cocoaforge

I know SVN shouldn't be discussed here... but to be fair, this applies to ALL builds of Adium. So once the Mac community really got a hold of the Firefox source code, G4 and G5 optimized builds surfaced... I feel like there are a few compiler extensions to produce this? Correct? Is it possible to do this on our own w/ Adium?

-Kevin

i don't see what needs to be optimized on adium?

There is no need for processor specific optimizations. The system itself is optimized as needed, not to mention Adium relies on a fair number of platform-agnostic code parts that won't processor optimize. Let's not go hunting for more work when it's not necessary.

G5-specific optimization should only occur in apps that need to work with the extended memory addresses (such as a massive scientific calculations and whatnot that can use more space in which to store and compute in memory and processor)

bgannin: all they're asking for is -mcpu=7450 -mtune=7450, nothing exotic (and nothing that would do anything noticeable...)

Those g5 optimized firefox builds have started an onslaught of these types of posts in all different types of forums.

Yah i was just looking for the -7450 command line arguments when compiling from source. Obviously AdiumX is pretty snappy on my system, but if there are optimizations to be made, I don't see the harm. My G4 optimized Firefox build is much faster than the standard one. But if you say it is not any faster, I'll take your word for it.
-Kevin

spaceballl wrote:Yah i was just looking for the -7450 command line arguments when compiling from source. Obviously AdiumX is pretty snappy on my system, but if there are optimizations to be made, I don't see the harm. My G4 optimized Firefox build is much faster than the standard one. But if you say it is not any faster, I'll take your word for it.
-Kevin

Most likely they're using some other optimization flags as well. -O3 or -Os instead of -O2 for example. Adium is built with -Os.

You are welcome to build your own G4 or G5 optimized build, but as Brian noted, you won't see much difference (really, you will probably feel NO difference -- we just don't have any code that can be optimized that way). For those reasons, we will not be releasing optimized builds of Adium -- it's unnecessary and even gives the wrong impression.

Man these g3 optimized adium builds rock!!


oh wait, it runs the same on a g4.... and on a g5..


Man my g3 rocks! It's as fast as a g5 is!

The_Tick wrote:Man these g3 optimized adium builds rock!!


oh wait, it runs the same on a g4.... and on a g5..


Man my g3 rocks! It's as fast as a g5 is!

lol got a chuckle out of that one

Okay so thanks for the reply... so maybe this is the wrong place to ask technical questions like this, but... why won't adding the command line options for G4/G5 optimization make it any faster? Is it because Adium is already so low on CPU usage in the first place? Would this make it less? Sorry to bother...

-Kevin

spaceballl wrote:Okay so thanks for the reply... so maybe this is the wrong place to ask technical questions like this, but... why won't adding the command line options for G4/G5 optimization make it any faster? Is it because Adium is already so low on CPU usage in the first place? Would this make it less? Sorry to bother...

No, it's because compiler optimization is essentially pointless. For more information, see http://funroll-loops.org/

G4 and G5 optimization basically helps with:

1) Things that could use AltiVec -- i.e. crazy math operations. We don't do a lot of heavy math.
2) Things which would benefit from being 64-bit optimized (G5's) -- definitely nothing in Adium.
3) Some graphics code -- but the code would probably not be anything in Adium. You might see your contact list update a bit faster, but is that really a problem?

dchoby98 wrote:G4 and G5 optimization basically helps with:

1) Things that could use AltiVec -- i.e. crazy math operations. We don't do a lot of heavy math.
2) Things which would benefit from being 64-bit optimized (G5's) -- definitely nothing in Adium.
3) Some graphics code -- but the code would probably not be anything in Adium. You might see your contact list update a bit faster, but is that really a problem?

Actually... not really. -mcpu and -mtune just change instruction choice and scheduling, so they apply to pretty much any kind of code. That said, the effect is extremely minimal in almost all situations. Altivec, at this point, has to be explicitly used, and is often explicitly scheduled, so it doesn't really apply to that.