Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
SiliconAddict said:
Dude you have "deep deep" issues.
Yes, but not about the hating of windows ;) I'd like to remind everyone it is perfectly normal to sit alone in the dark stroking a knife, whilst saying "must destroy.. windows" :D
 
SSE3 is nothing like Altivec. That said, any kind of Altivec emulation in Rosetta will probably significantly speed up some apps. I only fear that it might have exactly the opposite effect (i.e. Apps that are Altivec-aware now see Rosetta as a G4 and start using Altivec instructions which when emulated, turn out to be slower than a G3 instruction equivalent emulated through Rosetta...) My recollection, however, is that PearPC got a lot faster when the Altivec branch stabilised, so it should be good. I look forward to experiencing this new build.
 
SiliconAddict said:
A-freaking-men. Classic needs to die a horrible screaming death. If software development houses haven't updated their wares by now they aren't going to be updating them and if that is the case people should be seriously looking to migrate their data to another app. Anyone who sticks to legacy apps based on a legacy OS is just asking for trouble.
I can attest to this since we have one user still on a windows 3.11 system that runs some proprietary apps that use a reel to reel tape reader. Also we have one legacy app that was built around Windows 95. The company went out of business in '97. And when we upgraded to Windows 2000 with NTFS file system all of a sudden the app started getting glitchy with security rights in NTFS. No fix or work around because the vendor was no longer there. I warned the powers that been about this but nooo. So all of a sudden a panic happens because they can't use this app anymore.
Honestly I don't feel sorry for people who are fretting over this transition. 5 years of support is a perfectly acceptable timeframe. At this point either upgrade, migrate, or buy a crap load of spare parts and stay in the past forever.

Contrary I still think it is nice for customers to have the option of having some form of classic support. Surely with Altivec support built into Rosetta, doesn't the x86 branch of OSX has for all intents and purposes a G4 emulator? Why can't it just run the old classic emulator on rosetta?

Sure it might be slow and clunky, but at least make it an option available for those who might need it.ß
 
NickCharles said:
x86 stinks, it always has, and always will ... x86 needs to die, right along with Windows.

Makes you wonder whether Apple have asked Intel to produce an x86 chip range without legacy x86 support. They could argue that such a chip would have wide appeal to the Linux community and niche OS players as well as Apple (it would likely run faster, more efficiently and cooler). It probably woundn't be a huge effort for Intel (although designing any chip requires significant Engineering resources).

Another possibility might be to put an Intel die and an Altivec die in the same CPU package. If intel sell their processors as dies (do they?) and the Altivec is available as a die (is it?), packaging them together wouldn't be prohibitively expensive. This would mean that Apple would have a CPU that is different to most PCs and offers better performance for the same clock speed (AltiVec offers better performance than SSE3 under most circumstances).

A third possibility is that Apple will move as much of the AltiVec-dependant routines to the GPU as possible and that all future Apple products will feature a programmable GPU. SSE3 will be used only where it makes sense (where it can complete the operation faster than the GPU).

I think the first scenario is fairly unlikely - both Apple and Linux perceive value in allowing Windows applications to run in a virtual machine. Six months ago I would have said that the second scenario was likely, but if the reports are true, it seems that Apple are working hard to make OSX run on SSE3. The third scenario currently seems the most likely and would be an effective way to differentiate Apple products from the competition (particularly laptops and the AOpen miniPC).
 
mdavey said:
Another possibility might be to put an Intel die and an Altivec die in the same CPU package. If intel sell their processors as dies (do they?) and the Altivec is available as a die (is it?), packaging them together wouldn't be prohibitively expensive. This would mean that Apple would have a CPU that is different to most PCs and offers better performance for the same clock speed (AltiVec offers better performance than SSE3 under most circumstances).

Quite unlikely, Apple is too small... and that custom chip would be too expensive. Besides don't forget IBM and Freescale own the rights to the PPC design too. I doubt Intel will be able to incorporate Altivec without paying them.


mdavey said:
A third possibility is that Apple will move as much of the AltiVec-dependant routines to the GPU as possible and that all future Apple products will feature a programmable GPU. SSE3 will be used only where it makes sense (where it can complete the operation faster than the GPU).

Trouble is not every single altivec instruction call is necessarily for display purposes. What if I were running Seti@home, or any other similar number cruncher?

I would be AMAZED if the GPU would be used to augment the CPU for processing.
 
