[vsni.co.uk] Contact us
Author Message
Post new topic

  ASReml  ~  A faster ASReml for slow jobs.

Posted: Fri Oct 19, 2018 2:02 am Reply with quote
Joined: 05 Aug 2008 Posts: 471 Location: Orange, NSW
Dear ASReml Users.

This post is to advise that the current latest release of ASReml 4.1 has a
substantial speed advantage for slow jobs, jobs that involve dense matrices, whether
US or GRM.

There is also a beta release of ASReml 4.2 which is even faster for such jobs.

There is also often a speed gain from using factor analytic (XFA) variance structures over US
structures, and a further gain by modelling these as RR+DIAG when crossed with GRM/NRM matrices.

ASReml 4.1
The significant speed change here simply involves processing a loop in the reverse order
so the chance of this destabilising the program is small. The changed order means that
cells are mainly inserted at the beginning of a link list so the insertion point is
quicker to find when the list is long.

You can obtain the latest build of ASReml 4.1 (Standalone) with
where ?????? is your 6 character site reference.

This change has also been implemented ASreml-R v4 which should soon be available.
ASReml-R v4 involves a major rewrite of the R interface code as well as a new licensing

ASReml 4.2
The additional speed gain in ASReml 4.2 (over the latest 4.1) is because
the Absorption process has been completely reorganised to avoid link list
processing. There are also functionality extensions for HGLM models, and an
increase in workspace that can be accessed from a maximum of 32 Gbyte to 96Gbyte.
This remake has not yet been extensively used and so there is an increased risk of usage issues.
We are looking for beta testers who will report any issues in the interests of all users.
Again the speed gain in slow jobs is substantial.

A user in New Zealand reported a job that was crashing (under windows and linux). I suggested the
bug was probably fixed in the latest version, but recommeded he use 4.2 since his was a large GRM
job. It was taking 8 hours per iteration and crashed after 5 iterations (40 hours).
The feedback was
Thanks very much for your prompt reply.

I installed version 4.2 and got Mary to rerun her problematic job. When I got back from
lunch early, I saw that the job had completed successfully and put the output on Mary’s desk.
I wish you could have seen the look on Mary’s face when she calculated the elapsed time!
All power to parallel processing – I did get a hint when I looked at Task Manager in
Windows 10 and saw 99% against CPU usage."

To access the beta release of ASReml 4.2, goto

You will need your site reference to access the download page.
It should run with your existing ASReml license key.

We would be very grateful for your feedback, comments and suggestions -
and of course to hear about any problems as soon as possible!

Reduced Rank Factor Analytic models.
It is not unusual for users to have difficulty fitting unstructured
variance component models, especially with more than 10 parameters,
and these are also slow to fit when crossed
with a dense genomic relationship matrix (GRM).

We recommend considering the Factor Analytic model (XFA) as it allows the
fitted matrix to be reduced rank (positive semidefinite). The down side of using
a factor analytic is the need to build up from 1 to 2 ... to k factors.
That is replacing us(Env).grm(Id) with xfak(Env).grm(Id) where k is the number of factors.

Some users have discovered that these models run faster if actually fitted as two
terms, one estimating the loadings, and the other the specific variances.
That is replacing xfak(Env).grm(Id) with rrk(Env).grm(Id) + diag(Env).grm(Id).

Arthur Gilmour

Retired Principal Research Scientist (Biometrics)
View user's profile Send private message Send e-mail Visit poster's website
Posted: Wed Feb 27, 2019 12:59 am Reply with quote
Joined: 14 Dec 2010 Posts: 21
The link to the 4.2 beta version redirects to main vsni page. Is there a way to access the download link for 4.2 now?
Thanks - Jim
View user's profile Send private message

Display posts from previous:  

All times are GMT
Page 1 of 1
Post new topic

Jump to:  

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum