Convert with saturation modifier? cvt.sat.u8.u32 ...

So…is there any way to explicitly force the compiler to use the saturation modifier for the cvt instruction (see page 121 of the PTX ISA 2.1 document)? Right now I’m kind of stuck doing this…

int var = data_input;

var = min (var, 255);

char var_out = var;

Any idea? Or is this one of those features (like predicated execution) that the tools do not support yet?

I guess you have to use inline assembly for that.

That’s what I was thinking but until nVidia officially supports inline assembly we are not going to do it.