static void lshift_block (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4])
{
#ifdef IS_NV
- out0[0] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[1] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[2] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[3] = __byte_perm (in0[3], in1[0], 0x4321);
- out1[0] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[1] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[2] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[3] = __byte_perm (in1[3], 0, 0x4321);
+ out0[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out1[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[3] = __byte_perm_S (in1[3], 0, 0x4321);
#endif
#if defined IS_AMD || defined IS_GENERIC
- out0[0] = amd_bytealign (in0[1], in0[0], 1);
- out0[1] = amd_bytealign (in0[2], in0[1], 1);
- out0[2] = amd_bytealign (in0[3], in0[2], 1);
- out0[3] = amd_bytealign (in1[0], in0[3], 1);
- out1[0] = amd_bytealign (in1[1], in1[0], 1);
- out1[1] = amd_bytealign (in1[2], in1[1], 1);
- out1[2] = amd_bytealign (in1[3], in1[2], 1);
- out1[3] = amd_bytealign ( 0, in1[3], 1);
+ out0[0] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[1] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[2] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[3] = amd_bytealign_S (in1[0], in0[3], 1);
+ out1[0] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[1] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[2] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[3] = amd_bytealign_S ( 0, in1[3], 1);
#endif
}
static void rshift_block (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4])
{
#ifdef IS_NV
- out1[3] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[2] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[1] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[0] = __byte_perm (in0[3], in1[0], 0x6543);
- out0[3] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[2] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[1] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[0] = __byte_perm ( 0, in0[0], 0x6543);
+ out1[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out0[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[0] = __byte_perm_S ( 0, in0[0], 0x6543);
#endif
#if defined IS_AMD || defined IS_GENERIC
- out1[3] = amd_bytealign (in1[3], in1[2], 3);
- out1[2] = amd_bytealign (in1[2], in1[1], 3);
- out1[1] = amd_bytealign (in1[1], in1[0], 3);
- out1[0] = amd_bytealign (in1[0], in0[3], 3);
- out0[3] = amd_bytealign (in0[3], in0[2], 3);
- out0[2] = amd_bytealign (in0[2], in0[1], 3);
- out0[1] = amd_bytealign (in0[1], in0[0], 3);
- out0[0] = amd_bytealign (in0[0], 0, 3);
+ out1[3] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[2] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[1] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[0] = amd_bytealign_S (in1[0], in0[3], 3);
+ out0[3] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[2] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[1] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[0] = amd_bytealign_S (in0[0], 0, 3);
#endif
}
out1[2] = in1[2];
out1[3] = in1[3];
break;
- case 1: out0[0] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[1] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[2] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[3] = __byte_perm (in0[3], in1[0], 0x4321);
- out1[0] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[1] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[2] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[3] = __byte_perm (in1[3], 0, 0x4321);
- break;
- case 2: out0[0] = __byte_perm (in0[0], in0[1], 0x5432);
- out0[1] = __byte_perm (in0[1], in0[2], 0x5432);
- out0[2] = __byte_perm (in0[2], in0[3], 0x5432);
- out0[3] = __byte_perm (in0[3], in1[0], 0x5432);
- out1[0] = __byte_perm (in1[0], in1[1], 0x5432);
- out1[1] = __byte_perm (in1[1], in1[2], 0x5432);
- out1[2] = __byte_perm (in1[2], in1[3], 0x5432);
- out1[3] = __byte_perm (in1[3], 0, 0x5432);
- break;
- case 3: out0[0] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[1] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[2] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[3] = __byte_perm (in0[3], in1[0], 0x6543);
- out1[0] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[1] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[2] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[3] = __byte_perm (in1[3], 0, 0x6543);
+ case 1: out0[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out1[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[3] = __byte_perm_S (in1[3], 0, 0x4321);
+ break;
+ case 2: out0[0] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out0[1] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out0[2] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out0[3] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out1[0] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out1[1] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out1[2] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out1[3] = __byte_perm_S (in1[3], 0, 0x5432);
+ break;
+ case 3: out0[0] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[1] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[2] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[3] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out1[0] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[1] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[2] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[3] = __byte_perm_S (in1[3], 0, 0x6543);
break;
case 4: out0[0] = in0[1];
out0[1] = in0[2];
out1[2] = in1[3];
out1[3] = 0;
break;
- case 5: out0[0] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[1] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[2] = __byte_perm (in0[3], in1[0], 0x4321);
- out0[3] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[0] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[1] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[2] = __byte_perm (in1[3], 0, 0x4321);
+ case 5: out0[0] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[1] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[2] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out0[3] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[0] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[1] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[2] = __byte_perm_S (in1[3], 0, 0x4321);
out1[3] = 0;
break;
- case 6: out0[0] = __byte_perm (in0[1], in0[2], 0x5432);
- out0[1] = __byte_perm (in0[2], in0[3], 0x5432);
- out0[2] = __byte_perm (in0[3], in1[0], 0x5432);
- out0[3] = __byte_perm (in1[0], in1[1], 0x5432);
- out1[0] = __byte_perm (in1[1], in1[2], 0x5432);
- out1[1] = __byte_perm (in1[2], in1[3], 0x5432);
- out1[2] = __byte_perm (in1[3], 0, 0x5432);
+ case 6: out0[0] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out0[1] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out0[2] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out0[3] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out1[0] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out1[1] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out1[2] = __byte_perm_S (in1[3], 0, 0x5432);
out1[3] = 0;
break;
- case 7: out0[0] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[1] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[2] = __byte_perm (in0[3], in1[0], 0x6543);
- out0[3] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[0] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[1] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[2] = __byte_perm (in1[3], 0, 0x6543);
+ case 7: out0[0] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[1] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[2] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out0[3] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[0] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[1] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[2] = __byte_perm_S (in1[3], 0, 0x6543);
out1[3] = 0;
break;
case 8: out0[0] = in0[2];
out1[2] = 0;
out1[3] = 0;
break;
- case 9: out0[0] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[1] = __byte_perm (in0[3], in1[0], 0x4321);
- out0[2] = __byte_perm (in1[0], in1[1], 0x4321);
- out0[3] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[0] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[1] = __byte_perm (in1[3], 0, 0x4321);
+ case 9: out0[0] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[1] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out0[2] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out0[3] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[0] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[1] = __byte_perm_S (in1[3], 0, 0x4321);
out1[2] = 0;
out1[3] = 0;
break;
- case 10: out0[0] = __byte_perm (in0[2], in0[3], 0x5432);
- out0[1] = __byte_perm (in0[3], in1[0], 0x5432);
- out0[2] = __byte_perm (in1[0], in1[1], 0x5432);
- out0[3] = __byte_perm (in1[1], in1[2], 0x5432);
- out1[0] = __byte_perm (in1[2], in1[3], 0x5432);
- out1[1] = __byte_perm (in1[3], 0, 0x5432);
+ case 10: out0[0] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out0[1] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out0[2] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out0[3] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out1[0] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out1[1] = __byte_perm_S (in1[3], 0, 0x5432);
out1[2] = 0;
out1[3] = 0;
break;
- case 11: out0[0] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[1] = __byte_perm (in0[3], in1[0], 0x6543);
- out0[2] = __byte_perm (in1[0], in1[1], 0x6543);
- out0[3] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[0] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[1] = __byte_perm (in1[3], 0, 0x6543);
+ case 11: out0[0] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[1] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out0[2] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out0[3] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[0] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[1] = __byte_perm_S (in1[3], 0, 0x6543);
out1[2] = 0;
out1[3] = 0;
break;
out1[3] = 0;
break;
case 13:
- out0[0] = __byte_perm (in0[3], in1[0], 0x4321);
- out0[1] = __byte_perm (in1[0], in1[1], 0x4321);
- out0[2] = __byte_perm (in1[1], in1[2], 0x4321);
- out0[3] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[0] = __byte_perm (in1[3], 0, 0x4321);
+ out0[0] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out0[1] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out0[2] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out0[3] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[0] = __byte_perm_S (in1[3], 0, 0x4321);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 14: out0[0] = __byte_perm (in0[3], in1[0], 0x5432);
- out0[1] = __byte_perm (in1[0], in1[1], 0x5432);
- out0[2] = __byte_perm (in1[1], in1[2], 0x5432);
- out0[3] = __byte_perm (in1[2], in1[3], 0x5432);
- out1[0] = __byte_perm (in1[3], 0, 0x5432);
+ case 14: out0[0] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out0[1] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out0[2] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out0[3] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out1[0] = __byte_perm_S (in1[3], 0, 0x5432);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 15: out0[0] = __byte_perm (in0[3], in1[0], 0x6543);
- out0[1] = __byte_perm (in1[0], in1[1], 0x6543);
- out0[2] = __byte_perm (in1[1], in1[2], 0x6543);
- out0[3] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[0] = __byte_perm (in1[3], 0, 0x6543);
+ case 15: out0[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out0[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out0[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out0[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[0] = __byte_perm_S (in1[3], 0, 0x6543);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 17: out0[0] = __byte_perm (in1[0], in1[1], 0x4321);
- out0[1] = __byte_perm (in1[1], in1[2], 0x4321);
- out0[2] = __byte_perm (in1[2], in1[3], 0x4321);
- out0[3] = __byte_perm (in1[3], 0, 0x4321);
+ case 17: out0[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out0[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out0[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out0[3] = __byte_perm_S (in1[3], 0, 0x4321);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 18: out0[0] = __byte_perm (in1[0], in1[1], 0x5432);
- out0[1] = __byte_perm (in1[1], in1[2], 0x5432);
- out0[2] = __byte_perm (in1[2], in1[3], 0x5432);
- out0[3] = __byte_perm (in1[3], 0, 0x5432);
+ case 18: out0[0] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out0[1] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out0[2] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out0[3] = __byte_perm_S (in1[3], 0, 0x5432);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 19: out0[0] = __byte_perm (in1[0], in1[1], 0x6543);
- out0[1] = __byte_perm (in1[1], in1[2], 0x6543);
- out0[2] = __byte_perm (in1[2], in1[3], 0x6543);
- out0[3] = __byte_perm (in1[3], 0, 0x6543);
+ case 19: out0[0] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out0[1] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out0[2] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out0[3] = __byte_perm_S (in1[3], 0, 0x6543);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 21: out0[0] = __byte_perm (in1[1], in1[2], 0x4321);
- out0[1] = __byte_perm (in1[2], in1[3], 0x4321);
- out0[2] = __byte_perm (in1[3], 0, 0x4321);
+ case 21: out0[0] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out0[1] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out0[2] = __byte_perm_S (in1[3], 0, 0x4321);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 22: out0[0] = __byte_perm (in1[1], in1[2], 0x5432);
- out0[1] = __byte_perm (in1[2], in1[3], 0x5432);
- out0[2] = __byte_perm (in1[3], 0, 0x5432);
