;;;; This file implements the stack analysis phase in the compiler. We
-;;;; do a graph walk to determine which unknown-values continuations
-;;;; are on the stack at each point in the program, and then we insert
+;;;; do a graph walk to determine which unknown-values lvars are on
+;;;; the stack at each point in the program, and then we insert
;;;; cleanup code to pop off unused values.
;;;; This software is part of the SBCL system. See the README file for
;;;; files for more information.
(in-package "SB!C")
-
-(file-comment
- "$Header$")
\f
-;;; Scan through Block looking for uses of :Unknown continuations that have
-;;; their Dest outside of the block. We do some checking to verify the
+;;; Scan through BLOCK looking for uses of :UNKNOWN lvars that have
+;;; their DEST outside of the block. We do some checking to verify the
;;; invariant that all pushes come after the last pop.
-(defun find-pushed-continuations (block)
+(defun find-pushed-lvars (block)
(let* ((2block (block-info block))
(popped (ir2-block-popped 2block))
(last-pop (if popped
- (continuation-dest (car (last popped)))
+ (lvar-dest (car (last popped)))
nil)))
(collect ((pushed))
(let ((saw-last nil))
- (do-nodes (node cont block)
+ (do-nodes (node lvar block)
(when (eq node last-pop)
(setq saw-last t))
- (let ((dest (continuation-dest cont))
- (2cont (continuation-info cont)))
- (when (and dest
- (not (eq (node-block dest) block))
- 2cont
- (eq (ir2-continuation-kind 2cont) :unknown))
- (assert (or saw-last (not last-pop)))
- (pushed cont)))))
+ (when lvar
+ (let ((dest (lvar-dest lvar))
+ (2lvar (lvar-info lvar)))
+ (when (and (not (eq (node-block dest) block))
+ 2lvar
+ (eq (ir2-lvar-kind 2lvar) :unknown))
+ (aver (or saw-last (not last-pop)))
+ (pushed lvar))))))
(setf (ir2-block-pushed 2block) (pushed))))
(values))
\f
;;;; annotation graph walk
-;;; Do a backward walk in the flow graph simulating the run-time stack of
-;;; unknown-values continuations and annotating the blocks with the result.
+;;; Do a backward walk in the flow graph simulating the run-time stack
+;;; of unknown-values lvars and annotating the blocks with the result.
;;;
-;;; Block is the block that is currently being walked and Stack is the stack
-;;; of unknown-values continuations in effect immediately after block. We
-;;; simulate the stack by popping off the unknown-values generated by this
-;;; block (if any) and pushing the continuations for values received by this
-;;; block. (The role of push and pop are interchanged because we are doing a
-;;; backward walk.)
+;;; BLOCK is the block that is currently being walked and STACK is the
+;;; stack of unknown-values lvars in effect immediately after
+;;; block. We simulate the stack by popping off the unknown-values
+;;; generated by this block (if any) and pushing the lvars for
+;;; values received by this block. (The role of push and pop are
+;;; interchanged because we are doing a backward walk.)
;;;
-;;; If we run into a values generator whose continuation isn't on stack top,
-;;; then the receiver hasn't yet been reached on any walk to this use. In this
-;;; case, we ignore the push for now, counting on Annotate-Dead-Values to clean
-;;; it up if we discover that it isn't reachable at all.
+;;; If we run into a values generator whose lvar isn't on
+;;; stack top, then the receiver hasn't yet been reached on any walk
+;;; to this use. In this case, we ignore the push for now, counting on
+;;; Annotate-Dead-Values to clean it up if we discover that it isn't
+;;; reachable at all.
;;;
-;;; If our final stack isn't empty, then we walk all the predecessor blocks
-;;; that don't have all the continuations that we have on our Start-Stack on
-;;; their End-Stack. This is our termination condition for the graph walk. We
-;;; put the test around the recursive call so that the initial call to this
-;;; function will do something even though there isn't initially anything on
-;;; the stack.
+;;; If our final stack isn't empty, then we walk all the predecessor
+;;; blocks that don't have all the lvars that we have on our
+;;; START-STACK on their END-STACK. This is our termination condition
+;;; for the graph walk. We put the test around the recursive call so
+;;; that the initial call to this function will do something even
+;;; though there isn't initially anything on the stack.
