mm2-gb

mm2-gb accelerates the chaining step of Minimap2 on GPU without compromising mapping accuracy for long-read DNA mapping.

Key Features:

• GPU-accelerated chaining: Leverages GPU parallelism to significantly improve the chaining step, a performance bottleneck in the standard Minimap2 workflow.
• Finer-grained parallelism: Exploits intra- and inter-read parallelism by breaking down high latency large workloads into smaller independent segments that can be run in parallel.
• Workload balancing strategies: Implements split-kernels and prioritized scheduling of segments based on sorted size for better workload distribution.
• Maintained accuracy: Achieves performance improvements without losing mapping accuracy, unlike prior acceleration approaches.

Scientific Applications:

• Long nanopore read mapping: Achieves 2.57-5.33x performance improvement on long nanopore reads (10kb-100kb) compared to SIMD accelerated mm2-fast.
• Super long read mapping: Delivers up to 1.87x performance gain on super long reads (100kb-300kb) compared to SIMD accelerated mm2-fast.
• Genetic diagnostics: Supports long-read DNA sequencing applications increasingly used for genetic diagnostics.

Methodology:

mm2-gb accelerates the chaining step of Minimap2 on GPU by exploiting intra- and inter-read parallelism, breaking down high latency large workloads into smaller independent segments, and using split-kernels with prioritized scheduling of segments based on sorted size for workload balancing.