Merge branch 'master' into test/gromacs

Vasileios Karakasis · web-flow · commit f19d73c669bf · 2019-10-25T09:33:55.000+02:00
diff --git a/config/cscs-ci.py b/config/cscs-ci.py
@@ -157,6 +157,7 @@ class ReframeSettings:
                 'format': (
                     '%(asctime)s|reframe %(version)s|'
                     '%(check_info)s|jobid=%(check_jobid)s|'
+                    'num_tasks=%(check_num_tasks)s|'
                     '%(check_perf_var)s=%(check_perf_value)s|'
                     'ref=%(check_perf_ref)s '
                     '(l=%(check_perf_lower_thres)s, '
diff --git a/config/cscs-pbs.py b/config/cscs-pbs.py
@@ -140,6 +140,7 @@ class ReframeSettings:
                 'format': (
                     '%(asctime)s|reframe %(version)s|'
                     '%(check_info)s|jobid=%(check_jobid)s|'
+                    'num_tasks=%(check_num_tasks)s|'
                     '%(check_perf_var)s=%(check_perf_value)s|'
                     'ref=%(check_perf_ref)s '
                     '(l=%(check_perf_lower_thres)s, '
diff --git a/config/cscs.py b/config/cscs.py
@@ -588,6 +588,7 @@ class ReframeSettings:
                 'format': (
                     '%(asctime)s|reframe %(version)s|'
                     '%(check_info)s|jobid=%(check_jobid)s|'
+                    'num_tasks=%(check_num_tasks)s|'
                     '%(check_perf_var)s=%(check_perf_value)s|'
                     'ref=%(check_perf_ref)s '
                     '(l=%(check_perf_lower_thres)s, '
diff --git a/cscs-checks/apps/tensorflow/tensorflow_check.py b/cscs-checks/apps/tensorflow/tensorflow_check.py
@@ -15,7 +15,7 @@ def __init__(self, model_name):
         self.num_gpus_per_node = 1
         tf_version = '1.14.0'
         cuda_version = '10.1.168'
-        tc_version = '19.09'
+        tc_version = '19.10'
         self.modules = ['TensorFlow/%s-CrayGNU-%s-cuda-%s-python3' %
                         (tf_version, tc_version, cuda_version)]
 
diff --git a/cscs-checks/apps/tensorflow/tf_horovod_check.py b/cscs-checks/apps/tensorflow/tf_horovod_check.py
@@ -11,7 +11,7 @@ def __init__(self, variant):
         self.valid_prog_environs = ['PrgEnv-gnu']
         tfshortver = '1.14'
         self.sourcesdir = 'https://github.com/tensorflow/benchmarks'
-        self.modules = ['Horovod/0.16.4-CrayGNU-19.09-tf-%s.0' % tfshortver]
+        self.modules = ['Horovod/0.16.4-CrayGNU-19.10-tf-%s.0' % tfshortver]
         if variant == 'small':
             self.valid_systems += ['dom:gpu']
             self.num_tasks = 8
diff --git a/cscs-checks/cuda/multi_gpu.py b/cscs-checks/cuda/multi_gpu.py
@@ -39,7 +39,7 @@ def __init__(self):
                                 '--end=%d' % self.max_buffer_size, '--csv']
         self.num_tasks = 0
         self.num_tasks_per_node = 1
-        if self.current_system.name in ['daint', 'dom']:
+        if self.current_system.name in ['daint', 'dom', 'tiger']:
             self.modules = ['craype-accel-nvidia60']
             self.num_gpus_per_node = 1
         else:
@@ -59,9 +59,13 @@ def __init__(self):
             'dom:gpu:d2d': (499000, -0.1, None, 'MB/s'),
             'kesch:cn:h2d':   (7583, -0.1, None, 'MB/s'),
             'kesch:cn:d2h':   (7584, -0.1, None, 'MB/s'),
-            'kesch:cn:d2d': (137408, -0.1, None, 'MB/s')
+            'kesch:cn:d2d': (137408, -0.1, None, 'MB/s'),
+            'tiger:gpu:h2d': (0, None, None, 'MB/s'),
+            'tiger:gpu:d2h': (0, None, None, 'MB/s'),
+            'tiger:gpu:d2d': (0, None, None, 'MB/s'),
         }
-        self.tags = {'diagnostic', 'mch', 'craype', 'external-resources'}
+        self.tags = {'diagnostic', 'benchmark', 'mch',
+                     'craype', 'external-resources'}
         self.maintainers = ['AJ', 'VK']
 
