How do you use constant memory for very large kernel parameter lists?

Accelerated Computing CUDA CUDA Programming and Performance
1

There is apparently a 256 byte limit on kernel parameter lists, and it is suggested on [url=“http://forums.nvidia.com/index.php?showtopic=84440”]http://forums.nvidia.com/index.php?showtopic=84440[/url] that constant memory is used, but no example is given.

I’ve tried passing some parameters in a pointer to structure but that didn’t work, so how can constant memory be used? Is there an example in the guide/SDK projects etc?

2

The 'particles" sample in the SDK demonstrates this.

3

The 'particles" sample in the SDK demonstrates this.

4

I’ve realised my error and have amended to use that proposed in http://forums.nvidia.com/index.php?showtopic=106708 which seems to work.

But I now get a compiler advisory that “cannot tell what pointer points to, assuming global memory space”. Is this something to worry about?

5

I’ve realised my error and have amended to use that proposed in http://forums.nvidia.com/index.php?showtopic=106708 which seems to work.

But I now get a compiler advisory that “cannot tell what pointer points to, assuming global memory space”. Is this something to worry about?