This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: tree-ssa performance
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Richard dot Earnshaw at arm dot com
- Cc: Steven Bosscher <s dot bosscher at student dot tudelft dot nl>, Andrew Macleod <amacleod at redhat dot com>, Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>, Andrew Haley <aph at redhat dot com>, Jeff Sturm <jsturm at one-point dot com>, Tim Josling <tej at melbpc dot org dot au>, Zack Weinberg <zack at codesourcery dot com>, gcc mailing list <gcc at gcc dot gnu dot org>, Neil Booth <neil at daikokuya dot co dot uk>
- Date: 10 Jun 2003 13:45:13 -0400
- Subject: Re: tree-ssa performance
- Organization: Red Hat Canada
- References: <200306101736.h5AHaSr05055@pc960.cambridge.arm.com>
On Tue, 2003-06-10 at 13:36, Richard Earnshaw wrote:
> > Eh, the SPEC95 machine has 256Mb of RAM. That's not huge for today's
> > standards. Bootstrap times for that machine are in line with what I get
> > on the bigger machine (~10% slower than mainline).
> >
> > I also regularly bootstrap on a PPC with 128Mb of RAM. I'll try
> > bootstrapping mainline on that machine to see what the difference is.
>
> Which is still 4 times more memory than one of the machine I have
> available for bootstrapping the ARM compiler.
>
> This squandering of memory is getting silly IMO. There's no wonder gcc
> has a reputation for being slow.
>
I agree 100%. I was only addressing the mainline vs tree-ssa problem.
GCC is memory-hungry and I think it will get worse before it gets
better. Some of the heavy optimization passes we have in mind will use
all the memory you can throw at them.
That's not to say that we shouldn't try to limit memory usage either by
adding more smarts into the optimizers or by having the memory intensive
passes triggered at -O3 or something along those lines.
Diego.