All of lore.kernel.org
 help / color / mirror / Atom feed
* Increasing startup time
@ 2015-11-06  3:08 Matthew Eaton
  2015-11-08  0:29 ` Jens Axboe
  0 siblings, 1 reply; 5+ messages in thread
From: Matthew Eaton @ 2015-11-06  3:08 UTC (permalink / raw)
  To: fio

I noticed during testing recent fio releases it seemed bit laggy to
start, so I tried some tests and it seems fio has been increasingly
taking longer to start for some reason.  Don't know if this is
expected or if something is wrong?

These are some tests I ran on a few different release versions...

matt@matt-desktop:~$ time fio-2.2.3/fio -v
fio-2.2.3

real    0m0.008s
user    0m0.004s
sys    0m0.004s
matt@matt-desktop:~$ time fio-2.2.3/fio -v
fio-2.2.3

real    0m0.009s
user    0m0.009s
sys    0m0.000s
matt@matt-desktop:~$ time fio-2.2.4/fio -v
fio-2.2.4

real    0m0.009s
user    0m0.000s
sys    0m0.008s
matt@matt-desktop:~$ time fio-2.2.4/fio -v
fio-2.2.4

real    0m0.009s
user    0m0.009s
sys    0m0.000s
matt@matt-desktop:~$ time fio-2.2.5/fio -v
fio-2.2.5

real    0m0.015s
user    0m0.011s
sys    0m0.004s
matt@matt-desktop:~$ time fio-2.2.5/fio -v
fio-2.2.5

real    0m0.014s
user    0m0.003s
sys    0m0.010s
matt@matt-desktop:~$ time fio-2.2.6/fio -v
fio-2.2.6

real    0m0.015s
user    0m0.004s
sys    0m0.011s
matt@matt-desktop:~$ time fio-2.2.6/fio -v
fio-2.2.6

real    0m0.015s
user    0m0.000s
sys    0m0.015s
matt@matt-desktop:~$ time fio-2.2.7/fio -v
fio-2.2.7

real    0m0.083s
user    0m0.015s
sys    0m0.066s
matt@matt-desktop:~$ time fio-2.2.7/fio -v
fio-2.2.7

real    0m0.081s
user    0m0.013s
sys    0m0.067s
matt@matt-desktop:~$ time fio-2.2.11/fio -v
fio-2.2.11

real    0m0.079s
user    0m0.020s
sys    0m0.059s
matt@matt-desktop:~$ time fio-2.2.11/fio -v
fio-2.2.11

real    0m0.079s
user    0m0.000s
sys    0m0.078s

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: Increasing startup time
  2015-11-06  3:08 Increasing startup time Matthew Eaton
@ 2015-11-08  0:29 ` Jens Axboe
  2015-11-08  0:34   ` Jens Axboe
  0 siblings, 1 reply; 5+ messages in thread
From: Jens Axboe @ 2015-11-08  0:29 UTC (permalink / raw)
  To: Matthew Eaton; +Cc: fio

On Thu, Nov 05 2015, Matthew Eaton wrote:
> I noticed during testing recent fio releases it seemed bit laggy to
> start, so I tried some tests and it seems fio has been increasingly
> taking longer to start for some reason.  Don't know if this is
> expected or if something is wrong?

Try the below.


diff --git a/smalloc.c b/smalloc.c
index 8412e7518464..ce64d6b73df6 100644
--- a/smalloc.c
+++ b/smalloc.c
@@ -213,7 +213,6 @@ static int add_pool(struct pool *pool, unsigned int alloc_size)
 	if (ptr == MAP_FAILED)
 		goto out_fail;
 
-	memset(ptr, 0, alloc_size);
 	pool->map = ptr;
 	pool->bitmap = (void *) ptr + (pool->nr_blocks * SMALLOC_BPL);
 

-- 
Jens Axboe



^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: Increasing startup time
  2015-11-08  0:29 ` Jens Axboe
@ 2015-11-08  0:34   ` Jens Axboe
  2015-11-09 17:33     ` Matthew Eaton
  0 siblings, 1 reply; 5+ messages in thread
From: Jens Axboe @ 2015-11-08  0:34 UTC (permalink / raw)
  To: Matthew Eaton; +Cc: fio

On Sat, Nov 07 2015, Jens Axboe wrote:
> On Thu, Nov 05 2015, Matthew Eaton wrote:
> > I noticed during testing recent fio releases it seemed bit laggy to
> > start, so I tried some tests and it seems fio has been increasingly
> > taking longer to start for some reason.  Don't know if this is
> > expected or if something is wrong?
> 
> Try the below.

