It had more to do with the terminology you used than any bad english.
Quote:
Originally Posted by apfel
But that is the direction about I am speaking. And there was much hope for OpenGL3.
Thats true. And on mobile devices OpenGL ES will be big. But don't forget OpenGL ES != OpenGL.
I disagree on OpenGL ES != OpenGL. I would say it is more like: OpenGL ES ~ OpenGL. Meanwhile, I would say DirectX != OpenGL (nor ES). That seems like a more accurate assessment to me.
Quote:
Originally Posted by apfel
Don't understand me wrong. I would like to see more OpenGL instead of Direct-X. But unfortunately in Games I don't believe it will happen, also because they failed with OpenGL3. Windows is the biggest market on the Desktop and DirectX > OpenGL. I hope it will change (but there was an opportunity and they didn't use it) and that maybe Apple is doing some pressure in the future to make a cleanup in OpenGL.
Guess it depends on how you measure success. Obviously DirectX has a larger user base than OpenGL because DirectX is tied to Windows, which is the OS on around 90% of the world's computers. If we go by that metric, then obviously DirectX wins. But that's an overly simplistic answer that doesn't really mean anything.
Serious PC gaming is a niche market and its miniscule percentage of the overall market is shrinking thanks to the advent of console gaming and a few major casual PC games like WoW that don't require a high-end computer to run which have hurt PC game sales because people just play WoW all the time without investing in other games (and in particular, one's that make use of DX9 or DX10).
The PS2, which used an OpenGL-like API, was hugely successful last generation and the Wii, which also uses an OpenGL-like API, is hugely successful this generation. Microsoft's Xbox and Xbox 360 have both wound up fighting for a distant second place (with the GameCube last generation and with the PS3 this generation). In addition, DirectX PC ~ DirectX Xbox. Back on the computer side, Apple's Macs, which make exclusive use of OpenGL, are outpacing the flat PC industry 4 to 1 and their OpenGL ES iPhone has already become the #1 phone in the US, with thousands of OpenGL games available in the App Store after only a few months (not to mention 3D games are the most popular category on the App Store).
So considering the whole picture, OpenGL doesn't have to be perfect or even superior to unseat DirectX; in many ways it already has.
This is no surprise to anyone who's done a little curiosity digging in the iphone SDK, or in a jailbroken iPhone.
My suggestion for Apple Insider: A lot more accurate and timely information can be had from intelligent deduction and weak (always keep it weak) extrapolation from currently quite freely available data, and in a way that in no way would ever put you in legal trouble with Apple.
Are any of you familiar with the old "strings" command? Go find an old UNIX first or second generation UNIX head, a good one, and throw OS X at him (there are no her good ones anymore, the last one worked for IBM and retired just this year), and then ask him in a couple of days what's interesting.
... but there is a limited class of tasks for which this is useful. OpenCL is far more useful on the desktop than on the phone.
But then you make this statement which I can't agree with. Yes the class of apps where OpenCL is useful is limited but how does that make it less useful on iPhone? If anything it is more useful as it would allow for apps that might not be viable due to the processors performance.
In other words OpenCL on iPhone will enable a greater range of apps. That is very useful but moreso on newer generation Touch based devices that everyone is hoping for.
Quote:
C.
As a side note it appears that the current generation iPhone's will NOT be able to leverage OpenCL. I believe that Apple is very serious about OpenCL thus think that we will see the PA Semi chips a lot sooner than many here think. I'd say well before 2010. OpenCL might not be ready for them but Apple will want to get compatible hardware out in the field soon just like they have done with the laptops.
Apparently OpenCL on PowerVR requires the latest SDX core. This is not a big deal in my mind as Apple has several development paths they could follow. One would be the latest GPU on a tryed and true ARM11 core. This would be very useful for the lower end hardware and allow Apple to quickly do an incremental update to the current IPhone hardware. Think early 2009!
For the more powerful devices a Cortex based SoC would follow. This would be a far more powerful design with exceptional clock rate. The reasons for at least 2 SoC is one to reduce Apples exposure to new tech. Second to have power balanced SoC for the class of device in question. IE iPhone Nano doesn't need a high performance solution.
As to exposure, Cortex is a different instruction set and a vastly improved ARM chip. It might take awhile to wring out the bugs. I would suspect though that Apple already has the hardware in hand to do just that.
In my [rather limited] understanding, openCL is NOT designed to be a development language for graphics apps. That is the task of the openGL.
