[Mono-list] Performance benchmark

Miguel de Icaza miguel at novell.com
Sun Feb 28 17:00:21 EST 2010


Hello,

    The change to the old decimal code does not show any regressions using
either Mono's test suite (mscorlib, System and System.Data) or the test
suite that we received from Microsoft for Silverlight and Rotor.

     In our internal mailing list, I found a reference to a discussion in
May of 2004 where Atsushi and Uma were discussing all the pending failures
in the test suite, and Atsushi set out to fix both mscorlib as well as
System.Data.   Those correspond to the changes of May 11th 2004.

On Sun, Feb 28, 2010 at 4:21 PM, Miguel de Icaza <miguel at novell.com> wrote:

> I modified my local copy to use the unmanaged double to decimal conversion,
> these are the results:
>
> Trunk, with the current implementation:
> mono$ mono 5.exe
> first=00:00:01.9547310
> decimal=00:00:00.7654980
> justpow=00:00:00.1711550
> second=00:00:01.6576570
> Combined=00:00:03.5090000
>
> Trunk, with local changes to use the original double to decimal code:
> mono$ mono 5.exe
> first=00:00:00.3222750
> decimal=00:00:00.2351060
> justpow=00:00:00.1768560
> second=00:00:00.5998440
> Combined=00:00:00.5855580
>
> Now I have to find the actual regressions.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ximian.com/pipermail/mono-list/attachments/20100228/0fc7c015/attachment.html 


More information about the Mono-list mailing list