add qsort_r and make qsort a wrapper around it

we make qsort a wrapper by providing a wrapper_cmp function that uses
the extra argument as a function pointer. should be optimized to a tail
call on most architectures, as long as it's built with
-fomit-frame-pointer, so the performance impact should be minimal.

to keep the git history clean, for now qsort_r is implemented in qsort.c
and qsort is implemented in qsort_nr.c.  qsort.c also received a few
trivial cleanups, including replacing (*cmp)() calls with cmp().
qsort_nr.c contains only wrapper_cmp and qsort as a qsort_r wrapper
itself.

1 files changed, 20 insertions, 17 deletions

diff --git a/src/stdlib/qsort.c b/src/stdlib/qsort.c
index da58fd31..314ddc29 100644
--- a/src/stdlib/qsort.c
+++ b/src/stdlib/qsort.c
@@ -24,6 +24,7 @@
 /* Smoothsort, an adaptive variant of Heapsort.  Memory usage: O(1).
    Run time: Worst case O(n log n), close to O(n) in the mostly-sorted case. */
 
+#define _BSD_SOURCE
 #include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
@@ -31,7 +32,7 @@
 #include "atomic.h"
 #define ntz(x) a_ctz_l((x))
 
-typedef int (*cmpfun)(const void *, const void *);
+typedef int (*cmpfun)(const void *, const void *, void *);
 
 static inline int pntz(size_t p[2]) {
 	int r = ntz(p[0] - 1);
@@ -88,7 +89,7 @@ static inline void shr(size_t p[2], int n)
 	p[1] >>= n;
 }
 
-static void sift(unsigned char *head, size_t width, cmpfun cmp, int pshift, size_t lp[])
+static void sift(unsigned char *head, size_t width, cmpfun cmp, void *arg, int pshift, size_t lp[])
 {
 	unsigned char *rt, *lf;
 	unsigned char *ar[14 * sizeof(size_t) + 1];
@@ -99,10 +100,10 @@ static void sift(unsigned char *head, size_t width, cmpfun cmp, int pshift, size
 		rt = head - width;
 		lf = head - width - lp[pshift - 2];
 
-		if((*cmp)(ar[0], lf) >= 0 && (*cmp)(ar[0], rt) >= 0) {
+		if(cmp(ar[0], lf, arg) >= 0 && cmp(ar[0], rt, arg) >= 0) {
 			break;
 		}
-		if((*cmp)(lf, rt) >= 0) {
+		if(cmp(lf, rt, arg) >= 0) {
 			ar[i++] = lf;
 			head = lf;
 			pshift -= 1;
@@ -115,7 +116,7 @@ static void sift(unsigned char *head, size_t width, cmpfun cmp, int pshift, size
 	cycle(width, ar, i);
 }
 
-static void trinkle(unsigned char *head, size_t width, cmpfun cmp, size_t pp[2], int pshift, int trusty, size_t lp[])
+static void trinkle(unsigned char *head, size_t width, cmpfun cmp, void *arg, size_t pp[2], int pshift, int trusty, size_t lp[])
 {
 	unsigned char *stepson,
 	              *rt, *lf;
@@ -130,13 +131,13 @@ static void trinkle(unsigned char *head, size_t width, cmpfun cmp, size_t pp[2],
 	ar[0] = head;
 	while(p[0] != 1 || p[1] != 0) {
 		stepson = head - lp[pshift];
-		if((*cmp)(stepson, ar[0]) <= 0) {
+		if(cmp(stepson, ar[0], arg) <= 0) {
 			break;
 		}
 		if(!trusty && pshift > 1) {
 			rt = head - width;
 			lf = head - width - lp[pshift - 2];
-			if((*cmp)(rt, stepson) >= 0 || (*cmp)(lf, stepson) >= 0) {
+			if(cmp(rt, stepson, arg) >= 0 || cmp(lf, stepson, arg) >= 0) {
 				break;
 			}
 		}
@@ -150,11 +151,11 @@ static void trinkle(unsigned char *head, size_t width, cmpfun cmp, size_t pp[2],
 	}
 	if(!trusty) {
 		cycle(width, ar, i);
-		sift(head, width, cmp, pshift, lp);
+		sift(head, width, cmp, arg, pshift, lp);
 	}
 }
 
-void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
+void __qsort_r(void *base, size_t nel, size_t width, cmpfun cmp, void *arg)
 {
 	size_t lp[12*sizeof(size_t)];
 	size_t i, size = width * nel;
@@ -173,16 +174,16 @@ void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
 
 	while(head < high) {
 		if((p[0] & 3) == 3) {
-			sift(head, width, cmp, pshift, lp);
+			sift(head, width, cmp, arg, pshift, lp);
 			shr(p, 2);
 			pshift += 2;
 		} else {
 			if(lp[pshift - 1] >= high - head) {
-				trinkle(head, width, cmp, p, pshift, 0, lp);
+				trinkle(head, width, cmp, arg, p, pshift, 0, lp);
 			} else {
-				sift(head, width, cmp, pshift, lp);
+				sift(head, width, cmp, arg, pshift, lp);
 			}
-			
+
 			if(pshift == 1) {
 				shl(p, 1);
 				pshift = 0;
@@ -191,12 +192,12 @@ void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
 				pshift = 1;
 			}
 		}
-		
+
 		p[0] |= 1;
 		head += width;
 	}
 
-	trinkle(head, width, cmp, p, pshift, 0, lp);
+	trinkle(head, width, cmp, arg, p, pshift, 0, lp);
 
 	while(pshift != 1 || p[0] != 1 || p[1] != 0) {
 		if(pshift <= 1) {
@@ -208,11 +209,13 @@ void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
 			pshift -= 2;
 			p[0] ^= 7;
 			shr(p, 1);
-			trinkle(head - lp[pshift] - width, width, cmp, p, pshift + 1, 1, lp);
+			trinkle(head - lp[pshift] - width, width, cmp, arg, p, pshift + 1, 1, lp);
 			shl(p, 1);
 			p[0] |= 1;
-			trinkle(head - width, width, cmp, p, pshift, 1, lp);
+			trinkle(head - width, width, cmp, arg, p, pshift, 1, lp);
 		}
 		head -= width;
 	}
 }
+
+weak_alias(__qsort_r, qsort_r);