     def _xfer_pattern(self, xfer_kind, devno, nodename):
diff --git a/cscs-checks/libraries/boost/boost_python_check.py b/cscs-checks/libraries/boost/boost_python_check.py
@@ -2,7 +2,7 @@
 import reframe.utility.sanity as sn
 
 
-@rfm.parameterized_test(['1.70.0', '19.09', '2.7'], ['1.70.0', '19.09', '3.6'])
+@rfm.parameterized_test(['1.70.0', '19.10', '2.7'], ['1.70.0', '19.10', '3.6'])
 class BoostCrayGnuPythonTest(rfm.RegressionTest):
     def __init__(self, boost_version, cray_gnu_version, python_version):
         self.descr = ('Test for Boost-%s for CrayGnu-%s with python %s '
diff --git a/cscs-checks/microbenchmarks/spec-accel/spec.py b/cscs-checks/microbenchmarks/spec-accel/spec.py
@@ -15,6 +15,7 @@ def __init__(self, prg_envs):
         self.configs = {
             'PrgEnv-gnu':  'cscs-gnu',
             'PrgEnv-cray': 'cscs-cray',
+            'PrgEnv-pgi': 'cscs-pgi',
         }
 
         app_source = os.path.join(self.current_system.resourcesdir,
@@ -31,7 +32,7 @@ def __init__(self, prg_envs):
         self.refs = {
             env: {bench_name: (rt, None, 0.1, 'Seconds')
                   for (bench_name, rt) in
-                  zip(self.benchmarks, self.exec_times[env])}
+                  zip(self.benchmarks[env], self.exec_times[env])}
             for env in self.valid_prog_environs
         }
 
@@ -42,14 +43,18 @@ def __init__(self, prg_envs):
         self.executable = 'runspec'
 
         outfile = sn.getitem(sn.glob('result/ACCEL.*.log'), 0)
-        self.sanity_patterns = sn.all([sn.assert_found(
-            r'Success.*%s' % bn, outfile) for bn in self.benchmarks])
-
-        self.perf_patterns = {
-            bench_name: sn.avg(sn.extractall(
-                r'Success.*%s.*runtime=(?P<rt>[0-9.]+)' % bench_name,
-                outfile, 'rt', float))
-            for bench_name in self.benchmarks
+        self.sanity_patterns_ = {
+            env: sn.all([sn.assert_found(
+                r'Success.*%s' % bn, outfile) for bn in self.benchmarks[env]])
+            for env in self.valid_prog_environs
+        }
+
+        self.perf_patterns_ = {
+            env: {bench_name: sn.avg(sn.extractall(
+                  r'Success.*%s.*runtime=(?P<rt>[0-9.]+)' % bench_name,
+                  outfile, 'rt', float))
+                  for bench_name in self.benchmarks[env]}
+            for env in self.valid_prog_environs
         }
 
         self.maintainers = ['SK']
@@ -61,12 +66,15 @@ def setup(self, partition, environ, **job_opts):
         self.executable_opts = ['--config=%s' % self.configs[environ.name],
                                 '--platform NVIDIA',
                                 '--tune=base',
-                                '--device GPU'] + self.benchmarks
+                                '--device GPU'] + self.benchmarks[environ.name]
         self.reference = {
             'dom:gpu':   self.refs[environ.name],
             'daint:gpu': self.refs[environ.name]
         }
 
