| blob | d7b3a635503e6b55cef3e9a727fa0116ff9e3f0a |
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.
54 ;; Stack analysis requires that the CALL ends the block, so
55 ;; that MAP-BLOCK-NLXES sees the cleanup we insert here.
60 :mess-up entry
72 ;;; This function handles merging the tail sets if CALL is potentially
73 ;;; tail-recursive, and is a call to a function with a different
74 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
75 ;;; IR1 so as to place a local call in what might be a tail-recursive
76 ;;; context. Note that any call which returns its value to a RETURN is
77 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
78 ;;; might be optimized away.
79 ;;;
80 ;;; We destructively modify the set for the calling function to
81 ;;; represent both, and then change all the functions in callee's set
82 ;;; to reference the first. If we do merge, we reoptimize the
83 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
84 ;;; tail set type.
98 t)))))
100 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
101 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
102 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
103 ;;; the function in the REF node with the new function.
104 ;;;
105 ;;; We change the REF last, since changing the reference can trigger
106 ;;; LET conversion of the new function, but will only do so if the
107 ;;; call is local. Note that the replacement may trigger LET
108 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
109 ;;; with NEW-FUN before the substitution, since after the substitution
110 ;;; (and LET conversion), the call may no longer be recognizable as
111 ;;; tail-recursive.
125 \f
126 ;;;; external entry point creation
128 ;;; Return a LAMBDA form that can be used as the definition of the XEP
129 ;;; for FUN.
130 ;;;
131 ;;; If FUN is a LAMBDA, then we check the number of arguments
132 ;;; (conditional on policy) and call FUN with all the arguments.
133 ;;;
134 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
135 ;;; of supplied arguments by doing do an = test for each entry-point,
136 ;;; calling the entry with the appropriate prefix of the passed
137 ;;; arguments.
138 ;;;
139 ;;; If there is a &MORE arg, then there are a couple of optimizations
140 ;;; that we make (more for space than anything else):
141 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
142 ;;; no argument count error is possible.
143 ;;; -- We can omit the = clause for the last entry-point, allowing the
144 ;;; case of 0 more args to fall through to the more entry.
145 ;;;
146 ;;; We don't bother to policy conditionalize wrong arg errors in
147 ;;; optional dispatches, since the additional overhead is negligible
148 ;;; compared to the cost of everything else going on.
149 ;;;
150 ;;; Note that if policy indicates it, argument type declarations in
151 ;;; FUN will be verified. Since nothing is known about the type of the
152 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
153 ;;; are passed to the actual function.
174 ;; Force convertion of all entries
177 and n from min
185 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
186 ;; deftransforms and lambda-conversion.
195 ;;; Make an external entry point (XEP) for FUN and return it. We
196 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
197 ;;; then associate this lambda with FUN as its XEP. After the
198 ;;; conversion, we iterate over the function's associated lambdas,
199 ;;; redoing local call analysis so that the XEP calls will get
200 ;;; converted.
201 ;;;
202 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
203 ;;; discover an XEP after the initial local call analyze pass.
219 xep)))
232 ;;; Notice a REF that is not in a local-call context. If the REF is
233 ;;; already to an XEP, then do nothing, otherwise change it to the
234 ;;; XEP, making an XEP if necessary.
235 ;;;
236 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
237 ;;; it as though it was not an XEP reference (i.e. leave it alone).
245 \f
246 ;;; Attempt to convert all references to FUN to local calls. The
247 ;;; reference must be the function for a call, and the function lvar
248 ;;; must be used only once, since otherwise we cannot be sure what
249 ;;; function is to be called. The call lvar would be multiply used if
250 ;;; there is hairy stuff such as conditionals in the expression that
251 ;;; computes the function.
252 ;;;
253 ;;; If we cannot convert a reference, then we mark the referenced
254 ;;; function as an entry-point, creating a new XEP if necessary. We
255 ;;; don't try to convert calls that are in error (:ERROR kind.)
256 ;;;
257 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
258 ;;; can force analysis of newly introduced calls. Note that we don't
259 ;;; do LET conversion here.
284 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
285 ;;; calls into local calls when it is legal. We also attempt to
286 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
287 ;;; deletion of a function reference, but functions that start out
288 ;;; eligible for conversion must be noticed sometime.
