Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-14 Thread Peter Zijlstra
On Mon, Apr 14, 2014 at 09:34:19AM -0400, Steven Rostedt wrote: > On Mon, 14 Apr 2014 11:55:54 +0200 > Ingo Molnar wrote: > > > > Index: linux-rt.git/kernel/rtmutex.c > > > > Side note: could you please in general include diffstats with such > > patches, especially since you seem to be export

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-14 Thread Steven Rostedt
On Mon, 14 Apr 2014 11:55:54 +0200 Ingo Molnar wrote: > > Index: linux-rt.git/kernel/rtmutex.c > > Side note: could you please in general include diffstats with such > patches, especially since you seem to be exporting it from a Git repo? Sure, I'll try to remember to add them. I do my work

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-14 Thread Ingo Molnar
* Steven Rostedt wrote: > A while back ago I wrote a patch that would allow for reader/writer > locks like rwlock and rwsems to have multiple readers in PREEMPT_RT. It > was slick and fast but unfortunately it was way too complex and ridden > with nasty little critters which earned me my large c

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-11 Thread Daniel Bristot de Oliveira
On 04/10/2014 11:44 AM, Clark Williams wrote: On Wed, 9 Apr 2014 15:19:22 -0400 Steven Rostedt wrote: This patch is built on top of the two other patches that I posted earlier, which should not be as controversial. If you have any benchmark on large machines I would be very happy if you co

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Mike Galbraith
On Fri, 2014-04-11 at 05:52 +0200, Mike Galbraith wrote: > On Thu, 2014-04-10 at 23:25 -0400, Steven Rostedt wrote: > > On Fri, 11 Apr 2014 04:50:26 +0200 > > Mike Galbraith wrote: > > > > > Oh, there was another useful bit. > > > > > > 3.12.17-rt25 403991 384216 377786 > > >

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Mike Galbraith
On Thu, 2014-04-10 at 23:25 -0400, Steven Rostedt wrote: > On Fri, 11 Apr 2014 04:50:26 +0200 > Mike Galbraith wrote: > > > Oh, there was another useful bit. > > > > 3.12.17-rt25 403991 384216 377786 > >405466 394011 392850 > > > > 3.12.17-nopreempt

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Fri, 11 Apr 2014 04:50:26 +0200 Mike Galbraith wrote: > Oh, there was another useful bit. > > 3.12.17-rt25 403991 384216 377786 >405466 394011 392850 > > 3.12.17-nopreempt 109049 186133 118766 !?!.. ew, TTWU_QUEUE >3503

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Mike Galbraith
Oh, there was another useful bit. 3.12.17-rt25 403991 384216 377786 405466 394011 392850 3.12.17-nopreempt 109049 186133 118766 !?!.. ew, TTWU_QUEUE 350385 318786 367336 !TTWU_QUEUE 326009 35693

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Thu, 10 Apr 2014 14:30:03 -0700 "Paul E. McKenney" wrote: > On Thu, Apr 10, 2014 at 03:17:41PM -0400, Steven Rostedt wrote: > > On Thu, 10 Apr 2014 17:36:17 +0200 > > Peter Zijlstra wrote: > > > > > > > It defaults to the total number of CPUs in the system, given the default > > > setup (al

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Paul E. McKenney
On Thu, Apr 10, 2014 at 03:17:41PM -0400, Steven Rostedt wrote: > On Thu, 10 Apr 2014 17:36:17 +0200 > Peter Zijlstra wrote: > > > > It defaults to the total number of CPUs in the system, given the default > > setup (all CPUs in a single balance domain), this should result in all > > CPUs workin

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Peter Zijlstra
On Thu, Apr 10, 2014 at 03:17:41PM -0400, Steven Rostedt wrote: > On Thu, 10 Apr 2014 17:36:17 +0200 > Peter Zijlstra wrote: > > > > It defaults to the total number of CPUs in the system, given the default > > setup (all CPUs in a single balance domain), this should result in all > > CPUs workin

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Thu, 10 Apr 2014 17:36:17 +0200 Peter Zijlstra wrote: > It defaults to the total number of CPUs in the system, given the default > setup (all CPUs in a single balance domain), this should result in all > CPUs working concurrently on the boosted read sides. Unfortunately, it currently default

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Peter Zijlstra
On Thu, Apr 10, 2014 at 05:03:36PM +0200, Sebastian Andrzej Siewior wrote: > On 04/10/2014 04:44 PM, Clark Williams wrote: > > The means of each group of five test runs are: > > > > vanilla.log: 1210117 rt.log: 17210953 (14.2 x slower than > > vanilla) rt-fixes.log: 10062027 (8.3 x slower than

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Thu, 10 Apr 2014 17:03:36 +0200 Sebastian Andrzej Siewior wrote: > On 04/10/2014 04:44 PM, Clark Williams wrote: > > The means of each group of five test runs are: > > > > vanilla.log: 1210117 rt.log: 17210953 (14.2 x slower than > > vanilla) rt-fixes.log: 10062027 (8.3 x slower than vanil

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Sebastian Andrzej Siewior
On 04/10/2014 04:44 PM, Clark Williams wrote: > The means of each group of five test runs are: > > vanilla.log: 1210117 rt.log: 17210953 (14.2 x slower than > vanilla) rt-fixes.log: 10062027 (8.3 x slower than vanilla) > rt-multi.log: 3179582 (2.x x slower than vanilla) > > > As expected,

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Thu, 10 Apr 2014 09:44:30 -0500 Clark Williams wrote: > I wrote a program named whack_mmap_sem which creates a large (4GB) > buffer, then creates 2 x ncpus threads that are affined across all the > available cpus. These threads then randomly write into the buffer, > which should cause page fau

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Clark Williams
On Wed, 9 Apr 2014 15:19:22 -0400 Steven Rostedt wrote: > This patch is built on top of the two other patches that I posted > earlier, which should not be as controversial. > > If you have any benchmark on large machines I would be very happy if > you could test this patch against the unpatche

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Steven Rostedt
On Thu, 10 Apr 2014 16:28:43 +0200 Mike Galbraith wrote: > On Thu, 2014-04-10 at 16:18 +0200, Mike Galbraith wrote: > > On Wed, 2014-04-09 at 15:19 -0400, Steven Rostedt wrote: > > > > > If you have any benchmark on large machines I would be very happy if > > > you could test this patch against

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Mike Galbraith
On Thu, 2014-04-10 at 16:18 +0200, Mike Galbraith wrote: > On Wed, 2014-04-09 at 15:19 -0400, Steven Rostedt wrote: > > > If you have any benchmark on large machines I would be very happy if > > you could test this patch against the unpatched version of -rt. > > Too bad I don't have (and know ho

Re: [RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-10 Thread Mike Galbraith
On Wed, 2014-04-09 at 15:19 -0400, Steven Rostedt wrote: > If you have any benchmark on large machines I would be very happy if > you could test this patch against the unpatched version of -rt. Too bad I don't have (and know how to use) specjbb. I dug up old vmark, thinking I'd be able to get so

[RFC PATCH RT] rwsem: The return of multi-reader PI rwsems

2014-04-09 Thread Steven Rostedt
A while back ago I wrote a patch that would allow for reader/writer locks like rwlock and rwsems to have multiple readers in PREEMPT_RT. It was slick and fast but unfortunately it was way too complex and ridden with nasty little critters which earned me my large collection of frozen sharks in the f