Skip to content

Commit d861d27

Browse files
jartjart
committed
Import OpenBSD sorting algorithms
OpenBSD's qsort() function is more secure than the ones used by FreeBSD, NetBSD and MacOS. The best part is it goes faster too! This change also imports the OpenBSD mergesort() and heapsort()
1 parent dbf12c3 commit d861d27

File tree

11 files changed

+912
-215
lines changed

11 files changed

+912
-215
lines changed

‎libc/calls/timespec_add.c renamed to ‎libc/calls/_timespec_add.c

File renamed without changes.

‎libc/calls/timespec_get.c renamed to ‎libc/calls/_timespec_get.c

File renamed without changes.

‎libc/calls/timespec_getres.c renamed to ‎libc/calls/_timespec_getres.c

File renamed without changes.

‎libc/calls/timespec_sub.c renamed to ‎libc/calls/_timespec_sub.c

File renamed without changes.

‎libc/mem/alg.h

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -10,6 +10,8 @@ void *bsearch_r(const void *, const void *, size_t, size_t,
1010
int (*)(const void *, const void *, void *), void *)
1111
paramsnonnull((1, 2, 5)) dontthrow nosideeffect;
1212
void djbsort(int32_t *, size_t);
13+
void qsort3(void *, size_t, size_t, int (*)(const void *, const void *))
14+
paramsnonnull();
1315
void qsort(void *, size_t, size_t, int (*)(const void *, const void *))
1416
paramsnonnull();
1517
void qsort_r(void *, size_t, size_t,
@@ -18,6 +20,12 @@ void qsort_r(void *, size_t, size_t,
1820
void smoothsort(void *, size_t, size_t, int (*)(const void *, const void *));
1921
void smoothsort_r(void *, size_t, size_t,
2022
int (*)(const void *, const void *, void *), void *);
23+
int heapsort(void *, size_t, size_t, int (*)(const void *, const void *));
24+
int heapsort_r(void *, size_t, size_t,
25+
int (*)(const void *, const void *, void *), void *);
26+
int mergesort(void *, size_t, size_t, int (*)(const void *, const void *));
27+
int mergesort_r(void *, size_t, size_t,
28+
int (*)(const void *, const void *, void *), void *);
2129
int tarjan(int, const int (*)[2], int, int[], int[], int *)
2230
paramsnonnull((2, 4)) nocallback dontthrow;
2331

