Merge branch 'develop' into uif2-initial

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v7r2.hpp

@@ -132,9 +132,6 @@ struct ThreadwiseTensorSliceTransfer_v7r2
             Number<num>{});
     }
 
-    template <typename T>
-    using has_vec_len = decltype(std::declval<T&>().vec_len);
-
     // SrcDescs: Tuple<const SrcDesc0&, const SrcDesc1&, ...>
     // SrcBuffers: Tuple<const SrcBuffer0&, const SrcBuffer1&, ...>
     template <typename SrcBuffers,
@@ -159,94 +156,63 @@ struct ThreadwiseTensorSliceTransfer_v7r2
                                                            is_src_valid);
             });
 
-            if constexpr(is_detected<has_vec_len, decltype(element_op_)>::value)
-            {
-                constexpr auto elem_op_vec_len = decltype(element_op_)::vec_len;
-
-                static_assert(is_same<remove_cvref_t<decltype(elem_op_vec_len)>, index_t>::value,
-                              "vec_len in element_op_ type is not index_t");
+            constexpr auto get_elem_op_vec_len = []() {
+                if constexpr(is_detected<is_pack8_invocable_t, decltype(element_op_)>::value)
+                {
+                    if constexpr(decltype(element_op_)::is_pack8_invocable)
+                        return math::min(8, SrcScalarPerVector);
+                }
+                if constexpr(is_detected<is_pack4_invocable_t, decltype(element_op_)>::value)
+                {
+                    if constexpr(decltype(element_op_)::is_pack4_invocable)
+                        return math::min(4, SrcScalarPerVector);
+                }
+                if constexpr(is_detected<is_pack2_invocable_t, decltype(element_op_)>::value)
+                {
+                    if constexpr(decltype(element_op_)::is_pack2_invocable)
+                        return math::min(2, SrcScalarPerVector);
+                }
+                return 1;
+            };
+
+            constexpr index_t elem_op_vec_len = get_elem_op_vec_len();
+
+            // apply pointwise function
+            static_for<0, SrcScalarPerVector / elem_op_vec_len, 1>{}([&](auto i) {
+                // get reference to src data
+                const auto src_data_refs = generate_tie(
+                    // return type should be lvalue
+                    [&](auto iSrc) -> const auto& {
+                        using SrcData = remove_cvref_t<tuple_element_t<iSrc.value, SrcDatas>>;
+
+                        using elem_op_vec_t = typename vector_type<SrcData, elem_op_vec_len>::type;
+
+                        return src_vectors[iSrc].template AsType<elem_op_vec_t>()[i];
+                    },
+                    Number<nSrc>{});
+
+                // get reference to dst data
+                auto dst_data_refs = generate_tie(
+                    // return type should be lvalue
+                    [&](auto iDst) -> auto& {
+                        using DstData = remove_cvref_t<tuple_element_t<iDst.value, DstDatas>>;
+
+                        using elem_op_vec_t = typename vector_type<DstData, elem_op_vec_len>::type;
+
+                        return dst_vectors(iDst).template AsType<elem_op_vec_t>()(i);
+                    },
+                    Number<nDst>{});
 
-                static_assert(elem_op_vec_len == 1 || elem_op_vec_len == 2 ||
-                                  elem_op_vec_len == 4 || elem_op_vec_len == 8,
-                              "vec_len in element_op_ must be 1, 2, 4, 8");
-
-                static_assert(SrcScalarPerVector % elem_op_vec_len == 0,
-                              "vec_len in element_op_ cannot be divided by SrcScalarPerVector!");
-
-                // apply pointwise function
-                static_for<0, SrcScalarPerVector / elem_op_vec_len, 1>{}([&](auto i) {
-                    // get reference to src data
-                    const auto src_data_refs = generate_tie(
-                        // return type should be lvalue
-                        [&](auto iSrc) -> const auto& {
-                            using SrcData = remove_cvref_t<tuple_element_t<iSrc.value, SrcDatas>>;
-
-                            using elem_op_vec_t =
-                                typename vector_type<SrcData, elem_op_vec_len>::type;
-
-                            return src_vectors[iSrc].template AsType<elem_op_vec_t>()[i];
-                        },
-                        Number<nSrc>{});
-
-                    // get reference to dst data
-                    auto dst_data_refs = generate_tie(
-                        // return type should be lvalue
-                        [&](auto iDst) -> auto& {
-                            using DstData = remove_cvref_t<tuple_element_t<iDst.value, DstDatas>>;
-
-                            using elem_op_vec_t =
-                                typename vector_type<DstData, elem_op_vec_len>::type;
-
-                            return dst_vectors(iDst).template AsType<elem_op_vec_t>()(i);
-                        },
-                        Number<nDst>{});
-
-                    // apply pointwise function
-                    // pointwise function signature:
-                    // element_op_(dst_data_refs[I0],
-                    //             dst_data_refs[I1],
-                    //             ...,
-                    //             src_data_refs[I0],
-                    //             src_data_refs[I1],
-                    //             ...)
-                    unpack2(element_op_, dst_data_refs, src_data_refs);
-                });
-            }
-            else
-            {
                 // apply pointwise function
-                static_for<0, SrcScalarPerVector, 1>{}([&](auto i) {
-                    // get reference to src data
-                    const auto src_data_refs = generate_tie(
-                        // return type should be lvalue
-                        [&](auto iSrc) -> const auto& {
-                            using SrcData = remove_cvref_t<tuple_element_t<iSrc.value, SrcDatas>>;
-
-                            return src_vectors[iSrc].template AsType<SrcData>()[i];
-                        },
-                        Number<nSrc>{});
-
-                    // get reference to dst data
-                    auto dst_data_refs = generate_tie(
-                        // return type should be lvalue
-                        [&](auto iDst) -> auto& {
-                            using DstData = remove_cvref_t<tuple_element_t<iDst.value, DstDatas>>;
-
-                            return dst_vectors(iDst).template AsType<DstData>()(i);
-                        },
-                        Number<nDst>{});
-
-                    // apply pointwise function
-                    // pointwise function signature:
-                    // element_op_(dst_data_refs[I0],
-                    //             dst_data_refs[I1],
-                    //             ...,
-                    //             src_data_refs[I0],
-                    //             src_data_refs[I1],
-                    //             ...)
-                    unpack2(element_op_, dst_data_refs, src_data_refs);
-                });
-            }
+                // pointwise function signature:
+                // element_op_(dst_data_refs[I0],
+                //             dst_data_refs[I1],
+                //             ...,
+                //             src_data_refs[I0],
+                //             src_data_refs[I1],
+                //             ...)
+                unpack2(element_op_, dst_data_refs, src_data_refs);
+            });
 
             dst_vectors_tuple_(iAccess) = dst_vectors;
 

include/ck/utility/amd_buffer_addressing.hpp

@@ -299,584 +299,255 @@ enum struct AmdBufferCoherenceEnum
     GLC_SLC          = 3,
 };
 
