| blob | 31e813631a276e921875d2e0b75bc22a2bda5022 |
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 ;; Stack analysis wants DX value generators to end their
49 ;; blocks. Uses of mupltiple used LVARs already end their blocks,
50 ;; so we just need to process used-once 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.
83 :next call
89 :mess-up entry
103 ;;; This function handles merging the tail sets if CALL is potentially
104 ;;; tail-recursive, and is a call to a function with a different
105 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
106 ;;; IR1 so as to place a local call in what might be a tail-recursive
107 ;;; context. Note that any call which returns its value to a RETURN is
108 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
109 ;;; might be optimized away.
110 ;;;
111 ;;; We destructively modify the set for the calling function to
112 ;;; represent both, and then change all the functions in callee's set
113 ;;; to reference the first. If we do merge, we reoptimize the
114 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
115 ;;; tail set type.
129 t)))))
131 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
132 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
133 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
134 ;;; the function in the REF node with the new function.
135 ;;;
136 ;;; We change the REF last, since changing the reference can trigger
137 ;;; LET conversion of the new function, but will only do so if the
138 ;;; call is local. Note that the replacement may trigger LET
139 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
140 ;;; with NEW-FUN before the substitution, since after the substitution
141 ;;; (and LET conversion), the call may no longer be recognizable as
142 ;;; tail-recursive.
156 \f
157 ;;;; external entry point creation
159 ;;; Return a LAMBDA form that can be used as the definition of the XEP
160 ;;; for FUN.
161 ;;;
162 ;;; If FUN is a LAMBDA, then we check the number of arguments
163 ;;; (conditional on policy) and call FUN with all the arguments.
164 ;;;
165 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
166 ;;; of supplied arguments by doing do an = test for each entry-point,
167 ;;; calling the entry with the appropriate prefix of the passed
168 ;;; arguments.
169 ;;;
170 ;;; If there is a &MORE arg, then there are a couple of optimizations
171 ;;; that we make (more for space than anything else):
172 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
173 ;;; no argument count error is possible.
174 ;;; -- We can omit the = clause for the last entry-point, allowing the
175 ;;; case of 0 more args to fall through to the more entry.
176 ;;;
177 ;;; We don't bother to policy conditionalize wrong arg errors in
178 ;;; optional dispatches, since the additional overhead is negligible
179 ;;; compared to the cost of everything else going on.
180 ;;;
181 ;;; Note that if policy indicates it, argument type declarations in
182 ;;; FUN will be verified. Since nothing is known about the type of the
183 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
184 ;;; are passed to the actual function.
207 ;; Force convertion of all entries
210 and n from min
214 ;; FIXME: Make sure that INDEX type distinguishes between
215 ;; target and host. (Probably just make the SB!XC:DEFTYPE
216 ;; different from CL:DEFTYPE.)
221 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
222 ;; deftransforms and lambda-conversion.
234 ;;; Make an external entry point (XEP) for FUN and return it. We
235 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
236 ;;; then associate this lambda with FUN as its XEP. After the
237 ;;; conversion, we iterate over the function's associated lambdas,
238 ;;; redoing local call analysis so that the XEP calls will get
239 ;;; converted.
240 ;;;
241 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
242 ;;; discover an XEP after the initial local call analyze pass.
264 res)))
266 ;;; Notice a REF that is not in a local-call context. If the REF is
267 ;;; already to an XEP, then do nothing, otherwise change it to the
268 ;;; XEP, making an XEP if necessary.
269 ;;;
270 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
271 ;;; it as though it was not an XEP reference (i.e. leave it alone).
279 \f
280 ;;; Attempt to convert all references to FUN to local calls. The
281 ;;; reference must be the function for a call, and the function lvar
282 ;;; must be used only once, since otherwise we cannot be sure what
283 ;;; function is to be called. The call lvar would be multiply used if
284 ;;; there is hairy stuff such as conditionals in the expression that
285 ;;; computes the function.
286 ;;;
287 ;;; If we cannot convert a reference, then we mark the referenced
288 ;;; function as an entry-point, creating a new XEP if necessary. We
289 ;;; don't try to convert calls that are in error (:ERROR kind.)
290 ;;;
291 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
292 ;;; can force analysis of newly introduced calls. Note that we don't
293 ;;; do LET conversion here.
318 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
319 ;;; calls into local calls when it is legal. We also attempt to
320 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
321 ;;; deletion of a function reference, but functions that start out
322 ;;; eligible for conversion must be noticed sometime.
323 ;;;
324 ;;; Note that there is a lot of action going on behind the scenes
325 ;;; here, triggered by reference deletion. In particular, the
326 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
327 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
328 ;;; the COMPONENT-LAMBDAS when it is. Also, the
329 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
330 ;;; it is not updated when we delete functions, etc. Only
331 ;;; COMPONENT-LAMBDAS is updated.
