Yup, it was just posted to MaUpdate as well:

http://www.macupdate.com/download.php/14973/D … versal-0.65.dmg

This isn't official, though, is it? The application icon doesn't match the PC one and it apparently can read its preferences from the Library instead of a standard .config file like the PC (and Lamer0's build). I just tested that it still reads a -writeconf produced dosbox.conf normally, though.

I've noticed that both Lamer0's and this "punk" version uses different keyboard layouts, but none of the fully corresponds to my Macbook Pro's actual keyboard layout (Swedish).

Also, a weird thing is that both versions had a delay of about 2-3 seconds when starting them, but that has disappeared now.

Finally, and this is to me really weird now: At 12000 cycles, normal core, Dosbox takes 70% steady on OS X, and 50-70% on my 3200+ Windows XP machine running the same game. Yet, the OS X version sound stutters like crazy and runs very slowly!

Perhaps the a single core of the 1.83ghz core duo isn't faster than an AMD 3200+ XP cpu, I will have to test this when I install boot camp.

But as I was testing this, suddenly the game I was testing began crashing like mad using the dynamic core on that "punk" version, while it runs fine on lamer0's version!

Must test further...

Ok, a bit further testing...

No matter what I do, regardless of core type and number of CPU cycles, I cannot get the OS X version of 0.65 to stop sound stuttering (the best I could do was to get 0.5 seconds of silence every 2 seconds).

Also, the 0.65 OS X version crashes using dynamic core on the game I was testing, but version 0.65 for XP (and lamer0's version) does NOT crash.

I've noticed that both Lamer0's and this "punk" version uses different keyboard layouts, but none of the fully corresponds to my Macbook Pro's actual keyboard layout (Swedish).

the default keyboard layout is the US one (just like real dos), you probably need to load keyb (from freedos) to correctly use your swedish layout.

well there were some changes to the default mac os x layout after 0.65
not sure if which source each build uses.

Well, the keyboard is really a minor thing. The performance is more important, imo.

Has anyone else with a Macbook out there tested this yet?

I don't know how these Core Duos work yet, but it seems that though the activity monitor reports DosBox as taking 70% CPU resources, only about 50% of each core seems to be busy. Is it really dividing the workload between the cores? Still, the performance is awful, imo (and I don't really want to install boot camp to compare with XP until the next beta, in case my OS X partition gets trashed, as reported by a few).

For most apps not directly optimized for multiple processors (and this goes for dual core systems as well), you're likely to see the performance load fairly evenly divided among the processors. In general two processors/cores/etc is only a drastic improvement when the app is built multithreaded and with it in mind from the ground up. For poor performance though, what sort of settings are you using? Did you make sure you enabled the dynamic core? What output? etc. I can play lower end 486 games with no problems on my Mac mini G4. You should be able to go significantly higher than that on any of the Intel Mac systems.

I have tested the 0.65 "Punk"-version on my MacBook Pro, and it crashes all the time. Most games crash after 1-2 seconds before even showing any graphics!

This was also the case with the "Punk" 0.63 Universal version...

Starange thing is that Lamers Universal version was rock solid? Do they implement the dynamic core patches the same way? Must be something wrong with the "Punk"-version...

Thank you for confirming that the "punk" version is busted. I guess all we can do now is to wait until what was deemed wrong with lamer0's patches has been sorted out, because whatever he did, it apparently works, and whatever the "punk" people did, they failed.

Is there a Macbook Pro user out there with boot camp who can do speed comparison tests between XP and OS X? I still want to wait until the boot camp beta is more mature before I install it... (or perhaps I'll crumble and go ahead in a week or so, I don't know).

Can you use Ghost/Acronis or something to backup your drive?

Is there a Macbook Pro user out there with boot camp who can do speed comparison tests between XP and OS X?

The latest stable Windows DOSBox on my 1.83 Macbook w/ 1.5 gb ram is WAY faster than either the lamers or punk build. I played a few games: Red Baron, Shuttle, Might & Magic Xeen, and they all are really slow and the sound stutters on OS X, but run really good in XP. So yeah, something isn't working 100% in the unofficial UB versions.

steerpike wrote:

Is there a Macbook Pro user out there with boot camp who can do speed comparison tests between XP and OS X?

The latest stable Windows DOSBox on my 1.83 Macbook w/ 1.5 gb ram is WAY faster than either the lamers or punk build. I played a few games: Red Baron, Shuttle, Might & Magic Xeen, and they all are really slow and the sound stutters on OS X, but run really good in XP. So yeah, something isn't working 100% in the unofficial UB versions.

Thank you for confirming this. I was wondering why my old 3200 Athlon CPU seemed way faster than this new 1.83 Core Duo.

Then again, Tycho on PA saw a 25-50% increase in FPS in World of Warcraft when using boot camp to run it from Windows.

I hope OS X can go much faster, but I guess it's just that simply nobody noticed before because there were no good ways to compare.

I seem to remember something about Apple optimizing OSX for a small memory foot print over performance optimizations. Then again, maybe the x86 version of Darwin isn't as well optimized for the hardware as it is for PPC.

eobet wrote:

Then again, Tycho on PA saw a 25-50% increase in FPS in World of Warcraft when using boot camp to run it from Windows.

I hope OS X can go much faster, but I guess it's just that simply nobody noticed before because there were no good ways to compare.

The FPS increase is explicitly tied to Apple's OpenGL implementation. This shouldn't really be having much if any an impact on DOSBox. (And on the bright side, Apple and Blizzard are both working on this so performance improvements on both PPC and x86 side are expected to arrive~) I'm thinking something's either off with the UB builds you're using or maybe your settings are odd. I don't get any audio skipping after all on my Mac mini G4! My only limitation is that I'm stuck with for the most part older games since I'm having to do full blown emulation.

collector wrote:

I seem to remember something about Apple optimizing OSX for a small memory foot print over performance optimizations. Then again, maybe the x86 version of Darwin isn't as well optimized for the hardware as it is for PPC.