Azurael said:
SSE3 is nothing like Altivec. That said, any kind of Altivec emulation in Rosetta will probably significantly speed up some apps. I only fear that it might have exactly the opposite effect (i.e. Apps that are Altivec-aware now see Rosetta as a G4 and start using Altivec instructions which when emulated, turn out to be slower than a G3 instruction equivalent emulated through Rosetta...) My recollection, however, is that PearPC got a lot faster when the Altivec branch stabilised, so it should be good. I look forward to experiencing this new build.

I doubt so, Altivec is really just a pretty name over a bland technology. At the end if the day if you have 2 cores on a Powerbook and you throw one whole core for Altivec emulation, it will totally smoke our current G4 1.67Ghz chips many times over.

I doubt SSE3 would be that bad. G5s may be decent but they are not "magic".
 
mdavey said:
Makes you wonder whether Apple have asked Intel to produce an x86 chip range without legacy x86 support.

:confused: x86 without x86 architecture? That wouldn't be x86.
 
mdavey said:
Makes you wonder whether Apple have asked Intel to produce an x86 chip range without legacy x86 support. They could argue that such a chip would have wide appeal to the Linux community and niche OS players as well as Apple (it would likely run faster, more efficiently and cooler). It probably woundn't be a huge effort for Intel (although designing any chip requires significant Engineering resources).

Another possibility might be to put an Intel die and an Altivec die in the same CPU package. If intel sell their processors as dies (do they?) and the Altivec is available as a die (is it?), packaging them together wouldn't be prohibitively expensive. This would mean that Apple would have a CPU that is different to most PCs and offers better performance for the same clock speed (AltiVec offers better performance than SSE3 under most circumstances).

A third possibility is that Apple will move as much of the AltiVec-dependant routines to the GPU as possible and that all future Apple products will feature a programmable GPU. SSE3 will be used only where it makes sense (where it can complete the operation faster than the GPU).

I think the first scenario is fairly unlikely - both Apple and Linux perceive value in allowing Windows applications to run in a virtual machine. Six months ago I would have said that the second scenario was likely, but if the reports are true, it seems that Apple are working hard to make OSX run on SSE3. The third scenario currently seems the most likely and would be an effective way to differentiate Apple products from the competition (particularly laptops and the AOpen miniPC).

Ok i don't think you can just remove the X86 legacy parts and still have an X86 Processor. You can reduce it so you are talking directly to the internal RISC core (all X86 processors at heart are RISC) but the chips are made specifically to work with the X86 translation front end and would require a lot of work without it.

I think putting Altivec on die would be great idea but there are probably licensing agreements prohibiting this. Maybe Intel can release SSE3A (A for apple :) ) which would basically be the intel equivalent of Altivec. Any PPC program running on rosetta will translate the old altivec instructions to work on the new altivec like core (they'll probably do this anyway with altivec and SSE3) and all new Mactel binaries will work fine with the new altivec like core. This could be disabled in standard pentiums, or even left on and used by windows and linux to great effect.

As for everyone who says PPC is dead, the PPC is a great platform. The only problem is that no one has the resources to fund the R&D it so badly needs. If PPC had the R&D backing of Intel then X86 would not stand a chance against it. Most of work an X86 processor does is convert the old CISC X86 instructions into snappy RISC instructions.
 
SiliconAddict said:
Dude you have "deep deep" issues.

The best thing that could happen? All windows source code is burned and all traces of that horrible, crappy, bug ridden, piece of $#!+ operating system are erased from the face of the earth. It's the only thing Windows deserves. The only thing.
 
NickCharles said:
The best thing that could happen? All windows source code is burned and all traces of that horrible, crappy, bug ridden, piece of $#!+ operating system are erased from the face of the earth. It's the only thing Windows deserves. The only thing.

When you go away and program an operating system that runs on every new computer plus has enough legacy components to run on computers that are up to 5 years old, including even more legacy so it is compatible with programs written up to ten years ago and on top of that works with nearly every piece of hardware under the sun then you can come back to us and tell us how crap windows is. Until then, you really have no idea what you are talking about. Windows might not be as good as OSX or Linux for you but it is a good OS in it's own right. Step out of the box and you might appreciate how much work the windows programmers have done to keep the whole thing working with all the varying hardware it has to run on.
 