-template <typename T,
-          index_t N,
-          AmdBufferCoherenceEnum coherence = AmdBufferCoherenceEnum::DefaultCoherence>
-__device__ typename vector_type<T, N>::type amd_buffer_load_impl(int32x4_t src_wave_buffer_resource,
-                                                                 index_t src_thread_addr_offset,
-                                                                 index_t src_wave_addr_offset)
+template <index_t N, AmdBufferCoherenceEnum coherence = AmdBufferCoherenceEnum::DefaultCoherence>
+__device__ typename vector_type<int8_t, N>::type
+amd_buffer_load_impl_raw(int32x4_t src_wave_buffer_resource,
+                         index_t src_thread_addr_offset,
+                         index_t src_wave_addr_offset)
 {
-    static_assert(
-        (is_same<T, double>::value && (N == 1 || N == 2 || N == 4)) ||
-            (is_same<T, float>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, half_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, bhalf_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, int32_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, int8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)),
-        "wrong! not implemented");
+    static_assert(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32 || N == 64,
+                  "wrong! not implemented");
 
-    if constexpr(is_same<T, double>::value)
+    if constexpr(N == 1)
     {
-        // use fp32 load to mimic fp64 load
-        if constexpr(N == 1)
-        {
-            const float2_t tmp =
-                llvm_amdgcn_raw_buffer_load_fp32x2(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
-
-            return bit_cast<double>(tmp);
-        }
-        else if constexpr(N == 2)
-        {
-            const float4_t tmp =
-                llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
-
-            return bit_cast<double2_t>(tmp);
-        }
-        else if constexpr(N == 4)
-        {
-            const float4_t f32_0 =
-                llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
-
-            const float4_t f32_1 =
-                llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset + 4 * sizeof(float),
-                                                   static_cast<index_t>(coherence));
-            vector_type<double, 4> tmp;
-
-            tmp.AsType<double2_t>()(Number<0>{}) = bit_cast<double2_t>(f32_0);
-            tmp.AsType<double2_t>()(Number<1>{}) = bit_cast<double2_t>(f32_1);
-
-            return tmp.AsType<double4_t>()(Number<0>{});
-        }
+        return llvm_amdgcn_raw_buffer_load_i8(src_wave_buffer_resource,
+                                              src_thread_addr_offset,
+                                              src_wave_addr_offset,
+                                              static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, float>::value)
+    else if constexpr(N == 2)
     {
-        if constexpr(N == 1)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp32(src_wave_buffer_resource,
-                                                    src_thread_addr_offset,
-                                                    src_wave_addr_offset,
-                                                    static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp32x2(src_wave_buffer_resource,
-                                                      src_thread_addr_offset,
-                                                      src_wave_addr_offset,
-                                                      static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
+
+        int16_t tmp = llvm_amdgcn_raw_buffer_load_i16(src_wave_buffer_resource,
                                                       src_thread_addr_offset,
                                                       src_wave_addr_offset,
                                                       static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            vector_type<float, 8> tmp;
-
-            tmp.AsType<float4_t>()(Number<0>{}) =
-                llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
 
-            tmp.AsType<float4_t>()(Number<1>{}) =
-                llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset + 4 * sizeof(float),
-                                                   static_cast<index_t>(coherence));
-
-            return tmp.AsType<float8_t>()(Number<0>{});
-        }
+        return bit_cast<int8x2_t>(tmp);
     }
-    else if constexpr(is_same<T, half_t>::value)
+    else if constexpr(N == 4)
     {
-        if constexpr(N == 1)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp16(src_wave_buffer_resource,
-                                                    src_thread_addr_offset,
-                                                    src_wave_addr_offset,
-                                                    static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp16x2(src_wave_buffer_resource,
-                                                      src_thread_addr_offset,
-                                                      src_wave_addr_offset,
-                                                      static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            return llvm_amdgcn_raw_buffer_load_fp16x4(src_wave_buffer_resource,
+        int32_t tmp = llvm_amdgcn_raw_buffer_load_i32(src_wave_buffer_resource,
                                                       src_thread_addr_offset,
                                                       src_wave_addr_offset,
                                                       static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            // use fp32 load to mimic fp16 load
-            float4_t tmp = llvm_amdgcn_raw_buffer_load_fp32x4(src_wave_buffer_resource,
-                                                              src_thread_addr_offset,
-                                                              src_wave_addr_offset,
-                                                              static_cast<index_t>(coherence));
-
-            return bit_cast<half8_t>(tmp);
-        }
-    }
-    else if constexpr(is_same<T, bhalf_t>::value)
-    {
-        if constexpr(N == 1)
-        {
-            return llvm_amdgcn_raw_buffer_load_i16(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            return llvm_amdgcn_raw_buffer_load_i16x2(src_wave_buffer_resource,
-                                                     src_thread_addr_offset,
-                                                     src_wave_addr_offset,
-                                                     static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            return llvm_amdgcn_raw_buffer_load_i16x4(src_wave_buffer_resource,
-                                                     src_thread_addr_offset,
-                                                     src_wave_addr_offset,
-                                                     static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            int32x4_t tmp = llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
-                                                              src_thread_addr_offset,
-                                                              src_wave_addr_offset,
-                                                              static_cast<index_t>(coherence));
 
-            return bit_cast<bhalf8_t>(tmp);
-        }
+        return bit_cast<int8x4_t>(tmp);
     }
-    else if constexpr(is_same<T, int32_t>::value)
+    else if constexpr(N == 8)
     {
-        if constexpr(N == 1)
-        {
-            return llvm_amdgcn_raw_buffer_load_i32(src_wave_buffer_resource,
-                                                   src_thread_addr_offset,
-                                                   src_wave_addr_offset,
-                                                   static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            return llvm_amdgcn_raw_buffer_load_i32x2(src_wave_buffer_resource,
-                                                     src_thread_addr_offset,
-                                                     src_wave_addr_offset,
-                                                     static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            return llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
-                                                     src_thread_addr_offset,
-                                                     src_wave_addr_offset,
-                                                     static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            vector_type<int32_t, 8> tmp;
-
-            tmp.AsType<int32x4_t>()(Number<0>{}) =
-                llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
-                                                  src_thread_addr_offset,
-                                                  src_wave_addr_offset,
-                                                  static_cast<index_t>(coherence));
-
-            tmp.AsType<int32x4_t>()(Number<1>{}) =
-                llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
-                                                  src_thread_addr_offset,
-                                                  src_wave_addr_offset + 4 * sizeof(int32_t),
-                                                  static_cast<index_t>(coherence));
-            return tmp.AsType<int32x8_t>()(Number<0>{});
-        }
-    }
-    else if constexpr(is_same<T, int8_t>::value)
-    {
-        if constexpr(N == 1)
-        {
-            return llvm_amdgcn_raw_buffer_load_i8(src_wave_buffer_resource,
-                                                  src_thread_addr_offset,
-                                                  src_wave_addr_offset,
-                                                  static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            return llvm_amdgcn_raw_buffer_load_i8x2(src_wave_buffer_resource,
-                                                    src_thread_addr_offset,
-                                                    src_wave_addr_offset,
-                                                    static_cast<index_t>(coherence));
-#else
-            int16_t tmp = llvm_amdgcn_raw_buffer_load_i16(src_wave_buffer_resource,
+        int32x2_t tmp = llvm_amdgcn_raw_buffer_load_i32x2(src_wave_buffer_resource,
                                                           src_thread_addr_offset,
                                                           src_wave_addr_offset,
                                                           static_cast<index_t>(coherence));
 
