Today I am teaching my special one-day course in supercomputing to the student interns. It is becoming an annual event. (I taught it last year at about this time too.)
This year it is going to be even better. The people at the lab's supercomputing center offered me the use of one of their smaller machines for the students. Also, they're going to give them a 45-minute tour of the facility. How cool is that?!?!
Subscribe to:
Post Comments (Atom)
5 comments:
Sounds like fun. Can you post an online version for your friends in cyberspace?
Did you tell them that today's supercomputer is next decade's PC?
The first supercomputer I ever used (as the sole user) was a megaflop megabyte box circa the 70s. The last one I used (as the sole user) was a gigaflop gigabyte box circa the 90s.
I am writing this on a multi-GHz (not the same thing, particularly if running bloatware rather than Unix) multi-gigabyte desktop. That might give you a hint why you have to say you won't do certain things on this machine, or export it to those who would do them. Nuclear weapons were designed and codes were broken with those earlier ones.
Dr Pion, I did tell them about the ever-increasing power of supercomputers. I also showed them a picture of my favorite machine of all times, the Cray I (which also doubles as lobby furniture). If I could get my hands on one of them, I would totally set it up in my living room! Even though yeah, my underpowered laptop is probably more powerful.
It was pretty amazing what they could do back in the day with so little compute power (relatively speaking). Anybody who wants a nuke enough could easily design it without any export-controlled machines or software.
I have a Cray, but it is an X-MP and made of cardboard. (It is the coolest calendar you ever saw.) That is one more thing I don't have to fit into my meme. Thanks for the tag.
A few more:
That Cray 1 would heat your house in the dead of winter with the windows open.
The great fight was always over who got serial number 1, but someone got one that did not have a serial number. Or so I have heard.
Hans Bethe reportedly said that an "interesting" problem takes 100 hours on the fastest computer available. That changes every year, of course.
My favorite was the Cray-2, affectionately known as "bubbles" because it was cooled with clear synthetic blood in a clear case.
Oh, yes, and that the Cray 1 multiplication unit did not always guarantee that multiplication was commutative. (They left out part of the shift/add pyramid to save a lot of hardware.) Since division was "invert and multiply", there were certain simple divisions (I think one of them was 3) that would generate a flaw detected by the paranoia program.
Post a Comment