Build procedure¶

Linux / Mac / BSD¶

CUDA setup notes¶

The CUDA toolkit, by default, does not add its compiler or libraries to your search paths. Adding the following lines to ~/.bashrc should correct this:

export C_INCLUDE_PATH=$C_INCLUDE_PATH:/usr/local/cuda/include
export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/local/cuda/include
export PATH=$PATH:$HOME/cuda_sdk/bin:/usr/local/cuda/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib

Build procedure¶

Note that cmake only needs to be run once unless you are cross compiling - make will take care of recompiling changed source files.

cmake .
make

Cross-compiling¶

To cross-compile, set the TARGET environment variable to X86, AMD64, or PS3 before running cmake.

Example: to compile the 32 bit version on 64 bits, run

TARGET=X86 cmake .

NOTE: Before the September 3rd, 2009 release, the environment variable was name TARGET_ARCH instead of TARGET.

If compiling for multiple architectures in the same source tree (i.e. x86 and amd64), you should make clean and delete CMakeCache.txt before running cmake to reconfigure for another architecture. This will ensure that all build scripts are correctly regenerated for the new platform.

Windows¶

The Windows build procedure is currently being revised; this section will be updated soon.