Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-19 Thread Alan Cox
On 03/18/2015 12:58, Mateusz Guzik wrote: On Wed, Mar 18, 2015 at 10:17:22AM -0400, John Baldwin wrote: On Friday, March 13, 2015 06:32:03 AM Mateusz Guzik wrote: On Thu, Mar 12, 2015 at 06:13:00PM -0500, Alan Cox wrote: Below is partial results from a profile of a parallel (-j7) buildworld on

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-19 Thread Bryan Drewery
On 3/13/2015 10:23 AM, Ryan Stone wrote: It's almost 5% on the 32 core machine: This likely will harm package building. -- Regards, Bryan Drewery signature.asc Description: OpenPGP digital signature

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-18 Thread John Baldwin
On Friday, March 13, 2015 06:32:03 AM Mateusz Guzik wrote: On Thu, Mar 12, 2015 at 06:13:00PM -0500, Alan Cox wrote: Below is partial results from a profile of a parallel (-j7) buildworld on a 6-core machine that I did after the introduction of pmap_advise, so this is not a new profile.

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-18 Thread Mateusz Guzik
On Wed, Mar 18, 2015 at 10:17:22AM -0400, John Baldwin wrote: On Friday, March 13, 2015 06:32:03 AM Mateusz Guzik wrote: On Thu, Mar 12, 2015 at 06:13:00PM -0500, Alan Cox wrote: Below is partial results from a profile of a parallel (-j7) buildworld on a 6-core machine that I did

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-18 Thread Adrian Chadd
[snip] Hihi! Do you have a shell script or something that I can run on the power8 box to see if nathan's pmap locking changes eliminate at least that global pmap lock we're seeing on amd64? -a ___ freebsd-current@freebsd.org mailing list

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-17 Thread Mateusz Guzik
On Fri, Mar 13, 2015 at 11:23:06AM -0400, Ryan Stone wrote: On Thu, Mar 12, 2015 at 1:36 PM, Mateusz Guzik mjgu...@gmail.com wrote: Workloads like buildworld and the like (i.e. a lot of forks + execs) run into very severe contention in vm, which is orders of magnitude bigger than anything

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-13 Thread Adrian Chadd
[snip] someone emailed me privately - no tracking/priority lending is happening for readers. :( -a ___ freebsd-current@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-13 Thread Ryan Stone
On Thu, Mar 12, 2015 at 1:36 PM, Mateusz Guzik mjgu...@gmail.com wrote: Workloads like buildworld and the like (i.e. a lot of forks + execs) run into very severe contention in vm, which is orders of magnitude bigger than anything else. As such your result seems quite suspicious. You're

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-13 Thread Adrian Chadd
Again, why's it not loaning priority to the lock-owning thread when it's blocked? I thought that's what is supposed to happen. -adrian ___ freebsd-current@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Mateusz Guzik
On Thu, Mar 12, 2015 at 06:13:00PM -0500, Alan Cox wrote: Below is partial results from a profile of a parallel (-j7) buildworld on a 6-core machine that I did after the introduction of pmap_advise, so this is not a new profile. The results are sorted by total waiting time and only the top 20

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Adrian Chadd
Do you have access to any boxes that have more than 12 cores? (like 36, 64, 80+ ?) -adrian On 12 March 2015 at 08:14, Ryan Stone ryst...@gmail.com wrote: I've just submitted a patch to Differential[1] for review that converts the VFS cache to use an rmlock in place of the current rwlock.

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Adrian Chadd
Also, to ask a stupid question - why wasn't the reader gifted a temporary priority boost because you were trying to acquire the write lock? -adrian ___ freebsd-current@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current

[PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Ryan Stone
I've just submitted a patch to Differential[1] for review that converts the VFS cache to use an rmlock in place of the current rwlock. My main motivation for the change is to fix a priority inversion problem that I saw recently. A real-time priority thread attempted to acquire a write lock on

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Ryan Stone
On Thu, Mar 12, 2015 at 12:37 PM, Adrian Chadd adr...@freebsd.org wrote: Do you have access to any boxes that have more than 12 cores? I have a 14-core hyperthreaded machine (so 28 logical cores), but it has no disk (long story). I could do a build out of a memory disk though. Also, to ask a

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Mateusz Guzik
On Thu, Mar 12, 2015 at 11:14:42AM -0400, Ryan Stone wrote: I've just submitted a patch to Differential[1] for review that converts the VFS cache to use an rmlock in place of the current rwlock. My main motivation for the change is to fix a priority inversion problem that I saw recently. A

Re: [PATCH] Convert the VFS cache lock to an rmlock

2015-03-12 Thread Alan Cox
Below is partial results from a profile of a parallel (-j7) buildworld on a 6-core machine that I did after the introduction of pmap_advise, so this is not a new profile. The results are sorted by total waiting time and only the top 20 entries are listed. max wait_max total