Charm++ parallel programming system

[Prateek Jindal <jindal2 AT illinois.edu>]

---------- Forwarded message ----------
From: Wijayasiri,Malavi Pathirannahalage <adeeshaw AT ufl.edu>
Date: Sun, Nov 29, 2015 at 6:41 PM
Subject: Re: [ppl] [charm] Regarding building charm++ to run with CUDA
To: Prateek Jindal <jindal2 AT illinois.edu>, "ppl-local AT lists.illinois.edu" <ppl-local AT lists.illinois.edu>, "core-ppl AT lists.illinois.edu" <core-ppl AT lists.illinois.edu>

Hi,

I am trying to use Charm++ on multiGPU machine. My requirement is to divide work load to GPU cards. I could not find any documentation regarding this. 

I read the files in  hybridAPI in charm++ and as I understood there is no support available for my requirement. ( It can run many chares but all of them run on the same GPU device). 

is my understanding correct? or is there a way to distribute work load among gpu devices using Charm++?

Thank you.

---

From: prateekjindal.uiuc AT gmail.com <prateekjindal.uiuc AT gmail.com> on behalf of Prateek Jindal <jindal2 AT illinois.edu>
Sent: Tuesday, November 24, 2015 11:30 AM
To: Wijayasiri,Malavi Pathirannahalage; ppl-local AT lists.illinois.edu; core-ppl AT lists.illinois.edu
Subject: Re: [ppl] [charm] Regarding building charm++ to run with CUDA

 

Adeesha:

Good to know that it worked out for you! You may contact us anytime with Charm++ issues that you face.

--Prateek.

On Tue, Nov 24, 2015 at 10:26 AM, Wijayasiri,Malavi Pathirannahalage <adeeshaw AT ufl.edu> wrote:

Hi Prateek,

I rebuild charm++ with new arguments and now it is working. Thank you very much for the help.

I am currently working on parallel image processing algorithms and I am hoping to use charm++ to manage scheduling. I hope I can contact you if I get into some trouble about configuring.

Thank you again.

---

From: prateekjindal.uiuc AT gmail.com <prateekjindal.uiuc AT gmail.com> on behalf of Prateek Jindal <jindal2 AT illinois.edu>
Sent: Tuesday, November 24, 2015 11:04 AM
To: Wijayasiri,Malavi Pathirannahalage
Cc: charm AT cs.illinois.edu
Subject: Re: [ppl] [charm] Regarding building charm++ to run with CUDA

 

Hi Adeesha,

Can you please try the following build line for cuda builds:

./build charm++ net-linux-x86_64 cuda -j16 --with-production

Please let us know whether the issue gets resolved or not.

Thanks,

Prateek.

On Mon, Nov 23, 2015 at 7:47 PM, Wijayasiri,Malavi Pathirannahalage <adeeshaw AT ufl.edu> wrote:

Hi,

I am a phd student in University of Florida. I am trying to build charm++ to use with CUDA. But once I tried to run example cuda programs it gives me an errors saying that 

"undefined reference to `wrQueue' "

"undefined reference to `enqueue(workRequestQueue*, workRequest*)'"

undefined reference to `kernel_stream' .

When I was trying to build hybridAPI files which includes wr.h file it gives me some other include errors(cklists.h: No such file or directory).

I think I built charm++ using wrong parameters. I used 

./build ./build charm++  multicore-linux64 -j16

Could you please tell me what parameters should I use to build charm++ to run with CUDA.

Or is this because of some other thing that I have to do.

Any help on this is highly appreciated.

Thank you very much.

Best regards,

Adeesha.

--

Prateek Jindal.

Postdoctoral Research Associate,

Dept. of Computer Science, UIUC.

email: <jindal2 AT illinois.edu>

Phone: 217-607-4513

--

Prateek Jindal.

Postdoctoral Research Associate,

Dept. of Computer Science, UIUC.

email: <jindal2 AT illinois.edu>

Phone: 217-607-4513

--

Prateek Jindal.

Postdoctoral Research Associate,

Dept. of Computer Science, UIUC.

email: <jindal2 AT illinois.edu>

Phone: 217-607-4513