Merge branch 'develop' of https://github.com/reframe-hpc/reframe into feat/maint_nodes

Author: ekouts

docs/config_reference.rst

@@ -1141,8 +1141,24 @@ You may define different logger objects per system but *not* per partition.
    :default: :obj:`False`
 
    Emit a separate log record for each performance variable.
+
    Set this option to :obj:`True` if you want to keep compatibility with the performance logging prior to ReFrame 4.0.
 
+   .. deprecated:: 4.9
+
+      Please use :attr:`~config.logging.perflog_multiline` instead.
+
+
+.. py:attribute:: logging.perflog_multiline
+
+   :required: No
+   :default: :obj:`False`
+
+   Emit a separate log record for each performance variable.
+
+   .. versionadded:: 4.9
+
+
 .. py:attribute:: logging.target_systems
 
    :required: No

docs/tutorial.rst

@@ -2072,6 +2072,10 @@ Let's walk briefly through the most important parts of its configuration:
 
 Let's rerun our STREAM example using the new configuration:
 
+.. note::
+
+   ReFrame will now add entries in the perflogs for all tests, not only for performance tests.
+
 .. code-block:: bash
    :caption: Run with the Docker compose setup.
 

hpctestlib/microbenchmarks/gpu/dgemm/src/makefile.hip

@@ -4,7 +4,7 @@ RSMI_ROOT?=/opt/rocm/rocm_smi
 all: dgemm sgemm
 
 dgemm:
-	hipcc -O3 xgemm.cu -o $@.x -DTARGET_HIP -DGEMM_TYPE=double -DXBLAS_GEMM=XblasDgemm ${CXXFLAGS} -std=c++14 -I${ROCM_ROOT} -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas
+	hipcc -O3 xgemm.cu -o $@.x -DTARGET_HIP -DGEMM_TYPE=double -DXBLAS_GEMM=XblasDgemm ${CXXFLAGS} -std=c++14 -I${ROCM_ROOT}/include/rocblas/ -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas -lpthread
 
 sgemm:
-	hipcc -O3 xgemm.cu -o $@.x -DTARGET_HIP -DGEMM_TYPE=float -DXBLAS_GEMM=XblasSgemm ${CXXFLAGS} -std=c++14 -I${ROCM_ROOT} -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas
+	hipcc -O3 xgemm.cu -o $@.x -DTARGET_HIP -DGEMM_TYPE=float -DXBLAS_GEMM=XblasSgemm ${CXXFLAGS} -std=c++14 -I${ROCM_ROOT}/include/rocblas/ -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas -lpthread

hpctestlib/microbenchmarks/gpu/src/common/Xdevice/hip/smi.hpp

@@ -33,7 +33,7 @@ void Smi::setCpuAffinity(int id)
 {
   checkGpuIdIsSensible(id);
 
-  uint32_t numa_node;
+  int32_t numa_node;
   rsmiCheck( rsmi_topo_numa_affinity_get( id, &numa_node) );
   numa_run_on_node(numa_node);
 }

hpctestlib/microbenchmarks/gpu/src/gpu_burn/makefile.hip

@@ -1,4 +1,5 @@
 RSMI_ROOT?=/opt/rocm/rocm_smi
+ROCM_ROOT?=/opt/rocm/
 
 gpu_burn:
-	hipcc -O3 $@.cu -o $@.x -DTARGET_HIP ${CXXFLAGS} -std=c++14 -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas -pthreads
+	hipcc -O3 $@.cu -o $@.x -DTARGET_HIP ${CXXFLAGS} -std=c++14 -I${ROCM_ROOT}/include/rocblas/ -I${RSMI_ROOT}/include -lnuma -lrocm_smi64 -lrocblas -pthreads

reframe/core/logging.py

@@ -933,8 +933,8 @@ def _update_check_extras(self):
             self.extra['check_job_completion_time_unix'], r'%FT%T%:z'
         )
 
-    def log_performance(self, level, task, msg=None, multiline=False):
-        if self.check is None or not self.check.is_performance_check():
+    def log_result(self, level, task, msg=None, multiline=False):
+        if self.check is None:
             return
 
