tcg/i386: Implement vector minmax arithmetic
[qemu/ar7.git] / util / host-utils.c
blob7b9322071d1d5a74d86dd15985a36a50deb13016
1 /*
2 * Utility compute operations used by translated code.
4 * Copyright (c) 2003 Fabrice Bellard
5 * Copyright (c) 2007 Aurelien Jarno
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23 * THE SOFTWARE.
26 #include "qemu/osdep.h"
27 #include "qemu/host-utils.h"
29 #ifndef CONFIG_INT128
30 /* Long integer helpers */
31 static inline void mul64(uint64_t *plow, uint64_t *phigh,
32 uint64_t a, uint64_t b)
34 typedef union {
35 uint64_t ll;
36 struct {
37 #ifdef HOST_WORDS_BIGENDIAN
38 uint32_t high, low;
39 #else
40 uint32_t low, high;
41 #endif
42 } l;
43 } LL;
44 LL rl, rm, rn, rh, a0, b0;
45 uint64_t c;
47 a0.ll = a;
48 b0.ll = b;
50 rl.ll = (uint64_t)a0.l.low * b0.l.low;
51 rm.ll = (uint64_t)a0.l.low * b0.l.high;
52 rn.ll = (uint64_t)a0.l.high * b0.l.low;
53 rh.ll = (uint64_t)a0.l.high * b0.l.high;
55 c = (uint64_t)rl.l.high + rm.l.low + rn.l.low;
56 rl.l.high = c;
57 c >>= 32;
58 c = c + rm.l.high + rn.l.high + rh.l.low;
59 rh.l.low = c;
60 rh.l.high += (uint32_t)(c >> 32);
62 *plow = rl.ll;
63 *phigh = rh.ll;
66 /* Unsigned 64x64 -> 128 multiplication */
67 void mulu64 (uint64_t *plow, uint64_t *phigh, uint64_t a, uint64_t b)
69 mul64(plow, phigh, a, b);
72 /* Signed 64x64 -> 128 multiplication */
73 void muls64 (uint64_t *plow, uint64_t *phigh, int64_t a, int64_t b)
75 uint64_t rh;
77 mul64(plow, &rh, a, b);
79 /* Adjust for signs. */
80 if (b < 0) {
81 rh -= a;
83 if (a < 0) {
84 rh -= b;
86 *phigh = rh;
89 /* Unsigned 128x64 division. Returns 1 if overflow (divide by zero or */
90 /* quotient exceeds 64 bits). Otherwise returns quotient via plow and */
91 /* remainder via phigh. */
92 int divu128(uint64_t *plow, uint64_t *phigh, uint64_t divisor)
94 uint64_t dhi = *phigh;
95 uint64_t dlo = *plow;
96 unsigned i;
97 uint64_t carry = 0;
99 if (divisor == 0) {
100 return 1;
101 } else if (dhi == 0) {
102 *plow = dlo / divisor;
103 *phigh = dlo % divisor;
104 return 0;
105 } else if (dhi > divisor) {
106 return 1;
107 } else {
109 for (i = 0; i < 64; i++) {
110 carry = dhi >> 63;
111 dhi = (dhi << 1) | (dlo >> 63);
112 if (carry || (dhi >= divisor)) {
113 dhi -= divisor;
114 carry = 1;
115 } else {
116 carry = 0;
118 dlo = (dlo << 1) | carry;
121 *plow = dlo;
122 *phigh = dhi;
123 return 0;
127 int divs128(int64_t *plow, int64_t *phigh, int64_t divisor)
129 int sgn_dvdnd = *phigh < 0;
130 int sgn_divsr = divisor < 0;
131 int overflow = 0;
133 if (sgn_dvdnd) {
134 *plow = ~(*plow);
135 *phigh = ~(*phigh);
136 if (*plow == (int64_t)-1) {
137 *plow = 0;
138 (*phigh)++;
139 } else {
140 (*plow)++;
144 if (sgn_divsr) {
145 divisor = 0 - divisor;
148 overflow = divu128((uint64_t *)plow, (uint64_t *)phigh, (uint64_t)divisor);
150 if (sgn_dvdnd ^ sgn_divsr) {
151 *plow = 0 - *plow;
154 if (!overflow) {
155 if ((*plow < 0) ^ (sgn_dvdnd ^ sgn_divsr)) {
156 overflow = 1;
160 return overflow;
162 #endif
165 * urshift - 128-bit Unsigned Right Shift.
166 * @plow: in/out - lower 64-bit integer.
167 * @phigh: in/out - higher 64-bit integer.
168 * @shift: in - bytes to shift, between 0 and 127.
170 * Result is zero-extended and stored in plow/phigh, which are
171 * input/output variables. Shift values outside the range will
172 * be mod to 128. In other words, the caller is responsible to
173 * verify/assert both the shift range and plow/phigh pointers.
175 void urshift(uint64_t *plow, uint64_t *phigh, int32_t shift)
177 shift &= 127;
178 if (shift == 0) {
179 return;
182 uint64_t h = *phigh >> (shift & 63);
183 if (shift >= 64) {
184 *plow = h;
185 *phigh = 0;
186 } else {
187 *plow = (*plow >> (shift & 63)) | (*phigh << (64 - (shift & 63)));
188 *phigh = h;
193 * ulshift - 128-bit Unsigned Left Shift.
194 * @plow: in/out - lower 64-bit integer.
195 * @phigh: in/out - higher 64-bit integer.
196 * @shift: in - bytes to shift, between 0 and 127.
197 * @overflow: out - true if any 1-bit is shifted out.
199 * Result is zero-extended and stored in plow/phigh, which are
200 * input/output variables. Shift values outside the range will
201 * be mod to 128. In other words, the caller is responsible to
202 * verify/assert both the shift range and plow/phigh pointers.
204 void ulshift(uint64_t *plow, uint64_t *phigh, int32_t shift, bool *overflow)
206 uint64_t low = *plow;
207 uint64_t high = *phigh;
209 shift &= 127;
210 if (shift == 0) {
211 return;
214 /* check if any bit will be shifted out */
215 urshift(&low, &high, 128 - shift);
216 if (low | high) {
217 *overflow = true;
220 if (shift >= 64) {
221 *phigh = *plow << (shift & 63);
222 *plow = 0;
223 } else {
224 *phigh = (*plow >> (64 - (shift & 63))) | (*phigh << (shift & 63));
225 *plow = *plow << shift;