-            return bit_cast<int8x2_t>(tmp);
-#endif
-        }
-        else if constexpr(N == 4)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            return llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                    src_thread_addr_offset,
-                                                    src_wave_addr_offset,
-                                                    static_cast<index_t>(coherence));
-#else
-            int32_t tmp = llvm_amdgcn_raw_buffer_load_i32(src_wave_buffer_resource,
+        return bit_cast<int8x8_t>(tmp);
+    }
+    else if constexpr(N == 16)
+    {
+        int32x4_t tmp = llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
                                                           src_thread_addr_offset,
                                                           src_wave_addr_offset,
                                                           static_cast<index_t>(coherence));
+        return bit_cast<int8x16_t>(tmp);
+    }
+    else if constexpr(N == 32)
+    {
+        int32x4_t tmp0 = llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                                           src_thread_addr_offset,
+                                                           src_wave_addr_offset,
+                                                           static_cast<index_t>(coherence));
+        int32x4_t tmp1 =
+            llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                              src_thread_addr_offset,
+                                              src_wave_addr_offset + 4 * sizeof(int32_t),
+                                              static_cast<index_t>(coherence));
+        vector_type<int32_t, 8> tmp;
 
-            return bit_cast<int8x4_t>(tmp);
-#endif
-        }
-        else if constexpr(N == 8)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            vector_type<int8_t, 8> tmp;
-
-            tmp.AsType<int8x4_t>()(Number<0>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset,
-                                                 static_cast<index_t>(coherence));
-
-            tmp.AsType<int8x4_t>()(Number<1>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset + 4 * sizeof(int8_t),
-                                                 static_cast<index_t>(coherence));
-
-            return tmp.AsType<int8x8_t>()(Number<0>{});
-#else
-            int32x2_t tmp = llvm_amdgcn_raw_buffer_load_i32x2(src_wave_buffer_resource,
-                                                              src_thread_addr_offset,
-                                                              src_wave_addr_offset,
-                                                              static_cast<index_t>(coherence));
+        tmp.AsType<int32x4_t>()(Number<0>{}) = tmp0;
+        tmp.AsType<int32x4_t>()(Number<1>{}) = tmp1;
 
-            return bit_cast<int8x8_t>(tmp);
-#endif
-        }
-        else if constexpr(N == 16)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            vector_type<int8_t, 16> tmp;
-
-            tmp.AsType<int8x4_t>()(Number<0>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset,
-                                                 static_cast<index_t>(coherence));
-
-            tmp.AsType<int8x4_t>()(Number<1>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset + 4 * sizeof(int8_t),
-                                                 static_cast<index_t>(coherence));
-
-            tmp.AsType<int8x4_t>()(Number<2>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset + 8 * sizeof(int8_t),
-                                                 static_cast<index_t>(coherence));
-
-            tmp.AsType<int8x4_t>()(Number<3>{}) =
-                llvm_amdgcn_raw_buffer_load_i8x4(src_wave_buffer_resource,
-                                                 src_thread_addr_offset,
-                                                 src_wave_addr_offset + 12 * sizeof(int8_t),
-                                                 static_cast<index_t>(coherence));
-
-            return tmp.AsType<int8x16_t>()(Number<0>{});
-#else
-            int32x4_t tmp = llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
-                                                              src_thread_addr_offset,
-                                                              src_wave_addr_offset,
-                                                              static_cast<index_t>(coherence));
+        return bit_cast<int8x32_t>(tmp);
+    }
+    else if constexpr(N == 64)
+    {
+        int32x4_t tmp0 = llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                                           src_thread_addr_offset,
+                                                           src_wave_addr_offset,
+                                                           static_cast<index_t>(coherence));
+        int32x4_t tmp1 =
+            llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                              src_thread_addr_offset,
+                                              src_wave_addr_offset + 4 * sizeof(int32_t),
+                                              static_cast<index_t>(coherence));
+        int32x4_t tmp2 =
+            llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                              src_thread_addr_offset,
+                                              src_wave_addr_offset + 8 * sizeof(int32_t),
+                                              static_cast<index_t>(coherence));
+        int32x4_t tmp3 =
+            llvm_amdgcn_raw_buffer_load_i32x4(src_wave_buffer_resource,
+                                              src_thread_addr_offset,
+                                              src_wave_addr_offset + 12 * sizeof(int32_t),
+                                              static_cast<index_t>(coherence));
 
-            return bit_cast<int8x16_t>(tmp);
-#endif
-        }
+        vector_type<int32_t, 16> tmp;
+
+        tmp.AsType<int32x4_t>()(Number<0>{}) = tmp0;
+        tmp.AsType<int32x4_t>()(Number<1>{}) = tmp1;
+        tmp.AsType<int32x4_t>()(Number<2>{}) = tmp2;
+        tmp.AsType<int32x4_t>()(Number<3>{}) = tmp3;
+
+        return bit_cast<int8x64_t>(tmp);
     }
 }
 
 template <typename T,
           index_t N,
           AmdBufferCoherenceEnum coherence = AmdBufferCoherenceEnum::DefaultCoherence>
-__device__ void amd_buffer_store_impl(const typename vector_type<T, N>::type src_thread_data,
-                                      int32x4_t dst_wave_buffer_resource,
-                                      index_t dst_thread_addr_offset,
-                                      index_t dst_wave_addr_offset)
+__device__ typename vector_type<T, N>::type amd_buffer_load_impl(int32x4_t src_wave_buffer_resource,
+                                                                 index_t src_thread_addr_offset,
+                                                                 index_t src_wave_addr_offset)
 {
     static_assert(
-        (is_same<T, double>::value && (N == 1 || N == 2)) ||
-            (is_same<T, float>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, half_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, bhalf_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
-            (is_same<T, int32_t>::value && (N == 1 || N == 2 || N == 4)) ||
+        (is_same<T, double>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
+            (is_same<T, float>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, half_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, bhalf_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, int32_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, f8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, bf8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
             (is_same<T, int8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)),
         "wrong! not implemented");
 
