwiseOperation,
         MPerBlock,
         NPerBlock,
-        K0PerBlock,
+        K0PerBlock * K1,
+        K1, // AK1
+        K1, // BK1
         MPerXdl,
         NPerXdl,
-        K1,
         MXdlPerWave,
         NXdlPerWave,
         ABlockTransferThreadClusterLengths_K0_M_K1,

device_operation/include/device_gemm_xdl_c_shuffle.hpp

@@ -29,8 +29,9 @@ template <
     ck::index_t BlockSize,
     ck::index_t MPerBlock,
     ck::index_t NPerBlock,
-    ck::index_t K0PerBlock,
-    ck::index_t K1,
+    ck::index_t KPerBlock,
+    ck::index_t AK1,
+    ck::index_t BK1,
     ck::index_t MPerXDL,
     ck::index_t NPerXDL,
     ck::index_t MXdlPerWave,
@@ -61,13 +62,11 @@ struct DeviceGemmXdl_C_Shuffle
     static constexpr auto I1 = Number<1>{};
     static constexpr auto I2 = Number<2>{};
 
-    static constexpr auto K1Number = Number<K1>{};
-
     static auto MakeAGridDescriptor_K0_M_K1(index_t M, index_t K, index_t StrideA)
     {
-        assert(K % K1 == 0);
+        assert(K % AK1 == 0);
 
-        const index_t K0 = K / K1;
+        const index_t K0 = K / AK1;
 
         const auto a_grid_desc_m_k = [&]() {
             if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
@@ -80,21 +79,20 @@ struct DeviceGemmXdl_C_Shuffle
             }
         }();
 
-        const auto a_grid_desc_k0_m_k1 =
-            transform_tensor_descriptor(a_grid_desc_m_k,
-                                        make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
-                                                   make_pass_through_transform(M)),
-                                        make_tuple(Sequence<1>{}, Sequence<0>{}),
-                                        make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+        const auto a_grid_desc_k0_m_k1 = transform_tensor_descriptor(
+            a_grid_desc_m_k,
+            make_tuple(make_unmerge_transform(make_tuple(K0, AK1)), make_pass_through_transform(M)),
+            make_tuple(Sequence<1>{}, Sequence<0>{}),
+            make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
 
         return a_grid_desc_k0_m_k1;
     }
 
     static auto MakeBGridDescriptor_K0_N_K1(index_t K, index_t N, index_t StrideB)
     {
-        assert(K % K1 == 0);
+        assert(K % BK1 == 0);
 
-        const index_t K0 = K / K1;
+        const index_t K0 = K / BK1;
 
         const auto b_grid_desc_k_n = [&]() {
             if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
@@ -107,12 +105,11 @@ struct DeviceGemmXdl_C_Shuffle
             }
         }();
 
-        const auto b_grid_desc_k0_n_k1 =
-            transform_tensor_descriptor(b_grid_desc_k_n,
-                                        make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
-                                                   make_pass_through_transform(N)),
-                                        make_tuple(Sequence<0>{}, Sequence<1>{}),
-                                        make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+        const auto b_grid_desc_k0_n_k1 = transform_tensor_descriptor(
+            b_grid_desc_k_n,
+            make_tuple(make_unmerge_transform(make_tuple(K0, BK1)), make_pass_through_transform(N)),
+            make_tuple(Sequence<0>{}, Sequence<1>{}),
+            make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
 
         return b_grid_desc_k0_n_k1;
     }
@@ -148,10 +145,11 @@ struct DeviceGemmXdl_C_Shuffle
         CElementwiseOperation,
         MPerBlock,
         NPerBlock,
-        K0PerBlock,
+        KPerBlock,
+        AK1,
+        BK1,
         MPerXDL,
         NPerXDL,
-        K1,
         MXdlPerWave,
         NXdlPerWave,
         ABlockTransferThreadClusterLengths_K0_M_K1,
@@ -461,7 +459,9 @@ struct DeviceGemmXdl_C_Shuffle
             << BlockSize << ", "
             << MPerBlock << ", "
             << NPerBlock << ", "
-            << K0PerBlock
+            << KPerBlock << ", "
+            << AK1 << ", "
+            << BK1
             << ">";
         // clang-format on
 

device_operation/src/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -23,23 +23,23 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tuple<
     // clang-format off
-        //#####################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|      Num|
-        //#####################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector| Prefetch|
-        //#####################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         |
-        //#####################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |         |
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>,
-        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>
+        //#####################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|      Num|
+        //#####################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector| Prefetch|
+        //#####################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         |
+        //#####################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |         |
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>,
+        DeviceGemmXdl_C_Shuffle<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8,        2>
     // clang-format on
     >;
 

