Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Avie Tevanian Leaving Apple?
- Thread starter MacRumors
- Start date
- Sort by reaction score
rayz said:
Macworld UK said:
Macs are still a huge part and will increase, from what most analysts are saying.
edit/edit2: fixed link
boncellis said:
Apple is a HUGE company in regards to a single individual. Avie likely hasn't see or touched code in years and years. A person in his position spend most of his time in high-level meetings discussing trends, future goals and concepts 2+ years down the road and reviewing high-level product plans.
He is a guiding force in Apple but not involved in day to day development.
iMeowbot said:
It really is an issue of what the pathways Apple decided to optimize in Mac OS X. ... of course some can be improved.
You can come up with all kinds of measures of system "performance" that show good and bad aspect that often directly reflect which pathways Apple decided to optimize (or make fast at the costs of slowing down others) in order to best serve the target markets of Mac OS X (primarily consumers and media manipulation). Often those slow pathways are not an issue and are easily be dwarfed by other real world tasks (ones that cannot be avoided).
For example Mac OS X traditionally has excellent latency profiles (low latency and consistently low) allowing things like real-time audio processing in live performance situations.
Marlor said:
One aspect of the test is that Mac OS X does honor full file system sync requests which the database appears to be using in the test that Anandtech did while the Linux doesn't honor those. This easily degrades performance in the Mac OS X test because the database is asking the OS to write out all data to disk (including flushing disk cache) when doing atomic commits (database waits for this IO to complete before finishing the atomic operation).
Also for most Mac OS X users they don't attempt to fork/exec a huge amount of processes (which the Anandtech test mostly focused on) instead Mac OS X is optimized more for threads (use of processes with many threads instead of many processes). Linux was initially designed with processes as the scheduleable unit and so that aspect was and is heavily optimize.
Compile 'em all said:What the heck are you talking about?. Everybody knows that a monolithic
kernel beats the hell outta of microkernels preformance-wise. Besides, OS
X and NT both run a hybrid kernel (as another poster noted). I really can't
understand what do you mean by "modern" here. OS concepts evolve over
long periods of time and almost all the research in the area of Operating
systems has already been done.
Here take a look for yourself...modern Operating System my a**
All numbers are expressed in microseconds, lower is thus better.
source
You do realize that MySQL wasn't taking advantage of the native threading model for OS X, correct? From Kernel Programming.pdf
Since MySQL for Mac was developed around the pthreads on top of the native kernel threads it's not surprising the performance would be degraded, but then again this issue is addressed @Apple.com. Thread priority dynamically changes in OS X. To raise priority:
http://docs.info.apple.com/article.html?artnum=303225
dr_lha said:
Yes and you do realize CUPS on OS X is not setup for printing, by default. It works very easily, just as it does inside of Linux, but the interfaces are much cleaner and colorsync profiles for color printing are not available for Linux. The user can use either the printer driver provided by the manufacturer or a ppd inside of CUPS. The ipp:// features of CUPS aren't even touched upon by the printer driver manufacturers which goes into CUPS favor.
iBookie said:1. Apple switches to Intel
2. Windows Vista pushed back to 2007
3. Vista rumored to have 60% of code rewritten
4. MS fires a layer of Vista management
5. Avie "pursues other interests"
Is something going on here......
Ya, MS stands for Microscrap. I would just hate to see a technologicaly superior company go down for following the same footsteps as M$. Mickeysoft went after the buck, Apple is apparently going the same route, I know a company has to make a profit, but I hope they do not lose site of what we need. Technology, not M$ crap.
Marlor said:
Today Intel dual core processors and they promise quad core chips in early 2007. Today Sun has 8-core chips with four-way hyperthreading per core. Intel will have something lke this in maybe three years? Then what? The trend will continue. I jokingly was talking about "kilo-cores" a while back but I can envisin 1024 hardware supported threads in the next decade and 32 in the Intel CPU world in a few years.
OK so assume we have 8-core or 16-core computers, now the micro kernels are looking like they are a lot better.
I have been waiting a long time for "NON-symetric multiprocessing" That is a computer with two or more CPU chips wher the chips are not the same. Say one is a specialized vector processing or one was a G5, Itanium, hardware java bytecode machine and one is an X86. Mach would be ideal for this. Just think how smooth Aple's next transition will be when they can phase in a new CPU using non-symetric multiprocessing. (BTW I was once a system programmer on a non-symetric machine. Th old CDC 6600 series had both 60-bit CPUs and 12-bit "PPUs" CPU = Central Processor, PPU = Periferal Processor the later computers would have 2 CPUs and 20 PPUs. Most of the operating system ran in the little 12-bit machines leaving the CPUs free to crunch numbers.)
