Testing some singleton implementations

[Stanislav Blinov]

On Tuesday, 4 February 2014 at 14:23:51 UTC, Andrej Mitrovic 
wrote:
> On 2/4/14, Stanislav Blinov <stanislav.blinov at gmail.com> wrote:
>> Have you also included fixes from
>> http://forum.dlang.org/post/khidcgetalmguhassvqm@forum.dlang.org 
>> ?
>
> I haven't figured out exactly what you're trying to swap there. 
> Do you
> have a full example:

Both atomicLoad and atomicStore use raw MemoryOrder, and also the 
atomicStore is out of the synchronized {} section:

http://dpaste.dzfl.pl/291abc51bb0e

>> How do the test results look in multiple runs? Is 
>> AtomicSingleton
>> always faster than SyncSingleton on Windows?
>
> Pretty much. I'm getting reliable results.

Interesting. As you've seen, for me on Linux it's 50/50.

> But I'm not a statistics pro (and yeah I've read
> http://zedshaw.com/essays/programmer_stats.html - still doesn't 
> make me a pro).

Same here :)