device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp

@@ -21,23 +21,30 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[k, m] * b[k, n] = c[m, n]
-using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances =
-    std::tuple<
-        // clang-format off
-        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
-        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
-        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
-        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
-        // clang-format on
-        >;
+using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   4,   4,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   2,   2,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   2,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   2,   2,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   2,   2,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   2,   2,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
     std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)

device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp

@@ -21,23 +21,30 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[k, m] * b[n, k] = c[m, n]
-using device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances =
-    std::tuple<
-        // clang-format off
-        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
-        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
-        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
-        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
-        // clang-format on
-        >;
+using device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::tuple<
+    // clang-format off
+        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   2,   8,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   2,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   2,   8,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   2,   8,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   2,   8,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,     false,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
     std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)

device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -21,23 +21,30 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
-using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances =
-    std::tuple<
-        // clang-format off
-        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
-        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
-        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
-        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
-        // clang-format on
-        >;
+using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,     false,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
     std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)

device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -21,28 +21,27 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
-using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances =
-    std::tuple<
-        // clang-format off
-        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
-        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
-        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
-        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
-        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
-        // clang-format on
-        >;
+using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::tuple<
+    // clang-format off
+        //#####################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //#####################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //#####################|     |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>,
+        DeviceGemmXdl_C_Shuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
     std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)

example/1_gemm_xdl/gemm_xdl.cpp

@@ -45,7 +45,7 @@ static constexpr auto GemmDefault   = ck::tensor_operation::device::GemmSpeciali
 static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization_t::MNPadding;
 
 // clang-format off
-#if 1
+#if 0
 using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl
 //######| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|      Num|
 //######|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Spacialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|
@@ -53,6 +53,13 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl
 //######|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |
 //    [256, 128, 4, 8], 1 stage, 2 occupancy
         <   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1>;
+#elif 1
+using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl_C_Shuffle
+//######|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+//######| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+//######|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+//######|     |      |      |        |        |        |        |            |            |            |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        <  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>;
 #elif 0
 using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl
 //######| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|      Num|
@@ -82,14 +89,6 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl
 //    [ 64, 144, 4, 8], 2 stage, 2 occupancy
 //     85 TFlops
 //      <   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    64,   144,     4,  8,   16,   16,    1,    9,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,      true,               7,               1,        2>;
-#elif 1
-using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdl_C_Shuffle
-//######| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|      Num|
-//######|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector| Prefetch|
-//######|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         |
-//######|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |         |
-//    [128, 144, 8, 8], 1 stage, 1 occupancy, bounded by LDS size
-        <   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   144,     8,  8,   16,   16,    2,    9,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8,  8, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,      true,           1,           9,             S<1, 1,  8, 1, 9, 2>,               8,        1>;
 #endif
 // clang-format on
 
@@ -156,8 +155,8 @@ int main(int argc, char* argv[])
 
     Tensor<ADataType> a_m_k(f_host_tensor_descriptor(M, K, StrideA, ALayout{}));
     Tensor<BDataType> b_k_n(f_host_tensor_descriptor(K, N, StrideB, BLayout{}));
-    Tensor<BDataType> c_m_n_host_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
-    Tensor<BDataType> c_m_n_device_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+    Tensor<CDataType> c_m_n_host_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+    Tensor<CDataType> c_m_n_device_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
 
     std::cout << "a_m_k: " << a_m_k.mDesc << std::endl;
     std::cout << "b_k_n: " << b_k_n.mDesc << std::endl;
@@ -170,9 +169,13 @@ int main(int argc, char* argv[])
         a_m_k.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
         b_k_n.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
         break;
-    default:
+    case 2:
         a_m_k.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
         b_k_n.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
+        break;
+    default:
+        a_m_k.GenerateTensorValue(GeneratorTensor_Sequential<0>{});
+        b_k_n.GenerateTensorValue(GeneratorTensor_Sequential<1>{});
     }
 
     DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpace());

host/host_tensor/include/host_tensor_generator.hpp

@@ -154,4 +154,15 @@ struct GeneratorTensor_Checkboard
     }
 };
 
+template <ck::index_t Dim>
+struct GeneratorTensor_Sequential
+{
+    template <typename... Ts>
+    float operator()(Ts... Xs) const
+    {
+        std::array<ck::index_t, sizeof...(Ts)> dims = {{static_cast<ck::index_t>(Xs)...}};
+        return dims[Dim];
+    }
+};
+
 #endif