+        self.sanity_patterns = self.sanity_patterns_[environ.name]
+        self.perf_patterns = self.perf_patterns_[environ.name]
+
         super().setup(partition, environ, **job_opts)
         # The job launcher has to be changed since the `runspec`
         # script is not used with srun.
@@ -78,12 +86,22 @@ def setup(self, partition, environ, **job_opts):
 class SpecAccelCheckOpenCL(SpecAccelCheckBase):
     def __init__(self):
         self.descr = 'SPEC-accel benchmark OpenCL'
-        valid_prog_environs = ['PrgEnv-gnu', 'PrgEnv-cray']
+        valid_prog_environs = ['PrgEnv-gnu', 'PrgEnv-cray', 'PrgEnv-pgi']
 
-        self.benchmarks = ['systest', 'tpacf', 'stencil', 'lbm', 'fft',
+        self.benchmarks = {
+            'PrgEnv-gnu': ['systest', 'tpacf', 'stencil', 'lbm', 'fft',
                            'spmv', 'mriq', 'bfs', 'cutcp', 'kmeans',
                            'lavamd', 'cfd', 'nw', 'hotspot', 'lud',
-                           'ge', 'srad', 'heartwall', 'bplustree']
+                           'ge', 'srad', 'heartwall', 'bplustree'],
+            'PrgEnv-cray': ['systest', 'tpacf', 'stencil', 'lbm', 'fft',
+                            'spmv', 'mriq', 'bfs', 'cutcp', 'kmeans',
+                            'lavamd', 'cfd', 'nw', 'hotspot', 'lud',
+                            'ge', 'srad', 'heartwall', 'bplustree'],
+            'PrgEnv-pgi': ['systest', 'tpacf', 'stencil', 'lbm', 'fft',
+                           'spmv', 'mriq', 'bfs', 'kmeans',
+                           'lavamd', 'cfd', 'nw', 'hotspot', 'lud',
+                           'ge', 'srad', 'heartwall', 'bplustree'],
+        }
 
         self.exec_times = {
             'PrgEnv-gnu':  [10.7, 13.5, 17.0, 10.9, 11.91, 27.8,
@@ -92,6 +110,9 @@ def __init__(self):
             'PrgEnv-cray': [10.7, 13.5, 17.0, 10.9, 11.91, 27.8,
                             7.0, 23.1, 10.8, 24.9, 8.7, 24.4, 16.2,
                             15.7, 15.6, 11.1, 20.0, 41.9, 26.2],
+            'PrgEnv-pgi': [10.7, 30, 17.0, 10.9, 11.91, 27.8,
+                           7.0, 23.1, 24.9, 8.7, 24.4, 16.2,
+                           15.7, 15.6, 11.1, 20.0, 41.9, 26.2],
         }
 
         super().__init__(valid_prog_environs)
@@ -102,15 +123,22 @@ def __init__(self):
 class SpecAccelCheckOpenACC(SpecAccelCheckBase):
     def __init__(self):
         self.descr = 'SPEC-accel benchmark OpenACC'
-        valid_prog_environs = ['PrgEnv-cray']
+        valid_prog_environs = ['PrgEnv-cray', 'PrgEnv-pgi']
 
-        self.benchmarks = ['ostencil', 'olbm', 'omriq', 'md', 'ep',
+        self.benchmarks = {
+            'PrgEnv-cray': ['ostencil', 'olbm', 'omriq', 'md', 'ep',
+                            'clvrleaf', 'cg', 'seismic', 'sp', 'csp',
+                            'miniGhost', 'ilbdc', 'swim', 'bt'],
+            'PrgEnv-pgi': ['ostencil', 'olbm', 'omriq', 'md', 'ep',
                            'clvrleaf', 'cg', 'seismic', 'sp', 'csp',
-                           'miniGhost', 'ilbdc', 'swim', 'bt']
+                           'ilbdc', 'swim', 'bt'],
+        }
 