;;;
-;;; We can use the tailp test, since the only time we want to bottom out
-;;; with a non-empty stack is when we intersect with another path from the same
-;;; top-level call to this function that has more values receivers on that
-;;; path. When we bottom out in this way, we are counting on
-;;; DISCARD-UNUSED-VALUES doing its thing.
+;;; We can use the tailp test, since the only time we want to bottom
+;;; out with a non-empty stack is when we intersect with another path
+;;; from the same top level call to this function that has more values
+;;; receivers on that path. When we bottom out in this way, we are
+;;; counting on DISCARD-UNUSED-VALUES doing its thing.
;;;
;;; When we do recurse, we check that predecessor's END-STACK is a
-;;; subsequence of our START-STACK. There may be extra stuff on the top
-;;; of our stack because the last path to the predecessor may have discarded
-;;; some values that we use. There may be extra stuff on the bottom of our
-;;; stack because this walk may be from a values receiver whose lifetime
-;;; encloses that of the previous walk.
+;;; subsequence of our START-STACK. There may be extra stuff on the
+;;; top of our stack because the last path to the predecessor may have
+;;; discarded some values that we use. There may be extra stuff on the
+;;; bottom of our stack because this walk may be from a values
+;;; receiver whose lifetime encloses that of the previous walk.
;;;
-;;; If a predecessor block is the component head, then it must be the case
-;;; that this is a NLX entry stub. If so, we just stop our walk, since the
-;;; stack at the exit point doesn't have anything to do with our stack.
+;;; If a predecessor block is the component head, then it must be the
+;;; case that this is a NLX entry stub. If so, we just stop our walk,
+;;; since the stack at the exit point doesn't have anything to do with
+;;; our stack.
(defun stack-simulation-walk (block stack)
(declare (type cblock block) (list stack))
(let ((2block (block-info block)))
(dolist (push (reverse (ir2-block-pushed 2block)))
(if (eq (car new-stack) push)
(pop new-stack)
- (assert (not (member push new-stack)))))
+ (aver (not (member push new-stack)))))
(dolist (pop (reverse (ir2-block-popped 2block)))
(push pop new-stack))
(when new-stack
(dolist (pred (block-pred block))
(if (eq pred (component-head (block-component block)))
- (assert (find block
- (environment-nlx-info (block-environment block))
- :key #'nlx-info-target))
+ (aver (find block
+ (physenv-nlx-info (block-physenv block))
+ :key #'nlx-info-target))
(let ((pred-stack (ir2-block-end-stack (block-info pred))))
(unless (tailp new-stack pred-stack)
- (assert (search pred-stack new-stack))
+ (aver (search pred-stack new-stack))
(stack-simulation-walk pred new-stack))))))))
(values))
;;; Do stack annotation for any values generators in Block that were
-;;; unreached by all walks (i.e. the continuation isn't live at the point that
+;;; unreached by all walks (i.e. the lvar isn't live at the point that
;;; it is generated.) This will only happen when the values receiver cannot be
;;; reached from this particular generator (due to an unconditional control
;;; transfer.)
;;;
-;;; What we do is push on the End-Stack all continuations in Pushed that
+;;; What we do is push on the End-Stack all lvars in Pushed that
;;; aren't already present in the End-Stack. When we find any pushed
-;;; continuation that isn't live, it must be the case that all continuations
+;;; lvar that isn't live, it must be the case that all lvars
;;; pushed after (on top of) it aren't live.
;;;
-;;; If we see a pushed continuation that is the CONT of a tail call, then we
-;;; ignore it, since the tail call didn't actually push anything. The tail
-;;; call must always the last in the block.
+;;; If we see a pushed lvar that is the LVAR of a tail call, then we
+;;; ignore it, since the tail call didn't actually push anything. The
+;;; tail call must always the last in the block.
(defun annotate-dead-values (block)
(declare (type cblock block))
(let* ((2block (block-info block))
(stack (ir2-block-end-stack 2block))
(last (block-last block))
- (tailp-cont (if (node-tail-p last) (node-cont last))))
+ (tailp-lvar (if (node-tail-p last) (node-lvar last))))
(do ((pushes (ir2-block-pushed 2block) (rest pushes))
(popping nil))
((null pushes))
(let ((push (first pushes)))
(cond ((member push stack)
- (assert (not popping)))
- ((eq push tailp-cont)
- (assert (null (rest pushes))))
+ (aver (not popping)))
+ ((eq push tailp-lvar)
+ (aver (null (rest pushes))))
(t
(push push (ir2-block-end-stack 2block))
(setq popping t))))))
(values))
\f
-;;; Called when we discover that the stack-top unknown-values continuation
-;;; at the end of Block1 is different from that at the start of Block2 (its
-;;; successor.)
