| blob | b58d35b9272c63ae10e45efc5775dffdc91161ac |
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.
53 ;; The block may end up being deleted due to cast optimization
54 ;; caused by USE-GOOD-FOR-DX-P
57 ;; Stack analysis requires that the CALL ends the block, so
58 ;; that MAP-BLOCK-NLXES sees the cleanup we insert here.
63 :mess-up entry
75 ;;; This function handles merging the tail sets if CALL is potentially
76 ;;; tail-recursive, and is a call to a function with a different
77 ;;; TAIL-SET than CALL's FUN. This must be called whenever we alter
78 ;;; IR1 so as to place a local call in what might be a tail-recursive
79 ;;; context. Note that any call which returns its value to a RETURN is
80 ;;; considered potentially tail-recursive, since any implicit MV-PROG1
81 ;;; might be optimized away.
82 ;;;
83 ;;; We destructively modify the set for the calling function to
84 ;;; represent both, and then change all the functions in callee's set
85 ;;; to reference the first. If we do merge, we reoptimize the
86 ;;; RETURN-RESULT lvar to cause IR1-OPTIMIZE-RETURN to recompute the
87 ;;; tail set type.
101 t)))))
103 ;;; Convert a combination into a local call. We PROPAGATE-TO-ARGS, set
104 ;;; the combination kind to :LOCAL, add FUN to the CALLS of the
105 ;;; function that the call is in, call MERGE-TAIL-SETS, then replace
106 ;;; the function in the REF node with the new function.
107 ;;;
108 ;;; We change the REF last, since changing the reference can trigger
109 ;;; LET conversion of the new function, but will only do so if the
110 ;;; call is local. Note that the replacement may trigger LET
111 ;;; conversion or other changes in IR1. We must call MERGE-TAIL-SETS
112 ;;; with NEW-FUN before the substitution, since after the substitution
113 ;;; (and LET conversion), the call may no longer be recognizable as
114 ;;; tail-recursive.
128 \f
129 ;;;; external entry point creation
131 ;;; Return a LAMBDA form that can be used as the definition of the XEP
132 ;;; for FUN.
133 ;;;
134 ;;; If FUN is a LAMBDA, then we check the number of arguments
135 ;;; (conditional on policy) and call FUN with all the arguments.
136 ;;;
137 ;;; If FUN is an OPTIONAL-DISPATCH, then we dispatch off of the number
138 ;;; of supplied arguments by doing do an = test for each entry-point,
139 ;;; calling the entry with the appropriate prefix of the passed
140 ;;; arguments.
141 ;;;
142 ;;; If there is a &MORE arg, then there are a couple of optimizations
143 ;;; that we make (more for space than anything else):
144 ;;; -- If MIN-ARGS is 0, then we make the more entry a T clause, since
145 ;;; no argument count error is possible.
146 ;;; -- We can omit the = clause for the last entry-point, allowing the
147 ;;; case of 0 more args to fall through to the more entry.
148 ;;;
149 ;;; We don't bother to policy conditionalize wrong arg errors in
150 ;;; optional dispatches, since the additional overhead is negligible
151 ;;; compared to the cost of everything else going on.
152 ;;;
153 ;;; Note that if policy indicates it, argument type declarations in
154 ;;; FUN will be verified. Since nothing is known about the type of the
155 ;;; XEP arg vars, type checks will be emitted when the XEP's arg vars
156 ;;; are passed to the actual function.
164 #!+precise-arg-count-error
169 #!-precise-arg-count-error
183 ;; Force convertion of all entries
186 and n from min
189 #!+precise-arg-count-error
195 ;; Arg-checking is performed before this step,
196 ;; arranged by INIT-XEP-ENVIRONMENT,
197 ;; perform the last action unconditionally,
198 ;; and without this the function derived type will be bad.
205 #!-precise-arg-count-error
211 ;; KLUDGE: (NOT (< ...)) instead of >= avoids one round of
212 ;; deftransforms and lambda-conversion.
221 ;;; Make an external entry point (XEP) for FUN and return it. We
222 ;;; convert the result of MAKE-XEP-LAMBDA in the correct environment,
223 ;;; then associate this lambda with FUN as its XEP. After the
224 ;;; conversion, we iterate over the function's associated lambdas,
225 ;;; redoing local call analysis so that the XEP calls will get
226 ;;; converted.
227 ;;;
228 ;;; We set REANALYZE and REOPTIMIZE in the component, just in case we
229 ;;; discover an XEP after the initial local call analyze pass.
