Re: [PATCH] powerpc/64: Use tick accounting by default

2022-05-25 Thread Christophe Leroy
Le 22/05/2017 à 07:13, Anton Blanchard a écrit : Hi Michael, ppc64 is the only architecture that turns on VIRT_CPU_ACCOUNTING_NATIVE by default. The overhead of this option is extremely high - a context switch microbenchmark using sched_yield() is almost 20% slower. Running on what? It sho

Re: [PATCH] powerpc/64: Use tick accounting by default

2017-05-22 Thread Christophe LEROY
Le 19/05/2017 à 16:41, Anton Blanchard a écrit : From: Anton Blanchard ppc64 is the only architecture that turns on VIRT_CPU_ACCOUNTING_NATIVE by default. The overhead of this option is extremely high - a context switch microbenchmark using sched_yield() is almost 20% slower. To get finer gr

Re: [PATCH] powerpc/64: Use tick accounting by default

2017-05-21 Thread Anton Blanchard
Hi Michael, > > ppc64 is the only architecture that turns on > > VIRT_CPU_ACCOUNTING_NATIVE by default. The overhead of this option > > is extremely high - a context switch microbenchmark using > > sched_yield() is almost 20% slower. > > Running on what? It should all be nop'ed out unless you'r

Re: [PATCH] powerpc/64: Use tick accounting by default

2017-05-21 Thread Nicholas Piggin
On Mon, 22 May 2017 14:15:57 +1000 Michael Ellerman wrote: > Anton Blanchard writes: > > > From: Anton Blanchard > > > > ppc64 is the only architecture that turns on VIRT_CPU_ACCOUNTING_NATIVE > > by default. The overhead of this option is extremely high - a context > > switch microbenchmark u

Re: [PATCH] powerpc/64: Use tick accounting by default

2017-05-21 Thread Michael Ellerman
Anton Blanchard writes: > From: Anton Blanchard > > ppc64 is the only architecture that turns on VIRT_CPU_ACCOUNTING_NATIVE > by default. The overhead of this option is extremely high - a context > switch microbenchmark using sched_yield() is almost 20% slower. Running on what? It should all be

[PATCH] powerpc/64: Use tick accounting by default

2017-05-19 Thread Anton Blanchard
From: Anton Blanchard ppc64 is the only architecture that turns on VIRT_CPU_ACCOUNTING_NATIVE by default. The overhead of this option is extremely high - a context switch microbenchmark using sched_yield() is almost 20% slower. To get finer grained user/hardirq/softirq statitics, the IRQ_TIME_AC