1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/security/nss/lib/freebl/mpi/mpv_sparc.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,220 @@
1.4 +/* This Source Code Form is subject to the terms of the Mozilla Public
1.5 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1.6 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1.7 +
1.8 +#include "vis_proto.h"
1.9 +
1.10 +/***************************************************************/
1.11 +
1.12 +typedef int t_s32;
1.13 +typedef unsigned int t_u32;
1.14 +#if defined(__sparcv9)
1.15 +typedef long t_s64;
1.16 +typedef unsigned long t_u64;
1.17 +#else
1.18 +typedef long long t_s64;
1.19 +typedef unsigned long long t_u64;
1.20 +#endif
1.21 +typedef double t_d64;
1.22 +
1.23 +/***************************************************************/
1.24 +
1.25 +typedef union {
1.26 + t_d64 d64;
1.27 + struct {
1.28 + t_s32 i0;
1.29 + t_s32 i1;
1.30 + } i32s;
1.31 +} d64_2_i32;
1.32 +
1.33 +/***************************************************************/
1.34 +
1.35 +#define BUFF_SIZE 256
1.36 +
1.37 +#define A_BITS 19
1.38 +#define A_MASK ((1 << A_BITS) - 1)
1.39 +
1.40 +/***************************************************************/
1.41 +
1.42 +static t_u64 mask_cnst[] = {
1.43 + 0x8000000080000000ull
1.44 +};
1.45 +
1.46 +/***************************************************************/
1.47 +
1.48 +#define DEF_VARS(N) \
1.49 + t_d64 *py = (t_d64*)y; \
1.50 + t_d64 mask = *((t_d64*)mask_cnst); \
1.51 + t_d64 ca = (1u << 31) - 1; \
1.52 + t_d64 da = (t_d64)a; \
1.53 + t_s64 buff[N], s; \
1.54 + d64_2_i32 dy
1.55 +
1.56 +/***************************************************************/
1.57 +
1.58 +#define MUL_U32_S64_2(i) \
1.59 + dy.d64 = vis_fxnor(mask, py[i]); \
1.60 + buff[2*(i) ] = (ca - (t_d64)dy.i32s.i0) * da; \
1.61 + buff[2*(i)+1] = (ca - (t_d64)dy.i32s.i1) * da
1.62 +
1.63 +#define MUL_U32_S64_2_D(i) \
1.64 + dy.d64 = vis_fxnor(mask, py[i]); \
1.65 + d0 = ca - (t_d64)dy.i32s.i0; \
1.66 + d1 = ca - (t_d64)dy.i32s.i1; \
1.67 + buff[4*(i) ] = (t_s64)(d0 * da); \
1.68 + buff[4*(i)+1] = (t_s64)(d0 * db); \
1.69 + buff[4*(i)+2] = (t_s64)(d1 * da); \
1.70 + buff[4*(i)+3] = (t_s64)(d1 * db)
1.71 +
1.72 +/***************************************************************/
1.73 +
1.74 +#define ADD_S64_U32(i) \
1.75 + s = buff[i] + x[i] + c; \
1.76 + z[i] = s; \
1.77 + c = (s >> 32)
1.78 +
1.79 +#define ADD_S64_U32_D(i) \
1.80 + s = buff[2*(i)] +(((t_s64)(buff[2*(i)+1]))<<A_BITS) + x[i] + uc; \
1.81 + z[i] = s; \
1.82 + uc = ((t_u64)s >> 32)
1.83 +
1.84 +/***************************************************************/
1.85 +
1.86 +#define MUL_U32_S64_8(i) \
1.87 + MUL_U32_S64_2(i); \
1.88 + MUL_U32_S64_2(i+1); \
1.89 + MUL_U32_S64_2(i+2); \
1.90 + MUL_U32_S64_2(i+3)
1.91 +
1.92 +#define MUL_U32_S64_D_8(i) \
1.93 + MUL_U32_S64_2_D(i); \
1.94 + MUL_U32_S64_2_D(i+1); \
1.95 + MUL_U32_S64_2_D(i+2); \
