Re: [ck] [PATCH] sched: staircase deadline misc fixes
Am Sonntag 01 April 2007 schrieb michael chang: > On 4/1/07, Prakash Punnoor <[EMAIL PROTECTED]> wrote: > > Am Mittwoch 28 März 2007 schrieb Prakash Punnoor: > > > > > > Hi, I am using 2.6.21-rc5 with rsdl 0.37 and think I still see a > > > regression with my Athlon X2. Namely using this ac3 encoder > > > (http://aften.sourceforge.net/), which I parallelized in a simple way, > > > with my test sample I remember having encoding times of ~5.4sec with > > > vanilla and ~5.8 sec with rsdl - once the whole test wave is in cache. > > BTW, I confirmed this regression. With vanilla 2.76.21-rc5 I get back my > > 5.4 secs with the test sample and two threads. Furtmermore for me vanilla > > Which version of RSDL were you comparing to 2.6.21-rc5? Did you try > the patch in the first message (http://lkml.org/lkml/2007/3/28/146)? > The patch that _began_ this thread had SMP fixes in it... (Also, IIRC, > the latest version of the scheduler no longer has the rotating > component - so it's just SDl now.) As I said, I tried 0.37. Didn't it have the fix inside? Actually I am reluctant to go back to (r)sdl, as it didn't show improvements for me, yet. -- (°= =°) //\ Prakash Punnoor /\\ V_/ \_V pgplirkxW9eu6.pgp Description: PGP signature
Re: [ck] [PATCH] sched: staircase deadline misc fixes
Am Mittwoch 28 März 2007 schrieb Prakash Punnoor: > Am Mittwoch 28 März 2007 schrieb Con Kolivas: > > I'm cautiously optimistic that we're at the thin edge of the bugfix wedge > > now. > > > > --- > > set_load_weight() should be performed after p->quota is set. This fixes a > > large SMP performance regression. > > Hi, I am using 2.6.21-rc5 with rsdl 0.37 and think I still see a regression > with my Athlon X2. Namely using this ac3 encoder > (http://aften.sourceforge.net/), which I parallelized in a simple way, with > my test sample I remember having encoding times of ~5.4sec with vanilla and > ~5.8 sec with rsdl - once the whole test wave is in cache. Otherwise you > can easily I/O limit the encoder. ;-) You need to get sources from svn > though. The current 0.06 release doesn't have threads support. BTW, I confirmed this regression. With vanilla 2.76.21-rc5 I get back my 5.4 secs with the test sample and two threads. Furtmermore for me vanilla actually feels nicer on my dual core, even with load - just subjectively that's why I ditched rsdl... Cheers, -- (°= =°) //\ Prakash Punnoor /\\ V_/ \_V pgpRah6vrej2y.pgp Description: PGP signature
Re: [ck] [PATCH] sched: staircase deadline misc fixes
Am Mittwoch 28 März 2007 schrieb Con Kolivas: > I'm cautiously optimistic that we're at the thin edge of the bugfix wedge > now. > > --- > set_load_weight() should be performed after p->quota is set. This fixes a > large SMP performance regression. Hi, I am using 2.6.21-rc5 with rsdl 0.37 and think I still see a regression with my Athlon X2. Namely using this ac3 encoder (http://aften.sourceforge.net/), which I parallelized in a simple way, with my test sample I remember having encoding times of ~5.4sec with vanilla and ~5.8 sec with rsdl - once the whole test wave is in cache. Otherwise you can easily I/O limit the encoder. ;-) You need to get sources from svn though. The current 0.06 release doesn't have threads support. Cheers, -- (°= =°) //\ Prakash Punnoor /\\ V_/ \_V pgpPd1wincXFE.pgp Description: PGP signature
