| blob | d3eadd8163e2d3bd7a3b5342d86192f1fa36fa72 |
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 must be a regular
54 ;; combination, and its arguments are potentially DX as well.
78 :next call
84 :mess-up entry
98 ;;; This function handles merging the tail sets if CALL is potentially
99 ;;; tail-recursive, and is a call to a function with a different
100 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
101 ;;; IR1 so as to place a local call in what might be a tail-recursive
102 ;;; context. Note that any call which returns its value to a RETURN is
103 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
104 ;;; might be optimized away.
105 ;;;
106 ;;; We destructively modify the set for the calling function to
107 ;;; represent both, and then change all the functions in callee's set
108 ;;; to reference the first. If we do merge, we reoptimize the
109 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
110 ;;; tail set type.
124 t)))))
126 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
127 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
128 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
129 ;;; the function in the REF node with the new function.
130 ;;;
131 ;;; We change the REF last, since changing the reference can trigger
132 ;;; LET conversion of the new function, but will only do so if the
133 ;;; call is local. Note that the replacement may trigger LET
134 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
135 ;;; with NEW-FUN before the substitution, since after the substitution
136 ;;; (and LET conversion), the call may no longer be recognizable as
137 ;;; tail-recursive.
151 \f
152 ;;;; external entry point creation
154 ;;; Return a LAMBDA form that can be used as the definition of the XEP
155 ;;; for FUN.
156 ;;;
157 ;;; If FUN is a LAMBDA, then we check the number of arguments
158 ;;; (conditional on policy) and call FUN with all the arguments.
159 ;;;
160 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
161 ;;; of supplied arguments by doing do an = test for each entry-point,
162 ;;; calling the entry with the appropriate prefix of the passed
163 ;;; arguments.
164 ;;;
165 ;;; If there is a &MORE arg, then there are a couple of optimizations
166 ;;; that we make (more for space than anything else):
167 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
168 ;;; no argument count error is possible.
169 ;;; -- We can omit the = clause for the last entry-point, allowing the
170 ;;; case of 0 more args to fall through to the more entry.
171 ;;;
172 ;;; We don't bother to policy conditionalize wrong arg errors in
173 ;;; optional dispatches, since the additional overhead is negligible
174 ;;; compared to the cost of everything else going on.
175 ;;;
176 ;;; Note that if policy indicates it, argument type declarations in
177 ;;; FUN will be verified. Since nothing is known about the type of the
178 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
179 ;;; are passed to the actual function.
202 ;; Force convertion of all entries
205 and n from min
209 ;; FIXME: Make sure that INDEX type distinguishes between
210 ;; target and host. (Probably just make the SB!XC:DEFTYPE
211 ;; different from CL:DEFTYPE.)
216 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
217 ;; deftransforms and lambda-conversion.
229 ;;; Make an external entry point (XEP) for FUN and return it. We
230 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
231 ;;; then associate this lambda with FUN as its XEP. After the
232 ;;; conversion, we iterate over the function's associated lambdas,
233 ;;; redoing local call analysis so that the XEP calls will get
234 ;;; converted.
235 ;;;
236 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
237 ;;; discover an XEP after the initial local call analyze pass.
259 res)))
261 ;;; Notice a REF that is not in a local-call context. If the REF is
262 ;;; already to an XEP, then do nothing, otherwise change it to the
263 ;;; XEP, making an XEP if necessary.
264 ;;;
265 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
266 ;;; it as though it was not an XEP reference (i.e. leave it alone).
274 \f
275 ;;; Attempt to convert all references to FUN to local calls. The
276 ;;; reference must be the function for a call, and the function lvar
277 ;;; must be used only once, since otherwise we cannot be sure what
278 ;;; function is to be called. The call lvar would be multiply used if
279 ;;; there is hairy stuff such as conditionals in the expression that
280 ;;; computes the function.
281 ;;;
282 ;;; If we cannot convert a reference, then we mark the referenced
283 ;;; function as an entry-point, creating a new XEP if necessary. We
284 ;;; don't try to convert calls that are in error (:ERROR kind.)
285 ;;;
286 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
287 ;;; can force analysis of newly introduced calls. Note that we don't
288 ;;; do LET conversion here.
313 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
314 ;;; calls into local calls when it is legal. We also attempt to
315 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
316 ;;; deletion of a function reference, but functions that start out
317 ;;; eligible for conversion must be noticed sometime.
318 ;;;
319 ;;; Note that there is a lot of action going on behind the scenes
320 ;;; here, triggered by reference deletion. In particular, the
321 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
322 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
323 ;;; the COMPONENT-LAMBDAS when it is. Also, the
324 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
325 ;;; it is not updated when we delete functions, etc. Only
326 ;;; COMPONENT-LAMBDAS is updated.
