This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: reload, regcost problems [Was: Re: Math problems on PPC with gcc-2.96 20000928]
- To: Franz dot Sirl-kernel at lauterbach dot com
- Subject: Re: reload, regcost problems [Was: Re: Math problems on PPC with gcc-2.96 20000928]
- From: Geoff Keating <geoffk at cygnus dot com>
- Date: Sun, 1 Oct 2000 14:47:11 -0700
- CC: gcc-bugs at gcc dot gnu dot org
- References: <00093015550500.15331@enzo.bigblue.local> <200009302124.OAA11267@geoffk.org> <00100123123701.05125@enzo.bigblue.local>
> From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>
> Date: Sun, 1 Oct 2000 23:12:37 +0200
> Cc: gcc-bugs@gcc.gnu.org
> Any solution to this? And you are already sure reload is the culprit? I
> didn't get as far as you, but for me it seemed the -fpic was confusing the
> scheduler, see my other bug report "-fpic confuses scheduler" with the
> testcase.
From your mail, it looked like you had just switched off
-fschedule-insns and the bug went away and you concluded that the
scheduler is the culprit. That would be insufficient evidence on
which to base your conclusion.
The real way to find such a thing is to check that the RTL before
reload is OK, and the RTL after reload is bad, which is what I did.
I don't have a solution yet. I'm busy doing other stuff right now...
--
- Geoffrey Keating <geoffk@cygnus.com>