| blob | 3b57132bbb0664e302213c3582d869499eac853d |
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.
79 :next call
85 :mess-up entry
99 ;;; This function handles merging the tail sets if CALL is potentially
100 ;;; tail-recursive, and is a call to a function with a different
101 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
102 ;;; IR1 so as to place a local call in what might be a tail-recursive
103 ;;; context. Note that any call which returns its value to a RETURN is
104 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
105 ;;; might be optimized away.
106 ;;;
107 ;;; We destructively modify the set for the calling function to
108 ;;; represent both, and then change all the functions in callee's set
109 ;;; to reference the first. If we do merge, we reoptimize the
110 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
111 ;;; tail set type.
125 t)))))
127 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
128 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
129 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
130 ;;; the function in the REF node with the new function.
131 ;;;
132 ;;; We change the REF last, since changing the reference can trigger
133 ;;; LET conversion of the new function, but will only do so if the
134 ;;; call is local. Note that the replacement may trigger LET
135 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
136 ;;; with NEW-FUN before the substitution, since after the substitution
137 ;;; (and LET conversion), the call may no longer be recognizable as
138 ;;; tail-recursive.
152 \f
153 ;;;; external entry point creation
155 ;;; Return a LAMBDA form that can be used as the definition of the XEP
156 ;;; for FUN.
157 ;;;
158 ;;; If FUN is a LAMBDA, then we check the number of arguments
159 ;;; (conditional on policy) and call FUN with all the arguments.
160 ;;;
161 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
162 ;;; of supplied arguments by doing do an = test for each entry-point,
163 ;;; calling the entry with the appropriate prefix of the passed
164 ;;; arguments.
165 ;;;
166 ;;; If there is a &MORE arg, then there are a couple of optimizations
167 ;;; that we make (more for space than anything else):
168 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
169 ;;; no argument count error is possible.
170 ;;; -- We can omit the = clause for the last entry-point, allowing the
171 ;;; case of 0 more args to fall through to the more entry.
172 ;;;
173 ;;; We don't bother to policy conditionalize wrong arg errors in
174 ;;; optional dispatches, since the additional overhead is negligible
175 ;;; compared to the cost of everything else going on.
176 ;;;
177 ;;; Note that if policy indicates it, argument type declarations in
178 ;;; FUN will be verified. Since nothing is known about the type of the
179 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
180 ;;; are passed to the actual function.
203 ;; Force convertion of all entries
206 and n from min
210 ;; FIXME: Make sure that INDEX type distinguishes between
211 ;; target and host. (Probably just make the SB!XC:DEFTYPE
212 ;; different from CL:DEFTYPE.)
217 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
218 ;; deftransforms and lambda-conversion.
230 ;;; Make an external entry point (XEP) for FUN and return it. We
231 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
232 ;;; then associate this lambda with FUN as its XEP. After the
233 ;;; conversion, we iterate over the function's associated lambdas,
234 ;;; redoing local call analysis so that the XEP calls will get
235 ;;; converted.
236 ;;;
237 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
238 ;;; discover an XEP after the initial local call analyze pass.
260 res)))
262 ;;; Notice a REF that is not in a local-call context. If the REF is
263 ;;; already to an XEP, then do nothing, otherwise change it to the
264 ;;; XEP, making an XEP if necessary.
265 ;;;
266 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
267 ;;; it as though it was not an XEP reference (i.e. leave it alone).
275 \f
276 ;;; Attempt to convert all references to FUN to local calls. The
277 ;;; reference must be the function for a call, and the function lvar
278 ;;; must be used only once, since otherwise we cannot be sure what
279 ;;; function is to be called. The call lvar would be multiply used if
280 ;;; there is hairy stuff such as conditionals in the expression that
281 ;;; computes the function.
282 ;;;
283 ;;; If we cannot convert a reference, then we mark the referenced
284 ;;; function as an entry-point, creating a new XEP if necessary. We
285 ;;; don't try to convert calls that are in error (:ERROR kind.)
286 ;;;
287 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
288 ;;; can force analysis of newly introduced calls. Note that we don't
289 ;;; do LET conversion here.
314 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
315 ;;; calls into local calls when it is legal. We also attempt to
316 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
317 ;;; deletion of a function reference, but functions that start out
318 ;;; eligible for conversion must be noticed sometime.
319 ;;;
320 ;;; Note that there is a lot of action going on behind the scenes
321 ;;; here, triggered by reference deletion. In particular, the
322 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
323 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
324 ;;; the COMPONENT-LAMBDAS when it is. Also, the
325 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
326 ;;; it is not updated when we delete functions, etc. Only
327 ;;; COMPONENT-LAMBDAS is updated.
328 ;;;
329 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
330 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
331 ;;; LAMBDAS.
349 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
351 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
352 ;; apply: no-op.