327 ;;;
328 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
329 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
330 ;;; LAMBDAS.
348 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
350 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
351 ;; apply: no-op.
354 ; NEW-FUNCTIONALS, hence is new.
355 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
360 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
362 component)))))
373 ;; The original CMU CL code seemed to implicitly assume that
374 ;; COMPONENT is the only one here. Let's make that explicit.
385 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
386 ;;; to be in an infinite recursive loop, then change the reference to
387 ;;; reference a fresh copy. We return whichever function we decide to
388 ;;; reference.
406 original-functional)))))))
411 calls this LET-converted local function:~
412 ~% ~S"
416 expand because expansion refers to ~
417 the optimized away object ~S."
418 losing-local-object)))
425 original-functional)
428 converted-lambda))))
429 original-functional))
431 ;;; Dispatch to the appropriate function to attempt to convert a call.
432 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
433 ;;; optimization as well as in local call analysis. If the call is is
434 ;;; already :LOCAL, we do nothing. If the call is already scheduled
435 ;;; for deletion, also do nothing (in addition to saving time, this
436 ;;; also avoids some problems with optimizing collections of functions
437 ;;; that are partially deleted.)
438 ;;;
439 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
440 ;;; called on a :INITIAL component, we don't care whether the caller
441 ;;; and callee are in the same component. Afterward, we must stick
442 ;;; with whatever component division we have chosen.
443 ;;;
444 ;;; Before attempting to convert a call, we see whether the function
445 ;;; is supposed to be inline expanded. Call conversion proceeds as
446 ;;; before after any expansion.
447 ;;;
448 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
449 ;;; warnings will get the right context.
464 component))))
467 original-fun))
485 ;;; Attempt to convert a multiple-value call. The only interesting
486 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
487 ;;; exactly one reference and no XEP, and is called with one values
488 ;;; lvar.
489 ;;;
490 ;;; We change the call to be to the last optional entry point and
491 ;;; change the call to be local. Due to our preconditions, the call
492 ;;; should eventually be converted to a let, but we can't do that now,
493 ;;; since there may be stray references to the e-p lambda due to
494 ;;; optional defaulting code.
495 ;;;
496 ;;; We also use variable types for the called function to construct an
497 ;;; assertion for the values lvar.
498 ;;;
499 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
522 ;;; Attempt to convert a call to a lambda. If the number of args is
523 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
524 ;;; from future consideration. If the argcount is O.K. then we just
525 ;;; convert it.
534 'local-argument-mismatch
535 :format-control
536 "function called with ~R argument~:P, but wants exactly ~R"
539 \f
540 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
542 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
543 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
544 ;;; a call to the correct entry point. If &KEY args are supplied, then
545 ;;; dispatch to a subfunction. We don't convert calls to functions
546 ;;; that have a &MORE (or &REST) arg.
555 'local-argument-mismatch
556 :format-control
557 "function called with ~R argument~:P, but wants at least ~R"
569 'local-argument-mismatch
570 :format-control
571 "function called with ~R argument~:P, but wants at most ~R"
572 :format-arguments
577 ;;; This function is used to convert a call to an entry point when
578 ;;; complex transformations need to be done on the original arguments.
579 ;;; ENTRY is the entry point function that we are calling. VARS is a
580 ;;; list of variable names which are bound to the original call
581 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
582 ;;; is the list of arguments to the entry point function.
583 ;;;
584 ;;; In order to avoid gruesome graph grovelling, we introduce a new
585 ;;; function that rearranges the arguments and calls the entry point.
586 ;;; We analyze the new function and the entry point immediately so
587 ;;; that everything gets converted during the single pass.
604 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
605 ;;; function into a local call to the MAIN-ENTRY.
606 ;;;
607 ;;; First we verify that all keywords are constant and legal. If there
608 ;;; aren't, then we warn the user and don't attempt to convert the call.
609 ;;;
610 ;;; We massage the supplied &KEY arguments into the order expected
611 ;;; by the main entry. This is done by binding all the arguments to
612 ;;; the keyword call to variables in the introduced lambda, then
613 ;;; passing these values variables in the correct order when calling
614 ;;; the main entry. Unused arguments (such as the keywords themselves)
615 ;;; are discarded simply by not passing them along.
616 ;;;
617 ;;; If there is a &REST arg, then we bundle up the args and pass them
618 ;;; to LIST.
640 (:keyword
703 (:optional
707 (:rest
710 (:keyword
728 \f
729 ;;;; LET conversion
730 ;;;;
731 ;;;; Converting to a LET has differing significance to various parts
732 ;;;; of the compiler:
733 ;;;; -- The body of a LET is spliced in immediately after the
734 ;;;; corresponding combination node, making the control transfer
735 ;;;; explicit and allowing LETs to be mashed together into a single
736 ;;;; block. The value of the LET is delivered directly to the
737 ;;;; original lvar for the call, eliminating the need to
738 ;;;; propagate information from the dummy result lvar.
