| Message ID | 20210812154149.1061502-6-axboe@kernel.dk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Enable bio recycling for polled IO | expand |
On Thu, Aug 12, 2021 at 09:41:48AM -0600, Jens Axboe wrote: > Mark polled IO as being safe for dipping into the bio allocation > cache, in case the targeted bio_set has it enabled. > > This brings an IOPOLL gen2 Optane QD=128 workload from ~3.0M IOPS to > ~3.3M IOPS. Looks good, Reviewed-by: Christoph Hellwig <hch@lst.de> Didn't the cover letter say 3.5M+ IOPS, though?
On 8/12/21 10:39 AM, Christoph Hellwig wrote: > On Thu, Aug 12, 2021 at 09:41:48AM -0600, Jens Axboe wrote: >> Mark polled IO as being safe for dipping into the bio allocation >> cache, in case the targeted bio_set has it enabled. >> >> This brings an IOPOLL gen2 Optane QD=128 workload from ~3.0M IOPS to >> ~3.3M IOPS. > > Looks good, > > Reviewed-by: Christoph Hellwig <hch@lst.de> > > Didn't the cover letter say 3.5M+ IOPS, though? It does indeed, we've had some recent improvements so the range is now more in the 3.2M -> 3.5M IOPS with the cache. Didn't update it, as it's still roughly the same 10% bump.
diff --git a/fs/io_uring.c b/fs/io_uring.c index 6c65c90131cb..ea387b0741b8 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -2736,7 +2736,7 @@ static int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe) !kiocb->ki_filp->f_op->iopoll) return -EOPNOTSUPP; - kiocb->ki_flags |= IOCB_HIPRI; + kiocb->ki_flags |= IOCB_HIPRI | IOCB_ALLOC_CACHE; kiocb->ki_complete = io_complete_rw_iopoll; req->iopoll_completed = 0; } else {
Mark polled IO as being safe for dipping into the bio allocation cache, in case the targeted bio_set has it enabled. This brings an IOPOLL gen2 Optane QD=128 workload from ~3.0M IOPS to ~3.3M IOPS. Signed-off-by: Jens Axboe <axboe@kernel.dk> --- fs/io_uring.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)