CRI-MAP Users Forum Posted mail
From  Thu Mar 24 06:27:38 2016
From: Jill Maddox <>
To: Multiple Recipients of Cri-Map Users <>
Subject: Re: memory usage
Date: Thu, 24 Mar 2016 06:27:38 -0500

Hi John

Build starts off using only a small number of markers and with a large data
set would take a while to get a large number of markers. For a small
population, especially with not many markers, it doesn't use much memory
until it gets to having dense maps but the density will be limited by the
maximum number of informative meioses from the population and the lod score
threshold. To get a dense map with small populations you need to do several
iterations of build if you're building from scratch - start with high lod
to get good framework then use that as a starting point with a lower lod
etc. I assume that your fixed map probably has lots more markers than you
are getting positioned with build. You might be more cpu limited than
memory limited.


At 03:52 AM 24/03/2016, Littrell, John wrote: 

>Hi all, 
>When I run build (crimap2.507) on linux, it barely uses any memory on my 
>machine - top reports 0.1 for %MEM.  Right now we are testing using 60+ 
>nuclear families with 600+ total animals.  The low memory usage persists 
>whether I use dozens, hundreds, or thousands of markers (SNPs).  As a side 
>note, running fixed with the same dataset uses at least 6.0 %MEM.  I have set 
>nb_our_alloc to 20Gb (20000000000) in the .par file, and 
>MAX_PRODUCT_CACHE_MEMORY to 20Gb (20000000000LL) in the defs.h file before 
>compiling.  I did not set any of the other parameters in the defs.h file 
>because I don't really understand what they do.  My rig has 96Gb of  
>memory and 
>6 cores. Just wondering if this is to be expected or would you recommend 
>changing any of the settings in defs.h or something else? 
>Thank you. 
>Jack Littrell 
>Computer Dork III 
>Medical College of Wisconsin? 



© 2003-2024: USA · USDA · NRPSP8 · Program to Accelerate Animal Genomics Applications. Contact: Bioinformatics Team