         self.exec_times = {
             'PrgEnv-cray': [18, 26, 121, 20, 73, 59, 41,
-                            50, 71, 34, 72, 41, 34, 378]
+                            50, 71, 34, 72, 41, 34, 378],
+            'PrgEnv-pgi': [18, 40, 116, 20, 71, 57, 45,
+                           56, 33, 32, 41, 48, 15]
         }
 
         super().__init__(valid_prog_environs)
diff --git a/cscs-checks/microbenchmarks/spec-accel/src/cscs-pgi b/cscs-checks/microbenchmarks/spec-accel/src/cscs-pgi
@@ -0,0 +1,42 @@
+ext           = compsys
+output_format = asc
+#table         = 0
+teeout        = no
+teerunout     = yes
+tune          = base
+
+######################################################################
+# Compiler selection, versions, and dates.
+#
+# sw_compiler<xx> = additional information describing the compiler
+# CC  = how the C Compiler is invoked
+# FC  = how the FORTRAN compiler is invoked
+
+
+default=default=default=default:
+sw_compiler  = Computer System Compiler C and Fortran90
+CC           = cc
+CXX          = CC
+FC           = ftn
+
+strict_rundir_verify = 0
+
+#######################################################################
+# 
+# OPTIMIZE = flags applicable to all compilers
+# COPTIMIZE = flags applicable to the C compiler
+# FOPTIMIZE = flags applicable to the Fortran compiler
+#
+# See your compiler manual for information on the flags available
+# for your compiler
+#
+
+OPTIMIZE       = -O3
+LIBS           = -lOpenCL
+#EXTRA_LIBS     = 
+EXTRA_CXXFLAGS = -std=c++03
+#EXTRA_CFLAGS   = 
+
+#######################################################################
+# End of user-supplied information.  
+#######################################################################
diff --git a/cscs-checks/tools/profiling_and_debugging/cuda_gdb.py b/cscs-checks/tools/profiling_and_debugging/cuda_gdb.py
@@ -14,7 +14,10 @@ def __init__(self):
         self.num_gpus_per_node = 1
         self.num_tasks_per_node = 1
         self.sourcesdir = 'src/Cuda'
-        self.executable = 'cuda-gdb cuda_gdb_check'
+        self.executable = 'cuda-gdb'
+        self.executable_opts = ['-x .in.cudagdb ./cuda_gdb_check']
+        # unload xalt to avoid runtime error:
+        self.pre_run = ['unset LD_PRELOAD']
         if self.current_system.name == 'kesch':
             self.exclusive_access = True
             self.modules = ['cudatoolkit/8.0.61']
@@ -35,24 +38,13 @@ def __init__(self):
                                          '-lcudart', '-lm']
 
         self.sanity_patterns = sn.all([
-            sn.assert_found(r'^\(cuda-gdb\) Breakpoint 1 at .*: file ',
-                            self.stdout),
+            sn.assert_found(r'^Breakpoint 1 at .*: file ', self.stdout),
             sn.assert_found(r'_jacobi-cuda-kernel.cu, line 59\.', self.stdout),
-            sn.assert_found(r'^\(cuda-gdb\) Starting program:', self.stdout),
             sn.assert_found(r'^\(cuda-gdb\) quit', self.stdout),
             sn.assert_lt(sn.abs(sn.extractsingle(
-                r'^\(cuda-gdb\)\s+\$1\s+=\s+(?P<result>\S+)', self.stdout,
+                r'\$1\s+=\s+(?P<result>\S+)', self.stdout,
                 'result', float)), 1e-5)
         ])
 
