News for hackers. Stuff that matters.

A comment on Math Breakthrough Speeds Supercomputer Simulations.

Re:Math Breakthrough Speeds Supercomputer Simulations by marmaduke on Saturday October 19, 2019 @04:53PM (#21300697)
The article doesn't seem to say what achieved the speed up. If they take a larger time step, they've used a special feature of molecular dynamics or some other assumption to achieve that ('no free lunch' or some such)
[ Parent | on HN ]
Re:Math Breakthrough Speeds Supercomputer Simulations by garganzol on Sunday October 20, 2019 @04:44AM (#21302938)
I presume that instead of calculating each and every "point" and "particle" in discreet form, they find the statistically correct result instead.

Probably something similar to Vitter algorithm https://getkerf.wordpress.com/2016/03/30/the-best-algorithm-...

[ Parent | on HN ]
Re:Math Breakthrough Speeds Supercomputer Simulations by bonoboTP on Saturday October 19, 2019 @09:22PM (#21301758)
It doesn't say anything concrete because its purpose isn't to inform the reader. This is a PR marketing press release by the university advertising their impact.
[ Parent | on HN ]