Re: FW: floating point exception problem on apple PPC

On Tue, Aug 07, Giuliano Pochini wrote:


I'm not sure why you got that mail, but there is a bug.

We can sometimes build the current glibc and sometimes it fails with a
FPE in make check. Kernel is a 2.4.7 in that case.

We built it on a pseries 640, 4cpus, 2gig ram. It seems that it does not
occour on a single CPU machine.



Gruss Olaf

-----FW: [off-list ref]-----

Date: Tue, 7 Aug 2001 10:10:05 +0530
Sender: linux-kernel-owner@vger.kernel.org
From: mdaljeet@in.ibm.com
To: linux-kernel@vger.kernel.org
Subject: floating point exception problem on apple PPC

Hi,

I am using SuSE Linux 7.1 with kernel version 2.4.2 on Apple G4 dual
processor PPC machine.

Due to setting of MSR[FE0] and MSR[FE1] bits in the 'FPUnavailable'
exception handler, a user application that tries to do floating point
arthemetic with large floating point values gets a SIGFPE signal. By
setting the FPSCR register with appropriate value in the SIGFPE signal
handler the program runs fine with the expected values.

Now, what happens is that when I run such program (that does floating point
arthemetic with large floating point values) continuously for hours, the
other applications running on the system sometimes get terminated after
receiving a SIGFPE signal while my floating point program runs fine. Is
that due to some kernel problem or loading the FPSCR in the SIGFPE signal
handler of a process can lead to this problem?

regards,
Daljeet.



--------------End of forwarded message-------------------------

Bye.

--



--
 $ man clone

BUGS
       Main feature not yet implemented...

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/