Merge remote-tracking branch 'origin/develop' into aosewski/ggemm_multi_d2

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

@@ -658,27 +658,28 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceO
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
-                std::cout << "arg.Batch_ = " << arg.Batch_ << std::endl;
+                {
+                    std::cout << "arg.Batch_ = " << arg.Batch_ << std::endl;
 
-                std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
-                          << arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
-                          << arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) << ", "
-                          << arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) << "}" << std::endl;
+                    std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
+                              << arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
+                              << arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) << ", "
+                              << arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) << "}" << std::endl;
 
-                std::cout << "arg.b_grid_desc_bk0_n_bk1_{"
-                          << arg.b_grid_desc_bk0_n_bk1_.GetLength(I0) << ", "
-                          << arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) << ", "
-                          << arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) << "}" << std::endl;
+                    std::cout << "arg.b_grid_desc_bk0_n_bk1_{"
+                              << arg.b_grid_desc_bk0_n_bk1_.GetLength(I0) << ", "
+                              << arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) << ", "
+                              << arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) << "}" << std::endl;
 
-                std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
-                          << arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
+                    std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0)
+                              << ", " << arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
 
-                std::cout << "arg.reduce_grid_desc_m_{ " << arg.reduce_grid_desc_m_.GetLength(I0)
-                          << "}" << std::endl;
+                    std::cout << "arg.reduce_grid_desc_m_{ "
+                              << arg.reduce_grid_desc_m_.GetLength(I0) << "}" << std::endl;
+                }
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
                                             arg.b_grid_desc_bk0_n_bk1_,

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

@@ -858,7 +858,7 @@ struct DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle
 
     static bool IsSupportedArgument(const RawArg& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {
@@ -1435,7 +1435,7 @@ struct DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle
 #if 0
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {

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

@@ -719,9 +719,10 @@ struct DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-#if DEBUG_LOG
-        arg.Print();
-#endif
+        if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
+        {
+            arg.Print();
+        }
 
         if(!ck::is_xdl_supported())
         {

include/ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_cshuffle.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
 
@@ -627,7 +627,8 @@ struct DeviceContractionMultipleD_Xdl_CShuffle
             arg.a_max_read_elems_ % ABlockTransferSrcScalarPerVector == 0;
         const bool valid_a_access_dim_m = ABlockTransferSrcVectorDim == 1 && arg.a_mz_consecutive_;
         const bool valid_a_access_dim_k = ABlockTransferSrcVectorDim == 2 && arg.a_kz_consecutive_;
-        const bool valid_a_access_dim   = valid_a_access_dim_m || valid_a_access_dim_k;
+        const bool valid_a_access_dim =
+            valid_a_access_dim_m || valid_a_access_dim_k || ABlockTransferSrcScalarPerVector == 1;
         if(!(valid_a_vector_size && valid_a_access_dim))
         {
             return false;
@@ -637,7 +638,8 @@ struct DeviceContractionMultipleD_Xdl_CShuffle
             arg.b_max_read_elems_ % BBlockTransferSrcScalarPerVector == 0;
         const bool valid_b_access_dim_n = BBlockTransferSrcVectorDim == 1 && arg.b_nz_consecutive_;
         const bool valid_b_access_dim_k = BBlockTransferSrcVectorDim == 2 && arg.b_kz_consecutive_;
-        const bool valid_b_access_dim   = valid_b_access_dim_n || valid_b_access_dim_k;
+        const bool valid_b_access_dim =
+            valid_b_access_dim_n || valid_b_access_dim_k || BBlockTransferSrcScalarPerVector == 1;
         if(!(valid_b_vector_size && valid_b_access_dim))
         {
             return false;
@@ -648,7 +650,8 @@ struct DeviceContractionMultipleD_Xdl_CShuffle
             const bool valid_d_vector_size =
                 arg.ds_max_read_elems_[i] % CDEBlockTransferScalarPerVector_NPerBlock == 0;
             // Vector read of Ds is always on N dimension.
-            const bool valid_d_access_dim = arg.ds_nz_consecutive_[i];
+            const bool valid_d_access_dim =
+                arg.ds_nz_consecutive_[i] || CDEBlockTransferScalarPerVector_NPerBlock == 1;
             if(!(valid_d_vector_size && valid_d_access_dim))
             {
                 valid_ds_access = false;
@@ -662,7 +665,8 @@ struct DeviceContractionMultipleD_Xdl_CShuffle
         const bool valid_e_vector_size =
             arg.e_max_write_elems_ % CDEBlockTransferScalarPerVector_NPerBlock == 0;
         // Vector write of E is always on N dimension.
-        const bool valid_e_access_dim = arg.e_nz_consecutive_;
+        const bool valid_e_access_dim =
+            arg.e_nz_consecutive_ || CDEBlockTransferScalarPerVector_NPerBlock == 1;
         if(!(valid_e_vector_size && valid_e_access_dim))
         {
             return false;

include/ck/tensor_operation/gpu/device/impl/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp

@@ -516,26 +516,27 @@ struct DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
             float ave_time = 0;
             for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
             {
-#if DEBUG_LOG
+                if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
                 {
-                    std::cout << "arg.a_grid_desc_k0_m_k1_container_{"
-                              << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "
-                              << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I1) << ", "
-                              << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I2) << "}"
-                              << std::endl;
-
-                    std::cout << "arg.b_grid_desc_k0_n_k1_container_{"
-                              << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I0) << ", "
-                              << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I1) << ", "
-                              << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I2) << "}"
-                              << std::endl;
-
-                    std::cout << "arg.c_grid_desc_m_n_container_{ "
-                              << arg.c_grid_desc_m_n_container_[i].GetLength(I0) << ", "
-                              << arg.c_grid_desc_m_n_container_[i].GetLength(I1) << "}"
-                              << std::endl;
+                    {
+                        std::cout << "arg.a_grid_desc_k0_m_k1_container_{"
+                                  << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "
+                                  << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I1) << ", "
+                                  << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I2) << "}"
+                                  << std::endl;
+
+                        std::cout << "arg.b_grid_desc_k0_n_k1_container_{"
+                                  << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I0) << ", "
+                                  << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I1) << ", "
+                                  << arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I2) << "}"
+                                  << std::endl;
+
+                        std::cout << "arg.c_grid_desc_m_n_container_{ "
+                                  << arg.c_grid_desc_m_n_container_[i].GetLength(I0) << ", "
+                                  << arg.c_grid_desc_m_n_container_[i].GetLength(I1) << "}"
+                                  << std::endl;
+                    }
                 }
-#endif
 
                 if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_container_[i],
                                                 arg.b_grid_desc_k0_n_k1_container_[i],

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_add_nhwc_kyxc_nhwk.hpp

@@ -644,7 +644,7 @@ struct
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << DeviceOp{}.GetTypeString() << std::endl;
                 std::cout << "N " << arg.Conv_N_ << ", "
@@ -664,9 +664,7 @@ struct
                           << arg.input_left_pads_[1] << ", " << std::endl;
                 std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
                           << arg.input_right_pads_[1] << ", " << std::endl;
-            }
 
-            {
                 std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
                           << ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
                           << arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
@@ -684,7 +682,6 @@ struct
                 std::cout << "arg.c1_grid_desc_m_n_{ " << arg.c1_grid_desc_m_n_.GetLength(I0)
                           << ", " << arg.c1_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
                                             arg.b_grid_desc_k0_n_k1_,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_nhwc_kyxc_nhwk.hpp

@@ -614,7 +614,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << DeviceOp{}.GetTypeString() << std::endl;
                 std::cout << "N " << arg.Conv_N_ << ", "
@@ -634,9 +634,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X
                           << arg.input_left_pads_[1] << ", " << std::endl;
                 std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
                           << arg.input_right_pads_[1] << ", " << std::endl;
-            }
 
-            {
                 std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
                           << ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
                           << arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
@@ -651,7 +649,6 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X
                 std::cout << "arg.c0_grid_desc_m_n_{ " << arg.c0_grid_desc_m_n_.GetLength(I0)
                           << ", " << arg.c0_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
                                             arg.b_grid_desc_k0_n_k1_,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp

@@ -579,7 +579,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << DeviceOp{}.GetTypeString() << std::endl;
                 std::cout << "N " << arg.Conv_N_ << ", "
@@ -599,9 +599,7 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
                           << arg.input_left_pads_[1] << ", " << std::endl;
                 std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
                           << arg.input_right_pads_[1] << ", " << std::endl;
-            }
 
-            {
                 std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
                           << ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
                           << arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
@@ -635,7 +633,6 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
                            .GetLength(I5)
                     << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
                                             arg.b_grid_desc_k0_n_k1_,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk.hpp

@@ -431,7 +431,7 @@ struct DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
                           << ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
@@ -444,7 +444,7 @@ struct DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
                 std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
                           << arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
             }
-#endif
+
             if(!GridwiseGemm::CheckValidity(
                    arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.c_grid_desc_m_n_))
             {

include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp

@@ -401,7 +401,7 @@ struct DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << "num_batches_of_GEMM = " << arg.num_subbatches_ << std::endl;
                 std::cout << "a_grid_desc_k0_m_k1{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
@@ -415,7 +415,6 @@ struct DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_
                 std::cout << "c_grid_desc_m_n{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
                           << arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
                                             arg.b_grid_desc_k0_n_k1_,