289 ;;;
290 ;;; Note that there is a lot of action going on behind the scenes
291 ;;; here, triggered by reference deletion. In particular, the
292 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
293 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
294 ;;; the COMPONENT-LAMBDAS when it is. Also, the
295 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
296 ;;; it is not updated when we delete functions, etc. Only
297 ;;; COMPONENT-LAMBDAS is updated.
298 ;;;
299 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
300 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
301 ;;; LAMBDAS.
319 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
321 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
322 ;; apply: no-op.
325 ; NEW-FUNCTIONALS, hence is new.
326 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
331 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
333 component)))))
344 ;; The original CMU CL code seemed to implicitly assume that
345 ;; COMPONENT is the only one here. Let's make that explicit.
356 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
357 ;;; to be in an infinite recursive loop, then change the reference to
358 ;;; reference a fresh copy. We return whichever function we decide to
359 ;;; reference.
377 original-functional)))))))
382 calls this LET-converted local function:~
383 ~% ~S"
387 expand because expansion refers to ~
388 the optimized away object ~S."
389 losing-local-object)))
396 original-functional)
399 converted-lambda))))
400 original-functional))
402 ;;; Dispatch to the appropriate function to attempt to convert a call.
403 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
404 ;;; optimization as well as in local call analysis. If the call is is
405 ;;; already :LOCAL, we do nothing. If the call is already scheduled
406 ;;; for deletion, also do nothing (in addition to saving time, this
407 ;;; also avoids some problems with optimizing collections of functions
408 ;;; that are partially deleted.)
409 ;;;
410 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
411 ;;; called on a :INITIAL component, we don't care whether the caller
412 ;;; and callee are in the same component. Afterward, we must stick
413 ;;; with whatever component division we have chosen.
414 ;;;
415 ;;; Before attempting to convert a call, we see whether the function
416 ;;; is supposed to be inline expanded. Call conversion proceeds as
417 ;;; before after any expansion.
418 ;;;
419 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
420 ;;; warnings will get the right context.
435 component))))
438 original-fun))
456 ;;; Attempt to convert a multiple-value call. The only interesting
457 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
458 ;;; exactly one reference and no XEP, and is called with one values
459 ;;; lvar.
460 ;;;
461 ;;; We change the call to be to the last optional entry point and
462 ;;; change the call to be local. Due to our preconditions, the call
463 ;;; should eventually be converted to a let, but we can't do that now,
464 ;;; since there may be stray references to the e-p lambda due to
465 ;;; optional defaulting code.
466 ;;;
467 ;;; We also use variable types for the called function to construct an
468 ;;; assertion for the values lvar.
469 ;;;
470 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
493 ;;; Convenience function to mark local calls as known bad.
499 default-name))))
511 ;;; Attempt to convert a call to a lambda. If the number of args is
512 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
513 ;;; from future consideration. If the argcount is O.K. then we just
514 ;;; convert it.
523 'local-argument-mismatch
524 :format-control
525 "function called with ~R argument~:P, but wants exactly ~R"
527 \f
528 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
530 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
531 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
532 ;;; a call to the correct entry point. If &KEY args are supplied, then
533 ;;; dispatch to a subfunction. We don't convert calls to functions
534 ;;; that have a &MORE (or &REST) arg.
543 'local-argument-mismatch
544 :format-control
545 "function called with ~R argument~:P, but wants at least ~R"
556 'local-argument-mismatch
557 :format-control
558 "function called with ~R argument~:P, but wants at most ~R"
559 :format-arguments
563 ;;; This function is used to convert a call to an entry point when
564 ;;; complex transformations need to be done on the original arguments.
565 ;;; ENTRY is the entry point function that we are calling. VARS is a
566 ;;; list of variable names which are bound to the original call
567 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
568 ;;; is the list of arguments to the entry point function.
569 ;;;
570 ;;; In order to avoid gruesome graph grovelling, we introduce a new
571 ;;; function that rearranges the arguments and calls the entry point.
572 ;;; We analyze the new function and the entry point immediately so
573 ;;; that everything gets converted during the single pass.
592 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
593 ;;; function into a local call to the MAIN-ENTRY.
