Dolphin, the GameCube and Wii emulator - Forums

(10-04-2013, 06:16 PM)neobrain Wrote: [ -> ]

(10-04-2013, 11:07 AM)NaturalViolence Wrote: [ -> ]The reduction in performance from improved accuracy usually outweighs the improvements from optimizations. This will only become more true with time since they'll slowly run out of significant optimizations to make.

But then again you could say we'll run out of things to fix and could focus on improving performance...

improving performance? do you know how many people with average to slow cpu's would like this
you sir now have my interest

Yeah, just that it's not going to happen from my side :p

yay!

Quote:But then again you could say we'll run out of things to fix and could focus on improving performance...

Now your are talking...

Wouldn't it make more sense to do this with a PS3 or a Xbox 360? After all they both are based on the Power architecture.

No. PS3's CPU uses big endian which Dolphin doesn't support. https://forums.dolphin-emu.org/Thread-dolphin-for-little-endian-gnu-linux-powerpc-with-kvm

It also lacks RAM. PS4/XBox One is definitely a better option but they are still not good enough nevertheless.

All PowerPC CPUs are big-endian, including the one Dolphin's emulating. The only reason Dolphin requires a little-endian CPU is because so much of the code assumes it's running on one (woo, being written for x86 CPUs!), thus it'd take a fuckton of work to either rewrite the code to make it endian-independent or give it a whole separate codepath for big-endian.

neobrain Wrote:But then again you could say we'll run out of things to fix and could focus on improving performance...

Compare the following:
1. The number of things that still need to be fixed.
2. The slow emulation required to fix some of those issues.
3. The level of optimization already achieved.
4. Potential for significant future optimization. Particularly on the cpu side.

I'm hesitant to believe dolphin will become faster in the future.

1. Yes, the number of things that need to be fixed is comparatively low. Turn it around as you like, but it's low by any realistic definition, and I don't care for the definition you'll make up in your next comment to prove a point.
2. At least on the GPU rendering side of things we're getting to a point were the remaining unsupported features would be so ridiculously demanding that there'd be no point in using a hw renderer compared to a software renderer. Thus, that part of the code is pretty much as slow as it'll ever get.
3. You make it sound as if anyone capable* people had tried to optimize Dolphin in the last few years. If anything, Dolphin has the low hanging fruits of optimization done, but that's about it.
4. Yes. There's loads of potential for significant future optimization. Even on the CPU side. I can make the claim that it's possible, just like you can make the claim that it's not possible. Still, I'm likely right :p
5. (RE: "I'm hesitant...") Yes, me too, because people rarely bother doing the boring work of fixing the pile of crap that Dolphin's code base has become.

* knowing SSE doesn't count as "capable".

Has NV forgotten the possibility of better busy loop detection and macroblocks. If my understanding of JIT is right, macroblocks will cause anywhere between a 1% and 75% improvement.