TBi said:
When you go away and program an operating system that runs on every new computer plus has enough legacy components to run on computers that are up to 5 years old, including even more legacy so it is compatible with programs written up to ten years ago and on top of that works with nearly every piece of hardware under the sun then you can come back to us and tell us how crap windows is. Until then, you really have no idea what you are talking about. Windows might not be as good as OSX or Linux for you but it is a good OS in it's own right. Step out of the box and you might appreciate how much work the windows programmers have done to keep the whole thing working with all the varying hardware it has to run on.
Thank God someone here's isn't a complete zealot. It is one thing to dislike something but it is totally another to completely discount it because of personal opinion. It's already been said MANY times OS X would be almost exactly like Windows if it had to support everything the way Windows does. I'm not saying Windows is great but it certainly isn't as horrible as the way everyone says it is.
 
DeathChill said:
Thank God someone here's isn't a complete zealot. It is one thing to dislike something but it is totally another to completely discount it because of personal opinion. It's already been said MANY times OS X would be almost exactly like Windows if it had to support everything the way Windows does. I'm not saying Windows is great but it certainly isn't as horrible as the way everyone says it is.

You can defend it till you're blue in the face, it doesn't deserve it. You won't change my mind. Call me any name in the book. Windows will always be a piece of crap. Sadly, Corporate America and most consumers are content to let Microsoft and Windows drag them around by the balls.
 
Fiveos22 said:
Slightly off topic but...

After reading the Wikipedia article on Altivec and then reading their article on Power PC, it dawned on me that there may no longer be a Power Mac and a Power Book once Apple switches to Intel...

http://en.wikipedia.org/wiki/PowerPC


What is the replacement going to be, the Pentium Mac?

Powerbook predates the PowerPC processor by years. PowerMac may have have had an initial relationship to the PowerPC, but now it just signifies Apple's professional line just as "i" signifies the consumer line.
 
Photorun said:
Yeah, a 38 year old "young'un" who's been on computers since 78, had an Amiga, Commodore, started on Macs with the Apple][, worked for a major software company, have had over 30 computers, four of which can still boot up in conceivably everything but System 6 (though I could try installing it on my Classic SE), clearly by your snark know probably a crapload more than you

Oh my, sensitive aren't you. Okay, here we go, the battle of who's done what. I've been in computers since 1976, have designed and build computers using a wide variety of chip sets and processors. Beat you. :)

certainly I look better than your goofy avatar so what the H is your point?

Oh my, so if you can't win on a real level you reduce it to petty personal attacks. Hmm... from you're avatar you look like a small buck toothed mammal - a pest that my dogs enjoy killing and eating. :)

It is interesting that this topic of maintaining support for Classic applications raises such passion in those who oppose it. How can it possibly threaten you that someone else needs continued support for legacy applications. It doesn't take anything away from you. It requires minimal support on Apple's part. They have already done the hard work. Get over it.
 
generik said:
I doubt so, Altivec is really just a pretty name over a bland technology. At the end if the day if you have 2 cores on a Powerbook and you throw one whole core for Altivec emulation, it will totally smoke our current G4 1.67Ghz chips many times over.

I doubt SSE3 would be that bad. G5s may be decent but they are not "magic".

You are obviously not a developer or engineer of any kind. Altivec is pretty much a one-of-a-kind technology, in that was the first REAL vector processing unit to reach down to desktops. Previously, this technology was reserved for the likes of Cray computers, etc.

SSE has only 8 128-bit registers, whereas Altivec has 32 128-bit registers. Also, Altivec is dual-pipelined which means it can execute two of it's instructions at once.

Many of Altivec's benefits come in the ability to deal with vectors in one clock cycle, whereas a regular general-purpose CPU would take multiple clock cycles to complete the same task. Take for instance the permutation of an array (moving a bunch of data around in an array). Altivec can do this in one step, whereas a general-purpose CPU would have to copy each element, taking multiple clock cycles. For reference, SSE doesn't even have an instruction to handle doing this. (I give this example because I am currently working on a project to accelerate encryption algorithms using SIMD hardware like Altivec. Permutations are very common in encryption).
 
generik said:
Contrary I still think it is nice for customers to have the option of having some form of classic support. Surely with Altivec support built into Rosetta, doesn't the x86 branch of OSX has for all intents and purposes a G4 emulator? Why can't it just run the old classic emulator on rosetta?

Sure it might be slow and clunky, but at least make it an option available for those who might need it.ß
I guess the point is that if most users don't need these legacy resources, should we all have to pay for them (via higher prices for the newer computers).

There is always the possibility of a third party program (or even Apple's SW division, as a separate revenue source) to do the emulation - thus letting the people who need Classic pay for it but sparing the rest of us.

