| blob | 41985f2039500cdd284130dcc7e522eb996a0b9b |
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 ;; The block may end up being deleted due to cast optimization
55 ;; caused by USE-GOOD-FOR-DX-P
58 ;; Stack analysis requires that the CALL ends the block, so
59 ;; that MAP-BLOCK-NLXES sees the cleanup we insert here.
64 :mess-up entry
76 ;;; This function handles merging the tail sets if CALL is potentially
77 ;;; tail-recursive, and is a call to a function with a different
78 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
79 ;;; IR1 so as to place a local call in what might be a tail-recursive
80 ;;; context. Note that any call which returns its value to a RETURN is
81 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
82 ;;; might be optimized away.
83 ;;;
84 ;;; We destructively modify the set for the calling function to
85 ;;; represent both, and then change all the functions in callee's set
86 ;;; to reference the first. If we do merge, we reoptimize the
87 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
88 ;;; tail set type.
103 t)))))
105 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
106 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
107 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
108 ;;; the function in the REF node with the new function.
109 ;;;
110 ;;; We change the REF last, since changing the reference can trigger
111 ;;; LET conversion of the new function, but will only do so if the
112 ;;; call is local. Note that the replacement may trigger LET
113 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
114 ;;; with NEW-FUN before the substitution, since after the substitution
115 ;;; (and LET conversion), the call may no longer be recognizable as
116 ;;; tail-recursive.
130 \f
131 ;;;; external entry point creation
133 ;;; Return a LAMBDA form that can be used as the definition of the XEP
134 ;;; for FUN.
135 ;;;
136 ;;; If FUN is a LAMBDA, then we check the number of arguments
137 ;;; (conditional on policy) and call FUN with all the arguments.
138 ;;;
139 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
140 ;;; of supplied arguments by doing do an = test for each entry-point,
141 ;;; calling the entry with the appropriate prefix of the passed
142 ;;; arguments.
143 ;;;
144 ;;; If there is a &MORE arg, then there are a couple of optimizations
145 ;;; that we make (more for space than anything else):
146 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
147 ;;; no argument count error is possible.
148 ;;; -- We can omit the = clause for the last entry-point, allowing the
149 ;;; case of 0 more args to fall through to the more entry.
150 ;;;
151 ;;; We don't bother to policy conditionalize wrong arg errors in
152 ;;; optional dispatches, since the additional overhead is negligible
153 ;;; compared to the cost of everything else going on.
154 ;;;
155 ;;; Note that if policy indicates it, argument type declarations in
156 ;;; FUN will be verified. Since nothing is known about the type of the
157 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
158 ;;; are passed to the actual function.
171 ;; Force convertion of all entries
182 ;; Ignore only the entries at the tail, can't
183 ;; deal with the values being used by
184 ;; subsequent default forms at the moment
186 for last = t then nil
189 for n downfrom max
193 keyp))
204 collect
217 ;; The &rest var is unused, call the main entry point directly
222 with vars = temps
225 collect
227 t)
236 ;; Arg-checking is performed before this step,
237 ;; arranged by INIT-XEP-ENVIRONMENT,
238 ;; perform the last action unconditionally,
239 ;; and without this the function derived type will be bad.
244 for var in vars
247 return var)))
259 ;;; Make an external entry point (XEP) for FUN and return it. We
260 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
261 ;;; then associate this lambda with FUN as its XEP. After the
262 ;;; conversion, we iterate over the function's associated lambdas,
263 ;;; redoing local call analysis so that the XEP calls will get
264 ;;; converted.
265 ;;;
266 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
267 ;;; discover an XEP after the initial local call analyze pass.
283 xep)))
296 ;;; Notice a REF that is not in a local-call context. If the REF is
297 ;;; already to an XEP, then do nothing, otherwise change it to the
298 ;;; XEP, making an XEP if necessary.
299 ;;;
300 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
301 ;;; it as though it was not an XEP reference (i.e. leave it alone).
310 \f
311 ;;; Attempt to convert all references to FUN to local calls. The
312 ;;; reference must be the function for a call, and the function lvar
313 ;;; must be used only once, since otherwise we cannot be sure what
314 ;;; function is to be called. The call lvar would be multiply used if
315 ;;; there is hairy stuff such as conditionals in the expression that
316 ;;; computes the function.
317 ;;;
318 ;;; If we cannot convert a reference, then we mark the referenced
319 ;;; function as an entry-point, creating a new XEP if necessary. We
320 ;;; don't try to convert calls that are in error (:ERROR kind.)
321 ;;;
322 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
323 ;;; can force analysis of newly introduced calls. Note that we don't
324 ;;; do LET conversion here.
338 ;; It might have been deleted by CONVERT-CALL-IF-POSSIBLE
351 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
352 ;;; calls into local calls when it is legal. We also attempt to
353 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
354 ;;; deletion of a function reference, but functions that start out
355 ;;; eligible for conversion must be noticed sometime.
356 ;;;
357 ;;; Note that there is a lot of action going on behind the scenes
358 ;;; here, triggered by reference deletion. In particular, the
359 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
360 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
361 ;;; the COMPONENT-LAMBDAS when it is. Also, the
362 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
363 ;;; it is not updated when we delete functions, etc. Only
364 ;;; COMPONENT-LAMBDAS is updated.
365 ;;;
366 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
367 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
368 ;;; LAMBDAS.
386 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
388 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
389 ;; apply: no-op.
392 ; NEW-FUNCTIONALS, hence is new.
393 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
398 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
400 component)))))
411 ;; The original CMU CL code seemed to implicitly assume that
412 ;; COMPONENT is the only one here. Let's make that explicit.
423 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
424 ;;; to be in an infinite recursive loop, then change the reference to
425 ;;; reference a fresh copy. We return whichever function we decide to
426 ;;; reference.
444 original-functional)))))))
449 calls this LET-converted local function:~
450 ~% ~S"
454 expand because expansion refers to ~
455 the optimized away object ~S."
456 losing-local-object)))
463 original-functional)
466 converted-lambda))))
467 original-functional))
469 ;;; Dispatch to the appropriate function to attempt to convert a call.
470 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
471 ;;; optimization as well as in local call analysis. If the call is is
472 ;;; already :LOCAL, we do nothing. If the call is already scheduled
473 ;;; for deletion, also do nothing (in addition to saving time, this
474 ;;; also avoids some problems with optimizing collections of functions
475 ;;; that are partially deleted.)
476 ;;;
477 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
478 ;;; called on a :INITIAL component, we don't care whether the caller
479 ;;; and callee are in the same component. Afterward, we must stick
480 ;;; with whatever component division we have chosen.
481 ;;;
482 ;;; Before attempting to convert a call, we see whether the function
483 ;;; is supposed to be inline expanded. Call conversion proceeds as
484 ;;; before after any expansion.
485 ;;;
486 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
487 ;;; warnings will get the right context.
502 component))))
505 original-fun))
523 ;;; Attempt to convert a multiple-value call. The only interesting
524 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
525 ;;; exactly one reference and no XEP, and is called with one values
526 ;;; lvar.
527 ;;;
528 ;;; We change the call to be to the last optional entry point and
529 ;;; change the call to be local. Due to our preconditions, the call
530 ;;; should eventually be converted to a let, but we can't do that now,
531 ;;; since there may be stray references to the e-p lambda due to
532 ;;; optional defaulting code.
533 ;;;
534 ;;; We also use variable types for the called function to construct an
535 ;;; assertion for the values lvar.
536 ;;;
537 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
562 while vars
563 do
565 arg
570 for var in vars
575 ;;; Convenience function to mark local calls as known bad.
581 default-name))))
589 node
595 ;;; Attempt to convert a call to a lambda. If the number of args is
596 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
597 ;;; from future consideration. If the argcount is O.K. then we just
598 ;;; convert it.
607 'local-argument-mismatch
608 :format-control
609 "function called with ~R argument~:P, but wants exactly ~R"
611 \f
612 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
614 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
615 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
616 ;;; a call to the correct entry point. If &KEY args are supplied, then
617 ;;; dispatch to a subfunction. We don't convert calls to functions
618 ;;; that have a &MORE (or &REST) arg.
627 'local-argument-mismatch
628 :format-control
629 "function called with ~R argument~:P, but wants at least ~R"
640 'local-argument-mismatch
641 :format-control
642 "function called with ~R argument~:P, but wants at most ~R"
643 :format-arguments
647 ;;; This function is used to convert a call to an entry point when
648 ;;; complex transformations need to be done on the original arguments.
649 ;;; ENTRY is the entry point function that we are calling. VARS is a
650 ;;; list of variable names which are bound to the original call
651 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
652 ;;; is the list of arguments to the entry point function.
653 ;;;
654 ;;; In order to avoid gruesome graph grovelling, we introduce a new
655 ;;; function that rearranges the arguments and calls the entry point.
