Compiler ouldn't find double precision intrinsics commands

Hey, I was trying to use the __dadd_rd() function, but the compiler couldn’t find it. I was googling around for hours, trying to figure out what did I miss, what should I #include, but didn’t find any useful information.

The float version of the function (__fadd_rd()) works fine.

Any ideas are welcome.

What compute capability are you compiling for? Are you by any chance compiling for multiple compute capability and the compile only fails for some of them?

I don’t know for sure what do you mean by multiple compute capability. I’m a starter at CUDA, this is a single .cu file, and I have a 9600M GT. Compiling with a single nvcc command in cmd prompt.

In this case the answer is simple: the 9600M GT is of compute capability 1.1 and thus does not support double precision.

You can still compile the code ( but you will not be able to run it on your card) if you pass the flag “-arch sm_13” or higher
to nvcc.

Thank you very much!