Lazy question again, just curious, like, in Intel Fortran, with /qparallel, I believe gfortran have the same thing. Like below,
- Does /qparallel automatically parallelize something?
Or does it need to specify thing like !$OMP … ?
- Do you guys find sometime enabling /qopt-matmul + /qparallel , the code is actually slower? If so why?
Thank you very much!
Such options are indeed meant to automatically parallellise things, but only if the compiler can determine it is safe to do so (AFAIK). It can produce reports on whether that was done or not.
Parallellisation, alas, does not automatically mean that things go faster: it depends on the overhead induced by starting several threads and dispatching the work. So it makes sense for heavy-duty loops for instance where a lot goes on.