Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

pdtran upgrade

t_r_hopkinskent_ac_u
298 Views
In the list of upgrades to the latest release of the MKL (10.1 Update 2) it mentions

Fixed a load-balancing problem in the pdtran ScaLAPACK function by improving the communications
scheme improving performance by up to 3 times

Can anyone give me details of the changes that were made to the source code as I'd like to update my version
of this routine to see if it gets rid of a problem I seem to be having when I try and transpose very large matrices
under MPI.

Many thanks,

Tim
0 Kudos
1 Reply
Andrei_Moskalev__Int
298 Views
Some ourchanges were based on ideas described in the following paper: http://users.cs.cf.ac.uk/David.W.Walker/LINALG/transpose95.html
Also therearesome problem for big sizes in current implementation, for example instability of performance.
Try to use ITAC - Intel Trace Analyzer and Collector for obtaining data on your implementation load balance.

0 Kudos
Reply