forked from wrf-model/WTF
-
Notifications
You must be signed in to change notification settings - Fork 0
/
regTest_cuda_Yellowstone.wtf
167 lines (129 loc) · 8.41 KB
/
regTest_cuda_Yellowstone.wtf
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
##
## WRF SOFTWARE TEST SUITE CONFIGURATION FILE
##
## The WRF Software Test Suite is designed to automatically build and run WRF under a variety of
## configurations. People contributing changes to WRF are encouraged to run a small set of tests given in the
## "testNamelists/contributors" directory, as a way of confirming that their changes do not break existing code.
## This file should be the only one that requires modification by the test suite user.
##
## STEPS TO RUNNING THE TEST SUITE:
##
## 1) Familiarize yourself with the directory structure of the test suite by reading the README file in
## the top-level directory of the test suite.
## 2) Place one or more WRF source tar-files in the top-level "tarballs" directory. These tar-files
## must have the case-sensitive suffix ".tar". The rest of each tar-file name is used to label builds and test results.
## 3) Modify the variables in this WRF Test File. Variables near the top are most likely to require modification.
## 4) After checking the settings in this file to confirm they are correct for your needs, type the following at the command
## line: "scripts/run_WRF_Tests.ksh -R <this_file> >& run.log &".
##
## Author: Brian Bonnlander
##
## A helpful check to make sure that the compiler is ready to use on yellowstone.
export COMPILER_WTF=cuda
CHECK_MODULES=true
if $CHECK_MODULES; then
if [ `hostname | cut -c1-2` == "ys" -o \
`hostname | cut -c1-2` == "ca" -o \
`hostname | cut -c1-2` == "gs" ]; then
$WRF_TEST_ROOT/scripts/checkModules ${COMPILER_WTF}
if [ $? != 0 ]; then
echo "WTF: ${COMPILER_WTF} compiler not loaded; aborting."
exit 255
fi
fi
fi
# Specify which WRF target executables to build and test. Choose some or all of the following:
# "em_real nmm_nest em_chem em_chem_kpp em_b_wave em_quarter_ss". The final two cases are
# idealized cases, and if either is specified, then "em_real" must also be included earlier in the list.
export BUILD_TYPES="em_real"
# Select one of the directories below. This is the parent directory for all namelist.input files to be used for
# the tests. Below this directory should be subdirectories "em_real", "nmm_nest", etc with namelist.input files
# that should be run for that flavor of WRF.
#
# Code contributors: select the directory ending in "contributors".
#export NAMELIST_DIR=$WRF_TEST_ROOT/Namelists/contributors
#export NAMELIST_DIR=$WRF_TEST_ROOT/Namelists/internal
export NAMELIST_DIR=$WRF_TEST_ROOT/Namelists/cuda
#export NAMELIST_DIR=$WRF_TEST_ROOT/Namelists/caseStudies
# Choose the options that should be passed to the "configure" script used to build WRF. Each specified option
# builds a separate WRF executable that will be tested. Each option indicates a Fortran compiler to use for
# building WRF and a communication framework (serial, mpi, or openmp) to use
# when running WRF. If you are unsure of which options you need, unpack the WRF source code, start a terminal
# session, go to the top-level directory of the WRF source code, and type "./configure". The available options
# will be displayed. For example, on bluefire, the three options "1 2 3" specify the xlf compiler with the
# "serial", "openmp", and "mpi" communication frameworks, respectively. Leave a blank string for any options
# that you do not wish to build and test.
export CONFIGURE_SERIAL=""
export CONFIGURE_OPENMP=""
export CONFIGURE_MPI="11"
