Hi all,
I am trying to compile NEI 2020 using SMOKE in hopes I can modulate the emission inventory for a project. Every sector runs smoothly, except for the onroad_RPD sector.
The main issue is I am running SMOKE on NCAR’s Derecho server, which has an HPC walltime of 12 hours. Running the RPD sector on the command line doesn’t allocate enough memory, and thus, I have to submit it as an HPC job. The 12 hour wall time is a major hinderance, a single day barely finishes within that 12 hours. I tried the 7day mcip files and changing the DAYS_PER_RUN to a single day as well and both take a similar amount of time. I also tried setting OMP_NUM_THREADS in the submit script as well, with little improvement.
Any help is appreciated. Thank you in advance!
Runtime is, more or less, directly proportional to the number of emission factors that need to be read in a given run. So you could consider splitting up the onroad sector into halves or thirds (possibly thirds are better), and then run each half or third as a separate sector, e.g. onroad_east, onroad_central, and onroad_west. To implement this, you just need to split up the activity data by state, and edit the activity data filenames and sector names in the run scripts.
We aren’t sure if your configuration is for the separate gas and diesel outputs, but if so, we recommend turning that off (set SMK_SUB_SECTOR_OUTPUT_YN to “N”) and that should help with runtime and memory also.
If you split the onroad sector into thirds, then you -might- be able to use the DAYS_PER_RUN=7 approach and complete a week in under 12 hours.
Thank you for the advice! I took the VMT data and split it into geographical regions like you suggested (east, central, west) and changed the relevant paths in the run script, but I get an error when SPCMAT is ran, saying:
WARNING in subroutine CHKFIL3 >>>
Inconsistent file attribute NROWS for file MSMAT_S
Value from file: 699445
Value from caller: 145296
Could not open file set “MSMAT_S”.
*** ERROR ABORT in subroutine OPENSMAT
Could not open file set "MSMAT_S".
Since MSMAT_S is an intermediate file in the routine, I am not too sure how to proceed. Any thoughts? Thank you again!
@eyth.alison Hi again. I probably should have tagged you in my original reply. I am curious if you have any ideas what could be wrong with the way I subset the data for RPD from the above reply. Thank you!
We recommend that you try removing all files in the intermed/[sector]/RPD directory and try running again.
@eyth.alison That did it! It worked rather well too, I was able to generate a week for 1/3 of the states in less than 2 hours. Thank you again for the help.