I understand now - strangely this is one/. thread where I have learned something by reading all the posts! (though I have had to be sceptical about some of them, but there's been a broad consensus about how it works).
It seems that it is an easier way to implement threading in the apps too, so not only does the OS schedule these as needed, the developer gets to implement his threaded app very easily too. Someone said its "islands of serialisation in a sea of parallelism" (or something like that), and I get that too - its like the old message passing middleware system I used to work on.
All in all, I think its great, if it can be implemented in the linux kernel too (but with an obj-C interface) then we can all benefit. And Windows will be playing catch-up while Linux devs get all the parallel processing goodness and Windows devs have to kludge up their own mechanisms app by app. (yes, I'm a Windows dev too, but I like the idea of linux getting the better toys nowadays).
The question is... how is this different to Intel's Thread Building Bocks, or OpenMP, both of which are better supported and more widely available to non-Mac developers.
I guess having their own implemntation (for the mac) makes sense, as they can integrate it throughout the OS. I don't know anything about GCD either, but could it be used in Linux to make that a more parallel-friendly OS with less developer effort, and more standardisation of parallel execution?
Linux is always playing catch-up with Windows features, its could be nice it got something truly fancy that used all those 4096 CPUs it can support:)
There are people who think that the only software that should be used is Microsoft's. I suppose that's a similarly extreme view, however, in the case of the GPL licence(s) if he takes the view that all software should follow the open source methodology, then only having the GPL as a licence is not so extreme. Perhaps he means all OSS should be licenced under the GPL only, I think that's the view of the FSF.
Incriminating? That's an opinion piece, which barely describes what he's opinioning about. I think it had something to do with a discussion regarding the old hippie 'property is theft' mantra, but I'm not sure.
Mainstream companies are beginning to get with this anyway, very many are into open source as a free product that is monetised in non-standard ways. Its no longer about making money by selling software, but selling services. That's hardly an extreme viewpoint anymore, especially when you consider you never buy software in the first pace, but a licence to use it.
Or... it was more a marketing campaign right from the start. How else would you describe the large number of self-congratulatory 'Microsoft loves linux, really' blogs and announcements, while the actual code (the real work part of this) is so poor?
If Microsoft was serious about this code, about assisting Linux with the Hyper-V integration, then they would be at least helping to support the code themselves.
is that better than people looking at the packaging, notice it says no linux support, and sticking with Windows?
User mode drivers are obviously the way forward then, with a 'Linux certification' before you can install them in the kernel. Or a crash dump when things go wrong that firmly puts the blame on the faulting driver (in big letters).
Its the annual meeting apparently. A chance for Ballmer to high-five the devs who did such a great job bringing the experience of Vista service pack 3 to the masses, and helping to firm up their revenues.
Some of the comments from 'softies in that blog are interesting. They certainly don't like Zune, think Search has spent a bit too much money, and there are a few too many project managers at MS.
I think the point of a driver ABI is that people could write drivers entirely separately to the kernel, and they wouldn't need recompiling for each minor release. Just that would encourage moving the drivers out of the kernel to a separate repository. That would make it easier to manage as the number of them increases.
Windows does have a lot of crappy drivers, but they also have loads and loads of good ones. Enough good ones that a lot of people don't complain about them.
Besides, is it more important to have a buggy driver, or no driver at all? Remember that a buggy driver that has its source released can have its bugs fixed by others. You don't get that in the Windows world. I guess a lot of manufacturers will release binary-only drivers, but even if they are buggy, that doesn't leave in any worse state than we are already - those manufacturers aren't releasing drivers for Linux in the first place.
I think a stable ABI would be a good thing, and would only inconvenience a kernel dev a small amount. In the past it made sense not to have the ABI as things changed too much, but Linux is a much more stable beast nowadays (or should be!) so a stable ABI, even if its only for major releases, would be a benefit to Linux. If the ABI needs changing, then the kernel can be released with a major point increase.
