INSTALLATION

Untar tarball
tar -xzf TMscore-gpu.1.0.tgz

Change to directory
cd TMscore-gpu.1.0

Compile
make

This creates an executable tmscore_gpu_ava which calculates the TMscores between all pairs of structures (pdb format) given in a list

USAGE

tmscore_gpu_ava -i <input_file> -o <output_file> -c <optional use cpu - default is use gepu> 

The input file is a text list of the pdb files to be compared - one per line
The output is in a text file in the form of the two pdbfile names followed by the TMscore

A test_directory is included pdb_test which contains sample pdbs generated de novo from the Nutritious Rice for the World project.

Two test shell scripts are included.

cpu_test - runs tmscore_gpu_ava on a small set of 30 pdbs in cpu mode
gpu_test - runs tmscore_gpu_ava on a small set of 30 pdbs and on a larger set of 300 pdbs in gpu mode

To run - type cpu_test or gpu_test in the console.

Both of these should take no more than 30 seconds to run.

The output results from are in the results directory and can  be compared to the expected results in the results_comparisons

INSTALLATION  REQUIREMENTS

OpenMP is used for the timings. If the timers are deleted, it will compile without the OpenMP libaries. 
OpenCL is used for the GPU implementation. The implementation uses AMDAPP2.5 and has been tested on Catalyst 11.12 drivers.
These need to be downloaded and installed from AMD for compilation. 
The CPU code is not dependent on OpenCL.

The software is provided as is - to be used as desired. We only ask that if you find it useful that you cite the appropriate reference

Ling-Hong Hung 03/27/12
