Talk:Montgomery reduction: Difference between revisions
→Example numbers?: library
(→Example numbers?: library) |
|||
Line 33:
:: Well, you can't make a code vs. library comparison like this, just as you can't compare apples to Yorkshire Terriers. A serious implemention should have access to the innards to the <code>big</code> (your <code>math/big</code>--is that the weekly instead of release build?) package, which means it should be in nat.go and directly operate on the bits of the machine word slices. As presented, the line <code>a.Rsh(a, 1)</code> alone is enough to kill performance, not to mention <code>a.Add(a, m.m)</code>. --[[User:Ledrug|Ledrug]] 05:28, 24 December 2011 (UTC)
::: As long as both the code and library tests were done on the same hardware, comparison timings shouldn't be a problem. The real difficulty with timings comes when people try to use results from different environments. (One could even get rid of 'time' output entirely and only display relative percentages for in-example comparison) As for identifying which version of dependent software is used, that's what {{tmpl|works with}} is for. --[[User:Short Circuit|Michael Mol]] 15:38, 24 December 2011 (UTC)
| |||