[sel-sched] Fix substitution bugs, calling of reorder hooks, and bump alignment

Tehila Meyzels TEHILA@il.ibm.com
Mon Mar 5 13:30:00 GMT 2007


Andrey Belevantsev <abel@ispras.ru> wrote on 28/02/2007 20:24:23:
>Hello,
>
>These are four patches that were sitting in our local tree for a while.
I've decided to write about them in one >email, though they were >applied
in different commits.
>
>The first patch dumps the info about scheduled insns in the Haifa
scheduler the same way as in the selective >scheduling, making debug
easier. >The second patch bumps the default alignment of functions and
loops in the ia64 >backend for the same purpose, so we don't see
performance >losses due to the bad alignment.
>
>The third patch fixes the substitution transformation by disabling the
optimization in unsubstitution that works >incorrectly.
>
>The fourth patch is the fixes that were needed to get the compiler to
bootstrap on powerpc. With this patch, we >now call reorder* and
>variable_issue hooks properly; a few thinkos are also fixed.
>
>Bootstrapped and regtested on ia64. The last patch is also bootstrapped
and regtested on powerpc-linux with >--enable-languages=c (the only
>powerpc I have access to is powerpc750, which is not terribly fast.)

I've successfully bootstraped this branch on powerpc970
(--enable-languages=c only. when trying to bootstrap with all languages it
got stuck).
I'm trying now with c++,fortran.

Tehila.

>
>Committed to sel-sched branch.
>
>Andrey



More information about the Gcc-patches mailing list