+ case 22: out0[0] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out0[1] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out0[2] = __byte_perm_S (in1[3], 0, 0x5432);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 23: out0[0] = __byte_perm (in1[1], in1[2], 0x6543);
- out0[1] = __byte_perm (in1[2], in1[3], 0x6543);
- out0[2] = __byte_perm (in1[3], 0, 0x6543);
+ case 23: out0[0] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out0[1] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out0[2] = __byte_perm_S (in1[3], 0, 0x6543);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 25: out0[0] = __byte_perm (in1[2], in1[3], 0x4321);
- out0[1] = __byte_perm (in1[3], 0, 0x4321);
+ case 25: out0[0] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out0[1] = __byte_perm_S (in1[3], 0, 0x4321);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 26: out0[0] = __byte_perm (in1[2], in1[3], 0x5432);
- out0[1] = __byte_perm (in1[3], 0, 0x5432);
+ case 26: out0[0] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out0[1] = __byte_perm_S (in1[3], 0, 0x5432);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 27: out0[0] = __byte_perm (in1[2], in1[3], 0x6543);
- out0[1] = __byte_perm (in1[3], 0, 0x6543);
+ case 27: out0[0] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out0[1] = __byte_perm_S (in1[3], 0, 0x6543);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 29: out0[0] = __byte_perm (in1[3], 0, 0x4321);
+ case 29: out0[0] = __byte_perm_S (in1[3], 0, 0x4321);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 30: out0[0] = __byte_perm (in1[3], 0, 0x5432);
+ case 30: out0[0] = __byte_perm_S (in1[3], 0, 0x5432);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 31: out0[0] = __byte_perm (in1[3], 0, 0x6543);
+ case 31: out0[0] = __byte_perm_S (in1[3], 0, 0x6543);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out1[2] = in1[2];
out1[3] = in1[3];
break;
- case 1: out0[0] = amd_bytealign (in0[1], in0[0], 1);
- out0[1] = amd_bytealign (in0[2], in0[1], 1);
- out0[2] = amd_bytealign (in0[3], in0[2], 1);
- out0[3] = amd_bytealign (in1[0], in0[3], 1);
- out1[0] = amd_bytealign (in1[1], in1[0], 1);
- out1[1] = amd_bytealign (in1[2], in1[1], 1);
- out1[2] = amd_bytealign (in1[3], in1[2], 1);
- out1[3] = amd_bytealign ( 0, in1[3], 1);
- break;
- case 2: out0[0] = amd_bytealign (in0[1], in0[0], 2);
- out0[1] = amd_bytealign (in0[2], in0[1], 2);
- out0[2] = amd_bytealign (in0[3], in0[2], 2);
- out0[3] = amd_bytealign (in1[0], in0[3], 2);
- out1[0] = amd_bytealign (in1[1], in1[0], 2);
- out1[1] = amd_bytealign (in1[2], in1[1], 2);
- out1[2] = amd_bytealign (in1[3], in1[2], 2);
- out1[3] = amd_bytealign ( 0, in1[3], 2);
- break;
- case 3: out0[0] = amd_bytealign (in0[1], in0[0], 3);
- out0[1] = amd_bytealign (in0[2], in0[1], 3);
- out0[2] = amd_bytealign (in0[3], in0[2], 3);
- out0[3] = amd_bytealign (in1[0], in0[3], 3);
- out1[0] = amd_bytealign (in1[1], in1[0], 3);
- out1[1] = amd_bytealign (in1[2], in1[1], 3);
- out1[2] = amd_bytealign (in1[3], in1[2], 3);
- out1[3] = amd_bytealign ( 0, in1[3], 3);
+ case 1: out0[0] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[1] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[2] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[3] = amd_bytealign_S (in1[0], in0[3], 1);
+ out1[0] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[1] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[2] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[3] = amd_bytealign_S ( 0, in1[3], 1);
+ break;
+ case 2: out0[0] = amd_bytealign_S (in0[1], in0[0], 2);
+ out0[1] = amd_bytealign_S (in0[2], in0[1], 2);
+ out0[2] = amd_bytealign_S (in0[3], in0[2], 2);
+ out0[3] = amd_bytealign_S (in1[0], in0[3], 2);
+ out1[0] = amd_bytealign_S (in1[1], in1[0], 2);
+ out1[1] = amd_bytealign_S (in1[2], in1[1], 2);
+ out1[2] = amd_bytealign_S (in1[3], in1[2], 2);
+ out1[3] = amd_bytealign_S ( 0, in1[3], 2);
+ break;
+ case 3: out0[0] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[1] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[2] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[3] = amd_bytealign_S (in1[0], in0[3], 3);
+ out1[0] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[1] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[2] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[3] = amd_bytealign_S ( 0, in1[3], 3);
break;
case 4: out0[0] = in0[1];
out0[1] = in0[2];
out1[2] = in1[3];
out1[3] = 0;
break;
- case 5: out0[0] = amd_bytealign (in0[2], in0[1], 1);
- out0[1] = amd_bytealign (in0[3], in0[2], 1);
- out0[2] = amd_bytealign (in1[0], in0[3], 1);
- out0[3] = amd_bytealign (in1[1], in1[0], 1);
- out1[0] = amd_bytealign (in1[2], in1[1], 1);
- out1[1] = amd_bytealign (in1[3], in1[2], 1);
- out1[2] = amd_bytealign ( 0, in1[3], 1);
+ case 5: out0[0] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[1] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[2] = amd_bytealign_S (in1[0], in0[3], 1);
+ out0[3] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[0] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[1] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[2] = amd_bytealign_S ( 0, in1[3], 1);
out1[3] = 0;
break;
- case 6: out0[0] = amd_bytealign (in0[2], in0[1], 2);
- out0[1] = amd_bytealign (in0[3], in0[2], 2);
- out0[2] = amd_bytealign (in1[0], in0[3], 2);
- out0[3] = amd_bytealign (in1[1], in1[0], 2);
- out1[0] = amd_bytealign (in1[2], in1[1], 2);
- out1[1] = amd_bytealign (in1[3], in1[2], 2);
- out1[2] = amd_bytealign ( 0, in1[3], 2);
+ case 6: out0[0] = amd_bytealign_S (in0[2], in0[1], 2);
+ out0[1] = amd_bytealign_S (in0[3], in0[2], 2);
+ out0[2] = amd_bytealign_S (in1[0], in0[3], 2);
+ out0[3] = amd_bytealign_S (in1[1], in1[0], 2);
+ out1[0] = amd_bytealign_S (in1[2], in1[1], 2);
+ out1[1] = amd_bytealign_S (in1[3], in1[2], 2);
+ out1[2] = amd_bytealign_S ( 0, in1[3], 2);
out1[3] = 0;
break;
- case 7: out0[0] = amd_bytealign (in0[2], in0[1], 3);
- out0[1] = amd_bytealign (in0[3], in0[2], 3);
- out0[2] = amd_bytealign (in1[0], in0[3], 3);
- out0[3] = amd_bytealign (in1[1], in1[0], 3);
- out1[0] = amd_bytealign (in1[2], in1[1], 3);
- out1[1] = amd_bytealign (in1[3], in1[2], 3);
- out1[2] = amd_bytealign ( 0, in1[3], 3);
+ case 7: out0[0] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[1] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[2] = amd_bytealign_S (in1[0], in0[3], 3);
+ out0[3] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[0] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[1] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[2] = amd_bytealign_S ( 0, in1[3], 3);
out1[3] = 0;
break;
case 8: out0[0] = in0[2];
out1[2] = 0;
out1[3] = 0;
break;
- case 9: out0[0] = amd_bytealign (in0[3], in0[2], 1);
- out0[1] = amd_bytealign (in1[0], in0[3], 1);
- out0[2] = amd_bytealign (in1[1], in1[0], 1);
- out0[3] = amd_bytealign (in1[2], in1[1], 1);
- out1[0] = amd_bytealign (in1[3], in1[2], 1);
- out1[1] = amd_bytealign ( 0, in1[3], 1);
+ case 9: out0[0] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[1] = amd_bytealign_S (in1[0], in0[3], 1);
+ out0[2] = amd_bytealign_S (in1[1], in1[0], 1);
+ out0[3] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[0] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[1] = amd_bytealign_S ( 0, in1[3], 1);
out1[2] = 0;
out1[3] = 0;
break;
- case 10: out0[0] = amd_bytealign (in0[3], in0[2], 2);
- out0[1] = amd_bytealign (in1[0], in0[3], 2);
- out0[2] = amd_bytealign (in1[1], in1[0], 2);
- out0[3] = amd_bytealign (in1[2], in1[1], 2);
- out1[0] = amd_bytealign (in1[3], in1[2], 2);
- out1[1] = amd_bytealign ( 0, in1[3], 2);
+ case 10: out0[0] = amd_bytealign_S (in0[3], in0[2], 2);
+ out0[1] = amd_bytealign_S (in1[0], in0[3], 2);
+ out0[2] = amd_bytealign_S (in1[1], in1[0], 2);
+ out0[3] = amd_bytealign_S (in1[2], in1[1], 2);
+ out1[0] = amd_bytealign_S (in1[3], in1[2], 2);
+ out1[1] = amd_bytealign_S ( 0, in1[3], 2);
out1[2] = 0;
out1[3] = 0;
break;
- case 11: out0[0] = amd_bytealign (in0[3], in0[2], 3);
- out0[1] = amd_bytealign (in1[0], in0[3], 3);
- out0[2] = amd_bytealign (in1[1], in1[0], 3);
- out0[3] = amd_bytealign (in1[2], in1[1], 3);
- out1[0] = amd_bytealign (in1[3], in1[2], 3);
- out1[1] = amd_bytealign ( 0, in1[3], 3);
+ case 11: out0[0] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[1] = amd_bytealign_S (in1[0], in0[3], 3);
+ out0[2] = amd_bytealign_S (in1[1], in1[0], 3);
+ out0[3] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[0] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[1] = amd_bytealign_S ( 0, in1[3], 3);
out1[2] = 0;
out1[3] = 0;
break;
out1[2] = 0;
out1[3] = 0;
break;
- case 13: out0[0] = amd_bytealign (in1[0], in0[3], 1);
- out0[1] = amd_bytealign (in1[1], in1[0], 1);
- out0[2] = amd_bytealign (in1[2], in1[1], 1);
- out0[3] = amd_bytealign (in1[3], in1[2], 1);
- out1[0] = amd_bytealign ( 0, in1[3], 1);
+ case 13: out0[0] = amd_bytealign_S (in1[0], in0[3], 1);
+ out0[1] = amd_bytealign_S (in1[1], in1[0], 1);
+ out0[2] = amd_bytealign_S (in1[2], in1[1], 1);
+ out0[3] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[0] = amd_bytealign_S ( 0, in1[3], 1);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 14: out0[0] = amd_bytealign (in1[0], in0[3], 2);
- out0[1] = amd_bytealign (in1[1], in1[0], 2);
- out0[2] = amd_bytealign (in1[2], in1[1], 2);
- out0[3] = amd_bytealign (in1[3], in1[2], 2);
- out1[0] = amd_bytealign ( 0, in1[3], 2);
+ case 14: out0[0] = amd_bytealign_S (in1[0], in0[3], 2);
+ out0[1] = amd_bytealign_S (in1[1], in1[0], 2);
+ out0[2] = amd_bytealign_S (in1[2], in1[1], 2);
+ out0[3] = amd_bytealign_S (in1[3], in1[2], 2);
+ out1[0] = amd_bytealign_S ( 0, in1[3], 2);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 15: out0[0] = amd_bytealign (in1[0], in0[3], 3);
- out0[1] = amd_bytealign (in1[1], in1[0], 3);
- out0[2] = amd_bytealign (in1[2], in1[1], 3);
- out0[3] = amd_bytealign (in1[3], in1[2], 3);
- out1[0] = amd_bytealign ( 0, in1[3], 3);
+ case 15: out0[0] = amd_bytealign_S (in1[0], in0[3], 3);
+ out0[1] = amd_bytealign_S (in1[1], in1[0], 3);
+ out0[2] = amd_bytealign_S (in1[2], in1[1], 3);
+ out0[3] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[0] = amd_bytealign_S ( 0, in1[3], 3);
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 17: out0[0] = amd_bytealign (in1[1], in1[0], 1);
- out0[1] = amd_bytealign (in1[2], in1[1], 1);
- out0[2] = amd_bytealign (in1[3], in1[2], 1);
- out0[3] = amd_bytealign ( 0, in1[3], 1);
+ case 17: out0[0] = amd_bytealign_S (in1[1], in1[0], 1);
+ out0[1] = amd_bytealign_S (in1[2], in1[1], 1);
+ out0[2] = amd_bytealign_S (in1[3], in1[2], 1);
+ out0[3] = amd_bytealign_S ( 0, in1[3], 1);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 18: out0[0] = amd_bytealign (in1[1], in1[0], 2);
- out0[1] = amd_bytealign (in1[2], in1[1], 2);
- out0[2] = amd_bytealign (in1[3], in1[2], 2);
- out0[3] = amd_bytealign ( 0, in1[3], 2);
+ case 18: out0[0] = amd_bytealign_S (in1[1], in1[0], 2);
+ out0[1] = amd_bytealign_S (in1[2], in1[1], 2);
+ out0[2] = amd_bytealign_S (in1[3], in1[2], 2);
+ out0[3] = amd_bytealign_S ( 0, in1[3], 2);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 19: out0[0] = amd_bytealign (in1[1], in1[0], 3);
- out0[1] = amd_bytealign (in1[2], in1[1], 3);
- out0[2] = amd_bytealign (in1[3], in1[2], 3);
- out0[3] = amd_bytealign ( 0, in1[3], 3);
+ case 19: out0[0] = amd_bytealign_S (in1[1], in1[0], 3);
+ out0[1] = amd_bytealign_S (in1[2], in1[1], 3);
+ out0[2] = amd_bytealign_S (in1[3], in1[2], 3);
+ out0[3] = amd_bytealign_S ( 0, in1[3], 3);
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 21: out0[0] = amd_bytealign (in1[2], in1[1], 1);
- out0[1] = amd_bytealign (in1[3], in1[2], 1);
- out0[2] = amd_bytealign ( 0, in1[3], 1);
+ case 21: out0[0] = amd_bytealign_S (in1[2], in1[1], 1);
+ out0[1] = amd_bytealign_S (in1[3], in1[2], 1);
+ out0[2] = amd_bytealign_S ( 0, in1[3], 1);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 22: out0[0] = amd_bytealign (in1[2], in1[1], 2);
- out0[1] = amd_bytealign (in1[3], in1[2], 2);
- out0[2] = amd_bytealign ( 0, in1[3], 2);
+ case 22: out0[0] = amd_bytealign_S (in1[2], in1[1], 2);
+ out0[1] = amd_bytealign_S (in1[3], in1[2], 2);
+ out0[2] = amd_bytealign_S ( 0, in1[3], 2);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 23: out0[0] = amd_bytealign (in1[2], in1[1], 3);
- out0[1] = amd_bytealign (in1[3], in1[2], 3);
- out0[2] = amd_bytealign ( 0, in1[3], 3);
+ case 23: out0[0] = amd_bytealign_S (in1[2], in1[1], 3);
+ out0[1] = amd_bytealign_S (in1[3], in1[2], 3);
+ out0[2] = amd_bytealign_S ( 0, in1[3], 3);
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 25: out0[0] = amd_bytealign (in1[3], in1[2], 1);
- out0[1] = amd_bytealign ( 0, in1[3], 1);
+ case 25: out0[0] = amd_bytealign_S (in1[3], in1[2], 1);
+ out0[1] = amd_bytealign_S ( 0, in1[3], 1);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 26: out0[0] = amd_bytealign (in1[3], in1[2], 2);
- out0[1] = amd_bytealign ( 0, in1[3], 2);
+ case 26: out0[0] = amd_bytealign_S (in1[3], in1[2], 2);
+ out0[1] = amd_bytealign_S ( 0, in1[3], 2);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 27: out0[0] = amd_bytealign (in1[3], in1[2], 3);
- out0[1] = amd_bytealign ( 0, in1[3], 3);
+ case 27: out0[0] = amd_bytealign_S (in1[3], in1[2], 3);
+ out0[1] = amd_bytealign_S ( 0, in1[3], 3);
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 29: out0[0] = amd_bytealign ( 0, in1[3], 1);
+ case 29: out0[0] = amd_bytealign_S ( 0, in1[3], 1);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 30: out0[0] = amd_bytealign ( 0, in1[3], 2);
+ case 30: out0[0] = amd_bytealign_S ( 0, in1[3], 2);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out1[2] = 0;
out1[3] = 0;
break;
- case 31: out0[0] = amd_bytealign ( 0, in1[3], 3);
+ case 31: out0[0] = amd_bytealign_S ( 0, in1[3], 3);
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
out0[1] = in0[1];
out0[0] = in0[0];
break;
- case 1: out1[3] = __byte_perm (in1[2], in1[3], 0x6543);
- out1[2] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[1] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[0] = __byte_perm (in0[3], in1[0], 0x6543);
