3 ms·
Ah yes, pop count. Back in my days as a CPU logic designer on refrigerated, walk-in, scientific mainframes, I was working on a vector processor that was intend
by dbcurtis 3y ago
Ah yes, pop count. Back in my days as a CPU logic designer on refrigerated, walk-in, scientific mainframes, I was working on a vector processor that was intended to be a competitor to Cray-1, but sank due to poor marketing and a collapse in oil prospecting... but I digress. A few of us on the team had come from previous CPUs at other companies, and argued with the architecture committee that we should have a vector pop count instruction in order to get some interest from No Such Agency. Shot down, of course.
When serial No. 1 was in initial benchmarking, two guys named Frank and George with No Last Names, who lived somewhere near Fort Mead, said they would buy at least one if we added vector pop count.
- bruce343434 3y agoI've read this story before but can't remember where
- dbcurtis 3y agoI suspect it played out multiple times at different companies.
- angiosperm 3y agoPopcount eventually gets added to every architecture that lacks it, always at great expense. You would think people would have learned by now. And, no, not typically just because the spooks want it.
- NotYourLawyer 3y agoWhy, then?
- angiosperm 3y agoBecause there are numerous order-of-magnitude optimizations it enables, that are useful far beyond cryptanalysis. People mention spooks because the spooks have often had access to enough money to force the issue.