bugs in device pad

src/targets/gpu/device/pad.cpp

@@ -17,21 +17,28 @@ pad(hipStream_t stream, argument result, argument arg1, float value, std::vector
     std::size_t nelements = arg1.get_shape().elements();
     if(float_equal(value, std::numeric_limits<float>::lowest()))
     {
-        visit_all(result)([&](auto output) {
-            auto* outptr = device_cast(output.data());
-            auto val =
-                device_cast(std::numeric_limits<typename decltype(output)::value_type>::lowest());
+        auto val =
+                device_cast(std::numeric_limits<decltype(value)>::lowest());         
+        nary(stream, result)([=] { return val; });
+        // visit_all(result)([&](auto output) {
+        //     auto* outptr = device_cast(output.data());
+        //     auto val =
+        //         device_cast(std::numeric_limits<typename decltype(output)::value_type>::lowest());
 
-            gs_launch(stream, nelements)([=](auto i) { outptr[i] = val; });
-        });
+        //     gs_launch(stream, nelements)([=](auto i) { outptr[i] = val; });
+        // });
     }
 
     else
     {
-        visit_all(result)([&](auto output) {
-            auto* outptr = device_cast(output.data());
-            gs_launch(stream, nelements)([=](auto i) { outptr[i] = value; });
-        });
+        // visit_all(result)([&](auto output) {
+        //     auto* outptr = device_cast(output.data());
+        //     auto val =
+        //         device_cast(value);
+        //     gs_launch(stream, nelements)([=](auto i) { outptr[i] = val; });
+        // });'
+        auto val = device_cast(value);
+        nary(stream, result)([=] { return val; });
     }
 
     // nary(stream, result)([=] { return value; });