Sure but part of the problem with OpenGL is that people have tried to get it to do more things than the GPUs were capable of so they added parts to the language to support custom features on certain GPUs. OpenCL can signal a starting point where all GPGPUs can run the same code and simply scale it down if they aren't fast enough - this leaves OpenGL to be a much cleaner API as the complicated diverse stuff can be external OpenCL libraries that are bundled with the application rather than being GPU-specific. Leave all the basic common functionality to OpenGL and let OpenCL handle the more diverse stuff like shader computation, procedural textures etc as well as some calculations.
Shader computation is one of the slowest parts of the graphics pipeline and if the GPU can do it in parallel without choking up the CPU then it should do it.
They said there won't be any built-in functions to help though. This is why I think development will take a bit of time to get going. One of the examples in the OpenCL presentation was for Fourier Transform code or something. The building blocks need to be in place before really useful code can start to run.
Quote:
Originally Posted by jmmx
The end of this all is that an next gen. iPhone/Touch will have amazing compute capabilities. I am sure it will absolutely blow anything else out of the water. It will be near desktop compute power in your hand. Also add in that one of the main objectives of Snow Leopard is to reduce the memory Paw Print... WOW!
On the mobile end, I see it more as a load-balancing feature and power-saving feature than being able to achieve desktop performance from a mobile device. The next-gen PowerVR SGX spec says it can handle 16 threads:
The current iphone isn't that powerful and clock speed of the cores is quite low. The CPU is a single core ARM 400-500MHz chip. The GPU is only 133MHz. I don't think that it will come close to desktop performance but it depends on the benchmark used. Technically, the iphone is already faster than desktops 10-15 years ago. With OpenCL, I can see the current generation iphone coming close to a 1GHz desktop but at a stretch. Certainly not any modern Core Duo desktop.
Once they get multi-core ARM or Intel delivers an Atom-like processor that competes with it then that can happen. I personally don't think there's an urgency for desktop performance from the iphone. I find the iphone runs extremely well for its purpose.
OpenCL offers a standardized abstraction for offloading repetitive computational tasks onto different GPU hardware. OpenCL facilitates a certain class of computations. I can't think of many examples where it would be useful in a handheld. But I'd be interested to hear examples.
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
Ok, now I get it. It is as if someone slammed a hammer to my head (ouch!). OpenCL's rush out-the-door is not so much for the Macintosh lines but for the iPhone lines. OSX core API are used for the iPhone OS, so what is developed for OSX migrates to the iPhone OS.
By June 2009 we'll see an impressive iPhone with the fruits of the PA Semi acquisition in a new CPU core and a PowerVR core saddled up next to it. Apple wants to dominate the handheld video game arena obviously. Duh!
Snow Leopard will benefit but I was thinking how much whiz-bang do we need in the OS (Apple?). It makes more sense for the handheld market to take advantage of OpenCL.
Is iPhone gaming restricted more by processing power, or by battery life?
OpenCL offers a standardized abstraction for offloading repetitive computational tasks onto different GPU hardware. OpenCL facilitates a certain class of computations. I can't think of many examples where it would be useful in a handheld. But I'd be interested to hear examples.
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
They will be sitting on top of the same proprietary investment, but leaking on all seams. Direct X, MSIE, Office, and all the other shite they have been stuffing down our throats for years will eventually fail as the world returns to the open standards that prevailed before the Microsoft era.
Please forgive my ignorance on this but do you mind elaborating on what "open" standards there were before the Microsoft era? Are you referring to the 70's when Microsoft was just a bit-player or during the 90's when it started taking off? I don't recall much for open standards during those times. Companies back then had their own ideas but weren't necessarily open to the free world to take it and run with it.
OpenCL offers a standardized abstraction for offloading repetitive computational tasks onto different GPU hardware. OpenCL facilitates a certain class of computations. I can't think of many examples where it would be useful in a handheld. But I'd be interested to hear examples.
I can think of several ideas off the top of my head. Mind you I'm by a cell phone engineer so i'm not up on moser approaches. But here is a few ideas to throw in the wringer:
1.
Acceleration of camera functionality. This might include realtime application of corrections to pre sampled picture. That is apply white balance and other corrections in realtime without the smearing and tearing as is common today. If that can be done at 24 frames a second then all the better. There actually are a number of interesting possibilities here.
2.
Acceleration of voice and video compression for video phone service. Apparently though the new PowerVR chip has dedicated logic to handle this bit whatever it can't handle the GPU can help with.
