Re: [hwloc-users] unusual memory binding results

2019-01-28 Thread Biddiscombe, John A.
Brice >Can you print the pattern before and after thread 1 touched its pages, or even >in the middle ? >It looks like somebody is touching too many pages here. Experimenting with different threads touching one or more pages, I get unpredicatable results here on the 8 numa node device, the

Re: [hwloc-users] unusual memory binding results

2019-01-28 Thread Biddiscombe, John A.
If I disable thread 0 and allow thread 1 then I get this pattern on 1 machine (clearly wrong)

Re: [hwloc-users] unusual memory binding results

2019-01-28 Thread Brice Goglin
Le 28/01/2019 à 11:28, Biddiscombe, John A. a écrit : > If I disable thread 0 and allow thread 1 then I get this pattern on 1 machine > (clearly wrong) > >

Re: [hwloc-users] unusual memory binding results

2019-01-28 Thread Biddiscombe, John A.
Brice I might have been using the wrong params to hwloc_get_area_memlocation in my original version, but I bypassed it and have been calling int get_numa_domain(void *page) { HPX_ASSERT( (std::size_t(page) & 4095) ==0 ); void *pages[1] = { page };

Re: [hwloc-users] unusual memory binding results

2019-01-28 Thread Brice Goglin
Can you try again disabling the touching in one thread to check whether the other thread only touched its own pages? (others' status should be -2 (ENOENT)) Recent kernels have ways to migrate memory at runtime (CONFIG_NUMA_BALANCING) but this should only occur when it detects that some thread