AGAThA: Fast and Efficient GPU Acceleration of Guided Sequence Alignment for Long Read Mapping

TL;DR

AGAThA addresses the challenge of exact guided sequence alignment for long reads by introducing a GPU-accelerated implementation that preserves the reference algorithm while mitigating memory and workload bottlenecks. The method introduces rolling-window tracking of anti-diagonal maxima, a sliced diagonal tiling, subwarp rejoining, and uneven bucketing to balance intra- and inter-warp workloads, all backed by a simple performance model. Empirically, AGAThA achieves up to 18.8× speedup over CPU Minimap2 and substantial gains over other GPU baselines, with strong ablations showing the impact of each component. The approach maintains exactness, demonstrates cross-hardware robustness (including multi-GPU scalability), and applies to related tools like BWA-MEM, indicating broad practical impact for fast, accurate long-read mapping pipelines.

Abstract

With the advance in genome sequencing technology, the lengths of deoxyribonucleic acid (DNA) sequencing results are rapidly increasing at lower prices than ever. However, the longer lengths come at the cost of a heavy computational burden on aligning them. For example, aligning sequences to a human reference genome can take tens or even hundreds of hours. The current de facto standard approach for alignment is based on the guided dynamic programming method. Although this takes a long time and could potentially benefit from high-throughput graphic processing units (GPUs), the existing GPU-accelerated approaches often compromise the algorithm's structure, due to the GPU-unfriendly nature of the computational pattern. Unfortunately, such compromise in the algorithm is not tolerable in the field, because sequence alignment is a part of complicated bioinformatics analysis pipelines. In such circumstances, we propose AGAThA, an exact and efficient GPU-based acceleration of guided sequence alignment. We diagnose and address the problems of the algorithm being unfriendly to GPUs, which comprises strided/redundant memory accesses and workload imbalances that are difficult to predict. According to the experiments on modern GPUs, AGAThA achieves 18.8$\times$ speedup against the CPU-based baseline, 9.6$\times$ against the best GPU-based baseline, and 3.6$\times$ against GPU-based algorithms with different heuristics.

Figures (16)

• Figure 1: Guided sequence alignment.
• Figure 2: Existing GPU-accelerated sequence alignments.
• Figure 3: A motivational study. (a) represents the execution times of the existing CPU-based algorithm and two naive GPU-based alignments, and (b) represents the distribution of the accumulated workloads and alignment count for alignment tasks.
• Figure 4: Tracking the anti-diagonal maximums with rolling window. Cells on the same anti-diagonal are colored with the same hue.
• Figure 5: Sliced diagonal strategy.