‎libc/mem/heapsort.c

Lines changed: 198 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,198 @@
1+
/*-*- mode:c;indent-tabs-mode:t;c-basic-offset:8;tab-width:8;coding:utf-8 -*-│
2+
│vi: set et ft=c ts=8 tw=8 fenc=utf-8 :vi│
3+
╞══════════════════════════════════════════════════════════════════════════════╡
4+
│ Copyright (c) 1991, 1993 │
5+
│ The Regents of the University of California. All rights reserved. │
6+
│ │
7+
│ Redistribution and use in source and binary forms, with or without │
8+
│ modification, are permitted provided that the following conditions │
9+
│ are met: │
10+
│ 1. Redistributions of source code must retain the above copyright │
11+
│ notice, this list of conditions and the following disclaimer. │
12+
│ 2. Redistributions in binary form must reproduce the above copyright │
13+
│ notice, this list of conditions and the following disclaimer in the │
14+
│ documentation and/or other materials provided with the distribution. │
15+
│ 3. Neither the name of the University nor the names of its contributors │
16+
│ may be used to endorse or promote products derived from this software │
17+
│ without specific prior written permission. │
18+
│ │
19+
│ THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND │
20+
│ ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE │
21+
│ IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE │
22+
│ ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE │
23+
│ FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL │
24+
│ DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS │
25+
│ OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) │
26+
│ HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT │
27+
│ LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY │
28+
│ OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF │
29+
│ SUCH DAMAGE. │
30+
╚─────────────────────────────────────────────────────────────────────────────*/
31+
#include "libc/mem/alg.h"
32+
#include "libc/mem/mem.h"
33+
#include "libc/sysv/errfuns.h"
34+
// clang-format off
35+
36+
/*
37+
* Swap two areas of size number of bytes. Although qsort(3) permits random
38+
* blocks of memory to be sorted, sorting pointers is almost certainly the
39+
* common case (and, were it not, could easily be made so). Regardless, it
40+
* isn't worth optimizing; the SWAP's get sped up by the cache, and pointer
41+
* arithmetic gets lost in the time required for comparison function calls.
42+
*/
43+
#define SWAP(a, b, count, size, tmp) { \
44+
count = size; \
45+
do { \
46+
tmp = *a; \
47+
*a++ = *b; \
48+
*b++ = tmp; \
49+
} while (--count); \
50+
}
51+
52+
/* Copy one block of size size to another. */
53+
#define COPY(a, b, count, size, tmp1, tmp2) { \
54+
count = size; \
55+
tmp1 = a; \
56+
tmp2 = b; \
57+
do { \
58+
*tmp1++ = *tmp2++; \
59+
} while (--count); \
60+
}
61+
62+
/*
63+
* Build the list into a heap, where a heap is defined such that for
64+
* the records K1 ... KN, Kj/2 >= Kj for 1 <= j/2 <= j <= N.
65+
*
66+
* There are two cases. If j == nmemb, select largest of Ki and Kj. If
67+
* j < nmemb, select largest of Ki, Kj and Kj+1.
68+
*/
69+
#define CREATE(initval, nmemb, par_i, child_i, par, child, size, count, tmp) { \
70+
for (par_i = initval; (child_i = par_i * 2) <= nmemb; \
71+
par_i = child_i) { \
72+
child = base + child_i * size; \
73+
if (child_i < nmemb && compar(child, child + size, z) < 0) { \
74+
child += size; \
75+
++child_i; \
76+
} \
77+
par = base + par_i * size; \
78+
if (compar(child, par, z) <= 0) \
79+
break; \
80+
SWAP(par, child, count, size, tmp); \
81+
} \
82+
}
83+
84+
/*
85+
* Select the top of the heap and 'heapify'. Since by far the most expensive
86+
* action is the call to the compar function, a considerable optimization
87+
* in the average case can be achieved due to the fact that k, the displaced
88+
* element, is usually quite small, so it would be preferable to first
89+
* heapify, always maintaining the invariant that the larger child is copied
90+
* over its parent's record.
91+
*
92+
* Then, starting from the *bottom* of the heap, finding k's correct place,
93+
* again maintaining the invariant. As a result of the invariant no element
94+
* is 'lost' when k is assigned its correct place in the heap.
95+
*
96+
* The time savings from this optimization are on the order of 15-20% for the
97+
* average case. See Knuth, Vol. 3, page 158, problem 18.
98+
*
99+
* XXX Don't break the #define SELECT line, below. Reiser cpp gets upset.
100+
*/
101+
#define SELECT(par_i, child_i, nmemb, par, child, size, k, count, tmp1, tmp2) { \
102+
for (par_i = 1; (child_i = par_i * 2) <= nmemb; par_i = child_i) { \
103+
child = base + child_i * size; \
104+
if (child_i < nmemb && compar(child, child + size, z) < 0) { \
105+
child += size; \
106+
++child_i; \
107+
} \
108+
par = base + par_i * size; \
109+
COPY(par, child, count, size, tmp1, tmp2); \
110+
} \
111+
for (;;) { \
112+
child_i = par_i; \
113+
par_i = child_i / 2; \
114+
child = base + child_i * size; \
115+
par = base + par_i * size; \
116+
if (child_i == 1 || compar(k, par, z) < 0) { \
117+
COPY(child, k, count, size, tmp1, tmp2); \
118+
break; \
119+
} \
120+
COPY(child, par, count, size, tmp1, tmp2); \
121+
} \
122+
}
123+
124+
/**
125+
* Sorts array w/ optional callback argument.
126+
*
127+
* @param vbase is base of array
128+
* @param nmemb is item count
129+
* @param size is item width
130+
* @param compar is a callback returning <0, 0, or >0
131+
* @param z will optionally be passed as the third argument to cmp
132+
* @see heapsort()
133+
*/
134+
int
135+
heapsort_r(void *vbase, size_t nmemb, size_t size,
136+
int (*compar)(const void *, const void *, void *), void *z)
137+
{
138+
size_t cnt, i, j, l;
139+
char tmp, *tmp1, *tmp2;
140+
char *base, *k, *p, *t;
141+
142+
if (nmemb <= 1)
143+
return (0);
144+
145+
if (!size)
146+
return (einval());
147+
148+
if ((k = malloc(size)) == NULL)
149+
return (-1);
150+
151+
/*
152+
* Items are numbered from 1 to nmemb, so offset from size bytes
153+
* below the starting address.
154+
*/
155+
base = (char *)vbase - size;
156+
157+
for (l = nmemb / 2 + 1; --l;)
158+
CREATE(l, nmemb, i, j, t, p, size, cnt, tmp);
159+
160+
/*
161+
* For each element of the heap, save the largest element into its
162+
* final slot, save the displaced element (k), then recreate the
163+
* heap.
164+
*/
165+
while (nmemb > 1) {
166+
COPY(k, base + nmemb * size, cnt, size, tmp1, tmp2);
167+
COPY(base + nmemb * size, base + size, cnt, size, tmp1, tmp2);
168+
--nmemb;
169+
SELECT(i, j, nmemb, t, p, size, k, cnt, tmp1, tmp2);
170+
}
171+
free(k);
172+
return (0);
173+
}
174+
175+
/**
176+
* Sorts array.
177+
*
178+
* Runs in O (N lg N), both average and worst. While heapsort is faster
179+
* than the worst case of quicksort, the BSD quicksort does median
180+
* selection so that the chance of finding a data set that will trigger
181+
* the worst case is nonexistent. Heapsort's only advantage over
182+
* quicksort is that it requires little additional memory.
183+
*
184+
* @param vbase is base of array
185+
* @param nmemb is item count
186+
* @param size is item width
187+
* @param compar is a callback returning <0, 0, or >0
188+
* @see Knuth, Vol. 3, page 145.
189+
* @see heapsort_r()
190+
* @see mergesort()
191+
* @see qsort()
192+
*/
193+
int
194+
heapsort(void *vbase, size_t nmemb, size_t size,
195+
int (*compar)(const void *, const void *))
196+
{
197+
return heapsort_r(vbase, nmemb, size, (void *)compar, 0);
198+
}

0 commit comments

Comments
 (0)