Re: Co-Array Fortran
- From: gary.l.scott@xxxxxxxx
- Date: Tue, 30 Oct 2007 15:26:44 -0700
On Oct 29, 10:42 pm, Walter Spector <w6ws_xthiso...@xxxxxxxxxxxxx>
wrote:
Gary Scott wrote:
But isn't it important to at least know whether a process is "local" or
"remote"? That could impact how you structure the problem solution.
Of course. And that is the whole point of the coarray notation.
For "local" processes, you use the arrays normally. E.g.:
a(i) = b(i) + c(i)
All the above computations are "local".
For addressing "remote" data, you use the coarray notation (square
brackets) as needed. E.g.:
a(i) = b(i) + c(i)[np]
The A and B references are local. But C is a coarray, and so would
load values from coarray process 'np'. (Which may or may not be local -
depending on the value of np.)
So all communication is very visible. Nothing is hidden, except
for the actual communication mechanism. Might be mmapped data
areas on a shared memory machine, a get/put on a distributed memory
machine, even a socket read/write.
W.
I think I understand basically, but not sure why it isn't often
important to know with certainty whether a coarray is local or not.
(performance?) If a coarray COULD be local (more or less), i might
structure the problem differently than if it might be indeterminately
local or remote.
.
- Follow-Ups:
- Re: Co-Array Fortran
- From: Dan Nagle
- Re: Co-Array Fortran
- References:
- Co-Array Fortran
- From: gary . l . scott
- Re: Co-Array Fortran
- From: Dan Nagle
- Re: Co-Array Fortran
- From: Gary Scott
- Re: Co-Array Fortran
- From: Walter Spector
- Co-Array Fortran
- Prev by Date: Re: help with MPICH ? (please)
- Next by Date: Re: Co-Array Fortran
- Previous by thread: Re: Co-Array Fortran
- Next by thread: Re: Co-Array Fortran
- Index(es):
Relevant Pages
|