On 10/18/2013 05:48 PM, Matthew Wilcox wrote:
On Fri, Oct 18, 2013 at 03:14:19PM +0200, Matias Bjorling wrote:
Performance study:
System: HGST Research NVMe prototype, Haswell i7-4770 3.4Ghz, 32GB 1333Mhz
I don't have one of these. Can you provide more details about it,
such as:
- How many I/O queues does it have?
- Are the interrupts correctly bound to the CPUs (for both sets of tests)?
The driver isn't allowed to set the irq affinity correctly itself;
it can only set a hint and userspace (eg irqbalance) has to set the
affinity correctly
- Does it support interrupt coalescing? If so, have you managed to do any
testing using that feature?
Current implementation is a driven using a PCI-e Gen2x4 interface, two
I/O queues. One for admin and one for all other I/Os. I'll check up on
IRQ affinity when multiple queues are available. I haven't yet tested
interrupt coalescing.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
