Hi Sophocles,
1. The stand-alone version is 64 bit (if your system supports it, which is
generally true): it says OS_X_64/U at the top.
The one in OxMetrics is 32-bit (OS_X/U) because the interface code is still
32-bit. The 64-bit version is usually a bit faster.
2. The OS X version is not yet multithreaded. Linux and Windows are using the
Intel compiler for this; I still need to try it for the OS X version.
Jurgen
oxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxox
oxoxoxox 14th OxMetrics user conference 2014
oxoxoxox Washington DC, 20-21 March
oxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxox
Dr Jurgen A Doornik
James Martin Fellow, Institute for New Economic Thinking
at the Oxford Martin School, University of Oxford
http://www.doornik.com
http://www.oxmetrics.net
oxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxoxox
On 2013-12-06 15:30, Sophocles Mavroeidis wrote:
> Dear All,
>
> I encountered the following issues when running ox7 on an iMac OS X 10.7.5, with a quad core processor intel i7 (2.93GHz).
>
> 1) The same ox program (see below) runs about 30% faster when I use OxEdit versus OxMetrics on the Mac (there is no difference on a Windows server 2008). This is true for other ox programs I tried.
>
> 2) Making the for loop parallel doesn't make any difference to speed on the Mac -- in fact it slows it somewhat. (works as expected on the Windows server)
>
> Any hints? Are these known problems when running Ox on a Mac?
>
> Thanks,
>
> Sophocles
>
> ~~~~~~~~
> the code:
>
> #include <oxstd.h>
>
> main()
> {
> decl timer = time();
> decl nrep = 1e6;
> decl vs2 = new matrix[nrep][1];
> decl cn = 1000;
> decl i;
> parallel for(i = 0; i < nrep; ++i)
> vs2[i] = meanc(sqr(rann(cn,1)));
> println(cn~meanc(vs2));
> println("elapsed time: ", timespan(timer));
>
> }
>
|