332 ;;;
333 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
334 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
335 ;;; LAMBDAS.
353 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
355 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
356 ;; apply: no-op.
359 ; NEW-FUNCTIONALS, hence is new.
360 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
365 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
367 component)))))
378 ;; The original CMU CL code seemed to implicitly assume that
379 ;; COMPONENT is the only one here. Let's make that explicit.
390 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
391 ;;; to be in an infinite recursive loop, then change the reference to
392 ;;; reference a fresh copy. We return whichever function we decide to
393 ;;; reference.
411 original-functional)))))))
416 calls this LET-converted local function:~
417 ~% ~S"
421 expand because expansion refers to ~
422 the optimized away object ~S."
423 losing-local-object)))
430 original-functional)
433 converted-lambda))))
434 original-functional))
436 ;;; Dispatch to the appropriate function to attempt to convert a call.
437 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
438 ;;; optimization as well as in local call analysis. If the call is is
439 ;;; already :LOCAL, we do nothing. If the call is already scheduled
440 ;;; for deletion, also do nothing (in addition to saving time, this
441 ;;; also avoids some problems with optimizing collections of functions
442 ;;; that are partially deleted.)
443 ;;;
444 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
445 ;;; called on a :INITIAL component, we don't care whether the caller
446 ;;; and callee are in the same component. Afterward, we must stick
447 ;;; with whatever component division we have chosen.
448 ;;;
449 ;;; Before attempting to convert a call, we see whether the function
450 ;;; is supposed to be inline expanded. Call conversion proceeds as
451 ;;; before after any expansion.
452 ;;;
453 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
454 ;;; warnings will get the right context.
469 component))))
472 original-fun))
490 ;;; Attempt to convert a multiple-value call. The only interesting
491 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
492 ;;; exactly one reference and no XEP, and is called with one values
493 ;;; lvar.
494 ;;;
495 ;;; We change the call to be to the last optional entry point and
496 ;;; change the call to be local. Due to our preconditions, the call
497 ;;; should eventually be converted to a let, but we can't do that now,
498 ;;; since there may be stray references to the e-p lambda due to
499 ;;; optional defaulting code.
500 ;;;
501 ;;; We also use variable types for the called function to construct an
502 ;;; assertion for the values lvar.
503 ;;;
504 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
527 ;;; Attempt to convert a call to a lambda. If the number of args is
528 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
529 ;;; from future consideration. If the argcount is O.K. then we just
530 ;;; convert it.
539 'local-argument-mismatch
540 :format-control
541 "function called with ~R argument~:P, but wants exactly ~R"
544 \f
545 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
547 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
548 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
549 ;;; a call to the correct entry point. If &KEY args are supplied, then
550 ;;; dispatch to a subfunction. We don't convert calls to functions
551 ;;; that have a &MORE (or &REST) arg.
560 'local-argument-mismatch
561 :format-control
562 "function called with ~R argument~:P, but wants at least ~R"
574 'local-argument-mismatch
575 :format-control
576 "function called with ~R argument~:P, but wants at most ~R"
577 :format-arguments
582 ;;; This function is used to convert a call to an entry point when
583 ;;; complex transformations need to be done on the original arguments.
584 ;;; ENTRY is the entry point function that we are calling. VARS is a
585 ;;; list of variable names which are bound to the original call
586 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
587 ;;; is the list of arguments to the entry point function.
588 ;;;
589 ;;; In order to avoid gruesome graph grovelling, we introduce a new
590 ;;; function that rearranges the arguments and calls the entry point.
591 ;;; We analyze the new function and the entry point immediately so
592 ;;; that everything gets converted during the single pass.
609 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
610 ;;; function into a local call to the MAIN-ENTRY.
611 ;;;
612 ;;; First we verify that all keywords are constant and legal. If there
613 ;;; aren't, then we warn the user and don't attempt to convert the call.
614 ;;;
615 ;;; We massage the supplied &KEY arguments into the order expected
616 ;;; by the main entry. This is done by binding all the arguments to
617 ;;; the keyword call to variables in the introduced lambda, then
618 ;;; passing these values variables in the correct order when calling
619 ;;; the main entry. Unused arguments (such as the keywords themselves)
620 ;;; are discarded simply by not passing them along.
621 ;;;
622 ;;; If there is a &REST arg, then we bundle up the args and pass them
623 ;;; to LIST.
645 (:keyword
708 (:optional
712 (:rest
715 (:keyword
733 \f
734 ;;;; LET conversion
735 ;;;;
736 ;;;; Converting to a LET has differing significance to various parts
737 ;;;; of the compiler:
738 ;;;; -- The body of a LET is spliced in immediately after the
739 ;;;; corresponding combination node, making the control transfer
740 ;;;; explicit and allowing LETs to be mashed together into a single
741 ;;;; block. The value of the LET is delivered directly to the
742 ;;;; original lvar for the call, eliminating the need to
743 ;;;; propagate information from the dummy result lvar.
