I think that's the atomic operation you need.
On 23 Oct 2014 21:09, "Juergen Sauermann" <
address@hidden> wrote:
Hi,
I did some rework of the parallel stuff so that it might compile
now under OS/X and possibly other platforms. SVN 498.
On non-linux machines core affinities are not set explicitly.
However all GNU APL threads are 100% busy
unless GNU APL is blocked on input. The scheduler of the OS should
normally detect this and place
the threads on different cores even if their core affinities are
not set explicitly.
Please try the ScalarBenchmark,apl workspace.
The bare minimum that is needed for parallel APL on other
platforms than linux or OS/X (Solaris ?) is some
atomic fetch_and_add() function.
/// Jürgen