594 ;;;
595 ;;; First we verify that all keywords are constant and legal. If there
596 ;;; aren't, then we warn the user and don't attempt to convert the call.
597 ;;;
598 ;;; We massage the supplied &KEY arguments into the order expected
599 ;;; by the main entry. This is done by binding all the arguments to
600 ;;; the keyword call to variables in the introduced lambda, then
601 ;;; passing these values variables in the correct order when calling
602 ;;; the main entry. Unused arguments (such as the keywords themselves)
603 ;;; are discarded simply by not passing them along.
604 ;;;
605 ;;; If there is a &REST arg, then we bundle up the args and pass them
606 ;;; to LIST.
628 (:keyword
641 call
687 call
701 (:optional
705 (:rest
707 ;; &REST arguments may be accompanied by extra
708 ;; context and count arguments. We know this by
709 ;; the ARG-INFO-DEFAULT. Supply 0 and 0 or
710 ;; don't convert at all depending.
717 ;; We've already converted to use the more context
718 ;; instead of the rest list.
724 (:keyword
741 more-temps)))))
744 \f
745 ;;;; LET conversion
746 ;;;;
747 ;;;; Converting to a LET has differing significance to various parts
748 ;;;; of the compiler:
749 ;;;; -- The body of a LET is spliced in immediately after the
750 ;;;; corresponding combination node, making the control transfer
751 ;;;; explicit and allowing LETs to be mashed together into a single
752 ;;;; block. The value of the LET is delivered directly to the
753 ;;;; original lvar for the call, eliminating the need to
754 ;;;; propagate information from the dummy result lvar.
755 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
756 ;;;; that there is only one expression that the variable can be bound
757 ;;;; to, and this is easily substituted for.
758 ;;;; -- LETs are interesting to environment analysis and to the back
759 ;;;; end because in most ways a LET can be considered to be "the
760 ;;;; same function" as its home function.
761 ;;;; -- LET conversion has dynamic scope implications, since control
762 ;;;; transfers within the same environment are local. In a local
763 ;;;; control transfer, cleanup code must be emitted to remove
764 ;;;; dynamic bindings that are no longer in effect.
766 ;;; Set up the control transfer to the called CLAMBDA. We split the
767 ;;; call block immediately after the call, and link the head of
768 ;;; CLAMBDA to the call block. The successor block after splitting
769 ;;; (where we return to) is returned.
770 ;;;
771 ;;; If the lambda is is a different component than the call, then we
772 ;;; call JOIN-COMPONENTS. This only happens in block compilation
773 ;;; before FIND-INITIAL-DFO.
790 next-block)))
792 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
793 ;;; same set as; but leave CLAMBDA with a valid tail set value of
794 ;;; its own, for the benefit of code which might try to pull
795 ;;; something out of it (e.g. return type).
797 ;; Until sbcl-0.pre7.37.flaky5.2, we did
798 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
799 ;; (SETF (TAIL-SET-FUNS TAILS)
800 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
801 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
802 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
803 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
804 ;; inconsistent, and perhaps unsafe, for us to think we're in the
805 ;; tail set. Unfortunately..
806 ;;
807 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
808 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
809 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
810 ;; which you called in order to get the address of an external LAMBDA):
811 ;; the external function was defined in terms of internal function,
812 ;; which was LET-converted, and then things blew up downstream when
813 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
814 ;; the now-NILed-out TAIL-SET. So..
815 ;;
816 ;; To deal with this problem, we no longer NIL out
817 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
818 ;; * If we're the only function in TAIL-SET-FUNS, it should
819 ;; be safe to leave ourself linked to it, and it to you.
820 ;; * If there are other functions in TAIL-SET-FUNS, then we're
821 ;; afraid of future optimizations on those functions causing
822 ;; the TAIL-SET object no longer to be valid to describe our
823 ;; return value. Thus, we delete ourselves from that object;
824 ;; but we save a newly-allocated tail-set, derived from the old
825 ;; one, for ourselves, for the use of later code (e.g.
826 ;; FINALIZE-XEP-DEFINITION) which might want to
827 ;; know about our return type.
836 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
837 ;; remain valid in this case, so we nuke it on the theory that
838 ;; missing information tends to be less dangerous than incorrect
839 ;; information.
