diff --git a/loopy/statistics.py b/loopy/statistics.py
index e27a0f482885658888c97081e4fc1d97fcd149fd..0ff82e5abb30eca279c66f8ce1cc0ce08dfc68a2 100755
--- a/loopy/statistics.py
+++ b/loopy/statistics.py
@@ -1197,6 +1197,9 @@ def get_unused_hw_axes_factor(knl, insn, disregard_local_axes, space=None):
 def count_insn_runs(knl, insn, count_redundant_work, disregard_local_axes=False):
 
     insn_inames = knl.insn_inames(insn)
+    if not insn_inames:
+        p = isl.PwQPolynomial('{ 1 }')
+        return GuardedPwQPolynomial(p, isl.Set.universe(p.domain().space))
 
     if disregard_local_axes:
         from loopy.kernel.data import LocalIndexTag
diff --git a/test/test_statistics.py b/test/test_statistics.py
index ea0bdb62bb75d8a5bcf7dd987c00c33b848091fd..14944407c9573fcbe69acf28a8e83524a76a1c31 100644
--- a/test/test_statistics.py
+++ b/test/test_statistics.py
@@ -71,6 +71,24 @@ def test_op_counter_basic():
     assert i32add == n*m*2
 
 
+def test_op_outsite():
+
+    data = [lp.GlobalArg("a", shape=(10,), dtype=float),
+            lp.GlobalArg("b", shape=(10,), dtype=float),
+            lp.GlobalArg("c", shape=(10,), dtype=float)]
+
+    knl = lp.make_kernel(
+        "[] -> {[]}",
+        "a[0] = b[0] + c[0]",
+        kernel_data=data
+    )
+
+    op = lp.get_op_map(knl)
+    flops = op.sum().eval_with_dict({})
+
+    assert flops == 1
+
+
 def test_op_counter_reduction():
 
     knl = lp.make_kernel(