-    if constexpr(is_same<T, double>::value)
+    using r_t     = typename vector_type<T, N>::type;
+    auto raw_data = amd_buffer_load_impl_raw<sizeof(T) * N, coherence>(
+        src_wave_buffer_resource, src_thread_addr_offset, src_wave_addr_offset);
+    return bit_cast<r_t>(raw_data);
+}
+
+template <index_t N, AmdBufferCoherenceEnum coherence = AmdBufferCoherenceEnum::DefaultCoherence>
+__device__ void
+amd_buffer_store_impl_raw(const typename vector_type<int8_t, N>::type src_thread_data,
+                          int32x4_t dst_wave_buffer_resource,
+                          index_t dst_thread_addr_offset,
+                          index_t dst_wave_addr_offset)
+{
+    static_assert(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32 || N == 64,
+                  "wrong! not implemented");
+
+    if constexpr(N == 1)
     {
-        // use fp32 store to mimic fp64 store
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_fp32x2(bit_cast<float2_t>(src_thread_data),
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            llvm_amdgcn_raw_buffer_store_fp32x4(bit_cast<float4_t>(src_thread_data),
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
+        llvm_amdgcn_raw_buffer_store_i8(src_thread_data,
+                                        dst_wave_buffer_resource,
+                                        dst_thread_addr_offset,
+                                        dst_wave_addr_offset,
+                                        static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, float>::value)
+    else if constexpr(N == 2)
     {
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_fp32(src_thread_data,
-                                              dst_wave_buffer_resource,
-                                              dst_thread_addr_offset,
-                                              dst_wave_addr_offset,
-                                              static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            llvm_amdgcn_raw_buffer_store_fp32x2(src_thread_data,
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            llvm_amdgcn_raw_buffer_store_fp32x4(src_thread_data,
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            vector_type<float, 8> tmp{src_thread_data};
-            llvm_amdgcn_raw_buffer_store_fp32x4(tmp.AsType<float4_t>()[Number<0>{}],
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-            llvm_amdgcn_raw_buffer_store_fp32x4(tmp.AsType<float4_t>()[Number<1>{}],
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset + 4 * sizeof(float),
-                                                static_cast<index_t>(coherence));
-        }
+
+        llvm_amdgcn_raw_buffer_store_i16(bit_cast<int16_t>(src_thread_data),
+                                         dst_wave_buffer_resource,
+                                         dst_thread_addr_offset,
+                                         dst_wave_addr_offset,
+                                         static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, half_t>::value)
+    else if constexpr(N == 4)
     {
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_fp16(src_thread_data,
-                                              dst_wave_buffer_resource,
-                                              dst_thread_addr_offset,
-                                              dst_wave_addr_offset,
-                                              static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            llvm_amdgcn_raw_buffer_store_fp16x2(src_thread_data,
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            llvm_amdgcn_raw_buffer_store_fp16x4(src_thread_data,
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-#if 0
-            vector_type<half_t, 8> tmp{src_thread_data};
-
-            llvm_amdgcn_raw_buffer_store_fp16x4(tmp.AsType<half4_t>()[Number<0>{}],
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-
-            llvm_amdgcn_raw_buffer_store_fp16x4(tmp.AsType<half4_t>()[Number<1>{}],
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset + 4 * sizeof(half_t),
-                                                static_cast<index_t>(coherence));
-#else
-            llvm_amdgcn_raw_buffer_store_fp32x4(bit_cast<float4_t>(src_thread_data),
-                                                dst_wave_buffer_resource,
-                                                dst_thread_addr_offset,
-                                                dst_wave_addr_offset,
-                                                static_cast<index_t>(coherence));
-#endif
-        }
+        llvm_amdgcn_raw_buffer_store_i32(bit_cast<int32_t>(src_thread_data),
+                                         dst_wave_buffer_resource,
+                                         dst_thread_addr_offset,
+                                         dst_wave_addr_offset,
+                                         static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, bhalf_t>::value)
+    else if constexpr(N == 8)
     {
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_i16(src_thread_data,
-                                             dst_wave_buffer_resource,
-                                             dst_thread_addr_offset,
-                                             dst_wave_addr_offset,
-                                             static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            llvm_amdgcn_raw_buffer_store_i16x2(src_thread_data,
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            llvm_amdgcn_raw_buffer_store_i16x4(src_thread_data,
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 8)
-        {
-            vector_type<bhalf_t, 8> tmp{src_thread_data};
-
-            llvm_amdgcn_raw_buffer_store_i16x4(tmp.AsType<bhalf4_t>()[Number<0>{}],
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-
-            llvm_amdgcn_raw_buffer_store_i16x4(tmp.AsType<bhalf4_t>()[Number<1>{}],
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset + 4 * sizeof(bhalf_t),
-                                               static_cast<index_t>(coherence));
-        }
+        llvm_amdgcn_raw_buffer_store_i32x2(bit_cast<int32x2_t>(src_thread_data),
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset,
+                                           static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, int32_t>::value)
+    else if constexpr(N == 16)
     {
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_i32(src_thread_data,
-                                             dst_wave_buffer_resource,
-                                             dst_thread_addr_offset,
-                                             dst_wave_addr_offset,
-                                             static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-            llvm_amdgcn_raw_buffer_store_i32x2(src_thread_data,
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 4)
-        {
-            llvm_amdgcn_raw_buffer_store_i32x4(src_thread_data,
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
+        llvm_amdgcn_raw_buffer_store_i32x4(bit_cast<int32x4_t>(src_thread_data),
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset,
+                                           static_cast<index_t>(coherence));
     }
-    else if constexpr(is_same<T, int8_t>::value)
+    else if constexpr(N == 32)
     {
-        if constexpr(N == 1)
-        {
-            llvm_amdgcn_raw_buffer_store_i8(src_thread_data,
-                                            dst_wave_buffer_resource,
-                                            dst_thread_addr_offset,
-                                            dst_wave_addr_offset,
-                                            static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 2)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            llvm_amdgcn_raw_buffer_store_i8x2(src_thread_data,
-                                              dst_wave_buffer_resource,
-                                              dst_thread_addr_offset,
-                                              dst_wave_addr_offset,
-                                              static_cast<index_t>(coherence));
-#else
-            llvm_amdgcn_raw_buffer_store_i16(bit_cast<int16_t>(src_thread_data),
-                                             dst_wave_buffer_resource,
-                                             dst_thread_addr_offset,
-                                             dst_wave_addr_offset,
-                                             static_cast<index_t>(coherence));
-#endif
-        }
-        else if constexpr(N == 4)
-        {
-#if !CK_WORKAROUND_SWDEV_XXXXXX_INT8_BUFFER_LOAD_STORE_ISSUE
-            llvm_amdgcn_raw_buffer_store_i8x4(src_thread_data,
-                                              dst_wave_buffer_resource,
-                                              dst_thread_addr_offset,
-                                              dst_wave_addr_offset,
-                                              static_cast<index_t>(coherence));
-#else
-            llvm_amdgcn_raw_buffer_store_i32(bit_cast<int32_t>(src_thread_data),
-                                             dst_wave_buffer_resource,
-                                             dst_thread_addr_offset,
-                                             dst_wave_addr_offset,
-                                             static_cast<index_t>(coherence));
-#endif
-        }
-        else if constexpr(N == 8)
-        {
-            llvm_amdgcn_raw_buffer_store_i32x2(bit_cast<int32x2_t>(src_thread_data),
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
-        else if constexpr(N == 16)
-        {
-            llvm_amdgcn_raw_buffer_store_i32x4(bit_cast<int32x4_t>(src_thread_data),
-                                               dst_wave_buffer_resource,
-                                               dst_thread_addr_offset,
-                                               dst_wave_addr_offset,
-                                               static_cast<index_t>(coherence));
-        }
+        vector_type<int32_t, 8> tmp{bit_cast<int32x8_t>(src_thread_data)};
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<0>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset,
+                                           static_cast<index_t>(coherence));
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<1>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset + sizeof(int32_t) * 4,
+                                           static_cast<index_t>(coherence));
     }
+    else if constexpr(N == 64)
+    {
+        vector_type<int32_t, 16> tmp{bit_cast<int32x16_t>(src_thread_data)};
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<0>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset,
+                                           static_cast<index_t>(coherence));
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<1>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset + sizeof(int32_t) * 4,
+                                           static_cast<index_t>(coherence));
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<2>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset + sizeof(int32_t) * 8,
+                                           static_cast<index_t>(coherence));
+
+        llvm_amdgcn_raw_buffer_store_i32x4(tmp.template AsType<int32x4_t>()[Number<3>{}],
+                                           dst_wave_buffer_resource,
+                                           dst_thread_addr_offset,
+                                           dst_wave_addr_offset + sizeof(int32_t) * 12,
+                                           static_cast<index_t>(coherence));
+    }
+}
+
+template <typename T,
+          index_t N,
+          AmdBufferCoherenceEnum coherence = AmdBufferCoherenceEnum::DefaultCoherence>
+__device__ void amd_buffer_store_impl(const typename vector_type<T, N>::type src_thread_data,
+                                      int32x4_t dst_wave_buffer_resource,
+                                      index_t dst_thread_addr_offset,
+                                      index_t dst_wave_addr_offset)
+{
+    static_assert(
+        (is_same<T, double>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
+            (is_same<T, float>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, half_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, bhalf_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, int32_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, f8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, bf8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
+            (is_same<T, int8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)),
+        "wrong! not implemented");
+
+    using r_t = typename vector_type<int8_t, sizeof(T) * N>::type;
+
+    amd_buffer_store_impl_raw<sizeof(T) * N, coherence>(bit_cast<r_t>(src_thread_data),
+                                                        dst_wave_buffer_resource,
+                                                        dst_thread_addr_offset,
+                                                        dst_wave_addr_offset);
 }
 
