Subject: Re: GCC -O6
To: None <Guenther.Grau@bosch.com>
From: John Darrow <email@example.com>
Date: 09/09/1999 16:42:50
[Cc'ing to tech-toolchain as this is an egcs issue...]
Guenther Grau wrote:
>Adam Ciarcinski wrote:
>> Should I send this as a bug report via send-pr?
>> Is there any danger with using -O6 or optimizations greater than -O2?
>Yes. People have reported in the past that gcc still contains
>code generation bugs with -Ox , x > 2, or maybe even x >= 2
>on some architectures. Maybe we can spot all the problems
>in our code and at least document them so that the gcc-maintainters
>have a chance to fix them (if we don't have any gcc-savvy
>people in our team).
Yes. I've just submitted a bug report (I thought I had done so long ago,
but it was only in a followup message to another optimizer bug report)
regarding an optimizer problem on i386 which is tickled by rwho. Compiling
with -O2 results in erroneous printouts from rwho -a, while compiling with
-O is okay. This problem first appeared in the original egcs integration,
and has lasted through all the egcs imports and patches since then.
The pr is port-i386/8364. I don't know if any other arches are affected,
John Darrow - Senior Technical Specialist Office: 630/752-5201
Computing Services, Wheaton College, Wheaton, IL 60187 Fax: 630/752-5968
Alphapage: firstname.lastname@example.org Pager: 630/316-0707