Enable template max and min for HIP-Clang (#2028)

It was for HCC only. HIP-Clang also needs it for __fp16 since AMDMIGraphX uses it.

Change-Id: Id49322b7b89ef799accdf6b47627a6fce51d1ab5

include/hip/hcc_detail/math_functions.h

@@ -1410,12 +1410,18 @@ float func(float x, int y) \
 }
 __DEF_FLOAT_FUN2I(scalbn)
 
-#if __HCC__
 template<class T>
 __DEVICE__ inline static T min(T arg1, T arg2) {
   return (arg1 < arg2) ? arg1 : arg2;
 }
 
+template<class T>
+__DEVICE__ inline static T max(T arg1, T arg2) {
+  return (arg1 > arg2) ? arg1 : arg2;
+}
+
+#if __HCC__
+
 __DEVICE__ inline static uint32_t min(uint32_t arg1, int32_t arg2) {
   return min(arg1, (uint32_t) arg2);
 }
@@ -1437,11 +1443,6 @@ __DEVICE__ inline static unsigned long long min(long long arg1, unsigned long lo
   return min((unsigned long long) arg1, arg2);
 }*/
 
-template<class T>
-__DEVICE__ inline static T max(T arg1, T arg2) {
-  return (arg1 > arg2) ? arg1 : arg2;
-}
-
 __DEVICE__ inline static uint32_t max(uint32_t arg1, int32_t arg2) {
   return max(arg1, (uint32_t) arg2);
 }