3.
If the ARM and GPU combo become powerful enough signal processing for 3G could be offloaded to the combined resources. This would vastly alter the iPhone's architecture but there is a lot of potential gain. For example the iPhone doesn't have to power all the various CPUs in the outlying chips. Different people have estimated that there may be as many as six ARM processors in the current iPhone. The processors being embedded in various parts of the phone. Now I don't know the exact number but the point is if you can replace even some of these with an SMP ARM supplemented by the GPU you gain advantages in design and possible power usage. Plus when those chips are tied up with phone signal processing you have a lot more power for apps. Frankly Qualcomm hinted at this sort of appraoch in documents for one of it's up and coming processors, well at least the fast SMP approach, OpenCL wasn't even heArd of back then.
4.
We shouldn't forget that in simple terms the iPhone/Touch are audio devices and there is always room for innovation in audio processing. Here the GPU might be used to process recorded audio or maybe raw signals from analog inputs. One possibility is to have the iPhone function as a hearing aid with equalization.
5.
Kinda related but iPhone could handle signal processing for a software defined radio give the analog inputs to do so. If a software defined radio is to much, in the next performance increment, it certainly could handle morse code processing.
6.
Acceleration of compression codes. Especially useful if maps from google could be transmitted faster. In any event any off load here would be helpful.
7.
Resolution independence! Actually this may enable something Apple has been trying to deliver for some time now. But think this way if everything one IPhone suddenly became a vector graphic how mich CPU power do you think we would need. Further would having a bunch of real threads help? It isn't talked about much anymore but OpenCL may be one of those technologies that needs to happen before resolution independence. There is no doubt in my Mind that it will require considerable system resources. Let's just say that between ARM and PowerVR Apple can have thirty six threads available to them, would that do the resolution independence trick? Imagine too that thirty two of those processors are optimized for fast math operations.
8.
Maybe an extra thread will allow for keyboards that work right all the time on IPhone. I only bring this up because right now I'm not always getting the key hit visual feedback (another bug). Maybe get rid of the lag that still crops up from time to time. This is in the realm of trying to reduce processor loading.
Well that is a start and are things off the top of my head. I'm sure others could add to the list. It should also be noted that some of these things could be handled by another ARM core. At the low end I don't think Apple will go SMP when the required GPU is there to do some heavy lifting. The other thing is we really don't know just how general these 32threads will be. There is a lot of variation in GPU processors. Hopefully apple and Imagine worked together on this baby.
Quote:
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
Yes and it is widely used, contrary to comments by others in the various threads.
Quote:
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Well I have to say you got that wrong. GPUs have developed the way they have because MicroSoft dictated in detail exactly what they needed accelerated.
Now we can only hope that Apple has had some influence with respect to the way PowerVR has been developed. Considering the newest GPU has support for OpenCL one might think that is a bright spot.
Quote:
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
No body was forced to do anything. Frankly I suspect that these vendors liked the idea of proprietary code. It is the mentality of the lock in. In any event Apple seems to get by fine supporting OpenGL ES.
If there was a williness to work together then OpenGL might look different today. I see a positive sign though that efforts are afoot, lead by Apple, to make nice with OpenGL. Apple as a large company can have a major impact on moving the standard forward. They just have to be careful not to alienate a large subset of users.
Quote:
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
C.
Stabilized GPUs??? I honestly don't see that. Much is happening in GPU land even at Intel.
As for Kronos yeah at the moment it is looking better but they need to be careful not to become "Apples" standard body. A little bumping of heads was needed to push some things forward but Apple just needs to be careful not to ruin a good thing.
Is it possible AI could actually publish a new story????
I know you guys are USA focussed and basically shut down for the weekend, but geez! Nothin to write about? This story has been up for quite a bit over 24 hours now, which on the Internet is like a week.
MDN has five or six new articles this morning if you need to steal one or link to something.
Is it possible AI could actually publish a new story????
I know you guys are USA focussed and basically shut down for the weekend, but geez! Nothin to write about? This story has been up for quite a bit over 24 hours now, which on the Internet is like a week.
MDN has five or six new articles this morning if you need to steal one or link to something.
Comments
Yes - I am sorry for my bad english.
It had more to do with the terminology you used than any bad english.
But that is the direction about I am speaking. And there was much hope for OpenGL3.
Thats true. And on mobile devices OpenGL ES will be big. But don't forget OpenGL ES != OpenGL.
