| blob | d8122aa0220bec26f10977cf1413d46760a9339b |
1 ;;;; function call for the x86 VM
3 ;;;; This software is part of the SBCL system. See the README file for
4 ;;;; more information.
5 ;;;;
6 ;;;; This software is derived from the CMU CL system, which was
7 ;;;; written at Carnegie Mellon University and released into the
8 ;;;; public domain. The software is in the public domain and is
9 ;;;; provided with absolutely no warranty. See the COPYING and CREDITS
10 ;;;; files for more information.
13 \f
14 ;;;; interfaces to IR2 conversion
16 ;;; Return a wired TN describing the N'th full call argument passing
17 ;;; location.
35 ;;; Make a passing location TN for a local call return PC.
36 ;;;
37 ;;; Always wire the return PC location to the stack in its standard
38 ;;; location.
42 sap-stack-sc-number return-pc-save-offset))
47 ;;; This is similar to MAKE-RETURN-PC-PASSING-LOCATION, but makes a
48 ;;; location to pass OLD-FP in.
49 ;;;
50 ;;; This is wired in both the standard and the local-call conventions,
51 ;;; because we want to be able to assume it's always there. Besides,
52 ;;; the x86 doesn't have enough registers to really make it profitable
53 ;;; to pass it in a register.
57 ocfp-save-offset))
62 ;;; Make the TNs used to hold OLD-FP and RETURN-PC within the current
63 ;;; function. We treat these specially so that the debugger can find
64 ;;; them at a known location.
65 ;;;
66 ;;; Without using a save-tn - which does not make much sense if it is
67 ;;; wired to the stack?
70 control-stack-sc-number
71 ocfp-save-offset)
72 physenv))
76 sap-stack-sc-number return-pc-save-offset)
77 physenv))
79 ;;; Make a TN for the standard argument count passing location. We only
80 ;;; need to make the standard location, since a count is never passed when we
81 ;;; are using non-standard conventions.
85 ;;; Make a TN to hold the number-stack frame pointer. This is allocated
86 ;;; once per component, and is component-live.
96 ;;; Return a list of TNs that can be used to represent an unknown-values
97 ;;; continuation within a function.
102 ;;; This function is called by the ENTRY-ANALYZE phase, allowing
103 ;;; VM-dependent initialization of the IR2-COMPONENT structure. We
104 ;;; push placeholder entries in the CONSTANTS to leave room for
105 ;;; additional noise in the code object header.
108 ;; The 1+ here is because for the x86 the first constant is a
109 ;; pointer to a list of fixups, or NIL if the code object has none.
110 ;; (If I understand correctly, the fixups are needed at GC copy
111 ;; time because the X86 code isn't relocatable.)
112 ;;
113 ;; KLUDGE: It'd be cleaner to have the fixups entry be a named
114 ;; element of the CODE (aka component) primitive object. However,
115 ;; it's currently a large, tricky, error-prone chore to change
116 ;; the layout of any primitive object, so for the foreseeable future
117 ;; we'll just live with this ugliness. -- WHN 2002-01-02
122 \f
123 ;;;; frame hackery
125 ;;; This is used for setting up the Old-FP in local call.
131 ;;; We don't have a separate NFP, so we don't need to do anything here.
136 nil))
138 ;;; Accessing a slot from an earlier stack frame is definite hackery.
162 :base frame-pointer
184 :base frame-pointer
197 load)))
206 set)))))))
213 :base frame-pointer
220 complex-double-stack
234 complex-single-stack
253 collect
254 `(,name
260 fp value res))
263 fp new-val value)))))))))
265 ancestor-frame-ref/double-float
268 ancestor-frame-ref/single-float
271 ancestor-frame-ref/complex-double-float
274 ancestor-frame-ref/complex-single-float
277 ancestor-frame-ref/signed-byte-32
280 ancestor-frame-ref/unsigned-byte-32
283 ancestor-frame-ref/unsigned-byte-32
286 ancestor-frame-ref/system-area-pointer
294 ;; Skip space for the function header.
299 ;; The start of the actual code.
300 ;; Save the return-pc.
311 ;;; This is emitted directly before either a known-call-local, call-local,
312 ;;; or a multiple-call-local. All it does is allocate stack space for the
313 ;;; callee (who has the same size stack as us).
324 ;;; Allocate a partial frame for passing stack arguments in a full
325 ;;; call. NARGS is the number of arguments passed. We allocate at
326 ;;; least 3 slots, because the XEP noise is going to want to use them
327 ;;; before it can extend the stack.
335 \f
336 ;;; Emit code needed at the return-point from an unknown-values call
337 ;;; for a fixed number of values. Values is the head of the TN-REF
338 ;;; list for the locations that the values are to be received into.
339 ;;; Nvals is the number of values that are to be received (should
340 ;;; equal the length of Values).
341 ;;;
342 ;;; If 0 or 1 values are expected, then we just emit an instruction to
343 ;;; reset the SP (which will only be executed when other than 1 value
344 ;;; is returned.)
345 ;;;
346 ;;; In the general case we have to do three things:
347 ;;; -- Default unsupplied register values. This need only be done
348 ;;; when a single value is returned, since register values are
349 ;;; defaulted by the called in the non-single case.
350 ;;; -- Default unsupplied stack values. This needs to be done whenever
351 ;;; there are stack values.
352 ;;; -- Reset SP. This must be done whenever other than 1 value is
353 ;;; returned, regardless of the number of values desired.
363 register-arg-count)
366 ;; The function never returns, it may happen that the code
367 ;; ends right here leavig the :SINGLE-VALUE-RETURN note
368 ;; dangling. Let's emit a NOP.
384 ;; Default the unsupplied registers.
392 for count from 2 below register-arg-count
400 ;; The number of bytes depends on the relative jump instructions.
401 ;; Best case is 31+(n-3)*14, worst case is 35+(n-3)*18. For
402 ;; NVALS=6 that is 73/89 bytes, and for NVALS=7 that is 87/107
403 ;; bytes which is likely better than using the blt below.
408 ;; Branch off to the MV case.
410 ;; Do the single value case.
411 ;; Default the register args
417 ;; Fake other registers so it looks like we returned with all the
418 ;; registers filled in.
438 ;; There are stack args so the frame of the callee is
439 ;; still there, save EDX in its first slot temporalily.
453 ;; We are setting the first stack argument to NIL.
454 ;; The callee's stack frame is dead, save EDX by
455 ;; pushing it to the stack, it will end up at same
456 ;; place as in the (STOREW EDX-TN EBX-TN -1) case
457 ;; above.
462 ;; 91 bytes for this branch.
469 ;; Branch off to the MV case.
471 ;; Default the register args, and set up the stack as if we
472 ;; entered the MV return point.
476 ;; Compute a pointer to where to put the [defaulted] stack values.
483 ;; Load EAX with NIL so we can quickly store it, and set up
484 ;; stuff for the loop.
488 ;; Jump into the default loop.
490 ;; The regs are defaulted. We need to copy any stack arguments,
491 ;; and then default the remaining stack arguments.
493 ;; Compute the number of stack arguments, and if it's zero or
494 ;; less, don't copy any stack arguments.
497 ;; Save EDI.
499 ;; Throw away any unwanted args.
504 ;; Save the number of stack values.
506 ;; Compute a pointer to where the stack args go.
510 ;; Save ESI, and compute a pointer to where the args come from.
516 ;; Do the copy.
521 ;; Restore ESI.
523 ;; Now we have to default the remaining args. Find out how many.
526 ;; If none, then just blow out of here.
530 ;; Load EAX with NIL for fast storing.
532 ;; Do the store.
536 ;; Restore EDI, and reset the stack.
542 \f
543 ;;;; unknown values receiving
545 ;;; Emit code needed at the return point for an unknown-values call
546 ;;; for an arbitrary number of values.
547 ;;;
548 ;;; We do the single and non-single cases with no shared code: there
549 ;;; doesn't seem to be any potential overlap, and receiving a single
550 ;;; value is more important efficiency-wise.
