Computers

General Overview
MESA and NuGrid MPPNP need strong computer power, while NuGrid PPN needs pretty small power.

Available Computers (in the order of computation power)
- GC-MASTER (additional) (Ubuntu 16.04; 64bit, cpu: 2.2 GHz * 48; memory 132 GB); 2 threads/core, 12 cores/socket, 2 socket (2*12*2=48) --> Installed OpenMPI 2019/10/17 (with K. Hagel). I installed hdf5 by myself somehow (see details on MPPNP page). And succeeded installing mppnp and run on 2019/10/18. This is the most powerful computer available to me.

- WENDI (Ubuntu 18.04; 64 bit, cpu: 2.4 GHz * 16; memory 120 GB) --> (astrohub.uvic.ca) If necessary for very heavy calculations, possible to use this by asking Fork (I got access from him now). https://astrohub.uvic.ca. Now I'm using this mostly for MESA (since HD size is only 300 GB and not good to run MPPNP by uploading hdf5 input data). Note WENDI is likely to be shared with other users. When someone uses, the running speed slows down pretty much. WENDI IP address is 172.19.0.16 (but I don't know how to ssh. ssh user@ method didn't work).

- WENDI-b (Ubuntu 18.04; 64 bit, cpu: 2.4 GHz * 16; memory 120 GB) --> WENDI backup server (astrohub.phys.uvic.ca).

- GCA server (Ubuntu 16.04.6; 64 bit, cpu: 3.6 GHz * 8; memory 16 GB); 2 threads/core, 4 cores/socket, 1 socket (2*4*1=8) --> MESA and MPPNP (and PPN).

- Bash4.1 (cycdhcp190 (used to 39 until 2019 Oct.)) (Scientific Linux 6.10 (carbon); since 11/13/2019 CentOS 8.0, cpu: 3.4 GHz * 8; memory 8 GB) --> MPPNP doesn' work (get the same HDF5 error messsage when you run) MPPNP works now (see MPPNP page for details). For MESA, the memory may be insufficient (8M preMS to AGB test suite is interrupted). Note this doesn't work with MESA (maybe too few memories).

- Sony VAIO laptop (Ubuntu 18.04.3; 64 bit; since 11/20/2019 CentOS 8.0 (installed with Jerome so matches with Bash4.1), cpu: 2 GHz * 4; memory 8 GB); 2 threads/core, 2 cores/socket, 1 socket (2*2*1=4) --> Basically NuGrid (sometimes MESA for small load calculation). MPPNP couldn't be successfully installed. It looks like hdf5 is not installed well (maybe different versions 1.8. and 1.10 are mixed...??). So this is only used for PPN. '''VERY IMPORTANT!! Now MPPNP working. Problem was gcc/gfortran version. 7.2.0 (included in MESASDK 2018 Aug. version) works well, while ver 8 or 9 give HDF5 error message (see details in MPPNP page)'''

- Macbook Air (Sierra; 64 bit, cpu: 1.4 GHz (intel) * 1; memory 4 GB) --> NuGrid only (just for quick test). Naturally only PPN.

To check cpu threads and details, "lscpu | grep -E '^Thread|^Core|^Socket|^CPU\("'

Computation History
- MPPNP run_test (run_template); M25 Z0.02 from cycle 1 to last 23,000. @ Sony Linux (must be GCA because Sony Linux can't run MPPNP), took 24 hrs. 9/21/2019.

- MPPNP run; M5 Z0.02 from cycle 1 to last 125,000. @ GCA, took 4 days. 10/10-15/2019. (many times calculations down, so needed to restart).

- MPPNP run; same condition above, @ WENI, took 2.5 days. 10/17-19/2019. So, WENDI is about 1.6 times faster than GCA (NOTE! In fact, it took 5 days; 10/17-22 because someone used somewhere. If no one but me uses the computer, it is fast as I wrote above, but in reality it is likely someone also uses and slow down my calculations.

- MPPNP run; same condition above, @ GC-MASTER, took 1 day. 10/18-19/2019. So, GC-MASTER is about 4 (or >3) times faster than GCA.

Summary of Computation Capability
- If say GC-Master = 100, Wendi = 40, GCA = 25, Cycdhcp = 20, Sony Linux = 10. Cycdhcp + Sony Linux (Clustering) = 30.