hardware wish

[Janosik]

i wish the vray would run at gpu stream processor and will be a lot faster than on cpu

[vlado]

We are, of course, thinking in that direction. However the GPU technology is quite unstable yet, with completely new hardware coming out each year, which makes any long-term plas quite difficult to make. Hopefully the GPU technology will settle on a standard sometime soon, so that development can be more predictable.

Best regards,
Vlado

[Neilg]

So if you coded vray to work on an 8800, would it not work for many other new cards after that?

[vlado]

So if you coded vray to work on an 8800, would it not work for many other new cards after that?

I don't have the slightest idea... which is part of the problem What if someone comes out with a better hardware solution and the 8800 architecture becomes obsolete - we'd have to start all over again.

Best regards,
Vlado

[Neilg]

Excellent technology, potentially a revolution, but thanks to one little oversight in creating a standard it becomes completley and utterly useless.


More curiosity - would you have to re-write the way that vray does its calculations or is it just updating/optimising all the code to use the right parts of the hardware?

[studioDIM]

It's a MESS
I was reading some page (watch it, it's a pdf...) on mathematica accelerators: dedicated, parallel processing units that come on a pci-E card, from http://www.clearspeed.com .
The principle is VERY close to that touted by NVidia (the article on CUDA had a link to the "similar" clearspeed boards).

FIGURE 6: STANDARD C

#include <stdio.h>
#include <math.h>
#define PI 3.1415926535897932384f
#define SAMPLES 96
int main&#40;&#41; &#123;
double sine, angle;
int i;
for &#40;i = 0; i < SAMPLES; i++&#41; &#123;

// convert to an angle in
range 0 to Pi
angle = i * PI / SAMPLES;

// calculate sine of angle
sine = sin&#40;angle&#41;;
&#125;
&#125;

FIGURE 7: PARALLEL C

#include <lib_ext.h>
#define PI 3.1415926535897932384f
#define SAMPLES 96

int main&#40;&#41; &#123;
poly double sine, angle;
poly int i;

// get PE number&#58; 0...n-1
i = get_penum&#40;&#41;;

// convert to an angle in range 0 to Pi
angle = i * PI / SAMPLES;

// calculate all sine values simultaneously
sine = sinp&#40;angle&#41;;
&#125;

To make it simpler: forget the "For/Do" cycle entirely...
I don't know, lol.

Lele

[Neilg]

Thats not too difficult then, vlado just needs to open his code in word and do a find/replace to change 'for' to 'poly'



(I'm guessing its a pretty big job then...)