Coverage Report

Created: 2017-04-29 12:21

/Users/buildslave/jenkins/sharedspace/clang-stage2-coverage-R@2/llvm/tools/polly/lib/External/isl/isl_sort.c
Line
Count
Source (jump to first uncovered line)
1
/*
2
 * The code of this file was taken from http://jeffreystedfast.blogspot.be,
3
 * where it was posted in 2011 by Jeffrey Stedfast under the MIT license.
4
 * The MIT license text is as follows:
5
 *
6
 * Permission is hereby granted, free of charge, to any person obtaining a copy
7
 * of this software and associated documentation files (the "Software"), to
8
 * deal in the Software without restriction, including without limitation the
9
 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10
 * sell copies of the Software, and to permit persons to whom the Software is
11
 * furnished to do so, subject to the following conditions:
12
 *
13
 * The above copyright notice and this permission notice shall be included in
14
 * all copies or substantial portions of the Software.
15
 *
16
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
22
 * IN THE SOFTWARE.
23
 */
24
25
#include <errno.h>
26
#include <string.h>
27
#include <stdlib.h>
28
#include <isl_sort.h>
29
30
690k
#define MID(lo, hi) (lo + ((hi - lo) >> 1))
31
32
/* The code here is an optimized merge sort. Starting from a generic merge sort
33
 * the following optimizations were applied:
34
 *
35
 * o Batching of memcpy() calls: Instead of calling memcpy() to copy each and
36
 *   every element into a temporary buffer, blocks of elements are copied
37
 *   at a time.
38
 *
39
 * o To reduce the number of memcpy() calls further, copying leading
40
 *   and trailing elements into our temporary buffer is avoided, in case it is
41
 *   not necessary to merge them.
42
 *
43
 * A further optimization could be to specialize memcpy calls based on the
44
 * size of the types we compare. For now, this code does not include the
45
 * relevant optimization, as clang e.g. inlines a very efficient memcpy()
46
 * implementation. It is not clear, that the specialized version as provided in
47
 * the blog post, is really superior to the one that will be inlined by
48
 * default. So we decided to keep the code simple until this optimization was
49
 * proven to be beneficial.
50
 */
51
52
static void
53
msort (void *array, void *buf, size_t low, size_t high, size_t size,
54
       int (* compare) (const void *, const void *, void *), void *arg)
55
690k
{
56
690k
    char *a1, *al, *am, *ah, *ls, *hs, *lo, *hi, *b;
57
690k
    size_t copied = 0;
58
690k
    size_t mid;
59
690k
60
690k
    mid = MID (low, high);
61
690k
62
690k
    if (mid + 1 < high)
63
210k
        msort (array, buf, mid + 1, high, size, compare, arg);
64
690k
65
690k
    if (mid > low)
66
335k
        msort (array, buf, low, mid, size, compare, arg);
67
690k
68
690k
    ah = ((char *) array) + ((high + 1) * size);
69
690k
    am = ((char *) array) + ((mid + 1) * size);
70
690k
    a1 = al = ((char *) array) + (low * size);
71
690k
72
690k
    b = (char *) buf;
73
690k
    lo = al;
74
690k
    hi = am;
75
690k
76
769k
    do {
77
769k
        ls = lo;
78
769k
        hs = hi;
79
769k
80
769k
        if (
lo > al || 769k
hi > am707k
)
{78.4k
81
78.4k
            /* our last loop already compared lo & hi and found lo <= hi */
82
78.4k
            lo += size;
83
78.4k
        }
84
769k
85
1.93M
        while (
lo < am && 1.93M
compare (lo, hi, arg) <= 01.38M
)
86
1.17M
            lo += size;
87
769k
88
769k
        if (
lo < am769k
)
{215k
89
215k
            if (
copied == 0215k
)
{165k
90
165k
                /* avoid copying the leading items */
91
165k
                a1 = lo;
92
165k
                ls = lo;
93
165k
            }
94
215k
95
215k
            /* our last compare tells us hi < lo */
96
215k
            hi += size;
97
215k
98
360k
            while (
hi < ah && 360k
compare (hi, lo, arg) < 0223k
)
99
144k
                hi += size;
100
215k
101
215k
            if (
lo > ls215k
)
{50.7k
102
50.7k
                memcpy (b, ls, lo - ls);
103
50.7k
                copied += (lo - ls);
104
50.7k
                b += (lo - ls);
105
50.7k
            }
106
215k
107
215k
            memcpy (b, hs, hi - hs);
108
215k
            copied += (hi - hs);
109
215k
            b += (hi - hs);
110
553k
        } else 
if (553k
copied553k
)
{27.7k
111
27.7k
            memcpy (b, ls, lo - ls);
112
27.7k
            copied += (lo - ls);
113
27.7k
            b += (lo - ls);
114
27.7k
115
27.7k
            /* copy everything we needed to re-order back into array */
116
27.7k
            memcpy (a1, buf, copied);
117
27.7k
            return;
118
525k
        } else {
119
525k
            /* everything already in order */
120
525k
            return;
121
525k
        }
122
215k
    } while (hi < ah);
123
690k
124
137k
    
if (137k
lo < am137k
)
{137k
125
137k
        memcpy (b, lo, am - lo);
126
137k
        copied += (am - lo);
127
137k
    }
128
137k
129
137k
    memcpy (a1, buf, copied);
130
137k
}
131
132
static int
133
MergeSort (void *base, size_t nmemb, size_t size,
134
           int (* compare) (const void *, const void *, void *), void *arg)
135
172k
{
136
172k
    void *tmp;
137
172k
138
172k
    if (nmemb < 2)
139
26.8k
        return 0;
140
172k
141
145k
    
if (145k
!(tmp = malloc (nmemb * size))145k
)
{0
142
0
        errno = ENOMEM;
143
0
        return -1;
144
0
    }
145
145k
146
145k
    msort (base, tmp, 0, nmemb - 1, size, compare, arg);
147
145k
148
145k
    free (tmp);
149
145k
150
145k
    return 0;
151
145k
}
152
153
int isl_sort(void *const pbase, size_t total_elems, size_t size,
154
  int (*cmp)(const void *, const void *, void *arg), void *arg)
155
172k
{
156
172k
    return MergeSort (pbase, total_elems, size, cmp, arg);
157
172k
}