Latest / Deep End Era. Engineering with Sankalp

Many Hands, One Answer: Parallel and Concurrent Algorithms, from Work and Span to the ABA Problem
A fictional payments engineer watches a sixty-four-core server finish a nightly settlement job only three times faster than her laptop. A fictional game-engine programmer's lock-free job queue works perfectly for eleven months and then, once a fortnight, hands the same task to two threads. One problem is arithmetic about how much of the work can actually run at once; the other is a value that changed and changed back while nobody was looking. From Gene Amdahl's 1967 conference talk to a 1997 rover on Mars that kept rebooting itself, this is the art of getting many hands to produce one right…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.