[Mono-dev] Fundamental performance problems with Mono

[zvikag]

Robert Jordan wrote:
> 
> The new GC won't help in this special case.
> 

why not? Even when using the Begin\End API, a generational GC could get rid
of all the garbage this API creates during the fast gen 0 collections (at
least this is what happens with Microsoft's GC on Windows).


Robert Jordan wrote:
> 
> So you're actually blaming a wrapper whose solely intention was
> to gain .NET 2.0 compatibility.
> 

My point is that two things prevent writing a high-performance server in
Mono:
1. The current Begin\End async socket API (you call it 1.1) that creates
lots of garbage
2. The current GC that cannot handle all this garbage

Fix even one of the two (by properly implementing the 2.0 SP1 XXXAsync API
to eliminate garbage, or implement a generational GC) and you fixed the
problem. But better fix both... :)


-- 
View this message in context: http://old.nabble.com/Fundamental-performance-problems-with-Mono-tp27026906p27078968.html
Sent from the Mono - Dev mailing list archive at Nabble.com.