public inbox for newlib-cvs@sourceware.org
help / color / mirror / Atom feed
* [newlib-cygwin] Ensure qsort recursion depth is bounded
@ 2018-03-16  9:32 Corinna Vinschen
  0 siblings, 0 replies; only message in thread
From: Corinna Vinschen @ 2018-03-16  9:32 UTC (permalink / raw)
  To: newlib-cvs

https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;h=0045445ad6f5d2271d677fa77f896b549485e581

commit 0045445ad6f5d2271d677fa77f896b549485e581
Author: Hakan Lindqvist <hakan.lindqvist@enea.com>
Date:   Mon Mar 12 13:51:07 2018 +0100

    Ensure qsort recursion depth is bounded
    
    The qsort algorithm splits the input array in three parts. The
    left and right parts may need further sorting. One of them is
    sorted by recursion, the other by iteration. This update ensures
    that it is the smaller part that is chosen for recursion.
    
    By choosing the smaller part, each recursion level will handle
    less than half the array of the previous recursion level. Hence
    the recursion depth is bounded to be less than log2(n) i.e. 1
    level per significant bit in the array size n.
    
    The update also includes code comments explaining the algorithm.

Diff:
---
 newlib/libc/search/qsort.c | 68 ++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 56 insertions(+), 12 deletions(-)

diff --git a/newlib/libc/search/qsort.c b/newlib/libc/search/qsort.c
index db3f589..4dc61be 100644
--- a/newlib/libc/search/qsort.c
+++ b/newlib/libc/search/qsort.c
@@ -173,15 +173,18 @@ qsort (void *a,
 	int cmp_result;
 	int swaptype, swap_cnt;
 
-loop:	SWAPINIT(a, es);
-	swap_cnt = 0;
+	SWAPINIT(a, es);
+loop:	swap_cnt = 0;
 	if (n < 7) {
+		/* Short arrays are insertion sorted. */
 		for (pm = (char *) a + es; pm < (char *) a + n * es; pm += es)
 			for (pl = pm; pl > (char *) a && CMP(thunk, pl - es, pl) > 0;
 			     pl -= es)
 				swap(pl, pl - es);
 		return;
 	}
+
+	/* Select a pivot element, move it to the left. */
 	pm = (char *) a + (n / 2) * es;
 	if (n > 7) {
 		pl = a;
@@ -195,11 +198,17 @@ loop:	SWAPINIT(a, es);
 		pm = med3(pl, pm, pn, cmp, thunk);
 	}
 	swap(a, pm);
-	pa = pb = (char *) a + es;
 
+	/*
+	 * Sort the array relative the pivot in four ranges as follows:
+	 * { elems == pivot, elems < pivot, elems > pivot, elems == pivot }
+	 */
+	pa = pb = (char *) a + es;
 	pc = pd = (char *) a + (n - 1) * es;
 	for (;;) {
+		/* Scan left to right stopping at first element > pivot. */
 		while (pb <= pc && (cmp_result = CMP(thunk, pb, a)) <= 0) {
+			/* Move elements == pivot to the left (to pa) */
 			if (cmp_result == 0) {
 				swap_cnt = 1;
 				swap(pa, pb);
@@ -207,7 +216,9 @@ loop:	SWAPINIT(a, es);
 			}
 			pb += es;
 		}
+		/* Scan right to left stopping at first element < pivot. */
 		while (pb <= pc && (cmp_result = CMP(thunk, pc, a)) >= 0) {
+			/* Move elements == pivot to the right (to pd) */
 			if (cmp_result == 0) {
 				swap_cnt = 1;
 				swap(pc, pd);
@@ -217,6 +228,7 @@ loop:	SWAPINIT(a, es);
 		}
 		if (pb > pc)
 			break;
+		/* The scan has found two elements to swap with each other. */
 		swap(pb, pc);
 		swap_cnt = 1;
 		pb += es;
@@ -230,24 +242,56 @@ loop:	SWAPINIT(a, es);
 		return;
 	}
 
+	/*
+	 * Rearrange the array in three parts sorted like this:
+	 * { elements < pivot, elements == pivot, elements > pivot }
+	 */
 	pn = (char *) a + n * es;
 	r = min(pa - (char *)a, pb - pa);
 	vecswap(a, pb - r, r);
 	r = min(pd - pc, pn - pd - es);
 	vecswap(pb, pn - r, r);
-	if ((r = pb - pa) > es)
+	d = pb - pa; /* d = Size of left part. */
+	r = pd - pc; /* r = Size of right part. */
+	pn -= r;     /* pn = Base of right part. */
+
+	/*
+	 * Check which of the left and right parts are larger.
+	 * Set (a, n)  to (base, size) of the larger part.
+	 * Set (pa, r) to (base, size) of the smaller part.
+	 */
+	if (r > d) { /* Right part is the larger part */
+		pa = a;
+		a = pn;
+		n = r;
+		r = d;
+	}
+	else { /* Left part is the larger part, or both are equal. */
+		pa = pn;
+		n = d;
+	}
+
+	/*
+	 * The left and right parts each need further sorting if they
+	 * contain two elements or more. If both need sorting we use
+	 * recursion to sort the smaller part and save the larger part
+	 * to be sorted by iteration after the recursion.
+	 * Using recursion only for the smaller part guarantees a
+	 * recursion depth that is bounded to be less than (log2(n)).
+	 */
+	if (r > es) {  /* Smaller part > 1 element. Both parts need sorting. */
+		/* Sort smaller part using recursion. */
 #if defined(I_AM_QSORT_R)
-		__bsd_qsort_r(a, r / es, es, thunk, cmp);
+		__bsd_qsort_r(pa, r / es, es, thunk, cmp);
 #elif defined(I_AM_GNU_QSORT_R)
-		qsort_r(a, r / es, es, cmp, thunk);
+		qsort_r(pa, r / es, es, cmp, thunk);
 #else
-		qsort(a, r / es, es, cmp);
+		qsort(pa, r / es, es, cmp);
 #endif
-	if ((r = pd - pc) > es) {
-		/* Iterate rather than recurse to save stack space */
-		a = pn - r;
-		n = r / es;
+	}
+	if (n > es) {  /* The larger part needs sorting. Iterate to sort.  */
+		n = n / es;
 		goto loop;
 	}
-/*		qsort(pn - r, r / es, es, cmp);*/
+	/* Both left and right parts are one element or less - level done. */
 }


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2018-03-16  9:32 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-03-16  9:32 [newlib-cygwin] Ensure qsort recursion depth is bounded Corinna Vinschen

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).