 template <typename T, index_t N>
@@ -1127,54 +798,14 @@ amd_buffer_load_invalid_element_return_zero(const T* p_src_wave,
 
 #if CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK
     uint32_t src_addr_shift = src_thread_element_valid ? 0 : 0x80000000;
-#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-    if constexpr(is_same<scalar_t, f8_t>::value || is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 && !defined CK_ENABLE_BF8
-        if constexpr(is_same<scalar_t, f8_t>::value)
-#endif
-#if !defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-            if constexpr(is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            {
-                auto tmp = amd_buffer_load_impl<int8_t, vector_size, coherence>(
-                    src_wave_buffer_resource, src_addr_shift + src_thread_addr_offset, 0);
-                return bit_cast<vector_t>(tmp);
-            }
-            else
-            {
-#endif
-                return amd_buffer_load_impl<scalar_t, vector_size, coherence>(
-                    src_wave_buffer_resource, src_addr_shift + src_thread_addr_offset, 0);
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            }
-#endif
+    return amd_buffer_load_impl<scalar_t, vector_size, coherence>(
+        src_wave_buffer_resource, src_addr_shift + src_thread_addr_offset, 0);
+
 #else
-#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-    if constexpr(is_same<scalar_t, f8_t>::value || is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 && !defined CK_ENABLE_BF8
-        if constexpr(is_same<scalar_t, f8_t>::value)
-#endif
-#if !defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-            if constexpr(is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            {
-                auto tmp = amd_buffer_load_impl<int8_t, vector_size, coherence>(
-                    src_wave_buffer_resource, src_thread_addr_offset, 0);
-                return src_thread_element_valid ? bit_cast<vector_t>(tmp) : vector_t(0);
-            }
-            else
-            {
-#endif
-                vector_t tmp = amd_buffer_load_impl<scalar_t, vector_size, coherence>(
-                    src_wave_buffer_resource, src_thread_addr_offset, 0);
-                return src_thread_element_valid ? tmp : vector_t(0);
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            }
-#endif
+
+    vector_t tmp = amd_buffer_load_impl<scalar_t, vector_size, coherence>(
+        src_wave_buffer_resource, src_thread_addr_offset, 0);
+    return src_thread_element_valid ? tmp : vector_t(0);
 #endif
 }
 
@@ -1232,62 +863,13 @@ __device__ void amd_buffer_store(const typename vector_type_maker<T, N>::type::t
 
 #if CK_EXPERIMENTAL_USE_BUFFER_STORE_OOB_CHECK_OFFSET_TRICK
     uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x80000000;
-
-#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-    if constexpr(is_same<scalar_t, f8_t>::value || is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 && !defined CK_ENABLE_BF8
-        if constexpr(is_same<scalar_t, f8_t>::value)
-#endif
-#if !defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-            if constexpr(is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            {
-                auto tmp = bit_cast<typename vector_type_maker<int8_t, vector_size>::type::type>(
-                    src_thread_data);
-                amd_buffer_store_impl<int8_t, vector_size, coherence>(
-                    tmp, dst_wave_buffer_resource, dst_addr_shift + dst_thread_addr_offset, 0);
-            }
-            else
-            {
-#endif
-                amd_buffer_store_impl<scalar_t, vector_size, coherence>(src_thread_data,
-                                                                        dst_wave_buffer_resource,
-                                                                        dst_addr_shift +
-                                                                            dst_thread_addr_offset,
-                                                                        0);
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-            }
-#endif
+    amd_buffer_store_impl<scalar_t, vector_size, coherence>(
+        src_thread_data, dst_wave_buffer_resource, dst_addr_shift + dst_thread_addr_offset, 0);
 #else
     if(dst_thread_element_valid)
     {
-#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-        if constexpr(is_same<scalar_t, f8_t>::value || is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 && !defined CK_ENABLE_BF8
-            if constexpr(is_same<scalar_t, f8_t>::value)
-#endif
-#if !defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-                if constexpr(is_same<scalar_t, bf8_t>::value)
-#endif
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-                {
-                    auto tmp =
-                        bit_cast<typename vector_type_maker<int8_t, vector_size>::type::type>(
-                            src_thread_data);
-                    amd_buffer_store_impl<int8_t, vector_size, coherence>(
-                        tmp, dst_wave_buffer_resource, dst_thread_addr_offset, 0);
-                }
-                else
-                {
-#endif
-                    amd_buffer_store_impl<scalar_t, vector_size, coherence>(
-                        src_thread_data, dst_wave_buffer_resource, dst_thread_addr_offset, 0);
-#if defined CK_ENABLE_FP8 || defined CK_ENABLE_BF8
-                }
-#endif
+        amd_buffer_store_impl<scalar_t, vector_size, coherence>(
+            src_thread_data, dst_wave_buffer_resource, dst_thread_addr_offset, 0);
     }
 #endif
 }

include/ck/utility/is_detected.hpp

@@ -31,4 +31,13 @@ struct nonesuch
 template <template <class...> class Op, class... Args>
 using is_detected = typename detail::detector<nonesuch, void, Op, Args...>::value_t;
 
