GC.calloc with random bits causes slowdown, also seen in built in AA

[Steven Schveighoffer]

On Sun, 14 Mar 2010 21:03:38 -0400, Michael Rynn  
<michaelrynn at optusnet.com.au> wrote:

>
>> a major overhaul, and I introduced a nasty bug that causes this.  Please
>> apply the fix identified here:
>> http://www.digitalmars.com/webnews/newsgroups.php?
> art_group=digitalmars.D.announce&article_id=17841
>>
>> Sorry again.  I hope this can be fixed soon.
>>
>> -Steve
>
> This particular progressive slowdown of built-in AA is not due to the bug
> introduced with 2-040.  It was present in 2-039, and is also observable
> in the dmd, druntime (rev 265 for lifetime.d), phobos, built from latest
> SVN for the fixed 2-041
>
> The pydict performance still 25% worse than fastest builtin AA, but does
> not slow anything down.  Once the builtin AA tests have been run,
> everything slows down, including the pydict.

Sorry, I meant the crashing bug was introduced :)  I can't take credit for  
the original issue you were chasing.

-Steve