diff --git a/test/lib/test_allocators.py b/test/lib/test_allocators.py
index 17b32a2..41be86c 100644
--- a/test/lib/test_allocators.py
+++ b/test/lib/test_allocators.py
@@ -4,6 +4,7 @@
 from amaranth import *
 from transactron import *
 from transactron.lib.allocators import *
+from transactron.lib.allocators import PreservedOrderAllocator
 from transactron.testing import (
     SimpleTestCircuit,
     TestCaseWithSimulator,
@@ -52,3 +53,48 @@ async def process(sim: TestbenchContext):
             for i in range(ways):
                 sim.add_testbench(make_allocator(i))
                 sim.add_testbench(make_deallocator(i))
+
+
+class TestPreservedOrderAllocator(TestCaseWithSimulator):
+    @pytest.mark.parametrize("entries", [5, 8])
+    def test_allocator(self, entries: int):
+        dut = SimpleTestCircuit(PreservedOrderAllocator(entries))
+
+        iterations = 5 * entries
+
+        allocated: list[int] = []
+        free: list[int] = list(range(entries))
+
+        async def allocator(sim: TestbenchContext):
+            for _ in range(iterations):
+                val = (await dut.alloc.call(sim)).ident
+                sim.delay(1e-9)  # Runs after deallocator
+                free.remove(val)
+                allocated.append(val)
+                await self.random_wait_geom(sim, 0.5)
+
+        async def deallocator(sim: TestbenchContext):
+            for _ in range(iterations):
+                while not allocated:
+                    await sim.tick()
+                idx = random.randrange(len(allocated))
+                val = allocated[idx]
+                if random.randint(0, 1):
+                    await dut.free.call(sim, ident=val)
+                else:
+                    await dut.free_idx.call(sim, idx=idx)
+                free.append(val)
+                allocated.pop(idx)
+                await self.random_wait_geom(sim, 0.4)
+
+        async def order_verifier(sim: TestbenchContext):
+            while True:
+                val = await dut.order.call(sim)
+                sim.delay(2e-9)  # Runs after allocator and deallocator
+                assert val.used == len(allocated)
+                assert val.order == allocated + free
+
+        with self.run_simulation(dut) as sim:
+            sim.add_testbench(order_verifier, background=True)
+            sim.add_testbench(allocator)
+            sim.add_testbench(deallocator)
diff --git a/transactron/lib/allocators.py b/transactron/lib/allocators.py
index be3f44f..93b1845 100644
--- a/transactron/lib/allocators.py
+++ b/transactron/lib/allocators.py
@@ -1,7 +1,8 @@
 from amaranth import *
 
-from transactron.core import Methods, TModule, def_methods
+from transactron.core import Method, Methods, TModule, def_method, def_methods
 from transactron.utils.amaranth_ext.elaboratables import MultiPriorityEncoder
+from amaranth.lib.data import ArrayLayout
 
 
 __all__ = ["PriorityEncoderAllocator"]
@@ -60,3 +61,72 @@ def _(_, ident):
             m.d.sync += not_used.bit_select(ident, 1).eq(1)
 
         return m
+
+
+class PreservedOrderAllocator(Elaboratable):
+    """Allocator with allocation order information.
+
+    This module allows to allocate and deallocate identifiers from a
+    continuous range. The order of allocations is preserved in the form of
+    a permutation of identifiers. Smaller positions correspond to earlier
+    (older) allocations.
+
+    Attributes
+    ----------
+    alloc : Method
+        Allocates a fresh identifier.
+    free : Method
+        Frees a previously allocated identifier.
+    free_idx : Method
+        Frees a previously allocated identifier at the given index of the
+        allocation order.
+    order : Method
+        Returns the allocation order as a permutation of identifiers
+        and the number of allocated identifiers.
+    """
+
+    def __init__(self, entries: int):
+        self.entries = entries
+
+        self.alloc = Method(o=[("ident", range(entries))])
+        self.free = Method(i=[("ident", range(entries))])
+        self.free_idx = Method(i=[("idx", range(entries))])
+        self.order = Method(
+            o=[("used", range(entries + 1)), ("order", ArrayLayout(range(self.entries), self.entries))],
+            nonexclusive=True,
+        )
+
+    def elaborate(self, platform) -> TModule:
+        m = TModule()
+
+        order = Signal(ArrayLayout(range(self.entries), self.entries), init=list(range(self.entries)))
+        used = Signal(range(self.entries + 1))
+        incr_used = Signal(range(self.entries + 1))
+
+        m.d.comb += incr_used.eq(used + self.alloc.run)
+        m.d.sync += used.eq(incr_used - self.free_idx.run)
+
+        @def_method(m, self.alloc, ready=used != self.entries)
+        def _():
+            return {"ident": order[used]}
+
+        @def_method(m, self.free_idx)
+        def _(idx):
+            for i in range(self.entries - 1):
+                with m.If(i >= idx):
+                    m.d.sync += order[i].eq(order[i + 1])
+            m.d.sync += order[self.entries - 1].eq(order[idx])
+
+        @def_method(m, self.free)
+        def _(ident):
+            idx = Signal(range(self.entries))
+            for i in range(self.entries):
+                with m.If(order[i] == ident):
+                    m.d.comb += idx.eq(i)
+            self.free_idx(m, idx=idx)
+
+        @def_method(m, self.order)
+        def _():
+            return {"used": used, "order": order}
+
+        return m