July 23, 2004

The best part of a day

Sometimes you read or hear something, and realize that things have changed. You get that "we're not in Kansas anymore" feeling. At a project meeting this morning, one of the grad students was explaining how he'd spent his week:

"...so we ran [Ryan's] gene tagger and compared the results with Yang's list of alternative gene names. It was kind of slow, though -- we did about two million documents, and it took the best part of a day."

The documents were abstracts, so if you do the math, this is only about 12,000 words/second. And he probably ran it on our linux cluster, so you could divide 12K by N for however many nodes he used.

But all the same, when I think about my first experience with computers...

It was on an early PDP-9 with 8192 18-bit words of core memory, whose mass storage devices were paper tape and a magnetic drum whose storage capacity was 256KB, if I remember right... It could do a fixed-point addition in two microseconds -- multiplication was done in software. There was an CRT display driven by a special-purpose vector system. For a while after the machine arrived, it was fairly easy to get console time since there wasn't any software yet.

I wrote a program for measuring durations in speech waveforms.

Posted by Mark Liberman at July 23, 2004 02:59 PM