This is the mail archive of the
mailing list for the GCC project.
Re: Draft "Unsafe fp optimizations" project description.
- To: dewar at gnat dot com, toon at moene dot indiv dot nluug dot nl
- Subject: Re: Draft "Unsafe fp optimizations" project description.
- From: dewar at gnat dot com
- Date: Sun, 5 Aug 2001 18:30:44 -0400 (EDT)
- Cc: gcc at gcc dot gnu dot org, lucier at math dot purdue dot edu, tprince at computer dot org
<<It all depends on what you mean by "now". -ffast-math (no c :-) enables
underflow-to-zero on ia64 and alpha. On my alpha (I've never even seen
a ia64 workstation, but they seem to exist), forcing the processor to
treat underflow-as-zero meant an 1.8 times speed up on our Numerical
Weather Forecasting system.
Sure but that is exatly the case we know is highly significant on this
processor, what is interesting is to know if *any* of the other transformations
have any effect at all.
For my taste I would have underflow-to-zero be the default on alpha, and
only use the very expensive denormal stuff if -mieee is set.