* dead code in io_file_put()
@ 2019-04-13 17:27 Al Viro
2019-04-13 17:49 ` Jens Axboe
0 siblings, 1 reply; 2+ messages in thread
From: Al Viro @ 2019-04-13 17:27 UTC (permalink / raw)
To: Jens Axboe; +Cc: linux-fsdevel
It's never called with NULL first argument anymore.
Is there any reason to keep that check (and the second argument,
never used when state != NULL) there?
Frankly, I wonder if there is any point in that
fget_many()/fput_many() stuff. Does the benefit from that
really show up in any profiles?
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: dead code in io_file_put()
2019-04-13 17:27 dead code in io_file_put() Al Viro
@ 2019-04-13 17:49 ` Jens Axboe
0 siblings, 0 replies; 2+ messages in thread
From: Jens Axboe @ 2019-04-13 17:49 UTC (permalink / raw)
To: Al Viro; +Cc: linux-fsdevel
On 4/13/19 11:27 AM, Al Viro wrote:
> It's never called with NULL first argument anymore.
> Is there any reason to keep that check (and the second argument,
> never used when state != NULL) there?
Good catch, looks like that part can go now.
> Frankly, I wonder if there is any point in that
> fget_many()/fput_many() stuff. Does the benefit from that
> really show up in any profiles?
It certainly made a difference for higher IOPS workloads. Of course
we have the file registration now to negate that, but since we're
using the state anyway for iocb allocation, it'd be a shame to drop
the file ref cost amortization for the non-registered case.
--
Jens Axboe
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2019-04-13 17:50 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-04-13 17:27 dead code in io_file_put() Al Viro
2019-04-13 17:49 ` Jens Axboe
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.