Back to PCs .. Once you have 32-cores then you ask "Do all 32 cores need a floating point unit? No. Cores that run kernel threads don't. Compilres don't need FP either. So why not put that silicone to better use with eeither even more cores, less power used or more cache? The OS scheduler can look at some task atributes to see which core is "best" I'm sure it will happen
Apple's Max OS is looking to the future ten years out and is a better match to future hardware then to present hardware. Not a bad thing really.
For those who dont know who he is, he's the guy steve thanks at the end of keynotes for "all of his hard work".
I don't know but it seems to me that everybody is always talking about software as the real advantage in the MS vs Apple war (yes, aesthetics too, but would YOU pay a 2xprice premium for a Max with XP?)
If the guy behind the software is leaving, looks like a problem to me...
If the guy behind the software is leaving, looks like a problem to me...
i find it fascinating how many threads like this one end up being about how many angels can fit on the head of a pin.
To bring it back to the topic
1. Never heard of him (does that make me a bad person?)
2. The cult of individuals is always a bad thing - you have to give consideration to what happens next
To bring it back to the topic
1. Never heard of him (does that make me a bad person?)
2. The cult of individuals is always a bad thing - you have to give consideration to what happens next
Name change
I think we need to change the name of this website from "MacRumors" to "MacHumors"
mygoldens said:Ya, MS stands for Microscrap. I would just hate to see a technologicaly superior company go down for following the same footsteps as M$. Mickeysoft went after the buck, Apple is apparently going the same route, I know a company has to make a profit, but I hope they do not lose site of what we need. Technology, not M$ crap.
I think we need to change the name of this website from "MacRumors" to "MacHumors"
"What the heck are you talking about?. Everybody knows that a monolithic kernel beats the hell outta of microkernels preformance-wise."
Those benchmarks measure things like "fork()". So how many forks per second are done on a desktop OS? I doubt the long term average is in to the double digits so if each one takes an extra 4000 microseconds it just doesn't matter.
What is starting to mater today and will matter more and more is how well the OS scales to a large number of CPUs.
Those benchmarks measure things like "fork()". So how many forks per second are done on a desktop OS? I doubt the long term average is in to the double digits so if each one takes an extra 4000 microseconds it just doesn't matter.
What is starting to mater today and will matter more and more is how well the OS scales to a large number of CPUs.
peharri said:
I don't know what language you refer to, as I am offending no one...just putting forward my opinion as someone who used Linux for a while, and did NOT like it.
But it seems like some are proposing that Linux IS a better OS, which is FAR from true.
The mere fact that you feel like explaining that XNU is not GNU/Linux which is not who-knows-what-else just proves my point: Linux is NOT for normal users. And no, I don't even remember which distro I worked with at the Uni...you see? Too complex for normal people...
No, they're not. They're explaining that, in their view, Linux is a better kernel than XNU. That is, a component of the GNU/Linux operating system (of the RedHat operating system, if you'd prefer) is better than a component of the Mac OS X operating system.BRLawyer said:
No, it doesn't. I don't think you understand what's under discussion here. It's a little like comparing an electric car to a gasoline car, and thinking the electric car is harder to drive because of the discussion about efficiency and generators and all that stuff.
When you talk about Linux not being any good, you're (BRLawyer) most likely talking about an entire operating system. This is largely because of confusion around the term "Linux". Linux is a program written originally by Linus Torvalds which is what's termed a kernel. It essentially mediates between applications and hardware, dividing up memory between them, ensuring they don't step on each other's toes, and providing basic features such as the ability for different programs to communicate. Its counterpart in Mac OS X is called XNU.
Some time early on in the history of Linux, people started shipping distributions. These distributions consisted of Linux and a set of tools called GNU. For reasons which are still controvertial, most distributions named the entire package "Linux".
When you were criticising "Linux", you actually were refering to the latter, which the rest of us call GNU/Linux (because it's GNU and Linux.) However, when the people you were addressing were talking about "Linux", they were talking about the former. So it was a bit like someone saying "XNU sucks/rocks!" and you saying "No it doesn't/yes it does, Mac OS X has a great UI!", except that in the Linux world, the equivalents of Mac OS X and XNU keep being refered to be the same word, so it gets confusing.
Is "Linux" too difficult for ordinary people? Well, the question's meaningless, and in any case the same answer, whatever it is, would apply to XNU. Neither Linux nor XNU come with anything that interacts with end users (well, I think both have a serial debugger, so you can hang a terminal off the serial port if you want, but I'm not sure that's adequate in this context.) The stuff that interacts with users runs over Linux or XNU, and very little of Linux or XNU ends up defining how that stuff works.
