This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Sigh. Inlining heuristics.
- To: dewar at gnat dot com, kenner at vlsi1 dot ultra dot nyu dot edu
- Subject: Re: Sigh. Inlining heuristics.
- From: dewar at gnat dot com
- Date: Tue, 10 Jul 2001 07:45:11 -0400 (EDT)
- Cc: gcc at gcc dot gnu dot org
<<That strongly depends on the loop structure of that function. If we have
a function that's straight-line code and calls, say, ten functions that
aren't called elsewhere, inlining those calls, no matter what the sizes of
the functions, is a performance win, though admittedly probably a small one.
>>
Sure, that's at least theoretically true. Nevertheless, in the case of
GNAT with GCC 2.8.1, we have seen that -O3 is almost always a
pessimization, at least on the x86 where we have most data, compared
to -O2.