clang-format-12

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_blockscale_bpreshuffle.hpp

@@ -234,7 +234,8 @@ struct DeviceGemmMultiD_BlockScale_Xdl_CShuffle_V3_BPreshuffle
             };
 
             // unconditional 2 to remove agpr usage
-            constexpr index_t minimum_occupancy = MPerBlock * NPerBlock * KPerBlock > (128*128*128/sizeof(ADataType))? 1: 2;
+            constexpr index_t minimum_occupancy =
+                MPerBlock * NPerBlock * KPerBlock > (128 * 128 * 128 / sizeof(ADataType)) ? 1 : 2;
 
             if(has_main_k_block_loop)
             {

include/ck/utility/type_convert.hpp

@@ -1391,7 +1391,8 @@ inline __host__ __device__ f4x2_t f4_convert_rne(float2_t x, float scale = 1.0f)
     // If we keep origin order, error occured:
     value.bitwise = __builtin_amdgcn_cvt_scalef32_pk_fp4_f32(value.bitwise, x[0], x[1], scale, 0);
     // permute high bits and low bits to match the order of the original vector
-    // value.bitwise = __builtin_amdgcn_cvt_scalef32_pk_fp4_f32(value.bitwise, x[1], x[0], scale, 0);
+    // value.bitwise = __builtin_amdgcn_cvt_scalef32_pk_fp4_f32(value.bitwise, x[1], x[0], scale,
+    // 0);
     return value.f4x2_array[0];
 #else
     union

test/data_type/test_mx_fp4.cpp

@@ -91,7 +91,7 @@ test_mx_fp4_scaled_convert(uint64_t N, float* p_test, uint64_t* p_completed)
     {
         return;
     }
-    p_test[i++] = f32x2[1];   // 2* 0b0001(=0.5) = 1.0
+    p_test[i++] = f32x2[1]; // 2* 0b0001(=0.5) = 1.0
     if(i >= N)
     {
         return;