This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Re[2]: [GSoC: DDG export][RFC] Current status
- From: Zdenek Dvorak <rakdver at kam dot mff dot cuni dot cz>
- To: Alexander Monakov <amonakov at gmail dot com>
- Cc: gcc at gcc dot gnu dot org, Ian Lance Taylor <iant at google dot com>, Dmitry Melnik <dm at ispras dot ru>, Daniel Berlin <dannyb at google dot com>, Andrey Belevantsev <abel at ispras dot ru>, Ayal Zaks <ZAKS at il dot ibm dot com>, Sebastian Pop <sebpop at gmail dot com>
- Date: Tue, 4 Sep 2007 17:23:34 +0200
- Subject: Re: Re[2]: [GSoC: DDG export][RFC] Current status
- References: <3495f40d0707021147j4e8ee8a3ve1a3f859738cd589@mail.gmail.com> <m3oditftx5.fsf@localhost.localdomain> <1069995120.20070704183906@ispras.ru> <3495f40d0707131141o2357d5f5v81b26c4edfb86573@mail.gmail.com> <20070715200413.GC25014@kam.mff.cuni.cz> <3495f40d0708201108l76d947afo967fec5a584bb2f2@mail.gmail.com>
Hello,
> An important missing piece is correction of exported information for
> loop unrolling. As far as I can tell, for loop unrolled by factor N we
> need to clone MEM_ORIG_EXPRs and datarefs for newly-created MEMs, create
> no-dependence DDRs for those pairs, for which original DDR was
> no-dependence, and for DDRs with recorded distance vectors we will be
> able to recompute new distances from old distance(D) and N (but only if
> N % D == 0 || D % N == 0). Is that right, Zdenek?
yes, that looks correct to me.
>Where should I start to implement this?
In loop-unroll.c, look for apply_opt_in_copies; that is used to
post-process the unrolled loop body.
Zdenek