245 xep)))
258 ;;; Notice a REF that is not in a local-call context. If the REF is
259 ;;; already to an XEP, then do nothing, otherwise change it to the
260 ;;; XEP, making an XEP if necessary.
261 ;;;
262 ;;; If REF is to a special :CLEANUP or :ESCAPE function, then we treat
263 ;;; it as though it was not an XEP reference (i.e. leave it alone).
271 \f
272 ;;; Attempt to convert all references to FUN to local calls. The
273 ;;; reference must be the function for a call, and the function lvar
274 ;;; must be used only once, since otherwise we cannot be sure what
275 ;;; function is to be called. The call lvar would be multiply used if
276 ;;; there is hairy stuff such as conditionals in the expression that
277 ;;; computes the function.
278 ;;;
279 ;;; If we cannot convert a reference, then we mark the referenced
280 ;;; function as an entry-point, creating a new XEP if necessary. We
281 ;;; don't try to convert calls that are in error (:ERROR kind.)
282 ;;;
283 ;;; This is broken off from LOCALL-ANALYZE-COMPONENT so that people
284 ;;; can force analysis of newly introduced calls. Note that we don't
285 ;;; do LET conversion here.
310 ;;; We examine all NEW-FUNCTIONALS in COMPONENT, attempting to convert
311 ;;; calls into local calls when it is legal. We also attempt to
312 ;;; convert each LAMBDA to a LET. LET conversion is also triggered by
313 ;;; deletion of a function reference, but functions that start out
314 ;;; eligible for conversion must be noticed sometime.
315 ;;;
316 ;;; Note that there is a lot of action going on behind the scenes
317 ;;; here, triggered by reference deletion. In particular, the
318 ;;; COMPONENT-LAMBDAS are being hacked to remove newly deleted and LET
319 ;;; converted LAMBDAs, so it is important that the LAMBDA is added to
320 ;;; the COMPONENT-LAMBDAS when it is. Also, the
321 ;;; COMPONENT-NEW-FUNCTIONALS may contain all sorts of drivel, since
322 ;;; it is not updated when we delete functions, etc. Only
323 ;;; COMPONENT-LAMBDAS is updated.
324 ;;;
325 ;;; COMPONENT-REANALYZE-FUNCTIONALS is treated similarly to
326 ;;; COMPONENT-NEW-FUNCTIONALS, but we don't add lambdas to the
327 ;;; LAMBDAS.
345 ;; Fix/check FUNCTIONAL's relationship to COMPONENT-LAMDBAS.
347 ;; Since FUNCTIONAL isn't a LAMBDA, this doesn't
348 ;; apply: no-op.
351 ; NEW-FUNCTIONALS, hence is new.
352 ;; FUNCTIONAL becomes part of COMPONENT-LAMBDAS now.
357 ;; FUNCTIONAL should be in COMPONENT-LAMBDAS already.
359 component)))))
370 ;; The original CMU CL code seemed to implicitly assume that
371 ;; COMPONENT is the only one here. Let's make that explicit.
382 ;;; If policy is auspicious and CALL is not in an XEP and we don't seem
383 ;;; to be in an infinite recursive loop, then change the reference to
384 ;;; reference a fresh copy. We return whichever function we decide to
385 ;;; reference.
403 original-functional)))))))
408 calls this LET-converted local function:~
409 ~% ~S"
413 expand because expansion refers to ~
414 the optimized away object ~S."
415 losing-local-object)))
422 original-functional)
425 converted-lambda))))
426 original-functional))
428 ;;; Dispatch to the appropriate function to attempt to convert a call.
429 ;;; REF must be a reference to a FUNCTIONAL. This is called in IR1
430 ;;; optimization as well as in local call analysis. If the call is is
431 ;;; already :LOCAL, we do nothing. If the call is already scheduled
432 ;;; for deletion, also do nothing (in addition to saving time, this
433 ;;; also avoids some problems with optimizing collections of functions
434 ;;; that are partially deleted.)
435 ;;;
436 ;;; This is called both before and after FIND-INITIAL-DFO runs. When
437 ;;; called on a :INITIAL component, we don't care whether the caller
438 ;;; and callee are in the same component. Afterward, we must stick
439 ;;; with whatever component division we have chosen.
440 ;;;
441 ;;; Before attempting to convert a call, we see whether the function
442 ;;; is supposed to be inline expanded. Call conversion proceeds as
443 ;;; before after any expansion.
444 ;;;
445 ;;; We bind *COMPILER-ERROR-CONTEXT* to the node for the call so that
446 ;;; warnings will get the right context.
