On Thu, Dec 01, 2016 at 07:26:39PM +0000, Stefan Hajnoczi wrote: > v4: > * Added poll time self-tuning algorithm [Christian and Paolo] > * Try a single iteration of polling to avoid non-blocking > ppoll(2)/epoll_wait(2) [Paolo] > * Reordered patches to make performance analysis easier - see below > > v3: > * Avoid ppoll(2)/epoll_wait(2) if polling succeeded [Paolo] > * Disable guest->host virtqueue notification during polling [Christian] > * Rebased on top of my virtio-blk/scsi virtqueue notification disable patches
Karl Rister found a performance regression between v2 -> v3. There was a bug in the poll_disable_cnt code in aio_set_fd_handler() which meant that polling was disabled unnecessarily. This has been fixed in v4 and I suspect it was the cause for the regression. Stefan
signature.asc
Description: PGP signature
