[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: cellfun vs. parcellfun: speed
|
From: |
Martin Helm |
|
Subject: |
Re: cellfun vs. parcellfun: speed |
|
Date: |
Tue, 04 Sep 2012 13:47:42 +0200 |
|
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:15.0) Gecko/20120825 Thunderbird/15.0 |
Am 04.09.2012 13:41, schrieb c.:
> My experience is that it is usually almost useless to use more processes tha
> physical cores on your machine,
> whether you have hypertreading or not makes little difference.
That was my point, hence I showed the output for both nproc=2 and
nproc=4 where the result for nproc=4 is slower than nproc=2 (but still
faster than the serial version).
You always need to play with different settings combinations on
different machines.
The more trivial the calculations are the less it makes sense to use a
"ChunksPerProc" higher than one as a rule of thumb and agree to what you
say nprocs=number of physical cpu's or physical cpu's -1 (if there are
more than 2).