4
$\begingroup$

My program needs to perform pivoted QR decomposition on tall (e.g. 1e9 by 100) matrices. I run into the bottleneck that the major computational time of my program is spent on doing serial pivoted-QR decomposition, while the node has 20 threads. For some reason I am confined to one node to perform the decomposition, so that distributed strategies such as scalapack will not help. I am wondering if there is any library/package that provides threaded version of pivoted-QR decomposition. Right now my code calls lapack's sgeqp3 subroutine. Thanks!

$\endgroup$
1
$\begingroup$

The Intel MKL version of sgeqp3 uses multithreads.

I don't know how well it scales to 20 cores, however.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.