| blob | a65bae30a733193799ac601aa1e969fbc619c45f |
1 ;;;; This file implements local call analysis. A local call is a
2 ;;;; function call between functions being compiled at the same time.
3 ;;;; If we can tell at compile time that such a call is legal, then we
4 ;;;; change the combination to call the correct lambda, mark it as
5 ;;;; local, and add this link to our call graph. Once a call is local,
6 ;;;; it is then eligible for let conversion, which places the body of
7 ;;;; the function inline.
8 ;;;;
9 ;;;; We cannot always do a local call even when we do have the
10 ;;;; function being called. Calls that cannot be shown to have legal
11 ;;;; arg counts are not converted.
13 ;;;; This software is part of the SBCL system. See the README file for
14 ;;;; more information.
15 ;;;;
16 ;;;; This software is derived from the CMU CL system, which was
17 ;;;; written at Carnegie Mellon University and released into the
18 ;;;; public domain. The software is in the public domain and is
19 ;;;; provided with absolutely no warranty. See the COPYING and CREDITS
20 ;;;; files for more information.
24 ;;; This function propagates information from the variables in the
25 ;;; function FUN to the actual arguments in CALL. This is also called
26 ;;; by the VALUES IR1 optimizer when it sleazily converts MV-BINDs to
27 ;;; LETs.
28 ;;;
29 ;;; We flush all arguments to CALL that correspond to unreferenced
30 ;;; variables in FUN. We leave NILs in the COMBINATION-ARGS so that
31 ;;; the remaining args still match up with their vars.
32 ;;;
33 ;;; We also apply the declared variable type assertion to the argument
34 ;;; lvars.
48 ;; DX value generators must end their blocks: see UPDATE-UVL-LIVE-SETS.
49 ;; Uses of mupltiple-use LVARs already end their blocks, so we just need
50 ;; to process uses of single-use LVARs.
53 ;; If this LVAR's USE is good for DX, it is either a CAST, or it
54 ;; must be a regular combination whose arguments are potentially DX as well.
79 ;; Stack analysis requires that the CALL ends the block, so
80 ;; that MAP-BLOCK-NLXES sees the cleanup we insert here.
85 :mess-up entry
97 ;;; This function handles merging the tail sets if CALL is potentially
98 ;;; tail-recursive, and is a call to a function with a different
99 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
100 ;;; IR1 so as to place a local call in what might be a tail-recursive
101 ;;; context. Note that any call which returns its value to a RETURN is
102 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
103 ;;; might be optimized away.
104 ;;;
105 ;;; We destructively modify the set for the calling function to
106 ;;; represent both, and then change all the functions in callee's set
107 ;;; to reference the first. If we do merge, we reoptimize the
108 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
109 ;;; tail set type.
123 t)))))
125 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
126 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
127 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
128 ;;; the function in the REF node with the new function.
129 ;;;
130 ;;; We change the REF last, since changing the reference can trigger
131 ;;; LET conversion of the new function, but will only do so if the
132 ;;; call is local. Note that the replacement may trigger LET
133 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
134 ;;; with NEW-FUN before the substitution, since after the substitution
135 ;;; (and LET conversion), the call may no longer be recognizable as
136 ;;; tail-recursive.
150 \f
151 ;;;; external entry point creation
153 ;;; Return a LAMBDA form that can be used as the definition of the XEP
154 ;;; for FUN.
155 ;;;
156 ;;; If FUN is a LAMBDA, then we check the number of arguments
157 ;;; (conditional on policy) and call FUN with all the arguments.
158 ;;;
159 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
160 ;;; of supplied arguments by doing do an = test for each entry-point,
161 ;;; calling the entry with the appropriate prefix of the passed
162 ;;; arguments.
163 ;;;
164 ;;; If there is a &MORE arg, then there are a couple of optimizations
165 ;;; that we make (more for space than anything else):
166 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
167 ;;; no argument count error is possible.
168 ;;; -- We can omit the = clause for the last entry-point, allowing the
169 ;;; case of 0 more args to fall through to the more entry.
170 ;;;
171 ;;; We don't bother to policy conditionalize wrong arg errors in
172 ;;; optional dispatches, since the additional overhead is negligible
173 ;;; compared to the cost of everything else going on.
