[Mono-list] Why is mono so slow at math compared to .Net?

Jon Harrop jon at ffconsultancy.com
Wed Feb 18 16:40:04 EST 2009


On Wednesday 18 February 2009 21:09:42 you wrote:
> On Wed, Feb 18, 2009 at 6:03 PM, Jon Harrop <jon at ffconsultancy.com> wrote:
> > Mono 2.2 is >50% slower than .NET on four of the five subtests in
> > SciMark2:
> >
> >  http://flyingfrogblog.blogspot.com/2009/01/mono-22.html
> >
> > In particular, Mono is 2x slower than .NET on the FFT test.
>
> Your benchmarks were ran on windows...

That is incorrect. My Mono benchmarks were run on Linux.

> The F# version of Scimark uses a lot of tail recursion

That is also incorrect. The F# version of Scimark2 is a direct translation of 
the Java and uses imperative loops and mutation throughout.

> and this is an area that mono has issues.

While Mono certainly has problems with tail calls and garbage collection the 
problem here is actually the quality of the code generator in Mono 2.2 
compared to, say, LLVM.

> Most code on the .NET platform is written in C#. 

Most *numerical* code on the .NET platform is written in F# because that 
language was specifically designed by Microsoft for technical computing.

-- 
Dr Jon Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/?e


More information about the Mono-list mailing list