1.1 --- a/src/cpu/ppc/vm/stubGenerator_ppc.cpp Tue Jan 29 08:28:24 2019 -0500
1.2 +++ b/src/cpu/ppc/vm/stubGenerator_ppc.cpp Sun Aug 11 19:11:08 2019 -0400
1.3 @@ -2224,7 +2224,7 @@
1.4 return start;
1.5 }
1.6
1.7 - // Arguments for generated stub (little endian only):
1.8 + // Arguments for generated stub:
1.9 // R3_ARG1 - source byte array address
1.10 // R4_ARG2 - destination byte array address
1.11 // R5_ARG3 - round key array
1.12 @@ -2243,7 +2243,6 @@
1.13 Register keylen = R8;
1.14 Register temp = R9;
1.15 Register keypos = R10;
1.16 - Register hex = R11;
1.17 Register fifteen = R12;
1.18
1.19 VectorRegister vRet = VR0;
1.20 @@ -2263,164 +2262,170 @@
1.21 VectorRegister vTmp3 = VR11;
1.22 VectorRegister vTmp4 = VR12;
1.23
1.24 - VectorRegister vLow = VR13;
1.25 - VectorRegister vHigh = VR14;
1.26 -
1.27 - __ li (hex, 16);
1.28 __ li (fifteen, 15);
1.29 - __ vspltisb (fSplt, 0x0f);
1.30
1.31 // load unaligned from[0-15] to vsRet
1.32 __ lvx (vRet, from);
1.33 __ lvx (vTmp1, fifteen, from);
1.34 __ lvsl (fromPerm, from);
1.35 +#ifdef VM_LITTLE_ENDIAN
1.36 + __ vspltisb (fSplt, 0x0f);
1.37 __ vxor (fromPerm, fromPerm, fSplt);
1.38 +#endif
1.39 __ vperm (vRet, vRet, vTmp1, fromPerm);
1.40
1.41 // load keylen (44 or 52 or 60)
1.42 __ lwz (keylen, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT), key);
1.43
1.44 // to load keys
1.45 - __ lvsr (keyPerm, key);
1.46 - __ vxor (vTmp2, vTmp2, vTmp2);
1.47 + __ load_perm (keyPerm, key);
1.48 +#ifdef VM_LITTLE_ENDIAN
1.49 __ vspltisb (vTmp2, -16);
1.50 __ vrld (keyPerm, keyPerm, vTmp2);
1.51 __ vrld (keyPerm, keyPerm, vTmp2);
1.52 __ vsldoi (keyPerm, keyPerm, keyPerm, 8);
1.53 -
1.54 - // load the 1st round key to vKey1
1.55 - __ li (keypos, 0);
1.56 +#endif
1.57 +
1.58 + // load the 1st round key to vTmp1
1.59 + __ lvx (vTmp1, key);
1.60 + __ li (keypos, 16);
1.61 __ lvx (vKey1, keypos, key);
1.62 - __ addi (keypos, keypos, 16);
1.63 + __ vec_perm (vTmp1, vKey1, keyPerm);
1.64 +
1.65 + // 1st round
1.66 + __ vxor (vRet, vRet, vTmp1);
1.67 +
1.68 + // load the 2nd round key to vKey1
1.69 + __ li (keypos, 32);
1.70 + __ lvx (vKey2, keypos, key);
1.71 + __ vec_perm (vKey1, vKey2, keyPerm);
1.72 +
1.73 + // load the 3rd round key to vKey2
1.74 + __ li (keypos, 48);
1.75 + __ lvx (vKey3, keypos, key);
1.76 + __ vec_perm (vKey2, vKey3, keyPerm);
1.77 +
1.78 + // load the 4th round key to vKey3
1.79 + __ li (keypos, 64);
1.80 + __ lvx (vKey4, keypos, key);
1.81 + __ vec_perm (vKey3, vKey4, keyPerm);
1.82 +
1.83 + // load the 5th round key to vKey4
1.84 + __ li (keypos, 80);
1.85 __ lvx (vTmp1, keypos, key);
1.86 - __ vperm (vKey1, vTmp1, vKey1, keyPerm);
1.87 -
1.88 - // 1st round
1.89 - __ vxor (vRet, vRet, vKey1);
1.90 -
1.91 - // load the 2nd round key to vKey1
1.92 - __ addi (keypos, keypos, 16);
1.93 - __ lvx (vTmp2, keypos, key);
