also uploading the profiling data

benchmarks.json

@@ -69,6 +69,8 @@
             "project_id_regex": "",
             "after_script": [
                 "!reference [.export_variables, before_script]",
+                "!reference [.upload, script]",
+                "export DATAFILE=$(echo gpu_profile_*.sqlite3)",
                 "!reference [.upload, script]"
             ]
         }

jobscript/benchmarks/percolationgpu.sh

@@ -15,7 +15,7 @@ dev_name=$(nvidia-smi --id=${VISABLE_DEVICE} --query-gpu=gpu_name --format=csv,n
 
 # first test case channelflow
 basename_exe=$(basename ${EXECUTABLE})
-profile_file="ncuprofile_channelflow_${basename_exe}_${dev_name}"
+profile_file="ncuprofile_channelflow_${dev_name}"
 
 # with profling
 cmdChannelFlow="${PROFILING} -o $profile_file $MPIRUN $FULL_EXECUTABLE $FULL_ARGUMENT -NumericalSetup.timeSteps=2 -NumericalSetup.useParticles=false"
@@ -30,12 +30,12 @@ export DB_FILE="gpu_benchmark_percolation_${dev_name}.sqlite3"
 $cmdChannelFlow || echo "$cmdChannelFlow failed"
 
 # second test case percolation
-profile_file="ncuprofile_percolation_${basename_exe}_${dev_name}"
+profile_file="ncuprofile_percolation_${dev_name}"
 
 # with profling
 cmdPercolation="${PROFILING} -o $profile_file $MPIRUN $FULL_EXECUTABLE $FULL_ARGUMENT -NumericalSetup.timeSteps=2"
-export DB_FILE="gpu_profile_percolation_${dev_name}.sqlite3"
 echo "Runing $cmdPercolation"
+export DB_FILE="gpu_profile_percolation_${dev_name}.sqlite3"
 $cmdPercolation && ncu -i "${profile_file}.ncu-rep" --csv --page=raw > "$(basename -s ".ncu-rep" ${profile_file})".csv || echo "${cmdPercolation} failed"
 
 # without profiling

processors.py

@@ -49,7 +49,8 @@ def mesa_pd_sqlite(
             for timing in cb.build_iterate_query(
                 connection, from_table="timingPool", where_args=["runId", runid]
             ):
-                fields.update({timing["sweep"]: timing["average"] * timing["count"]})
+                fields.update(
+                    {timing["sweep"]: timing["average"] * timing["count"]})
 
             yield cb.DataPoint(
                 "MESA_PD_KernelBenchmark", time=time, fields=fields, tags=tags
@@ -78,9 +79,10 @@ def add_likwid_fields(search_pattern: str, perf_group: str) -> dict:
         return {}
 
 
-def add_ncu_profile(search_pattern: str):
+def add_ncu_profile(search_pattern: str, path="."):
     try:
-        ncu_file = find_profile_file(f"ncuprofile_*{search_pattern}*.csv")
+        ncu_file = find_profile_file(
+            f"ncuprofile_*{search_pattern}*.csv", path=path)
     except (FileNotFoundError,) as e:
         logger.info(e)
         return {}, {}
@@ -128,7 +130,8 @@ def _iter_runs(runs, *, measurement_name, perf_group: str, gpu_name=None):
         dp.tags.update(
             {
                 "mpi_num_processes": prod(
-                    [int(run[key]) for key in run.keys() if key.startswith("block")]
+                    [int(run[key])
+                     for key in run.keys() if key.startswith("block")]
                 )
             }
         )
@@ -263,6 +266,16 @@ def percolation_sqlite(
     *,
     perf_group: str,
 ):
+    measurement_name = "PercolationGPU"
+    basename = Path(db_file).stem
+    basename_splitted = basename.split("_")
+    gpu_name = basename_splitted[-1]
+    benchmark_case = ""
+    profile = False
+    if basename.startswith("gpu_profile"):
+        profile = True
+        measurement_name += "_profile"
+
     with cb.sqlite_context(db_file) as connection:
         runs = cb.query_complete_table(connection, "runs")
         field_keys = [
@@ -275,6 +288,7 @@ def percolation_sqlite(
             "time",
             "timeStepsPerSecond",
         ]
+
         for run in runs.fetchall():
             time = cb.util.get_time_stamp_from_env(
                 fallback=lambda: cb.util.time_conversion(run["timestamp"])
@@ -289,6 +303,17 @@ def percolation_sqlite(
                 for col in ["average", "min", "max", "count", "variance", "percentage"]:
                     fields.update({f'{timing["sweep"]}_{col}': timing[col]})
 
-            yield cb.DataPoint(
-                "PercolationGPU", time=time, fields=fields, tags=tags
-            ).asdict()
+            if profile:
+                benchmark_case = (
+                    "percolation" if tags["useParticles"] else "channelflow"
+                )
+                ncu_fields, ncu_tags = add_ncu_profile(
+                    f"{benchmark_case}*_{gpu_name}", path=os.path.dirname(db_file)
+                )
+                fields.update(ncu_fields)
+                tags.update(ncu_tags)
+
+            dp = cb.DataPoint(measurement_name, time=time,
+                              fields=fields, tags=tags)
+
+            yield dp.asdict()