         self.maintainers = ['MK', 'JG']
         self.tags = {'production', 'craype'}
-
-    def setup(self, partition, environ, **job_opts):
-        super().setup(partition, environ, **job_opts)
-        self.job.launcher = LauncherWrapper(
-            self.job.launcher, 'printf', [
-                r"'break _jacobi-cuda-kernel.cu:59\n",
-                r"run\n", r"print *residue_d'", ' | '
-            ]
-        )
diff --git a/cscs-checks/tools/profiling_and_debugging/gdb4hpc.py b/cscs-checks/tools/profiling_and_debugging/gdb4hpc.py
@@ -41,7 +41,7 @@ def __init__(self, lang, extension):
             'OMP_PROC_BIND': 'true',
         }
         self.maintainers = ['JG']
-        self.tags = {'production', 'craype'}
+        self.tags = {'craype'}
         # gdb4hpc has its own way to launch a debugging job and needs an
         # additional jobscript. The reframe jobscript can be copied for that
         # purpose, by adding the cray_debug_ comments around the job launch
diff --git a/cscs-checks/tools/profiling_and_debugging/intel_vtune.py b/cscs-checks/tools/profiling_and_debugging/intel_vtune.py
@@ -93,6 +93,7 @@ def setup(self, partition, environ, **job_opts):
             sn.assert_eq(sn.extractsingle(r'(?P<perfevent>\d)',
                          self.paranoid_rpt, 'perfevent'), '0'),
             # check the hotspots:
-            sn.assert_found(r'^[jJ]acobi.*\$omp\$parallel@\d+\s+\d+.\d+s',
+            sn.assert_found(r'^.*\$omp\$parallel.*@(?P<line>\d+)\s+'
+                            r'(?P<sec>\S+)s',
                             self.summary_rpt),
         ])
diff --git a/cscs-checks/tools/profiling_and_debugging/ipm_mpi_omp.py b/cscs-checks/tools/profiling_and_debugging/ipm_mpi_omp.py
@@ -22,7 +22,7 @@ def __init__(self, lang):
             'PrgEnv-pgi': ['-O2', '-g', '-mp']
         }
         ipm_ver = '2.0.6'
-        tc_ver = '19.09'
+        tc_ver = '19.10'
         self.ipm_modules = {
             'PrgEnv-gnu': ['IPM/%s-CrayGNU-%s' % (ipm_ver, tc_ver)],
             'PrgEnv-cray': ['IPM/%s-CrayCCE-%s' % (ipm_ver, tc_ver)],
diff --git a/cscs-checks/tools/profiling_and_debugging/mpip_mpi_omp.py b/cscs-checks/tools/profiling_and_debugging/mpip_mpi_omp.py
@@ -22,6 +22,8 @@ def __init__(self, lang):
             'PrgEnv-intel': ['-g', '-qopenmp', '-O2'],
             'PrgEnv-pgi': ['-g', '-mp', '-O2']
         }
+        # unload xalt to avoid _buffer_decode error:
+        self.prebuild_cmd = ['module rm 2.7.10 ;module list -t']
         self.modules = ['mpiP']
         self.build_system = 'Make'
         self.num_iterations = 500
@@ -37,6 +39,8 @@ def __init__(self, lang):
         if lang == 'F90':
             self.build_system.max_concurrency = 1
 
+        # unload xalt to avoid _buffer_decode error:
+        self.pre_run = ['module rm 2.7.10']
         self.num_tasks = 96
         self.num_tasks_per_node = 24
         self.num_cpus_per_task = 1
@@ -66,7 +70,7 @@ def __init__(self, lang):
             # check performance report:
             sn.assert_found('Single collector task', self.rpt_file),
             sn.assert_eq(sn.extractsingle(
-                r'^.*_jacobi.*\s+(?P<mpi_isendline>\d+)\s.*Isend',
+                r'^\s+\d\s+\d\s_jacobi.{4}\s+(?P<mpi_isendline>\d+)\s.*Isend',
                 self.rpt_file, 'mpi_isendline'), mpi_isendline),
         ])
         self.maintainers = ['JG']
diff --git a/cscs-checks/tools/profiling_and_debugging/scorep_mpi_omp.py b/cscs-checks/tools/profiling_and_debugging/scorep_mpi_omp.py
@@ -13,15 +13,14 @@ def __init__(self, lang):
         self.descr = 'SCORE-P %s check' % lang
         self.valid_systems = ['daint:gpu', 'daint:mc', 'dom:gpu', 'dom:mc']
 
