I need to reduce project build time in Yocto (open embedded) in Linux platform. So I decided to cluster computers for the best performance for compilation and building.

My question:
Is there any MPI which best suits this scenario?

  • Welcome to U&L. We want complete, self contained answers here on this site, so asking for links (or reference materials) is inappropriate. – Anthon Feb 7 '16 at 6:44
  • Sorry, I'm new to this forum. – GokulTechno Feb 7 '16 at 12:53

I am working on a new MPI platform, but this won't work for this scenario, because so far I understand, you have a lot of shared memory.

Maybe it could help you openmp with multithreading for local parallelism and according to my experience, for parallelism through the network, OpenMPI is the best choice and well documented for linux distributions.

P.S. Could you describe more the parallelism requirements?

  • First, thank you for the answer. My requirement is, I need to build source files parallel using yocto with Linux based computers connected through network and final output to the master computer. – GokulTechno Feb 7 '16 at 12:50
  • Have a look at this answer: stackoverflow.com/questions/14472175/… – Frank Meerkötter Feb 7 '16 at 14:24
  • @GokulTechno Ok, then you require a lot of shared memory. As Frank did mention, it is possible to use parallel build (using local parallelism). I don't have experience with yocto, however, according to my experience I recommend you to start trying to compile each header and source file per CPU and then join all precompiled files into the final binary on the master computer. Probably you will need to wirte a brief MP interface. – ncomputers Feb 7 '16 at 19:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.