1.94 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.95 -
1.96 - // load the 3rd round key to vKey2
1.97 - __ addi (keypos, keypos, 16);
1.98 + __ vec_perm (vKey4, vTmp1, keyPerm);
1.99 +
1.100 + // 2nd - 5th rounds
1.101 + __ vcipher (vRet, vRet, vKey1);
1.102 + __ vcipher (vRet, vRet, vKey2);
1.103 + __ vcipher (vRet, vRet, vKey3);
1.104 + __ vcipher (vRet, vRet, vKey4);
1.105 +
1.106 + // load the 6th round key to vKey1
1.107 + __ li (keypos, 96);
1.108 + __ lvx (vKey2, keypos, key);
1.109 + __ vec_perm (vKey1, vTmp1, vKey2, keyPerm);
1.110 +
1.111 + // load the 7th round key to vKey2
1.112 + __ li (keypos, 112);
1.113 + __ lvx (vKey3, keypos, key);
1.114 + __ vec_perm (vKey2, vKey3, keyPerm);
1.115 +
1.116 + // load the 8th round key to vKey3
1.117 + __ li (keypos, 128);
1.118 + __ lvx (vKey4, keypos, key);
1.119 + __ vec_perm (vKey3, vKey4, keyPerm);
1.120 +
1.121 + // load the 9th round key to vKey4
1.122 + __ li (keypos, 144);
1.123 __ lvx (vTmp1, keypos, key);
1.124 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.125 -
1.126 - // load the 4th round key to vKey3
1.127 - __ addi (keypos, keypos, 16);
1.128 - __ lvx (vTmp2, keypos, key);
1.129 - __ vperm (vKey3, vTmp2, vTmp1, keyPerm);
1.130 -
1.131 - // load the 5th round key to vKey4
1.132 - __ addi (keypos, keypos, 16);
1.133 + __ vec_perm (vKey4, vTmp1, keyPerm);
1.134 +
1.135 + // 6th - 9th rounds
1.136 + __ vcipher (vRet, vRet, vKey1);
1.137 + __ vcipher (vRet, vRet, vKey2);
1.138 + __ vcipher (vRet, vRet, vKey3);
1.139 + __ vcipher (vRet, vRet, vKey4);
1.140 +
1.141 + // load the 10th round key to vKey1
1.142 + __ li (keypos, 160);
1.143 + __ lvx (vKey2, keypos, key);
1.144 + __ vec_perm (vKey1, vTmp1, vKey2, keyPerm);
1.145 +
1.146 + // load the 11th round key to vKey2
1.147 + __ li (keypos, 176);
1.148 __ lvx (vTmp1, keypos, key);
1.149 - __ vperm (vKey4, vTmp1, vTmp2, keyPerm);
1.150 -
1.151 - // 2nd - 5th rounds
1.152 - __ vcipher (vRet, vRet, vKey1);
1.153 - __ vcipher (vRet, vRet, vKey2);
1.154 - __ vcipher (vRet, vRet, vKey3);
1.155 - __ vcipher (vRet, vRet, vKey4);
1.156 -
1.157 - // load the 6th round key to vKey1
1.158 - __ addi (keypos, keypos, 16);
1.159 - __ lvx (vTmp2, keypos, key);
1.160 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.161 -
1.162 - // load the 7th round key to vKey2
1.163 - __ addi (keypos, keypos, 16);
1.164 - __ lvx (vTmp1, keypos, key);
1.165 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.166 -
1.167 - // load the 8th round key to vKey3
1.168 - __ addi (keypos, keypos, 16);
1.169 - __ lvx (vTmp2, keypos, key);
1.170 - __ vperm (vKey3, vTmp2, vTmp1, keyPerm);
1.171 -
1.172 - // load the 9th round key to vKey4
1.173 - __ addi (keypos, keypos, 16);
1.174 - __ lvx (vTmp1, keypos, key);
1.175 - __ vperm (vKey4, vTmp1, vTmp2, keyPerm);
1.176 -
1.177 - // 6th - 9th rounds
