djiamnot

I will be using jitter in a project, mainly for analyzing videos. 2 (maybe more) at the same time and dispatching the analysis data to various receivers (thousands of them, it seems). I do not yet know the specifics (size, framerate etc) but right now I know that we will be choosing hardware specifically for that purpose.

I know, from the docs, that jitter uses the GPU and it can work with multiple CPUs and is both 32 and 64bit friendly. What would be an optimal setup for analysis (I suppose I will be mainly manipulating matrix data, and it is unclear to me whether this kind of work is done on the CPU or the GPU)? Perhaps there is a chart somewhere indicating which specific objects do their number crunching on a CPU vs GPU?

I had a quick read of jitter's tutorial 42 and 43. I am gathering that computationally intensive calculations can be "forced" to be done on a GPU. Right?

I still have a little catching up to do with docs reading but I would appreciate a headstart on the topic of choosing the right hardware.

The whole thing is supposed to be running with WinXP.

I will be using jitter in a project, mainly for analyzing videos.  2 (maybe more) at the same time and dispatching the analysis data to various receivers (thousands of them, it seems).  I do not yet know the specifics (size, framerate etc) but right now I know that we will be choosing hardware specifically for that purpose.

I know, from the docs, that jitter uses the GPU and it can work with multiple CPUs and is both 32 and 64bit friendly.  What would be an optimal setup for analysis (I suppose I will be mainly manipulating matrix data, and it is unclear to me whether this kind of work is done on the CPU or the GPU)? Perhaps there is a chart somewhere indicating which specific objects do their number crunching on a CPU vs GPU?  

I had a quick read of jitter's tutorial 42 and 43.  I am gathering that computationally intensive calculations can be "forced" to be done on a GPU.  Right?

what-hardware-for-jitter

> I had a quick read of jitter's tutorial 42 and 43. I am gathering that computationally intensive calculations can be "forced" to be done on a GPU. Right?

Yes, but you will incur a heavy cost reading it back if your intention

is to send the processed data to anything but the video out on your

> I had a quick read of jitter's tutorial 42 and 43.  I am gathering that computationally intensive calculations can be "forced" to be done on a GPU.  Right?
>

Yes, but you will incur a heavy cost reading it back if your intention
is to send the processed data to anything but the video out on your
graphics card.

Ah, yes, thanks for that, it makes sense. I will not be sending video out therefore I guess I will keep off GPU.

Therefore, WRT the CPU, do I gain performance on a 64bit platform? multicore? Threaded? Anything else?

Ah, yes, thanks for that, it makes sense.  I will not be sending video out therefore I guess I will keep off GPU.

Therefore, WRT the CPU, do I gain performance on a 64bit platform?  multicore?  Threaded?  Anything else?

> Ah, yes, thanks for that, it makes sense. I will not be sending video out therefore I guess I will keep off GPU.

> Therefore, WRT the CPU, do I gain performance on a 64bit platform?

Yes for certain operations although the impact might not be that

great. If you're writing your own externals, there's a fairly

straightforward (as these things go) multithreaded matrix processing

https://cycling74.com/download/JitterAPIReference/

On 4/10/07, Mike  wrote:
>
> Ah, yes, thanks for that, it makes sense.  I will not be sending video out therefore I guess I will keep off GPU.
>
> Therefore, WRT the CPU, do I gain performance on a 64bit platform?

Yes for certain operations although the impact might not be that
great.  If you're writing your own externals, there's a fairly
straightforward (as these things go) multithreaded matrix processing
library in the Jitter API.
https://cycling74.com/download/JitterAPIReference/

What hardware for jitter?