Re: 3.2GHz processor weirdness

"J.P. Delport" <[email protected]> Tue, 01 Mar 2005 15:16:32 +0200
Newsgroups gmane.linux.real-time.rtlinux.general
Message-ID <[email protected]>
Hi,

Masahiro ABE wrote:
> Hello,
> 
> On Tue, 01 Mar 2005 12:23:22 +0200
> "J.P. Delport" <[email protected]> wrote:
> 
> 
>>Hello,
>>
>>I've recently tried to use rtlinux-3.2-rc1 from cvs on a new machine. 
>>The machine has an Intel P4 processor on an Abit IC7 M/B (875P chipset).
>>
>>I compiled a vanilla 2.4.27 kernel, patched with bigphysarea and rtlinux 
>>under Debian, gcc-2.95.4.
>>
>>The 'make regression' fails the 'thread wait times' and 'Linux time is 
>>monotonically increasing' tests before it just idles inside the floating 
>>point test.
>>
>>Also when I run 'scripts/insrtl' the modules are inserted, but the PC 
>>starts behaving funny: when I type in X, a single character press is 
>>sometimes shown multiple times, as if the keyboard got stuck. When the 
>>system bell sounds, it does not stop again.
>>
>>When I change the processor speed in the BIOS to 1.6GHz (100*16 instead 
>>of 200*16), the regression tests all pass and I do not get the funny 
>>keyboard and speaker.
>>
>>Does anyone know what can cause this behaviour? Also, what is the 
>>preferred setting for APIC in both the BIOS and the kernel config?
>>
>>regards
>>jp
> 
> 
> You may take a look at my old post here:
> <http://hq.fsmlabs.com/pipermail/rtl/2004-March/013463.html>
> 
> I've seen schedulers/i386/rtl_time.c in 3.2-rc1 CVS, and "NLOOPS" is
> still 50, which is too high for 3.0+GHz CPU, I think (or, this problem
> is addressed/fixed at other place?).
> 
> =================================
> Masahiro ABE, A&D Co., Ltd. Japan

Like you suggested, I changed the NLOOPS define in
rtlinux/schedulers/i386/rtl_time.c
from 50 to 25.

After a recompile, the regression tests now complete correctly and the 
keyboard and bell weirdness have disappeared.

Thank you very much.

regards
jp


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
MailScanner thanks transtec Computers for their support.

_______________________________________________
Rtl mailing list
[email protected]
http://hq.fsmlabs.com/mailman/listinfo/rtl