842 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
843 ;;; its LETs to LETs for the CALL's home function. We merge the calls
844 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
845 ;;; in the process. We also merge the ENTRIES.
846 ;;;
847 ;;; We also unlink the function head from the component head and set
848 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
849 ;;; recomputed.
869 ;; CLAMBDA belongs to HOME now.
881 ;; All of CLAMBDA's LETs belong to HOME now.
887 ;; CLAMBDA no longer has an independent existence as an entity
888 ;; which has LETs.
891 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
892 ;; DFO dependencies.
896 ;; CLAMBDA no longer has an independent existence as an entity
897 ;; which calls things or has DFO dependencies.
900 ;; All of CLAMBDA's ENTRIES belong to HOME now.
904 ;; CLAMBDA no longer has an independent existence as an entity
905 ;; with ENTRIES.
910 ;;; Handle the value semantics of LET conversion. Delete FUN's return
911 ;;; node, and change the control flow to transfer to NEXT-BLOCK
912 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
931 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
938 ;;; We are converting FUN to be a LET when the call is in a non-tail
939 ;;; position. Any previously tail calls in FUN are no longer tail
940 ;;; calls, and must be restored to normal calls which transfer to
941 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
942 ;;; the RETURN-RESULT, because the return might have been deleted (if
943 ;;; all calls were TR.)
962 ;; The called function might be an assignment in the
963 ;; case where we are currently converting that function.
964 ;; In steady-state, assignments never appear as a called
965 ;; function.
966 (:assignment
970 ;;; Deal with returning from a LET or assignment that we are
971 ;;; converting. FUN is the function we are calling, CALL is a call to
972 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
973 ;;; NULL if call is a tail call.
974 ;;;
975 ;;; If the call is not a tail call, then we must do
976 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
977 ;;; its value out of the enclosing non-let function. When call is
978 ;;; non-TR, we must convert it back to an ordinary local call, since
979 ;;; the value must be delivered to the receiver of CALL's value.
980 ;;;
981 ;;; We do different things depending on whether the caller and callee
982 ;;; have returns left:
984 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
985 ;;; Either the function doesn't return, or all returns are via
986 ;;; tail-recursive local calls.
987 ;;; -- If CALL is a non-tail call, or if both have returns, then
988 ;;; we delete the callee's return, move its uses to the call's
989 ;;; result lvar, and transfer control to the appropriate
990 ;;; return point.
991 ;;; -- If the callee has a return, but the caller doesn't, then we
992 ;;; move the return to the caller.
1021 ;;; Actually do LET conversion. We call subfunctions to do most of the
1022 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1023 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1024 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1025 ;;; new references to it.
1030 nil
1031 next-block)))
1035 ;; If CALL has a derive type NIL, it means that "its return" is
1036 ;; unreachable, but the next BIND is still reachable; in order to
1037 ;; not confuse MAYBE-TERMINATE-BLOCK...
1040 ;;; Reoptimize all of CALL's args and its result.
1049 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1050 ;;; LET converted?
1052 ;; From the user's point of view, LET-converting something that
1053 ;; has a name is inlining it. (The user can't see what we're doing
1054 ;; with anonymous things, and suppressing inlining
1055 ;; for such things can easily give Python acute indigestion, so
1056 ;; we don't.)
1057 ;;
1058 ;; A functional that is already inline-expanded in this componsne definitely
1059 ;; deserves let-conversion -- and in case of main entry points for inline
1060 ;; expanded optional dispatch, the main-etry isn't explicitly marked :INLINE
1061 ;; even if the function really is.
1064 ;; ANSI requires that explicit NOTINLINE be respected.
1066 ;; If (= LET-CONVERSION 0) we can guess that inlining
1067 ;; generally won't be appreciated, but if the user
1068 ;; specifically requests inlining, that takes precedence over
1069 ;; our general guess.
1073 ;;; We also don't convert calls to named functions which appear in the
1074 ;;; initial component, delaying this until optimization. This
1075 ;;; minimizes the likelihood that we will LET-convert a function which
1076 ;;; may have references added due to later local inline expansion.
1083 ;;; ir1opt usually takes care of forwarding let-bound values directly
1084 ;;; to their destination when possible. However, locall analysis
1085 ;;; greatly benefits from that transformation, and is executed in a
1086 ;;; distinct phase from ir1opt. After let-conversion, variables
1087 ;;; bound to functional values are immediately substituted away.
