Feature/python3 slurm

benchmarks/case.py

@@ -6,7 +6,7 @@
 from jinja2 import Template
 
 from settings import *
-from utils import read, write
+from benchmarks.utils import read, write
 
 
 class Case(object):
@@ -25,10 +25,10 @@ def __init__(self, name, type, config, work_dir):
         self.work_dir = work_dir
         self.stdout = ".".join((self.name, "log"))
         self.stdout_file = os.path.join(self.work_dir, self.stdout)
-        self.config = deepcopy(self._get_default_config())
+        self.config = deepcopy(self._get_case_config(config))
         self.config.update(config)
 
-    def _get_default_config(self):
+    def _get_case_config(self, config):
         """
         Returns a default config for the case. It will be merged by the passed config.
 
@@ -59,7 +59,7 @@ def _get_rms_context(self):
             "WALLTIME": self.config["walltime"],
             "NOTIFY": self.config["notify"],
             "EMAIL": self.config["email"],
-            "MODULE": self.config["module"],
+            "ENVIRONMENT": self.config["environment"],
             "COMMAND": self.config["command"],
             "RUNTIME_FILE": RUNTIME_FILE,
             "CPU_INFO_FILE": CPU_INFO_FILE,

benchmarks/espresso/__init__.py

@@ -1,23 +1,23 @@
 import os
 
 from benchmarks.case import Case
-from settings import ESPRESSO_MODULE
+from settings import ESPRESSO_ENVIRONMENT
 
 
 class ESPRESSOCase(Case):
     def __init__(self, name, type, config, work_dir):
         super(ESPRESSOCase, self).__init__(name, type, config, work_dir)
 
-    def _get_default_config(self):
+    def _get_case_config(self, config):
         """
         Returns a default config for the case. It will be merged by the passed config.
 
         Returns:
             dict
         """
-        default_config = super(ESPRESSOCase, self)._get_default_config()
+        default_config = super(ESPRESSOCase, self)._get_case_config(config)
         default_config.update({
-            "module": ESPRESSO_MODULE
+            "environment": ESPRESSO_ENVIRONMENT
         })
         return default_config
 

benchmarks/gromacs/__init__.py

@@ -1,4 +1,4 @@
-from settings import GROMACS_MODULE
+from settings import GROMACS_ENVIRONMENT
 from benchmarks.case import Case, os
 from benchmarks.utils import read, write
 
@@ -7,16 +7,16 @@ class GROMACSCase(Case):
     def __init__(self, name, type, config, work_dir):
         super(GROMACSCase, self).__init__(name, type, config, work_dir)
 
-    def _get_default_config(self):
+    def _get_case_config(self, config):
         """
         Returns a default config for the case. It will be merged by the passed config.
 
         Returns:
             dict
         """
-        default_config = super(GROMACSCase, self)._get_default_config()
+        default_config = super(GROMACSCase, self)._get_case_config(config)
         default_config.update({
-            "module": GROMACS_MODULE
+            "environment": GROMACS_ENVIRONMENT
         })
         return default_config
 

benchmarks/gromacs/cases.py

@@ -1,6 +1,6 @@
 import os
 
-from settings import NODES_CONFIGURATION, PPN_CONFIGURATION
+from settings import PBS_NP, NODES_CONFIGURATION, PPN_CONFIGURATION
 
 GROMACS_CASES = []
 CWD = os.path.dirname(__file__)
@@ -22,7 +22,7 @@
                             "template": "benchmarks/gromacs/inputs/{}/md.tpr".format(input_)
                         }
                     ],
-                    "command": "source GMXRC.bash; mpirun -np $PBS_NP gmx_mpi_d mdrun -ntomp 1 -s md.tpr -deffnm md"
+                    "command": f"source GMXRC.bash; mpirun -np {PBS_NP} gmx_mpi_d mdrun -ntomp 1 -s md.tpr -deffnm md"
                 }
             }
             GROMACS_CASES.append(config)

benchmarks/hpl/__init__.py

@@ -3,24 +3,25 @@
 
 from benchmarks.case import Case
 from benchmarks.utils import read
-from settings import HPL_MODULE, HPL_RESULT_REGEX
+from settings import HPL_ENVIRONMENT, HPL_RESULT_REGEX
 
 
 class HPLCase(Case):
     def __init__(self, name, type, config, work_dir):
         super(HPLCase, self).__init__(name, type, config, work_dir)
 
-    def _get_default_config(self):
+    def _get_case_config(self, config):
         """
         Returns a default config for the case. It will be merged by the passed config.
 
