SekeRob, thank you for your help.

Optimizing software can be easy but for the purpose of this software, it has to go through a long validation process. Just one WU being successfully validated is not a thorough test of the software. The researchers spent many hours on the software and the validation process. This is not something that could easily be performed by anyone here.

I have no idea what the expected error rate is. Some WU's error out on some machines but not others and the researchers expect this. Hypothetically speaking, we'll say one out of every 10,000 errors out. What if someone optimized the code and it was one out of 1,000 or even 1 out of 5,000. How could one person do that in a timely fashion? You also have various versions of OS X, Windows, Linux and Android and different processor architectures.

If the software is proprietary, then if they give it to someone it opens the door for it to be given to others. Sure you can put an NDA and a license attached to it, but it would find its way to others.

What if someone optimized it and many users used it as it was set to use newer processor technology? Then you have hardcore crunchers using it and what if the error rate was high and you had the same WU given to two people running the optimized code and they both performed the same error? You now have a valid WU result per the system but it really isn't a valid WU result.

The MCM scientists don't want to deal with WU's being returned that could be suspect. It detracts from what they are trying to accomplish.

As previously mentioned, AD is open source and you could use that.

The project may not accept your result though. AD Vina supports multiple cores/threads, so one could build it and go through a WU extremely quickly. Rather than a quad core with a total of 8 threads running 8 WU's, you could have all 8 threads working on the same WU. If a WU on one thread took 32 minutes to complete, it would take 4 minutes (if the scalability was completely linear). The validation scheme could have a minimum threshold set and would reject that result.

The BOINC software is also open source, so you could build your own DC project. Obviously you will never use the results for science but you would get a complete working system. You might be able to use what you learned with other BOINC projects. You never know, in the end you could get a job. You could even setup your own project on something you are interested in.

Ian,

I don't think he wants to run the app for WCG. He just needs it for his thesis.

Yes, what I meant was that he does not wish to compile an optimized app and run it ON WCG.

He wants to compile for Intel GALILEO and run it locally, I believe.