Sign in to follow this  

random number generation based on time in HLSL

This topic is 2570 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

hi,

i will be passing timeGetTime values from CPU to a variable
in my shader and based on that possible seed value ,

I want to implement a random number generation function in HLSL , but actually currently have no idea about algorithms

As starting point what would you advise ?

And also what about producing something with same quality with rand() of C/C++ ,
is it a hard job or can it be done with a reasonable effort ?

many thanks in advance...

Share this post


Link to post
Share on other sites
Why would you want to do that in HLSL? That's not really HLSL's job, you should pass in those kind of parameters to the shader if you need them.

Share this post


Link to post
Share on other sites
Because I want to implement main `Update` method of my particle system
in GPU with vertex shaders ( since I am using D3D9 and i cant use geometry shaders ,i still have code in C++ side except `Update` part which is the meat of the job actually)

so that is why I want to have random number generation function in my shader

in current implementation , in CPU side update method of my particle system , I am already passing random numbers to the shader variables , but I want to keep CPU side as low as possible

Share this post


Link to post
Share on other sites
You might be able to leverage noise. Though that's not really suited for that purpose.

Any actually good random number generator is going to burn through too much instruction count to be of any good.

You can also write a bunch of random numbers to a texture and look that up in the shader. That's probably the best answer if you really want random numbers in your shader.

Share this post


Link to post
Share on other sites
I guess this is what you're looking for =)



#define RANDOM_IA 16807
#define RANDOM_IM 2147483647
#define RANDOM_AM (1.0f/float(RANDOM_IM))
#define RANDOM_IQ 127773
#define RANDOM_IR 2836
#define RANDOM_MASK 123459876


int random_x;


float random ()
{
int k;
float ans;

random_x ^= RANDOM_MASK;
k = random_x / RANDOM_IQ;
random_x = RANDOM_IA * (random_x - k * RANDOM_IQ ) - RANDOM_IR * k;

if ( random_x < 0 )
random_x += RANDOM_IM;

ans = RANDOM_AM * random_x;
random_x ^= RANDOM_MASK;

return ans;
}


float random ( float low, float high )
{
float v = random();
return low * ( 1.0f - v ) + high * v;
}


float2 random ( float2 low, float2 high )
{
float2 v = float2( random(), random() );
return low * ( 1.0f - v ) + high * v;
}


float3 random ( float3 low, float3 high )
{
float3 v = float3( random(), random(), random() );
return low * ( 1.0f - v ) + high * v;
}


void seed ( int value )
{
random_x = value;
random();
}

Share this post


Link to post
Share on other sites
glSmurf , thanks very much

yeah it seems as what i am looking for :)

just a quick question , by using shader model 1.1 2.0 and 4.0 , i could not compile it because of '^' operator , how can I resolve this issue ?

thanks in advance...

Share this post


Link to post
Share on other sites
From msdn:
"Bitwise operators are defined to operate only on int and uint data types. Attempting to use bitwise operators on float, or struct data types will result in an error."

Maybe you changed 'int random_x' to 'float random_x' ?

Share this post


Link to post
Share on other sites
@glSmurf: Can you please tell us exactly how you got your code to work (shader model, compile options) and in what context (pixel/vertex shader, software shader even ?) because I have some doubts: One thing I'm concerned is the bitwise operations (only available on SM >= 4), the other that your PRNG uses a global that gets changed. You can't do that with shaders IMHO.

Share this post


Link to post
Share on other sites
First of all, it's not my own code. I got it from the NVidia DX10 SDK. I've used it in a vertex shader and it works just fine (shader model 4.0).
For more information check out the SDK.
File: ... / NVIDIA Direct3D SDK 10 / Source / Lightning / Lightning.fx

Share this post


Link to post
Share on other sites
Thanks, that clarifies the problem: akhin uses DX9 so only SM 3 (or less) is available, unfortunately.

@akhin: Bear in mind that if you now go for a texture lookup in the vertex shader, this is not possible with SM 2.

Share this post


Link to post
Share on other sites
thanks for clarification ,

I need to ask one last question since I am new to shader programming ,

I am not crystal clear sure how mechanics behind the scenes are so ,

As it has been said , it is Shader Model 3 with DX9 which is the latest SM I can use, what about using DX9 and NvidiaCG together instead of HLSL ?

So it wouldnt change anything would it ?

Actually I am after to learn how the graphics API itself is a bottleneck for the latest shader model we can use , if the shaders are being compiled externally and targeted for GPU

Share this post


Link to post
Share on other sites
Haven't explicitly checked myself, but I really doubt it (these profiles are from an old version of Cg, couldn't find a newer table). You still have to (want to?) load your shader in a DX 9 context, so you're limited to SM 3. By the way: Cg can compile HLSL code.

To get more features from your card you have to use a different API (DX10, OpenGL, or even OpenCL, CUDA).

Share this post


Link to post
Share on other sites

This topic is 2570 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this