Closed lagphase closed 10 months ago
Dear @lagphase,
I'm going to assume you mean number of processors. If not, feel free to correct me. Are you using standard HyPhy or HYPHYMPI
?
For the former, try
hyphy CPU=1 gard
with 1
being the number of CPUs you'd like to use.
For the latter, try
mpirun -np 8 HYPHYMPI
.
Best, Steven
Hi Steven,
Thanks. Yes I meant CPUs and I'm using the standard HyPhy.
The code seems to work slightly faster. Any estimation on how long it would take for 250 sequences with 100 CPUs?
Dear @lagphase,
It really depends on the analysis. What are you running? HyPhy will not be able to load 100 CPUs (HYPHYMPI can for FEL, MEME, and other such analyses), unless you have a large and (more importantly) long alignment. In, fact, HyPhy will run a mini-benchmark at the beginning on time-intensive tasks to guess-timate how many CPUs is worth using.
You can add ENV="VERBOSITY_LEVEL=1;
to the command line, and for most analyses you will see a line that looks like this, which will also report the effective CPU load.
Current Max: -17515.127 (0 % done) LF Evals/Sec: 1571 CPU Load: 3.604
Best, Sergei
Hi Sergei,
I'm using hyphy gard to detect recombination sequences in the alignment. Yes my alignment is long and ~100 MB. Is there a way to speed up the process?
Thanks.
Dear @lagphase,
GARD will make better use of an MPI environment. That said, it is meant for screening "gene-sized" aligments (at most 10-20kb). While GARD will attempt to run for a 100 megabase alignment, it will be very slow. How many sequences do you have? Because GARD only considers variable sites, the effective search space could be manageable if you have a few (5-20) sequences.
Best, Sergei
Stale issue message
Hello I don't see any flag for the number of threads used in hyphy gard --help
I'm running 250 genomes and the dafault is taking forever.