[test] Add MPI Tool Information Interfaces (MPI_T) check

cscs-checks/prgenv/mpi_t.py

@@ -0,0 +1,63 @@
+# Copyright 2016-2020 Swiss National Supercomputing Centre (CSCS/ETH Zurich)
+# ReFrame Project Developers. See the top-level LICENSE file for details.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+import os
+import reframe as rfm
+import reframe.utility.sanity as sn
+
+
+@rfm.simple_test
+class MpiTCheck(rfm.RegressionTest):
+    def __init__(self):
+        self.descr = 'Checks MPI_T control/performance variables/categories'
+        self.valid_systems = ['daint:gpu', 'dom:gpu', 'daint:mc', 'dom:mc']
+        self.valid_prog_environs = ['PrgEnv-cray', 'PrgEnv-gnu', 'PrgEnv-pgi',
+                                    'PrgEnv-intel', 'PrgEnv-cray_classic']
+        self.build_system = 'SingleSource'
+        self.sourcepath = 'mpit_vars.c'
+        self.num_tasks_per_node = 1
+        self.variables = {'MPITEST_VERBOSE': '1', 'MPICH_VERSION_DISPLAY': '1'}
+        self.rpt = 'rpt'
+        self.executable_opts = [f'&> {self.rpt}']
+        self.maintainers = ['JG']
+        self.tags = {'production', 'craype', 'maintenance'}
+
+    @rfm.run_before('sanity')
+    def set_sanity(self):
+        # 1/ MPI Control Variables:
+        # --- extract reference data:
+        regex = r'^(?P<vars>MPIR\S+)$'
+        ref = os.path.join(self.stagedir, 'mpit_control_vars.ref')
+        self.ref_control_vars = sorted(sn.extractall(regex, ref, 'vars'))
+        # --- extract runtime data:
+        regex = r'^\t(?P<vars>MPIR\S+)\t'
+        rpt = os.path.join(self.stagedir, self.rpt)
+        self.run_control_vars = sorted(sn.extractall(regex, rpt, 'vars'))
+        # 2/ MPI Performance Variables:
+        # --- extract reference data:
+        regex = r'(?P<vars>\w+)'
+        ref = os.path.join(self.stagedir, 'mpit_perf_vars.ref')
+        self.ref_perf_vars = sorted(sn.extractall(regex, ref, 'vars'))
+        # --- extract runtime data:
+        regex = r'^\t(?P<vars>(nem_|rma_)\S+)\t'
+        rpt = os.path.join(self.stagedir, self.rpt)
+        self.run_perf_vars = sorted(sn.extractall(regex, rpt, 'vars'))
+        # 3/ MPI Category:
+        # --- extract reference data:
+        regex = r'^(?P<category>.*)$'
+        ref = os.path.join(self.stagedir, 'mpit_categories.ref')
+        ref_cat_vars = sorted(sn.extractall(regex, ref, 'category'))
+        self.ref_cat_vars = list(filter(None, ref_cat_vars))
+        # --- extract runtime data:
+        regex = (r'^(?P<category>Category \w+ has \d+ control variables, \d+'
+                 r' performance variables, \d+ subcategories)')
+        rpt = os.path.join(self.stagedir, self.rpt)
+        self.run_cat_vars = sorted(sn.extractall(regex, rpt, 'category'))
+        # 4/ Extracted lists can be compared (when sorted):
+        self.sanity_patterns = sn.all([
+            sn.assert_eq(self.ref_control_vars, self.run_control_vars),
+            sn.assert_eq(self.ref_perf_vars, self.run_perf_vars),
+            sn.assert_eq(self.ref_cat_vars, self.run_cat_vars),
+        ])

cscs-checks/prgenv/src/mpit_categories.ref

@@ -0,0 +1,19 @@
+Category CH3 has 1 control variables, 0 performance variables, 0 subcategories
+Category COLLECTIVE has 27 control variables, 0 performance variables, 0 subcategories
+Category COMMUNICATOR has 0 control variables, 0 performance variables, 0 subcategories
+Category CRAY_CONTROL has 17 control variables, 0 performance variables, 0 subcategories
+Category CRAY_DISPLAY has 7 control variables, 0 performance variables, 0 subcategories
+Category CRAY_DMAPP has 3 control variables, 0 performance variables, 0 subcategories
+Category CRAY_GNI has 32 control variables, 0 performance variables, 0 subcategories
+Category CRAY_GPU has 3 control variables, 0 performance variables, 0 subcategories
+Category CRAY_MPIIO has 18 control variables, 0 performance variables, 0 subcategories
+Category CRAY_RMA_STAT has 0 control variables, 7 performance variables, 0 subcategories
+Category DEBUGGER has 0 control variables, 0 performance variables, 0 subcategories
+Category DEVELOPER has 0 control variables, 0 performance variables, 0 subcategories
+Category DIMS has 0 control variables, 0 performance variables, 0 subcategories
+Category ERROR_HANDLING has 0 control variables, 0 performance variables, 0 subcategories
+Category FT has 0 control variables, 0 performance variables, 0 subcategories
+Category MEMORY has 0 control variables, 0 performance variables, 0 subcategories
+Category NEMESIS has 0 control variables, 1 performance variables, 0 subcategories
+Category PROCESS_MANAGER has 1 control variables, 0 performance variables, 0 subcategories
+Category THREADS has 0 control variables, 0 performance variables, 0 subcategories

