TroubleShooting.tex

\chapter{Troubleshooting}
\label{chap:troubleshooting}


This chapter describe the  meaning and solution of common 
\begin{enumerate}
\item \verb!ConfigureCall FAILED:9!: it can be
  \begin{enumerate}
  \item The deice has not been initialized yet (call cudaSetDevice())
  \item The chevron syntax is not properly set, check the value of the
    block size and grid size
  \end{enumerate}

\item
  \verb!0: DEV_MKDESC: allocate FAILED:4(unspecified launch failure)!:
  there is an unknown error in allocation, e.g. run out of memory or allocation
  size is too large or address is out of bound. The best way to know what
  exactly the problem is, we use \verb!cuda-memcheck!. 
  
  \item If we look at the driver type \verb!driver_types.h!, it states that
  ``this indicates that an uncorrectable ECC error was detected during
  execution''. It means cudaErrorECCUncorrectable denotes a double-bit error,
  and most likely it indicates a bad GPU board. 
  \begin{verbatim}
  copyout Memcpy FAILED: 39<uncorrectable>
  \end{verbatim}
  
\end{enumerate}

\section{{\bf -lGL} missing}
\label{sec:lGL}

When you compile the SDK and get the error
\begin{verbatim}
/usr/bin/ld:cannot find -lGL
\end{verbatim}
then you need to make sure /usr/lib/libGL.so and /usr/lib/libGL.so.1 both link
to the same and correct files, e.g. 
\begin{verbatim}
root@cea2:C\ $ls -l /usr/lib/libGL.so
lrwxrwxrwx 1 root root 13 2011-08-10 13:15 /usr/lib/libGL.so -> mesa/libGL.so
root@cea2:C\ $ls -l /usr/lib/libGL.so.1
lrwxrwxrwx 1 root root 15 2011-08-10 13:11 /usr/lib/libGL.so.1 -> libGL.so.280.13
\end{verbatim}
then you can fix
\begin{verbatim}
rm /usr/lib/libGL.so
cd /usr/lib
ln -s libGL.so.280.13 libGL.so
\end{verbatim}

If you get the error
\begin{verbatim}
/usr/bin/ld: cannot find -lXmu
\end{verbatim}
then you install package \verb!libxmu-dev!. 


%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "gpucomputing"
%%% End: