Hello,
I've a suggestion on my own, since MPI allows people to develop massively
parallel applications, libraries itself should also be taking advantage of
this. Today, we are at a point where we are talking about big data, but in
fact, we are about to face with huge data (I think this will be a fact in a
couple of years). So, people like me already working on implementing
parallel solutions for daily problems.Still, the data itself is not the
only one who becomes huge, but the source code too. With the newest
discoveries of algorithms, patterns, technologies in development, companies
and individuals are doing their best to take advantage of it and this
causes to produce huge source code piles which take hours to compile.
So, my suggestion is that, as standard, MPI compilers (like mpicc, mpif90)
should offer an option to the users which enables them to start a
distributed compile operation, it can be executed either on multiple cores
locally or multiple CPUs throughout the network.
With the best wishes,
--
Özgür Pekçağlıyan
Computer Engineer (M.Sc.)
