diff --git a/loopy/transform/iname.py b/loopy/transform/iname.py
index 83598dcc26646261703ce9b24fecebdd8a975774..ddc412d94abec693057d796f6a4cd5f248c51f9d 100644
--- a/loopy/transform/iname.py
+++ b/loopy/transform/iname.py
@@ -1750,4 +1750,41 @@ def add_inames_to_insn(knl, inames, insn_match):
 # }}}
 
 
+# {{{ nest_and_predicate_instructions
+
+def nest_and_predicate_instructions(kernel, iname, within):
+    """
+    Nests an instruction inside of a loop while only executing on one iteration.
+    Useful for instructions which don't require all local hardware axes.
+    :arg iname: loop iname to nest instruction inside
+    "arg within: instruction to nest and predicate
+    """
+    from loopy.match import parse_match
+    within = parse_match(within)
+
+    from pymbolic import var
+    from pymbolic.primitives import Comparison
+
+    def rewrite_insn(insn):
+        if within(kernel, insn):
+            if iname in insn.within_inames:
+                raise LoopyError("instruction '%s' is already nested "
+                    "within iname '%s'" % (insn.id, iname))
+            return insn.copy(
+                within_inames=insn.within_inames | frozenset([iname]),
+                predicates=(
+                    # FIXME Do a better job finding the lower bound
+                    insn.predicates
+                    | frozenset([Comparison(var(iname), "==", 0)]))
+                )
+        else:
+            return insn
+
+    return kernel.copy(
+        instructions=[rewrite_insn(insn) for insn in kernel.instructions]
+        )
+
+
+# }}}
+
 # vim: foldmethod=marker