         Returns:
             dict
         """
-        default_config = super(HPLCase, self)._get_default_config()
+        mpirun_np = config.get('P', 1) * config.get('Q', 1)
+        default_config = super(HPLCase, self)._get_case_config(config)
         default_config.update({
-            "module": HPL_MODULE,
-            "command": "mpirun -np $PBS_NP xhpl &> {}".format(self.stdout),
+            "environment": HPL_ENVIRONMENT,
+            "command": f"mpirun -np {mpirun_np} xhpl &> {self.stdout}",
             "inputs": [
                 {
                     "name": "HPL.dat",

benchmarks/hpl/cases.json

@@ -1,30 +1,30 @@
 [
     {
         "nodes": 1,
-        "N": 200448,
+        "N": 6144,
         "NB": 192,
-        "P": 6,
-        "Q": 6
+        "P": 1,
+        "Q": 1
     },
     {
-        "nodes": 2,
-        "N": 283776,
+        "nodes": 4,
+        "N": 12288,
         "NB": 192,
-        "P": 8,
-        "Q": 9
+        "P": 2,
+        "Q": 2
     },
     {
-        "nodes": 4,
-        "N": 401280,
+        "nodes": 9,
+        "N": 18432,
         "NB": 192,
-        "P": 12,
-        "Q": 12
+        "P": 3,
+        "Q": 3
     },
     {
-        "nodes": 8,
-        "N": 567552,
+        "nodes": 16,
+        "N": 24576,
         "NB": 192,
-        "P": 16,
-        "Q": 18
+        "P": 4,
+        "Q": 4
     }
 ]

benchmarks/utils.py

@@ -4,11 +4,28 @@
 
 
 def read(path):
-    with open(path) as f:
-        return f.read()
+    with open(path, "rb") as f:
+        # PROBLEM: Not all input file formats are utf-8
+        # SOLUTION: Read bytes from file, then attempt to decode to string
+        file_bytes = f.read()
+        try:
+            file_string = file_bytes.decode("utf-8")
+            return file_string
+        except UnicodeError:
+            #print(f"Unable to decode file '{path}' - returning bytes")
+            pass
+        return file_bytes
 
 
 def write(path, content, mode="w+"):
+    # PROBLEM: Not all output file content is utf-8
+    # SOLUTION: Encode string as bytes, then write bytes to file
+    if type(content) == str:
+        content = content.encode("utf-8")
+    elif type(content) != bytes:
+        raise TypeError()
+    if "b" not in mode:
+        mode = "b" + mode
     with open(path, mode) as f:
         f.write(content)
 

benchmarks/vasp/__init__.py

@@ -1,24 +1,24 @@
 import os
 
-from settings import VASP_MODULE
+from settings import PBS_NP, VASP_ENVIRONMENT
 from benchmarks.case import Case
 
 
 class VASPCase(Case):
     def __init__(self, name, type, config, work_dir):
         super(VASPCase, self).__init__(name, type, config, work_dir)
 
-    def _get_default_config(self):
+    def _get_case_config(self, config):
         """
         Returns a default config for the case. It will be merged by the passed config.
 
         Returns:
             dict
         """
-        default_config = super(VASPCase, self)._get_default_config()
+        default_config = super(VASPCase, self)._get_case_config(config)
         default_config.update({
-            "module": VASP_MODULE,
-            "command": "mpirun -np $PBS_NP vasp &> {}".format(self.stdout),
+            "environment": VASP_ENVIRONMENT,
+            "command": f"mpirun -np {PBS_NP} vasp &> {self.stdout}",
         })
         return default_config
 

job.rms

@@ -1,28 +1,27 @@
 #!/bin/bash
 
 ## Below are resource manager directives.
-## Replace them with directives for your system (e.g. #SBATCH), if needed.
+## Replace them with directives for your system (e.g. #PBS), if needed.
+# https://slurm.schedmd.com/rosetta.pdf
 
-#PBS -N {{ NAME }}
-#PBS -j oe
-#PBS -R n
-#PBS -r n
-#PBS -q {{ QUEUE }}
-#PBS -l nodes={{ NODES }}
-#PBS -l ppn={{ PPN }}
-#PBS -l walltime={{ WALLTIME }}
-#PBS -m {{ NOTIFY }}
-#PBS -M {{ EMAIL }}
+#SBATCH --job-name={{ NAME }}
+#SBATCH --no-requeue
+#SBATCH --partition={{ QUEUE }}
+#SBATCH --nodes={{ NODES }}
+#SBATCH --ntasks-per-node={{ PPN }}
+#SBATCH --time={{ WALLTIME }}
+#SBATCH --mail-type {{ NOTIFY }}
+#SBATCH --mail-user={{ EMAIL }}
 
-cd $PBS_O_WORKDIR
+cd $SLURM_SUBMIT_DIR
 
 ## The use of environment modules is assumed below.
 ## Replace the directive with your system convention if modules are not installed.
 
