tmp save

include/ck/utility/amd_buffer_addressing.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 #include "data_type.hpp"

include/ck/utility/dynamic_buffer.hpp

@@ -357,7 +357,8 @@ struct DynamicBuffer
 #if CK_USE_AMD_BUFFER_ATOMIC_ADD_INTEGER && CK_USE_AMD_BUFFER_ATOMIC_ADD_FLOAT
         bool constexpr use_amd_buffer_addressing =
             is_same_v<remove_cvref_t<scalar_t>, int32_t> ||
-            is_same_v<remove_cvref_t<scalar_t>, float> ||
+            is_same_v<remove_cvref_t<scalar_t>, float> || 
+            (is_same_v<remove_cvref_t<scalar_t>, bhalf_t> && scalar_per_x_vector % 2 == 0) ||
             (is_same_v<remove_cvref_t<scalar_t>, half_t> && scalar_per_x_vector % 2 == 0);
 #elif CK_USE_AMD_BUFFER_ATOMIC_ADD_INTEGER && (!CK_USE_AMD_BUFFER_ATOMIC_ADD_FLOAT)
         bool constexpr use_amd_buffer_addressing = is_same_v<remove_cvref_t<scalar_t>, int32_t>;

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_bf16_i8_bf16_mk_kn_mn_instance.cpp

@@ -33,7 +33,7 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
 
-using device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
+using device_grouped_gemm_xdl_fixed_nk_bf16_i8_bf16_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|
@@ -57,16 +57,16 @@ void add_device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_instances(
                                                          Row,
                                                          DsLayout,
                                                          Row,
-                                                         F16,
+                                                         BF16,
                                                          I8,
                                                          DsDataType,
-                                                         F16,
+                                                         BF16,
                                                          PassThrough,
                                                          PassThrough,
                                                          PassThrough>>>& instances)
 {
     add_device_operation_instances(
-        instances, device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_irregular_tile_instances{});
+        instances, device_grouped_gemm_xdl_fixed_nk_bf16_i8_bf16_mk_kn_mn_irregular_tile_instances{});
 }
 
 } // namespace instance