NPPS Leadership Team

[Torre Wenaus <wenaus AT gmail.com>]

FYI. The response I'm thinking of is that when we fill the joint position with HEP-CCE directed at GPU utilization we can make this activity (basically being a domain science guinea pig for their developments) part of that work program. But until then we have no effort, they have the effort, namely Meifeng et al who have received HEP-CCE support. And for completeness I will ask whether we can expect any support for our contributions. Comments and suggestions appreciated.

  Torre

---------- Forwarded message ---------
From: Barbara Chapman <barbara.chapman AT stonybrook.edu>
Date: Tue, Mar 30, 2021 at 5:38 PM
Subject: Application developer / user contributions to X-Stack proposal
To: Campbell, Stuart <scampbell AT bnl.gov>, Wenaus,Torre J <wenaus AT bnl.gov>
Cc: Malik,Abid <amalik AT bnl.gov>

Dear Stuart, dear Torre,

First I apologize for being so late in getting back to you regarding the X-Stack proposal.  

We submitted two pre proposals to the call and both of them got the green light, so Abid and I have been very busy in the past two weeks.

The final submission date is April 12.

One of the proposals is headed up by PNNL and they have  their target applications already.

We propose to include you in the BNL-led submission which is focused on extending the state of the art in technology for converting code from one parallel programming model to another.  Our approach has two parts:  a traditional one that extends today's conversion techniques and makes them easier to implement and a non-traditional one.   The non-traditional approach is a collaboration with our ML team and applies insights gained in NLP to the problem of translating code.  We have a colleague from ANL working on verification to complement this. 

For this proposal, it would be great if you could provide code that can help us manually describe translation patterns and also get code that can serve as data to the ML approaches.

It would be especially valuable if you have code fragments or regions written in two or more parallel programming APIs. 

Our primary focus is C++  and we are interested in a variety of parallel programming models - MPI, OpenMP. CUDA, OpenCL especially.  I recall that you also had some Kokkos code that we could potentially use.

I hope this will be interesting for you and that you would like to join us.

Best,

Barbara

--

-- Torre Wenaus, BNL NPPS Group, ATLAS Experiment

-- BNL 510A 1-222 | 631-681-7892 |  wenaus AT gmail.com | npps.bnl.gov | wenaus.com

-- NPPS Mattermost room: https://chat.sdcc.bnl.gov/npps/channels/town-square