This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: x86 code generation question
- From: Zack Weinberg <zack at codesourcery dot com>
- To: Vadim Lobanov <vadim at cs dot washington dot edu>
- Cc: Paul Koning <pkoning at equallogic dot com>, <gcc at gcc dot gnu dot org>
- Date: Thu, 29 Apr 2004 16:38:34 -0700
- Subject: Re: x86 code generation question
- References: <20040429162251.F8144-100000@attu2.cs.washington.edu>
Vadim Lobanov <vadim@cs.washington.edu> writes:
> Ah, okay. So essentially it is done this way to maintain generality and
> architecture-independence of the code that translates syntax trees into
> RTL.
Generality? Architecture-independence? snicker ... guffaw
... BWAHAHAHA!
It's more like "that code was written about a decade before processors
with conditional move existed, with no design, and no one has ever
dared to rewrite it."
But you are still correct - if someone *did* overhaul that code (which
might even happen one of these years, what with tree-ssa (potentially)
making its job so much easier) they probably wouldn't implement
conditional move.
It should be noted that the code generated by gcc at -O0 is very very
bad in general. This might change in the future, but for the time
being, we basically aren't interested in reports that the code
generated at -O0 is lame. -O1 or -O2, that's a different story.
zw