Re: [RFC] 2.4.0-test6-pre2 Merge softirq, local_irq_count, local_bh_count

Re: [RFC] 2.4.0-test6-pre2 Merge softirq, local_irq_count, local_bh_count

[Paul Mackerras]

Keith Owens writes:

> This patch hits every arch.  It is being cross mailed to every arch
> mailing list that has not already confirmed that the patch works.
> Please trim replies to the relevant mailing list.

Seems to work ok on uniprocessor PPC.  The 2.4 kernels have problems
on SMP PPC boxes at the moment so I haven't tested it on SMP.

Have you done any measurements, with lmbench or whatever, to measure
how much performance improvement you get with the patch?

--
Paul Mackerras, Senior Open Source Researcher, Linuxcare, Inc.
+61 2 6262 8990 tel, +61 2 6262 8991 fax
paulus@linuxcare.com.au, http://www.linuxcare.com.au/
Linuxcare.  Support for the revolution.

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

Re: [RFC] 2.4.0-test6-pre2 Merge softirq, local_irq_count, local_bh_count

[Keith Owens]

On Sat, 5 Aug 2000 17:14:36 +1000 (EST),
Paul Mackerras <paulus@linuxcare.com.au> wrote:
>Seems to work ok on uniprocessor PPC.  The 2.4 kernels have problems
>on SMP PPC boxes at the moment so I haven't tested it on SMP.

Thanks for that.  Linus rolled the patch into 2.4.0-test6-pre4.

>Have you done any measurements, with lmbench or whatever, to measure
>how much performance improvement you get with the patch?

No, it was not really driven by performance considerations.  SGI wanted
to add a syscall count for performance and fail over monitoring.  Linus
said he would accept the syscall count if we cleaned up softirq at the
same time, instead of having two statistic cache lines per cpu.


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