174 ;;;
175 ;;; Note that if policy indicates it, argument type declarations in
176 ;;; FUN will be verified. Since nothing is known about the type of the
177 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
178 ;;; are passed to the actual function.
201 ;; Force convertion of all entries
204 and n from min
208 ;; FIXME: Make sure that INDEX type distinguishes between
209 ;; target and host. (Probably just make the SB!XC:DEFTYPE
210 ;; different from CL:DEFTYPE.)
215 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
216 ;; deftransforms and lambda-conversion.
228 ;;; Make an external entry point (XEP) for FUN and return it. We
229 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
230 ;;; then associate this lambda with FUN as its XEP. After the
231 ;;; conversion, we iterate over the function's associated lambdas,
232 ;;; redoing local call analysis so that the XEP calls will get
233 ;;; converted.
234 ;;;
235 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
236 ;;; discover an XEP after the initial local call analyze pass.
258 res)))
260 ;;; Notice a REF that is not in a local-call context. If the REF is
261 ;;; already to an XEP, then do nothing, otherwise change it to the
262 ;;; XEP, making an XEP if necessary.
263 ;;;
264 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
265 ;;; it as though it was not an XEP reference (i.e. leave it alone).
273 \f
274 ;;; Attempt to convert all references to FUN to local calls. The
275 ;;; reference must be the function for a call, and the function lvar
276 ;;; must be used only once, since otherwise we cannot be sure what
277 ;;; function is to be called. The call lvar would be multiply used if
278 ;;; there is hairy stuff such as conditionals in the expression that
279 ;;; computes the function.
280 ;;;
281 ;;; If we cannot convert a reference, then we mark the referenced
282 ;;; function as an entry-point, creating a new XEP if necessary. We
283 ;;; don't try to convert calls that are in error (:ERROR kind.)
284 ;;;
285 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
286 ;;; can force analysis of newly introduced calls. Note that we don't
287 ;;; do LET conversion here.
312 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
313 ;;; calls into local calls when it is legal. We also attempt to
314 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
315 ;;; deletion of a function reference, but functions that start out
316 ;;; eligible for conversion must be noticed sometime.
317 ;;;
318 ;;; Note that there is a lot of action going on behind the scenes
319 ;;; here, triggered by reference deletion. In particular, the
320 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
321 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
322 ;;; the COMPONENT-LAMBDAS when it is. Also, the
323 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
324 ;;; it is not updated when we delete functions, etc. Only
325 ;;; COMPONENT-LAMBDAS is updated.
326 ;;;
327 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
328 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
329 ;;; LAMBDAS.
347 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
349 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
350 ;; apply: no-op.
353 ; NEW-FUNCTIONALS, hence is new.
354 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
359 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
361 component)))))
372 ;; The original CMU CL code seemed to implicitly assume that
373 ;; COMPONENT is the only one here. Let's make that explicit.
384 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
385 ;;; to be in an infinite recursive loop, then change the reference to
386 ;;; reference a fresh copy. We return whichever function we decide to
387 ;;; reference.
405 original-functional)))))))
410 calls this LET-converted local function:~
411 ~% ~S"
415 expand because expansion refers to ~
416 the optimized away object ~S."
417 losing-local-object)))
424 original-functional)
427 converted-lambda))))
428 original-functional))
430 ;;; Dispatch to the appropriate function to attempt to convert a call.
431 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
432 ;;; optimization as well as in local call analysis. If the call is is
433 ;;; already :LOCAL, we do nothing. If the call is already scheduled
434 ;;; for deletion, also do nothing (in addition to saving time, this
435 ;;; also avoids some problems with optimizing collections of functions
436 ;;; that are partially deleted.)
437 ;;;
438 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
439 ;;; called on a :INITIAL component, we don't care whether the caller
440 ;;; and callee are in the same component. Afterward, we must stick
441 ;;; with whatever component division we have chosen.
442 ;;;
443 ;;; Before attempting to convert a call, we see whether the function
444 ;;; is supposed to be inline expanded. Call conversion proceeds as
445 ;;; before after any expansion.
446 ;;;
447 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
448 ;;; warnings will get the right context.
463 component))))
466 original-fun))
484 ;;; Attempt to convert a multiple-value call. The only interesting
485 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
486 ;;; exactly one reference and no XEP, and is called with one values
487 ;;; lvar.
