Re: [patch] protecting MIPS prologues from scheduling.

[cgd at sibyte dot com (Chris G. Demetriou)]

Richard Henderson <rth@cygnus.com> writes:
> On Mon, Jul 31, 2000 at 05:55:23PM -0700, Chris G. Demetriou wrote:
> > * Is the issue at the root of my coworker's problem considered a
> >   gcc/optimization issue, or a GDB issue?  Specifically, where
> >   "should" it be solved (or worked around)?
> 
> IMO the problem should be solved -- at least to the extent that
> this patch solves anything -- at the keyboard level.  If you don't
> want scheduling, don't turn on the scheduler.  I.e. use -O1 or
> -O2 -fno-schedule-insns{,2}.  You'll have just as many problems
> with function body code as you will with the prologue.

Certainly, if you're enabling scheduling, you're opening yourself up
to a certain class of problems.  This seems to be one of the less
expected or perhaps more noticeable ones (when it actually occurs).
(My coworker apparently didn't have an issue with, or didn't notice,
the effects of scheduling when 'inside' functions, but he was annoyed
by problems caused by the effects being visible at the breakpoints at
the 'start' of functions.)

Anyway, it's an easy patch for me to maintain in my local tree, and if
anybody else needs it it's in the archive.  8-)


Thanks for the quick response.



chris