Roland Dreier wrote: > > Yes, but there are MSI devices in the field... > > But do any of them benefit from having multiple interrupt vectors available? > > If I recall correct, willy was motivated by AHCI performance to try this out, > and in the end things didn't go any faster using multiple MSI. > > The big win with multiple interrupt vectors is when you have independent > queues that you can point different CPUs at, and give each CPU its own > interrupt. And this is how AHCI is designed, but libahci.c currently wraps its interrupt handler in a device-wide lock. I don't know if that test also reorganized the driver code. Regards, Clemens -- 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
