Interpreting cudaGetErrorString

[Ian Goodfellow]

Is there any documentation of what the strings returned from
cudaGetErrorString mean?
I'm getting "invalid argument" returned for a kernel call.
The only documentation I can find does not explain the individual errors at all:
http://developer.download.nvidia.com/compute/cuda/4_2/rel/toolkit/docs/online/group__CUDART__ERROR_g38e5684c158c22144ad3c269ad61bc78.html#g38e5684c158c22144ad3c269ad61bc78
and I have no idea how an invalid argument to a kernel could be a
runtime error. Shouldn't
code with invalid arguments fail to compile?

[Ian Goodfellow]

OK, this turned out not to be an error that was raised due to my kernel call.
It was an error caused by fprint_CudaNdarray that was still hanging around.
If anyone does know of a good source of documentation for cuda errors,
please do let me know.
I found this and it is better but still doesn't contain exactly the
strin that was
raised:
http://developer.download.nvidia.com/compute/cuda/4_2/rel/toolkit/docs/online/group__CUDART__TYPES_g3f51e3575c2178246db0a94a430e0038.html#g3f51e3575c2178246db0a94a430e0038

[Frédéric Bastien]

I do not know a good source. If you find one, sent it here, it will
help many people I think.

Fred

On Thu, Sep 27, 2012 at 7:17 PM, Ian Goodfellow

[wang wei]

The message "invalid argument" shown by cudaGetErrorString() is corresponding to the Error 11 (cudaErrorInvalidValue) in cudaError_t

you can refer to this page https://sites.google.com/site/diesel56peko/Home/puroguramingu/cuda/rifarensu/cudageterrorstring