656 ;;; We analyze the new function and the entry point immediately so
657 ;;; that everything gets converted during the single pass.
676 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
677 ;;; function into a local call to the MAIN-ENTRY.
678 ;;;
679 ;;; First we verify that all keywords are constant and legal. If there
680 ;;; aren't, then we warn the user and don't attempt to convert the call.
681 ;;;
682 ;;; We massage the supplied &KEY arguments into the order expected
683 ;;; by the main entry. This is done by binding all the arguments to
684 ;;; the keyword call to variables in the introduced lambda, then
685 ;;; passing these values variables in the correct order when calling
686 ;;; the main entry. Unused arguments (such as the keywords themselves)
687 ;;; are discarded simply by not passing them along.
688 ;;;
689 ;;; If there is a &REST arg, then we bundle up the args and pass them
690 ;;; to LIST.
712 (:keyword
725 call
771 call
785 (:optional
789 t
791 (:rest
793 ;; &REST arguments may be accompanied by extra
794 ;; context and count arguments. We know this by
795 ;; the ARG-INFO-DEFAULT. Supply 0 and 0 or
796 ;; don't convert at all depending.
803 ;; We've already converted to use the more context
804 ;; instead of the rest list.
810 (:keyword
832 more-temps)))))
835 \f
836 ;;;; LET conversion
837 ;;;;
838 ;;;; Converting to a LET has differing significance to various parts
839 ;;;; of the compiler:
840 ;;;; -- The body of a LET is spliced in immediately after the
841 ;;;; corresponding combination node, making the control transfer
842 ;;;; explicit and allowing LETs to be mashed together into a single
843 ;;;; block. The value of the LET is delivered directly to the
844 ;;;; original lvar for the call, eliminating the need to
845 ;;;; propagate information from the dummy result lvar.
846 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
847 ;;;; that there is only one expression that the variable can be bound
848 ;;;; to, and this is easily substituted for.
849 ;;;; -- LETs are interesting to environment analysis and to the back
850 ;;;; end because in most ways a LET can be considered to be "the
851 ;;;; same function" as its home function.
852 ;;;; -- LET conversion has dynamic scope implications, since control
853 ;;;; transfers within the same environment are local. In a local
854 ;;;; control transfer, cleanup code must be emitted to remove
855 ;;;; dynamic bindings that are no longer in effect.
857 ;;; Set up the control transfer to the called CLAMBDA. We split the
858 ;;; call block immediately after the call, and link the head of
859 ;;; CLAMBDA to the call block. The successor block after splitting
860 ;;; (where we return to) is returned.
861 ;;;
862 ;;; If the lambda is is a different component than the call, then we
863 ;;; call JOIN-COMPONENTS. This only happens in block compilation
864 ;;; before FIND-INITIAL-DFO.
881 next-block)))
883 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
884 ;;; same set as; but leave CLAMBDA with a valid tail set value of
885 ;;; its own, for the benefit of code which might try to pull
886 ;;; something out of it (e.g. return type).
888 ;; Until sbcl-0.pre7.37.flaky5.2, we did
889 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
890 ;; (SETF (TAIL-SET-FUNS TAILS)
891 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
892 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
893 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
894 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
895 ;; inconsistent, and perhaps unsafe, for us to think we're in the
896 ;; tail set. Unfortunately..
897 ;;
898 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
899 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
900 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
901 ;; which you called in order to get the address of an external LAMBDA):
902 ;; the external function was defined in terms of internal function,
903 ;; which was LET-converted, and then things blew up downstream when
904 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
905 ;; the now-NILed-out TAIL-SET. So..
906 ;;
907 ;; To deal with this problem, we no longer NIL out
908 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
909 ;; * If we're the only function in TAIL-SET-FUNS, it should
910 ;; be safe to leave ourself linked to it, and it to you.
911 ;; * If there are other functions in TAIL-SET-FUNS, then we're
912 ;; afraid of future optimizations on those functions causing
913 ;; the TAIL-SET object no longer to be valid to describe our
914 ;; return value. Thus, we delete ourselves from that object;
915 ;; but we save a newly-allocated tail-set, derived from the old
916 ;; one, for ourselves, for the use of later code (e.g.
917 ;; FINALIZE-XEP-DEFINITION) which might want to
918 ;; know about our return type.
927 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
928 ;; remain valid in this case, so we nuke it on the theory that
929 ;; missing information tends to be less dangerous than incorrect
930 ;; information.
933 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
934 ;;; its LETs to LETs for the CALL's home function. We merge the calls
935 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
936 ;;; in the process. We also merge the ENTRIES.
937 ;;;
938 ;;; We also unlink the function head from the component head and set
939 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
940 ;;; recomputed.
960 ;; CLAMBDA belongs to HOME now.
972 ;; All of CLAMBDA's LETs belong to HOME now.
978 ;; CLAMBDA no longer has an independent existence as an entity
979 ;; which has LETs.
982 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
983 ;; DFO dependencies.
987 ;; CLAMBDA no longer has an independent existence as an entity
988 ;; which calls things or has DFO dependencies.
991 ;; All of CLAMBDA's ENTRIES belong to HOME now.
995 ;; CLAMBDA no longer has an independent existence as an entity
996 ;; with ENTRIES.
1001 ;;; Handle the value semantics of LET conversion. Delete FUN's return
1002 ;;; node, and change the control flow to transfer to NEXT-BLOCK
1003 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
1022 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
1029 ;;; We are converting FUN to be a LET when the call is in a non-tail
1030 ;;; position. Any previously tail calls in FUN are no longer tail
1031 ;;; calls, and must be restored to normal calls which transfer to
1032 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
1033 ;;; the RETURN-RESULT, because the return might have been deleted (if
1034 ;;; all calls were TR.)
1055 ;; The called function might be an assignment in the
1056 ;; case where we are currently converting that function.
1057 ;; In steady-state, assignments never appear as a called
1058 ;; function.
1059 (:assignment
1063 ;;; Deal with returning from a LET or assignment that we are
1064 ;;; converting. FUN is the function we are calling, CALL is a call to
1065 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
1066 ;;; NULL if call is a tail call.
1067 ;;;
1068 ;;; If the call is not a tail call, then we must do
1069 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
1070 ;;; its value out of the enclosing non-let function. When call is
1071 ;;; non-TR, we must convert it back to an ordinary local call, since
1072 ;;; the value must be delivered to the receiver of CALL's value.
1073 ;;;
1074 ;;; We do different things depending on whether the caller and callee
1075 ;;; have returns left:
1077 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
1078 ;;; Either the function doesn't return, or all returns are via
1079 ;;; tail-recursive local calls.
1080 ;;; -- If CALL is a non-tail call, or if both have returns, then
1081 ;;; we delete the callee's return, move its uses to the call's
1082 ;;; result lvar, and transfer control to the appropriate
1083 ;;; return point.
1084 ;;; -- If the callee has a return, but the caller doesn't, then we
1085 ;;; move the return to the caller.
1098 ;; A new return will be put into that lambda, don't want
1099 ;; DELETE-RETURN called by DELETE-BLOCK to delete the new return
1100 ;; from the lambda.
1101 ;; (Previously, UNLINK-NODE was called on the return, but it
1102 ;; doesn't work well on deleted blocks)
1104 call-return nil))
1120 ;;; Actually do LET conversion. We call subfunctions to do most of the
1121 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1122 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1123 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1124 ;;; new references to it.
1129 nil
1130 next-block)))
1134 ;; If CALL has a derive type NIL, it means that "its return" is
1135 ;; unreachable, but the next BIND is still reachable; in order to
1136 ;; not confuse MAYBE-TERMINATE-BLOCK...
1139 ;;; Reoptimize all of CALL's args and its result.
1148 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1149 ;;; LET converted?
1151 ;; From the user's point of view, LET-converting something that
1152 ;; has a name is inlining it. (The user can't see what we're doing
1153 ;; with anonymous things, and suppressing inlining
1154 ;; for such things can easily give Python acute indigestion, so
1155 ;; we don't.)
1156 ;;
1157 ;; A functional that is already inline-expanded in this componsne definitely
1158 ;; deserves let-conversion -- and in case of main entry points for inline
1159 ;; expanded optional dispatch, the main-etry isn't explicitly marked :INLINE
1160 ;; even if the function really is.
1163 ;; ANSI requires that explicit NOTINLINE be respected.
1165 ;; If (= LET-CONVERSION 0) we can guess that inlining
1166 ;; generally won't be appreciated, but if the user
1167 ;; specifically requests inlining, that takes precedence over
1168 ;; our general guess.
1172 ;;; We also don't convert calls to named functions which appear in the
1173 ;;; initial component, delaying this until optimization. This
1174 ;;; minimizes the likelihood that we will LET-convert a function which
1175 ;;; may have references added due to later local inline expansion.