I disagree on OpenGL ES != OpenGL. I would say it is more like: OpenGL ES ~ OpenGL. Meanwhile, I would say DirectX != OpenGL (nor ES). That seems like a more accurate assessment to me.
Don't understand me wrong. I would like to see more OpenGL instead of Direct-X. But unfortunately in Games I don't believe it will happen, also because they failed with OpenGL3. Windows is the biggest market on the Desktop and DirectX > OpenGL. I hope it will change (but there was an opportunity and they didn't use it) and that maybe Apple is doing some pressure in the future to make a cleanup in OpenGL.
Guess it depends on how you measure success. Obviously DirectX has a larger user base than OpenGL because DirectX is tied to Windows, which is the OS on around 90% of the world's computers. If we go by that metric, then obviously DirectX wins. But that's an overly simplistic answer that doesn't really mean anything.
Serious PC gaming is a niche market and its miniscule percentage of the overall market is shrinking thanks to the advent of console gaming and a few major casual PC games like WoW that don't require a high-end computer to run which have hurt PC game sales because people just play WoW all the time without investing in other games (and in particular, one's that make use of DX9 or DX10).
The PS2, which used an OpenGL-like API, was hugely successful last generation and the Wii, which also uses an OpenGL-like API, is hugely successful this generation. Microsoft's Xbox and Xbox 360 have both wound up fighting for a distant second place (with the GameCube last generation and with the PS3 this generation). In addition, DirectX PC ~ DirectX Xbox. Back on the computer side, Apple's Macs, which make exclusive use of OpenGL, are outpacing the flat PC industry 4 to 1 and their OpenGL ES iPhone has already become the #1 phone in the US, with thousands of OpenGL games available in the App Store after only a few months (not to mention 3D games are the most popular category on the App Store).
So considering the whole picture, OpenGL doesn't have to be perfect or even superior to unseat DirectX; in many ways it already has.
My suggestion for Apple Insider: A lot more accurate and timely information can be had from intelligent deduction and weak (always keep it weak) extrapolation from currently quite freely available data, and in a way that in no way would ever put you in legal trouble with Apple.
Are any of you familiar with the old "strings" command? Go find an old UNIX first or second generation UNIX head, a good one, and throw OS X at him (there are no her good ones anymore, the last one worked for IBM and retired just this year), and then ask him in a couple of days what's interesting.
Or, probably a little easier, just ask Stallman.
Exactly right...
Yes he hit the bulls eye!
... but there is a limited class of tasks for which this is useful. OpenCL is far more useful on the desktop than on the phone.
But then you make this statement which I can't agree with. Yes the class of apps where OpenCL is useful is limited but how does that make it less useful on iPhone? If anything it is more useful as it would allow for apps that might not be viable due to the processors performance.
In other words OpenCL on iPhone will enable a greater range of apps. That is very useful but moreso on newer generation Touch based devices that everyone is hoping for.
C.
As a side note it appears that the current generation iPhone's will NOT be able to leverage OpenCL. I believe that Apple is very serious about OpenCL thus think that we will see the PA Semi chips a lot sooner than many here think. I'd say well before 2010. OpenCL might not be ready for them but Apple will want to get compatible hardware out in the field soon just like they have done with the laptops.
Apparently OpenCL on PowerVR requires the latest SDX core. This is not a big deal in my mind as Apple has several development paths they could follow. One would be the latest GPU on a tryed and true ARM11 core. This would be very useful for the lower end hardware and allow Apple to quickly do an incremental update to the current IPhone hardware. Think early 2009!
For the more powerful devices a Cortex based SoC would follow. This would be a far more powerful design with exceptional clock rate. The reasons for at least 2 SoC is one to reduce Apples exposure to new tech. Second to have power balanced SoC for the class of device in question. IE iPhone Nano doesn't need a high performance solution.
As to exposure, Cortex is a different instruction set and a vastly improved ARM chip. It might take awhile to wring out the bugs. I would suspect though that Apple already has the hardware in hand to do just that.
Dave
In my [rather limited] understanding, openCL is NOT designed to be a development language for graphics apps. That is the task of the openGL.
Sure but part of the problem with OpenGL is that people have tried to get it to do more things than the GPUs were capable of so they added parts to the language to support custom features on certain GPUs. OpenCL can signal a starting point where all GPGPUs can run the same code and simply scale it down if they aren't fast enough - this leaves OpenGL to be a much cleaner API as the complicated diverse stuff can be external OpenCL libraries that are bundled with the application rather than being GPU-specific. Leave all the basic common functionality to OpenGL and let OpenCL handle the more diverse stuff like shader computation, procedural textures etc as well as some calculations.