744 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
745 ;;;; that there is only one expression that the variable can be bound
746 ;;;; to, and this is easily substituted for.
747 ;;;; -- LETs are interesting to environment analysis and to the back
748 ;;;; end because in most ways a LET can be considered to be "the
749 ;;;; same function" as its home function.
750 ;;;; -- LET conversion has dynamic scope implications, since control
751 ;;;; transfers within the same environment are local. In a local
752 ;;;; control transfer, cleanup code must be emitted to remove
753 ;;;; dynamic bindings that are no longer in effect.
755 ;;; Set up the control transfer to the called CLAMBDA. We split the
756 ;;; call block immediately after the call, and link the head of
757 ;;; CLAMBDA to the call block. The successor block after splitting
758 ;;; (where we return to) is returned.
759 ;;;
760 ;;; If the lambda is is a different component than the call, then we
761 ;;; call JOIN-COMPONENTS. This only happens in block compilation
762 ;;; before FIND-INITIAL-DFO.
779 next-block)))
781 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
782 ;;; same set as; but leave CLAMBDA with a valid tail set value of
783 ;;; its own, for the benefit of code which might try to pull
784 ;;; something out of it (e.g. return type).
786 ;; Until sbcl-0.pre7.37.flaky5.2, we did
787 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
788 ;; (SETF (TAIL-SET-FUNS TAILS)
789 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
790 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
791 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
792 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
793 ;; inconsistent, and perhaps unsafe, for us to think we're in the
794 ;; tail set. Unfortunately..
795 ;;
796 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
797 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
798 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
799 ;; which you called in order to get the address of an external LAMBDA):
800 ;; the external function was defined in terms of internal function,
801 ;; which was LET-converted, and then things blew up downstream when
802 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
803 ;; the now-NILed-out TAIL-SET. So..
804 ;;
805 ;; To deal with this problem, we no longer NIL out
806 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
807 ;; * If we're the only function in TAIL-SET-FUNS, it should
808 ;; be safe to leave ourself linked to it, and it to you.
809 ;; * If there are other functions in TAIL-SET-FUNS, then we're
810 ;; afraid of future optimizations on those functions causing
811 ;; the TAIL-SET object no longer to be valid to describe our
812 ;; return value. Thus, we delete ourselves from that object;
813 ;; but we save a newly-allocated tail-set, derived from the old
814 ;; one, for ourselves, for the use of later code (e.g.
815 ;; FINALIZE-XEP-DEFINITION) which might want to
816 ;; know about our return type.
825 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
826 ;; remain valid in this case, so we nuke it on the theory that
827 ;; missing information tends to be less dangerous than incorrect
828 ;; information.
831 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
832 ;;; its LETs to LETs for the CALL's home function. We merge the calls
833 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
834 ;;; in the process. We also merge the ENTRIES.
835 ;;;
836 ;;; We also unlink the function head from the component head and set
837 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
838 ;;; recomputed.
858 ;; CLAMBDA belongs to HOME now.
868 ;; All of CLAMBDA's LETs belong to HOME now.
874 ;; CLAMBDA no longer has an independent existence as an entity
875 ;; which has LETs.
878 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
879 ;; DFO dependencies.
883 ;; CLAMBDA no longer has an independent existence as an entity
884 ;; which calls things or has DFO dependencies.
887 ;; All of CLAMBDA's ENTRIES belong to HOME now.
891 ;; CLAMBDA no longer has an independent existence as an entity
892 ;; with ENTRIES.
897 ;;; Handle the value semantics of LET conversion. Delete FUN's return
898 ;;; node, and change the control flow to transfer to NEXT-BLOCK
899 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
918 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
925 ;;; We are converting FUN to be a LET when the call is in a non-tail
926 ;;; position. Any previously tail calls in FUN are no longer tail
927 ;;; calls, and must be restored to normal calls which transfer to
928 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
929 ;;; the RETURN-RESULT, because the return might have been deleted (if
930 ;;; all calls were TR.)
949 ;; The called function might be an assignment in the
950 ;; case where we are currently converting that function.
951 ;; In steady-state, assignments never appear as a called
952 ;; function.
953 (:assignment
957 ;;; Deal with returning from a LET or assignment that we are
958 ;;; converting. FUN is the function we are calling, CALL is a call to
959 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
960 ;;; NULL if call is a tail call.
961 ;;;
962 ;;; If the call is not a tail call, then we must do
963 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
964 ;;; its value out of the enclosing non-let function. When call is
965 ;;; non-TR, we must convert it back to an ordinary local call, since
966 ;;; the value must be delivered to the receiver of CALL's value.
