[Mono-dev] [PATCH] Mono DTrace provider

Joachim Ante joe at unity3d.com
Wed May 28 21:48:59 EDT 2008


Hi,

What is this patch capable of at the moment. Can I do managed code  
performance profiling with it by using Apple's Instruments tools?
Or does this require adding specific probes for performance profiling.

Best regards,
Joachim Ante

>
>> Some comments about the patch:
>> - I would vote for the --enable style instead of the --disable style,
>> cause enabling
>>  things by default is known to break builds on some environments. If
>> things look ok,
>>  we can enable it later.
>
> Let me add that I consider *both* the Solaris-related changes to the
> build system and the proposed 'mono' provider experimental.
>
> The lack of any pragmas in the .d provider file implies that the
> provider is "private" and "unstable". Especially I still consider the
> arguments and probe naming unstable. Once we've found an acceptable
> solution for the build issues, so that we can further experiment, it
> might be a good idea to post the provider on the DTrace list for  
> review.
>
> On my TODO list still is testing this on Solaris 10 i86/amd64. Solaris
> 10 will likely have an older version of DTrace, and in the worst case
> it doesn't support generating a header file. If so, that would be a
> good reason to proceed with manually architecting a header file, as
> inspired by Zoltan.
>
> Another point to consider is Novell's binary builds for the Mac. I
> haven't tested this, but I'd assume inserting probes on the Mac is not
> backwards compatible and thus would need to be disabled for the
> official builds, which target v10.4 (and were still inofficially
> compatible with v10.3.9 last time I checked), whereas DTrace is new in
> v10.5. If we'd choose to enable it by default, it would need to be
> disabled there.
>
> Andreas
>
> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list at lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list



More information about the Mono-devel-list mailing list