This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 3.2 vs 3.1 speed
"David S. Miller" wrote:
>
> From: "Vladimir N. Makarov" <vmakarov@redhat.com>
> Date: Wed, 05 Jun 2002 09:49:40 -0400
>
> Hi, Dan. I'll look at the 1st insn scheduling (the 2nd insn scheduling is
> actually faster) today. As David, I suspect the code is slower because
> scheduling lookahead. I think that the specific features of code is in long
> ready queues. I have some ideas how to improve such behavior.
>
> Vlad, remember yet another issue with applies to Sparc with
> DFA. The fact that we have DFA for many processor types and
> state machine must make all of these processor type checks on
> each state machine transition to see which DFA machines apply.
>
> Or did you fix this?
>
Yes, it is fixed now. Now only states of automata whose units are
used in the insn reservations are changed for given insn. So .md file
can contain as many descriptions of different processors (subtargets) as
you want without degradation in speed of pipeline hazard recognizer.
This is not issue anymore.
> It is real problem if it is not cured already. Ix86 will hit it
> too, as will MIPS when/if it gets DFA descriptions.