I see. I never saw that post, so I take umbrage that I was criticised for something I never even saw, let alone responded to. I don't think the parent of the post I replied to has anything much to do with my reply.
Not that I care too much, perhaps its a lesson that simply being critical of others is a bad idea.
There are plenty of music players out there that don't have the bells and whistles that are so annoying (get it, ha, get it? ok..) I recommend a Sony, I have a NW-S705F which is simple and very effective, but there are plenty of video-style ones as well as conventional music-only players. come with good noise-cancelling headphones too.
actually, just checking the parent post I replied to - it said:
"Pan's Labyrinth may not be for everyone, but it's is certainly among the top tier in movies. Maybe you should stick to Disney if you want something less depressing."
tell them that it sends their browsing habits to Microsoft (to make sure they don't download any pirate copies of Windows) and they'll instantly get *very* concerned.
While things may perform a vast amount better at idle you get worse performance under load which is when you want it.
not necessarily, today power saving and efficiency at idle is a more desirable thing than performance under load.
That's the point, there's almost certainly no 'right' answer, it depends too much on the task you want your computer to do. Having a single scheduler to fit all cases is never going to be right, you will always want something else. There's a post here from someone who runs high-performance computing using the noop scheduler because it gives than "a 30% boost", I say fair enough to that - a extreme case where you want a different scheduler. Back in the ordinary world, we want different schedulers for different uses, whether that be desktop, server, HPC, netbook or a mobile phone. 1 doesn't cut it well enough across *all* those platforms.
The kernel devs could be gods, I doubt they can still come up with something that runs on a both a supercomputer and a mobile phone without making compromises that affect both of them. Better to tune each of them according to their needs.
I think what you want is not a single scheduler designed for the desktop, but one designed for server processes. That's probably the whole argument here - there isn't a single scheduler that can work efficiently for the 2 wildly different types of work a user put a machine to, but currently you don't have a choice. This is all about giving users choice of what kind of scheduler they'd like to run. You might even find that a scheduler designed for lots of CPUs (at the expense of interactivity probably) would suit you much more than the current system, especially when you buy more cores.
I was hoping Windows 2000 would achieve that, then I was hoping Windows XP would achieve that, then I was hoping some of the newer 2.6 kernels in Linux coupled with innovations in X would achieve that - but I was always deeply, utterly disappointed. Then I kinda hoped Vista would get somewhat close to what BeOS did. Oh yeah, now that was a hope decisively smashed.
Yup, but apparently Microsoft 'listened' to the screams of outrage, hate and despair that the users of Vista cried out, and changed things in W7 to make user repsonsiveness much more important. I'd like to say they went and listened to feedback from their users, but it was hard not to hear it.
From some reviews I've read, the decision was more a marketing one - W7 is just as slow as Vista in many parts, but because the UI responds to the user quicker (even though the time the operation takes is just the same) everyone thinks that W7 is a much more usable, ans faster OS. It isn't, its just more responsive. But at the end of the day, who cares about that? Users do, and we really can't give them moire reasons to use Windows over Linux, please. (perhaps we need some reproducible benchmarks, like all the competitive effort that went into making Web servers work faster)
Anyway, Windows has had 2 schedulers for ages - you can select desktop or server style processing (and cache strategy) since NT4.
Slashdot Mirror
Android, WebOS, Moblin, Maemo... when are these 'convergent device' OSes going to converge to a single Linux distro?
LSB, we have a new problem for you to tackle!
(oh, and when are we going to see these as desktop OSs?)
I understand now - strangely this is one /. thread where I have learned something by reading all the posts! (though I have had to be sceptical about some of them, but there's been a broad consensus about how it works).
It seems that it is an easier way to implement threading in the apps too, so not only does the OS schedule these as needed, the developer gets to implement his threaded app very easily too. Someone said its "islands of serialisation in a sea of parallelism" (or something like that), and I get that too - its like the old message passing middleware system I used to work on.