488 ;;;
489 ;;; We change the call to be to the last optional entry point and
490 ;;; change the call to be local. Due to our preconditions, the call
491 ;;; should eventually be converted to a let, but we can't do that now,
492 ;;; since there may be stray references to the e-p lambda due to
493 ;;; optional defaulting code.
494 ;;;
495 ;;; We also use variable types for the called function to construct an
496 ;;; assertion for the values lvar.
497 ;;;
498 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
521 ;;; Attempt to convert a call to a lambda. If the number of args is
522 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
523 ;;; from future consideration. If the argcount is O.K. then we just
524 ;;; convert it.
533 'local-argument-mismatch
534 :format-control
535 "function called with ~R argument~:P, but wants exactly ~R"
538 \f
539 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
541 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
542 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
543 ;;; a call to the correct entry point. If &KEY args are supplied, then
544 ;;; dispatch to a subfunction. We don't convert calls to functions
545 ;;; that have a &MORE (or &REST) arg.
554 'local-argument-mismatch
555 :format-control
556 "function called with ~R argument~:P, but wants at least ~R"
568 'local-argument-mismatch
569 :format-control
570 "function called with ~R argument~:P, but wants at most ~R"
571 :format-arguments
576 ;;; This function is used to convert a call to an entry point when
577 ;;; complex transformations need to be done on the original arguments.
578 ;;; ENTRY is the entry point function that we are calling. VARS is a
579 ;;; list of variable names which are bound to the original call
580 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
581 ;;; is the list of arguments to the entry point function.
582 ;;;
583 ;;; In order to avoid gruesome graph grovelling, we introduce a new
584 ;;; function that rearranges the arguments and calls the entry point.
585 ;;; We analyze the new function and the entry point immediately so
586 ;;; that everything gets converted during the single pass.
603 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
604 ;;; function into a local call to the MAIN-ENTRY.
605 ;;;
606 ;;; First we verify that all keywords are constant and legal. If there
607 ;;; aren't, then we warn the user and don't attempt to convert the call.
608 ;;;
609 ;;; We massage the supplied &KEY arguments into the order expected
610 ;;; by the main entry. This is done by binding all the arguments to
611 ;;; the keyword call to variables in the introduced lambda, then
612 ;;; passing these values variables in the correct order when calling
613 ;;; the main entry. Unused arguments (such as the keywords themselves)
614 ;;; are discarded simply by not passing them along.
615 ;;;
616 ;;; If there is a &REST arg, then we bundle up the args and pass them
617 ;;; to LIST.
639 (:keyword
702 (:optional
706 (:rest
709 (:keyword
727 \f
728 ;;;; LET conversion
729 ;;;;
730 ;;;; Converting to a LET has differing significance to various parts
731 ;;;; of the compiler:
732 ;;;; -- The body of a LET is spliced in immediately after the
733 ;;;; corresponding combination node, making the control transfer
734 ;;;; explicit and allowing LETs to be mashed together into a single
735 ;;;; block. The value of the LET is delivered directly to the
736 ;;;; original lvar for the call, eliminating the need to
737 ;;;; propagate information from the dummy result lvar.
738 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
739 ;;;; that there is only one expression that the variable can be bound
740 ;;;; to, and this is easily substituted for.
741 ;;;; -- LETs are interesting to environment analysis and to the back
742 ;;;; end because in most ways a LET can be considered to be "the
743 ;;;; same function" as its home function.
744 ;;;; -- LET conversion has dynamic scope implications, since control
745 ;;;; transfers within the same environment are local. In a local
746 ;;;; control transfer, cleanup code must be emitted to remove
747 ;;;; dynamic bindings that are no longer in effect.
749 ;;; Set up the control transfer to the called CLAMBDA. We split the
750 ;;; call block immediately after the call, and link the head of
751 ;;; CLAMBDA to the call block. The successor block after splitting
752 ;;; (where we return to) is returned.
753 ;;;
754 ;;; If the lambda is is a different component than the call, then we
755 ;;; call JOIN-COMPONENTS. This only happens in block compilation
756 ;;; before FIND-INITIAL-DFO.