- out0[3] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[2] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[1] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[0] = __byte_perm ( 0, in0[0], 0x6543);
- break;
- case 2: out1[3] = __byte_perm (in1[2], in1[3], 0x5432);
- out1[2] = __byte_perm (in1[1], in1[2], 0x5432);
- out1[1] = __byte_perm (in1[0], in1[1], 0x5432);
- out1[0] = __byte_perm (in0[3], in1[0], 0x5432);
- out0[3] = __byte_perm (in0[2], in0[3], 0x5432);
- out0[2] = __byte_perm (in0[1], in0[2], 0x5432);
- out0[1] = __byte_perm (in0[0], in0[1], 0x5432);
- out0[0] = __byte_perm ( 0, in0[0], 0x5432);
- break;
- case 3: out1[3] = __byte_perm (in1[2], in1[3], 0x4321);
- out1[2] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[1] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[0] = __byte_perm (in0[3], in1[0], 0x4321);
- out0[3] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[2] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[1] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[0] = __byte_perm ( 0, in0[0], 0x4321);
+ case 1: out1[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
+ out1[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out0[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[0] = __byte_perm_S ( 0, in0[0], 0x6543);
+ break;
+ case 2: out1[3] = __byte_perm_S (in1[2], in1[3], 0x5432);
+ out1[2] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out1[1] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out1[0] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out0[3] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out0[2] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out0[1] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out0[0] = __byte_perm_S ( 0, in0[0], 0x5432);
+ break;
+ case 3: out1[3] = __byte_perm_S (in1[2], in1[3], 0x4321);
+ out1[2] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[1] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[0] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out0[3] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[2] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[1] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[0] = __byte_perm_S ( 0, in0[0], 0x4321);
break;
case 4: out1[3] = in1[2];
out1[2] = in1[1];
out0[1] = in0[0];
out0[0] = 0;
break;
- case 5: out1[3] = __byte_perm (in1[1], in1[2], 0x6543);
- out1[2] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[1] = __byte_perm (in0[3], in1[0], 0x6543);
- out1[0] = __byte_perm (in0[2], in0[3], 0x6543);
- out0[3] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[2] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[1] = __byte_perm ( 0, in0[0], 0x6543);
+ case 5: out1[3] = __byte_perm_S (in1[1], in1[2], 0x6543);
+ out1[2] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[1] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out1[0] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out0[3] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[2] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[1] = __byte_perm_S ( 0, in0[0], 0x6543);
out0[0] = 0;
break;
- case 6: out1[3] = __byte_perm (in1[1], in1[2], 0x5432);
- out1[2] = __byte_perm (in1[0], in1[1], 0x5432);
- out1[1] = __byte_perm (in0[3], in1[0], 0x5432);
- out1[0] = __byte_perm (in0[2], in0[3], 0x5432);
- out0[3] = __byte_perm (in0[1], in0[2], 0x5432);
- out0[2] = __byte_perm (in0[0], in0[1], 0x5432);
- out0[1] = __byte_perm ( 0, in0[0], 0x5432);
+ case 6: out1[3] = __byte_perm_S (in1[1], in1[2], 0x5432);
+ out1[2] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out1[1] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out1[0] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out0[3] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out0[2] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out0[1] = __byte_perm_S ( 0, in0[0], 0x5432);
out0[0] = 0;
break;
- case 7: out1[3] = __byte_perm (in1[1], in1[2], 0x4321);
- out1[2] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[1] = __byte_perm (in0[3], in1[0], 0x4321);
- out1[0] = __byte_perm (in0[2], in0[3], 0x4321);
- out0[3] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[2] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[1] = __byte_perm ( 0, in0[0], 0x4321);
+ case 7: out1[3] = __byte_perm_S (in1[1], in1[2], 0x4321);
+ out1[2] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[1] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out1[0] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out0[3] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[2] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[1] = __byte_perm_S ( 0, in0[0], 0x4321);
out0[0] = 0;
break;
case 8: out1[3] = in1[1];
out0[1] = 0;
out0[0] = 0;
break;
- case 9: out1[3] = __byte_perm (in1[0], in1[1], 0x6543);
- out1[2] = __byte_perm (in0[3], in1[0], 0x6543);
- out1[1] = __byte_perm (in0[2], in0[3], 0x6543);
- out1[0] = __byte_perm (in0[1], in0[2], 0x6543);
- out0[3] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[2] = __byte_perm ( 0, in0[0], 0x6543);
+ case 9: out1[3] = __byte_perm_S (in1[0], in1[1], 0x6543);
+ out1[2] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out1[1] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out1[0] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out0[3] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[2] = __byte_perm_S ( 0, in0[0], 0x6543);
out0[1] = 0;
out0[0] = 0;
break;
- case 10: out1[3] = __byte_perm (in1[0], in1[1], 0x5432);
- out1[2] = __byte_perm (in0[3], in1[0], 0x5432);
- out1[1] = __byte_perm (in0[2], in0[3], 0x5432);
- out1[0] = __byte_perm (in0[1], in0[2], 0x5432);
- out0[3] = __byte_perm (in0[0], in0[1], 0x5432);
- out0[2] = __byte_perm ( 0, in0[0], 0x5432);
+ case 10: out1[3] = __byte_perm_S (in1[0], in1[1], 0x5432);
+ out1[2] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out1[1] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out1[0] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out0[3] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out0[2] = __byte_perm_S ( 0, in0[0], 0x5432);
out0[1] = 0;
out0[0] = 0;
break;
- case 11: out1[3] = __byte_perm (in1[0], in1[1], 0x4321);
- out1[2] = __byte_perm (in0[3], in1[0], 0x4321);
- out1[1] = __byte_perm (in0[2], in0[3], 0x4321);
- out1[0] = __byte_perm (in0[1], in0[2], 0x4321);
- out0[3] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[2] = __byte_perm ( 0, in0[0], 0x4321);
+ case 11: out1[3] = __byte_perm_S (in1[0], in1[1], 0x4321);
+ out1[2] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out1[1] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out1[0] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out0[3] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[2] = __byte_perm_S ( 0, in0[0], 0x4321);
out0[1] = 0;
out0[0] = 0;
break;
out0[1] = 0;
out0[0] = 0;
break;
- case 13: out1[3] = __byte_perm (in0[3], in1[0], 0x6543);
- out1[2] = __byte_perm (in0[2], in0[3], 0x6543);
- out1[1] = __byte_perm (in0[1], in0[2], 0x6543);
- out1[0] = __byte_perm (in0[0], in0[1], 0x6543);
- out0[3] = __byte_perm ( 0, in0[0], 0x6543);
+ case 13: out1[3] = __byte_perm_S (in0[3], in1[0], 0x6543);
+ out1[2] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out1[1] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out1[0] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out0[3] = __byte_perm_S ( 0, in0[0], 0x6543);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 14: out1[3] = __byte_perm (in0[3], in1[0], 0x5432);
- out1[2] = __byte_perm (in0[2], in0[3], 0x5432);
- out1[1] = __byte_perm (in0[1], in0[2], 0x5432);
- out1[0] = __byte_perm (in0[0], in0[1], 0x5432);
- out0[3] = __byte_perm ( 0, in0[0], 0x5432);
+ case 14: out1[3] = __byte_perm_S (in0[3], in1[0], 0x5432);
+ out1[2] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out1[1] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out1[0] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out0[3] = __byte_perm_S ( 0, in0[0], 0x5432);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 15: out1[3] = __byte_perm (in0[3], in1[0], 0x4321);
- out1[2] = __byte_perm (in0[2], in0[3], 0x4321);
- out1[1] = __byte_perm (in0[1], in0[2], 0x4321);
- out1[0] = __byte_perm (in0[0], in0[1], 0x4321);
- out0[3] = __byte_perm ( 0, in0[0], 0x4321);
+ case 15: out1[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
+ out1[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out1[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out1[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out0[3] = __byte_perm_S ( 0, in0[0], 0x4321);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 17: out1[3] = __byte_perm (in0[2], in0[3], 0x6543);
- out1[2] = __byte_perm (in0[1], in0[2], 0x6543);
- out1[1] = __byte_perm (in0[0], in0[1], 0x6543);
- out1[0] = __byte_perm ( 0, in0[0], 0x6543);
+ case 17: out1[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
+ out1[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out1[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out1[0] = __byte_perm_S ( 0, in0[0], 0x6543);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 18: out1[3] = __byte_perm (in0[2], in0[3], 0x5432);
- out1[2] = __byte_perm (in0[1], in0[2], 0x5432);
- out1[1] = __byte_perm (in0[0], in0[1], 0x5432);
- out1[0] = __byte_perm ( 0, in0[0], 0x5432);
+ case 18: out1[3] = __byte_perm_S (in0[2], in0[3], 0x5432);
+ out1[2] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out1[1] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out1[0] = __byte_perm_S ( 0, in0[0], 0x5432);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 19: out1[3] = __byte_perm (in0[2], in0[3], 0x4321);
- out1[2] = __byte_perm (in0[1], in0[2], 0x4321);
- out1[1] = __byte_perm (in0[0], in0[1], 0x4321);
- out1[0] = __byte_perm ( 0, in0[0], 0x4321);
+ case 19: out1[3] = __byte_perm_S (in0[2], in0[3], 0x4321);
+ out1[2] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out1[1] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out1[0] = __byte_perm_S ( 0, in0[0], 0x4321);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 21: out1[3] = __byte_perm (in0[1], in0[2], 0x6543);
- out1[2] = __byte_perm (in0[0], in0[1], 0x6543);
- out1[1] = __byte_perm ( 0, in0[0], 0x6543);
+ case 21: out1[3] = __byte_perm_S (in0[1], in0[2], 0x6543);
+ out1[2] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out1[1] = __byte_perm_S ( 0, in0[0], 0x6543);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 22: out1[3] = __byte_perm (in0[1], in0[2], 0x5432);
- out1[2] = __byte_perm (in0[0], in0[1], 0x5432);
- out1[1] = __byte_perm ( 0, in0[0], 0x5432);
+ case 22: out1[3] = __byte_perm_S (in0[1], in0[2], 0x5432);
+ out1[2] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out1[1] = __byte_perm_S ( 0, in0[0], 0x5432);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 23: out1[3] = __byte_perm (in0[1], in0[2], 0x4321);
- out1[2] = __byte_perm (in0[0], in0[1], 0x4321);
- out1[1] = __byte_perm ( 0, in0[0], 0x4321);
+ case 23: out1[3] = __byte_perm_S (in0[1], in0[2], 0x4321);
+ out1[2] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out1[1] = __byte_perm_S ( 0, in0[0], 0x4321);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 25: out1[3] = __byte_perm (in0[0], in0[1], 0x6543);
- out1[2] = __byte_perm ( 0, in0[0], 0x6543);
+ case 25: out1[3] = __byte_perm_S (in0[0], in0[1], 0x6543);
+ out1[2] = __byte_perm_S ( 0, in0[0], 0x6543);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 26: out1[3] = __byte_perm (in0[0], in0[1], 0x5432);
- out1[2] = __byte_perm ( 0, in0[0], 0x5432);
+ case 26: out1[3] = __byte_perm_S (in0[0], in0[1], 0x5432);
+ out1[2] = __byte_perm_S ( 0, in0[0], 0x5432);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 27: out1[3] = __byte_perm (in0[0], in0[1], 0x4321);
- out1[2] = __byte_perm ( 0, in0[0], 0x4321);
+ case 27: out1[3] = __byte_perm_S (in0[0], in0[1], 0x4321);
+ out1[2] = __byte_perm_S ( 0, in0[0], 0x4321);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 29: out1[3] = __byte_perm ( 0, in0[0], 0x6543);
+ case 29: out1[3] = __byte_perm_S ( 0, in0[0], 0x6543);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 30: out1[3] = __byte_perm ( 0, in0[0], 0x5432);
+ case 30: out1[3] = __byte_perm_S ( 0, in0[0], 0x5432);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 31: out1[3] = __byte_perm ( 0, in0[0], 0x4321);
+ case 31: out1[3] = __byte_perm_S ( 0, in0[0], 0x4321);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
out0[1] = in0[1];
out0[0] = in0[0];
break;
- case 1: out1[3] = amd_bytealign (in1[3], in1[2], 3);
- out1[2] = amd_bytealign (in1[2], in1[1], 3);
- out1[1] = amd_bytealign (in1[1], in1[0], 3);
- out1[0] = amd_bytealign (in1[0], in0[3], 3);
- out0[3] = amd_bytealign (in0[3], in0[2], 3);
- out0[2] = amd_bytealign (in0[2], in0[1], 3);
- out0[1] = amd_bytealign (in0[1], in0[0], 3);
- out0[0] = amd_bytealign (in0[0], 0, 3);
- break;
- case 2: out1[3] = amd_bytealign (in1[3], in1[2], 2);
- out1[2] = amd_bytealign (in1[2], in1[1], 2);
- out1[1] = amd_bytealign (in1[1], in1[0], 2);
- out1[0] = amd_bytealign (in1[0], in0[3], 2);
- out0[3] = amd_bytealign (in0[3], in0[2], 2);
- out0[2] = amd_bytealign (in0[2], in0[1], 2);
- out0[1] = amd_bytealign (in0[1], in0[0], 2);
- out0[0] = amd_bytealign (in0[0], 0, 2);
- break;
- case 3: out1[3] = amd_bytealign (in1[3], in1[2], 1);
- out1[2] = amd_bytealign (in1[2], in1[1], 1);
- out1[1] = amd_bytealign (in1[1], in1[0], 1);
- out1[0] = amd_bytealign (in1[0], in0[3], 1);
- out0[3] = amd_bytealign (in0[3], in0[2], 1);
- out0[2] = amd_bytealign (in0[2], in0[1], 1);
- out0[1] = amd_bytealign (in0[1], in0[0], 1);
- out0[0] = amd_bytealign (in0[0], 0, 1);
+ case 1: out1[3] = amd_bytealign_S (in1[3], in1[2], 3);
+ out1[2] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[1] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[0] = amd_bytealign_S (in1[0], in0[3], 3);
+ out0[3] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[2] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[1] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[0] = amd_bytealign_S (in0[0], 0, 3);
+ break;
+ case 2: out1[3] = amd_bytealign_S (in1[3], in1[2], 2);
