clang

example/67_gemm_microscaling/gemm_mx_common.hpp

@@ -317,7 +317,6 @@ bool run_mx_gemm(const ProblemSize& problem_size, const ExecutionConfig& config)
             std::cout << "Computing GEMM on host..." << std::endl;
         }
 
-
         using ReferenceGemmInstance = ck::tensor_operation::host::ReferenceMXGemm<ADataType,
                                                                                   BDataType,
                                                                                   CDataType,
@@ -340,7 +339,6 @@ bool run_mx_gemm(const ProblemSize& problem_size, const ExecutionConfig& config)
                                                   PassThrough{},
                                                   PassThrough{});
 
-
         ref_invoker.Run(ref_argument);
 
         if(config.verbosity > 0)
@@ -357,12 +355,10 @@ bool run_mx_gemm(const ProblemSize& problem_size, const ExecutionConfig& config)
                       << ((res_verified) ? " (PASSED!)" : " (FAILED!)") << std::endl;
         }
 
-
         res_verified = res_verified && ck::utils::check_err(c_m_n_device_result,
                                                             c_m_n_host_result,
                                                             "Error: Incorrect results!");
 
-
         if(config.verbosity > 0 && res_verified)
             std::cout << "Done." << std::endl;
     }

example/ck_tile/03_gemm/run_gemm_example.inc

@@ -30,8 +30,13 @@ auto calculate_rtol_atol(const ck_tile::index_t K,
     return ck_tile::make_tuple(std::max(rtol, rtol_split_k), std::max(atol, atol_split_k));
 }
 
-template <typename ADataType, typename BDataType, typename AccDataType, typename CDataType, 
-          typename ALayout, typename BLayout, typename CLayout>
+template <typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CDataType,
+          typename ALayout,
+          typename BLayout,
+          typename CLayout>
 float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
                   ck_tile::DeviceMem& b_k_n_dev_buf,
                   ck_tile::DeviceMem& c_m_n_dev_buf,
@@ -57,8 +62,8 @@ float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
     args.stride_B = stride_B;
     args.stride_C = stride_C;
 
-    float ave_time = gemm_calc<ADataType, BDataType, AccDataType, CDataType, 
-                               ALayout, BLayout, CLayout>(
+    float ave_time =
+        gemm_calc<ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout>(
             args, ck_tile::stream_config{nullptr, true, 1, n_warmup, n_repeat});
 
     std::size_t flop = std::size_t(2) * M * N * K;
@@ -69,14 +74,11 @@ float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
 
     std::cout << "Run Gemm kernel with M =" << M << " N =" << N << " K =" << K
               << " StrideA =" << stride_A << " StrideB =" << stride_B << " StrideC =" << stride_C
-              << " A_Layout =" << ALayout::name 
-              << " B_Layout =" << BLayout::name 
-              << " C_Layout =" << CLayout::name
-              << " A Type = " << DataTypeTraits<ADataType>::name 
+              << " A_Layout =" << ALayout::name << " B_Layout =" << BLayout::name
+              << " C_Layout =" << CLayout::name << " A Type = " << DataTypeTraits<ADataType>::name
               << " B Type = " << DataTypeTraits<BDataType>::name
-              << " C Type = " << DataTypeTraits<CDataType>::name                             
-              << " : " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
-              << std::endl;
+              << " C Type = " << DataTypeTraits<CDataType>::name << " : " << ave_time << " ms, "
+              << tflops << " TFlops, " << gb_per_sec << " GB/s, " << std::endl;
 
     return ave_time;
 }
@@ -133,8 +135,8 @@ int run_gemm_example_with_layouts(int argc,
     c_m_n_dev_buf.SetZero();
     c_m_n_dev_result.SetZero();
 
-    invoke_gemm<ADataType, BDataType, AccDataType, CDataType, 
-                ALayout, BLayout, CLayout>(a_m_k_dev_buf,
+    invoke_gemm<ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout>(
+        a_m_k_dev_buf,
         b_k_n_dev_buf,
         c_m_n_dev_buf,
         M,
@@ -160,8 +162,8 @@ int run_gemm_example_with_layouts(int argc,
             a_m_k, b_k_n, c_m_n_host_ref);
         const float max_accumulated_value =
             *std::max_element(c_m_n_host_ref.mData.begin(), c_m_n_host_ref.mData.end());
-        const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>
-                                    (K, kbatch, max_accumulated_value);
+        const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>(
+            K, kbatch, max_accumulated_value);
         pass = ck_tile::check_err(c_m_n_dev_result,
                                   c_m_n_host_ref,
                                   "Error: Incorrect results!",
@@ -218,8 +220,8 @@ int run_gemm_example_with_layouts(int argc,
         c_m_n_gpu_buf_ref.FromDevice(c_m_n_gpu_ref.data());
         const float max_accumulated_value =
             *std::max_element(c_m_n_gpu_ref.mData.begin(), c_m_n_gpu_ref.mData.end());
-        const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>
-                                    (K, kbatch, max_accumulated_value);
+        const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>(
+            K, kbatch, max_accumulated_value);
         pass = ck_tile::check_err(c_m_n_dev_result,
                                   c_m_n_gpu_ref,
                                   "Error: Incorrect results!",

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_streamk_v3.hpp

@@ -224,8 +224,8 @@ struct GridwiseGemm_xdl_cshuffle_streamk_v3
         }();
 
         // Pad both M and K to be multiples of the block sizes
-        const auto a_grid_desc_m_k = transform_tensor_descriptor(
-            a_grid_desc_mraw_kraw,
+        const auto a_grid_desc_m_k =
+            transform_tensor_descriptor(a_grid_desc_mraw_kraw,
                                         make_tuple(make_right_pad_transform(M, MPad - M),
                                                    make_right_pad_transform(K, KPad - K)),
                                         make_tuple(Sequence<0>{}, Sequence<1>{}),
@@ -324,8 +324,8 @@ struct GridwiseGemm_xdl_cshuffle_streamk_v3
         }();
 
         // Pad both N and K to be multiples of the block sizes
-        const auto b_grid_desc_n_k = transform_tensor_descriptor(
-            b_grid_desc_nraw_kraw,
+        const auto b_grid_desc_n_k =
+            transform_tensor_descriptor(b_grid_desc_nraw_kraw,
                                         make_tuple(make_right_pad_transform(N, NPad - N),
                                                    make_right_pad_transform(K, KPad - K)),
                                         make_tuple(Sequence<0>{}, Sequence<1>{}),
@@ -1076,7 +1076,6 @@ struct GridwiseGemm_xdl_cshuffle_streamk_v3
                               << __LINE__ << ", in function: " << __func__ << std::endl;
                 }
 
-
                 return false;
             }
         }

profiler/src/profile_gemm_universal_streamk.cpp

@@ -56,7 +56,6 @@ int profile_gemm_universal_streamk(int argc, char* argv[])
         exit(1);
     }
 
-    
     int M;
     int N;
     int StrideA;