Enable multiple threads?

[waterml]

Currently, yass runs with only single threads, and is slow for large fasta files. Is it possible to speed up yass by running parallelly on more threads? Thanks.

[laurentnoe]

There are several parameters to compile it with threads

./configure --help

  --with-threads          compile with all threads
  --with-thread-fr        compile with two separate threads for Forward and Reverse sequence
  --with-thread-aa        compile with two separate threads for Assemble and Align steps
  --with-thread-qc        compile with threads for the query chunks

however, for multiple queries (multifasta for the first file) there is also a way to increase (inside the util.h file) the number of threads to process queries independently util.h:#define MAX_QUERY_CHUNK_THREADS 4

PS : if you use yass on the web server "https://bioinfo.univ-lille.fr" , THEN when large files are provided by the user, the single threaded version is used : the reason is here to reduce the "memory footprint" (only one index is created at a time) so the program is more likely to terminate without "being terminated" 🤖

[laurentnoe]

... is slow for large fasta files ...

😐

• for full sensitivity with "rather small (weighted)" spaced seeds, yass will (always) be slow ...
• to cope with that, we now propose Minimally overlapping words seeds, or "Mow" 🚜✂️🌱 seeds
  • they are less sensitive but an order of magnitude faster (as minimizers, but avoiding some of their artefacts 👍) : several are already provided in the https://bioinfo.univ-lille.fr/yass/seeds.txt file.
  • they could be selected and used for the "Seed pattern:" in the bottom of https://bioinfo.univ-lille.fr/yass/yass.php but by now are not in the [autoselect] mode, only manual select enables them ...