This is the mail archive of the
mailing list for the GCC project.
Re: ia64 performance regression
- To: smc at us dot ibm dot com (Steve Christiansen)
- Subject: Re: ia64 performance regression
- From: Joern Rennecke <amylaar at onetel dot net dot uk>
- Date: Sun, 14 Oct 2001 23:35:42 +0100 (BST)
- Cc: gcc at gcc dot gnu dot org
> for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
> regs->array[i].may_not_optimize = 1;
> regs->array[i].set_in_loop = 1;
I would guess that this is about not extending hard register lifetimes.
This looks like something that you want to make dependent on
SMALL_REGISTER_CLASSES and CLASS_LIKELY_SPILLED_P.
There could also be issues with fixed registers.