+template <typename T>
+using is_pack2_invocable_t = decltype(std::declval<T&>().is_pack2_invocable);
+
+template <typename T>
+using is_pack4_invocable_t = decltype(std::declval<T&>().is_pack4_invocable);
+
+template <typename T>
+using is_pack8_invocable_t = decltype(std::declval<T&>().is_pack8_invocable);
+
 } // namespace ck

include/ck/utility/sequence.hpp

@@ -897,14 +897,3 @@ template <index_t NSize, index_t I>
 using uniform_sequence_gen_t = typename uniform_sequence_gen<NSize, I>::type;
 
 } // namespace ck
-
-template <ck::index_t... Is>
-std::ostream& operator<<(std::ostream& os, const ck::Sequence<Is...>)
-{
-    using S = ck::Sequence<Is...>;
-    os << "{";
-    ck::static_for<0, S::Size() - ck::Number<1>{}, 1>{}(
-        [&](auto i) { os << S::At(i).value << ", "; });
-    os << S::At(S::Size() - ck::Number<1>{}).value << "}";
-    return os;
-}

library/include/ck/library/reference_tensor_operation/cpu/reference_conv_fwd.hpp

@@ -14,27 +14,27 @@ namespace ck {
 namespace tensor_operation {
 namespace host {
 
-///
-/// @brief      Reference implementation for forward convolution.
-///
-/// @paragraph
-///             Tensor descriptor in GNCHW/GKCXY/GNKHW dimensional order
-///             Supports both GNCHW/NGCHW as well as GNHWC/NHWGC physical layout
-///             as long as dimensions in tensor descriptor is in GNCHW order
-///
-/// @tparam     InDataType               Input tensor data type.
-/// @tparam     WeiDataType              Weights tensor data type.
-/// @tparam     OutDataType              Output tensor data type.
-/// @tparam     InElementwiseOperation   Functor for input tensor elementwise
-///                                      operation.
-/// @tparam     WeiElementwiseOperation  Functor for weights tensor elementwise
-///                                      operation.
-/// @tparam     NDimSpatial  Number of spatial dimensions.
-///
-/// input descriptor in [G, N, C, Do, Ho, Wo] order
-/// weight descriptor in [G, K, C, Z, Y, X] order
-/// output descriptor in [G, N, K, Di, Hi, Wi] order
-/// phyiscal layout is irrelavent
+//
+// @brief      Reference implementation for forward convolution.
+//
+// @paragraph
+//             Tensor descriptor in GNCHW/GKCXY/GNKHW dimensional order
+//             Supports both GNCHW/NGCHW as well as GNHWC/NHWGC physical layout
+//             as long as dimensions in tensor descriptor is in GNCHW order
+//
+// @tparam     InDataType               Input tensor data type.
+// @tparam     WeiDataType              Weights tensor data type.
+// @tparam     OutDataType              Output tensor data type.
+// @tparam     InElementwiseOperation   Functor for input tensor elementwise
+//                                      operation.
+// @tparam     WeiElementwiseOperation  Functor for weights tensor elementwise
+//                                      operation.
+// @tparam     NDimSpatial  Number of spatial dimensions.
+//
+// input descriptor in [G, N, C, Do, Ho, Wo] order
+// weight descriptor in [G, K, C, Z, Y, X] order
+// output descriptor in [G, N, K, Di, Hi, Wi] order
+// phyiscal layout is irrelavent
 template <ck::index_t NDimSpatial,
           typename InDataType,
           typename WeiDataType,

library/include/ck/library/tensor_operation_instance/gpu/gemm_splitk.hpp

@@ -98,6 +98,26 @@ void add_device_gemm_xdl_splitk_f16_f8_f16_mk_nk_mn_instances(
     std::vector<std::unique_ptr<
         DeviceGemmSplitK<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
         instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
 #endif
 
 template <typename ADataType,
@@ -105,7 +125,8 @@ template <typename ADataType,
           typename CDataType,
           typename ALayout,
           typename BLayout,
-          typename CLayout>
+          typename CLayout,
+          typename ComputeType>
 struct DeviceOperationInstanceFactory<
     ck::tensor_operation::device::DeviceGemmSplitK<ALayout,
                                                    BLayout,
@@ -115,7 +136,8 @@ struct DeviceOperationInstanceFactory<
                                                    CDataType,
                                                    ck::tensor_operation::element_wise::PassThrough,
                                                    ck::tensor_operation::element_wise::PassThrough,
-                                                   ck::tensor_operation::element_wise::PassThrough>>
+                                                   ck::tensor_operation::element_wise::PassThrough,
+                                                   ComputeType>>
 {
     using DeviceOp = DeviceGemmSplitK<ALayout,
                                       BLayout,
@@ -125,7 +147,8 @@ struct DeviceOperationInstanceFactory<
                                       CDataType,
                                       ck::tensor_operation::element_wise::PassThrough,
                                       ck::tensor_operation::element_wise::PassThrough,
-                                      ck::tensor_operation::element_wise::PassThrough>;
+                                      ck::tensor_operation::element_wise::PassThrough,
+                                      ComputeType>;
 
     static auto GetInstances()
     {
@@ -158,7 +181,7 @@ struct DeviceOperationInstanceFactory<
 #endif
 #ifdef CK_ENABLE_FP16
         else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
-                          is_same_v<CDataType, half_t>)
+                          is_same_v<CDataType, half_t> && is_same_v<ComputeType, half_t>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
@@ -231,6 +254,31 @@ struct DeviceOperationInstanceFactory<
                 add_device_gemm_xdl_splitk_f16_f8_f16_km_nk_mn_instances(op_ptrs);
             }
         }
+        else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
+                          is_same_v<CDataType, half_t> && is_same_v<ComputeType, f8_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(op_ptrs);
+            }
+        }
+
 #endif
         return op_ptrs;
     }

library/src/tensor_operation_instance/gpu/gemm_splitk/CMakeLists.txt

@@ -15,6 +15,10 @@ list(APPEND GEMM_SPLITK_INSTANCES device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_in
                                   device_gemm_xdl_splitk_f16_fp8_f16_mk_kn_mn_instance.cpp
                                   device_gemm_xdl_splitk_f16_fp8_f16_mk_nk_mn_instance.cpp
                                   device_gemm_xdl_splitk_f16_fp8_f16_km_kn_mn_instance.cpp
-                                  device_gemm_xdl_splitk_f16_fp8_f16_km_nk_mn_instance.cpp)
+                                  device_gemm_xdl_splitk_f16_fp8_f16_km_nk_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_kn_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_nk_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_kn_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_nk_mn_instance.cpp)
 
 add_instance_library(device_gemm_splitk_instance ${GEMM_SPLITK_INSTANCES})