Hmm? Everywhere I've read about OS X performance since I got my Macbook, everyone has always seemed to say that OS X "loves memory", ie. it's a memory hungry SOB and eats up any and all availible memory at all times, just so that it can perform faster.

neowolf wrote:

I'm thinking something's either off with the UB builds you're using or maybe your settings are odd. I don't get any audio skipping after all on my Mac mini G4! My only limitation is that I'm stuck with for the most part older games since I'm having to do full blown emulation.

I've even once copied the settings from my 0.65 XP dosbox installation, so I can say that I'm 99% certain that it's not my settings.

Good to hear about Blizzard & Apple, though. I actually hope they find some fundamental flaw with OS X, so that the entire system will benefit from this and perform faster, not just World of Warcraft...

eobet wrote:

collector wrote:

I seem to remember something about Apple optimizing OSX for a small memory foot print over performance optimizations. Then again, maybe the x86 version of Darwin isn't as well optimized for the hardware as it is for PPC.

Hmm? Everywhere I've read about OS X performance since I got my Macbook, everyone has always seemed to say that OS X "loves memory", ie. it's a memory hungry SOB and eats up any and all availible memory at all times, just so that it can perform faster.

I'm thinking this comes from the common statement that Darwin's compiled with the -OS flag. It just means that it's compiled to try to be quick but also to not sacrifice too much memory for that speed. In general OS X over all quite the memory hog and the kernel's compile options don't have nearly that much of an impact on over all performance. Especially in specifics like this.

eobet wrote:

I've even once copied the settings from my 0.65 XP dosbox installation, so I can say that I'm 99% certain that it's not my settings.

Good to hear about Blizzard & Apple, though. I actually hope they find some fundamental flaw with OS X, so that the entire system will benefit from this and perform faster, not just World of Warcraft...

Well settings will behave differently in different OSes. For example overlay runs terrible on my mini both in a Window and in full screen. Likewise standard direct runs poorly in a window and great full screen. Though with the most recent builds I get ideal performance with opengl both in a window and full. From what I've gathered Windows tends to provide rather different results than those in general.

As for the OpenGL optimizations, they're far from just for World of Warcraft. Though honestly, the fundamental problem here IS with Apple's OpenGL implementation. It's heavily optimized for the typical OS functions it's used for in OS X for window compositing and what not... but... for games it's lagging behind. This is obvious if you look at the performance most OpenGL games exhibit compared to the same games on Windows machines. However Apple does seek to improve this for games and professional apps alike.

To be honest such improvements have been promised for some time and I'd been starting to get pretty doubtful, however there is some light shining. There's a recent build of Cinebench taking advantage of some of the OpenGL changes Apple's been slipping in behind the scenes and suffice it to say it's pretty hopeful. (About a 50% improvement in general tests.) Of course the games do need to be updated to take this sort of thing into account and even then there are still plenty of other issues with the OpenGL implementation, but it's progress~

That being said, don't set yourself up with hopes for a miracle cure. OS X is pretty bulky, and there's not really anything "fundamentally wrong" with OS X that could be fixed easily. There could be years of arguing over how certain kernel decisions impact performance for certain activities. Things are mostly how they are now and that's just that. A drastic improvement would require a drastic change. And history tells us that in general, OSes don't get faster, they get slower. History with OS X does run the opposite, but each major release's improvements in performance have been less and less compared to the last as well.

I got this comment from Lamer0 (victor) about his universal binary dosbox 0.63 (a bit too technical for me though):

Hey now, If you check the dosbox patch submit section on sourceforge you will see my patch there. Most likely 'Punk' is doing a straight compile without any modification. Dosbox compiles fine out of the box, however there are alignment issues that come into play. The patch I currently have out is a hack. The real issue with the dynarec crashing is the fact that external function calls are not 16-byte aligned. If you get the calls 16-byte aligned it should work great without pushing dx/bp etc to the stack and back. I hope this made any sense. Oh and on a side note, I don't use osx86 anymore. It was fun but it was just too much work with very little productivity benefit.I should really put up my work-in-progress emulators up. I got mupen64 with a proper dynarec working.The only issue being I can't compile it anymore since I don't have osx86. The fix is in my svn repository. So if you do want to compile it and redistribute , feel free to do so. -Victor

Crap, he wasn't even a... ehm... "proper" Apple user? Typical, and too bad. He did some really nice ports really fast (how did he learn cocoa that quickly???).

If the dyncore problems are really just because of required stack
alignment in some (which??) functions, it shouldn't be too hard
to find a good way around this for somebody with a little experience
compiling stuff on osX, i guess. Maybe some compiler flags (if the
stack alignment problem doesn't occur in some internal libraries),
or a stack adjustment when entering dynamic core blocks).

I mailed macports@punknews.org (who compiled the crashing CVS universal binary of dosbox 0.65) about lamers very stable 0.63 version. But no answer so far...

I personally think that the all-in-all speed issue is more important than the dynamic core, though. As you can see in my posts above, I suspect that there somehow is a huge speed difference, regardless of core, between the OS X and XP versions (with XP being a whole lot faster).

I haven't installed boot camp yet, but surely someone on these forums must have done it so they perhaps could do some accurate speed tests (ie. check at what amount of cycles they get 30fps in a game, and at what amount of cycles the sound starts stuttering).

On my 1.83 MBP I still can't get perfect, stutter-free sound no matter what I do. 🙁