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

@@ -1272,7 +1272,7 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Dl
             float ave_time = 0;
             for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
             {
-#if DEBUG_LOG
+                if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
                 {
                     std::cout << "arg.a_grid_desc_k0_m_k1_container_{"
                               << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "
@@ -1305,7 +1305,6 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Dl
                               << arg.c_grid_desc_m0_m10_m11_n0_n10_n11_container_[i].GetLength(I5)
                               << " ) " << std::endl;
                 }
-#endif
 
                 if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_container_[i],
                                                 arg.b_grid_desc_k0_n_k1_container_[i],
@@ -1393,8 +1392,8 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Dl
     static bool IsSupportedArgument(const Argument& arg)
     {
         // check device
-        if(!(ck::get_device_name() == "gfx906" || ck::is_navi2_supported() ||
-             ck::is_navi3_supported()))
+        if(!(ck::get_device_name() == "gfx906" || ck::is_gfx103_supported() ||
+             ck::is_gfx11_supported()))
         {
             return false;
         }

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

@@ -1220,7 +1220,7 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Xdl
             float ave_time = 0;
             for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
             {
-#if DEBUG_LOG
+                if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
                 {
                     std::cout << "arg.a_grid_desc_k0_m_k1{"
                               << arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "
@@ -1239,7 +1239,6 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Xdl
                               << arg.c_grid_desc_m_n_container_[i].GetLength(I1) << "}"
                               << std::endl;
                 }
-#endif
 
                 if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_container_[i],
                                                 arg.b_grid_desc_k0_n_k1_container_[i],

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

@@ -509,7 +509,7 @@ struct DeviceFpAintBGemm_Wmma_CShuffle : public DeviceGemm_dequantB<ALayout,
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, ck::half_t> ||
                            is_same_v<AccDataType, int32_t>))

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

@@ -334,7 +334,7 @@ struct DeviceGemmDl : public DeviceGemm<ALayout,
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << "arg.a_grid_desc_k0_m0_m1_k1_{"
                           << arg.a_grid_desc_k0_m_k1_.GetLength(I0) << ", "
@@ -349,7 +349,6 @@ struct DeviceGemmDl : public DeviceGemm<ALayout,
                 std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
                           << arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(
                    arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.c_grid_desc_m_n_))
@@ -536,8 +535,8 @@ struct DeviceGemmDl : public DeviceGemm<ALayout,
             }
         }
 
-        if(ck::get_device_name() == "gfx906" || ck::is_navi2_supported() ||
-           ck::is_navi3_supported())
+        if(ck::get_device_name() == "gfx906" || ck::is_gfx103_supported() ||
+           ck::is_gfx11_supported())
         {
             return GridwiseGemm::CheckValidity(
                 arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.c_grid_desc_m_n_);

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

@@ -168,7 +168,7 @@ struct DeviceGemmDpp : public DeviceGemm<ALayout,
 
     static bool IsSupportedArgument(const Argument& karg)
     {
-        if(ck::is_navi2_supported() || ck::is_navi3_supported())
+        if(ck::is_gfx103_supported() || ck::is_gfx11_supported())
         {
             return GridwiseGemm::CheckValidity(karg);
         }

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

@@ -552,7 +552,7 @@ struct DeviceGemmMultipleD_Dl : public DeviceGemmMultipleD<ALayout,
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(ck::get_device_name() == "gfx906" || ck::is_xdl_supported() ||
-           ck::is_navi2_supported() || ck::is_navi3_supported())
+           ck::is_gfx103_supported() || ck::is_gfx11_supported())
         {
             return GridwiseGemm::CheckValidity(
                 arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.e_grid_desc_m_n_);

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

@@ -515,7 +515,7 @@ struct DeviceGemmMultipleD_Wmma_CShuffle : public DeviceGemmMultipleD<ALayout,
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
             {

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

@@ -510,7 +510,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceOperatio
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-#if DEBUG_LOG
+            if(ck::EnvIsEnabled(ENV(CK_LOGGING)))
             {
                 std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
                           << arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
@@ -528,7 +528,6 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceOperatio
                 std::cout << "arg.reduce_grid_desc_m_{ " << arg.reduce_grid_desc_m_.GetLength(I0)
                           << "}" << std::endl;
             }
-#endif
 
             if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
                                             arg.b_grid_desc_bk0_n_bk1_,

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

@@ -443,7 +443,7 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, ck::half_t> ||
                            is_same_v<AccDataType, int32_t>))