Tools
Namelist Support table
Variable Names |
Input Options |
Description |
---|---|---|
&domains |
dimensions, nesting, parameters |
|
feedback |
0 |
No feedback from nested domain to the parent domain. Also known as one-way nesting. |
1 |
Feedback from nested domain to the parent domain. Also known as two-way nesting. (1 = feedback). |
|
smooth_option |
smoothing option for parent domain; used only with feedback=1 |
|
0 |
Off |
|
1 |
1-2-1 smoothing option for parent domain; used only with feedback=1 |
|
2 |
(default) smoothing-desmoothing option for parent domain; used only with feedback=1 |
|
interp_method_type |
2 |
(default) SINT Horizontal interpolation option |
&physics |
||
Microphysics |
(mp_physics) |
|
1 |
Kessler scheme |
|
6 |
WSM6 scheme |
|
8 |
Thompson scheme |
|
28 |
Thompson aerosol-aware scheme |
|
Radiation |
(ra_lw_physics and ra_sw_physics) |
|
4 |
RRTMG LW and RRTMG SW schemes |
|
PBL |
(bl_pbl_physics) boundary layer option |
|
1 |
YSU scheme; use sf_sfclay_physics = 1 |
|
2 |
Mellor-Yamada-Janjic TKE Scheme; use sf_sfclay_physics = 2 |
|
5 |
MYNN 2.5 level TKE scheme; use sf_sfclay_physics = 1, 2, or 5 |
|
6 |
MYNN 3rd level TKE scheme; use sf_sfclay_physics = 5 |
|
Land Surface |
(sf_surface_physics) land-surface option (set this before running real.exe; also make sure num_soil_layers is set correctly). The same value should be used for all domains. |
|
1 |
Thermal Diffusion scheme |
|
2 |
Unified Noah land-surface model; Includes Noah LSM mosaic sub-option (sf_surface_mosaic = 1, num_soil_layers = 4) |
|
3 |
RUC land-surface model (num_soil_layers = 6) |
|
Surface Layer |
(sf_sfclay_physics) surface layer option. The same value should be used for all domains |
|
1 |
Revised MM5 Monin-Obukhov scheme (Jimenez, renamed in v3.6) |
|
2 |
Monin-Obukhov (Janjic Eta) scheme |
|
5 |
MYNN |
|
Cumulus |
(cu_physics) Cumulus parameterization option |
|
0 |
Off (On when dx/dy >10km; Off when dx/dy <4km) |
|
1 |
Kain-Fritsch (new Eta) scheme |
|
2 |
Betts-Miller-Janjic scheme |
|
11 |
Multi-scale Kain-Fritsch scheme (when 4km < dx/dy < 10km) |
|
Miscellaneous |
||
Gravity wave drag parameterization |
1 |
gravity wave drag option; use when grid size > 10 km (default is 0 = off) |
GRIMS Shallow Cumulus option |
3 |
GRIMS scheme; independent shallow cumulus option (not tied to deep convection); ; Subject to further performance and numerical validation tests |
1D mixed layer ocean model |
1 |
Activate a simple ocean mixed layer (oml) model; Only works with sf_surface_physics = 1; Subject to further performance and numerical validation tests |
Time-varying sea-surface temperature |
1 |
real.exe will create wrflowinp file(s) at the same time interval as the available input data. These files contain SST, XICE, ALBEDO, and VEGFRA. Option requires additional parameters to be set in time control section |
Radar reflectivity calculation |
1 |
allows radar reflectivity to be computed using mp-scheme specific parameters. 0 = Off; 1 = On Must use with mp_physics = 6 or 8 |
Lightning potential prediction |
3 |
Predicting the potential for lightning activity (based on Yair et al., 2010) |
&dynamics |
Diffusion, damping options,advection options |
|
Majority supported |
Note: Most options can be turned off with a value of 0, where applicable.
Benchmarks
Once AceCAST has been downloaded, there will be a folder called benchmarks in ../AceCASTv1.2/run directory.
It is recommended to use the test cases provided in the benchmarks directory for getting started with AceCAST.
- easter100:
2km horizontal resolution, 100x100x51 grid, 2 days simulation initialized with HRRR
Very small domain intended for quick validation testing.
Note that this case is only for testing and should not be used for performance benchmarking due to its small domain size.
- easter500:
2km horizontal resolution, 500x500x51 grid, 2 days simulation initialized with HRRR
Larger domain intended for benchmarking on small numbers of GPUs (ex. 1-4 V100 GPUs)
- easter1500 (not included by default):
1km horizontal resolution, 1500x1500x51 grid, 1 hour simulation initialized with HRRR
Large, high resolution domain intended for performance benchmarking on larger numbers of GPUs (ex. 8-32 V100 GPUs)
You must ensure minimum 11GB space to simulate this case
This case is not included in the distribution package
download here
AceCAST Scaling Advisor
The AceCAST Advisor script provides various tools to assist AceCAST users with configuring and running their AceCAST simulations. The primary functionalities are selected with the –tool flag and are summarized below. This should be the primary tool for users to transition their namelists to take advantage of AceCAST.This is the primary tool for users to check and refine their namelist settings to ensure they are compatible with AceCAST. If your namelist settings are not supported by AceCAST, this tool will tell you which namelist setting(s) is/are not compatible, and will provide you with options that are supported by AceCAST. Additionally, this tool also assists users with determining the minimum and maximum number of GPUs that should be used to run their simulation in an optimal amount of time.
To see more information about the AceCAST Advisor Script (acecast-advisor.sh), make sure you are in the ../AceCASTv1.2/run directory and enter the following in the command line and press enter:
$ ./acecast-advisor.sh -help
Invoking this command will list all of the AceCAST Advisor Script options
usage: acecast-advisor.sh [OPTIONS]
- OPTIONS:
-t, –tool <support-check|scaling-advisor> [Default: none]
-n, –namelist-file <FILE> [Default: ./namelist.input]
-v, –verbose
-h, –help
Tools
Namelist Support Check
The namelist support checking tool is used to ensure that all options in a given namelist are supported by AceCAST. The script will tell the user which options are not supported and what supported alternatives the user should consider to adapt their namelist for running with AceCAST.
Scaling Advisor
The scaling advisor tool is used to give AceCAST users a general sense of how many GPUs they should use to run their simulations. The scaling characteristics of AceCAST are relatively complex making this a challenging task otherwise. This tool will determine a minimum and maximum number of GPUs that the user should consider initially for a given namelist and GPU hardware specification.
- Examples:
./acecast-advisor.sh -t support-check -namelist-file namelist.input
./acecast-advisor.sh -tool support-check -namelist-file namelist.input
./acecast-advisor.sh -t scaling-advisor -namelist-file namelist.input
./acecast-advisor.sh -tool scale-advisor -namelist-file namelist.input
Note
When using the scaling advisor tool the script should be run on the intended runtime machine and runtime environment. Otherwise the script will not be able to determine critical information about the user’s GPU specifications.