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              1,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              1,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+// static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
+
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_generic_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,    64,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              1,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               2,    F8>
+    // clang-format on
+    >;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,    64,   192,     4,  8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   192,    64,     4,  8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,   192,     4,  8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   192,    32,     4,  8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,    64,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    64,    32,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances, device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_generic_instances{});
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_generic_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              1,              8,      true,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              1,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               2,    F8>
+    // clang-format on
+    >;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|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|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  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|    Type|
+        //#########################|     |      |      |        |        |        |        |   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|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,    64,    64,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    MNKPadding,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances, device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_generic_instances{});
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -33,26 +33,26 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 // a[m, k] * b[k, n] = e[m, n]
 using device_grouped_gemm_xdl_f16_f16_f16_mk_kn_mn_instances = std::tuple<
     // clang-format off
-    //###################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
-    //###################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
-    //###################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
-    //###################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>
+        //###################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //###################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //###################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>
     // clang-format on
     >;
 
@@ -60,20 +60,20 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 
 using device_grouped_gemm_xdl_f16_f16_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
     // clang-format off
-    //###################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
-    //###################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
-    //###################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
-    //###################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
-    DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
+        //###################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //###################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //###################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   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,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -33,41 +33,27 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 // a[m, k] * b[k, n] = e[m, n]
 using device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instances = std::tuple<
     // clang-format off
-    //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|     PipelineVersion|
-    //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|                    |
-    //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                    |
-    //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                    |
-    // Pipline v1
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-
-    // Pipline v2
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        // Currently AK1 must equal BK1 !
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 16,16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instance.cpp

@@ -31,41 +31,57 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 
 using device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
     // clang-format off
-    //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|     PipelineVersion|
-    //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|                    |
-    //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                    |
-    //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                    |
-    // Pipline v1
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-    
-    // Pipline v2
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        // Currently AK1 must equal BK1 !
+     // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 16,16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+     // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+     // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+     // DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
+
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -32,31 +32,23 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 // a[m, k] * b[n, k] = e[m, n]
 using device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_instances = std::tuple<
     // clang-format off
-    //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|         PipelineVersion|
-    //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|                        |
-    //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                        |
-    //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                        |
-    // Pipeline v1
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,    64,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    32,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, ck::PipelineVersion::v1>,
-    // Pipeline v2
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,    64,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    32,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, ck::PipelineVersion::v2>
-
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_irregular_instance.cpp

@@ -31,31 +31,30 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 
 using device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
     // clang-format off
-    //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|         PipelineVersion|
-    //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|                        |
-    //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                        |
-    //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                        |
-    // Pipeline v1
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v1>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v1>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, ck::PipelineVersion::v1>,
-    
-    // Pipeline v2
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v2>,
-  //DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    64,   8,   8,   32,   32,    1,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, ck::PipelineVersion::v2>,
-    DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, ck::PipelineVersion::v2>
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| 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|
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  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_MWaveMPerXdl| ScalarPerVector|
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              4>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,    
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    32,   8,   8,   32,   32,    1,    4,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>
     // clang-format on
     >;
 

profiler/include/profiler/profile_gemm_splitk_impl.hpp

@@ -30,7 +30,8 @@ template <typename ADataType,
           typename CDataType,
           typename ALayout,
           typename BLayout,
-          typename CLayout>
+          typename CLayout,
+          typename ComputeType = CDataType>
 bool profile_gemm_splitk_impl(int do_verification,
                               int init_method,
                               bool do_log,
@@ -103,7 +104,8 @@ bool profile_gemm_splitk_impl(int do_verification,
                                                                     CDataType,
                                                                     AElementOp,
                                                                     BElementOp,
-                                                                    CElementOp>;
+                                                                    CElementOp,
+                                                                    ComputeType>;
 
     // get device op instances
     const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
@@ -120,7 +122,8 @@ bool profile_gemm_splitk_impl(int do_verification,
                                                                                 AccDataType,
                                                                                 AElementOp,
                                                                                 BElementOp,
-                                                                                CElementOp>;
+                                                                                CElementOp,
+                                                                                ComputeType>;
 
         auto ref_gemm    = ReferenceGemmInstance{};
         auto ref_invoker = ref_gemm.MakeInvoker();

profiler/src/profile_gemm_splitk.cpp

@@ -25,6 +25,7 @@ enum struct GemmDataType
     INT8_INT8_INT8, // 3
     F8_F16_F16,     // 4
     F16_F8_F16,     // 5
+    F16_F16_F16_F8, // 6
 };
 
 #define OP_NAME "gemm_splitk"
@@ -35,7 +36,8 @@ int profile_gemm_splitk(int argc, char* argv[])
     if(argc != 15)
     {
         printf("arg1: tensor operation (" OP_NAME ": " OP_DESC ")\n");
-        printf("arg2: data type (0: fp32; 1: fp16; 2: bf16; 3: int8; 4: f8@f16; 5: f16@f8)\n");
+        printf("arg2: data type (0: fp32; 1: fp16; 2: bf16; 3: int8; 4: f8@f16; 5: f16@f8; 6: f16, "
+               "comp f8)\n");
         printf("arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n");
         printf("                     1: A[m, k] * B[n, k] = C[m, n];\n");
         printf("                     2: A[k, m] * B[k, n] = C[m, n];\n");
@@ -80,7 +82,8 @@ int profile_gemm_splitk(int argc, char* argv[])
                        auto c_type,
                        auto a_layout,
                        auto b_layout,
-                       auto c_layout) {
+                       auto c_layout,
+                       auto compute_type) {
         using ADataType   = decltype(a_type);
         using BDataType   = decltype(b_type);
         using AccDataType = decltype(acc_type);
@@ -90,6 +93,8 @@ int profile_gemm_splitk(int argc, char* argv[])
         using BLayout = decltype(b_layout);
         using CLayout = decltype(c_layout);
 