461 component))))
464 original-fun))
482 ;;; Attempt to convert a multiple-value call. The only interesting
483 ;;; case is a call to a function that LOOKS-LIKE-AN-MV-BIND, has
484 ;;; exactly one reference and no XEP, and is called with one values
485 ;;; lvar.
486 ;;;
487 ;;; We change the call to be to the last optional entry point and
488 ;;; change the call to be local. Due to our preconditions, the call
489 ;;; should eventually be converted to a let, but we can't do that now,
490 ;;; since there may be stray references to the e-p lambda due to
491 ;;; optional defaulting code.
492 ;;;
493 ;;; We also use variable types for the called function to construct an
494 ;;; assertion for the values lvar.
495 ;;;
496 ;;; See CONVERT-CALL for additional notes on MERGE-TAIL-SETS, etc.
521 while vars
522 do
524 arg
533 ;;; Convenience function to mark local calls as known bad.
539 default-name))))
546 node
552 ;;; Attempt to convert a call to a lambda. If the number of args is
553 ;;; wrong, we give a warning and mark the call as :ERROR to remove it
554 ;;; from future consideration. If the argcount is O.K. then we just
555 ;;; convert it.
564 'local-argument-mismatch
565 :format-control
566 "function called with ~R argument~:P, but wants exactly ~R"
568 \f
569 ;;;; &OPTIONAL, &MORE and &KEYWORD calls
571 ;;; This is similar to CONVERT-LAMBDA-CALL, but deals with
572 ;;; OPTIONAL-DISPATCHes. If only fixed args are supplied, then convert
573 ;;; a call to the correct entry point. If &KEY args are supplied, then
574 ;;; dispatch to a subfunction. We don't convert calls to functions
575 ;;; that have a &MORE (or &REST) arg.
584 'local-argument-mismatch
585 :format-control
586 "function called with ~R argument~:P, but wants at least ~R"
597 'local-argument-mismatch
598 :format-control
599 "function called with ~R argument~:P, but wants at most ~R"
600 :format-arguments
604 ;;; This function is used to convert a call to an entry point when
605 ;;; complex transformations need to be done on the original arguments.
606 ;;; ENTRY is the entry point function that we are calling. VARS is a
607 ;;; list of variable names which are bound to the original call
608 ;;; arguments. IGNORES is the subset of VARS which are ignored. ARGS
609 ;;; is the list of arguments to the entry point function.
610 ;;;
611 ;;; In order to avoid gruesome graph grovelling, we introduce a new
612 ;;; function that rearranges the arguments and calls the entry point.
613 ;;; We analyze the new function and the entry point immediately so
614 ;;; that everything gets converted during the single pass.
633 ;;; Use CONVERT-HAIRY-FUN-ENTRY to convert a &MORE-arg call to a known
634 ;;; function into a local call to the MAIN-ENTRY.
635 ;;;
636 ;;; First we verify that all keywords are constant and legal. If there
637 ;;; aren't, then we warn the user and don't attempt to convert the call.
638 ;;;
639 ;;; We massage the supplied &KEY arguments into the order expected
640 ;;; by the main entry. This is done by binding all the arguments to
641 ;;; the keyword call to variables in the introduced lambda, then
642 ;;; passing these values variables in the correct order when calling
643 ;;; the main entry. Unused arguments (such as the keywords themselves)
644 ;;; are discarded simply by not passing them along.
645 ;;;
646 ;;; If there is a &REST arg, then we bundle up the args and pass them
647 ;;; to LIST.
669 (:keyword
682 call
728 call
742 (:optional
746 t
748 (:rest
750 ;; &REST arguments may be accompanied by extra
751 ;; context and count arguments. We know this by
752 ;; the ARG-INFO-DEFAULT. Supply 0 and 0 or
753 ;; don't convert at all depending.
760 ;; We've already converted to use the more context
761 ;; instead of the rest list.
767 (:keyword
789 more-temps)))))
792 \f
793 ;;;; LET conversion
794 ;;;;
795 ;;;; Converting to a LET has differing significance to various parts
796 ;;;; of the compiler:
797 ;;;; -- The body of a LET is spliced in immediately after the
798 ;;;; corresponding combination node, making the control transfer
799 ;;;; explicit and allowing LETs to be mashed together into a single
800 ;;;; block. The value of the LET is delivered directly to the
801 ;;;; original lvar for the call, eliminating the need to
802 ;;;; propagate information from the dummy result lvar.
