diff --git a/pyperformance/data-files/benchmarks/MANIFEST b/pyperformance/data-files/benchmarks/MANIFEST
index a851218b..728bd984 100644
--- a/pyperformance/data-files/benchmarks/MANIFEST
+++ b/pyperformance/data-files/benchmarks/MANIFEST
@@ -17,6 +17,7 @@ gc_collect	<local>
 generators	<local>
 chameleon	<local>
 chaos	<local>
+comprehensions	<local>
 crypto_pyaes	<local>
 dask	<local>
 deepcopy	<local>
diff --git a/pyperformance/data-files/benchmarks/bm_comprehensions/pyproject.toml b/pyperformance/data-files/benchmarks/bm_comprehensions/pyproject.toml
new file mode 100644
index 00000000..4f47b6d2
--- /dev/null
+++ b/pyperformance/data-files/benchmarks/bm_comprehensions/pyproject.toml
@@ -0,0 +1,9 @@
+[project]
+name = "pyperformance_bm_comprehensions"
+requires-python = ">=3.7"
+dependencies = ["pyperf"]
+urls = {repository = "https://github.com/python/pyperformance"}
+dynamic = ["version"]
+
+[tool.pyperformance]
+name = "comprehensions"
diff --git a/pyperformance/data-files/benchmarks/bm_comprehensions/run_benchmark.py b/pyperformance/data-files/benchmarks/bm_comprehensions/run_benchmark.py
new file mode 100644
index 00000000..0315820b
--- /dev/null
+++ b/pyperformance/data-files/benchmarks/bm_comprehensions/run_benchmark.py
@@ -0,0 +1,95 @@
+"""
+Benchmark comprehensions.
+
+Author: Carl Meyer
+"""
+
+from dataclasses import dataclass
+from enum import Enum
+from typing import Iterable, List, Optional
+
+import pyperf
+
+
+class WidgetKind(Enum):
+    BIG = 1
+    SMALL = 2
+
+
+@dataclass
+class Widget:
+    widget_id: int
+    creator_id: int
+    derived_widget_ids: List[int]
+    kind: WidgetKind
+    has_knob: bool
+    has_spinner: bool
+
+
+class WidgetTray:
+    def __init__(self, owner_id: int, widgets: List[Widget]) -> None:
+        self.owner_id = owner_id
+        self.sorted_widgets: List[Widget] = []
+        self._add_widgets(widgets)
+
+    def _any_knobby(self, widgets: Iterable[Optional[Widget]]) -> bool:
+        return any(w.has_knob for w in widgets if w)
+
+    def _is_big_spinny(self, widget: Widget) -> bool:
+        return widget.kind == WidgetKind.BIG and widget.has_spinner
+
+    def _add_widgets(self, widgets: List[Widget]) -> None:
+        # sort order: mine first, then any widgets with derived knobby widgets in order of
+        # number derived, then other widgets in order of number derived, and we exclude
+        # big spinny widgets entirely
+        widgets = [w for w in widgets if not self._is_big_spinny(w)]
+        id_to_widget = {w.widget_id: w for w in widgets}
+        id_to_derived = {
+            w.widget_id: [id_to_widget.get(dwid) for dwid in w.derived_widget_ids]
+            for w in widgets
+        }
+        sortable_widgets = [
+            (
+                w.creator_id == self.owner_id,
+                self._any_knobby(id_to_derived[w.widget_id]),
+                len(id_to_derived[w.widget_id]),
+                w.widget_id,
+            )
+            for w in widgets
+        ]
+        sortable_widgets.sort()
+        self.sorted_widgets = [id_to_widget[sw[-1]] for sw in sortable_widgets]
+
+
+def make_some_widgets() -> List[Widget]:
+    widget_id = 0
+    widgets = []
+    for creator_id in range(3):
+        for kind in WidgetKind:
+            for has_knob in [True, False]:
+                for has_spinner in [True, False]:
+                    derived = [w.widget_id for w in widgets[::creator_id + 1]]
+                    widgets.append(
+                        Widget(
+                            widget_id, creator_id, derived, kind, has_knob, has_spinner
+                        )
+                    )
+                    widget_id += 1
+    assert len(widgets) == 24
+    return widgets
+
+
+def bench_comprehensions(loops: int) -> float:
+    range_it = range(loops)
+    widgets = make_some_widgets()
+    t0 = pyperf.perf_counter()
+    for _ in range_it:
+        tray = WidgetTray(1, widgets)
+        assert len(tray.sorted_widgets) == 18
+    return pyperf.perf_counter() - t0
+
+
+if __name__ == "__main__":
+    runner = pyperf.Runner()
+    runner.metadata["description"] = "Benchmark comprehensions"
+    runner.bench_time_func("comprehensions", bench_comprehensions)