1 /* Copyright (C) 2011 by Valentin Ochs
2 *
3 * Permission is hereby granted, free of charge, to any person obtaining a copy
4 * of this software and associated documentation files (the "Software"), to
5 * deal in the Software without restriction, including without limitation the
6 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
7 * sell copies of the Software, and to permit persons to whom the Software is
8 * furnished to do so, subject to the following conditions:
9 *
10 * The above copyright notice and this permission notice shall be included in
11 * all copies or substantial portions of the Software.
12 *
13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
18 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
19 * IN THE SOFTWARE.
20 */
21
22 /* Minor changes by Rich Felker for integration in musl, 2011-04-27. */
23
24 /* Smoothsort, an adaptive variant of Heapsort. Memory usage: O(1).
25 Run time: Worst case O(n log n), close to O(n) in the mostly-sorted case. */
26
27 #include <stdint.h>
28 #include <stdlib.h>
29 #include <string.h>
30
31 #include "atomic.h"
32 #define ntz(x) a_ctz_l((x))
33
34 typedef int (*cmpfun)(const void *, const void *);
35
pntz(size_t p[2])36 static inline int pntz(size_t p[2]) {
37 int r = ntz(p[0] - 1);
38 if(r != 0 || (r = 8*sizeof(size_t) + ntz(p[1])) != 8*sizeof(size_t)) {
39 return r;
40 }
41 return 0;
42 }
43
cycle(size_t width,unsigned char * ar[],int n)44 static void cycle(size_t width, unsigned char* ar[], int n)
45 {
46 unsigned char tmp[256];
47 size_t l;
48 int i;
49
50 if(n < 2) {
51 return;
52 }
53
54 ar[n] = tmp;
55 while(width) {
56 l = sizeof(tmp) < width ? sizeof(tmp) : width;
57 memcpy(ar[n], ar[0], l);
58 for(i = 0; i < n; i++) {
59 memcpy(ar[i], ar[i + 1], l);
60 ar[i] += l;
61 }
62 width -= l;
63 }
64 }
65
66 /* shl() and shr() need n > 0 */
shl(size_t p[2],int n)67 static inline void shl(size_t p[2], int n)
68 {
69 if(n >= 8 * sizeof(size_t)) {
70 n -= 8 * sizeof(size_t);
71 p[1] = p[0];
72 p[0] = 0;
73 }
74 p[1] <<= n;
75 p[1] |= p[0] >> (sizeof(size_t) * 8 - n);
76 p[0] <<= n;
77 }
78
shr(size_t p[2],int n)79 static inline void shr(size_t p[2], int n)
80 {
81 if(n >= 8 * sizeof(size_t)) {
82 n -= 8 * sizeof(size_t);
83 p[0] = p[1];
84 p[1] = 0;
85 }
86 p[0] >>= n;
87 p[0] |= p[1] << (sizeof(size_t) * 8 - n);
88 p[1] >>= n;
89 }
90
sift(unsigned char * head,size_t width,cmpfun cmp,int pshift,size_t lp[])91 static void sift(unsigned char *head, size_t width, cmpfun cmp, int pshift, size_t lp[])
92 {
93 unsigned char *rt, *lf;
94 unsigned char *ar[14 * sizeof(size_t) + 1];
95 int i = 1;
96
97 ar[0] = head;
98 while(pshift > 1) {
99 rt = head - width;
100 lf = head - width - lp[pshift - 2];
101
102 if((*cmp)(ar[0], lf) >= 0 && (*cmp)(ar[0], rt) >= 0) {
103 break;
104 }
105 if((*cmp)(lf, rt) >= 0) {
106 ar[i++] = lf;
107 head = lf;
108 pshift -= 1;
109 } else {
110 ar[i++] = rt;
111 head = rt;
112 pshift -= 2;
113 }
114 }
115 cycle(width, ar, i);
116 }
117
trinkle(unsigned char * head,size_t width,cmpfun cmp,size_t pp[2],int pshift,int trusty,size_t lp[])118 static void trinkle(unsigned char *head, size_t width, cmpfun cmp, size_t pp[2], int pshift, int trusty, size_t lp[])
119 {
120 unsigned char *stepson,
121 *rt, *lf;
122 size_t p[2];
123 unsigned char *ar[14 * sizeof(size_t) + 1];
124 int i = 1;
125 int trail;
126
127 p[0] = pp[0];
128 p[1] = pp[1];
129
130 ar[0] = head;
131 while(p[0] != 1 || p[1] != 0) {
132 stepson = head - lp[pshift];
133 if((*cmp)(stepson, ar[0]) <= 0) {
134 break;
135 }
136 if(!trusty && pshift > 1) {
137 rt = head - width;
138 lf = head - width - lp[pshift - 2];
139 if((*cmp)(rt, stepson) >= 0 || (*cmp)(lf, stepson) >= 0) {
140 break;
141 }
142 }
143
144 ar[i++] = stepson;
145 head = stepson;
146 trail = pntz(p);
147 shr(p, trail);
148 pshift += trail;
149 trusty = 0;
150 }
151 if(!trusty) {
152 cycle(width, ar, i);
153 sift(head, width, cmp, pshift, lp);
154 }
155 }
156
qsort(void * base,size_t nel,size_t width,cmpfun cmp)157 void qsort(void *base, size_t nel, size_t width, cmpfun cmp)
158 {
159 size_t lp[12*sizeof(size_t)];
160 size_t i, size = width * nel;
161 unsigned char *head, *high;
162 size_t p[2] = {1, 0};
163 int pshift = 1;
164 int trail;
165
166 if (!size) return;
167
168 head = base;
169 high = head + size - width;
170
171 /* Precompute Leonardo numbers, scaled by element width */
172 for(lp[0]=lp[1]=width, i=2; (lp[i]=lp[i-2]+lp[i-1]+width) < size; i++);
173
174 while(head < high) {
175 if((p[0] & 3) == 3) {
176 sift(head, width, cmp, pshift, lp);
177 shr(p, 2);
178 pshift += 2;
179 } else {
180 if(lp[pshift - 1] >= high - head) {
181 trinkle(head, width, cmp, p, pshift, 0, lp);
182 } else {
183 sift(head, width, cmp, pshift, lp);
184 }
185
186 if(pshift == 1) {
187 shl(p, 1);
188 pshift = 0;
189 } else {
190 shl(p, pshift - 1);
191 pshift = 1;
192 }
193 }
194
195 p[0] |= 1;
196 head += width;
197 }
198
199 trinkle(head, width, cmp, p, pshift, 0, lp);
200
201 while(pshift != 1 || p[0] != 1 || p[1] != 0) {
202 if(pshift <= 1) {
203 trail = pntz(p);
204 shr(p, trail);
205 pshift += trail;
206 } else {
207 shl(p, 2);
208 pshift -= 2;
209 p[0] ^= 7;
210 shr(p, 1);
211 trinkle(head - lp[pshift] - width, width, cmp, p, pshift + 1, 1, lp);
212 shl(p, 1);
213 p[0] |= 1;
214 trinkle(head - width, width, cmp, p, pshift, 1, lp);
215 }
216 head -= width;
217 }
218 }
219