Compiling and Installing GPU-HMMER for Linux

Prerequisites : You will need to have a C++ compiler (assumes g++), a C compiler, and the NVIDIA nvcc compiler in your path. Also take note of where you have installed the CUDA SDK. GPU-HMMER assumes that the CUDA SDK is in the default location (normally ${HOME}/NVIDIA_CUDA_SDK), but if it's not, you'll need to make a couple of changes, described later. GPU-HMMER was designed to support version 2 of the CUDA SDK.

Installation Steps:

1) Download and untar GPU-HMMER from the source archive.
 

Get GPU-HMMER from our downloads section. tar xvzf GPU-HMMER-0.9.tar.gz

2) Configure and compile GPU-HMMER.

run configure as you normally would : ./configure At this point, if you have any non-standard install options you need to apply those to the Makefile in the src directory. In particular, if you need to change the location of the CUDA SDK, you should change src/Makefile to point CUDAINCDIR to your custom location. Also, you can change the number of active sequences by changing CUDADEFS in src/Makefile. If you're using the 8800 GTX Ultra, you shouldn't need to make any changes. Otherwise, you can raise or lower the thread size and block size as your architecture allows. When you're done, build GPU-HMMER from the toplevel directory by executing: make Make should complete without errors

3) Installing GPU-HMMER.

You are not required to install the GPU-HMMER binaries in the typical system locations. Instead, GPU-HMMER can be left in-place and will run without errors. For users wishing to perform a true installation, please execute: make install