This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch for automaton based pipeline hazard recognizer (part #1).
Bernd Schmidt wrote:
>
> On Tue, 20 Feb 2001, Vladimir Makarov wrote:
> >
> > As for minimal cost, the current scheduler doesn't take into account
> > that there are processors where insn latency is zero (of course
> > logically). We could make that the scheduler using old pipeline
> > interface also could be adequate to such processors.
>
> What are the problems with that? I just ran into a case where priority()
> would take exponential time, but other than that I don't see any problems.
> The ia64 machine description has a few insns with zero latency.
>
Insn cost and priority equal to zero means undefined (not evaluated yet)
values for old pipeline interface. So scheduler may evaluate them again
and again for zero latency time insns. No big deal. But it is better
do not mix undefined and real cost values at least for maintaining the
scheduler.
Vlad