Disable ripple-fast vectorization on ARM

yatisht / usher

Ultrafast Sample Placement on Existing Trees

MIT License

121 stars 40 forks source link

Disable ripple-fast vectorization on ARM #311

Closed yceh closed 1 year ago

yceh commented 1 year ago

This reverts to the old non-vectorized ripple merging routine on ARM so that it may be usable with degraded performance. However, to make the filtration pipeline work on ARM, we may need to recompile and validate MAFFT and faSomeRecord for ARM, so I am giving up porting it to NEON SIMD for now.

yceh commented 1 year ago

Added --black_list_node_file to prevent moving certain nodes.