The code is hosted at github: (https://github.com/adrian-bl/libfairydust) and should work with cuda 4.x and ATI/AMDs OpenCL implementation.
How does it work?Simply LD_PRELOAD libfairydust.so (export LD_PRELOAD=/path/to/libfairydust.so) and start 'fairyd.pl'.
The library works by 'hijacking' some cuda/opencl calls:
Do i need it?Unless you are running a cluster: no. Libfairydust only makes sense while used in combination with a batchsystem, such as LSF.
Why is it called libfairydust?Because it works with ELFs and does magic!
comments powered by Disqus