1182 ;;; ir1opt usually takes care of forwarding let-bound values directly
1183 ;;; to their destination when possible. However, locall analysis
1184 ;;; greatly benefits from that transformation, and is executed in a
1185 ;;; distinct phase from ir1opt. After let-conversion, variables
1186 ;;; bound to functional values are immediately substituted away.
1187 ;;;
1188 ;;; When called from locall, component is non-nil, and the functionals
1189 ;;; are marked for reanalysis when appropriate.
1195 ;; only do that in the absence of assignment
1197 do
1202 ;; unlike propagate-let-args, we're only concerned with
1203 ;; functionals.
1205 ;; if the types match, we can mutate refs to point to
1206 ;; the functional instead of var
1214 leaf var)))
1215 ;; otherwise, we can still play LVAR-level tricks for single
1216 ;; destination variables.
1218 ;; Don't substitute single-ref variables on high-debug /
1219 ;; low speed, to improve the debugging experience.
1223 ;; if we've done something, the functional may now be used in
1224 ;; more analysis-friendly manners. Enqueue it if we're in
1225 ;; locall.
1227 component
1232 ;;; This function is called when there is some reason to believe that
1233 ;;; CLAMBDA might be converted into a LET. This is done after local
1234 ;;; call analysis, and also when a reference is deleted. We return
1235 ;;; true if we converted.
1236 ;;;
1237 ;;; COMPONENT is non-nil during local call analysis. It is used to
1238 ;;; re-enqueue functionals for reanalysis when they have been forwarded
1239 ;;; directly to destination nodes.
1245 ;; We only convert to a LET when the function is a normal local
1246 ;; function, has no XEP, and is referenced in exactly one local
1247 ;; call. Conversion is also inhibited if the only reference is in
1248 ;; a block about to be deleted.
1249 ;;
1250 ;; These rules limiting LET conversion may seem unnecessarily
1251 ;; restrictive, since there are some cases where we could do the
1252 ;; return with a jump that don't satisfy these requirements. The
1253 ;; reason for doing things this way is that it makes the concept
1254 ;; of a LET much more useful at the level of IR1 semantics. The
1255 ;; :ASSIGNMENT function kind provides another way to optimize
1256 ;; calls to single-return/multiple call functions.
1257 ;;
1258 ;; We don't attempt to convert calls to functions that have an
1259 ;; XEP, since we might be embarrassed later when we want to
1260 ;; convert a newly discovered local call. Also, see
1261 ;; OK-INITIAL-CONVERT-P.
1278 nil)))
1288 ; for me to handle - PK 2012-05-30
1290 t))))
1291 \f
1292 ;;;; tail local calls and assignments
1294 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1295 ;;; they definitely won't generate any cleanup code. Currently we
1296 ;;; recognize lexical entry points that are only used locally (if at
1297 ;;; all).
1311 ;;; If a potentially TR local call really is TR, then convert it to
1312 ;;; jump directly to the called function. We also call
1313 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1314 ;;; tail-convert. The second is the value of M-C-T-A.
1321 ;; this is a tail call
1325 ;; If the call is in an XEP, we might decide to make it
1326 ;; non-tail so that we can use known return inside the
1327 ;; component.
1339 ;;; This is called when we believe it might make sense to convert
1340 ;;; CLAMBDA to an assignment. All this function really does is
1341 ;;; determine when a function with more than one call can still be
1342 ;;; combined with the calling function's environment. We can convert
1343 ;;; when:
1344 ;;; -- The function is a normal, non-entry function, and
1345 ;;; -- Except for one call, all calls must be tail recursive calls
1346 ;;; in the called function (i.e. are self-recursive tail calls)
1347 ;;; -- OK-INITIAL-CONVERT-P is true.
1348 ;;;
1349 ;;; There may be one outside call, and it need not be tail-recursive.
1350 ;;; Since all tail local calls have already been converted to direct
1351 ;;; transfers, the only control semantics needed are to splice in the
1352 ;;; body at the non-tail call. If there is no non-tail call, then we
1353 ;;; need only merge the environments. Both cases are handled by
1354 ;;; LET-CONVERT.
1355 ;;;
1356 ;;; ### It would actually be possible to allow any number of outside
1357 ;;; calls as long as they all return to the same place (i.e. have the
1358 ;;; same conceptual continuation.) A special case of this would be
1359 ;;; when all of the outside calls are tail recursive.
1387 t)
1389 nil))))))