Free access to Xeon Phi clusters

Question

Apologies if this is not the right forum to ask, but there has been a somewhat related question here.

I am working on a piece of software (nonlinear constrained optimisation, coded in C++11 and Python3) that I would like to try out on a Xeon Phi. Fujitsu Europe apparently offers free remote access to a Knight's Landing-based small cluster which, in their words on the webpage,

[...] is freely available to users who register and agree with the conditions of usage.

The cluster environment enables users to participate in activities such as benchmarking, code development, as well as code optimization and validation either using the pre-installed applications or those brought by the user.

This sounds too good to be true. And probably it is not true :-( , because I have tried to sign up 3 (three) times in the past 2 weeks, but received no reply whatsoever. I also wrote them an e-mail, inquiring about what happened to my sign-up requests: again, total silence.

My questions:

1. Has anyone succeeded getting an account on this Fujitsu cluster?
2. If not, then can you suggest a similar freely available resource with Xeon Phi coprocessors for code development, testing and tuning?

If the tests show that my code can be accelerated by a Xeon Phi then I plan to convince my boss to purchase such a system. But without benchmarking numbers he won't open the purse strings (and rightly so, I must say).

Answer 1

Well, I can answer at least my first question: yes, it is indeed possible to get an account on Fujitsu Europe's KNL cluster. It just takes a veeeery long time.... but finally they managed it.

First impressions:

1. Access is granted only for 30 days, but apparently a prolongation is possible. You have to ask their support for an extension.
2. Access is only possible via a (nice) web interface which offers a virtual "remote terminal" and a graphical file manager, among other things. Quite attractive. Direct command-line SSH access is blocked, however.
3. Some of the software installed on the cluster is outdated. For instance cmake: they offer Version 2.8.12 (!). Also there is no Python3, only Python2 (which IMO should go the way of Python1, but I digress).
4. I had some initial questions about how to use the system, the help desk was very responsive and gave me useful answers.

TL;DR: worth trying out, but be prepared for a long initial "lag phase".