2 comments

[ 3.5 ms ] story [ 14.0 ms ] thread
"lie in plain sight"
http://journals.ametsoc.org/doi/abs/10.1175/MWR-D-12-00352.1

Of course, to draw any conclusion at all assumes that actually have "data" in their output, and haven't been sloppy with their floating point arithmetic - especially how re-ordering calculations (due to parallel processing) is non-transitive in FP.

anybody writing a big, comples model needs to know about fp:

https://randomascii.wordpress.com/2013/07/16/floating-point-...

http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.ht...