PR 1/8: Hwbench fixes

hwbench/bench/benchmark.py

@@ -80,6 +80,16 @@ def __init__(self, engine_module: EngineModuleBase, parameters: BenchmarkParamet
         self.engine_module = engine_module
         self.skip = False
 
+    @property
+    def output_basename(self) -> str:
+        # Prefix the output files with the per-benchmark id used as the
+        # results.json key (get_name_with_position()). A single job expands via
+        # the scaling matrix (hosting_cpu_cores_scaling / stressor_range) into
+        # many runs that share the same engine `name`; without this prefix they
+        # all write into the same out_dir under the same filenames and only the
+        # last iteration survives.
+        return f"{self.parameters.get_name_with_position()}_{self.name}"
+
     def get_taskset(self, args):
         # Let's pin the CPU if needed
         if self.parameters.get_pinned_cpu():

hwbench/bench/benchmarks.py

@@ -16,9 +16,10 @@
 class Benchmarks:
     """A class to list and execute benchmarks to run."""
 
-    def __init__(self, out_dir, jobs_config) -> None:
+    def __init__(self, out_dir, jobs_config, verbose: bool = False) -> None:
         self.jobs_config = jobs_config
         self.out_dir = out_dir
+        self.verbose = verbose
         self.benchs: list[Benchmark] = []
         self.monitoring: Monitoring = None  # type: ignore[assignment]
         self.hardware: BaseHardware | None = None
@@ -175,7 +176,7 @@ def __schedule_benchmark(self, job, pinned_cpu, engine_module_parameter, validat
             for pdu in self.get_hardware().vendor.get_pdus():
                 pdu.connect_redfish()
                 pdu.detect()
-            self.monitoring = Monitoring(self.out_dir, self.jobs_config, self.get_hardware())
+            self.monitoring = Monitoring(self.out_dir, self.jobs_config, self.get_hardware(), verbose=self.verbose)
 
         # For each stressor, add a benchmark object to the list
         for stressor_count in self.jobs_config.get_stressor_range(job):

hwbench/bench/monitoring.py

@@ -35,9 +35,10 @@ def join(self):
 class Monitoring:
     """A class to perform monitoring."""
 
-    def __init__(self, out_dir, config, hardware: BaseHardware):
+    def __init__(self, out_dir, config, hardware: BaseHardware, verbose: bool = False):
         self.config = config
         self.out_dir = out_dir
+        self.verbose = verbose
         self.hardware = hardware
         self.vendor = hardware.get_vendor()
         self.metrics = MonitoringData()
@@ -114,7 +115,8 @@ def preup(self, precision_s: int):
         if self.turbostat:
             # Reinitialize turbostat metrics after reset (fast, doesn't run turbostat)
             self.turbostat.reinitialize_metrics()
-            print("Monitoring/turbostat: starting background monitoring with EOL-triggered sampling")
+            if self.verbose:
+                print("Monitoring/turbostat: starting background monitoring with EOL-triggered sampling")
             self.turbostat.start_background()
 
     def predown(self):
@@ -124,7 +126,8 @@ def predown(self):
         the background turbostat process.
         """
         if self.turbostat:
-            print("Monitoring/turbostat: stopping background monitoring")
+            if self.verbose:
+                print("Monitoring/turbostat: stopping background monitoring")
             self.turbostat.stop_background()
 
     def __monitor_bmc(self):

hwbench/bench/parameters.py

@@ -1,9 +1,15 @@
+from __future__ import annotations
+
 import pathlib
+from typing import TYPE_CHECKING
 
 from hwbench.environment.hardware import BaseHardware
 
 from .monitoring import Monitoring
 
+if TYPE_CHECKING:
+    from .benchmark import Benchmark
+
 
 class BenchmarkParameters:
     """A class to host parameters attached to a benchmark."""
@@ -38,6 +44,9 @@ def __init__(
         self.skip_method = skip_method
         self.sync_start = sync_start
         self.custom_parameters: dict[str, str] = kwargs
+        # Set once the owning Benchmark is built (see set_benchmark); until then
+        # get_name_with_position() falls back to the bare job name.
+        self.benchmark: Benchmark | None = None
 
     def get_benchmark(self):
         return self.benchmark

hwbench/engines/stressng.py

@@ -94,6 +94,8 @@ def run_cmd(self) -> list[str]:
             "--timeout",
             str(self.parameters.get_runtime()),
             "--metrics",
+            "--yaml",
+            f"{self.output_basename}.yaml",
         ]
 
         return self.get_taskset(args)

hwbench/engines/stressng_stream.py

@@ -22,6 +22,8 @@ def run_cmd(self) -> list[str]:
             "--timeout",
             str(self.parameters.get_runtime()),
             "--metrics",
+            "--yaml",
+            f"{self.name}.yaml",
             "--stream",
             str(self.parameters.get_engine_instances_count()),
         ]

hwbench/hwbench.py

@@ -37,7 +37,7 @@ def main():
     init_logging(tuning_out_dir / "hwbench-tuning.log")
 
     hwbench_config = config.Config(args.jobs_config)
-    benches = benchmarks.Benchmarks(out_dir, hwbench_config)
+    benches = benchmarks.Benchmarks(out_dir, hwbench_config, verbose=args.verbose)
 
     problems = env_hw.check_requirements() + benches.check_requirements()
 
@@ -109,6 +109,13 @@ def parse_options():
         default=True,
         help="Enable or disable tuning: this is useful when you want to test the system as-is.",
     )
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        action="store_true",
+        default=False,
+        help="Enable verbose output",
+    )
     return parser.parse_args()
 
 

hwbench/tests/parsing/stressng/v02103b/output

@@ -0,0 +1,5 @@
+{
+	"bogo ops/s": 217929.51,
+	"effective_runtime": 14.95
+}
+

hwbench/tests/parsing/stressng/v02103b/stdout

@@ -0,0 +1,11 @@
+stress-ng: info:  [205686] setting to a 15 secs run per stressor
+stress-ng: info:  [205686] dispatching hogs: 640 cpu
+stress-ng: info:  [205686] note: boost is disabled and this may impact top performance; setting /sys/devices/system/cpu/cpufreq/boost to 1 may improve performance.
+stress-ng: metrc: [205686] stressor       bogo ops real time  usr time  sys time   bogo ops/s     bogo ops/s CPU used per       RSS Max
+stress-ng: metrc: [205686]                           (secs)    (secs)    (secs)   (real time) (usr+sys time) instance (%)          (KB)
+stress-ng: metrc: [205686] cpu             3257280     14.95   9544.59      0.38    217929.51         341.26        99.78             0
+stress-ng: info:  [205686] skipped: 0
+stress-ng: info:  [205686] passed: 640: cpu (640)
+stress-ng: info:  [205686] failed: 0
+stress-ng: info:  [205686] metrics untrustworthy: 0
+stress-ng: info:  [205686] successful run completed in 15.02 secs