Re: apple's objc runtime on linux?

[Benhur Stein]

Citando Alexander Malmberg <alexander@malmberg.org>:


> Interesting. I've done a fair amount of benchmarking of this at
> different times, and these numbers are fairly consistent with mine. On
> what architecture is this? How did you test it?
On a 2.4GHz P4, running linux.
I attach x.m and y.c used to do this.

Inside the main loop, there are, commented, the various options.
I compiled one of them with gcc -O3 -S -fno-unroll-loops, separated
the code for sending the message from the rest, and substituted it 
with the code generated from the same compiler line for each of the
other options. The function ml is basicly a copy of objc_msg_lookup,
copied in x.m so that the compiler inlines it. The idea was to see
what could be obtained from such a change in message sending code in
the compiler, without doing any manual assembly language optimization.
Sure, this says nothing on cache effects from such an augmentation
in the generated code (to say the least).

> > Sure, if the method/function does something, this numbers become
> > closer...
> 
> True. And when you really need the speed, IMP caching isn't hard to do.
> :)
Yes, but it has been a bit overused inside gnustep, I think.

Benhur