From: Jim Phillips (jim_at_ks.uiuc.edu)
Date: Mon Feb 02 2015 - 00:15:09 CST
It's best to actually run on the actual CPU counts you are considering in
order to measure parallel speedup. Also, run long enough that you see
several "Benchmark time:" lines, which are printed after initial load
balancing so should be more representative of a long run.
On Sun, 1 Feb 2015, Dawid das wrote:
> Dear NAMD experts,
> I have question with quite obvious answer but I want to make sure.
> How should I understand this kind of message in *log file:
> Info: Initial time: 1 CPUs 0.042815 s/step 0.247772 days/ns 62.082 MB memory
> Does it mean that when I use 1CPU then 0.25 days will take for 1 ns of
> So I can easily calculate how much time my simulation will take on e.g. 4
> CPUs and for 5 ns?
> Best wishes,
> Dawid Grabarek
This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:20:52 CST