format

example/65_gemm_multiply_multiply/gemm_multiply_multiply_xdl_fp8.cpp

@@ -72,7 +72,7 @@ using CDEElementOp = MultiplyMultiply;
 static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNPadding;
 
 using DeviceOpInstance = ck::tensor_operation::device::DeviceGemmMultiD_Xdl_CShuffle_V3
-// clang-format off
+    // clang-format off
          <      Row,      Col, DsLayout, ELayout,
          A0DataType, B0DataType, DsDataType, EDataType, AccDataType, CShuffleDataType,
          AElementOp,  BElementOp, CDEElementOp,       GemmSpec,
@@ -134,7 +134,8 @@ int main(int argc, char* argv[])
         printf("arg1: verification (0=no, 1=yes)\n");
         printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
         printf("arg3: time kernel (0=no, 1=yes)\n");
-        printf("arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideD, StrideE, KBatch\n");
+        printf(
+            "arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideD, StrideE, KBatch\n");
         exit(0);
     }
 

example/65_gemm_multiply_multiply/gemm_multiply_multiply_xdl_fp8_ab_scale.cpp

@@ -60,7 +60,7 @@ static constexpr ck::index_t Scale_Block_N = 128;
 static constexpr ck::index_t Scale_Block_K = 128;
 
 using DeviceOpInstance = ck::tensor_operation::device::DeviceGemmMultiD_ABScale_Xdl_CShuffle_V3
-// clang-format off
+    // clang-format off
          <Row, Col, DsLayout, ELayout,
           A0DataType, A1DataType, B0DataType, B1DataType, DsDataType, EDataType, AccDataType, CShuffleDataType, 
           AElementOp,  BElementOp, CDEElementOp, GemmSpec,

include/ck/tensor_operation/gpu/device/device_gemm_multiple_d.hpp

@@ -96,8 +96,6 @@ struct DeviceGemmMultipleDSplitK : public BaseOperator
     virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
 };
 
-
-
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3.hpp

@@ -70,16 +70,16 @@ template <typename ALayout,
           typename LDSTypeA                           = ComputeTypeA,
           typename LDSTypeB                           = ComputeTypeB>
 struct DeviceGemmMultiD_Xdl_CShuffle_V3 : public DeviceGemmMultipleDSplitK<ALayout,
-                                                                     BLayout,
-                                                                     DsLayout,
-                                                                     CLayout,
-                                                                     ADataType,
-                                                                     BDataType,
-                                                                     DsDataType,
-                                                                     CDataType,
-                                                                     AElementwiseOperation,
-                                                                     BElementwiseOperation,
-                                                                     CElementwiseOperation>
+                                                                           BLayout,
+                                                                           DsLayout,
+                                                                           CLayout,
+                                                                           ADataType,
+                                                                           BDataType,
+                                                                           DsDataType,
+                                                                           CDataType,
+                                                                           AElementwiseOperation,
+                                                                           BElementwiseOperation,
+                                                                           CElementwiseOperation>
 {
     static constexpr index_t NumDTensor = DsDataType::Size();
 
@@ -192,12 +192,11 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3 : public DeviceGemmMultipleDSplitK<ALayo
                         // rotating mem
                         rotating_mem.Next();
                         // clear c mem
-                            if(arg_.KBatch > 1)
-                                hipGetErrorString(
-                                    hipMemsetAsync(arg_.p_c_grid,
-                                                   0,
-                                                   arg_.M * arg_.N * sizeof(CDataType),
-                                                   stream_config.stream_id_));
+                        if(arg_.KBatch > 1)
+                            hipGetErrorString(hipMemsetAsync(arg_.p_c_grid,
+                                                             0,
+                                                             arg_.M * arg_.N * sizeof(CDataType),
+                                                             stream_config.stream_id_));
                     };
 
                     ave_time = ck::utility::launch_and_time_kernel_with_preprocess<false>(