1.178 - __ vcipher (vRet, vRet, vKey1);
1.179 - __ vcipher (vRet, vRet, vKey2);
1.180 - __ vcipher (vRet, vRet, vKey3);
1.181 - __ vcipher (vRet, vRet, vKey4);
1.182 -
1.183 - // load the 10th round key to vKey1
1.184 - __ addi (keypos, keypos, 16);
1.185 - __ lvx (vTmp2, keypos, key);
1.186 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.187 -
1.188 - // load the 11th round key to vKey2
1.189 - __ addi (keypos, keypos, 16);
1.190 - __ lvx (vTmp1, keypos, key);
1.191 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.192 + __ vec_perm (vKey2, vTmp1, keyPerm);
1.193
1.194 // if all round keys are loaded, skip next 4 rounds
1.195 __ cmpwi (CCR0, keylen, 44);
1.196 __ beq (CCR0, L_doLast);
1.197
1.198 // 10th - 11th rounds
1.199 - __ vcipher (vRet, vRet, vKey1);
1.200 - __ vcipher (vRet, vRet, vKey2);
1.201 + __ vcipher (vRet, vRet, vKey1);
1.202 + __ vcipher (vRet, vRet, vKey2);
1.203
1.204 // load the 12th round key to vKey1
1.205 - __ addi (keypos, keypos, 16);
1.206 - __ lvx (vTmp2, keypos, key);
1.207 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.208 + __ li (keypos, 192);
1.209 + __ lvx (vKey2, keypos, key);
1.210 + __ vec_perm (vKey1, vTmp1, vKey2, keyPerm);
1.211
1.212 // load the 13th round key to vKey2
1.213 - __ addi (keypos, keypos, 16);
1.214 + __ li (keypos, 208);
1.215 __ lvx (vTmp1, keypos, key);
1.216 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.217 + __ vec_perm (vKey2, vTmp1, keyPerm);
1.218
1.219 // if all round keys are loaded, skip next 2 rounds
1.220 __ cmpwi (CCR0, keylen, 52);
1.221 __ beq (CCR0, L_doLast);
1.222
1.223 // 12th - 13th rounds
1.224 - __ vcipher (vRet, vRet, vKey1);
1.225 - __ vcipher (vRet, vRet, vKey2);
1.226 + __ vcipher (vRet, vRet, vKey1);
1.227 + __ vcipher (vRet, vRet, vKey2);
1.228
1.229 // load the 14th round key to vKey1
1.230 - __ addi (keypos, keypos, 16);
1.231 - __ lvx (vTmp2, keypos, key);
1.232 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.233 + __ li (keypos, 224);
1.234 + __ lvx (vKey2, keypos, key);
1.235 + __ vec_perm (vKey1, vTmp1, vKey2, keyPerm);
1.236
1.237 // load the 15th round key to vKey2
1.238 - __ addi (keypos, keypos, 16);
1.239 + __ li (keypos, 240);
1.240 __ lvx (vTmp1, keypos, key);
1.241 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.242 + __ vec_perm (vKey2, vTmp1, keyPerm);
1.243
1.244 __ bind(L_doLast);
1.245
1.246 // last two rounds
1.247 - __ vcipher (vRet, vRet, vKey1);
1.248 - __ vcipherlast (vRet, vRet, vKey2);
1.249 -
1.250 - __ neg (temp, to);
1.251 - __ lvsr (toPerm, temp);
1.252 - __ vspltisb (vTmp2, -1);
1.253 - __ vxor (vTmp1, vTmp1, vTmp1);
1.254 - __ vperm (vTmp2, vTmp2, vTmp1, toPerm);
1.255 - __ vxor (toPerm, toPerm, fSplt);
1.256 + __ vcipher (vRet, vRet, vKey1);
1.257 + __ vcipherlast (vRet, vRet, vKey2);
1.258 +
1.259 + // store result (unaligned)
1.260 +#ifdef VM_LITTLE_ENDIAN