967 ;;;
968 ;;; We do different things depending on whether the caller and callee
969 ;;; have returns left:
971 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
972 ;;; Either the function doesn't return, or all returns are via
973 ;;; tail-recursive local calls.
974 ;;; -- If CALL is a non-tail call, or if both have returns, then
975 ;;; we delete the callee's return, move its uses to the call's
976 ;;; result lvar, and transfer control to the appropriate
977 ;;; return point.
978 ;;; -- If the callee has a return, but the caller doesn't, then we
979 ;;; move the return to the caller.
1008 ;;; Actually do LET conversion. We call subfunctions to do most of the
1009 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1010 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1011 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1012 ;;; new references to it.
1017 nil
1018 next-block)))
1022 ;; If CALL has a derive type NIL, it means that "its return" is
1023 ;; unreachable, but the next BIND is still reachable; in order to
1024 ;; not confuse MAYBE-TERMINATE-BLOCK...
1027 ;;; Reoptimize all of CALL's args and its result.
1036 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1037 ;;; LET converted?
1039 ;; From the user's point of view, LET-converting something that
1040 ;; has a name is inlining it. (The user can't see what we're doing
1041 ;; with anonymous things, and suppressing inlining
1042 ;; for such things can easily give Python acute indigestion, so
1043 ;; we don't.)
1045 ;; ANSI requires that explicit NOTINLINE be respected.
1047 ;; If (= LET-CONVERSION 0) we can guess that inlining
1048 ;; generally won't be appreciated, but if the user
1049 ;; specifically requests inlining, that takes precedence over
1050 ;; our general guess.
1054 ;;; We also don't convert calls to named functions which appear in the
1055 ;;; initial component, delaying this until optimization. This
1056 ;;; minimizes the likelihood that we will LET-convert a function which
1057 ;;; may have references added due to later local inline expansion.
1064 ;;; This function is called when there is some reason to believe that
1065 ;;; CLAMBDA might be converted into a LET. This is done after local
1066 ;;; call analysis, and also when a reference is deleted. We return
1067 ;;; true if we converted.
1072 ;; We only convert to a LET when the function is a normal local
1073 ;; function, has no XEP, and is referenced in exactly one local
1074 ;; call. Conversion is also inhibited if the only reference is in
1075 ;; a block about to be deleted.
1076 ;;
1077 ;; These rules limiting LET conversion may seem unnecessarily
1078 ;; restrictive, since there are some cases where we could do the
1079 ;; return with a jump that don't satisfy these requirements. The
1080 ;; reason for doing things this way is that it makes the concept
1081 ;; of a LET much more useful at the level of IR1 semantics. The
1082 ;; :ASSIGNMENT function kind provides another way to optimize
1083 ;; calls to single-return/multiple call functions.
1084 ;;
1085 ;; We don't attempt to convert calls to functions that have an
1086 ;; XEP, since we might be embarrassed later when we want to
1087 ;; convert a newly discovered local call. Also, see
1088 ;; OK-INITIAL-CONVERT-P.
1105 nil)))
1114 t))))
1115 \f
1116 ;;;; tail local calls and assignments
1118 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1119 ;;; they definitely won't generate any cleanup code. Currently we
1120 ;;; recognize lexical entry points that are only used locally (if at
1121 ;;; all).
1135 ;;; If a potentially TR local call really is TR, then convert it to
1136 ;;; jump directly to the called function. We also call
1137 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1138 ;;; tail-convert. The second is the value of M-C-T-A.
1145 ;; this is a tail call
1149 ;; If the call is in an XEP, we might decide to make it
1150 ;; non-tail so that we can use known return inside the
1151 ;; component.
1163 ;;; This is called when we believe it might make sense to convert
1164 ;;; CLAMBDA to an assignment. All this function really does is
1165 ;;; determine when a function with more than one call can still be
1166 ;;; combined with the calling function's environment. We can convert
1167 ;;; when:
1168 ;;; -- The function is a normal, non-entry function, and
1169 ;;; -- Except for one call, all calls must be tail recursive calls
1170 ;;; in the called function (i.e. are self-recursive tail calls)
1171 ;;; -- OK-INITIAL-CONVERT-P is true.
1172 ;;;
1173 ;;; There may be one outside call, and it need not be tail-recursive.
1174 ;;; Since all tail local calls have already been converted to direct
1175 ;;; transfers, the only control semantics needed are to splice in the
1176 ;;; body at the non-tail call. If there is no non-tail call, then we
1177 ;;; need only merge the environments. Both cases are handled by
1178 ;;; LET-CONVERT.
1179 ;;;
1180 ;;; ### It would actually be possible to allow any number of outside
1181 ;;; calls as long as they all return to the same place (i.e. have the
1182 ;;; same conceptual continuation.) A special case of this would be
1183 ;;; when all of the outside calls are tail recursive.
1210 t)
1212 nil))))))