551 ;;;
552 ;;; When there is a single value, we just push it on the stack,
553 ;;; returning the old SP and 1.
554 ;;;
555 ;;; When there is a variable number of values, we move all of the
556 ;;; argument registers onto the stack, and return ARGS and NARGS.
557 ;;;
558 ;;; ARGS and NARGS are TNs wired to the named locations. We must
559 ;;; explicitly allocate these TNs, since their lifetimes overlap with
560 ;;; the results start and count. (Also, it's nice to be able to target
561 ;;; them.)
578 ;; The stack frame is burnt and RETurned from if there are no
579 ;; stack values. In this case quickly reallocate sufficient space.
581 register-arg-count)
586 ;; dtc: this writes the registers onto the stack even if they are
587 ;; not needed, only the number specified in ecx are used and have
588 ;; stack allocated to them. No harm is done.
591 for i downfrom -1
600 ;;; VOP that can be inherited by unknown values receivers. The main thing this
601 ;;; handles is allocation of the result temporaries.
605 values-start)
608 nvals)
611 \f
612 ;;;; local call with unknown values convention return
615 #+nil
619 #+nil
631 ;;; The local call convention doesn't fit that well with x86-style
632 ;;; calls. Emit a header for local calls to pop the return address
633 ;;; in the right place.
643 ;;; Non-TR local call for a fixed number of values passed according to
644 ;;; the unknown values convention.
645 ;;;
646 ;;; FP is the frame pointer in install before doing the call.
647 ;;;
648 ;;; NFP would be the number-stack frame pointer if we had a separate
649 ;;; number stack.
650 ;;;
651 ;;; Args are the argument passing locations, which are specified only
652 ;;; to terminate their lifetimes in the caller.
653 ;;;
654 ;;; VALUES are the return value locations (wired to the standard
655 ;;; passing locations). NVALS is the number of values received.
656 ;;;
657 ;;; Save is the save info, which we can ignore since saving has been
658 ;;; done.
659 ;;;
660 ;;; TARGET is a continuation pointing to the start of the called
661 ;;; function.
679 ;;; Non-TR local call for a variable number of return values passed according
680 ;;; to the unknown values convention. The results are the start of the values
681 ;;; glob and the number of values received.
698 \f
699 ;;;; local call with known values return
701 ;;; Non-TR local call with known return locations. Known-value return
702 ;;; works just like argument passing in local call.
703 ;;;
704 ;;; Note: we can't use normal load-tn allocation for the fixed args,
705 ;;; since all registers may be tied up by the more operand. Instead,
706 ;;; we use MAYBE-LOAD-STACK-TN.
722 \f
723 ;;; From Douglas Crosher
724 ;;; Return from known values call. We receive the return locations as
725 ;;; arguments to terminate their lifetimes in the returning function. We
726 ;;; restore FP and CSP and jump to the Return-PC.
737 ;; Zot all of the stack except for the old-fp and return-pc.
741 \f
742 ;;;; full call
743 ;;;
744 ;;; There is something of a cross-product effect with full calls.
745 ;;; Different versions are used depending on whether we know the
746 ;;; number of arguments or the name of the called function, and
747 ;;; whether we want fixed values, unknown values, or a tail call.
748 ;;;
749 ;;; In full call, the arguments are passed creating a partial frame on
750 ;;; the stack top and storing stack arguments into that frame. On
751 ;;; entry to the callee, this partial frame is pointed to by FP.
753 ;;; This macro helps in the definition of full call VOPs by avoiding
754 ;;; code replication in defining the cross-product VOPs.
755 ;;;
756 ;;; NAME is the name of the VOP to define.
757 ;;;
758 ;;; NAMED is true if the first argument is an fdefinition object whose
759 ;;; definition is to be called.
760 ;;;
761 ;;; RETURN is either :FIXED, :UNKNOWN or :TAIL:
762 ;;; -- If :FIXED, then the call is for a fixed number of values, returned in
763 ;;; the standard passing locations (passed as result operands).