1.261 + __ lvsl (toPerm, to);
1.262 +#else
1.263 + __ lvsr (toPerm, to);
1.264 +#endif
1.265 + __ vspltisb (vTmp3, -1);
1.266 + __ vspltisb (vTmp4, 0);
1.267 __ lvx (vTmp1, to);
1.268 - __ vperm (vRet, vRet, vRet, toPerm);
1.269 - __ vsel (vTmp1, vTmp1, vRet, vTmp2);
1.270 - __ lvx (vTmp4, fifteen, to);
1.271 + __ lvx (vTmp2, fifteen, to);
1.272 +#ifdef VM_LITTLE_ENDIAN
1.273 + __ vperm (vTmp3, vTmp3, vTmp4, toPerm); // generate select mask
1.274 + __ vxor (toPerm, toPerm, fSplt); // swap bytes
1.275 +#else
1.276 + __ vperm (vTmp3, vTmp4, vTmp3, toPerm); // generate select mask
1.277 +#endif
1.278 + __ vperm (vTmp4, vRet, vRet, toPerm); // rotate data
1.279 + __ vsel (vTmp2, vTmp4, vTmp2, vTmp3);
1.280 + __ vsel (vTmp1, vTmp1, vTmp4, vTmp3);
1.281 + __ stvx (vTmp2, fifteen, to); // store this one first (may alias)
1.282 __ stvx (vTmp1, to);
1.283 - __ vsel (vRet, vRet, vTmp4, vTmp2);
1.284 - __ stvx (vRet, fifteen, to);
1.285
1.286 __ blr();
1.287 return start;
1.288 }
1.289
1.290 - // Arguments for generated stub (little endian only):
1.291 + // Arguments for generated stub:
1.292 // R3_ARG1 - source byte array address
1.293 // R4_ARG2 - destination byte array address
1.294 // R5_ARG3 - K (key) in little endian int array
1.295 @@ -2442,7 +2447,6 @@
1.296 Register keylen = R8;
1.297 Register temp = R9;
1.298 Register keypos = R10;
1.299 - Register hex = R11;
1.300 Register fifteen = R12;
1.301
1.302 VectorRegister vRet = VR0;
1.303 @@ -2463,30 +2467,30 @@
1.304 VectorRegister vTmp3 = VR12;
1.305 VectorRegister vTmp4 = VR13;
1.306
1.307 - VectorRegister vLow = VR14;
1.308 - VectorRegister vHigh = VR15;
1.309 -
1.310 - __ li (hex, 16);
1.311 __ li (fifteen, 15);
1.312 - __ vspltisb (fSplt, 0x0f);
1.313
1.314 // load unaligned from[0-15] to vsRet
1.315 __ lvx (vRet, from);
1.316 __ lvx (vTmp1, fifteen, from);
1.317 __ lvsl (fromPerm, from);
1.318 +#ifdef VM_LITTLE_ENDIAN
1.319 + __ vspltisb (fSplt, 0x0f);
1.320 __ vxor (fromPerm, fromPerm, fSplt);
1.321 +#endif
1.322 __ vperm (vRet, vRet, vTmp1, fromPerm); // align [and byte swap in LE]
1.323
1.324 // load keylen (44 or 52 or 60)
1.325 __ lwz (keylen, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT), key);
1.326
1.327 // to load keys
1.328 - __ lvsr (keyPerm, key);
1.329 + __ load_perm (keyPerm, key);
1.330 +#ifdef VM_LITTLE_ENDIAN
1.331 __ vxor (vTmp2, vTmp2, vTmp2);
1.332 __ vspltisb (vTmp2, -16);
1.333 __ vrld (keyPerm, keyPerm, vTmp2);
1.334 __ vrld (keyPerm, keyPerm, vTmp2);
1.335 __ vsldoi (keyPerm, keyPerm, keyPerm, 8);
1.336 +#endif
1.337
1.338 __ cmpwi (CCR0, keylen, 44);
1.339 __ beq (CCR0, L_do44);
1.340 @@ -2494,32 +2498,32 @@
1.341 __ cmpwi (CCR0, keylen, 52);
1.342 __ beq (CCR0, L_do52);
1.343
1.344 - // load the 15th round key to vKey11
1.345 + // load the 15th round key to vKey1
1.346 __ li (keypos, 240);
