On 06/02/2013 06:44 PM, Andrei Alexandrescu wrote: > On 6/2/13 12:16 PM, John Colvin wrote: >> A lot of HPC scientific code is, at best, horribly fragile. > > Reminds me of that internal joke at www.llnl.gov. "If we make a mistake, > millions of people will live." True story.
A story I heard from a lecturer of mine -- one of the interesting little factors of the Cold War was how Soviet Russia managed to keep parity with NATO in terms of missile guidance and other computer-related military technologies. The US had far superior hardware, so it was both a concern and a mystery. What came out after the end of the Cold War was quite impressive -- Soviet scientists had realized very well that they couldn't compete on the hardware front and so had focused a very intense effort on really, really efficient algorithms that squeezed every drop of performance out of the hardware they had available, far greater performance than Western computer scientists had ever imagined possible.