Shader computation is one of the slowest parts of the graphics pipeline and if the GPU can do it in parallel without choking up the CPU then it should do it.
They said there won't be any built-in functions to help though. This is why I think development will take a bit of time to get going. One of the examples in the OpenCL presentation was for Fourier Transform code or something. The building blocks need to be in place before really useful code can start to run.
The end of this all is that an next gen. iPhone/Touch will have amazing compute capabilities. I am sure it will absolutely blow anything else out of the water. It will be near desktop compute power in your hand. Also add in that one of the main objectives of Snow Leopard is to reduce the memory Paw Print... WOW!
On the mobile end, I see it more as a load-balancing feature and power-saving feature than being able to achieve desktop performance from a mobile device. The next-gen PowerVR SGX spec says it can handle 16 threads:
http://74.125.77.132/search?q=cache:...k&cd=3&gl=uk#7
The current iphone isn't that powerful and clock speed of the cores is quite low. The CPU is a single core ARM 400-500MHz chip. The GPU is only 133MHz. I don't think that it will come close to desktop performance but it depends on the benchmark used. Technically, the iphone is already faster than desktops 10-15 years ago. With OpenCL, I can see the current generation iphone coming close to a 1GHz desktop but at a stretch. Certainly not any modern Core Duo desktop.
Once they get multi-core ARM or Intel delivers an Atom-like processor that competes with it then that can happen. I personally don't think there's an urgency for desktop performance from the iphone. I find the iphone runs extremely well for its purpose.
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
C.
Exactly!
I second your views. I think apple tablet "technology" will be a new standard in mobile computing. Oh the anticipation
Ok, now I get it. It is as if someone slammed a hammer to my head (ouch!). OpenCL's rush out-the-door is not so much for the Macintosh lines but for the iPhone lines. OSX core API are used for the iPhone OS, so what is developed for OSX migrates to the iPhone OS.
By June 2009 we'll see an impressive iPhone with the fruits of the PA Semi acquisition in a new CPU core and a PowerVR core saddled up next to it. Apple wants to dominate the handheld video game arena obviously. Duh!
Snow Leopard will benefit but I was thinking how much whiz-bang do we need in the OS (Apple?). It makes more sense for the handheld market to take advantage of OpenCL.
Is iPhone gaming restricted more by processing power, or by battery life?
OpenCL offers a standardized abstraction for offloading repetitive computational tasks onto different GPU hardware. OpenCL facilitates a certain class of computations. I can't think of many examples where it would be useful in a handheld. But I'd be interested to hear examples.
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
C.
http://www.mpi-inf.mpg.de/resources/hdr/
For the Medical Industry: being able to render quality X-rays, MRI views and more for consultation?
SVG->Bitmap rasterization on-the-fly.
GIS Mapping with varying layers of complexity.
They will be sitting on top of the same proprietary investment, but leaking on all seams. Direct X, MSIE, Office, and all the other shite they have been stuffing down our throats for years will eventually fail as the world returns to the open standards that prevailed before the Microsoft era.
Please forgive my ignorance on this but do you mind elaborating on what "open" standards there were before the Microsoft era? Are you referring to the 70's when Microsoft was just a bit-player or during the 90's when it started taking off? I don't recall much for open standards during those times. Companies back then had their own ideas but weren't necessarily open to the free world to take it and run with it.
OpenCL offers a standardized abstraction for offloading repetitive computational tasks onto different GPU hardware. OpenCL facilitates a certain class of computations. I can't think of many examples where it would be useful in a handheld. But I'd be interested to hear examples.
I can think of several ideas off the top of my head. Mind you I'm by a cell phone engineer so i'm not up on moser approaches. But here is a few ideas to throw in the wringer:
1.
Acceleration of camera functionality. This might include realtime application of corrections to pre sampled picture. That is apply white balance and other corrections in realtime without the smearing and tearing as is common today. If that can be done at 24 frames a second then all the better. There actually are a number of interesting possibilities here.
2.
Acceleration of voice and video compression for video phone service. Apparently though the new PowerVR chip has dedicated logic to handle this bit whatever it can't handle the GPU can help with.
3.
