This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: Serious code size regression from 3.0.2 to now
- From: Joern Rennecke <joern dot rennecke at superh dot com>
- To: tm <tm at mail dot kloo dot net>
- Cc: Geoff Keating <geoffk at redhat dot com>, gcc-bugs at gcc dot gnu dot org, stephen dot clarke at superh dot com, shumpei dot kawasaki at hsa dot hitachi dot com
- Date: Wed, 24 Jul 2002 12:11:17 +0100
- Subject: Re: Serious code size regression from 3.0.2 to now
- Organization: SuperH UK Ltd.
- References: <Pine.LNX.4.21.0207231700010.5505-100000@mail.kloo.net>
tm wrote:
> > > I can see two possible solutions for this problem:
> > >
> > > 1) Tweak BBRO heuristics to prevent out-of-lining short code sequences
> > > of less than n insns
> >
Geoff Keating:
> > That would be a significant de-optimisation.
tm wrote:
> Not necessarily. On the SH4 we can branch over single instructions
> with no penalty because it conditionally executes the one instruction on
> the inverse of the branch condition.
More to the point, putting short sequences out-of-line is a de-optimization
even for the path that does not use them, because it needs a long branch - i.e.
actually multiple instructions and a data entry in the text segment - to
reach these out-of-line instructions. So we even blow up the size of the
likely path, and cause Icache / Dcache conflicts.
--
--------------------------
SuperH
2430 Aztec West / Almondsbury / BRISTOL / BS32 4AQ
T:+44 1454 462330