On Mon, Sep 03, 2012 at 11:53:39AM -0700, Yinghai Lu wrote:
> On Mon, Sep 3, 2012 at 2:17 AM, Alexander Gordeev <agordeev@xxxxxxxxxx> wrote:
> You may update create_irq_nr to be __create_irq_nr, and it could take
> extra count.
>
> and later have create_irq_nr to be __create_irq_nr(,1,)
> and create_irqs to be __create_irq_nr(,count,)
> ....
Indeed. Will do.
> BTW, in short, how much performance benefits for adding 500 lines code?
Unfortunatelly, I do not have a short answer here. There are three types
of performance this series deals with - I'll try to summarize:
- devices - 3 SATA HDDs generate roughly one interrupt every 273 us while
it get handled in less than 5 us. So there is/could be no increase here;
- the hardware context interrupt handler - its performance dropped 2.5 times
(little bit more in fact) at the expense of increase of 1.3 times in
overall interrupt handling time (hardware context + threaded context);
- overall system performance - I *assume* it should increase, because:
(a) AHCI interrupt handlers keep local interrupts disabled 2.5 times less
(b) separate AHCI IRQs become subjects of IRQ balancing
(c) threaded handlers are per-device, per-CPU (well, up to irqbalanced)
and executed with local interrupts enabled;
--
Regards,
Alexander Gordeev
agordeev@xxxxxxxxxx
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