355 ; NEW-FUNCTIONALS, hence is new.
356 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
361 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
363 component)))))
374 ;; The original CMU CL code seemed to implicitly assume that
375 ;; COMPONENT is the only one here. Let's make that explicit.
386 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
387 ;;; to be in an infinite recursive loop, then change the reference to
388 ;;; reference a fresh copy. We return whichever function we decide to
389 ;;; reference.
407 original-functional)))))))
412 calls this LET-converted local function:~
413 ~% ~S"
417 expand because expansion refers to ~
418 the optimized away object ~S."
419 losing-local-object)))
426 original-functional)
429 converted-lambda))))
430 original-functional))
432 ;;; Dispatch to the appropriate function to attempt to convert a call.
433 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
434 ;;; optimization as well as in local call analysis. If the call is is
435 ;;; already :LOCAL, we do nothing. If the call is already scheduled
436 ;;; for deletion, also do nothing (in addition to saving time, this
437 ;;; also avoids some problems with optimizing collections of functions
438 ;;; that are partially deleted.)
439 ;;;
440 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
441 ;;; called on a :INITIAL component, we don't care whether the caller
442 ;;; and callee are in the same component. Afterward, we must stick
443 ;;; with whatever component division we have chosen.
444 ;;;
445 ;;; Before attempting to convert a call, we see whether the function
446 ;;; is supposed to be inline expanded. Call conversion proceeds as
447 ;;; before after any expansion.
448 ;;;
449 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
450 ;;; warnings will get the right context.
465 component))))
468 original-fun))
486 ;;; Attempt to convert a multiple-value call. The only interesting
487 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
488 ;;; exactly one reference and no XEP, and is called with one values
489 ;;; lvar.
490 ;;;
491 ;;; We change the call to be to the last optional entry point and
492 ;;; change the call to be local. Due to our preconditions, the call
493 ;;; should eventually be converted to a let, but we can't do that now,
494 ;;; since there may be stray references to the e-p lambda due to
495 ;;; optional defaulting code.
496 ;;;
497 ;;; We also use variable types for the called function to construct an
498 ;;; assertion for the values lvar.
499 ;;;
500 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
523 ;;; Attempt to convert a call to a lambda. If the number of args is
524 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
525 ;;; from future consideration. If the argcount is O.K. then we just
526 ;;; convert it.
535 'local-argument-mismatch
536 :format-control
537 "function called with ~R argument~:P, but wants exactly ~R"
540 \f
541 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
543 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
544 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
545 ;;; a call to the correct entry point. If &KEY args are supplied, then
546 ;;; dispatch to a subfunction. We don't convert calls to functions
547 ;;; that have a &MORE (or &REST) arg.
556 'local-argument-mismatch
557 :format-control
558 "function called with ~R argument~:P, but wants at least ~R"
570 'local-argument-mismatch
571 :format-control
572 "function called with ~R argument~:P, but wants at most ~R"
573 :format-arguments
578 ;;; This function is used to convert a call to an entry point when
579 ;;; complex transformations need to be done on the original arguments.
580 ;;; ENTRY is the entry point function that we are calling. VARS is a
581 ;;; list of variable names which are bound to the original call
582 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
583 ;;; is the list of arguments to the entry point function.
584 ;;;
585 ;;; In order to avoid gruesome graph grovelling, we introduce a new
586 ;;; function that rearranges the arguments and calls the entry point.
587 ;;; We analyze the new function and the entry point immediately so
588 ;;; that everything gets converted during the single pass.
605 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
606 ;;; function into a local call to the MAIN-ENTRY.
607 ;;;
608 ;;; First we verify that all keywords are constant and legal. If there
609 ;;; aren't, then we warn the user and don't attempt to convert the call.
610 ;;;
611 ;;; We massage the supplied &KEY arguments into the order expected
612 ;;; by the main entry. This is done by binding all the arguments to
613 ;;; the keyword call to variables in the introduced lambda, then
614 ;;; passing these values variables in the correct order when calling
615 ;;; the main entry. Unused arguments (such as the keywords themselves)
616 ;;; are discarded simply by not passing them along.
617 ;;;
618 ;;; If there is a &REST arg, then we bundle up the args and pass them
619 ;;; to LIST.
641 (:keyword
704 (:optional
708 (:rest
711 (:keyword
729 \f
730 ;;;; LET conversion
731 ;;;;
732 ;;;; Converting to a LET has differing significance to various parts
733 ;;;; of the compiler:
734 ;;;; -- The body of a LET is spliced in immediately after the
735 ;;;; corresponding combination node, making the control transfer
736 ;;;; explicit and allowing LETs to be mashed together into a single
737 ;;;; block. The value of the LET is delivered directly to the
738 ;;;; original lvar for the call, eliminating the need to
739 ;;;; propagate information from the dummy result lvar.
740 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
741 ;;;; that there is only one expression that the variable can be bound
742 ;;;; to, and this is easily substituted for.
743 ;;;; -- LETs are interesting to environment analysis and to the back
744 ;;;; end because in most ways a LET can be considered to be "the
745 ;;;; same function" as its home function.
746 ;;;; -- LET conversion has dynamic scope implications, since control
747 ;;;; transfers within the same environment are local. In a local
748 ;;;; control transfer, cleanup code must be emitted to remove
749 ;;;; dynamic bindings that are no longer in effect.
751 ;;; Set up the control transfer to the called CLAMBDA. We split the
752 ;;; call block immediately after the call, and link the head of
753 ;;; CLAMBDA to the call block. The successor block after splitting
754 ;;; (where we return to) is returned.
755 ;;;
756 ;;; If the lambda is is a different component than the call, then we
757 ;;; call JOIN-COMPONENTS. This only happens in block compilation
758 ;;; before FIND-INITIAL-DFO.
775 next-block)))
777 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
778 ;;; same set as; but leave CLAMBDA with a valid tail set value of
779 ;;; its own, for the benefit of code which might try to pull
780 ;;; something out of it (e.g. return type).
782 ;; Until sbcl-0.pre7.37.flaky5.2, we did
783 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
784 ;; (SETF (TAIL-SET-FUNS TAILS)
785 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
786 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
787 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
788 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
789 ;; inconsistent, and perhaps unsafe, for us to think we're in the
790 ;; tail set. Unfortunately..
791 ;;
792 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
793 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
794 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
795 ;; which you called in order to get the address of an external LAMBDA):
796 ;; the external function was defined in terms of internal function,
797 ;; which was LET-converted, and then things blew up downstream when
798 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
799 ;; the now-NILed-out TAIL-SET. So..
800 ;;
801 ;; To deal with this problem, we no longer NIL out
802 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
803 ;; * If we're the only function in TAIL-SET-FUNS, it should
804 ;; be safe to leave ourself linked to it, and it to you.
805 ;; * If there are other functions in TAIL-SET-FUNS, then we're
806 ;; afraid of future optimizations on those functions causing
807 ;; the TAIL-SET object no longer to be valid to describe our
808 ;; return value. Thus, we delete ourselves from that object;
809 ;; but we save a newly-allocated tail-set, derived from the old
810 ;; one, for ourselves, for the use of later code (e.g.
811 ;; FINALIZE-XEP-DEFINITION) which might want to
812 ;; know about our return type.
821 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
822 ;; remain valid in this case, so we nuke it on the theory that
823 ;; missing information tends to be less dangerous than incorrect
824 ;; information.
827 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
828 ;;; its LETs to LETs for the CALL's home function. We merge the calls
829 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
830 ;;; in the process. We also merge the ENTRIES.
831 ;;;
832 ;;; We also unlink the function head from the component head and set
833 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
834 ;;; recomputed.
854 ;; CLAMBDA belongs to HOME now.
864 ;; All of CLAMBDA's LETs belong to HOME now.
870 ;; CLAMBDA no longer has an independent existence as an entity
871 ;; which has LETs.
874 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
875 ;; DFO dependencies.
879 ;; CLAMBDA no longer has an independent existence as an entity
880 ;; which calls things or has DFO dependencies.
883 ;; All of CLAMBDA's ENTRIES belong to HOME now.
887 ;; CLAMBDA no longer has an independent existence as an entity
888 ;; with ENTRIES.
893 ;;; Handle the value semantics of LET conversion. Delete FUN's return
894 ;;; node, and change the control flow to transfer to NEXT-BLOCK
895 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
914 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
921 ;;; We are converting FUN to be a LET when the call is in a non-tail
922 ;;; position. Any previously tail calls in FUN are no longer tail
923 ;;; calls, and must be restored to normal calls which transfer to
924 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
925 ;;; the RETURN-RESULT, because the return might have been deleted (if
926 ;;; all calls were TR.)
945 ;; The called function might be an assignment in the
946 ;; case where we are currently converting that function.
947 ;; In steady-state, assignments never appear as a called
948 ;; function.
949 (:assignment
953 ;;; Deal with returning from a LET or assignment that we are
954 ;;; converting. FUN is the function we are calling, CALL is a call to
955 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
956 ;;; NULL if call is a tail call.
957 ;;;
958 ;;; If the call is not a tail call, then we must do
959 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
960 ;;; its value out of the enclosing non-let function. When call is
961 ;;; non-TR, we must convert it back to an ordinary local call, since
962 ;;; the value must be delivered to the receiver of CALL's value.