Committed, with a fixup to still clear the actual bitmap. Should work
faster for you than even the earlier versions you tested.


commit 9c3e13e3314da394698ca32f21cc46d46b7cfe47
Author: Jens Axboe <axboe@fb.com>
Date:   Sat Nov 7 17:33:38 2015 -0700

    smalloc: only clear the bitmap, not the whole pool
    
    This reduces startup time, and the memset() isn't useful nor
    needed on the full range.
    
    Signed-off-by: Jens Axboe <axboe@fb.com>

diff --git a/smalloc.c b/smalloc.c
index 8412e7518464..5047cda5e343 100644
--- a/smalloc.c
+++ b/smalloc.c
@@ -213,9 +213,9 @@ static int add_pool(struct pool *pool, unsigned int alloc_size)
 	if (ptr == MAP_FAILED)
 		goto out_fail;
 
-	memset(ptr, 0, alloc_size);
 	pool->map = ptr;
 	pool->bitmap = (void *) ptr + (pool->nr_blocks * SMALLOC_BPL);
+	memset(pool->bitmap, 0, bitmap_blocks * sizeof(unsigned int));
 
 	pool->lock = fio_mutex_init(FIO_MUTEX_UNLOCKED);
 	if (!pool->lock)

-- 
Jens Axboe



^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: Increasing startup time
  2015-11-08  0:34   ` Jens Axboe
@ 2015-11-09 17:33     ` Matthew Eaton
  2015-11-09 17:40       ` Jens Axboe
  0 siblings, 1 reply; 5+ messages in thread
From: Matthew Eaton @ 2015-11-09 17:33 UTC (permalink / raw)
  To: Jens Axboe; +Cc: fio

> Committed, with a fixup to still clear the actual bitmap. Should work
> faster for you than even the earlier versions you tested.


Thank you for looking at this Jens!  You are correct, startup times
are even faster now.

matt@matt-work:~$ time fio -v
fio-2.1.14

real    0m0.008s
user    0m0.000s
sys    0m0.004s
matt@matt-work:~$ time fio -v
fio-2.1.14

real    0m0.008s
user    0m0.004s
sys    0m0.000s
matt@matt-work:~$ time fio -v
fio-2.1.14

real    0m0.008s
user    0m0.000s
sys    0m0.004s
matt@matt-work:~$ time fio/fio -v
fio-2.2.11-3-g9c3e

real    0m0.003s
user    0m0.000s
sys    0m0.000s
matt@matt-work:~$ time fio/fio -v
fio-2.2.11-3-g9c3e

real    0m0.003s
user    0m0.000s
sys    0m0.000s
matt@matt-work:~$ time fio/fio -v
fio-2.2.11-3-g9c3e

real    0m0.003s
user    0m0.004s
sys    0m0.000s


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: Increasing startup time
  2015-11-09 17:33     ` Matthew Eaton
@ 2015-11-09 17:40       ` Jens Axboe
  0 siblings, 0 replies; 5+ messages in thread
From: Jens Axboe @ 2015-11-09 17:40 UTC (permalink / raw)
  To: Matthew Eaton; +Cc: fio

On 11/09/2015 10:33 AM, Matthew Eaton wrote:
>> Committed, with a fixup to still clear the actual bitmap. Should work
>> faster for you than even the earlier versions you tested.
>
>
> Thank you for looking at this Jens!  You are correct, startup times
> are even faster now.

Great, thanks for (re)testing.

-- 
Jens Axboe



^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2015-11-09 17:40 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-11-06  3:08 Increasing startup time Matthew Eaton
2015-11-08  0:29 ` Jens Axboe
2015-11-08  0:34   ` Jens Axboe
2015-11-09 17:33     ` Matthew Eaton
2015-11-09 17:40       ` 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.