+;;; This is called when we discover that the stack-top unknown-values
+;;; lvar at the end of BLOCK1 is different from that at the start of
+;;; BLOCK2 (its successor).
;;;
-;;; We insert a call to a funny function in a new cleanup block introduced
-;;; between Block1 and Block2. Since control analysis and LTN have already
-;;; run, we must do make an IR2 block, then do ADD-TO-EMIT-ORDER and
-;;; LTN-ANALYZE-BLOCK on the new block. The new block is inserted after Block1
-;;; in the emit order.
+;;; We insert a call to a funny function in a new cleanup block
+;;; introduced between BLOCK1 and BLOCK2. Since control analysis and
+;;; LTN have already run, we must do make an IR2 block, then do
+;;; ADD-TO-EMIT-ORDER and LTN-ANALYZE-BELATED-BLOCK on the new
+;;; block. The new block is inserted after BLOCK1 in the emit order.
;;;
-;;; If the control transfer between Block1 and Block2 represents a
-;;; tail-recursive return (:Deleted IR2-continuation) or a non-local exit, then
-;;; the cleanup code will never actually be executed. It doesn't seem to be
-;;; worth the risk of trying to optimize this, since this rarely happens and
+;;; If the control transfer between BLOCK1 and BLOCK2 represents a
+;;; tail-recursive return or a non-local exit, then the cleanup code
+;;; will never actually be executed. It doesn't seem to be worth the
+;;; risk of trying to optimize this, since this rarely happens and
;;; wastes only space.
(defun discard-unused-values (block1 block2)
(declare (type cblock block1 block2))
(- (length block1-stack)
(length block2-stack)
1))))
- (assert (tailp block2-stack block1-stack))
+ (aver (tailp block2-stack block1-stack))
(let* ((block (insert-cleanup-code block1 block2
- (continuation-next (block-start block2))
+ (block-start-node block2)
`(%pop-values ',last-popped)))
(2block (make-ir2-block block)))
(setf (block-info block) 2block)
(add-to-emit-order 2block (block-info block1))
- (ltn-analyze-block block)))
+ (ltn-analyze-belated-block block)))
(values))
\f
;;;; stack analysis
-;;; Return a list of all the blocks containing genuine uses of one of the
-;;; Receivers. Exits are excluded, since they don't drop through to the
-;;; receiver.
+;;; Return a list of all the blocks containing genuine uses of one of
+;;; the RECEIVERS. Exits are excluded, since they don't drop through
+;;; to the receiver.
(defun find-values-generators (receivers)
(declare (list receivers))
(collect ((res nil adjoin))
(res (node-block use))))))
(res)))
-;;; Analyze the use of unknown-values continuations in Component, inserting
-;;; cleanup code to discard values that are generated but never received. This
-;;; phase doesn't need to be run when Values-Receivers is null, i.e. there are
-;;; no unknown-values continuations used across block boundaries.
+;;; Analyze the use of unknown-values lvars in COMPONENT, inserting
+;;; cleanup code to discard values that are generated but never
+;;; received. This phase doesn't need to be run when Values-Receivers
+;;; is null, i.e. there are no unknown-values lvars used across block
+;;; boundaries.
;;;
-;;; Do the backward graph walk, starting at each values receiver. We ignore
-;;; receivers that already have a non-null Start-Stack. These are nested
-;;; values receivers that have already been reached on another walk. We don't
-;;; want to clobber that result with our null initial stack.
+;;; Do the backward graph walk, starting at each values receiver. We
+;;; ignore receivers that already have a non-null START-STACK. These
+;;; are nested values receivers that have already been reached on
+;;; another walk. We don't want to clobber that result with our null
+;;; initial stack.
(defun stack-analyze (component)
(declare (type component component))
(let* ((2comp (component-info component))
(generators (find-values-generators receivers)))
(dolist (block generators)
- (find-pushed-continuations block))
+ (find-pushed-lvars block))
(dolist (block receivers)
(unless (ir2-block-start-stack (block-info block))
projects / sbcl.git / blobdiff