X-Git-Url: https://git.gag.com/?p=fw%2Faltos;a=blobdiff_plain;f=src%2Fcore%2Fao_int64.c;fp=src%2Fcore%2Fao_int64.c;h=1fe717dcd6ca16fab19db928b3aa179448eb8a62;hp=0000000000000000000000000000000000000000;hb=3114baef45803250a2e5cdd2ee4a9171f2045b0c;hpb=d0b4e926ecececa7499a301b6135189be119512e diff --git a/src/core/ao_int64.c b/src/core/ao_int64.c new file mode 100644 index 00000000..1fe717dc --- /dev/null +++ b/src/core/ao_int64.c @@ -0,0 +1,119 @@ +/* + * Copyright © 2013 Keith Packard + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; version 2 of the License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA. + */ + +#include + +void ao_plus64(ao_int64_t *r, ao_int64_t *a, ao_int64_t *b) { + uint32_t t; + + r->low = t = a->low + b->low; + r->high = a->high + b->high; + if (t < a->low) + r->high++; +} + +void ao_rshift64(ao_int64_t *r, ao_int64_t *a, uint8_t d) { + if (d < 32) { + r->high = (int32_t) a->high >> d; + r->low = a->low >> d; + if (d) + r->low |= a->high << (32 - d); + } else { + d &= 0x1f; + r->high = 0; + r->low = (int32_t) a->high >> d; + } +} + +void ao_lshift64(ao_int64_t *r, ao_int64_t *a, uint8_t d) { + if (d < 32) { + r->high = a->high << d; + r->low = a->low << d; + if (d) + r->high |= a->low >> (32 - d); + } else { + d &= 0x1f; + r->low = 0; + r->high = a->low << d; + } +} + +void ao_umul64(ao_int64_t *r, uint32_t a, uint32_t b) +{ + uint32_t r1; + uint32_t r2, r3, r4; + ao_int64_t s,t,u,v; + r1 = (uint32_t) (uint16_t) a * (uint16_t) b; + r2 = (uint32_t) (uint16_t) (a >> 16) * (uint16_t) b; + r3 = (uint32_t) (uint16_t) a * (uint16_t) (b >> 16); + r4 = (uint32_t) (uint16_t) (a >> 16) * (uint16_t) (b >> 16); + + s.low = r1; + s.high = r4; + + t.high = (uint32_t) r2 >> 16; + t.low = r2 << 16; + ao_plus64(&u, &s, &t); + + v.high = (int32_t) r3 >> 16; + v.low = r3 << 16; + ao_plus64(r, &u, &v); +} + +void ao_neg64(ao_int64_t *r, ao_int64_t *a) { + r->high = ~a->high; + r->low = ~a->low; + if (!++r->low) + r->high++; +} + +void ao_mul64(ao_int64_t *r, int32_t a, int32_t b) { + uint8_t negative = 0; + + if (a < 0) { + a = -a; + negative = 1; + } + if (b < 0) { + b = -b; + negative = !negative; + } + ao_umul64(r, a, b); + if (negative) + ao_neg64(r, r); +} + +void ao_umul64_16(ao_int64_t *r, ao_int64_t *a, uint16_t b) { + uint32_t low = a->low; + ao_umul64(r, (uint32_t) low >> 1, (uint32_t) b << 1); + if (low & 1) { + if ((uint32_t) (r->low += b) < (uint32_t) b) + r->high++; + } + r->high += a->high * b; +} + +void ao_mul64_16(ao_int64_t *r, ao_int64_t *a, uint16_t b) { + if ((int32_t) a->high < 0) { + ao_int64_t t; + + ao_neg64(&t, a); + ao_umul64_16(r, &t, b); + ao_neg64(r, r); + } else + ao_umul64_16(r, a, b); +}