In all my years of Mac ownership since the original 128K back in 1984, one constant thread is that Apple decides to drop support for older programs (after a few years of transitional support) when it moves platforms. So it's hard to see why Apple would do something different in this case.
 
generik said:
Why can't classic run on rosetta?
I'll bet that it's the difference between "runs at all" and "runs acceptably". Much of the magic in Rosetta is that calls to the OS can be run as native code. Where Classic brings its own OS to the party, that bit of magic is going to be seriously impaired.
 
generik said:
Contrary I still think it is nice for customers to have the option of having some form of classic support. Surely with Altivec support built into Rosetta, doesn't the x86 branch of OSX has for all intents and purposes a G4 emulator? Why can't it just run the old classic emulator on rosetta?

Sure it might be slow and clunky, but at least make it an option available for those who might need it.ß
Sorry, but Apple clearly states that Classic isn't supported in Rosetta, and I for one am not too worried about it. Its time to move on, and I wouldn't want Apple expending significant resources making Classic work, when they can be working on other projects.

Classic itself seems to be a dead project anyway - as far as I can see it hasn't been changed or updated in any significant way since the release of Tiger.
 
Has anyone done any testing yet? :confused:

Some performance numbers for Altvec capable programs under this new version of OSX for x86? Just to see what kind of performance it offers?...

Sorry, but Apple clearly states that Classic isn't supported in Rosetta, and I for one am not too worried about it. Its time to move on, and I wouldn't want Apple expending significant resources making Classic work, when they can be working on other projects.

Classic itself seems to be a dead project anyway - as far as I can see it hasn't been changed or updated in any significant way since the release of Tiger.

Maybe im just talking out of my ass here, but its my understanding that the majority of Mac sales are to 'professional' customers. People who use their systems for serious work, or in business (or part of the business itself) and with that id imagine Classic support is important. I don't think Apple could dare neglect it, or brush it off .
 
RobHague said:
Maybe im just talking out of my ass here, but its my understanding that the majority of Mac sales are to 'professional' customers. People who use their systems for serious work, or in business (or part of the business itself) and with that id imagine Classic support is important. I don't think Apple could dare neglect it, or brush it off .
Out of interest, what Classic software would these pro users be running?
 
Not emulation, but yes altivec support (sorta)

SSE3 doesn't have the registers to "emulate" altivec with any kind of reasonable performance. What this likely is is Rosetta translating Altivec Accelerate.framework function calls to the SSE3 native Accelerate.framework calls on the fly and then doing some kind of swizzle on the data being passed. So yes, the vector optimization gets a benefit, but no it really isn't "emulation".

You'll notice that in prior builds things like iTunes use the Accelerate.framework still run because the calls get translated by rosetta to the non-vectorized function implemention. That's why iTunes under Rosetta (OSx86 10.4.1) runs like an emulated G3. Likely (and I don't mean the new universal binary) iTunes under the new Rosetta with "altivec support"will be significantly faster encoding simply because of the vectorized Accelerate.framework advantage.
 
Kelvin said:
SSE3 doesn't have the registers to "emulate" altivec with any kind of reasonable performance. What this likely is is Rosetta translating Altivec Accelerate.framework function calls to the SSE3 native Accelerate.framework calls on the fly and then doing some kind of swizzle on the data being passed. So yes, the vector optimization gets a benefit, but no it really isn't "emulation".

You'll notice that in prior builds things like iTunes use the Accelerate.framework still run because the calls get translated by rosetta to the non-vectorized function implemention. That's why iTunes under Rosetta (OSx86 10.4.1) runs like an emulated G3. Likely (and I don't mean the new universal binary) iTunes under the new Rosetta with "altivec support"will be significantly faster encoding simply because of the vectorized Accelerate.framework advantage.

Thanks for setting that straight. Obviously you have worked with Altivec and are now working with SSE, and understand the serious shortcomings of SSE when compared directly with Altivec (see my post above).
 
artifex said:
This also helps in explaining why some emulators perform much better than the originals, too, right? :)

If you can find one place that a software emulation offer more power then the hardware it is emulating I will give you a lollypop. And that is not a gameboy emulation on a 32 nod grind with 240 dualcore Operton.

Altivec is a hardware feature in the PPC, a X86 CPU does not have the same feature and never will. The SSE3 is not by far as good as Altivec but other then that most X86 spanks PPC to the ground.

I look forward to the Yohan CPU :) that will look very good in a mac.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.