1088 ;;;
1089 ;;; When called from locall, component is non-nil, and the functionals
1090 ;;; are marked for reanalysis when appropriate.
1096 ;; only do that in the absence of assignment
1098 do
1103 ;; unlike propagate-let-args, we're only concerned with
1104 ;; functionals.
1106 ;; if the types match, we can mutate refs to point to
1107 ;; the functional instead of var
1117 nil)))
1118 leaf var)))
1119 ;; otherwise, we can still play LVAR-level tricks for single
1120 ;; destination variables.
1122 ;; Don't substitute single-ref variables on high-debug /
1123 ;; low speed, to improve the debugging experience.
1127 ;; if we've done something, the functional may now be used in
1128 ;; more analysis-friendly manners. Enqueue it if we're in
1129 ;; locall.
1131 component
1136 ;;; This function is called when there is some reason to believe that
1137 ;;; CLAMBDA might be converted into a LET. This is done after local
1138 ;;; call analysis, and also when a reference is deleted. We return
1139 ;;; true if we converted.
1140 ;;;
1141 ;;; COMPONENT is non-nil during local call analysis. It is used to
1142 ;;; re-enqueue functionals for reanalysis when they have been forwarded
1143 ;;; directly to destination nodes.
1149 ;; We only convert to a LET when the function is a normal local
1150 ;; function, has no XEP, and is referenced in exactly one local
1151 ;; call. Conversion is also inhibited if the only reference is in
1152 ;; a block about to be deleted.
1153 ;;
1154 ;; These rules limiting LET conversion may seem unnecessarily
1155 ;; restrictive, since there are some cases where we could do the
1156 ;; return with a jump that don't satisfy these requirements. The
1157 ;; reason for doing things this way is that it makes the concept
1158 ;; of a LET much more useful at the level of IR1 semantics. The
1159 ;; :ASSIGNMENT function kind provides another way to optimize
1160 ;; calls to single-return/multiple call functions.
1161 ;;
1162 ;; We don't attempt to convert calls to functions that have an
1163 ;; XEP, since we might be embarrassed later when we want to
1164 ;; convert a newly discovered local call. Also, see
1165 ;; OK-INITIAL-CONVERT-P.
1182 nil)))
1192 ; for me to handle - PK 2012-05-30
1194 t))))
1195 \f
1196 ;;;; tail local calls and assignments
1198 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1199 ;;; they definitely won't generate any cleanup code. Currently we
1200 ;;; recognize lexical entry points that are only used locally (if at
1201 ;;; all).
1215 ;;; If a potentially TR local call really is TR, then convert it to
1216 ;;; jump directly to the called function. We also call
1217 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1218 ;;; tail-convert. The second is the value of M-C-T-A.
1225 ;; this is a tail call
1229 ;; If the call is in an XEP, we might decide to make it
1230 ;; non-tail so that we can use known return inside the
1231 ;; component.
1243 ;;; This is called when we believe it might make sense to convert
1244 ;;; CLAMBDA to an assignment. All this function really does is
1245 ;;; determine when a function with more than one call can still be
1246 ;;; combined with the calling function's environment. We can convert
1247 ;;; when:
1248 ;;; -- The function is a normal, non-entry function, and
1249 ;;; -- Except for one call, all calls must be tail recursive calls
1250 ;;; in the called function (i.e. are self-recursive tail calls)
1251 ;;; -- OK-INITIAL-CONVERT-P is true.
1252 ;;;
1253 ;;; There may be one outside call, and it need not be tail-recursive.
1254 ;;; Since all tail local calls have already been converted to direct
1255 ;;; transfers, the only control semantics needed are to splice in the
1256 ;;; body at the non-tail call. If there is no non-tail call, then we
1257 ;;; need only merge the environments. Both cases are handled by
1258 ;;; LET-CONVERT.
1259 ;;;
1260 ;;; ### It would actually be possible to allow any number of outside
1261 ;;; calls as long as they all return to the same place (i.e. have the
1262 ;;; same conceptual continuation.) A special case of this would be
1263 ;;; when all of the outside calls are tail recursive.
1290 t)
1292 nil))))))