cscs-checks/prgenv/src/mpit_control_vars.ref

@@ -0,0 +1,109 @@
+MPIR_CVAR_MPIIO_ABORT_ON_RW_ERROR
+MPIR_CVAR_MPIIO_AGGREGATOR_PLACEMENT_DISPLAY=0
+MPIR_CVAR_MPIIO_AGGREGATOR_PLACEMENT_STRIDE=-1
+MPIR_CVAR_MPIIO_CB_ALIGN=2
+MPIR_CVAR_MPIIO_DVS_MAXNODES=-1
+MPIR_CVAR_MPIIO_HINTS
+MPIR_CVAR_MPIIO_HINTS_DISPLAY=0
+MPIR_CVAR_MPIIO_MAX_NUM_IRECV=50
+MPIR_CVAR_MPIIO_MAX_NUM_ISEND=50
+MPIR_CVAR_MPIIO_MAX_SIZE_ISEND=10485760
+MPIR_CVAR_MPIIO_STATS=0
+MPIR_CVAR_MPIIO_STATS_FILE
+MPIR_CVAR_MPIIO_STATS_INTERVAL_MSEC
+MPIR_CVAR_MPIIO_TIMERS=0
+MPIR_CVAR_MPIIO_TIMERS_SCALE=0
+MPIR_CVAR_MPIIO_TIME_WAITS=1
+MPIR_CVAR_MPIIO_WRITE_EXIT_BARRIER=1
+MPIR_CVAR_MPIIO_DS_WRITE_CRAY=1
+MPIR_CVAR_SCATTERV_SHORT_MSG=-1
+MPIR_CVAR_DMAPP_A2A_SYMBUF_SIZE
+MPIR_CVAR_DMAPP_A2A_SHORT_MSG=4096
+MPIR_CVAR_DMAPP_A2A_USE_PUTS=0
+MPIR_CVAR_USE_DMAPP_COLL
+MPIR_CVAR_ALLGATHER_VSHORT_MSG=-1
+MPIR_CVAR_ALLGATHERV_VSHORT_MSG=-1
+MPIR_CVAR_ALLREDUCE_NO_SMP=0
+MPIR_CVAR_ALLTOALL_SHORT_MSG=-1
+MPIR_CVAR_ALLTOALLV_THROTTLE=8
+MPIR_CVAR_BCAST_ONLY_TREE=1
+MPIR_CVAR_BCAST_INTERNODE_RADIX=4
+MPIR_CVAR_BCAST_INTRANODE_RADIX=4
+MPIR_CVAR_COLL_BAL_INJECTION=-1
+MPIR_CVAR_COLL_OPT_OFF
+MPIR_CVAR_COLL_SYNC
+MPIR_CVAR_DMAPP_COLL_RADIX=64
+MPIR_CVAR_DMAPP_HW_CE
+MPIR_CVAR_GATHERV_SHORT_MSG=16384
+MPIR_CVAR_REDUCE_NO_SMP=0
+MPIR_CVAR_SCATTERV_SYNCHRONOUS=0
+MPIR_CVAR_SHARED_MEM_COLL_OPT
+MPIR_CVAR_NETWORK_BUFFER_COLL_OPT=0
+MPIR_CVAR_DMAPP_A2A_ARIES=0
+MPIR_CVAR_REDSCAT_COMMUTATIVE_LONG_MSG_SIZE=524288
+MPIR_CVAR_REDSCAT_MAX_COMMSIZE=6144
+MPIR_CVAR_DPM_DIR
+MPIR_CVAR_G2G_PIPELINE=0
+MPIR_CVAR_NO_GPU_DIRECT=0
+MPIR_CVAR_RDMA_ENABLED_CUDA=0
+MPIR_CVAR_RMA_FALLBACK=0
+MPIR_CVAR_SMP_SINGLE_COPY_OFF=0
+MPIR_CVAR_SMP_SINGLE_COPY_SIZE=8192
+MPIR_CVAR_GNI_SUPPRESS_PROC_FILE_WARNINGS=0
+MPIR_CVAR_GNI_BTE_MULTI_CHANNEL
+MPIR_CVAR_GNI_DATAGRAM_TIMEOUT
+MPIR_CVAR_GNI_DMAPP_INTEROP
+MPIR_CVAR_GNI_DYNAMIC_CONN
+MPIR_CVAR_GNI_FMA_SHARING
