On Fri 20-11-20 19:44:24, Feng Tang wrote:
> On Fri, Nov 13, 2020 at 03:34:36PM +0800, Feng Tang wrote:
> > On Thu, Nov 12, 2020 at 03:16:54PM +0100, Michal Hocko wrote:
> > > > > > I add one phony page_counter after the union and re-test, the
> > > > > > regression
> > > > > > reduced to -1.2%. It looks like the regression caused by the data
> > > > > > structure
> > > > > > layout change.
> > > > >
> > > > > Thanks for double checking. Could you try to cache align the
> > > > > page_counter struct? If that helps then we should figure which
> > > > > counters
> > > > > acks against each other by adding the alignement between the
> > > > > respective
> > > > > counters.
> > > >
> > > > We tried below patch to make the 'page_counter' aligned.
> > > >
> > > > diff --git a/include/linux/page_counter.h
> > > > b/include/linux/page_counter.h
> > > > index bab7e57..9efa6f7 100644
> > > > --- a/include/linux/page_counter.h
> > > > +++ b/include/linux/page_counter.h
> > > > @@ -26,7 +26,7 @@ struct page_counter {
> > > > /* legacy */
> > > > unsigned long watermark;
> > > > unsigned long failcnt;
> > > > -};
> > > > +} ____cacheline_internodealigned_in_smp;
> > > >
> > > > and with it, the -22.7% peformance change turns to a small -1.7%, which
> > > > confirms the performance bump is caused by the change to data alignment.
> > > >
> > > > After the patch, size of 'page_counter' increases from 104 bytes to 128
> > > > bytes, and the size of 'mem_cgroup' increases from 2880 bytes to 3008
> > > > bytes(with our kernel config). Another major data structure which
> > > > contains 'page_counter' is 'hugetlb_cgroup', whose size will change
> > > > from 912B to 1024B.
> > > >
> > > > Should we make these page_counters aligned to reduce cacheline conflict?
> > >
> > > I would rather focus on a more effective mem_cgroup layout. It is very
> > > likely that we are just stumbling over two counters here.
> > >
> > > Could you try to add cache alignment of counters after memory and see
> > > which one makes the difference? I do not expect memsw to be the one
> > > because that one is used together with the main counter. But who knows
> > > maybe the way it crosses the cache line has the exact effect. Hard to
> > > tell without other numbers.
> >
> > I added some alignments change around the 'memsw', but neither of them can
> > restore the -22.7%. Following are some log showing how the alignments
> > are:
> >
> > tl: memcg=0x7cd1000 memory=0x7cd10d0 memsw=0x7cd1140 kmem=0x7cd11b0
> > tcpmem=0x7cd1220
> > t2: memcg=0x7cd0000 memory=0x7cd00d0 memsw=0x7cd0140 kmem=0x7cd01c0
> > tcpmem=0x7cd0230
> >
> > So both of the 'memsw' are aligned, but t2's 'kmem' is aligned while
> > t1's is not.
> >
> > I will check more on the perf data about detailed hotspots.
>
> Some more check updates about it:
>
> Waiman's patch is effectively removing one 'struct page_counter' between
> 'memory' and "memsw'. And the mem_cgroup is:
>
> struct mem_cgroup {
>
> ...
>
> struct page_counter memory; /* Both v1 & v2 */
>
> union {
> struct page_counter swap; /* v2 only */
> struct page_counter memsw; /* v1 only */
> };
>
> /* Legacy consumer-oriented counters */
> struct page_counter kmem; /* v1 only */
> struct page_counter tcpmem; /* v1 only */
>
> ...
> ...
>
> MEMCG_PADDING(_pad1_);
>
> atomic_t moving_account;
> struct task_struct *move_lock_task;
>
> ...
> };
>
>
> I do experiments by inserting a 'page_counter' between 'memory'
> and the 'MEMCG_PADDING(_pad1_)', no matter where I put it, the
> benchmark result can be recovered from 145K to 185K, which is
> really confusing, as adding a 'page_counter' right before the
> '_pad1_' doesn't change cache alignment of any members.
Have you checked the result of pahole before and after your modification
whether something stands out?
Btw. is this reproducible an different CPU models?
--
Michal Hocko
SUSE Labs
