This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: bzero optimization rarely does
- From: Roger Sayle <roger at eyesopen dot com>
- To: Hans-Peter Nilsson <hp at bitrange dot com>
- Cc: Paul Koning <pkoning at equallogic dot com>, Richard Henderson <rth at redhat dot com>, <gcc at gcc dot gnu dot org>
- Date: Thu, 11 Jul 2002 22:42:06 -0600 (MDT)
- Subject: Re: bzero optimization rarely does
> > The problem is fixed by appropriately defining CLEAR_RATIO and
> > CLEAR_BY_PIECES_P in an analogous way to MOVE_RATIO and
> > MOVE_BY_PIECES_P, and then use this latter macro in clear_storage.
>
> Why not default CLEAR_RATIO expressed in terms of MOVE_RATIO?
The default value of MOVE_RATIO is affected by HAVE_movstrsi,
whereas the default value of CLEAR_RATIO is affected by
HAVE_clrstrsi.
I'm sorry if this wasn't clear from the patches context, but the
"#if defined" condition is quite different from the original.
Backends that have defined MOVE_RATIO may have done so accounting
for the presence or absence of block move instructions. These
factors may be quite different for CLEAR_RATIO, depending upon
block clear instructions.
Certainly on the MIPS :>
Roger
--