Mapping Cancer Markers - GPU coding

All: The folks at Mapping Cancer Markers use BOINC for distributed computing: Mapping Cancer Markers | Research | World Community Grid

How would they go about getting some help with GPU coding for the research project? Is this forum appropriate, or would another be better?

Thank you for any advice.

Could you be more specific? Help as in “we are porting our code base to CUDA and have hit a snag, what the best way forward?” or help as in “how could we motivate NVIDIA to donate the time of some of their engineers for some hands-on work on our CUDA port?”

Questions of the first kind is what we routinely deal with here. Questions of the second kind are typically handled through different channels at NVIDIA developer relations. No idea where to start that process, but your post here may be sufficient for someone to get in touch. In general NVIDIA appears to have been very supportive of GPU-based applications in the life sciences.

My memory is hazy, but I seem to recall that NVIDIA provided some hands-on support to Folding@Home to get them started on GPUs. Since I started folding at the time, I am pretty sure that was in mid 2008, when expertise in GPUs and CUDA was hard to find. Things may be different now (GPUs and CUDA are ubiquitous), but it would not hurt to ask.

Thanks very much for responding. The MCM research team includes at least 3 programmers:
( Jurisica Lab - Mapping Cancer Markers ),
but they are overloaded and likely need some help of the second kind.

However, they did successfully port a previous project to GPUs, so they may only need short-term assistance with the latest methods.

Yes, Folding@Home is still going strong, mainly on Nvidia cards; legend has it that the originator of that project Vijay Pande and Nvidia’s founder Jensen Huang were Stanford friends … tho my memory could be hazy, too.

This MCM research is exceptionally important, to anyone who could develop cancer, (that is, anyone), so whomever who can lend a hand would be making a real difference.

Thanks again for your thoughts.

Family members of mine passed away from cancer and other neoplasms, which is why I am still folding 14.5 years later. In the non-official member rankings for Folding@Home I am currently at #2596. I am on team whoopass which was founded by people on the early CUDA team. I am also participating in Rosetta@home on BOINC, but that is purely CPU based.

Is your code base open source? I understand, of course, that not everything can be open sourced, but it might make it easier to attract help from third parties.

njuffa: Thanks again for your reply - the fact that you are interested will no doubt give some hope to the team at Mapping Cancer Markers, so Thank you.

I am just a volunteer cruncher, trying to help beat cancer, so I really have no idea about the details of the MCM research code. Please send an email to the director Dr. Jurisica, his address is on the project’s site: https://www.cs.toronto.edu/~juris/jlab/contact.html

He will be very happy to hear from you!

Thanks as well for your contributions to F@H, everybody’s help is always welcome. My computers finished their first billion points last year, (which put me at #2655), then I took a break and have been helping with Einstein@home for my GPUs, with MCM on my CPUs. Once MCM is ported to GPUs, I will DEFINITELY switch to that.

Thanks again for your encouragement.

I’m not aware of a forum for requests for NVIDIA developers to work on a code. NVIDIA definitely works on various strategic 3rd party codes, but the selection process for this isn’t based on making requests on a forum somewhere.

If the folks at U of Toronto want to make contact with NVIDIA, we have a HER (Higher Education and Research) team that covers Canada. This would be the logical point of connection. I’ll drop a note to some members of that team who work in eastern Canada, but I can’t guarantee any specific outcomes. Any representatives of that team are welcome to reach out to me using this forum and I will also make a personal contact that way. I won’t be able to do it based on anything or anyone so far in this thread. I would need someone to contact me who has a U of Toronto email address before I would be able to take that any further.

None of this is any sort of commitment or guarantee; that would have to be evaluated by the team, and I am not on that team.

I would also note that CUDA programming has become a mainstream teaching vehicle for HPC courses at many major universities. I would be surprised if there were no such animal at U of Toronto. The Jurisica team there probably has various additional paths they could investigate to tap into knowledgeable CUDA programmers as well as other faculty and staff who have similar interests.

1 Like

Robert_Crovella: Thanks so much for your response. It is well understood that your comments are not any type of commitment - at this point I am just trying to make some sort of connection, to encourage the investigation of possibilities.

I have forwarded to Dr, Jurisica a link to this conversation - he does welcome assistance, so ideally you will hear from him soon.

Thanks again so much for your interest and insights.

Robert_Crovella:

“I would need someone to contact me who has a U of Toronto email address before I would be able to take that any further.”

How does Dr. Jurisica contact you? His email is on his project’s website, if you are able touch base with him: https://www.cs.toronto.edu/~juris/jlab/contact.html

Thanks again

reach out to me using the private message feature of this forum. Click on the avatar/icon to the left

<—

and select the message button