ONLY CUDA :-"
TBH it can be a bit tricky to install CUDA and run your CUDA code properly. Aside from using tutorials for installing and etc, here are some NOTES worth mentioning in my opinion.
- Try to run the deviceQuery program on cuda-samples. The results are important specs of your GPU and some of them are crucial to know.
- Your GPU CUDA compatibility version (can be known with previous program or from Wiki, Nvidia or etc) matters a lot! It determines the features and technical specifications of your GPU (compute-capabilities).
When compiling your CUDA program using
nvcc, sometimes you need to specify the compatiblity version (with -arch or -gencode options) to prevent some unseen errors without logs!
to be done!
# compiling
nvcc <flags-required-for-nvcc> -o gpu_exe LocalGaplessAlignmentGPU.cu Utils.cpp ScoreMatrix.cpp -D{REDUCE_ALIGNMENT_RESULT,REDUCE_ON_COLUMNS}
# running (refer to the code for the options!)
./gpu_exe <targets-file> <is-fasta> <alignment-method> <query-file>