Charm++ parallel programming system

[Vinicius Freitas <vinicius.mct.freitas AT gmail.com>]

Hello, fellow charm programmers,

I am currently working on a load balancing algorithms for the Charm++ platform, but have found difficult to run some applications in order to benchmark them.

I saw in a recent PPL paper (CCGrid 2018) that you guys have experimented with the LLNL application, Lassen. I wonder if you could provide the source code of this application, so we can execute some performance evaluation with it.

Among the applications we tested, LB Test has worked fine, as well as LeanMD. However, I have found difficult to use both ChaNGa and AMR base codes to run alongside a load balancer. What usually happens is an execution error, prompted by the Charm runtime, usually because of weird migration attempts. I've seen that AMR examples (those found in charm/examples) are creating load balancers on startup, and never calling AtSync(). Does this work? It is causing problems even with default LBs, at least for me.

For AMR, I executed the ./build AMR multicore-x86_64 script, while for ChaNGa, I followed the instructions in this link, but have had the same problems, with even standard LBs causing runtime errors.

This happened with Charm++ version 6.8.1 and 6.8.2 alike. Is there anything different you guys @ppl do when experimenting with this applications?

Thank you for your help and cooperation,
--

Vinicius Marino Calvo Torres de Freitas Computer Science Graduate Student (Aluno de pós-graduação em Ciência da Computação)

Research Assistant at the Embedded Computing Laboratory at UFSC

UFSC - CTC - INE - ECL, Brazil Email: vinicius.mctf AT posgrad.ufsc.br or vinicius.mct.freitas AT gmail.com Tel: +55 (48) 996 163 803