+        using ComputeType = decltype(compute_type);
+
         const int DefaultStrideA = ck::is_same_v<ALayout, Row> ? K : M;
         const int DefaultStrideB = ck::is_same_v<BLayout, Row> ? N : K;
         const int DefaultStrideC = ck::is_same_v<CLayout, Row> ? N : M;
@@ -100,7 +105,8 @@ int profile_gemm_splitk(int argc, char* argv[])
                                                            CDataType,
                                                            ALayout,
                                                            BLayout,
-                                                           CLayout>(
+                                                           CLayout,
+                                                           ComputeType>(
             do_verification,
             init_method,
             do_log,
@@ -118,68 +124,84 @@ int profile_gemm_splitk(int argc, char* argv[])
 
     if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::MK_KN_MN)
     {
-        return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Row{}, Row{});
+        return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Row{}, Row{}, F32{});
     }
     else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::MK_NK_MN)
     {
-        return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Col{}, Row{});
+        return profile(F32{}, F32{}, F32{}, F32{}, Row{}, Col{}, Row{}, F32{});
     }
     else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::KM_KN_MN)
     {
-        return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Row{}, Row{});
+        return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Row{}, Row{}, F32{});
     }
     else if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::KM_NK_MN)
     {
-        return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Col{}, Row{});
+        return profile(F32{}, F32{}, F32{}, F32{}, Col{}, Col{}, Row{}, F32{});
     }
     else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::MK_KN_MN)
     {
-        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{});
+        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::MK_NK_MN)
     {
-        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{});
+        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::KM_KN_MN)
     {
-        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{});
+        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::KM_NK_MN)
     {
-        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{});
+        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{}, F16{});
     }
 #if defined CK_ENABLE_FP8
     else if(data_type == GemmDataType::F8_F16_F16 && layout == GemmMatrixLayout::MK_KN_MN)
     {
-        return profile(F8{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{});
+        return profile(F8{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F8_F16_F16 && layout == GemmMatrixLayout::MK_NK_MN)
     {
-        return profile(F8{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{});
+        return profile(F8{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F8_F16_F16 && layout == GemmMatrixLayout::KM_KN_MN)
     {
-        return profile(F8{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{});
+        return profile(F8{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F8_F16_F16 && layout == GemmMatrixLayout::KM_NK_MN)
     {
-        return profile(F8{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{});
+        return profile(F8{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F8_F16 && layout == GemmMatrixLayout::MK_KN_MN)
     {
-        return profile(F16{}, F8{}, F32{}, F16{}, Row{}, Row{}, Row{});
+        return profile(F16{}, F8{}, F32{}, F16{}, Row{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F8_F16 && layout == GemmMatrixLayout::MK_NK_MN)
     {
-        return profile(F16{}, F8{}, F32{}, F16{}, Row{}, Col{}, Row{});
+        return profile(F16{}, F8{}, F32{}, F16{}, Row{}, Col{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F8_F16 && layout == GemmMatrixLayout::KM_KN_MN)
     {
-        return profile(F16{}, F8{}, F32{}, F16{}, Col{}, Row{}, Row{});
+        return profile(F16{}, F8{}, F32{}, F16{}, Col{}, Row{}, Row{}, F16{});
     }
     else if(data_type == GemmDataType::F16_F8_F16 && layout == GemmMatrixLayout::KM_NK_MN)
     {
-        return profile(F16{}, F8{}, F32{}, F16{}, Col{}, Col{}, Row{});
+        return profile(F16{}, F8{}, F32{}, F16{}, Col{}, Col{}, Row{}, F16{});
+    }
+    else if(data_type == GemmDataType::F16_F16_F16_F8 && layout == GemmMatrixLayout::MK_KN_MN)
+    {
+        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Row{}, Row{}, F8{});
+    }
+    else if(data_type == GemmDataType::F16_F16_F16_F8 && layout == GemmMatrixLayout::MK_NK_MN)
+    {
+        return profile(F16{}, F16{}, F32{}, F16{}, Row{}, Col{}, Row{}, F8{});
+    }
+    else if(data_type == GemmDataType::F16_F16_F16_F8 && layout == GemmMatrixLayout::KM_KN_MN)
+    {
+        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Row{}, Row{}, F8{});
+    }
+    else if(data_type == GemmDataType::F16_F16_F16_F8 && layout == GemmMatrixLayout::KM_NK_MN)
+    {
+        return profile(F16{}, F16{}, F32{}, F16{}, Col{}, Col{}, Row{}, F8{});
     }
 #endif
     else

script/cmake-ck-dev.sh

@@ -12,9 +12,7 @@ cmake
 -save-temps=$PWD"                                                                                 \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=ON                                                                                   \
--D GPU_TARGETS="gfx90a"                                                             \
+-D GPU_TARGETS="gfx908;gfx90a;gfx940"                                                             \
 -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON                                                                 \
 -D USE_BITINT_EXTENSION_INT4=OFF                                                                  \
 ${MY_PROJECT_SOURCE}
-
-#-D GPU_TARGETS="gfx908;gfx90a;gfx940"                                                             \

test/grouped_gemm/test_grouped_gemm_ut_cases.inc

@@ -4,7 +4,7 @@ TEST_P(RRR_F16_F16_F16, TinyCases)
 {
     const std::vector<int> Ms{0, 1};
     constexpr int N = 768;
-    constexpr int K = 1088;
+    constexpr int K = 544;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -17,9 +17,9 @@ TEST_P(RRR_F16_F16_F16, TinyCases)
 
 TEST_P(RRR_F16_F16_F16, SmallCases)
 {
-    const std::vector<int> Ms{2, 3, 4, 5};
+    const std::vector<int> Ms{2, 1, 3, 4, 5, 0};
     constexpr int N = 768;
-    constexpr int K = 1088;
+    constexpr int K = 544;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -34,7 +34,7 @@ TEST_P(RRR_F16_F16_F16, MidCases)
 {
     const std::vector<int> Ms{167, 183, 177, 153, 139, 204};
     constexpr int N = 768;
-    constexpr int K = 1088;
+    constexpr int K = 544;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -49,7 +49,7 @@ TEST_P(RRR_F16_F16_F16, Regular)
 {
     const std::vector<int> Ms{64, 128, 256};
     constexpr int N = 768;
-    constexpr int K = 640;
+    constexpr int K = 320;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -79,7 +79,7 @@ TEST_P(RCR_F16_F16_F16, TinyCases)
 {
     const std::vector<int> Ms{0, 1};
     constexpr int N = 768;
-    constexpr int K = 1088;
+    constexpr int K = 544;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -91,9 +91,9 @@ TEST_P(RCR_F16_F16_F16, TinyCases)
 
 TEST_P(RCR_F16_F16_F16, SmallCases)
 {
-    const std::vector<int> Ms{2, 3, 4, 5};
+    const std::vector<int> Ms{2, 1, 3, 4, 5, 0};
     constexpr int N = 768;
-    constexpr int K = 1088;
+    constexpr int K = 544;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -123,7 +123,7 @@ TEST_P(RCR_F16_F16_F16, Regular)
 {
     const std::vector<int> Ms{32, 64, 128, 256};
     constexpr int N = 768;
-    constexpr int K = 640;
+    constexpr int K = 320;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);
@@ -151,9 +151,9 @@ TEST_P(RCR_F16_F16_F16, MNKPadded)
 
 TEST_P(RRR_F16_F16_F16_LargeK, TestLargeKBatch)
 {
-    const std::vector<int> Ms{127, 150, 188, 210};
+    const std::vector<int> Ms{188, 210};
     constexpr int N = 768;
-    constexpr int K = 8192;
+    constexpr int K = 4096;
 
     const std::vector<int> Ns(Ms.size(), N);
     const std::vector<int> Ks(Ms.size(), K);