CUDA compiler bug ?


I have .cu file that I used to compile for 20 CUDA architecture (I’m using CUDA 5.0). I have added two more destination architectures (30 and 35) and tried to compile this file, compilation for 35 architecture failed:

1>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include\cuda_runtime.h(1106): error: expected a “)”
1>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include\cuda_runtime.h(1108): error: expected a “)”
1>2 errors detected in the compilation of “MulticoreHelpers.cpp2.i”.
1>Project : error PRJ0019: A tool returned an error code from “Compiling with CUDA Build Rule…”

An error is reported in one of CUDA headers, I have no clue why this happens. The very same file compiled without problems for 20 and 30 architectures.

Is it a bug ? How to overcome this issue ?

Thanks in advance.

Can you compile SDK samples?

Yes, I have just recompiled them from scratch - everything is fine.