On Wed, Feb 25, 2009 at 03:12:33PM -0700, Moore, Eric wrote: > On Wednesday, February 25, 2009 2:12 PM, Matthew Wilcox wrote: > > It's also worth considering a mode where each CPU gets its own > > interrupt. That lets us complete IOs on the CPU which submitted them > > and can be a real performance win. > > right, I was wondering if we could do that. Perhaps we could discuss that further in how do to that. It's something that's being worked on. I'm not quite sure where we are on it with the SCSI side. From the PCI side, we're at a point where you can request however many vectors you want and then assign them however you want. I think we need a higher-level API than that, but we're not quite sure what it should look like yet (and I want to redesign the lower-level API anyway, because it's horrible for drivers to use). > > I'm somewhat puzzled that you request four MSI-X interrupts, then only > > use one of them. Why not just request one? > > I wasn't sure if we pci_enable_msix would work when passed 1 in the 2nd parameter when we have 15 vectors available. Yes, that works fine. -- Matthew Wilcox Intel Open Source Technology Centre "Bill, look, we understand that you're interested in selling us this operating system, but compare it to ours. We can't possibly take such a retrograde step." -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