1.96 + MUL_U32_S64_2_D(i+3)
1.97 +
1.98 +/***************************************************************/
1.99 +
1.100 +#define ADD_S64_U32_8(i) \
1.101 + ADD_S64_U32(i); \
1.102 + ADD_S64_U32(i+1); \
1.103 + ADD_S64_U32(i+2); \
1.104 + ADD_S64_U32(i+3); \
1.105 + ADD_S64_U32(i+4); \
1.106 + ADD_S64_U32(i+5); \
1.107 + ADD_S64_U32(i+6); \
1.108 + ADD_S64_U32(i+7)
1.109 +
1.110 +#define ADD_S64_U32_D_8(i) \
1.111 + ADD_S64_U32_D(i); \
1.112 + ADD_S64_U32_D(i+1); \
1.113 + ADD_S64_U32_D(i+2); \
1.114 + ADD_S64_U32_D(i+3); \
1.115 + ADD_S64_U32_D(i+4); \
1.116 + ADD_S64_U32_D(i+5); \
1.117 + ADD_S64_U32_D(i+6); \
1.118 + ADD_S64_U32_D(i+7)
1.119 +
1.120 +/***************************************************************/
1.121 +
1.122 +t_u32 mul_add(t_u32 *z, t_u32 *x, t_u32 *y, int n, t_u32 a)
1.123 +{
1.124 + if (a < (1 << A_BITS)) {
1.125 +
1.126 + if (n == 8) {
1.127 + DEF_VARS(8);
1.128 + t_s32 c = 0;
1.129 +
1.130 + MUL_U32_S64_8(0);
1.131 + ADD_S64_U32_8(0);
1.132 +
1.133 + return c;
1.134 +
1.135 + } else if (n == 16) {
1.136 + DEF_VARS(16);
1.137 + t_s32 c = 0;
1.138 +
1.139 + MUL_U32_S64_8(0);
1.140 + MUL_U32_S64_8(4);
1.141 + ADD_S64_U32_8(0);
1.142 + ADD_S64_U32_8(8);
1.143 +
1.144 + return c;
1.145 +
1.146 + } else {
1.147 + DEF_VARS(BUFF_SIZE);
1.148 + t_s32 i, c = 0;
1.149 +
1.150 +#pragma pipeloop(0)
1.151 + for (i = 0; i < (n+1)/2; i ++) {
1.152 + MUL_U32_S64_2(i);
1.153 + }
1.154 +
1.155 +#pragma pipeloop(0)
1.156 + for (i = 0; i < n; i ++) {
1.157 + ADD_S64_U32(i);
1.158 + }
1.159 +
1.160 + return c;
1.161 +
1.162 + }
1.163 + } else {
1.164 +
1.165 + if (n == 8) {
1.166 + DEF_VARS(2*8);
1.167 + t_d64 d0, d1, db;
1.168 + t_u32 uc = 0;
1.169 +
1.170 + da = (t_d64)(a & A_MASK);
1.171 + db = (t_d64)(a >> A_BITS);
1.172 +
1.173 + MUL_U32_S64_D_8(0);
1.174 + ADD_S64_U32_D_8(0);
1.175 +
1.176 + return uc;
1.177 +
1.178 + } else if (n == 16) {
1.179 + DEF_VARS(2*16);
1.180 + t_d64 d0, d1, db;
1.181 + t_u32 uc = 0;
1.182 +
1.183 + da = (t_d64)(a & A_MASK);
1.184 + db = (t_d64)(a >> A_BITS);
1.185 +
1.186 + MUL_U32_S64_D_8(0);
1.187 + MUL_U32_S64_D_8(4);
1.188 + ADD_S64_U32_D_8(0);
1.189 + ADD_S64_U32_D_8(8);
1.190 +
1.191 + return uc;
1.192 +
1.193 + } else {
1.194 + DEF_VARS(2*BUFF_SIZE);
1.195 + t_d64 d0, d1, db;
1.196 + t_u32 i, uc = 0;
1.197 +
1.198 + da = (t_d64)(a & A_MASK);
1.199 + db = (t_d64)(a >> A_BITS);
1.200 +
1.201 +#pragma pipeloop(0)
1.202 + for (i = 0; i < (n+1)/2; i ++) {
1.203 + MUL_U32_S64_2_D(i);
1.204 + }
1.205 +
1.206 +#pragma pipeloop(0)
1.207 + for (i = 0; i < n; i ++) {
1.208 + ADD_S64_U32_D(i);
1.209 + }
1.210 +
1.211 + return uc;
1.212 + }
1.213 + }
1.214 +}
1.215 +
1.216 +/***************************************************************/
1.217 +
1.218 +t_u32 mul_add_inp(t_u32 *x, t_u32 *y, int n, t_u32 a)
1.219 +{
1.220 + return mul_add(x, x, y, n, a);
1.221 +}
1.222 +
1.223 +/***************************************************************/