-module add {{ MODULE }}
-
 ## DO NOT MODIFY LINES BELOW HERE
 
+{{ ENVIRONMENT }}
+
 start=`date +%s`
 
 {{ COMMAND }}

requirements.txt

@@ -1,4 +1,4 @@
-jinja2==2.8
+jinja2==3.1.2
 tabulate==0.7.5
 requests==2.20.1
-matplotlib==2.2.3
+matplotlib==3.7.1

settings.py

@@ -2,11 +2,11 @@
 
 # Name of the site where benchmarks are run. Should not contain space.
 # Should contain the cloud provider, company and cluster name and should be unique to identify the site.
-# Example: "AWS-C5-Exabyte-Cluster007
+# SITE_NAME = "AWS-t2.nano-16nodes"
 SITE_NAME = ""
 
 # Name of the location where the benchmarks are run. Should contain information about the location of the SITE.
-# Example: "East-US-zone-a"
+# SITE_LOCATION = "us-west-1b"
 SITE_LOCATION = ""
 
 # Name of the files to store the benchmark runtime, CPU and Memory info respectively.
@@ -22,37 +22,46 @@
 RESULTS_FILE_NAME = "results.json"
 RESULTS_FILE_PATH = os.path.join(os.path.dirname(__file__), "results", RESULTS_FILE_NAME)
 
-# Whether to publish the results. Set it to False to disable it.
-PUBLISH_RESULTS = True
+# Whether to publish the results to:
+# https://docs.google.com/spreadsheets/d/1oBHR8bp9q86MOxGYXcvUWHZa8KiC-uc_qX-UF1iqf-Y/edit
+# Set it to True to enable it.
+PUBLISH_RESULTS = False
 
 # Google Cloud Function endpoint to send the benchmark results to.
 REMOTE_SOURCE_API_URL = "https://us-central1-exabyte-io.cloudfunctions.net/Exabyte-Benchmarks-Results"
 
 # Resource Management System (RMS) settings for PBS/Torque. Adjust settings to accommodate for other resource managers.
 # http://docs.adaptivecomputing.com/torque/6-1-0/adminGuide/help.htm#topics/torque/2-jobs/submittingManagingJobs.htm
-PPN = 36
-QUEUE = "OF"
-NOTIFY = "abe"
+PPN = 1
+QUEUE = "queue1"
 WALLTIME = "05:00:00"
+NOTIFY = "BEGIN, END, FAIL"
 EMAIL = "[email protected]"
 DEFAULT_RMS_CONFIG = {
     "NODES": 1,
     "PPN": PPN,
     "QUEUE": QUEUE,
     "WALLTIME": WALLTIME,
-    "NOTIFY": NOTIFY,
-    "EMAIL": EMAIL,
     "RUNTIME_FILE": RUNTIME_FILE,
+    "NOTIFY": NOTIFY,
+    "EMAIL": EMAIL
 }
 
 # Name of qsub command
-QSUB_COMMAND = "qsub"
+QSUB_COMMAND = "sbatch"
+
+# Argument of -np option of mpirun (computed from P*Q for HPL benchmark)
+PBS_NP = 1
 
-# Environment Modules Settings
-HPL_MODULE = "hpl/22-i-174-impi-044"
-VASP_MODULE = "vasp/535-i-174-impi-044"
-ESPRESSO_MODULE = "espresso/540-i-174-impi-044"
-GROMACS_MODULE = "gromacs/20183-i-174-impi-044-md"
+# Environment Settings
+# HPL_ENVIRONMENT = '''
+# module load intelmpi
+# export PATH=$PATH:/home/ubuntu/hpl/bin/linux
+# '''
+HPL_ENVIRONMENT = "module add hpl/22-i-174-impi-044"
+VASP_ENVIRONMENT = "module add vasp/535-i-174-impi-044"
+ESPRESSO_ENVIRONMENT = "module add espresso/540-i-174-impi-044"
+GROMACS_ENVIRONMENT = "module add gromacs/20183-i-174-impi-044-md"
 
 # Regular expressions to extract the HPL results (N, NB, P, Q, TIME and GFLOPS)
 REGEX = {
@@ -68,6 +77,6 @@
 }
 
 # Specifies nodes and ppn configurations.
-# Cases are generated for all combinations.
+# Cases are generated for gromacs and vasp all combinations.
 NODES_CONFIGURATION = [1, 2, 4, 8]
 PPN_CONFIGURATION = [4, 8, 12, 16]