Parallel performance of the XL Fortran random{_}number intrinsic function on Seaborg Page: 3 of 8
This report is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to UNT Digital Library by the UNT Libraries Government Documents Department.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
Specifies the number of threads for parallel execution of the MATMUL and
RANDOM_NUMBER intrinsic procedures. The default for num threads equals the number of
processors online.
As mentioned in the report2 on the BeamBeam3D code, a web search located the following text at the URL
http://msgs.sp2.net/cgi-bin/get/sp207/45.html:
APAR: IY35729 COMPID: 5765F7100 REL: 810
ABSTRACT: ADD A RUNTIME OPTION TO CONTROL THE NUMBER OF THREADS
PROBLEM DESCRIPTION:
ERROR DESCRIPTION:
random _number() and matmul() are 2 Fortran instrinsic functions
that are multi-threaded. It has been requested by
customers to allow users to control the number of
threads (eg. 1) used in these functions.
PROBLEM CONCLUSION:
A number of customers have experienced the problem that
multi-threaded intrinsic function floods their systems with
uncontrollable number of threads, which can result in
unacceptable run-time performance. In order to
solve this problem, we have to provide a new run-time option to
give them the ability to control the number of threads used in
those functions. Currently only matmul and random-number have
the multi-threaded version.
Syntax:
intrinthds={ numthreads }
This runtime option has a significant effect on the performance of random _number on NERSC's SP as is
show below. A test code was written and used to investigate how execution times are affected by choices
for the value of intrinthds and the number of MPI tasks used per node.
The Test Code
A test simple test code was created that does little but initialize MPI and then call RANDOM_NUMBER a
number of times. Here is the code:
program ranno
implicit none
include 'mpif.h'
integer: : i
integer, parameter:: times=100
integer, parameter:: idim=10000000
real, dimension(idim):: realArrayl,realArray2
integer: : ierr
integer:: totTasks, task idcall MPI INIT(ierr)
Upcoming Pages
Here’s what’s next.
Search Inside
This report can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Report.
Gerber, Richard A. Parallel performance of the XL Fortran random{_}number intrinsic function on Seaborg, report, July 30, 2003; Berkeley, California. (https://digital.library.unt.edu/ark:/67531/metadc740183/m1/3/: accessed March 19, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.