FastCuCodeML

This package is an speed optimization version of CuCodeML (https://github.com/rmingming/CuCodeML)

The source code was written by Ziheng Yang (https://github.com/abacus-gene/paml)

Instructions for compiling in Ubuntu 18.04

you need install CUDA first

step 1:

    your_nvcc_localation/nvcc -arch=sm_75 -DCUDA -DSINGLE_GPU_ID=0 -O3 -c cuda-codeml.cu

this is mine RTX2070 sm arch(sm_75) and you need change based on your GPU

step 2:

    cc -DCUDA -DSSE -O3 -funroll-loops -fomit-frame-pointer -c tools.c
    cc -DCUDA -DSSE -O3 -funroll-loops -fomit-frame-pointer -c codeml.c

step 3:

    cc -DCUDA -DSSE -O3 -funroll-loops -fomit-frame-pointer cuda-codeml.o tools.o codeml.o -(your_cuda_lib64_location)lib64 -(your_cuda_lib_location)lib -lcudart -lstdc++ -lm -o FastCuCodeML

Then run the program with the following command: your-path/FastCuCodeML

We have tested this package in Windows10 using Visual Studio and Ubuntu 18.04 LTS

Speed test comparing to CuCodeML