Annotation of OpenXM/src/kan96xx/gmp-2.0.2/mpn/tests/mul_1.c, Revision 1.1.1.1
1.1 maekawa 1: #include <stdio.h>
2: #include "gmp.h"
3: #include "gmp-impl.h"
4: #include "longlong.h"
5:
6: #ifndef USG
7: #include <sys/time.h>
8: #include <sys/resource.h>
9:
10: unsigned long
11: cputime ()
12: {
13: struct rusage rus;
14:
15: getrusage (0, &rus);
16: return rus.ru_utime.tv_sec * 1000 + rus.ru_utime.tv_usec / 1000;
17: }
18: #else
19: #include <time.h>
20:
21: #ifndef CLOCKS_PER_SEC
22: #define CLOCKS_PER_SEC 1000000
23: #endif
24:
25: #if CLOCKS_PER_SEC >= 10000
26: #define CLOCK_TO_MILLISEC(cl) ((cl) / (CLOCKS_PER_SEC / 1000))
27: #else
28: #define CLOCK_TO_MILLISEC(cl) ((cl) * 1000 / CLOCKS_PER_SEC)
29: #endif
30:
31: unsigned long
32: cputime ()
33: {
34: return CLOCK_TO_MILLISEC (clock ());
35: }
36: #endif
37:
38: #define M * 1000000
39:
40: #ifndef CLOCK
41: #if defined (__m88k__)
42: #define CLOCK 20 M
43: #elif defined (__i386__)
44: #define CLOCK (16.666667 M)
45: #elif defined (__m68k__)
46: #define CLOCK (20 M)
47: #elif defined (_IBMR2)
48: #define CLOCK (25 M)
49: #elif defined (__sparc__)
50: #define CLOCK (20 M)
51: #elif defined (__sun__)
52: #define CLOCK (20 M)
53: #elif defined (__mips)
54: #define CLOCK (40 M)
55: #elif defined (__hppa__)
56: #define CLOCK (50 M)
57: #elif defined (__alpha)
58: #define CLOCK (133 M)
59: #else
60: #error "Don't know CLOCK of your machine"
61: #endif
62: #endif
63:
64: #ifndef OPS
65: #define OPS 20000000
66: #endif
67: #ifndef SIZE
68: #define SIZE 496
69: #endif
70: #ifndef TIMES
71: #define TIMES OPS/SIZE
72: #else
73: #undef OPS
74: #define OPS (SIZE*TIMES)
75: #endif
76:
77: mp_limb_t
78: refmpn_mul_1 (res_ptr, s1_ptr, s1_size, s2_limb)
79: register mp_ptr res_ptr;
80: register mp_srcptr s1_ptr;
81: mp_size_t s1_size;
82: register mp_limb_t s2_limb;
83: {
84: register mp_limb_t cy_limb;
85: register mp_size_t j;
86: register mp_limb_t prod_high, prod_low;
87:
88: /* The loop counter and index J goes from -S1_SIZE to -1. This way
89: the loop becomes faster. */
90: j = -s1_size;
91:
92: /* Offset the base pointers to compensate for the negative indices. */
93: s1_ptr -= j;
94: res_ptr -= j;
95:
96: cy_limb = 0;
97: do
98: {
99: umul_ppmm (prod_high, prod_low, s1_ptr[j], s2_limb);
100:
101: prod_low += cy_limb;
102: cy_limb = (prod_low < cy_limb) + prod_high;
103:
104: res_ptr[j] = prod_low;
105: }
106: while (++j != 0);
107:
108: return cy_limb;
109: }
110:
111: main (argc, argv)
112: int argc;
113: char **argv;
114: {
115: mp_limb_t s1[SIZE];
116: mp_limb_t dx[SIZE+2];
117: mp_limb_t dy[SIZE+2];
118: mp_limb_t cyx, cyy;
119: int i;
120: long t0, t;
121: int test;
122: mp_limb_t xlimb;
123: mp_size_t size;
124:
125: for (test = 0; ; test++)
126: {
127: #ifdef RANDOM
128: size = (random () % SIZE + 1);
129: #else
130: size = SIZE;
131: #endif
132:
133: mpn_random2 (s1, size);
134: mpn_random2 (dy+1, size);
135:
136: if (random () % 0x100 == 0)
137: xlimb = 0;
138: else
139: mpn_random2 (&xlimb, 1);
140:
141: dy[size+1] = 0x12345678;
142: dy[0] = 0x87654321;
143:
144: #if defined (PRINT) || defined (XPRINT)
145: printf ("xlimb=%*lX\n", (int) (2 * sizeof(mp_limb_t)), xlimb);
146: #endif
147: #ifdef PRINT
148: mpn_print (s1, size);
149: #endif
150:
151: MPN_COPY (dx, dy, size+2);
152: t0 = cputime();
153: for (i = 0; i < TIMES; i++)
154: cyx = refmpn_mul_1 (dx+1, s1, size, xlimb);
155: t = cputime() - t0;
156: #if TIMES != 1
157: printf ("refmpn_mul_1: %5ldms (%.2f cycles/limb)\n",
158: t,
159: ((double) t * CLOCK) / (OPS * 1000.0));
160: #endif
161:
162: MPN_COPY (dx, dy, size+2);
163: t0 = cputime();
164: for (i = 0; i < TIMES; i++)
165: cyy = mpn_mul_1 (dx+1, s1, size, xlimb);
166: t = cputime() - t0;
167: #if TIMES != 1
168: printf ("mpn_mul_1: %5ldms (%.2f cycles/limb)\n",
169: t,
170: ((double) t * CLOCK) / (OPS * 1000.0));
171: #endif
172:
173: cyx = refmpn_mul_1 (dx+1, s1, size, xlimb);
174: cyy = mpn_mul_1 (dy+1, s1, size, xlimb);
175:
176: #ifdef PRINT
177: printf ("%*lX ", (int) (2 * sizeof(mp_limb_t)), cyx);
178: mpn_print (dx+1, size);
179: printf ("%*lX ", (int) (2 * sizeof(mp_limb_t)), cyy);
180: mpn_print (dy+1, size);
181: #endif
182:
183: #ifndef NOCHECK
184: if (cyx != cyy || mpn_cmp (dx, dy, size+2) != 0
185: || dx[size+1] != 0x12345678 || dx[0] != 0x87654321)
186: {
187: #ifndef PRINT
188: printf ("%*lX ", (int) (2 * sizeof(mp_limb_t)), cyx);
189: mpn_print (dx+1, size);
190: printf ("%*lX ", (int) (2 * sizeof(mp_limb_t)), cyy);
191: mpn_print (dy+1, size);
192: #endif
193: abort();
194: }
195: #endif
196: }
197: }
198:
199: mpn_print (mp_ptr p, mp_size_t size)
200: {
201: mp_size_t i;
202:
203: for (i = size - 1; i >= 0; i--)
204: {
205: printf ("%0*lX", (int) (2 * sizeof(mp_limb_t)), p[i]);
206: #ifdef SPACE
207: if (i != 0)
208: printf (" ");
209: #endif
210: }
211: puts ("");
212: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>