1.347 + __ lvx (vKey1, keypos, key);
1.348 + __ li (keypos, 224);
1.349 + __ lvx (vKey2, keypos, key);
1.350 + __ vec_perm (vKey1, vKey2, vKey1, keyPerm);
1.351 +
1.352 + // load the 14th round key to vKey2
1.353 + __ li (keypos, 208);
1.354 + __ lvx (vKey3, keypos, key);
1.355 + __ vec_perm (vKey2, vKey3, vKey2, keyPerm);
1.356 +
1.357 + // load the 13th round key to vKey3
1.358 + __ li (keypos, 192);
1.359 + __ lvx (vKey4, keypos, key);
1.360 + __ vec_perm (vKey3, vKey4, vKey3, keyPerm);
1.361 +
1.362 + // load the 12th round key to vKey4
1.363 + __ li (keypos, 176);
1.364 + __ lvx (vKey5, keypos, key);
1.365 + __ vec_perm (vKey4, vKey5, vKey4, keyPerm);
1.366 +
1.367 + // load the 11th round key to vKey5
1.368 + __ li (keypos, 160);
1.369 __ lvx (vTmp1, keypos, key);
1.370 - __ addi (keypos, keypos, -16);
1.371 - __ lvx (vTmp2, keypos, key);
1.372 - __ vperm (vKey1, vTmp1, vTmp2, keyPerm);
1.373 -
1.374 - // load the 14th round key to vKey10
1.375 - __ addi (keypos, keypos, -16);
1.376 - __ lvx (vTmp1, keypos, key);
1.377 - __ vperm (vKey2, vTmp2, vTmp1, keyPerm);
1.378 -
1.379 - // load the 13th round key to vKey10
1.380 - __ addi (keypos, keypos, -16);
1.381 - __ lvx (vTmp2, keypos, key);
1.382 - __ vperm (vKey3, vTmp1, vTmp2, keyPerm);
1.383 -
1.384 - // load the 12th round key to vKey10
1.385 - __ addi (keypos, keypos, -16);
1.386 - __ lvx (vTmp1, keypos, key);
1.387 - __ vperm (vKey4, vTmp2, vTmp1, keyPerm);
1.388 -
1.389 - // load the 11th round key to vKey10
1.390 - __ addi (keypos, keypos, -16);
1.391 - __ lvx (vTmp2, keypos, key);
1.392 - __ vperm (vKey5, vTmp1, vTmp2, keyPerm);
1.393 + __ vec_perm (vKey5, vTmp1, vKey5, keyPerm);
1.394
1.395 // 1st - 5th rounds
1.396 __ vxor (vRet, vRet, vKey1);
1.397 @@ -2532,22 +2536,22 @@
1.398
1.399 __ bind (L_do52);
1.400
1.401 - // load the 13th round key to vKey11
1.402 + // load the 13th round key to vKey1
1.403 __ li (keypos, 208);
1.404 + __ lvx (vKey1, keypos, key);
1.405 + __ li (keypos, 192);
1.406 + __ lvx (vKey2, keypos, key);
1.407 + __ vec_perm (vKey1, vKey2, vKey1, keyPerm);
1.408 +
1.409 + // load the 12th round key to vKey2
1.410 + __ li (keypos, 176);
1.411 + __ lvx (vKey3, keypos, key);
1.412 + __ vec_perm (vKey2, vKey3, vKey2, keyPerm);
1.413 +
1.414 + // load the 11th round key to vKey3
1.415 + __ li (keypos, 160);
1.416 __ lvx (vTmp1, keypos, key);
1.417 - __ addi (keypos, keypos, -16);
1.418 - __ lvx (vTmp2, keypos, key);
1.419 - __ vperm (vKey1, vTmp1, vTmp2, keyPerm);
1.420 -
1.421 - // load the 12th round key to vKey10
1.422 - __ addi (keypos, keypos, -16);
1.423 - __ lvx (vTmp1, keypos, key);
1.424 - __ vperm (vKey2, vTmp2, vTmp1, keyPerm);
1.425 -
1.426 - // load the 11th round key to vKey10
1.427 - __ addi (keypos, keypos, -16);
1.428 - __ lvx (vTmp2, keypos, key);
1.429 - __ vperm (vKey3, vTmp1, vTmp2, keyPerm);