If the ARM and GPU combo become powerful enough signal processing for 3G could be offloaded to the combined resources. This would vastly alter the iPhone's architecture but there is a lot of potential gain. For example the iPhone doesn't have to power all the various CPUs in the outlying chips. Different people have estimated that there may be as many as six ARM processors in the current iPhone. The processors being embedded in various parts of the phone. Now I don't know the exact number but the point is if you can replace even some of these with an SMP ARM supplemented by the GPU you gain advantages in design and possible power usage. Plus when those chips are tied up with phone signal processing you have a lot more power for apps. Frankly Qualcomm hinted at this sort of appraoch in documents for one of it's up and coming processors, well at least the fast SMP approach, OpenCL wasn't even heArd of back then.
4.
We shouldn't forget that in simple terms the iPhone/Touch are audio devices and there is always room for innovation in audio processing. Here the GPU might be used to process recorded audio or maybe raw signals from analog inputs. One possibility is to have the iPhone function as a hearing aid with equalization.
5.
Kinda related but iPhone could handle signal processing for a software defined radio give the analog inputs to do so. If a software defined radio is to much, in the next performance increment, it certainly could handle morse code processing.
6.
Acceleration of compression codes. Especially useful if maps from google could be transmitted faster. In any event any off load here would be helpful.
7.
Resolution independence! Actually this may enable something Apple has been trying to deliver for some time now. But think this way if everything one IPhone suddenly became a vector graphic how mich CPU power do you think we would need. Further would having a bunch of real threads help? It isn't talked about much anymore but OpenCL may be one of those technologies that needs to happen before resolution independence. There is no doubt in my Mind that it will require considerable system resources. Let's just say that between ARM and PowerVR Apple can have thirty six threads available to them, would that do the resolution independence trick? Imagine too that thirty two of those processors are optimized for fast math operations.
8.
Maybe an extra thread will allow for keyboards that work right all the time on IPhone. I only bring this up because right now I'm not always getting the key hit visual feedback (another bug). Maybe get rid of the lag that still crops up from time to time. This is in the realm of trying to reduce processor loading.
Well that is a start and are things off the top of my head. I'm sure others could add to the list. It should also be noted that some of these things could be handled by another ARM core. At the low end I don't think Apple will go SMP when the required GPU is there to do some heavy lifting. The other thing is we really don't know just how general these 32threads will be. There is a lot of variation in GPU processors. Hopefully apple and Imagine worked together on this baby.
OpenGL offers a standardized abstraction for realtime 2D and 3D rendering on different GPU hardware.
Yes and it is widely used, contrary to comments by others in the various threads.
The problem with OpenGL has been that it has evolved much slower than the GPU hardware. Microsoft was able to evolve D3D in parallel with the hardware - and without the bureaucracy of a heavyweight academic / industrial group. This lack of agility caused Open GL fall behind the times.
Well I have to say you got that wrong. GPUs have developed the way they have because MicroSoft dictated in detail exactly what they needed accelerated.
Now we can only hope that Apple has had some influence with respect to the way PowerVR has been developed. Considering the newest GPU has support for OpenCL one might think that is a bright spot.
Hardware vendors were forced to end-run around the standard to expose their hardware features via extensions. And this in turn resulted in some fairly crappy looking APIs.
No body was forced to do anything. Frankly I suspect that these vendors liked the idea of proprietary code. It is the mentality of the lock in. In any event Apple seems to get by fine supporting OpenGL ES.
If there was a williness to work together then OpenGL might look different today. I see a positive sign though that efforts are afoot, lead by Apple, to make nice with OpenGL. Apple as a large company can have a major impact on moving the standard forward. They just have to be careful not to alienate a large subset of users.
The functionality of GPUs seems to have stabilized, and the Khronos Group seem to be getting their sh*t together.
C.
Stabilized GPUs??? I honestly don't see that. Much is happening in GPU land even at Intel.
As for Kronos yeah at the moment it is looking better but they need to be careful not to become "Apples" standard body. A little bumping of heads was needed to push some things forward but Apple just needs to be careful not to ruin a good thing.
Dave
I know you guys are USA focussed and basically shut down for the weekend, but geez! Nothin to write about? This story has been up for quite a bit over 24 hours now, which on the Internet is like a week.
MDN has five or six new articles this morning if you need to steal one or link to something.
Is it possible AI could actually publish a new story????
I know you guys are USA focussed and basically shut down for the weekend, but geez! Nothin to write about? This story has been up for quite a bit over 24 hours now, which on the Internet is like a week.
MDN has five or six new articles this morning if you need to steal one or link to something.
Or, you could just get a life.