NVIDIA-Genomics-Research / GenomeWorks

SDK for GPU accelerated genome assembly and analysis
https://clara-parabricks.github.io/GenomeWorks/
Apache License 2.0
281 stars 76 forks source link

[cudapoa] Traceback Buffer #573

Closed r-mafi closed 3 years ago

r-mafi commented 3 years ago

Added a new banded alignment: static_banded_traceback In this method, data required to perform backtracking in NW is stored in a traceback buffer and score matrix is stored partially. The data type for traceback, TraceT, generally can be more compact compared to score data size, ScoreT. TraceT is int8_t or int16_t, which is determined based on maximum distance of predecessors that are taken into account in backtracking (option -D in the cudapoabin API). This maximum distance also determines the height of score matrix. Using 1B TraceT for long-reads can reduce data storage of score matrices by x4 times.