Cipres Science Gateway | Code Running Parameters

Return to Login Page
How CSG Codes are Run:
This page describes how each of the CSG codes is run on TG resources. These run parameters were creating based on Benchmarks for each of the codes. How to read this page: For each tool, the table lists parameters that affect how jobs are run. For each condition, the table lists the number of nodes and CPUs used for the job. If the job is run in a Shared queue, the run is charged only for the number of requested CPUs. If the job is run in the Normal queue, the job receives the resources of each requested Node. As a result, some job runs (for example, MrBayes nruns*nchains=6) actually use only 30 CPUs, but they will be charged for 32, because the entire Node is reserved for that job run. While this appears slightly inefficient, by reserving an entire node we prevent competition for resources that might slow the job down.
Mr Bayes
nruns*nchains =	Nodes	CPUs	Queue	Charge
2 - 4	1	32	Normal	32*time
6	1	30	Normal	32*time
8	1	32	Normal	32*time
10	1	30	Normal	32*time
12	2	60	Normal	64*time
14	2	56	Normal	64*time
16	2	64	Normal	64*time

RAxML
Runs with no bootstrapping; any options other than -f a or -f d
Code version	Nodes	CPUs	Queue	Charge
pthreads only	1	8	Shared	8 * time
Runs with < 50 bootstraps
Code version	Nodes	CPUs	Queue	Charge
hybrid mpi/pthreads	1	30	Normal	32 * time
Runs with 50 or more bootstraps, or automatic termination of bootstraps
Code version	Nodes	CPUs	Queue	Charge
hybrid mpi/pthreads	2	60	Normal	64 * time

GARLI
Number of Bootstraps	Nodes	CPUs	Queue	Charge
0	1	1	Shared	1 * time
2-31	1	#bootstraps	Shared	#bootstraps * time
50 +	1	32	Normal	32 * time

BEAST
All runs	Nodes	CPUs	Queue	Charge
	1	8	Shared	8 * time
MAFFT
All runs	Nodes	CPUs	Queue	Charge
	1	32	Normal	32*time


Return to Login Page

If there is a tool or a feature you need, you can add it yourself or let us know.