All in all, I think its great, if it can be implemented in the linux kernel too (but with an obj-C interface) then we can all benefit. And Windows will be playing catch-up while Linux devs get all the parallel processing goodness and Windows devs have to kludge up their own mechanisms app by app. (yes, I'm a Windows dev too, but I like the idea of linux getting the better toys nowadays).
The question is... how is this different to Intel's Thread Building Bocks, or OpenMP, both of which are better supported and more widely available to non-Mac developers.
I guess having their own implemntation (for the mac) makes sense, as they can integrate it throughout the OS. I don't know anything about GCD either, but could it be used in Linux to make that a more parallel-friendly OS with less developer effort, and more standardisation of parallel execution?
Linux is always playing catch-up with Windows features, its could be nice it got something truly fancy that used all those 4096 CPUs it can support :)
There are people who think that the only software that should be used is Microsoft's. I suppose that's a similarly extreme view, however, in the case of the GPL licence(s) if he takes the view that all software should follow the open source methodology, then only having the GPL as a licence is not so extreme. Perhaps he means all OSS should be licenced under the GPL only, I think that's the view of the FSF.
Incriminating? That's an opinion piece, which barely describes what he's opinioning about. I think it had something to do with a discussion regarding the old hippie 'property is theft' mantra, but I'm not sure.
Mainstream companies are beginning to get with this anyway, very many are into open source as a free product that is monetised in non-standard ways. Its no longer about making money by selling software, but selling services. That's hardly an extreme viewpoint anymore, especially when you consider you never buy software in the first pace, but a licence to use it.
Or... it was more a marketing campaign right from the start. How else would you describe the large number of self-congratulatory 'Microsoft loves linux, really' blogs and announcements, while the actual code (the real work part of this) is so poor?
If Microsoft was serious about this code, about assisting Linux with the Hyper-V integration, then they would be at least helping to support the code themselves.
is that better than people looking at the packaging, notice it says no linux support, and sticking with Windows?
User mode drivers are obviously the way forward then, with a 'Linux certification' before you can install them in the kernel. Or a crash dump when things go wrong that firmly puts the blame on the faulting driver (in big letters).
Its the annual meeting apparently. A chance for Ballmer to high-five the devs who did such a great job bringing the experience of Vista service pack 3 to the masses, and helping to firm up their revenues.
Some of the comments from 'softies in that blog are interesting. They certainly don't like Zune, think Search has spent a bit too much money, and there are a few too many project managers at MS.
really? Would you like to back that up with a link to somewhere slightly reputable?
Nilfs? Nerds I'd Like to .. erm ... oh dear.
I think the point of a driver ABI is that people could write drivers entirely separately to the kernel, and they wouldn't need recompiling for each minor release. Just that would encourage moving the drivers out of the kernel to a separate repository. That would make it easier to manage as the number of them increases.
Windows does have a lot of crappy drivers, but they also have loads and loads of good ones. Enough good ones that a lot of people don't complain about them.
Besides, is it more important to have a buggy driver, or no driver at all? Remember that a buggy driver that has its source released can have its bugs fixed by others. You don't get that in the Windows world.
I guess a lot of manufacturers will release binary-only drivers, but even if they are buggy, that doesn't leave in any worse state than we are already - those manufacturers aren't releasing drivers for Linux in the first place.
I think a stable ABI would be a good thing, and would only inconvenience a kernel dev a small amount. In the past it made sense not to have the ABI as things changed too much, but Linux is a much more stable beast nowadays (or should be!) so a stable ABI, even if its only for major releases, would be a benefit to Linux. If the ABI needs changing, then the kernel can be released with a major point increase.
The thing that struck me with the memory graph is the labels: If I have 2Gb RAM installed, does that show up on the 1-2Gb or the 2-3Gb bar?
I see. I never saw that post, so I take umbrage that I was criticised for something I never even saw, let alone responded to. I don't think the parent of the post I replied to has anything much to do with my reply.
Not that I care too much, perhaps its a lesson that simply being critical of others is a bad idea.