803 ;;;; -- As far as IR1 optimization is concerned, it is interesting in
804 ;;;; that there is only one expression that the variable can be bound
805 ;;;; to, and this is easily substituted for.
806 ;;;; -- LETs are interesting to environment analysis and to the back
807 ;;;; end because in most ways a LET can be considered to be "the
808 ;;;; same function" as its home function.
809 ;;;; -- LET conversion has dynamic scope implications, since control
810 ;;;; transfers within the same environment are local. In a local
811 ;;;; control transfer, cleanup code must be emitted to remove
812 ;;;; dynamic bindings that are no longer in effect.
814 ;;; Set up the control transfer to the called CLAMBDA. We split the
815 ;;; call block immediately after the call, and link the head of
816 ;;; CLAMBDA to the call block. The successor block after splitting
817 ;;; (where we return to) is returned.
818 ;;;
819 ;;; If the lambda is is a different component than the call, then we
820 ;;; call JOIN-COMPONENTS. This only happens in block compilation
821 ;;; before FIND-INITIAL-DFO.
838 next-block)))
840 ;;; Remove CLAMBDA from the tail set of anything it used to be in the
841 ;;; same set as; but leave CLAMBDA with a valid tail set value of
842 ;;; its own, for the benefit of code which might try to pull
843 ;;; something out of it (e.g. return type).
845 ;; Until sbcl-0.pre7.37.flaky5.2, we did
846 ;; (LET ((TAILS (LAMBDA-TAIL-SET CLAMBDA)))
847 ;; (SETF (TAIL-SET-FUNS TAILS)
848 ;; (DELETE CLAMBDA (TAIL-SET-FUNS TAILS))))
849 ;; (SETF (LAMBDA-TAIL-SET CLAMBDA) NIL)
850 ;; here. Apparently the idea behind the (SETF .. NIL) was that since
851 ;; TAIL-SET-FUNS no longer thinks we're in the tail set, it's
852 ;; inconsistent, and perhaps unsafe, for us to think we're in the
853 ;; tail set. Unfortunately..
854 ;;
855 ;; The (SETF .. NIL) caused problems in sbcl-0.pre7.37.flaky5.2 when
856 ;; I was trying to get Python to emit :EXTERNAL LAMBDAs directly
857 ;; (instead of only being able to emit funny little :TOPLEVEL stubs
858 ;; which you called in order to get the address of an external LAMBDA):
859 ;; the external function was defined in terms of internal function,
860 ;; which was LET-converted, and then things blew up downstream when
861 ;; FINALIZE-XEP-DEFINITION tried to find out its DEFINED-TYPE from
862 ;; the now-NILed-out TAIL-SET. So..
863 ;;
864 ;; To deal with this problem, we no longer NIL out
865 ;; (LAMBDA-TAIL-SET CLAMBDA) here. Instead:
866 ;; * If we're the only function in TAIL-SET-FUNS, it should
867 ;; be safe to leave ourself linked to it, and it to you.
868 ;; * If there are other functions in TAIL-SET-FUNS, then we're
869 ;; afraid of future optimizations on those functions causing
870 ;; the TAIL-SET object no longer to be valid to describe our
871 ;; return value. Thus, we delete ourselves from that object;
872 ;; but we save a newly-allocated tail-set, derived from the old
873 ;; one, for ourselves, for the use of later code (e.g.
874 ;; FINALIZE-XEP-DEFINITION) which might want to
875 ;; know about our return type.
884 ;; The documentation on TAIL-SET-INFO doesn't tell whether it could
885 ;; remain valid in this case, so we nuke it on the theory that
886 ;; missing information tends to be less dangerous than incorrect
887 ;; information.
890 ;;; Handle the PHYSENV semantics of LET conversion. We add CLAMBDA and
891 ;;; its LETs to LETs for the CALL's home function. We merge the calls
892 ;;; for CLAMBDA with the calls for the home function, removing CLAMBDA
893 ;;; in the process. We also merge the ENTRIES.
894 ;;;
895 ;;; We also unlink the function head from the component head and set
896 ;;; COMPONENT-REANALYZE to true to indicate that the DFO should be
897 ;;; recomputed.
917 ;; CLAMBDA belongs to HOME now.
929 ;; All of CLAMBDA's LETs belong to HOME now.
935 ;; CLAMBDA no longer has an independent existence as an entity
936 ;; which has LETs.
939 ;; HOME no longer calls CLAMBDA, and owns all of CLAMBDA's old
940 ;; DFO dependencies.
