This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH]: Handle pow10 like exp10 optab
- From: "Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>
- To: kazu at cs dot umass dot edu, roger at eyesopen dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 12 Mar 2004 09:13:48 -0500 (EST)
- Subject: Re: [PATCH]: Handle pow10 like exp10 optab
- References: <email@example.com>
> From: Kazu Hirata <firstname.lastname@example.org>
> Hi Roger and Kaveh,
> > > * builtins.c (expand_builtin_mathfn): Add pow10* to the
> > > existing exp10* case.
> > > (expand_builtin): Likewise.
> You might be interested in knowing that these builtin optimizations
> probably never trigger on tree-ssa. (See PR 14541.)
> According to Andrew Pinski and Daniel Berlin, an expression like
> "a = log (exp (x))" is gimplified first like
> T = exp (x);
> a = log (T);
> before fold() [CALL_EXPR] gets to see "log (exp (x))".
I see your point, and I guess something will need to be done about
that. However FYI this particular pow10 optimization is expanded at
the RTL level irrespective of what it's argument would be.
Kaveh R. Ghazi email@example.com