1.430 + __ vec_perm (vKey3, vTmp1, vKey3, keyPerm);
1.431
1.432 // 1st - 3rd rounds
1.433 __ vxor (vRet, vRet, vKey1);
1.434 @@ -2558,42 +2562,42 @@
1.435
1.436 __ bind (L_do44);
1.437
1.438 - // load the 11th round key to vKey11
1.439 + // load the 11th round key to vKey1
1.440 __ li (keypos, 176);
1.441 + __ lvx (vKey1, keypos, key);
1.442 + __ li (keypos, 160);
1.443 __ lvx (vTmp1, keypos, key);
1.444 - __ addi (keypos, keypos, -16);
1.445 - __ lvx (vTmp2, keypos, key);
1.446 - __ vperm (vKey1, vTmp1, vTmp2, keyPerm);
1.447 + __ vec_perm (vKey1, vTmp1, vKey1, keyPerm);
1.448
1.449 // 1st round
1.450 __ vxor (vRet, vRet, vKey1);
1.451
1.452 __ bind (L_doLast);
1.453
1.454 - // load the 10th round key to vKey10
1.455 - __ addi (keypos, keypos, -16);
1.456 + // load the 10th round key to vKey1
1.457 + __ li (keypos, 144);
1.458 + __ lvx (vKey2, keypos, key);
1.459 + __ vec_perm (vKey1, vKey2, vTmp1, keyPerm);
1.460 +
1.461 + // load the 9th round key to vKey2
1.462 + __ li (keypos, 128);
1.463 + __ lvx (vKey3, keypos, key);
1.464 + __ vec_perm (vKey2, vKey3, vKey2, keyPerm);
1.465 +
1.466 + // load the 8th round key to vKey3
1.467 + __ li (keypos, 112);
1.468 + __ lvx (vKey4, keypos, key);
1.469 + __ vec_perm (vKey3, vKey4, vKey3, keyPerm);
1.470 +
1.471 + // load the 7th round key to vKey4
1.472 + __ li (keypos, 96);
1.473 + __ lvx (vKey5, keypos, key);
1.474 + __ vec_perm (vKey4, vKey5, vKey4, keyPerm);
1.475 +
1.476 + // load the 6th round key to vKey5
1.477 + __ li (keypos, 80);
1.478 __ lvx (vTmp1, keypos, key);
1.479 - __ vperm (vKey1, vTmp2, vTmp1, keyPerm);
1.480 -
1.481 - // load the 9th round key to vKey10
1.482 - __ addi (keypos, keypos, -16);
1.483 - __ lvx (vTmp2, keypos, key);
1.484 - __ vperm (vKey2, vTmp1, vTmp2, keyPerm);
1.485 -
1.486 - // load the 8th round key to vKey10
1.487 - __ addi (keypos, keypos, -16);
1.488 - __ lvx (vTmp1, keypos, key);
1.489 - __ vperm (vKey3, vTmp2, vTmp1, keyPerm);
1.490 -
1.491 - // load the 7th round key to vKey10
1.492 - __ addi (keypos, keypos, -16);
1.493 - __ lvx (vTmp2, keypos, key);
1.494 - __ vperm (vKey4, vTmp1, vTmp2, keyPerm);
1.495 -
1.496 - // load the 6th round key to vKey10
1.497 - __ addi (keypos, keypos, -16);
1.498 - __ lvx (vTmp1, keypos, key);
1.499 - __ vperm (vKey5, vTmp2, vTmp1, keyPerm);
1.500 + __ vec_perm (vKey5, vTmp1, vKey5, keyPerm);
1.501
1.502 // last 10th - 6th rounds
1.503 __ vncipher (vRet, vRet, vKey1);
1.504 @@ -2602,30 +2606,29 @@
1.505 __ vncipher (vRet, vRet, vKey4);
1.506 __ vncipher (vRet, vRet, vKey5);
1.507
1.508 - // load the 5th round key to vKey10
1.509 - __ addi (keypos, keypos, -16);
1.510 - __ lvx (vTmp2, keypos, key);
1.511 - __ vperm (vKey1, vTmp1, vTmp2, keyPerm);
1.512 -
1.513 - // load the 4th round key to vKey10
1.514 - __ addi (keypos, keypos, -16);
1.515 - __ lvx (vTmp1, keypos, key);
