Here are some benchmarks comparing the performance of Array(...) and Debox's
array(...) macro for varargs lists of various lengths.

benchmark    ns linear runtime
   Apply0  14.1 =
   Macro0  13.7 =
   Apply1  31.4 ==
   Macro1  14.7 =
   Apply2  27.6 =
   Macro2  14.8 =
   Apply4  35.7 ==
   Macro4  15.5 =
   Apply8  46.0 ===
   Macro8  15.9 =
  Apply16  66.5 ====
  Macro16  18.1 =
  Apply32 111.3 =======
  Macro32  26.7 =
  Apply64 222.9 ==============
  Macro64  48.5 ===
 Apply128 450.0 ==============================
 Macro128  92.6 ======

As you can see, for very small arrays the macro version in Debox is about
twice as fast (excepting the 0 case). As the number of arguments increases,
this advantage grows: ~3x at 8 items, ~4x at 32 items, and ~5x at 128 items.

(Of course, given that these times are in nanoseconds, you have to do this
sort of thing a lot before it adds up to an appreciable amount of user time.)

Debox is at: https://github.com/non/debox