topic In OpenMP, Reduction Clause in Intel® Moderncode for Parallel Architectures

In OpenMP, Reduction Clause

zhangzhe65 — Thu, 26 Mar 2009 14:46:15 GMT

Why?

#pragma omp parallel for
for (i=0; i{
sum+ = a * b;
}
and

#pragma omp parallel for reduction(+:sum)
for (i=0; i{
sum+ = a * b;
}

Please tell me why the two running results are identical?
Thank you.

Re: In OpenMP, Reduction Clause

jimdempseyatthecove — Thu, 26 Mar 2009 17:13:35 GMT

a) You got lucky
b) You compiled with OpenMP off
c) You compiled with OpenMP on, Nested levels off, and loop is in nested level.
d) You compiled with OpenMP on, but only 1 thread is used
e) N <= 1
f) all of a or all of b or both are 0
g) N is sufficiently small such that each thread finishes before the other starts (assuming multiple threads run the loop)

Jim Dempsey

Re: In OpenMP, Reduction Clause

zhangzhe65 — Fri, 27 Mar 2009 09:19:28 GMT

Quoting - jimdempseyatthecove

Re: In OpenMP, Reduction Clause

zhangzhe65 — Fri, 27 Mar 2009 09:25:38 GMT

Thank you very much indeed for your reply.