[Mono-list] Performance benchmark

[Rodrigo Kumpera]

On Sat, Feb 27, 2010 at 5:59 AM, James Mansion <james at mansionfamily.plus.com
> wrote:

> Rodrigo Kumpera wrote:
>
>> Things like page faults and hardware traps (raising a NRE) are
>> significantly slower when running under a VMM.
>>
> But this code is looping in user space and should not make many. Why do you
> think it will make
> many in this case?
>
>
>>        Mono runs faster on linux than windows, but the virtualization
>>        has it's costs.
>>
>>    For compute tasks that aren't dependent on glibc vs MS libc issues
>>    or kernel calls? Why?
>>
>>
>> Because mono is better optimized for linux, simple as that. Some codepaths
>> are slightly faster under linux.
>>
>
> In its runtime library interaction with system services - yes. This example
> is compute
> bound in user space.
>
> How is the code generation 'better optimized for linux' such that the
> generated code for
> this source will be different? Doesn't make sense to me.
>

The differences are in areas such as synchronization and thread-local
access.
We can generate more efficient thread-local access code on linux than on
windows,
for example.

Memory allocation and call into native/runtime functions both require fast
thread local access.
So a lot of under the hood code that the runtime needs to generate is faster
under linux.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ximian.com/pipermail/mono-list/attachments/20100227/c82eb980/attachment-0001.html