1.516 - __ vperm (vKey2, vTmp2, vTmp1, keyPerm);
1.517 -
1.518 - // load the 3rd round key to vKey10
1.519 - __ addi (keypos, keypos, -16);
1.520 - __ lvx (vTmp2, keypos, key);
1.521 - __ vperm (vKey3, vTmp1, vTmp2, keyPerm);
1.522 -
1.523 - // load the 2nd round key to vKey10
1.524 - __ addi (keypos, keypos, -16);
1.525 - __ lvx (vTmp1, keypos, key);
1.526 - __ vperm (vKey4, vTmp2, vTmp1, keyPerm);
1.527 -
1.528 - // load the 1st round key to vKey10
1.529 - __ addi (keypos, keypos, -16);
1.530 - __ lvx (vTmp2, keypos, key);
1.531 - __ vperm (vKey5, vTmp1, vTmp2, keyPerm);
1.532 + // load the 5th round key to vKey1
1.533 + __ li (keypos, 64);
1.534 + __ lvx (vKey2, keypos, key);
1.535 + __ vec_perm (vKey1, vKey2, vTmp1, keyPerm);
1.536 +
1.537 + // load the 4th round key to vKey2
1.538 + __ li (keypos, 48);
1.539 + __ lvx (vKey3, keypos, key);
1.540 + __ vec_perm (vKey2, vKey3, vKey2, keyPerm);
1.541 +
1.542 + // load the 3rd round key to vKey3
1.543 + __ li (keypos, 32);
1.544 + __ lvx (vKey4, keypos, key);
1.545 + __ vec_perm (vKey3, vKey4, vKey3, keyPerm);
1.546 +
1.547 + // load the 2nd round key to vKey4
1.548 + __ li (keypos, 16);
1.549 + __ lvx (vKey5, keypos, key);
1.550 + __ vec_perm (vKey4, vKey5, vKey4, keyPerm);
1.551 +
1.552 + // load the 1st round key to vKey5
1.553 + __ lvx (vTmp1, key);
1.554 + __ vec_perm (vKey5, vTmp1, vKey5, keyPerm);
1.555
1.556 // last 5th - 1th rounds
1.557 __ vncipher (vRet, vRet, vKey1);
1.558 @@ -2634,19 +2637,27 @@
1.559 __ vncipher (vRet, vRet, vKey4);
1.560 __ vncipherlast (vRet, vRet, vKey5);
1.561
1.562 - __ neg (temp, to);
1.563 - __ lvsr (toPerm, temp);
1.564 - __ vspltisb (vTmp2, -1);
1.565 - __ vxor (vTmp1, vTmp1, vTmp1);
1.566 - __ vperm (vTmp2, vTmp2, vTmp1, toPerm);
1.567 - __ vxor (toPerm, toPerm, fSplt);
1.568 + // store result (unaligned)
1.569 +#ifdef VM_LITTLE_ENDIAN
1.570 + __ lvsl (toPerm, to);
1.571 +#else
1.572 + __ lvsr (toPerm, to);
1.573 +#endif
1.574 + __ vspltisb (vTmp3, -1);
1.575 + __ vspltisb (vTmp4, 0);
1.576 __ lvx (vTmp1, to);
1.577 - __ vperm (vRet, vRet, vRet, toPerm);
1.578 - __ vsel (vTmp1, vTmp1, vRet, vTmp2);
1.579 - __ lvx (vTmp4, fifteen, to);
1.580 + __ lvx (vTmp2, fifteen, to);
1.581 +#ifdef VM_LITTLE_ENDIAN
1.582 + __ vperm (vTmp3, vTmp3, vTmp4, toPerm); // generate select mask
1.583 + __ vxor (toPerm, toPerm, fSplt); // swap bytes
1.584 +#else
1.585 + __ vperm (vTmp3, vTmp4, vTmp3, toPerm); // generate select mask
1.586 +#endif
1.587 + __ vperm (vTmp4, vRet, vRet, toPerm); // rotate data
1.588 + __ vsel (vTmp2, vTmp4, vTmp2, vTmp3);
1.589 + __ vsel (vTmp1, vTmp1, vTmp4, vTmp3);
1.590 + __ stvx (vTmp2, fifteen, to); // store this one first (may alias)
1.591 __ stvx (vTmp1, to);
1.592 - __ vsel (vRet, vRet, vTmp4, vTmp2);
1.593 - __ stvx (vRet, fifteen, to);
1.594
1.595 __ blr();
1.596 return start;