+ out1[2] = amd_bytealign_S (in1[2], in1[1], 2);
+ out1[1] = amd_bytealign_S (in1[1], in1[0], 2);
+ out1[0] = amd_bytealign_S (in1[0], in0[3], 2);
+ out0[3] = amd_bytealign_S (in0[3], in0[2], 2);
+ out0[2] = amd_bytealign_S (in0[2], in0[1], 2);
+ out0[1] = amd_bytealign_S (in0[1], in0[0], 2);
+ out0[0] = amd_bytealign_S (in0[0], 0, 2);
+ break;
+ case 3: out1[3] = amd_bytealign_S (in1[3], in1[2], 1);
+ out1[2] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[1] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[0] = amd_bytealign_S (in1[0], in0[3], 1);
+ out0[3] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[2] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[1] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[0] = amd_bytealign_S (in0[0], 0, 1);
break;
case 4: out1[3] = in1[2];
out1[2] = in1[1];
out0[1] = in0[0];
out0[0] = 0;
break;
- case 5: out1[3] = amd_bytealign (in1[2], in1[1], 3);
- out1[2] = amd_bytealign (in1[1], in1[0], 3);
- out1[1] = amd_bytealign (in1[0], in0[3], 3);
- out1[0] = amd_bytealign (in0[3], in0[2], 3);
- out0[3] = amd_bytealign (in0[2], in0[1], 3);
- out0[2] = amd_bytealign (in0[1], in0[0], 3);
- out0[1] = amd_bytealign (in0[0], 0, 3);
+ case 5: out1[3] = amd_bytealign_S (in1[2], in1[1], 3);
+ out1[2] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[1] = amd_bytealign_S (in1[0], in0[3], 3);
+ out1[0] = amd_bytealign_S (in0[3], in0[2], 3);
+ out0[3] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[2] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[1] = amd_bytealign_S (in0[0], 0, 3);
out0[0] = 0;
break;
- case 6: out1[3] = amd_bytealign (in1[2], in1[1], 2);
- out1[2] = amd_bytealign (in1[1], in1[0], 2);
- out1[1] = amd_bytealign (in1[0], in0[3], 2);
- out1[0] = amd_bytealign (in0[3], in0[2], 2);
- out0[3] = amd_bytealign (in0[2], in0[1], 2);
- out0[2] = amd_bytealign (in0[1], in0[0], 2);
- out0[1] = amd_bytealign (in0[0], 0, 2);
+ case 6: out1[3] = amd_bytealign_S (in1[2], in1[1], 2);
+ out1[2] = amd_bytealign_S (in1[1], in1[0], 2);
+ out1[1] = amd_bytealign_S (in1[0], in0[3], 2);
+ out1[0] = amd_bytealign_S (in0[3], in0[2], 2);
+ out0[3] = amd_bytealign_S (in0[2], in0[1], 2);
+ out0[2] = amd_bytealign_S (in0[1], in0[0], 2);
+ out0[1] = amd_bytealign_S (in0[0], 0, 2);
out0[0] = 0;
break;
- case 7: out1[3] = amd_bytealign (in1[2], in1[1], 1);
- out1[2] = amd_bytealign (in1[1], in1[0], 1);
- out1[1] = amd_bytealign (in1[0], in0[3], 1);
- out1[0] = amd_bytealign (in0[3], in0[2], 1);
- out0[3] = amd_bytealign (in0[2], in0[1], 1);
- out0[2] = amd_bytealign (in0[1], in0[0], 1);
- out0[1] = amd_bytealign (in0[0], 0, 1);
+ case 7: out1[3] = amd_bytealign_S (in1[2], in1[1], 1);
+ out1[2] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[1] = amd_bytealign_S (in1[0], in0[3], 1);
+ out1[0] = amd_bytealign_S (in0[3], in0[2], 1);
+ out0[3] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[2] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[1] = amd_bytealign_S (in0[0], 0, 1);
out0[0] = 0;
break;
case 8: out1[3] = in1[1];
out0[1] = 0;
out0[0] = 0;
break;
- case 9: out1[3] = amd_bytealign (in1[1], in1[0], 3);
- out1[2] = amd_bytealign (in1[0], in0[3], 3);
- out1[1] = amd_bytealign (in0[3], in0[2], 3);
- out1[0] = amd_bytealign (in0[2], in0[1], 3);
- out0[3] = amd_bytealign (in0[1], in0[0], 3);
- out0[2] = amd_bytealign (in0[0], 0, 3);
+ case 9: out1[3] = amd_bytealign_S (in1[1], in1[0], 3);
+ out1[2] = amd_bytealign_S (in1[0], in0[3], 3);
+ out1[1] = amd_bytealign_S (in0[3], in0[2], 3);
+ out1[0] = amd_bytealign_S (in0[2], in0[1], 3);
+ out0[3] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[2] = amd_bytealign_S (in0[0], 0, 3);
out0[1] = 0;
out0[0] = 0;
break;
- case 10: out1[3] = amd_bytealign (in1[1], in1[0], 2);
- out1[2] = amd_bytealign (in1[0], in0[3], 2);
- out1[1] = amd_bytealign (in0[3], in0[2], 2);
- out1[0] = amd_bytealign (in0[2], in0[1], 2);
- out0[3] = amd_bytealign (in0[1], in0[0], 2);
- out0[2] = amd_bytealign (in0[0], 0, 2);
+ case 10: out1[3] = amd_bytealign_S (in1[1], in1[0], 2);
+ out1[2] = amd_bytealign_S (in1[0], in0[3], 2);
+ out1[1] = amd_bytealign_S (in0[3], in0[2], 2);
+ out1[0] = amd_bytealign_S (in0[2], in0[1], 2);
+ out0[3] = amd_bytealign_S (in0[1], in0[0], 2);
+ out0[2] = amd_bytealign_S (in0[0], 0, 2);
out0[1] = 0;
out0[0] = 0;
break;
- case 11: out1[3] = amd_bytealign (in1[1], in1[0], 1);
- out1[2] = amd_bytealign (in1[0], in0[3], 1);
- out1[1] = amd_bytealign (in0[3], in0[2], 1);
- out1[0] = amd_bytealign (in0[2], in0[1], 1);
- out0[3] = amd_bytealign (in0[1], in0[0], 1);
- out0[2] = amd_bytealign (in0[0], 0, 1);
+ case 11: out1[3] = amd_bytealign_S (in1[1], in1[0], 1);
+ out1[2] = amd_bytealign_S (in1[0], in0[3], 1);
+ out1[1] = amd_bytealign_S (in0[3], in0[2], 1);
+ out1[0] = amd_bytealign_S (in0[2], in0[1], 1);
+ out0[3] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[2] = amd_bytealign_S (in0[0], 0, 1);
out0[1] = 0;
out0[0] = 0;
break;
out0[1] = 0;
out0[0] = 0;
break;
- case 13: out1[3] = amd_bytealign (in1[0], in0[3], 3);
- out1[2] = amd_bytealign (in0[3], in0[2], 3);
- out1[1] = amd_bytealign (in0[2], in0[1], 3);
- out1[0] = amd_bytealign (in0[1], in0[0], 3);
- out0[3] = amd_bytealign (in0[0], 0, 3);
+ case 13: out1[3] = amd_bytealign_S (in1[0], in0[3], 3);
+ out1[2] = amd_bytealign_S (in0[3], in0[2], 3);
+ out1[1] = amd_bytealign_S (in0[2], in0[1], 3);
+ out1[0] = amd_bytealign_S (in0[1], in0[0], 3);
+ out0[3] = amd_bytealign_S (in0[0], 0, 3);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 14: out1[3] = amd_bytealign (in1[0], in0[3], 2);
- out1[2] = amd_bytealign (in0[3], in0[2], 2);
- out1[1] = amd_bytealign (in0[2], in0[1], 2);
- out1[0] = amd_bytealign (in0[1], in0[0], 2);
- out0[3] = amd_bytealign (in0[0], 0, 2);
+ case 14: out1[3] = amd_bytealign_S (in1[0], in0[3], 2);
+ out1[2] = amd_bytealign_S (in0[3], in0[2], 2);
+ out1[1] = amd_bytealign_S (in0[2], in0[1], 2);
+ out1[0] = amd_bytealign_S (in0[1], in0[0], 2);
+ out0[3] = amd_bytealign_S (in0[0], 0, 2);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 15: out1[3] = amd_bytealign (in1[0], in0[3], 1);
- out1[2] = amd_bytealign (in0[3], in0[2], 1);
- out1[1] = amd_bytealign (in0[2], in0[1], 1);
- out1[0] = amd_bytealign (in0[1], in0[0], 1);
- out0[3] = amd_bytealign (in0[0], 0, 1);
+ case 15: out1[3] = amd_bytealign_S (in1[0], in0[3], 1);
+ out1[2] = amd_bytealign_S (in0[3], in0[2], 1);
+ out1[1] = amd_bytealign_S (in0[2], in0[1], 1);
+ out1[0] = amd_bytealign_S (in0[1], in0[0], 1);
+ out0[3] = amd_bytealign_S (in0[0], 0, 1);
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 17: out1[3] = amd_bytealign (in0[3], in0[2], 3);
- out1[2] = amd_bytealign (in0[2], in0[1], 3);
- out1[1] = amd_bytealign (in0[1], in0[0], 3);
- out1[0] = amd_bytealign (in0[0], 0, 3);
+ case 17: out1[3] = amd_bytealign_S (in0[3], in0[2], 3);
+ out1[2] = amd_bytealign_S (in0[2], in0[1], 3);
+ out1[1] = amd_bytealign_S (in0[1], in0[0], 3);
+ out1[0] = amd_bytealign_S (in0[0], 0, 3);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 18: out1[3] = amd_bytealign (in0[3], in0[2], 2);
- out1[2] = amd_bytealign (in0[2], in0[1], 2);
- out1[1] = amd_bytealign (in0[1], in0[0], 2);
- out1[0] = amd_bytealign (in0[0], 0, 2);
+ case 18: out1[3] = amd_bytealign_S (in0[3], in0[2], 2);
+ out1[2] = amd_bytealign_S (in0[2], in0[1], 2);
+ out1[1] = amd_bytealign_S (in0[1], in0[0], 2);
+ out1[0] = amd_bytealign_S (in0[0], 0, 2);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 19: out1[3] = amd_bytealign (in0[3], in0[2], 1);
- out1[2] = amd_bytealign (in0[2], in0[1], 1);
- out1[1] = amd_bytealign (in0[1], in0[0], 1);
- out1[0] = amd_bytealign (in0[0], 0, 1);
+ case 19: out1[3] = amd_bytealign_S (in0[3], in0[2], 1);
+ out1[2] = amd_bytealign_S (in0[2], in0[1], 1);
+ out1[1] = amd_bytealign_S (in0[1], in0[0], 1);
+ out1[0] = amd_bytealign_S (in0[0], 0, 1);
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 21: out1[3] = amd_bytealign (in0[2], in0[1], 3);
- out1[2] = amd_bytealign (in0[1], in0[0], 3);
- out1[1] = amd_bytealign (in0[0], 0, 3);
+ case 21: out1[3] = amd_bytealign_S (in0[2], in0[1], 3);
+ out1[2] = amd_bytealign_S (in0[1], in0[0], 3);
+ out1[1] = amd_bytealign_S (in0[0], 0, 3);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 22: out1[3] = amd_bytealign (in0[2], in0[1], 2);
- out1[2] = amd_bytealign (in0[1], in0[0], 2);
- out1[1] = amd_bytealign (in0[0], 0, 2);
+ case 22: out1[3] = amd_bytealign_S (in0[2], in0[1], 2);
+ out1[2] = amd_bytealign_S (in0[1], in0[0], 2);
+ out1[1] = amd_bytealign_S (in0[0], 0, 2);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 23: out1[3] = amd_bytealign (in0[2], in0[1], 1);
- out1[2] = amd_bytealign (in0[1], in0[0], 1);
- out1[1] = amd_bytealign (in0[0], 0, 1);
+ case 23: out1[3] = amd_bytealign_S (in0[2], in0[1], 1);
+ out1[2] = amd_bytealign_S (in0[1], in0[0], 1);
+ out1[1] = amd_bytealign_S (in0[0], 0, 1);
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 25: out1[3] = amd_bytealign (in0[1], in0[0], 3);
- out1[2] = amd_bytealign (in0[0], 0, 3);
+ case 25: out1[3] = amd_bytealign_S (in0[1], in0[0], 3);
+ out1[2] = amd_bytealign_S (in0[0], 0, 3);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 26: out1[3] = amd_bytealign (in0[1], in0[0], 2);
- out1[2] = amd_bytealign (in0[0], 0, 2);
+ case 26: out1[3] = amd_bytealign_S (in0[1], in0[0], 2);
+ out1[2] = amd_bytealign_S (in0[0], 0, 2);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 27: out1[3] = amd_bytealign (in0[1], in0[0], 1);
- out1[2] = amd_bytealign (in0[0], 0, 1);
+ case 27: out1[3] = amd_bytealign_S (in0[1], in0[0], 1);
+ out1[2] = amd_bytealign_S (in0[0], 0, 1);
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 29: out1[3] = amd_bytealign (in0[0], 0, 3);
+ case 29: out1[3] = amd_bytealign_S (in0[0], 0, 3);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 30: out1[3] = amd_bytealign (in0[0], 0, 2);
+ case 30: out1[3] = amd_bytealign_S (in0[0], 0, 2);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
out0[1] = 0;
out0[0] = 0;
break;
- case 31: out1[3] = amd_bytealign (in0[0], 0, 1);
+ case 31: out1[3] = amd_bytealign_S (in0[0], 0, 1);
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
break;
case 1:
- dst0[0] = __byte_perm (src_l0[0], src_r0[0], 0x6540);
- dst0[1] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst0[2] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
- dst0[3] = __byte_perm (src_r0[2], src_r0[3], 0x6543);
- dst1[0] = __byte_perm (src_r0[3], src_r1[0], 0x6543);
- dst1[1] = __byte_perm (src_r1[0], src_r1[1], 0x6543);
- dst1[2] = __byte_perm (src_r1[1], src_r1[2], 0x6543);
- dst1[3] = __byte_perm (src_r1[2], src_r1[3], 0x6543);
+ dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x6540);
+ dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
+ dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
+ dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
+ dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
+ dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x6543);
+ dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x6543);
break;
case 2:
- dst0[0] = __byte_perm (src_l0[0], src_r0[0], 0x5410);
- dst0[1] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst0[2] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
- dst0[3] = __byte_perm (src_r0[2], src_r0[3], 0x5432);
- dst1[0] = __byte_perm (src_r0[3], src_r1[0], 0x5432);
- dst1[1] = __byte_perm (src_r1[0], src_r1[1], 0x5432);
- dst1[2] = __byte_perm (src_r1[1], src_r1[2], 0x5432);
- dst1[3] = __byte_perm (src_r1[2], src_r1[3], 0x5432);
+ dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x5410);
+ dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
+ dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
+ dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
+ dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
+ dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x5432);
+ dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x5432);
break;
case 3:
- dst0[0] = __byte_perm (src_l0[0], src_r0[0], 0x4210);
- dst0[1] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst0[2] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
- dst0[3] = __byte_perm (src_r0[2], src_r0[3], 0x4321);
- dst1[0] = __byte_perm (src_r0[3], src_r1[0], 0x4321);
- dst1[1] = __byte_perm (src_r1[0], src_r1[1], 0x4321);
- dst1[2] = __byte_perm (src_r1[1], src_r1[2], 0x4321);
- dst1[3] = __byte_perm (src_r1[2], src_r1[3], 0x4321);
+ dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x4210);
+ dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
+ dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
+ dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
+ dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
+ dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x4321);
+ dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x4321);
break;
case 4:
break;
case 5:
- dst0[1] = __byte_perm (src_l0[1], src_r0[0], 0x6540);
- dst0[2] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst0[3] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
- dst1[0] = __byte_perm (src_r0[2], src_r0[3], 0x6543);
- dst1[1] = __byte_perm (src_r0[3], src_r1[0], 0x6543);
- dst1[2] = __byte_perm (src_r1[0], src_r1[1], 0x6543);
- dst1[3] = __byte_perm (src_r1[1], src_r1[2], 0x6543);
+ dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x6540);
+ dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
+ dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
+ dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
+ dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
+ dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x6543);
break;
case 6:
- dst0[1] = __byte_perm (src_l0[1], src_r0[0], 0x5410);
- dst0[2] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst0[3] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
- dst1[0] = __byte_perm (src_r0[2], src_r0[3], 0x5432);
- dst1[1] = __byte_perm (src_r0[3], src_r1[0], 0x5432);
- dst1[2] = __byte_perm (src_r1[0], src_r1[1], 0x5432);
- dst1[3] = __byte_perm (src_r1[1], src_r1[2], 0x5432);
+ dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x5410);
+ dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
+ dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
+ dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
+ dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
+ dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x5432);
break;
case 7:
- dst0[1] = __byte_perm (src_l0[1], src_r0[0], 0x4210);
- dst0[2] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst0[3] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
- dst1[0] = __byte_perm (src_r0[2], src_r0[3], 0x4321);
- dst1[1] = __byte_perm (src_r0[3], src_r1[0], 0x4321);
- dst1[2] = __byte_perm (src_r1[0], src_r1[1], 0x4321);
- dst1[3] = __byte_perm (src_r1[1], src_r1[2], 0x4321);
+ dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x4210);
+ dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
+ dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
+ dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
+ dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
+ dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x4321);
break;
case 8:
break;
case 9:
- dst0[2] = __byte_perm (src_l0[2], src_r0[0], 0x6540);
- dst0[3] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst1[0] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
- dst1[1] = __byte_perm (src_r0[2], src_r0[3], 0x6543);
- dst1[2] = __byte_perm (src_r0[3], src_r1[0], 0x6543);
- dst1[3] = __byte_perm (src_r1[0], src_r1[1], 0x6543);
+ dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x6540);
+ dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
+ dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
+ dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
+ dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
break;
case 10:
- dst0[2] = __byte_perm (src_l0[2], src_r0[0], 0x5410);
- dst0[3] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst1[0] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
- dst1[1] = __byte_perm (src_r0[2], src_r0[3], 0x5432);
- dst1[2] = __byte_perm (src_r0[3], src_r1[0], 0x5432);
- dst1[3] = __byte_perm (src_r1[0], src_r1[1], 0x5432);
+ dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x5410);
+ dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
+ dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
+ dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
+ dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
break;
case 11:
- dst0[2] = __byte_perm (src_l0[2], src_r0[0], 0x4210);
- dst0[3] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst1[0] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
- dst1[1] = __byte_perm (src_r0[2], src_r0[3], 0x4321);
- dst1[2] = __byte_perm (src_r0[3], src_r1[0], 0x4321);
- dst1[3] = __byte_perm (src_r1[0], src_r1[1], 0x4321);
+ dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x4210);
+ dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
+ dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
+ dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
+ dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
break;
case 12:
break;
case 13:
- dst0[3] = __byte_perm (src_l0[3], src_r0[0], 0x6540);
- dst1[0] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst1[1] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
- dst1[2] = __byte_perm (src_r0[2], src_r0[3], 0x6543);
- dst1[3] = __byte_perm (src_r0[3], src_r1[0], 0x6543);
+ dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x6540);
+ dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
+ dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
+ dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
break;
case 14:
- dst0[3] = __byte_perm (src_l0[3], src_r0[0], 0x5410);
- dst1[0] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst1[1] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
- dst1[2] = __byte_perm (src_r0[2], src_r0[3], 0x5432);
- dst1[3] = __byte_perm (src_r0[3], src_r1[0], 0x5432);
+ dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x5410);
+ dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
+ dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
+ dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
break;
case 15:
- dst0[3] = __byte_perm (src_l0[3], src_r0[0], 0x4210);
- dst1[0] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst1[1] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
- dst1[2] = __byte_perm (src_r0[2], src_r0[3], 0x4321);
- dst1[3] = __byte_perm (src_r0[3], src_r1[0], 0x4321);
+ dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x4210);
+ dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
+ dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
+ dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
break;
case 16:
break;
case 17:
- dst1[0] = __byte_perm (src_l1[0], src_r0[0], 0x6540);
- dst1[1] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst1[2] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
- dst1[3] = __byte_perm (src_r0[2], src_r0[3], 0x6543);
+ dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x6540);
+ dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
+ dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
break;
case 18:
- dst1[0] = __byte_perm (src_l1[0], src_r0[0], 0x5410);
- dst1[1] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst1[2] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
- dst1[3] = __byte_perm (src_r0[2], src_r0[3], 0x5432);
+ dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x5410);
+ dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
+ dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
break;
case 19:
- dst1[0] = __byte_perm (src_l1[0], src_r0[0], 0x4210);
- dst1[1] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst1[2] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
- dst1[3] = __byte_perm (src_r0[2], src_r0[3], 0x4321);
+ dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x4210);
+ dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
+ dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
break;
case 20:
break;
case 21:
- dst1[1] = __byte_perm (src_l1[1], src_r0[0], 0x6540);
- dst1[2] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
- dst1[3] = __byte_perm (src_r0[1], src_r0[2], 0x6543);
+ dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x6540);
+ dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
+ dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
break;
case 22:
- dst1[1] = __byte_perm (src_l1[1], src_r0[0], 0x5410);
- dst1[2] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
- dst1[3] = __byte_perm (src_r0[1], src_r0[2], 0x5432);
+ dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x5410);
+ dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
+ dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
break;
case 23:
- dst1[1] = __byte_perm (src_l1[1], src_r0[0], 0x4210);
- dst1[2] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
- dst1[3] = __byte_perm (src_r0[1], src_r0[2], 0x4321);
+ dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x4210);
+ dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
+ dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
break;
case 24:
break;
case 25:
- dst1[2] = __byte_perm (src_l1[2], src_r0[0], 0x6540);
- dst1[3] = __byte_perm (src_r0[0], src_r0[1], 0x6543);
+ dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x6540);
+ dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
break;
case 26:
- dst1[2] = __byte_perm (src_l1[2], src_r0[0], 0x5410);
- dst1[3] = __byte_perm (src_r0[0], src_r0[1], 0x5432);
+ dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x5410);
+ dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
break;
case 27:
- dst1[2] = __byte_perm (src_l1[2], src_r0[0], 0x4210);
- dst1[3] = __byte_perm (src_r0[0], src_r0[1], 0x4321);
+ dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x4210);
+ dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
break;
case 28:
break;
case 29:
- dst1[3] = __byte_perm (src_l1[3], src_r0[0], 0x6540);
+ dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x6540);
break;
case 30:
- dst1[3] = __byte_perm (src_l1[3], src_r0[0], 0x5410);
+ dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x5410);
break;
case 31:
- dst1[3] = __byte_perm (src_l1[3], src_r0[0], 0x4210);
+ dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x4210);
break;
}
#endif
dst1[3] = src_r0[0];
break;
case 27:
- dst1[3] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst1[2] = src_l1[2] | src_r0[0] << 24;
break;
case 26:
- dst1[3] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst1[2] = src_l1[2] | src_r0[0] << 16;
break;
case 25:
- dst1[3] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst1[2] = src_l1[2] | src_r0[0] << 8;
break;
case 24:
dst1[2] = src_r0[0];
break;
case 23:
- dst1[3] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst1[2] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst1[1] = src_l1[1] | src_r0[0] << 24;
break;
case 22:
- dst1[3] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst1[2] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst1[1] = src_l1[1] | src_r0[0] << 16;
break;
case 21:
- dst1[3] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst1[2] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst1[1] = src_l1[1] | src_r0[0] << 8;
break;
case 20:
dst1[1] = src_r0[0];
break;
case 19:
- dst1[3] = amd_bytealign (src_r0[3], src_r0[2], 1);
- dst1[2] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst1[1] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
+ dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst1[0] = src_l1[0] | src_r0[0] << 24;
break;
case 18:
- dst1[3] = amd_bytealign (src_r0[3], src_r0[2], 2);
- dst1[2] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst1[1] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
+ dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst1[0] = src_l1[0] | src_r0[0] << 16;
break;
case 17:
- dst1[3] = amd_bytealign (src_r0[3], src_r0[2], 3);
- dst1[2] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst1[1] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
+ dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst1[0] = src_l1[0] | src_r0[0] << 8;
break;
case 16:
dst1[0] = src_r0[0];
break;
case 15:
- dst1[3] = amd_bytealign (src_r1[0], src_r0[3], 1);
- dst1[2] = amd_bytealign (src_r0[3], src_r0[2], 1);
- dst1[1] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst1[0] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
+ dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
+ dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst0[3] = src_l0[3] | src_r0[0] << 24;
break;
case 14:
- dst1[3] = amd_bytealign (src_r1[0], src_r0[3], 2);
- dst1[2] = amd_bytealign (src_r0[3], src_r0[2], 2);
- dst1[1] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst1[0] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
+ dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
+ dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst0[3] = src_l0[3] | src_r0[0] << 16;
break;
case 13:
- dst1[3] = amd_bytealign (src_r1[0], src_r0[3], 3);
- dst1[2] = amd_bytealign (src_r0[3], src_r0[2], 3);
- dst1[1] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst1[0] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
+ dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
+ dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst0[3] = src_l0[3] | src_r0[0] << 8;
break;
case 12:
dst0[3] = src_r0[0];
break;
case 11:
- dst1[3] = amd_bytealign (src_r1[1], src_r1[0], 1);
- dst1[2] = amd_bytealign (src_r1[0], src_r0[3], 1);
- dst1[1] = amd_bytealign (src_r0[3], src_r0[2], 1);
- dst1[0] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst0[3] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
+ dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
+ dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
+ dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst0[2] = src_l0[2] | src_r0[0] << 24;
break;
case 10:
- dst1[3] = amd_bytealign (src_r1[1], src_r1[0], 2);
- dst1[2] = amd_bytealign (src_r1[0], src_r0[3], 2);
- dst1[1] = amd_bytealign (src_r0[3], src_r0[2], 2);
- dst1[0] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst0[3] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
+ dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
+ dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
+ dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst0[2] = src_l0[2] | src_r0[0] << 16;
break;
case 9:
- dst1[3] = amd_bytealign (src_r1[1], src_r1[0], 3);
- dst1[2] = amd_bytealign (src_r1[0], src_r0[3], 3);
- dst1[1] = amd_bytealign (src_r0[3], src_r0[2], 3);
- dst1[0] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst0[3] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
+ dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
+ dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
+ dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst0[2] = src_l0[2] | src_r0[0] << 8;
break;
case 8:
dst0[2] = src_r0[0];
break;
case 7:
- dst1[3] = amd_bytealign (src_r1[2], src_r1[1], 1);
- dst1[2] = amd_bytealign (src_r1[1], src_r1[0], 1);
- dst1[1] = amd_bytealign (src_r1[0], src_r0[3], 1);
- dst1[0] = amd_bytealign (src_r0[3], src_r0[2], 1);
- dst0[3] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst0[2] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 1);
+ dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
+ dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
+ dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
+ dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst0[1] = src_l0[1] | src_r0[0] << 24;
break;
case 6:
- dst1[3] = amd_bytealign (src_r1[2], src_r1[1], 2);
- dst1[2] = amd_bytealign (src_r1[1], src_r1[0], 2);
- dst1[1] = amd_bytealign (src_r1[0], src_r0[3], 2);
- dst1[0] = amd_bytealign (src_r0[3], src_r0[2], 2);
- dst0[3] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst0[2] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 2);
+ dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
+ dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
+ dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
+ dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst0[1] = src_l0[1] | src_r0[0] << 16;
break;
case 5:
- dst1[3] = amd_bytealign (src_r1[2], src_r1[1], 3);
- dst1[2] = amd_bytealign (src_r1[1], src_r1[0], 3);
- dst1[1] = amd_bytealign (src_r1[0], src_r0[3], 3);
- dst1[0] = amd_bytealign (src_r0[3], src_r0[2], 3);
- dst0[3] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst0[2] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 3);
+ dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
+ dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
+ dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
+ dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst0[1] = src_l0[1] | src_r0[0] << 8;
break;
case 4:
dst0[1] = src_r0[0];
break;
case 3:
- dst1[3] = amd_bytealign (src_r1[3], src_r1[2], 1);
- dst1[2] = amd_bytealign (src_r1[2], src_r1[1], 1);
- dst1[1] = amd_bytealign (src_r1[1], src_r1[0], 1);
- dst1[0] = amd_bytealign (src_r1[0], src_r0[3], 1);
- dst0[3] = amd_bytealign (src_r0[3], src_r0[2], 1);
- dst0[2] = amd_bytealign (src_r0[2], src_r0[1], 1);
- dst0[1] = amd_bytealign (src_r0[1], src_r0[0], 1);
+ dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 1);
+ dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 1);
+ dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
+ dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
+ dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
+ dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
+ dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
dst0[0] = src_l0[0] | src_r0[0] << 24;
break;
case 2:
- dst1[3] = amd_bytealign (src_r1[3], src_r1[2], 2);
- dst1[2] = amd_bytealign (src_r1[2], src_r1[1], 2);
- dst1[1] = amd_bytealign (src_r1[1], src_r1[0], 2);
- dst1[0] = amd_bytealign (src_r1[0], src_r0[3], 2);
- dst0[3] = amd_bytealign (src_r0[3], src_r0[2], 2);
- dst0[2] = amd_bytealign (src_r0[2], src_r0[1], 2);
- dst0[1] = amd_bytealign (src_r0[1], src_r0[0], 2);
+ dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 2);
+ dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 2);
+ dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
+ dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
+ dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
+ dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
+ dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
dst0[0] = src_l0[0] | src_r0[0] << 16;
break;
case 1:
- dst1[3] = amd_bytealign (src_r1[3], src_r1[2], 3);
- dst1[2] = amd_bytealign (src_r1[2], src_r1[1], 3);
- dst1[1] = amd_bytealign (src_r1[1], src_r1[0], 3);
- dst1[0] = amd_bytealign (src_r1[0], src_r0[3], 3);
- dst0[3] = amd_bytealign (src_r0[3], src_r0[2], 3);
- dst0[2] = amd_bytealign (src_r0[2], src_r0[1], 3);
- dst0[1] = amd_bytealign (src_r0[1], src_r0[0], 3);
+ dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 3);
+ dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 3);
+ dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
+ dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
+ dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
+ dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
+ dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
dst0[0] = src_l0[0] | src_r0[0] << 8;
break;
case 0:
tib41[2] = out0[1];
tib41[3] = out0[0];
- out0[0] = swap32 (tib40[0]);
- out0[1] = swap32 (tib40[1]);
- out0[2] = swap32 (tib40[2]);
- out0[3] = swap32 (tib40[3]);
- out1[0] = swap32 (tib41[0]);
- out1[1] = swap32 (tib41[1]);
- out1[2] = swap32 (tib41[2]);
- out1[3] = swap32 (tib41[3]);
+ out0[0] = swap32_S (tib40[0]);
+ out0[1] = swap32_S (tib40[1]);
+ out0[2] = swap32_S (tib40[2]);
+ out0[3] = swap32_S (tib40[3]);
+ out1[0] = swap32_S (tib41[0]);
+ out1[1] = swap32_S (tib41[1]);
+ out1[2] = swap32_S (tib41[2]);
+ out1[3] = swap32_S (tib41[3]);
}
static u32 rule_op_mangle_lrest (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
{
switch (i)
{
- case 0: if ((__byte_perm (buf0[0], 0, 0x6540)) == p0) buf0[0] = __byte_perm (p1, buf0[0], 0x7650);
+ case 0: if ((__byte_perm_S (buf0[0], 0, 0x6540)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7650);
break;
- case 1: if ((__byte_perm (buf0[0], 0, 0x6541)) == p0) buf0[0] = __byte_perm (p1, buf0[0], 0x7604);
+ case 1: if ((__byte_perm_S (buf0[0], 0, 0x6541)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7604);
break;
- case 2: if ((__byte_perm (buf0[0], 0, 0x6542)) == p0) buf0[0] = __byte_perm (p1, buf0[0], 0x7054);
+ case 2: if ((__byte_perm_S (buf0[0], 0, 0x6542)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7054);
break;
- case 3: if ((__byte_perm (buf0[0], 0, 0x6543)) == p0) buf0[0] = __byte_perm (p1, buf0[0], 0x0654);
+ case 3: if ((__byte_perm_S (buf0[0], 0, 0x6543)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x0654);
break;
- case 4: if ((__byte_perm (buf0[1], 0, 0x6540)) == p0) buf0[1] = __byte_perm (p1, buf0[1], 0x7650);
+ case 4: if ((__byte_perm_S (buf0[1], 0, 0x6540)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7650);
break;
- case 5: if ((__byte_perm (buf0[1], 0, 0x6541)) == p0) buf0[1] = __byte_perm (p1, buf0[1], 0x7604);
+ case 5: if ((__byte_perm_S (buf0[1], 0, 0x6541)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7604);
break;
- case 6: if ((__byte_perm (buf0[1], 0, 0x6542)) == p0) buf0[1] = __byte_perm (p1, buf0[1], 0x7054);
+ case 6: if ((__byte_perm_S (buf0[1], 0, 0x6542)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7054);
break;
- case 7: if ((__byte_perm (buf0[1], 0, 0x6543)) == p0) buf0[1] = __byte_perm (p1, buf0[1], 0x0654);
+ case 7: if ((__byte_perm_S (buf0[1], 0, 0x6543)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x0654);
break;
- case 8: if ((__byte_perm (buf0[2], 0, 0x6540)) == p0) buf0[2] = __byte_perm (p1, buf0[2], 0x7650);
+ case 8: if ((__byte_perm_S (buf0[2], 0, 0x6540)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7650);
break;
- case 9: if ((__byte_perm (buf0[2], 0, 0x6541)) == p0) buf0[2] = __byte_perm (p1, buf0[2], 0x7604);
+ case 9: if ((__byte_perm_S (buf0[2], 0, 0x6541)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7604);
break;
- case 10: if ((__byte_perm (buf0[2], 0, 0x6542)) == p0) buf0[2] = __byte_perm (p1, buf0[2], 0x7054);
+ case 10: if ((__byte_perm_S (buf0[2], 0, 0x6542)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7054);
break;
- case 11: if ((__byte_perm (buf0[2], 0, 0x6543)) == p0) buf0[2] = __byte_perm (p1, buf0[2], 0x0654);
+ case 11: if ((__byte_perm_S (buf0[2], 0, 0x6543)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x0654);
break;
- case 12: if ((__byte_perm (buf0[3], 0, 0x6540)) == p0) buf0[3] = __byte_perm (p1, buf0[3], 0x7650);
+ case 12: if ((__byte_perm_S (buf0[3], 0, 0x6540)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7650);
break;
- case 13: if ((__byte_perm (buf0[3], 0, 0x6541)) == p0) buf0[3] = __byte_perm (p1, buf0[3], 0x7604);
+ case 13: if ((__byte_perm_S (buf0[3], 0, 0x6541)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7604);
break;
- case 14: if ((__byte_perm (buf0[3], 0, 0x6542)) == p0) buf0[3] = __byte_perm (p1, buf0[3], 0x7054);
+ case 14: if ((__byte_perm_S (buf0[3], 0, 0x6542)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7054);
break;
- case 15: if ((__byte_perm (buf0[3], 0, 0x6543)) == p0) buf0[3] = __byte_perm (p1, buf0[3], 0x0654);
+ case 15: if ((__byte_perm_S (buf0[3], 0, 0x6543)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x0654);
break;
- case 16: if ((__byte_perm (buf1[0], 0, 0x6540)) == p0) buf1[0] = __byte_perm (p1, buf1[0], 0x7650);
+ case 16: if ((__byte_perm_S (buf1[0], 0, 0x6540)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7650);
break;
- case 17: if ((__byte_perm (buf1[0], 0, 0x6541)) == p0) buf1[0] = __byte_perm (p1, buf1[0], 0x7604);
+ case 17: if ((__byte_perm_S (buf1[0], 0, 0x6541)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7604);
break;
- case 18: if ((__byte_perm (buf1[0], 0, 0x6542)) == p0) buf1[0] = __byte_perm (p1, buf1[0], 0x7054);
+ case 18: if ((__byte_perm_S (buf1[0], 0, 0x6542)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7054);
break;
- case 19: if ((__byte_perm (buf1[0], 0, 0x6543)) == p0) buf1[0] = __byte_perm (p1, buf1[0], 0x0654);
+ case 19: if ((__byte_perm_S (buf1[0], 0, 0x6543)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x0654);
break;
- case 20: if ((__byte_perm (buf1[1], 0, 0x6540)) == p0) buf1[1] = __byte_perm (p1, buf1[1], 0x7650);
+ case 20: if ((__byte_perm_S (buf1[1], 0, 0x6540)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7650);
break;
- case 21: if ((__byte_perm (buf1[1], 0, 0x6541)) == p0) buf1[1] = __byte_perm (p1, buf1[1], 0x7604);
+ case 21: if ((__byte_perm_S (buf1[1], 0, 0x6541)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7604);
break;
- case 22: if ((__byte_perm (buf1[1], 0, 0x6542)) == p0) buf1[1] = __byte_perm (p1, buf1[1], 0x7054);
+ case 22: if ((__byte_perm_S (buf1[1], 0, 0x6542)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7054);
break;
- case 23: if ((__byte_perm (buf1[1], 0, 0x6543)) == p0) buf1[1] = __byte_perm (p1, buf1[1], 0x0654);
+ case 23: if ((__byte_perm_S (buf1[1], 0, 0x6543)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x0654);
break;
- case 24: if ((__byte_perm (buf1[2], 0, 0x6540)) == p0) buf1[2] = __byte_perm (p1, buf1[2], 0x7650);
+ case 24: if ((__byte_perm_S (buf1[2], 0, 0x6540)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7650);
break;
- case 25: if ((__byte_perm (buf1[2], 0, 0x6541)) == p0) buf1[2] = __byte_perm (p1, buf1[2], 0x7604);
+ case 25: if ((__byte_perm_S (buf1[2], 0, 0x6541)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7604);
break;
- case 26: if ((__byte_perm (buf1[2], 0, 0x6542)) == p0) buf1[2] = __byte_perm (p1, buf1[2], 0x7054);
+ case 26: if ((__byte_perm_S (buf1[2], 0, 0x6542)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7054);
break;
- case 27: if ((__byte_perm (buf1[2], 0, 0x6543)) == p0) buf1[2] = __byte_perm (p1, buf1[2], 0x0654);
+ case 27: if ((__byte_perm_S (buf1[2], 0, 0x6543)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x0654);
break;
- case 28: if ((__byte_perm (buf1[3], 0, 0x6540)) == p0) buf1[3] = __byte_perm (p1, buf1[3], 0x7650);
+ case 28: if ((__byte_perm_S (buf1[3], 0, 0x6540)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7650);
break;
- case 29: if ((__byte_perm (buf1[3], 0, 0x6541)) == p0) buf1[3] = __byte_perm (p1, buf1[3], 0x7604);
+ case 29: if ((__byte_perm_S (buf1[3], 0, 0x6541)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7604);
break;
- case 30: if ((__byte_perm (buf1[3], 0, 0x6542)) == p0) buf1[3] = __byte_perm (p1, buf1[3], 0x7054);
+ case 30: if ((__byte_perm_S (buf1[3], 0, 0x6542)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7054);
break;
- case 31: if ((__byte_perm (buf1[3], 0, 0x6543)) == p0) buf1[3] = __byte_perm (p1, buf1[3], 0x0654);
+ case 31: if ((__byte_perm_S (buf1[3], 0, 0x6543)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x0654);
break;
}
}
{
case 1: buf0[0] |= tmp;
break;
- case 2: buf0[0] |= __byte_perm (tmp, 0, 0x5400);
+ case 2: buf0[0] |= __byte_perm_S (tmp, 0, 0x5400);
break;
- case 3: buf0[0] |= __byte_perm (tmp, 0, 0x4000);
+ case 3: buf0[0] |= __byte_perm_S (tmp, 0, 0x4000);
break;
- case 4: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
+ case 4: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
break;
- case 5: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
+ case 5: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= tmp;
break;
- case 6: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x5400);
+ case 6: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x5400);
break;
- case 7: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x4000);
+ case 7: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x4000);
break;
- case 8: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
+ case 8: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
break;
- case 9: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
+ case 9: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= tmp;
break;
- case 10: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x5400);
+ case 10: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x5400);
break;
- case 11: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x4000);
+ case 11: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x4000);
break;
- case 12: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
+ case 12: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
break;
- case 13: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
+ case 13: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= tmp;
break;
- case 14: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x5400);
- break;
- case 15: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x4000);
- break;
- case 16: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- break;
- case 17: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
+ case 14: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x5400);
+ break;
+ case 15: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x4000);
+ break;
+ case 16: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ break;
+ case 17: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= tmp;
break;
- case 18: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x5400);
- break;
- case 19: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x4000);
- break;
- case 20: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- break;
- case 21: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
+ case 18: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x5400);
+ break;
+ case 19: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x4000);
+ break;
+ case 20: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ break;
+ case 21: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= tmp;
break;
- case 22: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x5400);
- break;
- case 23: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x4000);
- break;
