Sync production v15.3.1 + Sharrow

[jpn--]

This pull request builds on #37, adding compatibility with sharrow.

Several scripts are included to run various legacy and sharrow tests: https://github.com/driftlesslabs/sandag-abm3-example/tree/sync_production_v15.3.1/scripts

[jpn--]

@bhargavasana It would be great if someone at SANDAG could review this PR, including (especially) running some performance test on the "large" data on SANDAG servers, to confirm not only that it runs but also that turning on sharrow makes it run significantly faster, on the platform that matters. I've tested locally and this code works for me (with ~50% time savings over the legacy code) but no one really cares if it works on a macOS laptop, the real proof is when it works on a large Windows server. Thanks!

[bhargavasana]

@bhargavasana It would be great if someone at SANDAG could review this PR, including (especially) running some performance test on the "large" data on SANDAG servers, to confirm not only that it runs but also that turning on sharrow makes it run significantly faster, on the platform that matters. I've tested locally and this code works for me (with ~50% time savings over the legacy code) but no one really cares if it works on a macOS laptop, the real proof is when it works on a large Windows server. Thanks!

Sure, I will get this tested on a SANDAG server this week.

[jpn--]

Alexander Berryhill-Williams [email protected] via email ...

I got this crash when running run-large-sharrow.py. The OpenBLAS warning is the only suspicious thing that appears in the
console, and no errors appear in either the console or log file. I’ve attached the log file here, and here are the last few lines in the console:

[37:25.59] INFO: Running cdap_simulate with 256854 persons
[37:27.29] INFO: using existing flow code XW74XDW3RBW6QKTJ2Z4XK4ULKNOKOY5X
[37:27.81] INFO: using existing flow code CYPLPVKVDFO3KALIFDSNDPWDW3LLKLOZ
[37:27.98] INFO: using existing flow code C3OVYLQFAG3K2ED2YQTYNOLHEBWSHZLM
[37:28.48] INFO: using existing flow code VABUMWUS3JBFKR3YA5EP2R2X5ZG3PFMA
[37:28.80] INFO: using existing flow code 62MTG4HW4EFQOWWIMNI3GF3ZZBAWVJHB
[37:29.38] INFO: using existing flow code GLPVAYOYY7S6TQDNAK7N6XPS5B35WKKM
OpenBLAS warning: precompiled NUM_THREADS exceeded, adding auxiliary array for thread metadata.
To avoid this warning, please rebuild your copy of OpenBLAS with a larger NUM_THREADS setting
or set the environment variable OPENBLAS_NUM_THREADS to 24 or lower
[37:30.10] INFO: using existing flow code HNHYJPKU5T4CZDEOC3UWI6DSVZNFVOX2
[37:31.01] INFO: using existing flow code HFRWLDH35H3JXI7E54OYVGEQESKMVBUN
[37:32.45] INFO: using existing flow code FPR6VMZKH23EAT6GVK7R27243LEDGKRW

I attempted running run-large-sharrow.py twice and encountered this crash both times. I do not have the log from the first attempt, but the OpenBLAS warning also appeared just before that crash, so I believe they occurred in the same place. I did not encounter any issues with run-small-sharrow.py.