This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
How to best fold __builtin_powi?
- From: Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>
- To: gcc at gcc dot gnu dot org
- Date: Wed, 2 Feb 2005 16:56:22 +0100 (CET)
- Subject: How to best fold __builtin_powi?
Before coming up with the obvious follow-up patch after
introducing __builtin_powi[lf], I'd like to know wether
it would be accepted (at this stage) to move expansion
of pow(x, integer_cst) to the tree level. This would
reduce overall (source) code. A quick check produces
very slightly better code than expanding at the rtl
level (for a 5th degree polynomial calculation there is
one fxch less).
If this is too invasive for 4.0.0 I'll queue that for later
and just do the obvious stuff from fold_builtin_pow in fold_builtin_powi.
Thanks,
Richard.
--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/