739 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
740 ;;;; that there is only one expression that the variable can be bound
741 ;;;; to, and this is easily substituted for.
742 ;;;; -- LETs are interesting to environment analysis and to the back
743 ;;;; end because in most ways a LET can be considered to be "the
744 ;;;; same function" as its home function.
745 ;;;; -- LET conversion has dynamic scope implications, since control
746 ;;;; transfers within the same environment are local. In a local
747 ;;;; control transfer, cleanup code must be emitted to remove
748 ;;;; dynamic bindings that are no longer in effect.
750 ;;; Set up the control transfer to the called CLAMBDA. We split the
751 ;;; call block immediately after the call, and link the head of
752 ;;; CLAMBDA to the call block. The successor block after splitting
753 ;;; (where we return to) is returned.
754 ;;;
755 ;;; If the lambda is is a different component than the call, then we
756 ;;; call JOIN-COMPONENTS. This only happens in block compilation
757 ;;; before FIND-INITIAL-DFO.
774 next-block)))
776 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
777 ;;; same set as; but leave CLAMBDA with a valid tail set value of
778 ;;; its own, for the benefit of code which might try to pull
779 ;;; something out of it (e.g. return type).
781 ;; Until sbcl-0.pre7.37.flaky5.2, we did
782 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
783 ;; (SETF (TAIL-SET-FUNS TAILS)
784 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
785 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
786 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
787 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
788 ;; inconsistent, and perhaps unsafe, for us to think we're in the
789 ;; tail set. Unfortunately..
790 ;;
791 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
792 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
793 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
794 ;; which you called in order to get the address of an external LAMBDA):
795 ;; the external function was defined in terms of internal function,
796 ;; which was LET-converted, and then things blew up downstream when
797 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
798 ;; the now-NILed-out TAIL-SET. So..
799 ;;
800 ;; To deal with this problem, we no longer NIL out
801 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
802 ;; * If we're the only function in TAIL-SET-FUNS, it should
803 ;; be safe to leave ourself linked to it, and it to you.
804 ;; * If there are other functions in TAIL-SET-FUNS, then we're
805 ;; afraid of future optimizations on those functions causing
806 ;; the TAIL-SET object no longer to be valid to describe our
807 ;; return value. Thus, we delete ourselves from that object;
808 ;; but we save a newly-allocated tail-set, derived from the old
809 ;; one, for ourselves, for the use of later code (e.g.
810 ;; FINALIZE-XEP-DEFINITION) which might want to
811 ;; know about our return type.
820 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
821 ;; remain valid in this case, so we nuke it on the theory that
822 ;; missing information tends to be less dangerous than incorrect
823 ;; information.
826 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
827 ;;; its LETs to LETs for the CALL's home function. We merge the calls
828 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
829 ;;; in the process. We also merge the ENTRIES.
830 ;;;
831 ;;; We also unlink the function head from the component head and set
832 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
833 ;;; recomputed.
853 ;; CLAMBDA belongs to HOME now.
863 ;; All of CLAMBDA's LETs belong to HOME now.
869 ;; CLAMBDA no longer has an independent existence as an entity
870 ;; which has LETs.
873 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
874 ;; DFO dependencies.
878 ;; CLAMBDA no longer has an independent existence as an entity
879 ;; which calls things or has DFO dependencies.
882 ;; All of CLAMBDA's ENTRIES belong to HOME now.
886 ;; CLAMBDA no longer has an independent existence as an entity
887 ;; with ENTRIES.
892 ;;; Handle the value semantics of LET conversion. Delete FUN's return
893 ;;; node, and change the control flow to transfer to NEXT-BLOCK
894 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
913 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
920 ;;; We are converting FUN to be a LET when the call is in a non-tail
921 ;;; position. Any previously tail calls in FUN are no longer tail
922 ;;; calls, and must be restored to normal calls which transfer to
923 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
924 ;;; the RETURN-RESULT, because the return might have been deleted (if
925 ;;; all calls were TR.)
944 ;; The called function might be an assignment in the
945 ;; case where we are currently converting that function.
946 ;; In steady-state, assignments never appear as a called
947 ;; function.
948 (:assignment
952 ;;; Deal with returning from a LET or assignment that we are
953 ;;; converting. FUN is the function we are calling, CALL is a call to
954 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
955 ;;; NULL if call is a tail call.
956 ;;;
957 ;;; If the call is not a tail call, then we must do
958 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
959 ;;; its value out of the enclosing non-let function. When call is
960 ;;; non-TR, we must convert it back to an ordinary local call, since
961 ;;; the value must be delivered to the receiver of CALL's value.
