Slashdot Mirror

So whatever your politics are think about this one for a minute.

I worked for the Army this summer in the Army Research Lab (ARL). They have work for all sorts of areas from CS (me), Chem, Chip Fab, Acustics, etc. Just about anything that you could want to do the Army does, and they are some of the best at what they do. The area I worked they were simply the best in the world. I worked on a defensive technology, something which will actually e saving peoples lifes. Very cool stuff.

The pay was good (can be very good if you get into the correct program), mine is turning into a long term project/employment. If that happens one can end up with a security clearence, which gets one working on VERY interesting projects (and can be worth big $$$).

I got to live in another part of the US (MD/DC) and very much enoyed it!

Check out the below if you are a undergrad, or high school in MN. They run a great program.
http://www.ahpcrc.org/education/

There are also PET interns who are gov't workers. And one of the nice parts is that they can't ship you ass out to an interesting place to kill people.

bmayer 9@9 cs.umn.edu if you want to get ahold of me.

http://images.google.com/imgres?imgurl=http://www. teknoel.com/theory/connectionGrid/cm-2-andreaM.jpg &imgrefurl=http://www.teknoel.com/theory/connectio nGrid/cm-2.htm&h=240&w=240&sz=18&tbnid=_MOrYciW4rw J:&tbnh=104&tbnw=104&start=48&prev=/images%3Fq%3Dc m-2%26start%3D40%26hl%3Den%26lr%3D%26c2coff%3D1%26 client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:offic ial%26sa%3DN.

The machine was a massive SIMD processor. It didn't have one light per CPU, but one light per 16 or 32 (or was it 64?) CPUs. Depending on the setup, the light was on if the group of processors it represented was active.

Their other machine, the CM-5 also looked way cool:

http://www.ahpcrc.org/~ajohn/ARCHIVE/cm5.gif.

I still think the CM-2 is more 'space age' and I think it was better looking.

The Cray1/XMP/YMP was also a great looking box, even though it had no blinking lights. The benches were warm, becasue that's where the power supplies were housed, and even though it wasn't exactly comfortable, you could take a nap on it.

Darn...forgot the newlines. "Granted, it is more difficult to program something (from the ground up) that runs distributed, than it is to program something that runs on a giant 2048-way box."

Not quite. You don't use threads on Crays. I've done some work at the AHPCRC and the parallel programming done there is some combination of the following: MPI, UPC, Co-Array Fortran, OpenMP. Writing an MPI app for 2 processors is as easy as writing one for 4096.

FWIW, the AHPCRC's Cray X1 is 152 in the Top 500.

"Granted, it is more difficult to program something (from the ground up) that runs distributed, than it is to program something that runs on a giant 2048-way box." Not quite. You don't use threads on Crays. I've done some work at the AHPCRC and the parallel programming done there is some combination of the following: MPI, UPC, Co-Array Fortran, OpenMP. Writing an MPI app for 2 processors is as easy as writing one for 4096.

I found this interesting article on google when looking up some X1 specs. Brings up some other majpor factors (space, electricity, heat, etc.) for comparing the two.

What is the Linpack's "Highly Parallel Computing" benchmark? The third benchmark is called the Highly Parallel Computing Benchmark and can be found in Table 3 of the Benchmark Report. (This is the benchmark use for the Top500 report). This benchmark attempts to measure the best performance of a machine in solving a system of equations. The problem size and software can be chosen to produce the best performance. http://www.netlib.org/benchmark/hpl/

Nonetheless, with some restrictive assumptions on the interconnection network, the algorithm described here and its attached implementation are scalable in the sense that their parallel efficiency is maintained constant with respect to the per processor memory usage.