[Mono-devel-list] Performance Test

[Paolo Molaro]

On 06/30/03 Marcus wrote:
> On my machine, Mono's new JIT is 6X faster than another (non-JIT) 
> implementation (and 7X with -O=all). I do not have a Windows platform to 

Thanks for the info. Your results are consistent with the results I get
also in several other benchmarks: the mono JIT is from 3 to 12-15 times
faster than the interpreter from Portable.net.

> verify .NET's performance myself, but based on someone else's testing, Mono 
> takes about 50% longer than .NET.

We basically haven't optimized floating point operations yet, so a consistent
speed difference is expected. Running your program on my system,
the current jit is already 15% faster than the old jit and 8% faster
that the new jit from the 0.24 release. We have lots of room for
improvements in the jit, since we haven't yet implemented some important
optimization and some of the existing ones still need refining.
We also need to spend more time looking at the generated code
for code patterns that can be optimized better.
It looks like, though, that we're improving steadily with each release
and that is good:-)

lupus

-- 
-----------------------------------------------------------------
lupus at debian.org                                     debian/rules
lupus at ximian.com                             Monkeys do it better