From mboxrd@z Thu Jan 1 00:00:00 1970 From: hch@infradead.org (hch@infradead.org) Date: Thu, 12 May 2016 00:08:05 -0700 Subject: [PATCH 0/2] Block: Give option to force io polling In-Reply-To: References: <1459455554-2794-1-git-send-email-jonathan.derrick@intel.com> <20160508090257.GG15458@infradead.org> Message-ID: <20160512070805.GE20802@infradead.org> On Mon, May 09, 2016@02:53:30PM +0000, Stephen Bates wrote: > Christoph, this is a DRAM based NVMe device, the code for polling in NVMe was merged in 4.5 right? We are using the inbox NVMe driver. Here is some performance data: > > QD=1, single thread, random 4KB reads > > Polling Off: 12us Avg / 40us 99.99% ; > Polling On: 9.5us Avg / 25us 99.99% > > Both the average and 99.99% reduction are of interest. How does CPU usage look for common workloads with polling force enabled? If it's really an overall win we should just add a quirk to the NVMe driver to always force polling for this device based on the PCI ID.