RabbitSAlign is a GPU-accelerated short-read aligner based on strobealign. It doubles the processing speed on real biological data by utilizing GPU to accelerate the extending phase and optimizing inefficient operations in the seeding process.
- gcc 9.4.0 or newer
- nvcc 12.0 or newer
git clone https://github.com/RabbitBio/RabbitSAlign
cd RabbitSAlign
bash build.sh <path-to-cuda-installation-directory> <GPU-compute-capability>
(eg: bash build.sh /usr/local/cuda 86)
The resulting binary is in build/rabbitsalign.
Detailed usage can refer to the strobealign. RabbitSAlign can share the index file (.sti) with strobealign.
Note: If there is an error saying GPU out of memory, you can adjust the STREAM_BATCH_SIZE variable in src/gasal2_ssw.h, which is proportional to the size of the GPU memory.