Merge branch 'amd-develop' into amd-master

include/ck/tensor_operation/gpu/device/impl/device_image_to_column_impl.hpp

@@ -57,8 +57,8 @@ struct DeviceImageToColumnImpl
     static constexpr auto I1 = Number<1>{};
     static constexpr auto I2 = Number<2>{};
 
-    static constexpr auto conv_to_gemm_transformer =
-        TransformConvFwdToGemm<NDimSpatial, ConvolutionForwardSpecialization::Default>{};
+    using ConvToGemmFwdTransformer =
+        TransformConvFwdToGemm<NDimSpatial, ConvolutionForwardSpecialization::Default>;
 
     static constexpr auto matrix_padder =
         MatrixPadder<GemmSpecialization::MKPadding, index_t, index_t, index_t>{
@@ -97,9 +97,7 @@ struct DeviceImageToColumnImpl
         b_g_k_c_xs_lengths[I2]  = C;
         c_g_n_k_wos_lengths[I1] = N;
 
-        const auto in_gemmmraw_gemmkraw_desc =
-            conv_to_gemm_transformer.template MakeADescriptor_M_K<ImageLayout>(
-                a_g_n_c_wis_lengths,
+        ConvToGemmFwdTransformer conv_to_gemm_transformer{a_g_n_c_wis_lengths,
                                                           image_g_n_c_wis_strides,
                                                           b_g_k_c_xs_lengths,
                                                           {}, // not needed for A Descriptor
@@ -108,8 +106,10 @@ struct DeviceImageToColumnImpl
                                                           conv_filter_strides,
                                                           conv_filter_dilations,
                                                           input_left_pads,
-                input_right_pads,
-                N);
+                                                          input_right_pads};
+
+        const auto in_gemmmraw_gemmkraw_desc =
+            conv_to_gemm_transformer.template MakeADescriptor_M_K<ImageLayout>();
 
         const auto in_gemmm_gemmk_desc =
             matrix_padder.PadADescriptor_M_K(in_gemmmraw_gemmkraw_desc);

include/ck/tensor_operation/gpu/grid/gridwise_2d_multiple_reduction_multiblock.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
 
@@ -258,7 +258,7 @@ struct GridwiseMultipleReduction_mk_to_m_multiblock
 
             if(thread_k_cluster_id == 0)
             {
-                if(block_group_size == 0 && !float_equal_zero{}(beta_values[iR]))
+                if(!float_equal_zero{}(beta_values[iR]))
                 {
                     StaticBuffer<AddressSpaceEnum::Vgpr, OutDataType, MThreadSliceSize, true>
                         priorDstValueBuf;

include/ck/tensor_operation/gpu/grid/gridwise_2d_reduction_multiblock.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
 
@@ -244,7 +244,7 @@ struct GridwiseReduction_mk_to_m_multiblock
 
         if(thread_k_cluster_id == 0)
         {
-            if(block_group_size == 0 && !float_equal_zero{}(beta))
+            if(!float_equal_zero{}(beta))
             {
                 StaticBuffer<AddressSpaceEnum::Vgpr, OutDataType, MThreadSliceSize, true>
                     priorDstValueBuf;