k-user AT lists.cs.illinois.edu

Subject: K-user mailing list

List archive

[[K-user] ] Modeling performance as semantics in simulation

From: Eric Huber <echuber2 AT illinois.edu>
To: k-user AT lists.cs.illinois.edu
Subject: [[K-user] ] Modeling performance as semantics in simulation
Date: Thu, 25 Oct 2018 00:34:45 -0500
Authentication-results: illinois.edu; spf=neutral smtp.mailfrom=echuber2 AT illinois.edu; dkim=pass header.d=illinois-edu.20150623.gappssmtp.com header.s=20150623; dmarc=none header.from=illinois.edu

I'm wondering about something and maybe you all can help me settle a bet, so to speak. Would it be theoretically possible to create an execution model for OpenMP in K, such that the system can simulate performance benchmarks for different code samples?

The use case would be, instead of running students' parallel programming code on actual multi-threaded systems and measuring real performance, to execute the code in a simulator that could give a somewhat accurate measurement of the relative efficiency of their implementation.

Thanks,

Eric Huber

[[K-user] ] Modeling performance as semantics in simulation, Eric Huber, 10/25/2018