clang format

src/targets/gpu/include/migraphx/gpu/device/arg_op.hpp

@@ -21,13 +21,13 @@ struct val_index
     int64_t index;
 };
 
-template<class T>
+template <class T>
 MIGRAPHX_DEVICE_CONSTEXPR val_index<T> make_val_index(T v)
 {
     return {v, -1};
 }
 
-template<class T>
+template <class T>
 MIGRAPHX_DEVICE_CONSTEXPR val_index<T> make_val_index(T v, int64_t i)
 {
     return {v, i};
@@ -85,14 +85,15 @@ void arg_op(Op op, hipStream_t stream, const argument& result, const argument& a
         // use one block for items in one batch.
         const size_t max_block_size  = 256;
         const std::size_t block_size = compute_block_size(batch_item_num, max_block_size);
-        gs_launch(stream, batch_shape.elements() * block_size, block_size)(
-            [=](auto i, auto idx) __device__ {
+        gs_launch(stream,
+                  batch_shape.elements() * block_size,
+                  block_size)([=](auto i, auto idx) __device__ {
             auto batch_idx = batch_s.multi(i / block_size);
             auto data_idx  = batch_idx;
             auto init      = make_val_index<type>(op.init());
 
-                auto op_output = block_reduce<max_block_size>(
-                    idx, op, init, batch_item_num, [&](auto j) __device__ {
+            auto op_output =
+                block_reduce<max_block_size>(idx, op, init, batch_item_num, [&](auto j) __device__ {
                     data_idx[axis] = j;
                     return make_val_index(input[arg_s.index(data_idx)], j);
                 });