Fixed StaticFor for loops with a large iterations count

src/TNL/StaticFor.h

@@ -14,25 +14,47 @@
 
 namespace TNL {
 
+// Manual unrolling does not make sense for loops with a large iterations
+// count. For a very large iterations count it would trigger the compiler's
+// limit on recursive template instantiation. Also note that the compiler
+// will (at least partially) unroll loops with static bounds anyway.
+template< int Begin, int End, bool unrolled = (End - Begin <= 8) >
+struct StaticFor;
+
 template< int Begin, int End >
-struct StaticFor
+struct StaticFor< Begin, End, true >
 {
-    template< typename Function, typename... Args >
-    __cuda_callable__
-    static void exec( const Function& f, Args... args )
-    {
-        static_assert( Begin < End, "Wrong index interval for StaticFor. Being must be lower than end." );
-        f( Begin, args... );
-        StaticFor< Begin + 1, End >::exec( f, args... );
-    };
+   static_assert( Begin < End, "Wrong index interval for StaticFor. Begin must be less than end." );
+
+   template< typename Function, typename... Args >
+   __cuda_callable__
+   static void exec( const Function& f, Args... args )
+   {
+      f( Begin, args... );
+      StaticFor< Begin + 1, End >::exec( f, args... );
+   }
 };
 
 template< int End >
-struct StaticFor< End, End >
+struct StaticFor< End, End, true >
 {
-    template< typename Function, typename... Args >
-    __cuda_callable__
-    static void exec( const Function& f, Args... args ){};
+   template< typename Function, typename... Args >
+   __cuda_callable__
+   static void exec( const Function& f, Args... args ) {}
+};
+
+template< int Begin, int End >
+struct StaticFor< Begin, End, false >
+{
+   static_assert( Begin <= End, "Wrong index interval for StaticFor. Begin must be less than or equal to end." );
+
+   template< typename Function, typename... Args >
+   __cuda_callable__
+   static void exec( const Function& f, Args... args )
+   {
+      for( int i = Begin; i < End; i++ )
+         f( i, args... );
+   }
 };
 
 } //namespace TNL