764 ;;; -- If :UNKNOWN, then the result values are pushed on the stack, and the
765 ;;; result values are specified by the Start and Count as in the
766 ;;; unknown-values continuation representation.
767 ;;; -- If :TAIL, then do a tail-recursive call. No values are returned.
768 ;;; The Old-Fp and Return-PC are passed as the second and third arguments.
769 ;;;
770 ;;; In non-tail calls, the pointer to the stack arguments is passed as
771 ;;; the last fixed argument. If Variable is false, then the passing
772 ;;; locations are passed as a more arg. Variable is true if there are
773 ;;; a variable number of arguments passed on the stack. Variable
774 ;;; cannot be specified with :TAIL return. TR variable argument call
775 ;;; is implemented separately.
776 ;;;
777 ;;; In tail call with fixed arguments, the passing locations are
778 ;;; passed as a more arg, but there is no new-FP, since the arguments
779 ;;; have been set up in the current frame.
785 (:args
807 (:info
811 step-instrumenting)
813 (:ignore
817 ;; We pass either the fdefn object (for named call) or
818 ;; the actual function object (for unnamed call) in
819 ;; EAX. With named call, closure-tramp will replace it
820 ;; with the real function and invoke the real function
821 ;; for closures. Non-closures do not need this value,
822 ;; so don't care what shows up in it.
823 (:temporary
824 (:sc descriptor-reg
825 :offset eax-offset
828 eax)
830 ;; We pass the number of arguments in ECX.
833 ;; With variable call, we have to load the
834 ;; register-args out of the (new) stack frame before
835 ;; doing the call. Therefore, we have to tell the
836 ;; lifetime stuff that we need to use them.
850 old-fp-tmp)))
857 15
860 ;; This has to be done before the frame pointer is
861 ;; changed! EAX stores the 'lexical environment' needed
862 ;; for closures.
867 ;; For variable call, compute the number of
868 ;; arguments and move some of the arguments to
869 ;; registers.
871 ;; Compute the number of arguments.
874 ;; Move the necessary args to registers,
875 ;; this moves them all even if they are
876 ;; not all needed.
879 for index downfrom -1
887 ;; return to so might as well use that clue.
888 ;; This seems really important to the
889 ;; implementation of things like
890 ;; (without-interrupts ...)
891 ;;
892 ;; dtc; Could be doing a tail call from a
893 ;; known-local-call etc in which the old-fp
894 ;; or ret-pc are in regs or in non-standard
895 ;; places. If the passing location were
896 ;; wired to the stack in standard locations
897 ;; then these moves will be un-necessary;
898 ;; this is probably best for the x86.
903 ;; FIXME: FORMAT T for stale
904 ;; diagnostic output (several of
905 ;; them around here), ick
909 ebp-tn
914 ebp-tn
917 ;; For tail call, we have to push the
918 ;; return-pc so that it looks like we CALLed
919 ;; despite the fact that we are going to JMP.
921 ))
923 ;; For non-tail call, we have to save our
924 ;; frame pointer and install the new frame
925 ;; pointer. We can't load stack tns after this
926 ;; point.
928 ;; here which doesn't leave room for the
929 ;; ofp/ret stuff.
931 ;; The variable args are on the stack and
932 ;; become the frame, but there may be <3
933 ;; args and 3 stack slots are assumed
934 ;; allocate on the call. So need to ensure
935 ;; there are at least 3 slots. This hack
936 ;; just adds 3 more.
940 ;; Bias the new-fp for use as an fp
944 ;; Save the fp
949 )))
955 DONE
962 other-pointer-lowtag)
964 fun-pointer-lowtag)))
966 (:fixed
968 (:unknown
971 node)))
984 ;;; This is defined separately, since it needs special code that BLT's
985 ;;; the arguments down. All the real work is done in the assembly
986 ;;; routine. We just set things up so that it can find what it needs.
996 ;; Move these into the passing locations if they are not already there.
999 ;; And jump to the assembly routine.
1001 \f
1002 ;;;; unknown values return
1004 ;;; Return a single-value using the Unknown-Values convention.
