Open serjl opened 1 year ago
Hi,
I tried your example with and without templates on Cuda 11.4, both compile just fine. You have a full example with calling kernel code you can post (only the __device__
-function will just be optimized away anyhow)? What is the output of rtc.GetLogAsString()
?
Hi, I tried your example with and without templates on Cuda 11.4, both compile just fine. You have a full example with calling kernel code you can post (only the
__device__
-function will just be optimized away anyhow)? What is the output ofrtc.GetLogAsString()
?
Hi, Thanks a lot for the reply. From rtc.GetLogAsString() I get that "template template <typename T> __device__ __forceinline__ unsigned char HasBit(T bitmap, unsigned char pos) { return (bitmap >> pos) & 1; }
inside extern "C"{}
block - my fault. After taking it out of it all works perfectly.
Many thanks again!
Hello Michael, I am using the latest version of the wrapper (Cuda 11.4). I get into a strange failure while compiling a kernel with a template attribute, e.g.
ManagedCuda.NVRTC.NVRTCException HResult=0x80131500 Message=ErrorCompilation: Compilation error. Source=NVRTC
Without it , it works fine, e.g.
I use the following compilation code:
Do you have any idea what is wrong here?
Your help as well as you amazing wrapper are highly appreciated and valuable.