963 ;;;
964 ;;; We do different things depending on whether the caller and callee
965 ;;; have returns left:
967 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
968 ;;; Either the function doesn't return, or all returns are via
969 ;;; tail-recursive local calls.
970 ;;; -- If CALL is a non-tail call, or if both have returns, then
971 ;;; we delete the callee's return, move its uses to the call's
972 ;;; result lvar, and transfer control to the appropriate
973 ;;; return point.
974 ;;; -- If the callee has a return, but the caller doesn't, then we
975 ;;; move the return to the caller.
1004 ;;; Actually do LET conversion. We call subfunctions to do most of the
1005 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1006 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1007 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1008 ;;; new references to it.
1013 nil
1014 next-block)))
1018 ;; If CALL has a derive type NIL, it means that "its return" is
1019 ;; unreachable, but the next BIND is still reachable; in order to
1020 ;; not confuse MAYBE-TERMINATE-BLOCK...
1023 ;;; Reoptimize all of CALL's args and its result.
1032 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1033 ;;; LET converted?
1035 ;; From the user's point of view, LET-converting something that
1036 ;; has a name is inlining it. (The user can't see what we're doing
1037 ;; with anonymous things, and suppressing inlining
1038 ;; for such things can easily give Python acute indigestion, so
1039 ;; we don't.)
1041 ;; ANSI requires that explicit NOTINLINE be respected.
1043 ;; If (= LET-CONVERSION 0) we can guess that inlining
1044 ;; generally won't be appreciated, but if the user
1045 ;; specifically requests inlining, that takes precedence over
1046 ;; our general guess.
1050 ;;; We also don't convert calls to named functions which appear in the
1051 ;;; initial component, delaying this until optimization. This
1052 ;;; minimizes the likelihood that we will LET-convert a function which
1053 ;;; may have references added due to later local inline expansion.
1060 ;;; This function is called when there is some reason to believe that
1061 ;;; CLAMBDA might be converted into a LET. This is done after local
1062 ;;; call analysis, and also when a reference is deleted. We return
1063 ;;; true if we converted.
1068 ;; We only convert to a LET when the function is a normal local
1069 ;; function, has no XEP, and is referenced in exactly one local
1070 ;; call. Conversion is also inhibited if the only reference is in
1071 ;; a block about to be deleted.
1072 ;;
1073 ;; These rules limiting LET conversion may seem unnecessarily
1074 ;; restrictive, since there are some cases where we could do the
1075 ;; return with a jump that don't satisfy these requirements. The
1076 ;; reason for doing things this way is that it makes the concept
1077 ;; of a LET much more useful at the level of IR1 semantics. The
1078 ;; :ASSIGNMENT function kind provides another way to optimize
1079 ;; calls to single-return/multiple call functions.
1080 ;;
1081 ;; We don't attempt to convert calls to functions that have an
1082 ;; XEP, since we might be embarrassed later when we want to
1083 ;; convert a newly discovered local call. Also, see
1084 ;; OK-INITIAL-CONVERT-P.
1101 nil)))
1110 t))))
1111 \f
1112 ;;;; tail local calls and assignments
1114 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1115 ;;; they definitely won't generate any cleanup code. Currently we
1116 ;;; recognize lexical entry points that are only used locally (if at
1117 ;;; all).
1131 ;;; If a potentially TR local call really is TR, then convert it to
1132 ;;; jump directly to the called function. We also call
1133 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1134 ;;; tail-convert. The second is the value of M-C-T-A.
1141 ;; this is a tail call
1145 ;; If the call is in an XEP, we might decide to make it
1146 ;; non-tail so that we can use known return inside the
1147 ;; component.
1159 ;;; This is called when we believe it might make sense to convert
1160 ;;; CLAMBDA to an assignment. All this function really does is
1161 ;;; determine when a function with more than one call can still be
1162 ;;; combined with the calling function's environment. We can convert
1163 ;;; when:
1164 ;;; -- The function is a normal, non-entry function, and
1165 ;;; -- Except for one call, all calls must be tail recursive calls
1166 ;;; in the called function (i.e. are self-recursive tail calls)
1167 ;;; -- OK-INITIAL-CONVERT-P is true.
1168 ;;;
1169 ;;; There may be one outside call, and it need not be tail-recursive.
1170 ;;; Since all tail local calls have already been converted to direct
1171 ;;; transfers, the only control semantics needed are to splice in the
1172 ;;; body at the non-tail call. If there is no non-tail call, then we
1173 ;;; need only merge the environments. Both cases are handled by
1174 ;;; LET-CONVERT.
1175 ;;;
1176 ;;; ### It would actually be possible to allow any number of outside
1177 ;;; calls as long as they all return to the same place (i.e. have the
1178 ;;; same conceptual continuation.) A special case of this would be
1179 ;;; when all of the outside calls are tail recursive.
1206 t)
1208 nil))))))