Hi,
I was wondering if there is any way to reduce the cpu usage the
openmpi seems to spend in the busy wait loop.
Thanks,
/David
On Thu, Feb 28, 2013 at 4:34 PM, Bokassa <boka...@gmail.com> wrote:
> Hi,
> I notice that a simple MPI program in which rank 0 sends 4 bytes to
> each rank and receives a reply uses a
> considerable amount of CPU in system call.s
>
> % time seconds usecs/call calls errors syscall
> ------ ----------- ----------- --------- --------- ----------------
> 61.10 0.016719 3 5194 gettimeofday
> 20.77 0.005683 2 2596 epoll_wait
> 18.13 0.004961 2 2595 sched_yield
> 0.00 0.000000 0 4 write
> 0.00 0.000000 0 4 stat
> 0.00 0.000000 0 2 readv
> 0.00 0.000000 0 2 writev
> ------ ----------- ----------- --------- --------- ----------------
> 100.00 0.027363 10397 total
>
> and
>
> Process 2512 attached - interrupt to quit
> 16:32:17.793039 sched_yield() = 0 <0.000078>
> 16:32:17.793276 gettimeofday({1362065537, 793330}, NULL) = 0 <0.000070>
> 16:32:17.793460 epoll_wait(4, {}, 32, 0) = 0 <0.000114>
> 16:32:17.793712 gettimeofday({1362065537, 793773}, NULL) = 0 <0.000097>
> 16:32:17.793914 sched_yield() = 0 <0.000089>
> 16:32:17.794107 gettimeofday({1362065537, 794157}, NULL) = 0 <0.000083>
> 16:32:17.794292 epoll_wait(4, {}, 32, 0) = 0 <0.000072>
> 16:32:17.794457 gettimeofday({1362065537, 794541}, NULL) = 0 <0.000115>
> 16:32:17.794695 sched_yield() = 0 <0.000079>
> 16:32:17.794877 gettimeofday({1362065537, 794927}, NULL) = 0 <0.000081>
> 16:32:17.795062 epoll_wait(4, {}, 32, 0) = 0 <0.000079>
> 16:32:17.795244 gettimeofday({1362065537, 795294}, NULL) = 0 <0.000082>
> 16:32:17.795432 sched_yield() = 0 <0.000096>
> 16:32:17.795761 gettimeofday({1362065537, 795814}, NULL) = 0 <0.000079>
> 16:32:17.795940 epoll_wait(4, {}, 32, 0) = 0 <0.000080>
> 16:32:17.796123 gettimeofday({1362065537, 796191}, NULL) = 0 <0.000121>
> 16:32:17.796388 sched_yield() = 0 <0.000127>
> 16:32:17.796635 gettimeofday({1362065537, 796722}, NULL) = 0 <0.000121>
> 16:32:17.796951 epoll_wait(4, {}, 32, 0) = 0 <0.000089>
>
> What is the purpose of this behavior.
>
> Thanks,
> David
>
>