+MPIR_CVAR_GNI_FORK_MODE
+MPIR_CVAR_GNI_HUGEPAGE_SIZE
+MPIR_CVAR_GNI_LMT_GET_PATH
+MPIR_CVAR_GNI_LMT_PATH
+MPIR_CVAR_GNI_LOCAL_CQ_SIZE=8192
+MPIR_CVAR_GNI_MALLOC_FALLBACK=0
+MPIR_CVAR_GNI_MAX_EAGER_MSG_SIZE=8192
+MPIR_CVAR_GNI_MAX_NUM_RETRIES=16
+MPIR_CVAR_GNI_MAX_VSHORT_MSG_SIZE=-1
+MPIR_CVAR_GNI_MBOX_PLACEMENT
+MPIR_CVAR_GNI_MBOXES_PER_BLOCK=-1
+MPIR_CVAR_GNI_MDD_SHARING
+MPIR_CVAR_GNI_MEM_DEBUG_FNAME
+MPIR_CVAR_GNI_MAX_PENDING_GETS=-1
+MPIR_CVAR_GNI_GET_MAXSIZE=-1
+MPIR_CVAR_GNI_NDREG_ENTRIES=-1
+MPIR_CVAR_GNI_NDREG_LAZYMEM
+MPIR_CVAR_GNI_NDREG_MAXSIZE=-1
+MPIR_CVAR_GNI_NUM_BUFS=64
+MPIR_CVAR_GNI_NUM_MBOXES=-1
+MPIR_CVAR_GNI_RDMA_THRESHOLD=1024
+MPIR_CVAR_GNI_RECV_CQ_SIZE=40960
+MPIR_CVAR_GNI_ROUTING_MODE
+MPIR_CVAR_GNI_USE_UNASSIGNED_CPUS
+MPIR_CVAR_GNI_VC_MSG_PROTOCOL
+MPIR_CVAR_NEMESIS_ASYNC_PROGRESS
+MPIR_CVAR_NEMESIS_ON_NODE_ASYNC_OPT=-1
+MPIR_CVAR_GNI_NUM_DPM_CONNECTIONS=128
+MPIR_CVAR_ABORT_ON_ERROR=1
+MPIR_CVAR_CPUMASK_DISPLAY=0
+MPIR_CVAR_ENV_DISPLAY=0
+MPIR_CVAR_OPTIMIZED_MEMCPY=1
+MPIR_CVAR_STATS_DISPLAY=0
+MPIR_CVAR_STATS_VERBOSITY=1
+MPIR_CVAR_STATS_FILE
+MPIR_CVAR_RANK_REORDER_DISPLAY=0
+MPIR_CVAR_RANK_REORDER_METHOD=1
+MPIR_CVAR_USE_SYSTEM_MEMCPY=0
+MPIR_CVAR_VERSION_DISPLAY=1
+MPIR_CVAR_DMAPP_APP_IS_WORLD=0
+MPIR_CVAR_MEMCPY_MEM_CHECK=0
+MPIR_CVAR_MAX_THREAD_SAFETY
+MPIR_CVAR_MSG_QUEUE_DBG=0
+MPIR_CVAR_NO_BUFFER_ALIAS_CHECK=0
+MPIR_CVAR_DYNAMIC_VCS
+MPIR_CVAR_ALLOC_MEM_AFFINITY
+MPIR_CVAR_INTERNAL_MEM_AFFINITY
+MPIR_CVAR_ALLOC_MEM_POLICY
+MPIR_CVAR_ALLOC_MEM_PG_SZ
+MPIR_CVAR_CRAY_OPT_THREAD_SYNC=1
+MPIR_CVAR_OPT_THREAD_SYNC=1
+MPIR_CVAR_THREAD_YIELD_FREQ=10000

cscs-checks/prgenv/src/mpit_perf_vars.ref

@@ -0,0 +1,8 @@
+nem_fbox_fall_back_to_queue_count
+rma_basic_comm_ops_counter
+rma_basic_get_ops_counter
+rma_basic_put_ops_counter
+rma_basic_acc_ops_counter
+rma_basic_gacc_ops_counter
+rma_basic_cas_ops_counter
+rma_basic_fetch_ops_counter