773 next-block)))
775 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
776 ;;; same set as; but leave CLAMBDA with a valid tail set value of
777 ;;; its own, for the benefit of code which might try to pull
778 ;;; something out of it (e.g. return type).
780 ;; Until sbcl-0.pre7.37.flaky5.2, we did
781 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
782 ;; (SETF (TAIL-SET-FUNS TAILS)
783 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
784 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
785 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
786 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
787 ;; inconsistent, and perhaps unsafe, for us to think we're in the
788 ;; tail set. Unfortunately..
789 ;;
790 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
791 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
792 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
793 ;; which you called in order to get the address of an external LAMBDA):
794 ;; the external function was defined in terms of internal function,
795 ;; which was LET-converted, and then things blew up downstream when
796 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
797 ;; the now-NILed-out TAIL-SET. So..
798 ;;
799 ;; To deal with this problem, we no longer NIL out
800 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
801 ;; * If we're the only function in TAIL-SET-FUNS, it should
802 ;; be safe to leave ourself linked to it, and it to you.
803 ;; * If there are other functions in TAIL-SET-FUNS, then we're
804 ;; afraid of future optimizations on those functions causing
805 ;; the TAIL-SET object no longer to be valid to describe our
806 ;; return value. Thus, we delete ourselves from that object;
807 ;; but we save a newly-allocated tail-set, derived from the old
808 ;; one, for ourselves, for the use of later code (e.g.
809 ;; FINALIZE-XEP-DEFINITION) which might want to
810 ;; know about our return type.
819 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
820 ;; remain valid in this case, so we nuke it on the theory that
821 ;; missing information tends to be less dangerous than incorrect
822 ;; information.
825 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
826 ;;; its LETs to LETs for the CALL's home function. We merge the calls
827 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
828 ;;; in the process. We also merge the ENTRIES.
829 ;;;
830 ;;; We also unlink the function head from the component head and set
831 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
832 ;;; recomputed.
852 ;; CLAMBDA belongs to HOME now.
862 ;; All of CLAMBDA's LETs belong to HOME now.
868 ;; CLAMBDA no longer has an independent existence as an entity
869 ;; which has LETs.
872 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
873 ;; DFO dependencies.
877 ;; CLAMBDA no longer has an independent existence as an entity
878 ;; which calls things or has DFO dependencies.
881 ;; All of CLAMBDA's ENTRIES belong to HOME now.
885 ;; CLAMBDA no longer has an independent existence as an entity
886 ;; with ENTRIES.
891 ;;; Handle the value semantics of LET conversion. Delete FUN's return
892 ;;; node, and change the control flow to transfer to NEXT-BLOCK
893 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
912 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
919 ;;; We are converting FUN to be a LET when the call is in a non-tail
920 ;;; position. Any previously tail calls in FUN are no longer tail
921 ;;; calls, and must be restored to normal calls which transfer to
922 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
923 ;;; the RETURN-RESULT, because the return might have been deleted (if
924 ;;; all calls were TR.)
943 ;; The called function might be an assignment in the
944 ;; case where we are currently converting that function.
945 ;; In steady-state, assignments never appear as a called
946 ;; function.
947 (:assignment
951 ;;; Deal with returning from a LET or assignment that we are
952 ;;; converting. FUN is the function we are calling, CALL is a call to
953 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
954 ;;; NULL if call is a tail call.
955 ;;;
956 ;;; If the call is not a tail call, then we must do
957 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
958 ;;; its value out of the enclosing non-let function. When call is
959 ;;; non-TR, we must convert it back to an ordinary local call, since
960 ;;; the value must be delivered to the receiver of CALL's value.
961 ;;;
962 ;;; We do different things depending on whether the caller and callee
963 ;;; have returns left:
965 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
966 ;;; Either the function doesn't return, or all returns are via
967 ;;; tail-recursive local calls.
968 ;;; -- If CALL is a non-tail call, or if both have returns, then
969 ;;; we delete the callee's return, move its uses to the call's
970 ;;; result lvar, and transfer control to the appropriate
971 ;;; return point.
972 ;;; -- If the callee has a return, but the caller doesn't, then we
973 ;;; move the return to the caller.
1002 ;;; Actually do LET conversion. We call subfunctions to do most of the
1003 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1004 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1005 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1006 ;;; new references to it.
