Damascene, version 0.1.5

This software is a reexamination of the gPb image contour detector,
using Nvidia Graphics Processors.  The software builds under Linux,
with CUDA version 2.3, with gcc 4.3.
Damascene depends on an external LAPACK library for part of the
generalized eigensolver used in Normalized Cuts.  We have had the most
success using AMD's ACML LAPACK, which is freely available from
http://www.amd.com/acml
Please modify the common.mk file so that ACMLDIR points to the root
directory where you installed ACML, or modify the common.mk file to
link against another lapack.

The defaults in the common.mk file assume you are compiling for an
Nvidia processor with compute capability >= 1.2 (ie, GTX280, etc.) To
compile for processors with compute capability 1.1 (ie, 9800GTX,
etc.), please change the SMVERSIONFLAGS variable in common.mk to
-arch sm_11.

You will also need to modify the CUDA_SDK_PATH and CUDA_INSTALL_PATH
variables to point to your CUDA installation.

Please direct any questions or comments to Bryan Catanzaro,
[email protected]

--------
Changelog

0.1.4: Fixed bug in k-means, pointed out by Andreas Mueller ([email protected])
0.1.3: Updated to enable compilation under CUDA 2.3