Is Parallel Computing Essential Beyond Science and Graphics?

[Monte_Carlo]

Hi,

I'm curious what is a general view on parallel computing? Is it something that's important only for science and computer graphics professionals or does it also have applications in business, etc. Are there any particular skills in parallel computing that employers are looking for, any degrees/certificates or specific knowledge that's in demand? Do you think there is a growing demand for people who know MPI?

Thanks,

Monte

 

[twofish-quant]

I'm curious what is a general view on parallel computing? Is it something that's important only for science and computer graphics professionals or does it also have applications in business, etc.

All new high performance computing is parallel computing.

Are there any particular skills in parallel computing that employers are looking for, any degrees/certificates or specific knowledge that's in demand?

Any specific knowledge is "flavor of the month". Changes from month to month. However, the number of people that can do real time multithreading isn't that large.

Do you think there is a growing demand for people who know MPI?

Like most computer things, the important knowledge tends to be platform general. Also there is a difference between people that can program with MPI, and people that can program MPI.

 

[chiro]

Hi,

I'm curious what is a general view on parallel computing? Is it something that's important only for science and computer graphics professionals or does it also have applications in business, etc. Are there any particular skills in parallel computing that employers are looking for, any degrees/certificates or specific knowledge that's in demand? Do you think there is a growing demand for people who know MPI?

Thanks,

Monte

There are always applications for new techniques in computation.

One interesting application is doing bitcoin mining. There is actually a distro that allows you to to bitcoin mining using your GPU, since GPU's are great to doing things in parallel, since they are designed that way (since you deal with textures with individual picture elements as well as vertices for 3D data that are all independent and can be processed in this way).

I don't think you should study so called "MPI" in isolation. Just pick a domain, and an existing repository in that domain, and do something specific. It's the best way to get experience because it is 1) directly applicable to something that is tangible (at least in terms of results) and 2) It will contribute to your knowledge of that domain.

If you are interested in surveying the parallel stuff without wanting to do anything specific, what I recommend you do is focus on the different types of computation and then think about parallelism in the context of these things.

One way to identify these different paradigms is to be aware of the different types:

http://en.wikipedia.org/wiki/Comparison_of_programming_paradigms

Then on the hardware front, look at the different hardware architectures:

http://en.wikipedia.org/wiki/Parallel_computing

[I'm sure you have seen the above link though :)]