Use 'DevicePermute' device op in example

example/36_elementwise_permute/common.hpp

@@ -13,7 +13,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/device/device_elementwise.hpp"
+#include "ck/tensor_operation/gpu/device/device_permute.hpp"
 #include "ck/utility/type.hpp"
 
 #include "ck/library/utility/check_err.hpp"

example/36_elementwise_permute/elementwise_permute_fp16.cpp

@@ -7,7 +7,7 @@ using ADataType = F16;
 using BDataType = F16;
 
 using DeviceElementwisePermuteInstance = ck::tensor_operation::device::
-    DeviceElementwise<ck::Tuple<ADataType>, ck::Tuple<BDataType>, PassThrough, 4, 8, S<8>, S<1>>;
+    DevicePermute<ck::Tuple<ADataType>, ck::Tuple<BDataType>, PassThrough, 4, 8, S<8>, S<1>>;
 
 #include "run_elementwise_permute_example.inc"