GPGPU (Cuda,Ati Stream)

[DevilBlade]

GPGPU is when you use a Graphics card to process something you would normally run on a CPU. If Algdoo/Phun could take advantage of this we could run much more complicated Scenes. Here are a few examples
ATI Stream
http://www.youtube.com/watch?v=E3ANHaajCPo
Nvidia Cuda/Physix
http://www.youtube.com/watch?v=RqduA7myZok

As you can seen these are rendering 3D Fluid Simulations with ease Unfortunately if you have a Intel Gpu your out of luck. These demo are running on pretty low priced graphics cards and even with lower end it would realeave some of the load from the cpu. Next year we will have 28nm Gpus. GPU are many times faster. They compute many more Flops then CPUs
What are Flops?
Floating point Operations Per Second. The FLOPS is a measure of a computer's performance, especially in fields of scientific calculations that make heavy use of floating point calculations.

Examples
Gpus
ATI 5870: 2 TeraFlops 437$
ATI 5970: 4 TeraFlops $649.99
Nvidia GTX 295: 1.7 Terraflops 519$
Nvidia 8600GT: 113.28 Gigaflops 40$
Cpus
Intel Core i7 CPU: 100 Gigaflops $989.99
AMD 965 Cpu: 80 gigflops 190$

As you can see The performance increase should be good if this could be implemented into phun/algodoo.

[gradyfitz]

GPGPU is when you use a Graphics card to process something you would normally run on a CPU. If Algdoo/Phun could take advantage of this we could run much more complicated Scenes. Here are a few examples
ATI Stream
http://www.youtube.com/watch?v=E3ANHaajCPo
Nvidia Cuda/Physix
http://www.youtube.com/watch?v=RqduA7myZok

As you can seen these are rendering 3D Fluid Simulations with ease Unfortunately if you have a Intel Gpu your out of luck. These demo are running on pretty low priced graphics cards and even with lower end it would realeave some of the load from the cpu. Next year we will have 28nm Gpus. GPU are many times faster. They compute many more Flops then CPUs
What are Flops?
Floating point Operations Per Second. The FLOPS is a measure of a computer's performance, especially in fields of scientific calculations that make heavy use of floating point calculations.

Examples
Gpus
ATI 5870: 2 TeraFlops 437$
ATI 5970: 4 TeraFlops $649.99
Nvidia GTX 295: 1.7 Terraflops 519$
Nvidia 8600GT: 113.28 Gigaflops 40$
Cpus
Intel Core i7 CPU: 100 Gigaflops $989.99
AMD 965 Cpu: 80 gigflops 190$

As you can see The performance increase should be good if this could be implemented into phun/algodoo.

Algoryx are implementing this already (you might probably imagine what kind of work goes into making a program very parallel), also, what makes GPUs more difficult to use as general processors is that you have a very limited amount of functions (compared to that of a CPU), also, they aren't too good compared to CPUs because of their connection, it is often faster to redo a calculation than to store it to memory. Anyway, I don't see too much of a lag problem in Algodoo, but still, I'm always up for being able to do things that are ridiculous otherwise (like simulating 20 million particles or something crazy like that (for submarine machines)).

CUDA is already being implemented and is on the Roadmap for Algodoo, implementing parallel computing methods not only allows for GPGPU on Graphics Cards but also allows for using all cores in all the CPUs a user may have .

Although this is a good suggestion, it's already been suggested and is well on the way to being implemented (at least that's what I understand of it).

--
Grady

[DevilBlade]

Ok i dug a little and never found anything about this but thats cool there already working on it