944 ;; CLAMBDA no longer has an independent existence as an entity
945 ;; which calls things or has DFO dependencies.
948 ;; All of CLAMBDA's ENTRIES belong to HOME now.
952 ;; CLAMBDA no longer has an independent existence as an entity
953 ;; with ENTRIES.
958 ;;; Handle the value semantics of LET conversion. Delete FUN's return
959 ;;; node, and change the control flow to transfer to NEXT-BLOCK
960 ;;; instead. Move all the uses of the result lvar to CALL's lvar.
979 ;; FIXME: Replace the call with unsafe CAST. -- APD, 2003-01-26
986 ;;; We are converting FUN to be a LET when the call is in a non-tail
987 ;;; position. Any previously tail calls in FUN are no longer tail
988 ;;; calls, and must be restored to normal calls which transfer to
989 ;;; NEXT-BLOCK (FUN's return point.) We can't do this by DO-USES on
990 ;;; the RETURN-RESULT, because the return might have been deleted (if
991 ;;; all calls were TR.)
1010 ;; The called function might be an assignment in the
1011 ;; case where we are currently converting that function.
1012 ;; In steady-state, assignments never appear as a called
1013 ;; function.
1014 (:assignment
1018 ;;; Deal with returning from a LET or assignment that we are
1019 ;;; converting. FUN is the function we are calling, CALL is a call to
1020 ;;; FUN, and NEXT-BLOCK is the return point for a non-tail call, or
1021 ;;; NULL if call is a tail call.
1022 ;;;
1023 ;;; If the call is not a tail call, then we must do
1024 ;;; UNCONVERT-TAIL-CALLS, since a tail call is a call which returns
1025 ;;; its value out of the enclosing non-let function. When call is
1026 ;;; non-TR, we must convert it back to an ordinary local call, since
1027 ;;; the value must be delivered to the receiver of CALL's value.
1028 ;;;
1029 ;;; We do different things depending on whether the caller and callee
1030 ;;; have returns left:
1032 ;;; -- If the callee has no return we just do MOVE-LET-CALL-CONT.
1033 ;;; Either the function doesn't return, or all returns are via
1034 ;;; tail-recursive local calls.
1035 ;;; -- If CALL is a non-tail call, or if both have returns, then
1036 ;;; we delete the callee's return, move its uses to the call's
1037 ;;; result lvar, and transfer control to the appropriate
1038 ;;; return point.
1039 ;;; -- If the callee has a return, but the caller doesn't, then we
1040 ;;; move the return to the caller.
1069 ;;; Actually do LET conversion. We call subfunctions to do most of the
1070 ;;; work. We do REOPTIMIZE-LVAR on the args and CALL's lvar so that
1071 ;;; LET-specific IR1 optimizations get a chance. We blow away any
1072 ;;; entry for the function in *FREE-FUNS* so that nobody will create
1073 ;;; new references to it.
1078 nil
1079 next-block)))
1083 ;; If CALL has a derive type NIL, it means that "its return" is
1084 ;; unreachable, but the next BIND is still reachable; in order to
1085 ;; not confuse MAYBE-TERMINATE-BLOCK...
1088 ;;; Reoptimize all of CALL's args and its result.
1097 ;;; Are there any declarations in force to say CLAMBDA shouldn't be
1098 ;;; LET converted?
1100 ;; From the user's point of view, LET-converting something that
1101 ;; has a name is inlining it. (The user can't see what we're doing
1102 ;; with anonymous things, and suppressing inlining
1103 ;; for such things can easily give Python acute indigestion, so
1104 ;; we don't.)
1105 ;;
1106 ;; A functional that is already inline-expanded in this componsne definitely
1107 ;; deserves let-conversion -- and in case of main entry points for inline
1108 ;; expanded optional dispatch, the main-etry isn't explicitly marked :INLINE
1109 ;; even if the function really is.
1112 ;; ANSI requires that explicit NOTINLINE be respected.
1114 ;; If (= LET-CONVERSION 0) we can guess that inlining
1115 ;; generally won't be appreciated, but if the user
1116 ;; specifically requests inlining, that takes precedence over
1117 ;; our general guess.
1121 ;;; We also don't convert calls to named functions which appear in the
1122 ;;; initial component, delaying this until optimization. This
1123 ;;; minimizes the likelihood that we will LET-convert a function which
1124 ;;; may have references added due to later local inline expansion.
