lkml.org 
[lkml]   [2019]   [Mar]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [External] Re: mm/cma.c: High latency for cma allocation
[Please do not use html emails to the mailing list and try to fix your
email client to not break quoating. Fixed for this email]

On Mon 18-03-19 13:28:50, Pankaj Suryawanshi wrote:
> On Mon 18-03-19 12:58:28, Pankaj Suryawanshi wrote:
> > > Hello,
> > >
> > > I am facing issue of high latency in CMA allocation of large size buffer.
> > >
> > > I am frequently allocating/deallocation CMA memory, latency of allocation is very high.
> > >
> > > Below are the stat for allocation/deallocation latency issue.
> > >
> > > (390100 kB), latency 29997 us
> > > (390100 kB), latency 22957 us
> > > (390100 kB), latency 25735 us
> > > (390100 kB), latency 12736 us
> > > (390100 kB), latency 26009 us
> > > (390100 kB), latency 18058 us
> > > (390100 kB), latency 27997 us
> > > (16 kB), latency 560 us
> > > (256 kB), latency 280 us
> > > (4 kB), latency 311 us
> > >
> > > I am using kernel 4.14.65 with android pie(9.0).
> > >
> > > Is there any workaround or solution for this(cma_alloc latency) issue ?
> >
> > Do you have any more detailed information on where the time is spent?
> > E.g. migration tracepoints?
> >
> > Hello Michal,
>
> I have the system(vanilla kernel) with 2GB of RAM, reserved 1GB for CMA. No swap or zram.
> Sorry, I don't have information where the time is spent.
> time is calculated in between cma_alloc call.
> I have just cma_alloc trace information/function graph.

Then please collect that data because it is really hard to judge
anything from the numbers you have provided.
--
Michal Hocko
SUSE Labs

\
 
 \ /
  Last update: 2019-03-18 14:43    [W:0.830 / U:0.108 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site