1005 ;;;
1006 ;;; pfw--get wired-tn conflicts sometimes if register sc specd for args
1007 ;;; having problems targeting args to regs -- using temps instead.
1008 ;;;
1009 ;;; First off, modifying the return-pc defeats the branch-prediction
1010 ;;; optimizations on modern CPUs quite handily. Second, we can do all
1011 ;;; this without needing a temp register. Fixed the latter, at least.
1012 ;;; -- AB 2006/Feb/04
1020 ;; Drop stack above old-fp
1022 ;; Clear the multiple-value return flag
1024 ;; Restore the old frame pointer
1026 ;; And return.
1029 ;;; Do unknown-values return of a fixed (other than 1) number of
1030 ;;; values. The VALUES are required to be set up in the standard
1031 ;;; passing locations. NVALS is the number of values returned.
1032 ;;;
1033 ;;; Basically, we just load ECX with the number of values returned and
1034 ;;; EBX with a pointer to the values, set ESP to point to the end of
1035 ;;; the values, and jump directly to return-pc.
1042 ;; In the case of other than one value, we need these registers to
1043 ;; tell the caller where they are and how many there are.
1046 ;; We need to stretch the lifetime of return-pc past the argument
1047 ;; registers so that we can default the argument registers without
1048 ;; trashing return-pc.
1059 ;; This is handled in RETURN-SINGLE.
1061 ;; Establish the values pointer and values count.
1067 ;; Pre-default any argument register that need it.
1074 ;; Set the multiple value return flag.
1076 ;; And away we go. Except that return-pc is still on the
1077 ;; stack and we've changed the stack pointer. So we have to
1078 ;; tell it to index off of EBX instead of EBP.
1084 ;; Some values are on the stack after RETURN-PC and OLD-FP,
1085 ;; can't return normally and some slots of the frame will
1086 ;; be used as temporaries by the receiver.
1087 ;;
1088 ;; Clear as much of the stack as possible, but not past the
1089 ;; old frame address.
1100 ;;; Do unknown-values return of an arbitrary number of values (passed
1101 ;;; on the stack.) We check for the common case of a single return
1102 ;;; value, and do that inline using the normal single value return
1103 ;;; convention. Otherwise, we branch off to code that calls an
1104 ;;; assembly-routine.
1105 ;;;
1106 ;;; The assembly routine takes the following args:
1107 ;;; ECX -- number of values to find there.
1108 ;;; ESI -- pointer to where to find the values.
1122 ;; Check for the single case.
1126 ;; Return with one value.
1128 ;; Clear the stack until ocfp.
1130 ;; clear the multiple-value return flag
1132 ;; Out of here.
1135 ;; Nope, not the single case. Jump to the assembly routine.
1140 \f
1141 ;;;; XEP hackery
1143 ;;; Get the lexical environment from its passing location.
1149 ;; Get result.
1152 ;;; Copy a &MORE arg from the argument area to the end of the current
1153 ;;; frame. FIXED is the number of non-&MORE arguments.
1154 ;;;
1155 ;;; The tricky part is doing this without trashing any of the calling
1156 ;;; convention registers that are still needed. This vop is emitted
1157 ;;; directly after the xep-allocate frame. That means the registers
1158 ;;; are in use as follows:
1159 ;;;
1160 ;;; EAX -- The lexenv.
1161 ;;; EBX -- Available.
1162 ;;; ECX -- The total number of arguments * N-WORD-BYTES.
1163 ;;; EDX -- The first arg.
1164 ;;; EDI -- The second arg.
1165 ;;; ESI -- The third arg.
1166 ;;;
1167 ;;; So basically, we have one register available for our use: EBX.
1168 ;;;
1169 ;;; What we can do is push the other regs onto the stack, and then
1170 ;;; restore their values by looking directly below where we put the
1171 ;;; more-args.
1175 ;; Avoid the copy if there are no more args.
1182 ;; Allocate the space on the stack.
