linux-ide.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Re: multiple irqs per card on the sata_sil24
@ 2006-09-27  0:31 Trent George
  2006-09-27  0:46 ` Jeff Garzik
  2006-09-27  1:07 ` Tejun Heo
  0 siblings, 2 replies; 4+ messages in thread
From: Trent George @ 2006-09-27  0:31 UTC (permalink / raw)
  To: Tejun Heo; +Cc: linux-ide


While reading the specs, I noticed that the pci-x cards support splitting each port to a unique irq pin.
If you have 6 cards with 4 ports, can you spread the load over 24 irqs ?
I dont know if the motherboard can support this, or the driver could be modified to do this.
It would be nice when there is a lot of of disk traffic on lots of drives.... adding port multiplier drives to each channel.
I am trying to build a VERY large duplication system using LOTS of bandwidth. the irq can really put stress on the system.

the final system would be 100+ drives all going over 22mb/s.
It should give a nice supermicro board with 3 buses a real workout.

Trent


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: multiple irqs per card on the sata_sil24
  2006-09-27  0:31 multiple irqs per card on the sata_sil24 Trent George
@ 2006-09-27  0:46 ` Jeff Garzik
  2006-09-27  1:07 ` Tejun Heo
  1 sibling, 0 replies; 4+ messages in thread
From: Jeff Garzik @ 2006-09-27  0:46 UTC (permalink / raw)
  To: Trent George; +Cc: Tejun Heo, linux-ide

Trent George wrote:
> While reading the specs, I noticed that the pci-x cards support splitting each port to a unique irq pin.
> If you have 6 cards with 4 ports, can you spread the load over 24 irqs ?
> I dont know if the motherboard can support this, or the driver could be modified to do this.
> It would be nice when there is a lot of of disk traffic on lots of drives.... adding port multiplier drives to each channel.
> I am trying to build a VERY large duplication system using LOTS of bandwidth. the irq can really put stress on the system.
> 
> the final system would be 100+ drives all going over 22mb/s.
> It should give a nice supermicro board with 3 buses a real workout.

You're still going to max out PCI bus bandwidth first.  Best solution is 
PCI Express with MSI...

	Jeff




^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: multiple irqs per card on the sata_sil24
  2006-09-27  0:31 multiple irqs per card on the sata_sil24 Trent George
  2006-09-27  0:46 ` Jeff Garzik
@ 2006-09-27  1:07 ` Tejun Heo
  2006-09-27  1:09   ` Tejun Heo
  1 sibling, 1 reply; 4+ messages in thread
From: Tejun Heo @ 2006-09-27  1:07 UTC (permalink / raw)
  To: Trent George; +Cc: linux-ide

Trent George wrote:
> While reading the specs, I noticed that the pci-x cards support splitting each port to a unique irq pin.

3124 doesn't seem to support multiple MSI interrupts either.  Multiple 
Message Capable field is 000b.  Where does the spec say that it can 
split each port to a unique irq pin?

> If you have 6 cards with 4 ports, can you spread the load over 24 irqs ?
> I dont know if the motherboard can support this, or the driver could be modified to do this.
> It would be nice when there is a lot of of disk traffic on lots of drives.... adding port multiplier drives to each channel.
> I am trying to build a VERY large duplication system using LOTS of bandwidth. the irq can really put stress on the system.

Theoretically, yes, but linux currently cannot do multiple MSI 
interrupts and I don't think it's gonna be supported anytime soon. 
pci-e + multiple MSIX interrupts are supported but AFAIK 3132 supports that.

sil24 is pretty efficient and it only takes one more register access to 
determine which port has raised interrupt.  Also, if you have multiple 
busy ports w/ PMP attached, single interrupt is likely to serve multiple 
ports, so I don't think it's gonna matter performance-wise.

> the final system would be 100+ drives all going over 22mb/s.
> It should give a nice supermicro board with 3 buses a real workout.

Impressive.  Good luck and please keep us posted.

-- 
tejun

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: multiple irqs per card on the sata_sil24
  2006-09-27  1:07 ` Tejun Heo
@ 2006-09-27  1:09   ` Tejun Heo
  0 siblings, 0 replies; 4+ messages in thread
From: Tejun Heo @ 2006-09-27  1:09 UTC (permalink / raw)
  To: Tejun Heo; +Cc: Trent George, linux-ide

Tejun Heo wrote:
> Trent George wrote:
>> While reading the specs, I noticed that the pci-x cards support 
>> splitting each port to a unique irq pin.
> 
> 3124 doesn't seem to support multiple MSI interrupts either.  Multiple 

s/either//

> Theoretically, yes, but linux currently cannot do multiple MSI 
> interrupts and I don't think it's gonna be supported anytime soon. pci-e 
> + multiple MSIX interrupts are supported but AFAIK 3132 supports that.

s/supports/doesn't support/

Sorry.

-- 
tejun

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2006-09-27  1:09 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-09-27  0:31 multiple irqs per card on the sata_sil24 Trent George
2006-09-27  0:46 ` Jeff Garzik
2006-09-27  1:07 ` Tejun Heo
2006-09-27  1:09   ` Tejun Heo

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).