1131 ;;; ir1opt usually takes care of forwarding let-bound values directly
1132 ;;; to their destination when possible. However, locall analysis
1133 ;;; greatly benefits from that transformation, and is executed in a
1134 ;;; distinct phase from ir1opt. After let-conversion, variables
1135 ;;; bound to functional values are immediately substituted away.
1136 ;;;
1137 ;;; When called from locall, component is non-nil, and the functionals
1138 ;;; are marked for reanalysis when appropriate.
1144 ;; only do that in the absence of assignment
1146 do
1151 ;; unlike propagate-let-args, we're only concerned with
1152 ;; functionals.
1154 ;; if the types match, we can mutate refs to point to
1155 ;; the functional instead of var
1165 nil)))
1166 leaf var)))
1167 ;; otherwise, we can still play LVAR-level tricks for single
1168 ;; destination variables.
1170 ;; Don't substitute single-ref variables on high-debug /
1171 ;; low speed, to improve the debugging experience.
1175 ;; if we've done something, the functional may now be used in
1176 ;; more analysis-friendly manners. Enqueue it if we're in
1177 ;; locall.
1179 component
1184 ;;; This function is called when there is some reason to believe that
1185 ;;; CLAMBDA might be converted into a LET. This is done after local
1186 ;;; call analysis, and also when a reference is deleted. We return
1187 ;;; true if we converted.
1188 ;;;
1189 ;;; COMPONENT is non-nil during local call analysis. It is used to
1190 ;;; re-enqueue functionals for reanalysis when they have been forwarded
1191 ;;; directly to destination nodes.
1197 ;; We only convert to a LET when the function is a normal local
1198 ;; function, has no XEP, and is referenced in exactly one local
1199 ;; call. Conversion is also inhibited if the only reference is in
1200 ;; a block about to be deleted.
1201 ;;
1202 ;; These rules limiting LET conversion may seem unnecessarily
1203 ;; restrictive, since there are some cases where we could do the
1204 ;; return with a jump that don't satisfy these requirements. The
1205 ;; reason for doing things this way is that it makes the concept
1206 ;; of a LET much more useful at the level of IR1 semantics. The
1207 ;; :ASSIGNMENT function kind provides another way to optimize
1208 ;; calls to single-return/multiple call functions.
1209 ;;
1210 ;; We don't attempt to convert calls to functions that have an
1211 ;; XEP, since we might be embarrassed later when we want to
1212 ;; convert a newly discovered local call. Also, see
1213 ;; OK-INITIAL-CONVERT-P.
1230 nil)))
1240 ; for me to handle - PK 2012-05-30
1242 t))))
1243 \f
1244 ;;;; tail local calls and assignments
1246 ;;; Return T if there are no cleanups between BLOCK1 and BLOCK2, or if
1247 ;;; they definitely won't generate any cleanup code. Currently we
1248 ;;; recognize lexical entry points that are only used locally (if at
1249 ;;; all).
1263 ;;; If a potentially TR local call really is TR, then convert it to
1264 ;;; jump directly to the called function. We also call
1265 ;;; MAYBE-CONVERT-TO-ASSIGNMENT. The first value is true if we
1266 ;;; tail-convert. The second is the value of M-C-T-A.
1273 ;; this is a tail call
1277 ;; If the call is in an XEP, we might decide to make it
1278 ;; non-tail so that we can use known return inside the
1279 ;; component.
1291 ;;; This is called when we believe it might make sense to convert
1292 ;;; CLAMBDA to an assignment. All this function really does is
1293 ;;; determine when a function with more than one call can still be
1294 ;;; combined with the calling function's environment. We can convert
1295 ;;; when:
1296 ;;; -- The function is a normal, non-entry function, and
1297 ;;; -- Except for one call, all calls must be tail recursive calls
1298 ;;; in the called function (i.e. are self-recursive tail calls)
1299 ;;; -- OK-INITIAL-CONVERT-P is true.
1300 ;;;
1301 ;;; There may be one outside call, and it need not be tail-recursive.
1302 ;;; Since all tail local calls have already been converted to direct
1303 ;;; transfers, the only control semantics needed are to splice in the
1304 ;;; body at the non-tail call. If there is no non-tail call, then we
1305 ;;; need only merge the environments. Both cases are handled by
1306 ;;; LET-CONVERT.
1307 ;;;
1308 ;;; ### It would actually be possible to allow any number of outside
1309 ;;; calls as long as they all return to the same place (i.e. have the
1310 ;;; same conceptual continuation.) A special case of this would be
1311 ;;; when all of the outside calls are tail recursive.
1338 t)
1340 nil))))))