charm AT lists.siebelschool.illinois.edu
Subject: Charm++ parallel programming system
List archive
- From: "Kale, Laxmikant V" <kale AT illinois.edu>
- To: Raghuvar Prajapati <201351003 AT iiitvadodara.ac.in>, "charm AT lists.cs.illinois.edu" <charm AT lists.cs.illinois.edu>
- Subject: Re: [charm] Urgent : Queries Regarding Charm++
- Date: Fri, 7 Apr 2017 16:18:19 +0000
- Accept-language: en-US
People mean different things by “heterogeneous environments”. Charm allows you to use machines with different configurations (that you mention in Q3). Using machines with different data representations (e.g. big-endian vs little endian… this is rare these days) will require a data conversion. Number of cores being different is something that requires additional effort currently. For speed variation, a simple suggestion is to use a refinement-based load balancing strategy. More advanced will be rate-aware load balancers. Cache and RAM size variations are no problem (and the impact on speed can be handled by load balancing). (and if you mean use of accelerators such as GPGPUs, that is possible via a specialized interface).
What is your time frame for the project? This semester? Or later in the year.
-Sanjay
From:
Raghuvar Prajapati <201351003 AT iiitvadodara.ac.in>
Hello, I am working on the charm++ parallel programming framework for my Undergrad project and I am facing some problems and want to know more about it. So, can you people help me out with these queries?
Regards. Raghuvar Prajapati Final Year, B.Tech (CSE)
|
- [charm] Urgent : Queries Regarding Charm++, Raghuvar Prajapati, 04/06/2017
- Re: [charm] Urgent : Queries Regarding Charm++, Kale, Laxmikant V, 04/07/2017
- Re: [charm] Urgent : Queries Regarding Charm++, Raghuvar Prajapati, 04/11/2017
- Re: [charm] Urgent : Queries Regarding Charm++, Kale, Laxmikant V, 04/07/2017
Archive powered by MHonArc 2.6.19.