This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH][modulo-sched] Fix generation of reg-moves (was: Fix generation of prolog and epilog)
- From: Ayal Zaks <ZAKS at il dot ibm dot com>
- To: Revital1 Eres <ERES at il dot ibm dot com>
- Cc: "Andrey Belevantsev" <abel at ispras dot ru>, Alexander Monakov <amonakov at ispras dot ru>, gcc-patches at gcc dot gnu dot org
- Date: Tue, 18 Dec 2007 21:24:33 +0200
- Subject: Re: [PATCH][modulo-sched] Fix generation of reg-moves (was: Fix generation of prolog and epilog)
Revital1 Eres/Haifa/IBM wrote on 18/12/2007 13:49:14:
> Following our off-line discussion regarding the implementation of this
> patch; I re-submit it with the following change: the reg-moves of the
> kernel are inserted right before the notes which precede the insn they
> relates to. This will also prevent the problematic case were redundant
> reg-moves are generated while creating the prolog/epilog as reported
> I am now re-testing the patch on powerpc64-linux, x86_64-linux and SPU.
> OK for mainline once testing completes?
Yes, this wrong-code fix is ok.
In general it's better to keep dump info additions separate from the actual
bug fixing line(s).
> 2007-12-17 Alexander Monakov <firstname.lastname@example.org>
> Revital Eres <email@example.com>
> * modulo-sched.c (generate_reg_moves): Insert the reg-moves
> right before the notes which precedes the insn, if they exists.
> (loop_canon_p): Add dump info.
> (sms_schedule): Likewise.
> * gcc.dg/sms-5.c: New testcase.
> [attachment "sms-5.c.txt" deleted by Ayal Zaks/Haifa/IBM] [attachment
> "patch_fix_regmoves_17.txt" deleted by Ayal Zaks/Haifa/IBM]