1011 nil
1012 next-block)))
1016 ;; If CALL has a derive type NIL, it means that "its return" is
1017 ;; unreachable, but the next BIND is still reachable; in order to
1018 ;; not confuse MAYBE-TERMINATE-BLOCK...
1021 ;;; Reoptimize all of CALL's args and its result.
1030 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1031 ;;; LET converted?
1033 ;; From the user's point of view, LET-converting something that
1034 ;; has a name is inlining it. (The user can't see what we're doing
1035 ;; with anonymous things, and suppressing inlining
1036 ;; for such things can easily give Python acute indigestion, so
1037 ;; we don't.)
1039 ;; ANSI requires that explicit NOTINLINE be respected.
1041 ;; If (= LET-CONVERSION 0) we can guess that inlining
1042 ;; generally won't be appreciated, but if the user
1043 ;; specifically requests inlining, that takes precedence over
1044 ;; our general guess.
1048 ;;; We also don't convert calls to named functions which appear in the
1049 ;;; initial component, delaying this until optimization. This
1050 ;;; minimizes the likelihood that we will LET-convert a function which
1051 ;;; may have references added due to later local inline expansion.
1058 ;;; This function is called when there is some reason to believe that
1059 ;;; CLAMBDA might be converted into a LET. This is done after local
1060 ;;; call analysis, and also when a reference is deleted. We return
1061 ;;; true if we converted.
1066 ;; We only convert to a LET when the function is a normal local
1067 ;; function, has no XEP, and is referenced in exactly one local
1068 ;; call. Conversion is also inhibited if the only reference is in
1069 ;; a block about to be deleted.
1070 ;;
1071 ;; These rules limiting LET conversion may seem unnecessarily
1072 ;; restrictive, since there are some cases where we could do the
1073 ;; return with a jump that don't satisfy these requirements. The
1074 ;; reason for doing things this way is that it makes the concept
1075 ;; of a LET much more useful at the level of IR1 semantics. The
1076 ;; :ASSIGNMENT function kind provides another way to optimize
1077 ;; calls to single-return/multiple call functions.
1078 ;;
1079 ;; We don't attempt to convert calls to functions that have an
1080 ;; XEP, since we might be embarrassed later when we want to
1081 ;; convert a newly discovered local call. Also, see
1082 ;; OK-INITIAL-CONVERT-P.
1099 nil)))
1108 t))))
1109 \f
1110 ;;;; tail local calls and assignments
1112 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1113 ;;; they definitely won't generate any cleanup code. Currently we
1114 ;;; recognize lexical entry points that are only used locally (if at
1115 ;;; all).
1129 ;;; If a potentially TR local call really is TR, then convert it to
1130 ;;; jump directly to the called function. We also call
1131 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1132 ;;; tail-convert. The second is the value of M-C-T-A.
1139 ;; this is a tail call
1143 ;; If the call is in an XEP, we might decide to make it
1144 ;; non-tail so that we can use known return inside the
1145 ;; component.
1157 ;;; This is called when we believe it might make sense to convert
1158 ;;; CLAMBDA to an assignment. All this function really does is
1159 ;;; determine when a function with more than one call can still be
1160 ;;; combined with the calling function's environment. We can convert
1161 ;;; when:
1162 ;;; -- The function is a normal, non-entry function, and
1163 ;;; -- Except for one call, all calls must be tail recursive calls
1164 ;;; in the called function (i.e. are self-recursive tail calls)
1165 ;;; -- OK-INITIAL-CONVERT-P is true.
1166 ;;;
1167 ;;; There may be one outside call, and it need not be tail-recursive.
1168 ;;; Since all tail local calls have already been converted to direct
1169 ;;; transfers, the only control semantics needed are to splice in the
1170 ;;; body at the non-tail call. If there is no non-tail call, then we
1171 ;;; need only merge the environments. Both cases are handled by
1172 ;;; LET-CONVERT.
1173 ;;;
1174 ;;; ### It would actually be possible to allow any number of outside
1175 ;;; calls as long as they all return to the same place (i.e. have the
1176 ;;; same conceptual continuation.) A special case of this would be
1177 ;;; when all of the outside calls are tail recursive.
1204 t)
1206 nil))))))