MGS4 is on Dual-layer blu ray disc. XBOX360 can't  hold that on 1 disc , so 3 or 4 disc is not so bogus.

 I listened throught the whole podcast, and there is no mention of the game sizeor anything like this

"Guns of the Patriots is the first PS3 game that uses a full 50GB dual layer" < From Wikipedia

Around 17:00 they did mention the sound languajes and limitations of disc size, but none of them refers to the 50Gb thing...

I read through this entire thread and didn't see any mention of this so sorry if somebody pointed this out already.

WTF are you talking about? x264 isn't a compression codec, it is an AVC encoder. The compression codec it works with is called h264, which happens to be the same codec that is used in Blu-ray movies.

That's because you said nothing apart from a short cheapshot, as you don't want to go into detail. What he said supports what I said, you know the meaning of the word "allow", don't you? IMO let's not go there, let's stay polite and try to discuss things more in depth.

You agree, he's an experienced and knowledgeable developer, don't you?

Don't you agree that using the approach of branch hints and elemination is of benefit to the 360's cores as well? So actually the best measures to be taken to use the SPUs efficiently also result into code which will run more efficiently on PPE/Xenon?

@MikeB: You must have missed this post in which I explained you why you're wrong and misinterpreting what that developer said:

"Dynamic branch hints are similar to delayed jumps, something I use when programming with DSPs. It's not the same thing as never mispredicting branches, it just means that sometimes you can decide a branch early."

Seriously, until you're sure you fully understand the quotes you see, don't copy/paste them all the time. I'd much rather talk with someone who understands what they're talking about, such as the developers you so often misinterpret to fit your goals.

PS: Branch elimination is an optimization strategy like any other; sometimes it works, sometimes it doesn't. An experienced developer can surely use it to mitigate *some* of the problems of inexisting dynamic branch prediction. That's not what you were saying though.

I'm not really a codec expert to argue with you, but I am quite familiar with the movie rip scene on the net and there is a lot of torrents available that take a bluray movie and take it down to 2-3 gigs. Discussions on some torrent forums established that very little quality is lost...

Most of those torrents have name such as MovieName.720p.BluRay.x264

So I assumed x264 is the codec used. So you may be right...however, the size reduction using this compression codec is quite dramatic.

@ NJ5

Branch hints and branch prediction serve the same purpose which is to tell the CPU ahead of the branch execution time what code to fetch. Some code is very unpredictable for the branch predictor, then there's the advantage towards the manual approach best used for the SPUs, resulting into gains on for other CPUs as well.

It's often compared to changing gears in cars, if you are a good driver you will be able to yield better results changing gears manually.

Branch elimination approaches is beneficial as well, like Mike Acton pointed out. Redesigning your code, you can achieve exceptional results with the SPUs and better results with other CPUs as well.

What you fail to understand is that dynamic branch hints aren't always possible to use. As the developer you quoted said, it only works if you can put valuable instructions in between the branch hint and the actual branch (which is why I said dynamic branch hints are similar to delayed jumps, something which has existed for ages and doesn't solve all the problems of static branch prediction).

Regarding your last paragraph, I reiterate what I said - branch elimination is sometimes useful, even in processors with dynamic branch prediction. As with branch hints, branch elimination is nothing new, nor a silver bullet.

The key word in guidelines for optimizing software is "sometimes", everything must be seen on a case by case basis.

PS: Just tell me, what was the last time you programmed anything in Assembly? How many different CPUs have you programmed for?

@ NJ5

He states two things:

- Oportunity to never 'mispredict' a branch

- and to not pay any cost as long as ...

And neither is branch prediction. Understand?

Amiga days, some time ago (unless you count VP code, a few years back which was more like a mix of C/C++ and ASM). Programming isn't my profession, but I know some good developers I discuss things with.

The "as long as" and what's after it is precisely what you failed to understand. Is it that hard to admit you were wrong, or don't you even understand why you were wrong yet?

No, branch prediction isn't a silver bullet either, which only makes your initial assertion even more laughable. Effectively, you were saying that dynamic branch hints are perfect for every situation.

I figured you were not really a programmer. Your posts show that most of your knowledge comes from hearsay and misinterpretation of valid sources. My work in this thread is more than done, if you hadn't continued dodging what I said I would have stopped even earlier...