comp.lang.idl-pvwave archive
Messages from Usenet group comp.lang.idl-pvwave, compiled by Paulo Penteado

Home » Public Forums » archive » Specifying a particular LAPACK implementation using the standard IDL7 LAPACK DLM?
Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend 
Return to the default flat view Create a new topic Submit Reply
Specifying a particular LAPACK implementation using the standard IDL7 LAPACK DLM? [message #82605] Sun, 06 January 2013 19:32 Go to previous message
Matt Francis is currently offline  Matt Francis
Messages: 94
Registered: May 2010
Member
I have a task that has a bottleneck inverting large matrices. My understanding is that the LAPACK DLM that comes with IDL7 that is loaded when LA_INVERT is called for the first time is a serial implementation? I am running my code on an 8 core machine so would expect a reasonable speed up in the matrix inversion routine could utilise multiple cores to do the operation in parallel.

There are multi threaded LAPACK implementations, such as ATLAS, that I could use. If possible though, I'd like to avoid creating a new DLM from scratch. It would seem to me to be conceivable that if I could point IDL to the ATLAS (or whatever) multi-threaded LAPACK implementation instead of wherever it currently points to that this might neatly achieve what I want to do.

So, does anyone know if this is possible and if so how to do it? If not, does anyone have any suggestions for the simplest way to achieve what I want, which is parallel (dense) matrix inversion in IDL?
[Message index]
 
Read Message
Read Message
Previous Topic: Re: how to over plot after using "cgZplot"
Next Topic: Passing objects and parenthesis... doing something wrong here?

-=] Back to Top [=-
[ Syndicate this forum (XML) ] [ RSS ] [ PDF ]

Current Time: Wed Oct 08 19:39:14 PDT 2025

Total time taken to generate the page: 0.00450 seconds