Is that possible to call functions inside a kernel modules??
When I try I get the error “function undefined” . Probably it’s not possible but i wanna be 100% sure.
Thanks
Yes. These functions need to be defined as: device
notice that these aren’t real functions more like macros since the compiler will simply replace the function call with the function code.
With CUDA 1.1 it is possible to disable function inlining.
How do we do that Andrei? Is it an option to the nvcc?
noinline qualifier. Page 22 of 1.1 Programming Manual.
Function inlining is good for performance. What situations would it be useful to disable it?
Thanks.
The 1.1 Programming Manual is probably only on the developer login site?
I have not been able to find it.
The documentation for 1.1 is included in the beta installer (now available at [url=“CUDA Toolkit 1.1 (December 2007) | NVIDIA Developer”]http://developer.nvidia.com/object/cuda_1_1_beta.html[/url] ).
That will teach me to write from home before checking at work ;)