I was wondering how someone could be so ignorant on GCD and OS X, then I looked at the post count. D'oh
Grand Central is a parallel computation framework. Please, do explain how it fixes substandard (from a parallelism point of view) locking schemes in OSX's I/O, IPC, Process, and other subsystems. It doesn't do you much good to efficiently spawn a hundred GPU cores running vector computations if you can't get all that data back to disk. Apple never did have a big iron mentality, if they did they wouldn't have used Mach, they'd have started with any of the several variants of Linux that SGI, IBM, Cray, et al were developing for multi-hundred core applications.
Grand Central is a parallel computation framework. Please, do explain how it fixes substandard (from a parallelism point of view) locking schemes in OSX's I/O, IPC, Process, and other subsystems. It doesn't do you much good to efficiently spawn a hundred GPU cores running vector computations if you can't get all that data back to disk. Apple never did have a big iron mentality, if they did they wouldn't have used Mach, they'd have started with any of the several variants of Linux that SGI, IBM, Cray, et al were developing for multi-hundred core applications.
But that's not what you said.
Let me refresh your memory:"One of the reasons Apple doesn't go with huge numbers of cores is because OSX itself is (or at least was a couple of years ago) a dog at multithreading."
But hey, I give you credit for trying to move the goal posts to a more defensible position. Keep up the good work! Its obvious you are a master at trolling.
Let me refresh your memory:"One of the reasons Apple doesn't go with huge numbers of cores is because OSX itself is (or at least was a couple of years ago) a dog at multithreading."
But hey, I give you credit for trying to move the goal posts to a more defensible position. Keep up the good work! Its obvious you are a master at trolling.
No goalposts have been moved. Parallel computation is but a subset of multithreading. Efficiently doing large parallel computations on disjoint data sets is easy. Saving your results efficiently without locking contention killing you is hard. That's why I'm talking about I/O and IPC: they're the bottlenecks in OSX, which GC doesn't address.
It would seem to me that Apple would be well aware of any such constraints and are also looking toward the future of computing when determining how they are working with OS X. They know that all of their apps are going to go 64-bit. They know that all of their apps are going to require far more resources than they ever did before because technology is advancing. This isn't new to any of the folks in Cupertino.
I think that more cores are the way of the future. I think we're going to see far more than two chips per computer. I still remember the first time I heard that Apple was dropping two chips into their Mac Pro. It blew my mind. Now, here I am in 2010, using a Mac Pro that's already a couple of years old, and I've got 8 cores at my disposal at 3 GHz apiece. I've got 13 GB of RAM to give it some elbow room as well. It stands to reason that in a few years time, I'll look back on this Mac just like I currently look back on my 600 MHz G3 iBook with 2 GB of RAM in it. We're progressing.
If Apple goes with AMD and significantly steps up the amount of cores available to OS X, then it stands to reason that they are also looking into making the OS handle that capacity the very best way possible. I'm not someone that deals with these issues, but I know those people at Cupertino get paid a buck or two more than I do in order for them to make those decisions. I know that I love using my 8-cores as opposed to my brother's 2-cores when it comes to even simple things like editing HD 1080i footage in iMovie. I don't think Intel is really looking to make our lives easier with their pricing schemes. I think AMD is going to be where it's at when it comes to getting the most cores for our buck.
I think in a few years I'll look back and wonder how it is that I was able to get anything done with just 8 cores.
I very much like Phenom X6 1090T because its benchmarks is higher. Because its current entry Mac Pro is quite significantly while costing is same. AMD have scrapped the 4P tax, the higher end models could be turned into a 4P machine.
Comments
Don't feed the trolls. They're on a strict diet of FUD.
I was wondering how someone could be so ignorant on GCD and OS X, then I looked at the post count. D'oh
I was wondering how someone could be so ignorant on GCD and OS X, then I looked at the post count. D'oh
Grand Central is a parallel computation framework. Please, do explain how it fixes substandard (from a parallelism point of view) locking schemes in OSX's I/O, IPC, Process, and other subsystems. It doesn't do you much good to efficiently spawn a hundred GPU cores running vector computations if you can't get all that data back to disk. Apple never did have a big iron mentality, if they did they wouldn't have used Mach, they'd have started with any of the several variants of Linux that SGI, IBM, Cray, et al were developing for multi-hundred core applications.
Grand Central is a parallel computation framework. Please, do explain how it fixes substandard (from a parallelism point of view) locking schemes in OSX's I/O, IPC, Process, and other subsystems. It doesn't do you much good to efficiently spawn a hundred GPU cores running vector computations if you can't get all that data back to disk. Apple never did have a big iron mentality, if they did they wouldn't have used Mach, they'd have started with any of the several variants of Linux that SGI, IBM, Cray, et al were developing for multi-hundred core applications.
But that's not what you said.
Let me refresh your memory:"One of the reasons Apple doesn't go with huge numbers of cores is because OSX itself is (or at least was a couple of years ago) a dog at multithreading."
But hey, I give you credit for trying to move the goal posts to a more defensible position. Keep up the good work! Its obvious you are a master at trolling.
But that's not what you said.
Let me refresh your memory:"One of the reasons Apple doesn't go with huge numbers of cores is because OSX itself is (or at least was a couple of years ago) a dog at multithreading."
But hey, I give you credit for trying to move the goal posts to a more defensible position. Keep up the good work! Its obvious you are a master at trolling.
No goalposts have been moved. Parallel computation is but a subset of multithreading. Efficiently doing large parallel computations on disjoint data sets is easy. Saving your results efficiently without locking contention killing you is hard. That's why I'm talking about I/O and IPC: they're the bottlenecks in OSX, which GC doesn't address.
I think that more cores are the way of the future. I think we're going to see far more than two chips per computer. I still remember the first time I heard that Apple was dropping two chips into their Mac Pro. It blew my mind. Now, here I am in 2010, using a Mac Pro that's already a couple of years old, and I've got 8 cores at my disposal at 3 GHz apiece. I've got 13 GB of RAM to give it some elbow room as well. It stands to reason that in a few years time, I'll look back on this Mac just like I currently look back on my 600 MHz G3 iBook with 2 GB of RAM in it. We're progressing.
If Apple goes with AMD and significantly steps up the amount of cores available to OS X, then it stands to reason that they are also looking into making the OS handle that capacity the very best way possible. I'm not someone that deals with these issues, but I know those people at Cupertino get paid a buck or two more than I do in order for them to make those decisions. I know that I love using my 8-cores as opposed to my brother's 2-cores when it comes to even simple things like editing HD 1080i footage in iMovie. I don't think Intel is really looking to make our lives easier with their pricing schemes. I think AMD is going to be where it's at when it comes to getting the most cores for our buck.
I think in a few years I'll look back and wonder how it is that I was able to get anything done with just 8 cores.