CuCodeML is a GPU version of CodeML for the branch-site model.
The following instructions take you through a sequence of steps to get the default configuration of CuCodeML up and running.
(a) Prerequisites
(b) Configuration
Edit Makefile with your favorite editor and edit the following options in configuration part to fit your need:
WARNING: DO NOT EDIT ANY OTHER PART OF THE MAKEFILE
Summary: If you want a quick configuration, just modifying the CUDA_INSTALL_PATH and SINGLE_GPU_ID configuration is usually enough.
If you just want to use one GPU card and not cooperate with CPU, edit the configuration just like this:
MULTICARD_ONLY_GPU = no
HYBRID = no
SINGLE_GPU_ID = #GPUID you want to use
MAXGPU = 1
If you want to use one GPU card and cooperate with CPU, edit like this:
MULTICARD_ONLY_GPU = no
HYBRID = yes
MAXGPU = 1
If you want to use multiple GPU cards and not cooperate with CPU, edit like this:
MULTICARD_ONLY_GPU = yes
HYBRID = no
MAXGPU = #GPUs you want to use ( bigger than 1)
If you want to use multiple GPU cards and cooperate with CPU, edit like this:
MULTICARD_ONLY_GPU = no
HYBRID = yes
MAXGPU = #GPUs you want to use( bigger than 1)
(c) Build
make
(d) Installation
(e) Execution
your-CuCodeML-path/CuCodeML
CuCodeML is a free, open-source software under GNU GPL v3 license.