Re: forall and do loop

Wade Ward wrote:
"Richard Maine" <nospam@xxxxxxxxxxxxx> wrote in message news:1i1wrdp.1rrv09etp3izwN%nospam@xxxxxxxxxxxxxxxx
highegg <highegg@xxxxxxxxx> wrote:
[about FORALL]

Yes, I've read somewhere that it was regarded as a sort of "parallel
DO" by many, even in the standardization committe.
I should hope not. That would be a pretty serious misunderstanding of
it. While people on the committee are certainly not immune to
misunderstanding things (by a long shot), they would have had to sleep
through an awful lot in order to have avoided getting this one hammered
into them.

It is an assignment and is not much like a DO at all. The fact that some
people confuse it for a parallel DO and thus try to do things other than
assignment with it, is perhaps one of its flaws (in addition to its
usually poor performance).
Gosh, having now read this thread and compared to MR&C 6.9, I would think that the latter needs some revision. Are there other advantages to using a forall that don't have to do with speed, where apparently, it's a washout? Maybe something to do with pureness?
I suppose you could use it as a purely Fortran syntax check for some of the prerequisites for OpenMP parallel, before switching to OMP PARALLEL DO.
.

Quantcast