Re: Performance issue in 2.5.32+

From: James Morris (jmorris@intercode.com.au)
Date: Mon Sep 09 2002 - 11:06:04 EST


On Sun, 8 Sep 2002, Linus Torvalds wrote:

> I don't really see anythign very suspicious in the 31->32 changelogs.
> Somebody else already mentioned TLS etc, but that shouldn't really be
> noticeable...

I've now tracked the problem to the introduction of the new mtrr code, and
backing it out for 2.5.32 restores original performance. From what I can
tell, the new code is not handling a bios/hardware bug that the old code
was.

The old mtrr code says this during bootup:
 mtrr: your CPUs had inconsistent fixed MTRR settings
 mtrr: your CPUs had inconsistent variable MTRR settings
 mtrr: probably your BIOS does not setup all CPUs

While the new code just says:
 mtrr: SMP support incomplete for this vendor

The peformance hit is basically the same as when the old mtrr code is
disabled. I'll have a poke around and see if I can find out why the new
mtrr code isn't handling the problem.

- James

-- 
James Morris
<jmorris@intercode.com.au>

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sun Sep 15 2002 - 22:00:17 EST