[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <9051.1157971604@warthog.cambridge.redhat.com>
Date: Mon, 11 Sep 2006 11:46:44 +0100
From: David Howells <dhowells@...hat.com>
To: Ingo Molnar <mingo@...e.hu>
Cc: David Howells <dhowells@...hat.com>,
john stultz <johnstul@...ibm.com>,
Adrian Bunk <bunk@...sta.de>, Andrew Morton <akpm@...l.org>,
Arjan van de Ven <arjan@...ux.intel.com>,
linux-kernel@...r.kernel.org, Jeff Garzik <jeff@...zik.org>,
netdev@...r.kernel.org, Thomas Gleixner <tglx@...utronix.de>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>
Subject: Re: [PATCH] FRV: do_gettimeofday() should no longer use tickadj
Ingo Molnar <mingo@...e.hu> wrote:
> i cannot find Thomas' recent 2.6 one (Thomas, do you have a link to
> it?), but i did one 5 years ago:
>
> http://people.redhat.com/mingo/irq-rewrite-patches/irq-cleanup-2.4.15-B1.bz2
>
> in general it's a large but otherwise pretty dumb patch.
I wrote my own patch to test this last Friday. I found that removing all the
regs pointer passing from the interrupt code reduced interrupt entry with a
warm cache by 1 cpu cycle out of 87, and interrupt exit by 19 cycles out of 99.
I can't tell from that exactly how many instructions/memory accesses have been
removed since the FRV permits two instructions to be executed in one cycle
under some circumstances, and two registers to be stored/loaded in one
instruction.
But the main gain in the exit path has to be due to recovery of the clobbered
regs parameter due to a call inside a loop, possibly in handle_IRQ_event().
I'd expect i386 to do better in cycle reduction because it has fewer registers
and so getting one back should gain more.
David
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists