At job startup the rank 0 process calculates the effective
search space for each query. The process is somewhat
time consuming and should be profiled and optimized. A
likely suspect is the time spent indexing the query
sequence (unnecessarily).
We implemented a patch to skip indexing the query sequence
but the startup time is still high. Profiling with oprofile
indicates that bioseq_dust() is the offending function. My
understanding is that we need queries to be dusted for correct
calculation of effective search space. So this process can't be
skipped. It could be parallelized, but that's not a high priority
target at the moment. A workaround would be to shut off
dusting with the -F F command line option.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Logged In: YES
user_id=756995
We implemented a patch to skip indexing the query sequence
but the startup time is still high. Profiling with oprofile
indicates that bioseq_dust() is the offending function. My
understanding is that we need queries to be dusted for correct
calculation of effective search space. So this process can't be
skipped. It could be parallelized, but that's not a high priority
target at the moment. A workaround would be to shut off
dusting with the -F F command line option.