- case 24: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- break;
- case 25: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
+ case 22: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x5400);
+ break;
+ case 23: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x4000);
+ break;
+ case 24: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ break;
+ case 25: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= tmp;
break;
- case 26: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x5400);
- break;
- case 27: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x4000);
- break;
- case 28: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x0000);
- break;
- case 29: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x0000);
+ case 26: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x5400);
+ break;
+ case 27: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x4000);
+ break;
+ case 28: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ break;
+ case 29: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[3] |= tmp;
break;
- case 30: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x0000);
- buf1[3] |= __byte_perm (tmp, 0, 0x5400);
- break;
- case 31: buf0[0] |= __byte_perm (tmp, 0, 0x0000);
- buf0[1] |= __byte_perm (tmp, 0, 0x0000);
- buf0[2] |= __byte_perm (tmp, 0, 0x0000);
- buf0[3] |= __byte_perm (tmp, 0, 0x0000);
- buf1[0] |= __byte_perm (tmp, 0, 0x0000);
- buf1[1] |= __byte_perm (tmp, 0, 0x0000);
- buf1[2] |= __byte_perm (tmp, 0, 0x0000);
- buf1[3] |= __byte_perm (tmp, 0, 0x4000);
+ case 30: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[3] |= __byte_perm_S (tmp, 0, 0x5400);
+ break;
+ case 31: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
+ buf1[3] |= __byte_perm_S (tmp, 0, 0x4000);
break;
}
#endif
u32 tib41[4];
#ifdef IS_NV
- tib40[0] = __byte_perm (buf0[0], 0, 0x1100);
- tib40[1] = __byte_perm (buf0[0], 0, 0x3322);
- tib40[2] = __byte_perm (buf0[1], 0, 0x1100);
- tib40[3] = __byte_perm (buf0[1], 0, 0x3322);
- tib41[0] = __byte_perm (buf0[2], 0, 0x1100);
- tib41[1] = __byte_perm (buf0[2], 0, 0x3322);
- tib41[2] = __byte_perm (buf0[3], 0, 0x1100);
- tib41[3] = __byte_perm (buf0[3], 0, 0x3322);
+ tib40[0] = __byte_perm_S (buf0[0], 0, 0x1100);
+ tib40[1] = __byte_perm_S (buf0[0], 0, 0x3322);
+ tib40[2] = __byte_perm_S (buf0[1], 0, 0x1100);
+ tib40[3] = __byte_perm_S (buf0[1], 0, 0x3322);
+ tib41[0] = __byte_perm_S (buf0[2], 0, 0x1100);
+ tib41[1] = __byte_perm_S (buf0[2], 0, 0x3322);
+ tib41[2] = __byte_perm_S (buf0[3], 0, 0x1100);
+ tib41[3] = __byte_perm_S (buf0[3], 0, 0x3322);
buf0[0] = tib40[0];
buf0[1] = tib40[1];
if (in_len < 2) return (in_len);
#ifdef IS_NV
- buf0[0] = __byte_perm (buf0[0], 0, 0x3201);
+ buf0[0] = __byte_perm_S (buf0[0], 0, 0x3201);
#endif
#if defined IS_AMD || defined IS_GENERIC
#ifdef IS_NV
switch (in_len)
{
- case 2: buf0[0] = __byte_perm (buf0[0], 0, 0x5401);
+ case 2: buf0[0] = __byte_perm_S (buf0[0], 0, 0x5401);
break;
- case 3: buf0[0] = __byte_perm (buf0[0], 0, 0x4120);
+ case 3: buf0[0] = __byte_perm_S (buf0[0], 0, 0x4120);
break;
- case 4: buf0[0] = __byte_perm (buf0[0], 0, 0x2310);
+ case 4: buf0[0] = __byte_perm_S (buf0[0], 0, 0x2310);
break;
- case 5: buf0[1] = __byte_perm (buf0[1], buf0[0], 0x7210);
- buf0[0] = __byte_perm (buf0[0], buf0[1], 0x4210);
- buf0[1] = __byte_perm (buf0[1], 0, 0x6543);
+ case 5: buf0[1] = __byte_perm_S (buf0[1], buf0[0], 0x7210);
+ buf0[0] = __byte_perm_S (buf0[0], buf0[1], 0x4210);
+ buf0[1] = __byte_perm_S (buf0[1], 0, 0x6543);
break;
- case 6: buf0[1] = __byte_perm (buf0[1], 0, 0x5401);
+ case 6: buf0[1] = __byte_perm_S (buf0[1], 0, 0x5401);
break;
- case 7: buf0[1] = __byte_perm (buf0[1], 0, 0x4120);
+ case 7: buf0[1] = __byte_perm_S (buf0[1], 0, 0x4120);
break;
- case 8: buf0[1] = __byte_perm (buf0[1], 0, 0x2310);
+ case 8: buf0[1] = __byte_perm_S (buf0[1], 0, 0x2310);
break;
- case 9: buf0[2] = __byte_perm (buf0[2], buf0[1], 0x7210);
- buf0[1] = __byte_perm (buf0[1], buf0[2], 0x4210);
- buf0[2] = __byte_perm (buf0[2], 0, 0x6543);
+ case 9: buf0[2] = __byte_perm_S (buf0[2], buf0[1], 0x7210);
+ buf0[1] = __byte_perm_S (buf0[1], buf0[2], 0x4210);
+ buf0[2] = __byte_perm_S (buf0[2], 0, 0x6543);
break;
- case 10: buf0[2] = __byte_perm (buf0[2], 0, 0x5401);
+ case 10: buf0[2] = __byte_perm_S (buf0[2], 0, 0x5401);
break;
- case 11: buf0[2] = __byte_perm (buf0[2], 0, 0x4120);
+ case 11: buf0[2] = __byte_perm_S (buf0[2], 0, 0x4120);
break;
- case 12: buf0[2] = __byte_perm (buf0[2], 0, 0x2310);
+ case 12: buf0[2] = __byte_perm_S (buf0[2], 0, 0x2310);
break;
- case 13: buf0[3] = __byte_perm (buf0[3], buf0[2], 0x7210);
- buf0[2] = __byte_perm (buf0[2], buf0[3], 0x4210);
- buf0[3] = __byte_perm (buf0[3], 0, 0x6543);
+ case 13: buf0[3] = __byte_perm_S (buf0[3], buf0[2], 0x7210);
+ buf0[2] = __byte_perm_S (buf0[2], buf0[3], 0x4210);
+ buf0[3] = __byte_perm_S (buf0[3], 0, 0x6543);
break;
- case 14: buf0[3] = __byte_perm (buf0[3], 0, 0x5401);
+ case 14: buf0[3] = __byte_perm_S (buf0[3], 0, 0x5401);
break;
- case 15: buf0[3] = __byte_perm (buf0[3], 0, 0x4120);
+ case 15: buf0[3] = __byte_perm_S (buf0[3], 0, 0x4120);
break;
- case 16: buf0[3] = __byte_perm (buf0[3], 0, 0x2310);
+ case 16: buf0[3] = __byte_perm_S (buf0[3], 0, 0x2310);
break;
- case 17: buf1[0] = __byte_perm (buf1[0], buf0[3], 0x7210);
- buf0[3] = __byte_perm (buf0[3], buf1[0], 0x4210);
- buf1[0] = __byte_perm (buf1[0], 0, 0x6543);
+ case 17: buf1[0] = __byte_perm_S (buf1[0], buf0[3], 0x7210);
+ buf0[3] = __byte_perm_S (buf0[3], buf1[0], 0x4210);
+ buf1[0] = __byte_perm_S (buf1[0], 0, 0x6543);
break;
- case 18: buf1[0] = __byte_perm (buf1[0], 0, 0x5401);
+ case 18: buf1[0] = __byte_perm_S (buf1[0], 0, 0x5401);
break;
- case 19: buf1[0] = __byte_perm (buf1[0], 0, 0x4120);
+ case 19: buf1[0] = __byte_perm_S (buf1[0], 0, 0x4120);
break;
- case 20: buf1[0] = __byte_perm (buf1[0], 0, 0x2310);
+ case 20: buf1[0] = __byte_perm_S (buf1[0], 0, 0x2310);
break;
- case 21: buf1[1] = __byte_perm (buf1[1], buf1[0], 0x7210);
- buf1[0] = __byte_perm (buf1[0], buf1[1], 0x4210);
- buf1[1] = __byte_perm (buf1[1], 0, 0x6543);
+ case 21: buf1[1] = __byte_perm_S (buf1[1], buf1[0], 0x7210);
+ buf1[0] = __byte_perm_S (buf1[0], buf1[1], 0x4210);
+ buf1[1] = __byte_perm_S (buf1[1], 0, 0x6543);
break;
- case 22: buf1[1] = __byte_perm (buf1[1], 0, 0x5401);
+ case 22: buf1[1] = __byte_perm_S (buf1[1], 0, 0x5401);
break;
- case 23: buf1[1] = __byte_perm (buf1[1], 0, 0x4120);
+ case 23: buf1[1] = __byte_perm_S (buf1[1], 0, 0x4120);
break;
- case 24: buf1[1] = __byte_perm (buf1[1], 0, 0x2310);
+ case 24: buf1[1] = __byte_perm_S (buf1[1], 0, 0x2310);
break;
- case 25: buf1[2] = __byte_perm (buf1[2], buf1[1], 0x7210);
- buf1[1] = __byte_perm (buf1[1], buf1[2], 0x4210);
- buf1[2] = __byte_perm (buf1[2], 0, 0x6543);
+ case 25: buf1[2] = __byte_perm_S (buf1[2], buf1[1], 0x7210);
+ buf1[1] = __byte_perm_S (buf1[1], buf1[2], 0x4210);
+ buf1[2] = __byte_perm_S (buf1[2], 0, 0x6543);
break;
- case 26: buf1[2] = __byte_perm (buf1[2], 0, 0x5401);
+ case 26: buf1[2] = __byte_perm_S (buf1[2], 0, 0x5401);
break;
- case 27: buf1[2] = __byte_perm (buf1[2], 0, 0x4120);
+ case 27: buf1[2] = __byte_perm_S (buf1[2], 0, 0x4120);
break;
- case 28: buf1[2] = __byte_perm (buf1[2], 0, 0x2310);
+ case 28: buf1[2] = __byte_perm_S (buf1[2], 0, 0x2310);
break;
- case 29: buf1[3] = __byte_perm (buf1[3], buf1[2], 0x7210);
- buf1[2] = __byte_perm (buf1[2], buf1[3], 0x4210);
- buf1[3] = __byte_perm (buf1[3], 0, 0x6543);
+ case 29: buf1[3] = __byte_perm_S (buf1[3], buf1[2], 0x7210);
+ buf1[2] = __byte_perm_S (buf1[2], buf1[3], 0x4210);
+ buf1[3] = __byte_perm_S (buf1[3], 0, 0x6543);
break;
- case 30: buf1[3] = __byte_perm (buf1[3], 0, 0x5401);
+ case 30: buf1[3] = __byte_perm_S (buf1[3], 0, 0x5401);
break;
- case 31: buf1[3] = __byte_perm (buf1[3], 0, 0x4120);
+ case 31: buf1[3] = __byte_perm_S (buf1[3], 0, 0x4120);
break;
}
#endif
#ifdef IS_NV
switch (p0)
{
- case 0: tmp0 = __byte_perm (buf0[0], 0, 0x6540);
+ case 0: tmp0 = __byte_perm_S (buf0[0], 0, 0x6540);
break;
- case 1: tmp0 = __byte_perm (buf0[0], 0, 0x6541);
+ case 1: tmp0 = __byte_perm_S (buf0[0], 0, 0x6541);
break;
- case 2: tmp0 = __byte_perm (buf0[0], 0, 0x6542);
+ case 2: tmp0 = __byte_perm_S (buf0[0], 0, 0x6542);
break;
- case 3: tmp0 = __byte_perm (buf0[0], 0, 0x6543);
+ case 3: tmp0 = __byte_perm_S (buf0[0], 0, 0x6543);
break;
- case 4: tmp0 = __byte_perm (buf0[1], 0, 0x6540);
+ case 4: tmp0 = __byte_perm_S (buf0[1], 0, 0x6540);
break;
- case 5: tmp0 = __byte_perm (buf0[1], 0, 0x6541);
+ case 5: tmp0 = __byte_perm_S (buf0[1], 0, 0x6541);
break;
- case 6: tmp0 = __byte_perm (buf0[1], 0, 0x6542);
+ case 6: tmp0 = __byte_perm_S (buf0[1], 0, 0x6542);
break;
- case 7: tmp0 = __byte_perm (buf0[1], 0, 0x6543);
+ case 7: tmp0 = __byte_perm_S (buf0[1], 0, 0x6543);
break;
- case 8: tmp0 = __byte_perm (buf0[2], 0, 0x6540);
+ case 8: tmp0 = __byte_perm_S (buf0[2], 0, 0x6540);
break;
- case 9: tmp0 = __byte_perm (buf0[2], 0, 0x6541);
+ case 9: tmp0 = __byte_perm_S (buf0[2], 0, 0x6541);
break;
- case 10: tmp0 = __byte_perm (buf0[2], 0, 0x6542);
+ case 10: tmp0 = __byte_perm_S (buf0[2], 0, 0x6542);
break;
- case 11: tmp0 = __byte_perm (buf0[2], 0, 0x6543);
+ case 11: tmp0 = __byte_perm_S (buf0[2], 0, 0x6543);
break;
- case 12: tmp0 = __byte_perm (buf0[3], 0, 0x6540);
+ case 12: tmp0 = __byte_perm_S (buf0[3], 0, 0x6540);
break;
- case 13: tmp0 = __byte_perm (buf0[3], 0, 0x6541);
+ case 13: tmp0 = __byte_perm_S (buf0[3], 0, 0x6541);
break;
- case 14: tmp0 = __byte_perm (buf0[3], 0, 0x6542);
+ case 14: tmp0 = __byte_perm_S (buf0[3], 0, 0x6542);
break;
- case 15: tmp0 = __byte_perm (buf0[3], 0, 0x6543);
+ case 15: tmp0 = __byte_perm_S (buf0[3], 0, 0x6543);
break;
- case 16: tmp0 = __byte_perm (buf1[0], 0, 0x6540);
+ case 16: tmp0 = __byte_perm_S (buf1[0], 0, 0x6540);
break;
- case 17: tmp0 = __byte_perm (buf1[0], 0, 0x6541);
+ case 17: tmp0 = __byte_perm_S (buf1[0], 0, 0x6541);
break;
- case 18: tmp0 = __byte_perm (buf1[0], 0, 0x6542);
+ case 18: tmp0 = __byte_perm_S (buf1[0], 0, 0x6542);
break;
- case 19: tmp0 = __byte_perm (buf1[0], 0, 0x6543);
+ case 19: tmp0 = __byte_perm_S (buf1[0], 0, 0x6543);
break;
- case 20: tmp0 = __byte_perm (buf1[1], 0, 0x6540);
+ case 20: tmp0 = __byte_perm_S (buf1[1], 0, 0x6540);
break;
- case 21: tmp0 = __byte_perm (buf1[1], 0, 0x6541);
+ case 21: tmp0 = __byte_perm_S (buf1[1], 0, 0x6541);
break;
- case 22: tmp0 = __byte_perm (buf1[1], 0, 0x6542);
+ case 22: tmp0 = __byte_perm_S (buf1[1], 0, 0x6542);
break;
- case 23: tmp0 = __byte_perm (buf1[1], 0, 0x6543);
+ case 23: tmp0 = __byte_perm_S (buf1[1], 0, 0x6543);
break;
- case 24: tmp0 = __byte_perm (buf1[2], 0, 0x6540);
+ case 24: tmp0 = __byte_perm_S (buf1[2], 0, 0x6540);
break;
- case 25: tmp0 = __byte_perm (buf1[2], 0, 0x6541);
+ case 25: tmp0 = __byte_perm_S (buf1[2], 0, 0x6541);
break;
- case 26: tmp0 = __byte_perm (buf1[2], 0, 0x6542);
+ case 26: tmp0 = __byte_perm_S (buf1[2], 0, 0x6542);
break;
- case 27: tmp0 = __byte_perm (buf1[2], 0, 0x6543);
+ case 27: tmp0 = __byte_perm_S (buf1[2], 0, 0x6543);
break;
- case 28: tmp0 = __byte_perm (buf1[3], 0, 0x6540);
+ case 28: tmp0 = __byte_perm_S (buf1[3], 0, 0x6540);
break;
- case 29: tmp0 = __byte_perm (buf1[3], 0, 0x6541);
+ case 29: tmp0 = __byte_perm_S (buf1[3], 0, 0x6541);
break;
- case 30: tmp0 = __byte_perm (buf1[3], 0, 0x6542);
+ case 30: tmp0 = __byte_perm_S (buf1[3], 0, 0x6542);
break;
- case 31: tmp0 = __byte_perm (buf1[3], 0, 0x6543);
+ case 31: tmp0 = __byte_perm_S (buf1[3], 0, 0x6543);
break;
}
switch (p1)
{
- case 0: tmp1 = __byte_perm (buf0[0], 0, 0x6540);
- buf0[0] = __byte_perm (tmp0, buf0[0], 0x7650);
+ case 0: tmp1 = __byte_perm_S (buf0[0], 0, 0x6540);
+ buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7650);
break;
- case 1: tmp1 = __byte_perm (buf0[0], 0, 0x6541);
- buf0[0] = __byte_perm (tmp0, buf0[0], 0x7604);
+ case 1: tmp1 = __byte_perm_S (buf0[0], 0, 0x6541);
+ buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7604);
break;
- case 2: tmp1 = __byte_perm (buf0[0], 0, 0x6542);
- buf0[0] = __byte_perm (tmp0, buf0[0], 0x7054);
+ case 2: tmp1 = __byte_perm_S (buf0[0], 0, 0x6542);
+ buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7054);
break;
- case 3: tmp1 = __byte_perm (buf0[0], 0, 0x6543);
- buf0[0] = __byte_perm (tmp0, buf0[0], 0x0654);
+ case 3: tmp1 = __byte_perm_S (buf0[0], 0, 0x6543);
+ buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x0654);
break;
- case 4: tmp1 = __byte_perm (buf0[1], 0, 0x6540);
- buf0[1] = __byte_perm (tmp0, buf0[1], 0x7650);
+ case 4: tmp1 = __byte_perm_S (buf0[1], 0, 0x6540);
+ buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7650);
break;
- case 5: tmp1 = __byte_perm (buf0[1], 0, 0x6541);
- buf0[1] = __byte_perm (tmp0, buf0[1], 0x7604);
+ case 5: tmp1 = __byte_perm_S (buf0[1], 0, 0x6541);
+ buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7604);
break;
- case 6: tmp1 = __byte_perm (buf0[1], 0, 0x6542);
- buf0[1] = __byte_perm (tmp0, buf0[1], 0x7054);
+ case 6: tmp1 = __byte_perm_S (buf0[1], 0, 0x6542);
+ buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7054);
break;
- case 7: tmp1 = __byte_perm (buf0[1], 0, 0x6543);
- buf0[1] = __byte_perm (tmp0, buf0[1], 0x0654);
+ case 7: tmp1 = __byte_perm_S (buf0[1], 0, 0x6543);
+ buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x0654);
break;
- case 8: tmp1 = __byte_perm (buf0[2], 0, 0x6540);
- buf0[2] = __byte_perm (tmp0, buf0[2], 0x7650);
+ case 8: tmp1 = __byte_perm_S (buf0[2], 0, 0x6540);
+ buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7650);
break;
- case 9: tmp1 = __byte_perm (buf0[2], 0, 0x6541);
- buf0[2] = __byte_perm (tmp0, buf0[2], 0x7604);
+ case 9: tmp1 = __byte_perm_S (buf0[2], 0, 0x6541);
+ buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7604);
break;
- case 10: tmp1 = __byte_perm (buf0[2], 0, 0x6542);
- buf0[2] = __byte_perm (tmp0, buf0[2], 0x7054);
+ case 10: tmp1 = __byte_perm_S (buf0[2], 0, 0x6542);
+ buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7054);
break;
- case 11: tmp1 = __byte_perm (buf0[2], 0, 0x6543);
