Re: [Mono-dev] IOC performance much slower than Windows

2015-12-17 Thread jchannon
Here's a link to the results from the profiler:

https://gist.github.com/jchannon/e0ef11aba9d29fc06268



--
View this message in context: 
http://mono.1490590.n4.nabble.com/IOC-performance-much-slower-than-Windows-tp4667103p4667176.html
Sent from the Mono - Dev mailing list archive at Nabble.com.
___
Mono-devel-list mailing list
Mono-devel-list@lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list


Re: [Mono-dev] IOC performance much slower than Windows

2015-12-17 Thread jchannon
Miguel responded but doesnt look like it came into the mailing list so here's
whats missing


> To be honest I don't k is where to start. Here's a poor attempt and repro.
> https://github.com/jchannon/MediatRTest
> 
> In another test I also tried registering and resolving types with some IOC
> Containers but the first hit was 10ms on Windows/100ms on Mono but all
> subsequent calls were 0ms. I assume due to some sort of run time caching.
> However load testing a web app where types are resolved it's much slower
> on mono compared to Windows 


> Hello,
> 
> Or the first run triggers a lot more work on the part of the JIT.   
> 
> It would be nice if we can exercise the slow codepath and profile that
> with the statistical profiler.
> 
> Also, interfaces methods in general are much slower to invoke than virtual
> methods, and .NET might have a better implementation of that.





--
View this message in context: 
http://mono.1490590.n4.nabble.com/IOC-performance-much-slower-than-Windows-tp4667103p4667175.html
Sent from the Mono - Dev mailing list archive at Nabble.com.
___
Mono-devel-list mailing list
Mono-devel-list@lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list


Re: [Mono-dev] IOC performance much slower than Windows

2015-12-14 Thread Miguel de Icaza
Hello,

The first step would be to isolate the root cause for the problem.

Once we have that information, from this email alone it is not possible to
determine what the culprit is.

On Mon, Dec 7, 2015 at 9:49 AM, Jonathan Channon  wrote:

> Hi,
>
> I am deploying HTTP API apps into production on Linux with Mono.
>
> With some load testing on OSX I have discovered that when I inject a
> dependency into a HTTP handler the requests per second(req/s) are much
> slower than when I have none.
>
> I have done some test console apps using a couple of containers
> registering some interfaces and types and then calling Resolve. I have
> timed these and I can see mono is slower than Windows.
>
> Is there anything I can give you to make this a priority for mono
> development as there clearly is a problem on mono with reflection and/or
> expressions.
>
> Thanks
>
> ___
> Mono-devel-list mailing list
> Mono-devel-list@lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>
>
___
Mono-devel-list mailing list
Mono-devel-list@lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list


[Mono-dev] IOC performance much slower than Windows

2015-12-07 Thread Jonathan Channon
Hi,

I am deploying HTTP API apps into production on Linux with Mono.

With some load testing on OSX I have discovered that when I inject a
dependency into a HTTP handler the requests per second(req/s) are much
slower than when I have none.

I have done some test console apps using a couple of containers registering
some interfaces and types and then calling Resolve. I have timed these and
I can see mono is slower than Windows.

Is there anything I can give you to make this a priority for mono
development as there clearly is a problem on mono with reflection and/or
expressions.

Thanks
___
Mono-devel-list mailing list
Mono-devel-list@lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list