1183 ;; stack = ebp + sp->fp-offset - (max 3 frame-size) - (nargs - fixed)
1184 ;;
1185 ;; Problem: this might leave some &more args outside esp, so
1186 ;; clamp the movement for now. If fixed > frame-size, reset
1187 ;; esp to the end of the current &more args (which *should*
1188 ;; be a noop?), and only set esp to its final value after the
1189 ;; stack-stack memmove loop. Otherwise, an unlucky signal
1190 ;; could end up overwriting the &more arguments before they're
1191 ;; moved in their final place.
1198 fixed))))))
1202 ;; Now: nargs>=1 && nargs>fixed
1204 ;; Save the original count of args.
1208 ;; We must stop when we run out of stack args, not when we
1209 ;; run out of more args.
1210 ;; Number to copy = nargs-3
1212 ;; Everything of interest in registers.
1215 ;; Number to copy = nargs-fixed
1220 fixed)))
1223 ;; nothing to move!
1224 )
1226 ;; stack frame smaller than fixed; moving args to higher
1227 ;; addresses (stack grows downard), so copy from the
1228 ;; end. Moreover, because we'd have to shrink the frame,
1229 ;; esp currently points at the end of the source args.
1236 ;; compensate for PUSH above
1240 ;; compensate for PUSH, and
1241 ;; add (abs delta)
1243 ebx-tn)
1248 ;; stack frame larger than fixed. Moving args to lower
1249 ;; addresses, so copy from the lowest address. esp
1250 ;; already points to the lowest address of the destination.
1258 ;; PUSHed 2 words
1260 delta)))
1264 esi-tn)
1271 DO-REGS
1272 ;; stack can now be set to its final size
1278 ;; Restore ECX
1281 ;; Here: nargs>=1 && nargs>fixed
1283 ;; Now we have to deposit any more args that showed up in
1284 ;; registers.
1287 ;; Store it relative to ebp
1301 ;; Don't deposit any more than there are.
1309 JUST-ALLOC-FRAME
1316 DONE))
1363 ;;; Turn more arg (context, count) into a list.
1383 ;; Check to see whether there are no args, and just return NIL if so.
1389 ;; Set decrement mode (successive args at lower addresses)
1391 ;; Set up the result.
1393 ;; Jump into the middle of the loop, 'cause that's where we want
1394 ;; to start.
1397 ;; Compute a pointer to the next cons.
1399 ;; Store a pointer to this cons in the CDR of the previous cons.
1402 ;; Grab one value and stash it in the car of this cons.
1405 ;; Go back for more.
1408 ;; NIL out the last cons.
1413 ;;; Return the location and size of the &MORE arg glob created by
1414 ;;; COPY-MORE-ARG. SUPPLIED is the total number of arguments supplied
1415 ;;; (originally passed in ECX). FIXED is the number of non-rest
1416 ;;; arguments.
1417 ;;;
1418 ;;; We must duplicate some of the work done by COPY-MORE-ARG, since at
1419 ;;; that time the environment is in a pretty brain-damaged state,
1420 ;;; preventing this info from being returned as values. What we do is
1421 ;;; compute supplied - fixed, and return a pointer that many words
1422 ;;; below the current stack top.
1435 ;; SP at this point points at the last arg pushed.
1436 ;; Point to the first more-arg, not above it.
1443 ;;; Signal wrong argument count error if NARGS isn't equal to COUNT.
1469 ;;; Various other error signallers.
1477 args))
1485 object type)
1487 object layout)
1494 ;;; Single-stepping
1497 ;; We use different ways of representing whether stepping is on on
1498 ;; +SB-THREAD / -SB-THREAD: on +SB-THREAD, we use a slot in the
1499 ;; thread structure. On -SB-THREAD we use the value of a static
1500 ;; symbol. Things are done this way, since reading a thread-local
1501 ;; slot from a symbol would require an extra register on +SB-THREAD,
1502 ;; and reading a slot from a thread structure would require an extra
1503 ;; register on -SB-THREAD.
1504 #!+sb-thread
1512 #!-sb-thread
1514 nil-value))
1523 DONE