-        # Score-P fails with latest clang based cce compiler:
+        # Score-P fails with latest clang based cce and pgi compilers:
         # src/measurement/thread/fork_join/scorep_thread_fork_join_omp.c:402:
         # Fatal: Bug 'TPD == 0': Invalid OpenMP thread specific data object.
         # -> removing cce from supported compiler for now.
-        self.valid_prog_environs = ['PrgEnv-gnu', 'PrgEnv-intel', 'PrgEnv-pgi']
+        self.valid_prog_environs = ['PrgEnv-gnu', 'PrgEnv-intel']
         self.prgenv_flags = {
             'PrgEnv-gnu': ['-g', '-fopenmp'],
             'PrgEnv-intel': ['-g', '-openmp'],
-            'PrgEnv-pgi': ['-g', '-mp']
         }
         self.sourcesdir = os.path.join('src', lang)
         self.executable = 'jacobi'
@@ -62,7 +61,7 @@ def __init__(self, lang):
 
     def setup(self, partition, environ, **job_opts):
         scorep_ver = '6.0'
-        tc_ver = '19.09'
+        tc_ver = '19.10'
         cu_ver = '10.1'
         self.scorep_modules = {
             'PrgEnv-gnu': ['Score-P/%s-CrayGNU-%s' % (scorep_ver, tc_ver)],
@@ -81,6 +80,4 @@ def setup(self, partition, environ, **job_opts):
         self.build_system.cxxflags = prgenv_flags
         self.build_system.fflags = prgenv_flags
         self.build_system.ldflags = ['-lm']
-        self.build_system.options = [
-            "PREP='scorep --nopreprocess --mpp=mpi --thread=omp'"
-        ]
+        self.build_system.options = ["PREP='scorep --mpp=mpi --thread=omp'"]
diff --git a/cscs-checks/tools/profiling_and_debugging/src/Cuda/.in.cudagdb b/cscs-checks/tools/profiling_and_debugging/src/Cuda/.in.cudagdb
@@ -0,0 +1,3 @@
+break _jacobi-cuda-kernel.cu:59
+run
+print *residue_d
diff --git a/cscs-checks/tools/profiling_and_debugging/src/F90/_main.F90 b/cscs-checks/tools/profiling_and_debugging/src/F90/_main.F90
@@ -118,8 +118,13 @@ subroutine Init (myData)
         endif
         call MPI_Get_version(version, subversion, iErr)
         write (6,'(3(A,I1))') 'MPI-', version, '.', subversion, '#', provided
+#ifdef _OPENMP
         write (6,*) 'Jacobi', myData%iNumProcs, 'MPI process(es) with', &
-                    omp_get_max_threads(), 'OpenMP-', _OPENMP, ' thread(s)/process'
+              omp_get_max_threads(), 'OpenMP-', _OPENMP, ' thread(s)/process'
+#else
+        write (6,*) 'Jacobi', myData%iNumProcs, 'MPI process(es) with', &
+               ' 0 OpenMP-xxx thread(s)/process'
+#endif
 
 ! MPI_Get_library_version:
         call MPI_Get_library_version(mpilibversion, resultlen, iErr)
diff --git a/docs/running.rst b/docs/running.rst
diff --git a/reframe/core/containers.py b/reframe/core/containers.py
diff --git a/reframe/core/logging.py b/reframe/core/logging.py
diff --git a/reframe/frontend/statistics.py b/reframe/frontend/statistics.py