There are plenty of music players out there that don't have the bells and whistles that are so annoying (get it, ha, get it? ok..) I recommend a Sony, I have a NW-S705F which is simple and very effective, but there are plenty of video-style ones as well as conventional music-only players. come with good noise-cancelling headphones too.
Or you could get a Zune I suppose :0
actually, just checking the parent post I replied to - it said:
"Pan's Labyrinth may not be for everyone, but it's is certainly among the top tier in movies. Maybe you should stick to Disney if you want something less depressing."
Where did he say "Pan's Labyrinth sucked."?
You need to brush up on *your* reading before you start casting aspersions toward others.
na, I just look at the pretty pictures :)
Tom Bombadil could be a very good, fun movie.
The Silmarillion on the other hand... might be a tiny, little bit too confusing for today's audiences.
nobody said Pan's Labyrinth isn't a good film. What they said was, its not the style required for the film adaptation of The Hobbit.
The last thing we want is for Smaug to rear up, spread his wings and look like an evil peacock.
tell them that it sends their browsing habits to Microsoft (to make sure they don't download any pirate copies of Windows) and they'll instantly get *very* concerned.
If you made her work late by 20 minutes, or dock her pay, I doubt she'd consider that "not a big deal". Try it :)
While things may perform a vast amount better at idle you get worse performance under load which is when you want it.
not necessarily, today power saving and efficiency at idle is a more desirable thing than performance under load.
That's the point, there's almost certainly no 'right' answer, it depends too much on the task you want your computer to do. Having a single scheduler to fit all cases is never going to be right, you will always want something else. There's a post here from someone who runs high-performance computing using the noop scheduler because it gives than "a 30% boost", I say fair enough to that - a extreme case where you want a different scheduler. Back in the ordinary world, we want different schedulers for different uses, whether that be desktop, server, HPC, netbook or a mobile phone. 1 doesn't cut it well enough across *all* those platforms.
The kernel devs could be gods, I doubt they can still come up with something that runs on a both a supercomputer and a mobile phone without making compromises that affect both of them. Better to tune each of them according to their needs.
OOh. I've just seen the 'thought for the day' at the bottom of the page:
"One size fits all": Doesn't fit anyone.
Even the gods of slashdot are getting in on the debate.
Everyone considers the netbook market. Why won't anyone think of the mobile phones?!
I think what you want is not a single scheduler designed for the desktop, but one designed for server processes. That's probably the whole argument here - there isn't a single scheduler that can work efficiently for the 2 wildly different types of work a user put a machine to, but currently you don't have a choice. This is all about giving users choice of what kind of scheduler they'd like to run. You might even find that a scheduler designed for lots of CPUs (at the expense of interactivity probably) would suit you much more than the current system, especially when you buy more cores.
I was hoping Windows 2000 would achieve that, then I was hoping Windows XP would achieve that, then I was hoping some of the newer 2.6 kernels in Linux coupled with innovations in X would achieve that - but I was always deeply, utterly disappointed. Then I kinda hoped Vista would get somewhat close to what BeOS did. Oh yeah, now that was a hope decisively smashed.
Yup, but apparently Microsoft 'listened' to the screams of outrage, hate and despair that the users of Vista cried out, and changed things in W7 to make user repsonsiveness much more important. I'd like to say they went and listened to feedback from their users, but it was hard not to hear it.
From some reviews I've read, the decision was more a marketing one - W7 is just as slow as Vista in many parts, but because the UI responds to the user quicker (even though the time the operation takes is just the same) everyone thinks that W7 is a much more usable, ans faster OS. It isn't, its just more responsive. But at the end of the day, who cares about that? Users do, and we really can't give them moire reasons to use Windows over Linux, please. (perhaps we need some reproducible benchmarks, like all the competitive effort that went into making Web servers work faster)
Anyway, Windows has had 2 schedulers for ages - you can select desktop or server style processing (and cache strategy) since NT4.