« 4973, the prime of the beast | Main | Too much work at work to do have any Mulle progress »

Safari.app benchmark

I ran Safari with the tracer, produced 20 MBs worth of malloc library info and then let Apple's malloc compete against the Mullocator. The results are a bit disappointing as the mullocator is not twice as fast as Safari, only 60% faster, which means that i will have to look into tweaking some parameters, as twice as fast really should be doable.

I suspect that I can improve realloc some, as the memcpy routine is never inlined apparently on PPC, by wrtiting my own memcpy. Looked around the net and already found an Altivec memcpy. I wonder how much memory one needs to move, so that it becomes a win, as context switches for non-Altivec code are bound to become much slower. (Sames goes for using doubles in memcpy).

Comments (4)

Why invest work in such a project? As you may have noticed, Apple was actively searching for someone extending the Objective-C runtime for garbage collection!

Obviously this is going to use BoehmGC:

http://www.hpl.hp.com/personal/Hans_Boehm/gc/index.html

which is actually faster than most system malloc implementation! So just ensure that Boehm GC works as a drop in replacement and we get:
FAST malloc,
garbage collection
and
leak tracking
for free.

Blub:

What an excellent idea! Instead of wasting resources on a reimplementation of malloc, we can reuse a well proven malloc library.
Hell, this is even used for the gcj Java runtime, so it has to
be rock solid, being taken as a base for a language relying
on GC.

SpiderMonkey:

Hm. All this seems quite nice, but does Boehm GC actually use DOM to represent the malloc management structures like Mullocator?
A malloc library without the ability to serialize it's management structures into XML for further processing using XSLT into XUL UIs seems pretty useless to me!
And doesn't JavaScript already come with J2EE compliant memory management anyway?

Nat!:

Garbage collection! Phoeey :)

Who cares about "faster than most", when you can be "faster than all" ? To quote ZNeK, "Being Nr. 1 isn't everything, but being Nr. 2 is nothing."

Thanks for the DOM hint, i think I'll secretly rewrite my code.

About

This page contains a single entry from the blog posted on July 16, 2003 1:44 AM.

The previous post in this blog was 4973, the prime of the beast.

The next post in this blog is Too much work at work to do have any Mulle progress.

Many more can be found on the main index page or by looking through the archives.

Powered by
Movable Type 3.34