[Todd Madson]

I just found out today there is a G5 optimized version of Firefox.
(see this url: http://www.beatnikpad.com/archives/2.../20/firefox106)

Is there a list of G5 optimized applications out there? There was once a G4
optimized applications list at apple.com. I didn't see (or didn't find) a G5
version.

Anyone with G5s have the scoop?

[sonicularulus]

i find a g5 optimized version of firefox very funny.
why would an internet browser use a 64 bit, dual processor system to surf the internet.lol
you can surf the internet just fine with a computer built in 1998.

[Todd Madson]

Right. Still, every little optimization helps I suppose.

Here's Cinebenchs' G5 optimized benchmarker:
http://www.cinebench.com/

[OogaBooga]

Originally Posted by sonicularulus

i find a g5 optimized version of firefox very funny.
why would an internet browser use a 64 bit, dual processor system to surf the internet.lol
you can surf the internet just fine with a computer built in 1998.

I'm not sure if you've used OS X lately, but apparently it takes a fast machine just to get smooth scrolling.

[Millennium]

Originally Posted by sonicularulus

i find a g5 optimized version of firefox very funny.
why would an internet browser use a 64 bit, dual processor system to surf the internet.lol

Particularly since the "G5-optimization" of Firefox doesn't take any more advantage of the 64-bit or dual-processor architectures than the nonoptimized versions. That said, I've used it before, and it does seem to start up significantly faster. That's just about the only area where I see a speed advantage, though.

you can surf the internet just fine with a computer built in 1998.

Indeed you can, except for things which require Flash and such.

[yukon]

I agree, GCC optimizations (for speed) usually aren't any advantage unless the processor is slow to begin with, optimized versions for the Apple G3s and the earlier G4s would be of more use.

But still, I've found Firefox to be insanely slow on OS X, so slow that I could count out 5 seconds to open a new window, three seconds for a new tab, amazing friends of mine. I also use Firefox on Windows and Linux, on Linux it's much faster, but on Windows it's blazing fast and I don't think it could be improved much there. Been waiting for 1.1 to fix the OS X version, it's pushed back constantly, so I'm annoyed.

Anything OSS that you use, you can recompile with optimizations for your specific machine. Like Adium, Camino/firefox/mozilla, MPlayer (ooh, painful), etc. I've been trying to compile Camino for a while, got it at one point but it kept filling the console with JS errors so I stupidly trashed it, now I can't compile Camino or Firefox (compiling NSPR is failing). I'm told that Apple optimizes OS X with -Os, so theoretically you could compile GNU and Darwin stuff with -O3 and replace some.

[Catfish_Man]

Apple claims, iirc, that -Os often results in better performance, due to treating caches and bus bandwidth more nicely. I haven't actually attempted to test this though.

[yukon]

I had thought that -Os was -O3 but with additional space saving optimizations, but lately I've been told by multiple people that Os removes some optimizations since they increase file size, and that -Os really is for embedded systems rather than desktops. OS X comes on a DVD and disk space isn't a big issue anymore, so I was surprised when I was told Apple used Os....they have their own fork so Os might do something different for them, who knows what they do in-house.

You could be right, my poor 100mhz bus needs all the help it can get. The most I've tested it on was Simpletext, I noticed no difference in speed or even size....I figure I use Simpletext so often that I can allow myself to go crazy on optimizations, it's a joke really for a 188k program. In any case, I think I might find my G4 someday running Gentoo, since I found recompiling things like Samba resulted in a lot less CPU usage.

[wataru]

Originally Posted by yukon

But still, I've found Firefox to be insanely slow on OS X, so slow that I could count out 5 seconds to open a new window, three seconds for a new tab

Sounds like something's wrong with your setup. Opening a new window is a little slow for me, but tabs are instantaneous on an 800MHz iBook G4.

Try a recent nightly. They're much faster than 1.0.x.

[Millennium]

Originally Posted by yukon

OS X comes on a DVD and disk space isn't a big issue anymore, so I was surprised when I was told Apple used Os....they have their own fork so Os might do something different for them, who knows what they do in-house.

Their source for GCC is still open, even though it's forked (they have to do this, because of the GPL). The fork you speak of doesn't actually have anything to do with -Os; it's more about the Objective-C and Objective-C++ compilers.

That said, I'm also surprised about the -Os usage. It shouldn't cause much in the way of speed gains; if anything there should be a performance penalty. The only way around that would be if the Objective-C runtime were to have become so bloated that the speed gains from -O2 or -O3 were outweighed by the performance penalty of constant VM swapping. I find it difficult to believe that this would have happened, though. The Objective-C runtime is by necessity one of the best-optimized parts of the entire operating system.

That said, there may be something in the architecture which makes -Os optimization give some kind of speed boost. I have no idea how this would work, but if it did then it would be some kind of optimization holy grail, because it would get around the speed-versus-space tradeoff that one usually has to make when optimizing. If Apple had really done that, it would have made far bigger news than we've seen, and there would be a mad flurry of patents that doesn't seem to have materialized.

[yukon]

With the GCC fork, I had figured it could be "open source" like XNU is open source in OS X, it's believed that Apple compiles in extra closed source bits so that a freshly compiled version of XNU might not be able to run OS X (with drivers installed). Of course GCC is GPL while XNU is APSL (google says so anyway), so I guess I was wrong.

That stuff about Apple using -Os could be fake, I believe the story was "A guy working at Apple witnessed it and told someone", so that's really reliable....still possible though. I don't even know where I read it, hopefully not /.

[Ganesha]

Well webkit/webcore/javascriptcore all build with -Os.

[proton]

Apple does recommend that in some cases optimising for size will produce faster code than other optimisation options. One of the places this is stated is The GNU Compiler Collection on Mac OS X (look after the "Optimization" header).

The basic theory is fairly simple, smaller code is more likely to fit in the CPU's L1 instruction cache or L2 cache, and therefore can be faster in some cases. It's very much dependent on a particular piece of code if this is true, but in a surprising number of cases it does actually help.

As to the theories that XNU is compiled with special closed Apple additions, this isn't true. What Apple does do is provide some of the various kernel extensions (kexts) only in binary forms. This is particularly true for a large number of the device drivers (things like Video cards, Airport and Sound cards for example).

- proton