- buf0[2] = __byte_perm (tmp0, buf0[2], 0x0654);
+ case 11: tmp1 = __byte_perm_S (buf0[2], 0, 0x6543);
+ buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x0654);
break;
- case 12: tmp1 = __byte_perm (buf0[3], 0, 0x6540);
- buf0[3] = __byte_perm (tmp0, buf0[3], 0x7650);
+ case 12: tmp1 = __byte_perm_S (buf0[3], 0, 0x6540);
+ buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7650);
break;
- case 13: tmp1 = __byte_perm (buf0[3], 0, 0x6541);
- buf0[3] = __byte_perm (tmp0, buf0[3], 0x7604);
+ case 13: tmp1 = __byte_perm_S (buf0[3], 0, 0x6541);
+ buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7604);
break;
- case 14: tmp1 = __byte_perm (buf0[3], 0, 0x6542);
- buf0[3] = __byte_perm (tmp0, buf0[3], 0x7054);
+ case 14: tmp1 = __byte_perm_S (buf0[3], 0, 0x6542);
+ buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7054);
break;
- case 15: tmp1 = __byte_perm (buf0[3], 0, 0x6543);
- buf0[3] = __byte_perm (tmp0, buf0[3], 0x0654);
+ case 15: tmp1 = __byte_perm_S (buf0[3], 0, 0x6543);
+ buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x0654);
break;
- case 16: tmp1 = __byte_perm (buf1[0], 0, 0x6540);
- buf1[0] = __byte_perm (tmp0, buf1[0], 0x7650);
+ case 16: tmp1 = __byte_perm_S (buf1[0], 0, 0x6540);
+ buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7650);
break;
- case 17: tmp1 = __byte_perm (buf1[0], 0, 0x6541);
- buf1[0] = __byte_perm (tmp0, buf1[0], 0x7604);
+ case 17: tmp1 = __byte_perm_S (buf1[0], 0, 0x6541);
+ buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7604);
break;
- case 18: tmp1 = __byte_perm (buf1[0], 0, 0x6542);
- buf1[0] = __byte_perm (tmp0, buf1[0], 0x7054);
+ case 18: tmp1 = __byte_perm_S (buf1[0], 0, 0x6542);
+ buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7054);
break;
- case 19: tmp1 = __byte_perm (buf1[0], 0, 0x6543);
- buf1[0] = __byte_perm (tmp0, buf1[0], 0x0654);
+ case 19: tmp1 = __byte_perm_S (buf1[0], 0, 0x6543);
+ buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x0654);
break;
- case 20: tmp1 = __byte_perm (buf1[1], 0, 0x6540);
- buf1[1] = __byte_perm (tmp0, buf1[1], 0x7650);
+ case 20: tmp1 = __byte_perm_S (buf1[1], 0, 0x6540);
+ buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7650);
break;
- case 21: tmp1 = __byte_perm (buf1[1], 0, 0x6541);
- buf1[1] = __byte_perm (tmp0, buf1[1], 0x7604);
+ case 21: tmp1 = __byte_perm_S (buf1[1], 0, 0x6541);
+ buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7604);
break;
- case 22: tmp1 = __byte_perm (buf1[1], 0, 0x6542);
- buf1[1] = __byte_perm (tmp0, buf1[1], 0x7054);
+ case 22: tmp1 = __byte_perm_S (buf1[1], 0, 0x6542);
+ buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7054);
break;
- case 23: tmp1 = __byte_perm (buf1[1], 0, 0x6543);
- buf1[1] = __byte_perm (tmp0, buf1[1], 0x0654);
+ case 23: tmp1 = __byte_perm_S (buf1[1], 0, 0x6543);
+ buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x0654);
break;
- case 24: tmp1 = __byte_perm (buf1[2], 0, 0x6540);
- buf1[2] = __byte_perm (tmp0, buf1[2], 0x7650);
+ case 24: tmp1 = __byte_perm_S (buf1[2], 0, 0x6540);
+ buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7650);
break;
- case 25: tmp1 = __byte_perm (buf1[2], 0, 0x6541);
- buf1[2] = __byte_perm (tmp0, buf1[2], 0x7604);
+ case 25: tmp1 = __byte_perm_S (buf1[2], 0, 0x6541);
+ buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7604);
break;
- case 26: tmp1 = __byte_perm (buf1[2], 0, 0x6542);
- buf1[2] = __byte_perm (tmp0, buf1[2], 0x7054);
+ case 26: tmp1 = __byte_perm_S (buf1[2], 0, 0x6542);
+ buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7054);
break;
- case 27: tmp1 = __byte_perm (buf1[2], 0, 0x6543);
- buf1[2] = __byte_perm (tmp0, buf1[2], 0x0654);
+ case 27: tmp1 = __byte_perm_S (buf1[2], 0, 0x6543);
+ buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x0654);
break;
- case 28: tmp1 = __byte_perm (buf1[3], 0, 0x6540);
- buf1[3] = __byte_perm (tmp0, buf1[3], 0x7650);
+ case 28: tmp1 = __byte_perm_S (buf1[3], 0, 0x6540);
+ buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7650);
break;
- case 29: tmp1 = __byte_perm (buf1[3], 0, 0x6541);
- buf1[3] = __byte_perm (tmp0, buf1[3], 0x7604);
+ case 29: tmp1 = __byte_perm_S (buf1[3], 0, 0x6541);
+ buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7604);
break;
- case 30: tmp1 = __byte_perm (buf1[3], 0, 0x6542);
- buf1[3] = __byte_perm (tmp0, buf1[3], 0x7054);
+ case 30: tmp1 = __byte_perm_S (buf1[3], 0, 0x6542);
+ buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7054);
break;
- case 31: tmp1 = __byte_perm (buf1[3], 0, 0x6543);
- buf1[3] = __byte_perm (tmp0, buf1[3], 0x0654);
+ case 31: tmp1 = __byte_perm_S (buf1[3], 0, 0x6543);
+ buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x0654);
break;
}
switch (p0)
{
- case 0: buf0[0] = __byte_perm (tmp1, buf0[0], 0x7650);
+ case 0: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7650);
break;
- case 1: buf0[0] = __byte_perm (tmp1, buf0[0], 0x7604);
+ case 1: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7604);
break;
- case 2: buf0[0] = __byte_perm (tmp1, buf0[0], 0x7054);
+ case 2: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7054);
break;
- case 3: buf0[0] = __byte_perm (tmp1, buf0[0], 0x0654);
+ case 3: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x0654);
break;
- case 4: buf0[1] = __byte_perm (tmp1, buf0[1], 0x7650);
+ case 4: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7650);
break;
- case 5: buf0[1] = __byte_perm (tmp1, buf0[1], 0x7604);
+ case 5: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7604);
break;
- case 6: buf0[1] = __byte_perm (tmp1, buf0[1], 0x7054);
+ case 6: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7054);
break;
- case 7: buf0[1] = __byte_perm (tmp1, buf0[1], 0x0654);
+ case 7: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x0654);
break;
- case 8: buf0[2] = __byte_perm (tmp1, buf0[2], 0x7650);
+ case 8: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7650);
break;
- case 9: buf0[2] = __byte_perm (tmp1, buf0[2], 0x7604);
+ case 9: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7604);
break;
- case 10: buf0[2] = __byte_perm (tmp1, buf0[2], 0x7054);
+ case 10: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7054);
break;
- case 11: buf0[2] = __byte_perm (tmp1, buf0[2], 0x0654);
+ case 11: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x0654);
break;
- case 12: buf0[3] = __byte_perm (tmp1, buf0[3], 0x7650);
+ case 12: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7650);
break;
- case 13: buf0[3] = __byte_perm (tmp1, buf0[3], 0x7604);
+ case 13: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7604);
break;
- case 14: buf0[3] = __byte_perm (tmp1, buf0[3], 0x7054);
+ case 14: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7054);
break;
- case 15: buf0[3] = __byte_perm (tmp1, buf0[3], 0x0654);
+ case 15: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x0654);
break;
- case 16: buf1[0] = __byte_perm (tmp1, buf1[0], 0x7650);
+ case 16: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7650);
break;
- case 17: buf1[0] = __byte_perm (tmp1, buf1[0], 0x7604);
+ case 17: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7604);
break;
- case 18: buf1[0] = __byte_perm (tmp1, buf1[0], 0x7054);
+ case 18: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7054);
break;
- case 19: buf1[0] = __byte_perm (tmp1, buf1[0], 0x0654);
+ case 19: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x0654);
break;
- case 20: buf1[1] = __byte_perm (tmp1, buf1[1], 0x7650);
+ case 20: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7650);
break;
- case 21: buf1[1] = __byte_perm (tmp1, buf1[1], 0x7604);
+ case 21: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7604);
break;
- case 22: buf1[1] = __byte_perm (tmp1, buf1[1], 0x7054);
+ case 22: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7054);
break;
- case 23: buf1[1] = __byte_perm (tmp1, buf1[1], 0x0654);
+ case 23: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x0654);
break;
- case 24: buf1[2] = __byte_perm (tmp1, buf1[2], 0x7650);
+ case 24: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7650);
break;
- case 25: buf1[2] = __byte_perm (tmp1, buf1[2], 0x7604);
+ case 25: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7604);
break;
- case 26: buf1[2] = __byte_perm (tmp1, buf1[2], 0x7054);
+ case 26: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7054);
break;
- case 27: buf1[2] = __byte_perm (tmp1, buf1[2], 0x0654);
+ case 27: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x0654);
break;
- case 28: buf1[3] = __byte_perm (tmp1, buf1[3], 0x7650);
+ case 28: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7650);
break;
- case 29: buf1[3] = __byte_perm (tmp1, buf1[3], 0x7604);
+ case 29: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7604);
break;
- case 30: buf1[3] = __byte_perm (tmp1, buf1[3], 0x7054);
+ case 30: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7054);
break;
- case 31: buf1[3] = __byte_perm (tmp1, buf1[3], 0x0654);
+ case 31: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x0654);
break;
}
#endif
switch (i)
{
- case 0: tmp0 = __byte_perm (buf0[0], 0, 0x6540);
+ case 0: tmp0 = __byte_perm_S (buf0[0], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf0[0])); break;
- case 1: tmp0 = __byte_perm (buf0[0], 0, 0x6541);
+ case 1: tmp0 = __byte_perm_S (buf0[0], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf0[0])); break;
- case 2: tmp0 = __byte_perm (buf0[0], 0, 0x6542);
+ case 2: tmp0 = __byte_perm_S (buf0[0], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf0[0])); break;
- case 3: tmp0 = __byte_perm (buf0[0], 0, 0x6543);
+ case 3: tmp0 = __byte_perm_S (buf0[0], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf0[1])); break;
- case 4: tmp0 = __byte_perm (buf0[1], 0, 0x6540);
+ case 4: tmp0 = __byte_perm_S (buf0[1], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf0[1])); break;
- case 5: tmp0 = __byte_perm (buf0[1], 0, 0x6541);
+ case 5: tmp0 = __byte_perm_S (buf0[1], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf0[1])); break;
- case 6: tmp0 = __byte_perm (buf0[1], 0, 0x6542);
+ case 6: tmp0 = __byte_perm_S (buf0[1], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf0[1])); break;
- case 7: tmp0 = __byte_perm (buf0[1], 0, 0x6543);
+ case 7: tmp0 = __byte_perm_S (buf0[1], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf0[2])); break;
- case 8: tmp0 = __byte_perm (buf0[2], 0, 0x6540);
+ case 8: tmp0 = __byte_perm_S (buf0[2], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf0[2])); break;
- case 9: tmp0 = __byte_perm (buf0[2], 0, 0x6541);
+ case 9: tmp0 = __byte_perm_S (buf0[2], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf0[2])); break;
- case 10: tmp0 = __byte_perm (buf0[2], 0, 0x6542);
+ case 10: tmp0 = __byte_perm_S (buf0[2], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf0[2])); break;
- case 11: tmp0 = __byte_perm (buf0[2], 0, 0x6543);
+ case 11: tmp0 = __byte_perm_S (buf0[2], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf0[3])); break;
- case 12: tmp0 = __byte_perm (buf0[3], 0, 0x6540);
+ case 12: tmp0 = __byte_perm_S (buf0[3], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf0[3])); break;
- case 13: tmp0 = __byte_perm (buf0[3], 0, 0x6541);
+ case 13: tmp0 = __byte_perm_S (buf0[3], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf0[3])); break;
- case 14: tmp0 = __byte_perm (buf0[3], 0, 0x6542);
+ case 14: tmp0 = __byte_perm_S (buf0[3], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf0[3])); break;
- case 15: tmp0 = __byte_perm (buf0[3], 0, 0x6543);
+ case 15: tmp0 = __byte_perm_S (buf0[3], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf1[0])); break;
- case 16: tmp0 = __byte_perm (buf1[0], 0, 0x6540);
+ case 16: tmp0 = __byte_perm_S (buf1[0], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf1[0])); break;
- case 17: tmp0 = __byte_perm (buf1[0], 0, 0x6541);
+ case 17: tmp0 = __byte_perm_S (buf1[0], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf1[0])); break;
- case 18: tmp0 = __byte_perm (buf1[0], 0, 0x6542);
+ case 18: tmp0 = __byte_perm_S (buf1[0], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf1[0])); break;
- case 19: tmp0 = __byte_perm (buf1[0], 0, 0x6543);
+ case 19: tmp0 = __byte_perm_S (buf1[0], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf1[1])); break;
- case 20: tmp0 = __byte_perm (buf1[1], 0, 0x6540);
+ case 20: tmp0 = __byte_perm_S (buf1[1], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf1[1])); break;
- case 21: tmp0 = __byte_perm (buf1[1], 0, 0x6541);
+ case 21: tmp0 = __byte_perm_S (buf1[1], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf1[1])); break;
- case 22: tmp0 = __byte_perm (buf1[1], 0, 0x6542);
+ case 22: tmp0 = __byte_perm_S (buf1[1], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf1[1])); break;
- case 23: tmp0 = __byte_perm (buf1[1], 0, 0x6543);
+ case 23: tmp0 = __byte_perm_S (buf1[1], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf1[2])); break;
- case 24: tmp0 = __byte_perm (buf1[2], 0, 0x6540);
+ case 24: tmp0 = __byte_perm_S (buf1[2], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf1[2])); break;
- case 25: tmp0 = __byte_perm (buf1[2], 0, 0x6541);
+ case 25: tmp0 = __byte_perm_S (buf1[2], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf1[2])); break;
- case 26: tmp0 = __byte_perm (buf1[2], 0, 0x6542);
+ case 26: tmp0 = __byte_perm_S (buf1[2], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf1[2])); break;
- case 27: tmp0 = __byte_perm (buf1[2], 0, 0x6543);
+ case 27: tmp0 = __byte_perm_S (buf1[2], 0, 0x6543);
tmp1 = ~(0x00000020 & generate_cmask (buf1[3])); break;
- case 28: tmp0 = __byte_perm (buf1[3], 0, 0x6540);
+ case 28: tmp0 = __byte_perm_S (buf1[3], 0, 0x6540);
tmp1 = ~(0x00002000 & generate_cmask (buf1[3])); break;
- case 29: tmp0 = __byte_perm (buf1[3], 0, 0x6541);
+ case 29: tmp0 = __byte_perm_S (buf1[3], 0, 0x6541);
tmp1 = ~(0x00200000 & generate_cmask (buf1[3])); break;
- case 30: tmp0 = __byte_perm (buf1[3], 0, 0x6542);
+ case 30: tmp0 = __byte_perm_S (buf1[3], 0, 0x6542);
tmp1 = ~(0x20000000 & generate_cmask (buf1[3])); break;
}