962 ;;;
963 ;;; We do different things depending on whether the caller and callee
964 ;;; have returns left:
966 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
967 ;;; Either the function doesn't return, or all returns are via
968 ;;; tail-recursive local calls.
969 ;;; -- If CALL is a non-tail call, or if both have returns, then
970 ;;; we delete the callee's return, move its uses to the call's
971 ;;; result lvar, and transfer control to the appropriate
972 ;;; return point.
973 ;;; -- If the callee has a return, but the caller doesn't, then we
974 ;;; move the return to the caller.
1003 ;;; Actually do LET conversion. We call subfunctions to do most of the
1004 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1005 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1006 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1007 ;;; new references to it.
1012 nil
1013 next-block)))
1017 ;; If CALL has a derive type NIL, it means that "its return" is
1018 ;; unreachable, but the next BIND is still reachable; in order to
1019 ;; not confuse MAYBE-TERMINATE-BLOCK...
1022 ;;; Reoptimize all of CALL's args and its result.
1031 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1032 ;;; LET converted?
1034 ;; From the user's point of view, LET-converting something that
1035 ;; has a name is inlining it. (The user can't see what we're doing
1036 ;; with anonymous things, and suppressing inlining
1037 ;; for such things can easily give Python acute indigestion, so
1038 ;; we don't.)
1040 ;; ANSI requires that explicit NOTINLINE be respected.
1042 ;; If (= LET-CONVERSION 0) we can guess that inlining
1043 ;; generally won't be appreciated, but if the user
1044 ;; specifically requests inlining, that takes precedence over
1045 ;; our general guess.
1049 ;;; We also don't convert calls to named functions which appear in the
1050 ;;; initial component, delaying this until optimization. This
1051 ;;; minimizes the likelihood that we will LET-convert a function which
1052 ;;; may have references added due to later local inline expansion.
1059 ;;; This function is called when there is some reason to believe that
1060 ;;; CLAMBDA might be converted into a LET. This is done after local
1061 ;;; call analysis, and also when a reference is deleted. We return
1062 ;;; true if we converted.
1067 ;; We only convert to a LET when the function is a normal local
1068 ;; function, has no XEP, and is referenced in exactly one local
1069 ;; call. Conversion is also inhibited if the only reference is in
1070 ;; a block about to be deleted.
1071 ;;
1072 ;; These rules limiting LET conversion may seem unnecessarily
1073 ;; restrictive, since there are some cases where we could do the
1074 ;; return with a jump that don't satisfy these requirements. The
1075 ;; reason for doing things this way is that it makes the concept
1076 ;; of a LET much more useful at the level of IR1 semantics. The
1077 ;; :ASSIGNMENT function kind provides another way to optimize
1078 ;; calls to single-return/multiple call functions.
1079 ;;
1080 ;; We don't attempt to convert calls to functions that have an
1081 ;; XEP, since we might be embarrassed later when we want to
1082 ;; convert a newly discovered local call. Also, see
1083 ;; OK-INITIAL-CONVERT-P.
1100 nil)))
1109 t))))
1110 \f
1111 ;;;; tail local calls and assignments
1113 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1114 ;;; they definitely won't generate any cleanup code. Currently we
1115 ;;; recognize lexical entry points that are only used locally (if at
1116 ;;; all).
1130 ;;; If a potentially TR local call really is TR, then convert it to
1131 ;;; jump directly to the called function. We also call
1132 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1133 ;;; tail-convert. The second is the value of M-C-T-A.
1140 ;; this is a tail call
1144 ;; If the call is in an XEP, we might decide to make it
1145 ;; non-tail so that we can use known return inside the
1146 ;; component.
1158 ;;; This is called when we believe it might make sense to convert
1159 ;;; CLAMBDA to an assignment. All this function really does is
1160 ;;; determine when a function with more than one call can still be
1161 ;;; combined with the calling function's environment. We can convert
1162 ;;; when:
1163 ;;; -- The function is a normal, non-entry function, and
1164 ;;; -- Except for one call, all calls must be tail recursive calls
1165 ;;; in the called function (i.e. are self-recursive tail calls)
1166 ;;; -- OK-INITIAL-CONVERT-P is true.
1167 ;;;
1168 ;;; There may be one outside call, and it need not be tail-recursive.
1169 ;;; Since all tail local calls have already been converted to direct
1170 ;;; transfers, the only control semantics needed are to splice in the
1171 ;;; body at the non-tail call. If there is no non-tail call, then we
1172 ;;; need only merge the environments. Both cases are handled by
1173 ;;; LET-CONVERT.
1174 ;;;
1175 ;;; ### It would actually be possible to allow any number of outside
1176 ;;; calls as long as they all return to the same place (i.e. have the
1177 ;;; same conceptual continuation.) A special case of this would be
1178 ;;; when all of the outside calls are tail recursive.
1205 t)
1207 nil))))))