# If set to true, the tasks associated with building WRF and/or running tests using WRF will be submitted to a
# batch queue. All logically independent WRF builds (em_real, nmm_nest, em_chem, em_chem_kpp) will
# be built concurrently, each using NUM_PROC_BUILDS processors to build WRF. Then the remaining builds
# (em_b_wave, em_quarter_ss) will be built consecutively in the em_real directory tree; these depend on the
# existence of the "em_real" version of the WRF executable. If BATCH_COMPILE is set to false, then each
# WRF executable is built consecutively using NUM_PROC_BUILDS processors. If BATCH_TEST is false, all
# WRF tests are performed consecutively on the local machine.
# Recommended values: both false for personal computers, both true for mainframe computers with batch queues.
export BATCH_COMPILE=true
export BATCH_TEST=true
# Can be set to "LSF" or "NQS". Right now, "NQS" has been tested on janus.colorado.edu only.
# Ignored when BATCH_COMPILE and BATCH_TEST are false.
export BATCH_QUEUE_TYPE=LSF
# Names of batch queues to use for building and testing.
# Ignored when BATCH_COMPILE and BATCH_TEST are false.
export BUILD_QUEUE=caldera
export TEST_QUEUE=caldera
#export TEST_QUEUE=gpgpu
# Account charge code string for queue-managed computers. Ignored when BATCH_COMPILE and
# BATCH_TEST are false.
export BATCH_ACCOUNT=NMMM0054
# Number of processors/tasks to use for building WRF. Because of PGI license manager issues on
# yellowstone/caldera, we set this to a small number.
export NUM_PROC_BUILD=4
# Number of processors to use for running MPI-based tests,
# and checking test outcomes. On batch computers, this is not the *total* used at once, but the number
# used for each test or compilation. Recommended value for personal computers: the number of available
# processors on the machine. Recommended value for batch computers: at least 4, and no more than 12.
# NOTE: this value can be overridden by putting "NUM_PROCESSORS=<value>" in an individual test namelist.input file.
export NUM_PROC_TEST=6
# If true, all script commands are echoed to the standard error output stream. To capture this output to a file
# "run.log", run the tests using the command "scripts/run_WRF_Tests.ksh -R <this_file> >& run.log &".
# Recommended value: true for at least the first time using the test suite, false once you're confident the
# scripts are working.
export DEBUG_WTF=true
# Whether to build WRF optimized or not. Recommended value: false for code contributors, since turning on full
# optimization can cause bit-for-bit differences when comparing output from serial and non-serial versions of WRF.
export OPTIMIZE_WRF=false
# Whether to build the "em_chem" or "em_chem_kpp" versions of WRF optimized. Recommended value: false, since
# optimizing the large Fortran modules associated with these codes has taken up to 14 hours on some machines.
export OPTIMIZE_WRF_CHEM=false
# If set to true, WRF is built in such a way that checks array bounds, uninitialized variables, and serious
# floating point exceptions. Execution stops when one of these errors is encountered, and a debugger can be
# used to track down the problem. For Intel builds, "idb" is the debugger. For PGI builds, "pgdbg" is the
# debugger. For gfortran builds, "gdb" is the debugger.
#
# Recommended value for contributors: false, because several of the WRF microphysics options
# (e.g., CAM shortwave radiation) have spurious numerical exceptions occur as part of their expected behavior.
export TRAP_ERRORS=false
#
# ========================== STUFF BELOW HERE PROBABLY NEEDS NO MODIFICATION ==============================
#
# Directory where the WRF source tar files should be placed for testing.
export TARFILE_DIR=$WRF_TEST_ROOT/tarballs
# Directory where all the WRF source builds take place.
export BUILD_DIR=$WRF_TEST_ROOT/Builds
# The parent directory for all the regression test runs created and performed by this script.
export TEST_DIR=$WRF_TEST_ROOT/Runs
# The parent directory for all meteorological input data files to be used for the regression tests. Below this directory
# should be subdirectories "em_real", "nmm_nest", etc with data files that should be part of the test for each
# flavor of WRF. For idealized versions of WRF that use no observation data (e.g. em_b_wave), no input data are needed,
# so the directories can be empty.
export METDATA_DIR=$WRF_TEST_ROOT/Data
# Location of the library libfl.a; needed for building em_chem_kpp version of WRF. The location of this library
# is often in /usr/lib, /usr/local/lib, or /lib, but it could be anywhere the computer administrator chooses to
# install it. If unsure of its location, ask your system administrator. This option is ignored if the "em_chem_kpp"
# version of WRF is not built.
export FLEX_LIB_DIR='/usr/lib64'
# This probably does not need to be modified. Most WRF tests expect only basic nesting to be available within WRF.
# Recommended value: 1 (basic nesting).
export NEST_OPTION=1