Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: AmigaPixel on February 14, 2009, 09:09:29 AM
-
I have been delving into the architecture and applications of Cell processor
by Sony Computer Entertainment, Toshiba Corporation,and IBM, and also the GPGPU
or use of graphics processing units for general computing applications.
There are two thoughts that come to mind, It seems to me there is a way to take advantage of this in
Amiga OS and hardware development, maybe even in a cost effective way. The other thing I noticed is that these current type of developments are very much inline with the original Amiga philosophy. The idea of off loading to tasks to other chips is much like the way the Amiga used custom chips. The Amiga had much less latency issues because of the way it handled operations efficiently. Take a look at the links below, it's a lot of reading but well worth it.
http://en.wikipedia.org/wiki/CUDA
http://en.wikipedia.org/wiki/GPGPU
http://en.wikipedia.org/wiki/Cell_processor
-
AmigaPixel wrote:
I have been delving into the architecture and applications of Cell processor
by Sony Computer Entertainment, Toshiba Corporation,and IBM, and also the GPGPU
or use of graphics processing units for general computing applications.
Have you missed OpenCL?
http://en.wikipedia.org/wiki/OpenCL
read read read...
There are two thoughts that come to mind, It seems to me there is a way to take advantage of this in
Amiga OS and hardware development, maybe even in a cost effective way. The other thing I noticed is that these current type of developments are very much inline with the original Amiga philosophy. The idea of off loading to tasks to other chips is much like the way the Amiga used custom chips.
The Amiga custom chips were very specialized devices... Much more like the gfx chips in older gfx cards, than these new GP devices...
The Amiga had much less latency issues because of the way it handled operations efficiently. Take a look at the links below, it's a lot of reading but well worth it.
The Amiga was efficient because it disregarded things like effective memory management... which causes problems like memory fragmentation and serious security issues!
-
Thanks for the link great info. I wouldn't be surprised if Open CL pushes CUDA out since it is across platforms and brands. These GPU's may be off the shelf and not specialized devices in the old sense, but they are programmable and scalable and with API's like Open CL a lot of power will be harnessed. The Cell cpu offers some interesting possibilities too, it will be interesting to keep watch.
-
AmigaPixel wrote:
Thanks for the link great info. I wouldn't be surprised if Open CL pushes CUDA out since it is across platforms and brands.
OpenCL will and in fact has already displaced all other forms of parallel execution APIs...
These GPU's may be off the shelf and not specialized devices in the old sense, but they are programmable and scalable and with API's like Open CL a lot of power will be harnessed.
There is a lot of power, but it is useful for high power applications... the types of applications that would never been run on a system that doesn't even offer memory protection.
The Cell cpu offers some interesting possibilities too, it will be interesting to keep watch.
If development work continues then it might be able to keep up with the big players; AMD, Intel and Nvidia... but I'm not sure where the motivation will come from for the Cell...
-
One disadvantage of Open CL it is a lower level language that requires coders to write their own memory protection.
-
bloodline wrote:
AmigaPixel wrote:
Thanks for the link great info. I wouldn't be surprised if Open CL pushes CUDA out since it is across platforms and brands.
OpenCL will and in fact has already displaced all other forms of parallel execution APIs...
"Already displaced..." by something that doesn't yet exist? Seems a little to early to call that. IMHO, two reasons this won't ever get the traction it needs.
1) Khronos - Bad track record for OpenGL release/refresh problems, now adding this? Good luck.
2) Microsoft - DirectX 11 has GPGPU features too, so I doubt anybody will switch to a slower OpenGL/OpenCl implementation on Windows.
-
bloodline wrote:
AmigaPixel wrote:
I have been delving into the architecture and applications of Cell processor
by Sony Computer Entertainment, Toshiba Corporation,and IBM, and also the GPGPU
or use of graphics processing units for general computing applications.
Have you missed OpenCL?
http://en.wikipedia.org/wiki/OpenCL
You'd like to see some AmigaOS/AROS written in OpenCL?
The Amiga was efficient because it disregarded things like effective memory management... which causes problems like memory fragmentation and serious security issues!
Hmyes, :juggler:
I wonder if this memory management could have been, well, accelerated with a blitter kind of chip.
I think there just should be more dedicated silicium.