         _, part, env = task.testcase
@@ -951,7 +951,7 @@ def log_performance(self, level, task, msg=None, multiline=False):
         if msg is None:
             msg = 'sent by ' + self.extra['osuser']
 
-        if multiline:
+        if self.check.is_performance_check() and multiline:
             # Log one record for each performance variable
             for var, info in self.check.perfvalues.items():
                 val, ref, lower, upper, unit, result = info

reframe/frontend/cli.py

@@ -938,6 +938,13 @@ def restrict_logging():
         site_config = config.load_config(*conf_files)
         site_config.validate()
 
+        # Issue deprecation warnings
+        if site_config.get('logging/0/perflog_compat'):
+            printer.warning(
+                'the `perflog_compat` configuration option is deprecated; '
+                'please use `perflog_multiline` instead'
+            )
+
         if options.autodetect_method:
             printer.warning('RFM_AUTODETECT_METHOD is deprecated; '
                             'please use RFM_AUTODETECT_METHODS instead')

reframe/frontend/executors/__init__.py

@@ -232,9 +232,9 @@ def __init__(self, case, listeners=None, timeout=None):
 
         # Performance logging
         self._perflogger = logging.null_logger
-        self._perflog_compat = runtime.runtime().get_option(
-            'logging/0/perflog_compat'
-        )
+        self._perflog_multiline = runtime.runtime().get_option(
+            'logging/0/perflog_multiline'
+        ) or runtime.runtime().get_option('logging/0/perflog_compat')
 
     def duration(self, phase):
         # Treat pseudo-phases first
@@ -525,8 +525,8 @@ def finalize(self):
         self._current_stage = 'finalize'
         self._notify_listeners('on_task_success')
         try:
-            self._perflogger.log_performance(logging.INFO, self,
-                                             multiline=self._perflog_compat)
+            self._perflogger.log_result(logging.INFO, self,
+                                        multiline=self._perflog_multiline)
         except LoggingError as e:
             getlogger().warning(
                 f'could not log performance data for {self.testcase}: {e}'
@@ -550,8 +550,8 @@ def _wait_job(job):
         self._exc_info = exc_info or sys.exc_info()
         self._notify_listeners(callback)
         try:
-            self._perflogger.log_performance(logging.INFO, self,
-                                             multiline=self._perflog_compat)
+            self._perflogger.log_result(logging.INFO, self,
+                                        multiline=self._perflog_multiline)
         except LoggingError as e:
             getlogger().warning(
                 f'could not log performance data for {self.testcase}: {e}'

reframe/schemas/config.json

@@ -440,6 +440,7 @@
                 "properties": {
                     "level": {"$ref": "#/defs/loglevel"},
                     "perflog_compat": {"type": "boolean"},
+                    "perflog_multiline": {"type": "boolean"},
                     "handlers": {
                         "type": "array",
                         "items": {
@@ -621,6 +622,7 @@
         "general/verbose": 0,
         "logging/level": "undefined",
         "logging/perflog_compat": false,
+        "logging/perflog_multiline": false,
         "logging/target_systems": ["*"],
         "logging/handlers": [],
         "logging/handlers_perflog": [],

requirements.txt

@@ -1,6 +1,6 @@
 archspec==0.2.5
 argcomplete==3.1.2; python_version < '3.8'
-argcomplete==3.6.2; python_version >= '3.8'
+argcomplete==3.6.3; python_version >= '3.8'
 ClusterShell==1.9.3
 fasteners==0.19; python_version < '3.10'
 fasteners==0.20; python_version >= '3.10'
@@ -20,7 +20,8 @@ pytest-parallel==0.1.1
 pytest-rerunfailures==10.3; python_version == '3.6'
 pytest-rerunfailures==13.0; python_version == '3.7'
 pytest-rerunfailures==14.0; python_version == '3.8'
-pytest-rerunfailures==16.0.1; python_version >= '3.9'
+pytest-rerunfailures==16.0.1; python_version == '3.9'
+pytest-rerunfailures==16.1; python_version >= '3.10'
 PyYAML==6.0.1; python_version < '3.8'
 PyYAML==6.0.3; python_version >= '3.8'
 requests==2.27.1; python_version == '3.6'