If Apple went crazy (they have no real reason to do this) and removed XNU from Mac OS X, and put in Linux, would the result be an unfriendly operating system? Well, truth be told, no.
Most applications would "just work" with nothing more than a recompile. Other than that, from an end user's perspective, the kernel change wouldn't make the slightest difference. It would actually be as "different" as OS X for Intel is from OS X for PPC. There'd be performance affects, obviously. But Linux wouldn't change the user interface - Linux (and XNU) is not where you implement user interfaces, those get built on top of the kernels.
The big problem is you've walked into a very old flame war. Linux is a so-called "monolithic" kernel. That means all the device drivers and other "programs" that provide low level services can all "see" and touch each other. This means they can communicate with one another very quickly, but it also means a bug in one can cause havok with the rest.
XNU is based on Mach, which was originally a "microkernel", but isn't much of one any more. In a microkernel, all of these subsystems are protected from one another, so, for example, the USB keyboard driver can't write over the memory of the disk driver. In XNU, these subsystems still have a degree of independence, but they all have some degree of freedom to interfere with the running of other subsystems. This was done to increase efficiency.
People have strong views on what constitutes the correct approach. Academia favours microkernels. Both Apple and Microsoft have come down on the side of "Hybrids", which is what I've described XNU as being - both NT and XNU have this kind of modularity. On the other hand, Linus Torvalds favours the monolithic approach, and, early on in Linux's history, was told by at least one academic (the author of Minix) that it was an awful, outdated, design.
So anyway, I hope that makes it a little more clear. When you talk about "Linux", you're not talking about the same thing that those who are posting benchmarks are talking about. You're comparing GNU/Linux to Mac OS X. They're comparing the Linux kernel to XNU.
I think it would be a rather pointless thing for Apple to remove XNU from Mac OS X and put in Linux, and I doubt they'd do it anyway, for licensing reasons. But if they did, your user experience wouldn't change. You'd still get Aqua and Quartz, not X11 and GTK. You'd still get the Finder and Dock, not GNOME or KDE. The OS would continue to "just work" on Apple hardware. It might be fractionally faster, or fractionally slower, that's open to debate. I doubt anyone will notice to be quite honest.
Well, thanks anyway for your explanation; but still what we see in Anandtech's already old test is a clear mention to OS X versus Linux, thus giving us lotsa food for war...I understand the core of the debate, but just can't believe that a single test on MySQL is able to bring such bad feelings to some users here on MR...
Lead Engineers unhappy w/PPC->x86 transition?
Wow, it sure seems clear that both Rubenstein & Tevanian were unhappy with the transition to x86/Intel -- of course I could be missing something... but NeXT did the same thing -- went from 68k to x86 (NeXT even did some PowerPC testing at one point, but I'm unsure why PPC wasn't used, some ex-NeXTers can chime in here) and in the process sold their hardware division to Canon who went on to make Canon object.station computers (I had one). NeXT then became a software only company... WebObjects became their biggest cash cow, and Apple has largely neglected WO...
I would really like to know what companies Rubenstein and Tevanian are going to and the reasons for ther moves. Funny how this all coincides with a M$ shakeup and Vista delays... Lots of interesting things happening in the telecom/computing industries lately!
Standing Ovations for
AARON RUSSO's new movie!
America
FROM FREEDOM TO FASCISM
VIEW THE TRAILER:
http://freedomtofascism.com
General Release this summer, pre-screenings nationwide now!
Wow, it sure seems clear that both Rubenstein & Tevanian were unhappy with the transition to x86/Intel -- of course I could be missing something... but NeXT did the same thing -- went from 68k to x86 (NeXT even did some PowerPC testing at one point, but I'm unsure why PPC wasn't used, some ex-NeXTers can chime in here) and in the process sold their hardware division to Canon who went on to make Canon object.station computers (I had one). NeXT then became a software only company... WebObjects became their biggest cash cow, and Apple has largely neglected WO...
I would really like to know what companies Rubenstein and Tevanian are going to and the reasons for ther moves. Funny how this all coincides with a M$ shakeup and Vista delays... Lots of interesting things happening in the telecom/computing industries lately!
bored said:
Standing Ovations for
AARON RUSSO's new movie!
America
FROM FREEDOM TO FASCISM
VIEW THE TRAILER:
http://freedomtofascism.com
General Release this summer, pre-screenings nationwide now!
