Add helper functions to run specifc profiling analysis, starting with…

… parallel scaling.

om3utils/profiling_analyses.py

@@ -0,0 +1,23 @@
+from pathlib import Path
+import xarray as xr
+
+from om3utils.payu_config_yaml import read_payu_config_yaml
+
+
+def scaling_ncpus(run_dir):
+    config_file = Path(run_dir) / "config.yaml"
+    config = read_payu_config_yaml(config_file.as_posix())
+    return config["ncpus"]
+
+
+def scaling_speedup(stats: xr.Dataset) -> xr.Dataset:
+    speedup = stats.tavg.sel(ncpus=stats["ncpus"].min()) / stats.tavg
+    speedup.name = "speedup"
+    return speedup
+
+
+def scaling_efficiency(stats: xr.Dataset) -> xr.Dataset:
+    speedup = scaling_speedup(stats)
+    eff = speedup / speedup.ncpus * 100 * stats["ncpus"].min()
+    eff.name = "parallel efficiency [%]"
+    return eff

tests/conftest.py

@@ -0,0 +1,34 @@
+import pytest
+import xarray as xr
+
+
+@pytest.fixture()
+def profiling_data():
+    regions = ["Total runtime", "Ocean Initialization"]
+    ncpus = [1, 2, 4]
+    hits = []
+    tmin = []
+    tmax = []
+    tavg = []
+    for n in ncpus:
+        hits.append([1, 2])
+        tmin.append([value / min(n, 2) for value in [138.600364, 2.344926]])
+        tmax.append([value / min(n, 2) for value in [138.600366, 2.345701]])
+        tavg.append([value / min(n, 2) for value in [600365, 2.345388]])
+
+    return regions, ncpus, hits, tmin, tmax, tavg
+
+
+@pytest.fixture()
+def simple_scaling_data(profiling_data):
+    regions, ncpus, hits, tmin, tmax, tavg = profiling_data
+
+    return xr.Dataset(
+        data_vars=dict(
+            hits=(["ncpus", "region"], hits),
+            tmin=(["ncpus", "region"], tmin),
+            tmax=(["ncpus", "region"], tmax),
+            tavg=(["ncpus", "region"], tavg),
+        ),
+        coords=dict(region=regions, ncpus=ncpus),
+    )

tests/test_profiling.py

@@ -21,23 +21,6 @@ def read(self, path: Path) -> dict:
         return self._data[path.name]
 
 
-@pytest.fixture()
-def profiling_data():
-    regions = ["Total runtime", "Ocean Initialization"]
-    ncpus = [1, 2, 4]
-    hits = []
-    tmin = []
-    tmax = []
-    tavg = []
-    for n in ncpus:
-        hits.append([1, 2])
-        tmin.append([value / min(n, 2) for value in [138.600364, 2.344926]])
-        tmax.append([value / min(n, 2) for value in [138.600366, 2.345701]])
-        tavg.append([value / min(n, 2) for value in [600365, 2.345388]])
-
-    return regions, ncpus, hits, tmin, tmax, tavg
-
-
 @pytest.fixture()
 def simple_scaling_runs(tmp_path, profiling_data):
     regions, ncpus, hits, tmin, tmax, tavg = profiling_data
@@ -55,21 +38,6 @@ def simple_scaling_runs(tmp_path, profiling_data):
     return run_dirs, parser
 
 
-@pytest.fixture()
-def simple_scaling_data(profiling_data):
-    regions, ncpus, hits, tmin, tmax, tavg = profiling_data
-
-    return xr.Dataset(
-        data_vars=dict(
-            hits=(["ncpus", "region"], hits),
-            tmin=(["ncpus", "region"], tmin),
-            tmax=(["ncpus", "region"], tmax),
-            tavg=(["ncpus", "region"], tavg),
-        ),
-        coords=dict(region=regions, ncpus=ncpus),
-    )
-
-
 def test_parse_profiling_data(tmp_path, simple_scaling_runs, simple_scaling_data):
     run_dirs, parser = simple_scaling_runs
 

tests/test_scaling.py

@@ -0,0 +1,27 @@
+import pytest
+
+from om3utils.profiling_analyses import scaling_ncpus, scaling_speedup, scaling_efficiency
+
+
+@pytest.fixture()
+def scaling_run_dir(tmp_path):
+    file = tmp_path / "config.yaml"
+    file.write_text("ncpus: 4")
+
+
+def test_scaling_ncpus(tmp_path, scaling_run_dir):
+    ncpus = scaling_ncpus(tmp_path)
+
+    assert ncpus == 4
+
+
+def test_scaling_speedup(simple_scaling_data):
+    speedup = scaling_speedup(simple_scaling_data)
+
+    assert (speedup.sel(region="Total runtime").values[:] == [1.0, 2.0, 2.0]).all()
+
+
+def test_efficiency(simple_scaling_data):
+    efficiency = scaling_efficiency(simple_scaling_data)
+
+    assert (efficiency.sel(region="Total runtime").values[:] == [100.0, 100.0, 50.0]).all()