This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Sigh. Inlining heuristics.
- To: ak at suse dot de, dewar at gnat dot com
- Subject: Re: Sigh. Inlining heuristics.
- From: dewar at gnat dot com
- Date: Tue, 10 Jul 2001 08:12:45 -0400 (EDT)
- Cc: gcc at gcc dot gnu dot org
<<I can easily imagine such a program. e.g. when you inline a 1000 instruction
function and 900 instructions of that are removed during constant
propagation and dead code remove later because the argument input values
are known then. Of course 100 is still a fine default.
>>
Yes, well compiler writers are expert at "imagining" progrags that would
be helped by their pet optimizations, and that is a lot of the reason that
optimizations tend to prove disappointing. What is important is not what you
can imagine, but what actually occurs in real programs significantly often.