CODEWITHSUNDEEP
cudanvcc
YesThatIsMyName
YesThatIsMyName

Reputation: 1636

CUDA 4.2 nvcc (cicc) error while compiling with sm_21

I have a strange problem! I have a GTX 560 Ti card (compute capability 2.1) where I develop my CUDA apps.

Now after some hundred times of compiling, I got an error. That is, the compilation phase stops, with the cicc process not terminating. But when I change my number of threads per block number, it sometimes may compile. (For 64 threads it crashes for 1024 it works).

Does anyone had such a strange error. The other thing is, that with sm_13 it compiles every time, but with -arch=sm_21 it won't. But as I said it ran hundreds of times with this option, but now it will not work any more. Any suggestions?

Upvotes: 0

Views: 2719

Answers (1)

YesThatIsMyName
YesThatIsMyName

Reputation: 1636

Haha!

Here it runs again!

Commenting everything (global function) -> then compile -> uncomment again -> compiles as nothing happened before!!!

Such a strange behaviour! Maybe some intern compiler flags were set false, or something else.

Upvotes: 1

Related Questions