Le 26 nov. 2010 à 16:05, Alastair Houghton a écrit :
> That article is pretty poor, actually, and it doesn't show anything about the thread design since it doesn't actually test the threading system. Instead, it tests process creation (fork()-ing). On Linux, last I looked, threads were really just processes that shared address space with their parent, but on Mac OS X that is very much not the case; processes are implemented on top of Mach tasks, and have a set of Mach threads associated with them.
Indeed, this doesn't seem fair.
> It is also comparing G5 PowerPC Mac OS X with Linux on x86, which is just plain ridiculous. It doesn't make sense to compare except on the same hardware (and yes, Linux will run on a G5).
Btw, this kernel speed thing is really just for the gallery. Just like BeOS, which actually had a quite slow kernel compared to Linux, but blows it in terms of perceived reactivity. But this is about the design of the rest of the OS, threading used at the right places and so on... (and until they multithread Xorg they will still have issues)
If you really want to benchmark thread context switching though here is an IBM test that used to be published on their site, that I ported to BeOS here:
You might want to add the code for OSX (though it might actually work already using pthread).
But as you said, this must be run on the same machine for all OSes you test, else there is no point.
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden