Is std.perf still useful?

[Jonathan M Davis]

On 2011-05-25 11:27, Sean Kelly wrote:
> On May 24, 2011, at 10:37 PM, Jonathan M Davis wrote:
> > On 2011-05-24 22:15, Andrej Mitrovic wrote:
> >> It seems to be largely replaced by std.datetime.stopwatch and friends,
> >> right? Although I'm not sure IF those thread-specific timers are
> >> available in datetime.
> >> 
> >> If it's not going to be deprecated let's at least make its
> >> documentation linked from somewhere. This is a 2 year old bug:
> >> http://d.puremagic.com/issues/show_bug.cgi?id=3145
> > 
> > It's going away. If it core.time and std.datetime are missing
> > functionality that it provides that you think should be provided in
> > Phobos, then create an enhancement request.
> 
> std.datetime is so large that I had a hard time finding documentation for
> how to use it for perf timings.  It might be nice if there were a topical
> index at the top of the page.

It's on my todo list, though I don't know when I'll get to it. Certainly, the 
links that ddoc sets up are horrible, since it has no concept of hiearchy. 
Howevr, the two key timing-related items are StopWatch and benchmark. The 
timing stuff was written by Shoo and is really the only thing in std.datetime 
that I didn't write. So, I'm not super-familiar with how they work.

In any case, there will be something similar to what std.algorithm now has at 
its top at the top of std.datetime at some point (hopefully by the next 
release), but it hasn't been done yet.

- Jonathan M Davis