Re: Shared Generic Instance Code

[dewar@merv.cs.nyu.edu (Robert Dewar)]

<<Though I haven't done a lot of timing studies, I have the distinct
impression that big generics can *drastically* slow down compilation
(Gnat 3.04a NT/95).  I presume the macro is essentially expanded and
recompiled each time it's instantiated?>>

Compiling a macro expanded generic is indeed like compiling the source,
it should be no slower or no faster. Big generics are best avoided if
you are working with a compiler without shared generics and you plan
to do many instantiations, as I mentioned before, have a look at the
GNAT implementation of Integer_IO or Float_IO to see what I mean by
this.

But in any case I was talking about runtime efficiency, not compile
time speed, when I said that macro instantiated generics will generally
be significantly more efficient.