GOSTCoin CUDA miner project, compatible with most nvidia cards, containing only gostd algo
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

1426 lines
35 KiB

/* $Id: luffa.c 219 2010-06-08 17:24:41Z tp $ */
/*
* Luffa implementation.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#include <stddef.h>
#include <string.h>
#include <limits.h>
#include "sph_luffa.h"
#ifdef __cplusplus
extern "C"{
#endif
#if SPH_64_TRUE && !defined SPH_LUFFA_PARALLEL
#define SPH_LUFFA_PARALLEL 1
#endif
#ifdef _MSC_VER
#pragma warning (disable: 4146)
#endif
static const sph_u32 V_INIT[5][8] = {
{
SPH_C32(0x6d251e69), SPH_C32(0x44b051e0),
SPH_C32(0x4eaa6fb4), SPH_C32(0xdbf78465),
SPH_C32(0x6e292011), SPH_C32(0x90152df4),
SPH_C32(0xee058139), SPH_C32(0xdef610bb)
}, {
SPH_C32(0xc3b44b95), SPH_C32(0xd9d2f256),
SPH_C32(0x70eee9a0), SPH_C32(0xde099fa3),
SPH_C32(0x5d9b0557), SPH_C32(0x8fc944b3),
SPH_C32(0xcf1ccf0e), SPH_C32(0x746cd581)
}, {
SPH_C32(0xf7efc89d), SPH_C32(0x5dba5781),
SPH_C32(0x04016ce5), SPH_C32(0xad659c05),
SPH_C32(0x0306194f), SPH_C32(0x666d1836),
SPH_C32(0x24aa230a), SPH_C32(0x8b264ae7)
}, {
SPH_C32(0x858075d5), SPH_C32(0x36d79cce),
SPH_C32(0xe571f7d7), SPH_C32(0x204b1f67),
SPH_C32(0x35870c6a), SPH_C32(0x57e9e923),
SPH_C32(0x14bcb808), SPH_C32(0x7cde72ce)
}, {
SPH_C32(0x6c68e9be), SPH_C32(0x5ec41e22),
SPH_C32(0xc825b7c7), SPH_C32(0xaffb4363),
SPH_C32(0xf5df3999), SPH_C32(0x0fc688f1),
SPH_C32(0xb07224cc), SPH_C32(0x03e86cea)
}
};
static const sph_u32 RC00[8] = {
SPH_C32(0x303994a6), SPH_C32(0xc0e65299),
SPH_C32(0x6cc33a12), SPH_C32(0xdc56983e),
SPH_C32(0x1e00108f), SPH_C32(0x7800423d),
SPH_C32(0x8f5b7882), SPH_C32(0x96e1db12)
};
static const sph_u32 RC04[8] = {
SPH_C32(0xe0337818), SPH_C32(0x441ba90d),
SPH_C32(0x7f34d442), SPH_C32(0x9389217f),
SPH_C32(0xe5a8bce6), SPH_C32(0x5274baf4),
SPH_C32(0x26889ba7), SPH_C32(0x9a226e9d)
};
static const sph_u32 RC10[8] = {
SPH_C32(0xb6de10ed), SPH_C32(0x70f47aae),
SPH_C32(0x0707a3d4), SPH_C32(0x1c1e8f51),
SPH_C32(0x707a3d45), SPH_C32(0xaeb28562),
SPH_C32(0xbaca1589), SPH_C32(0x40a46f3e)
};
static const sph_u32 RC14[8] = {
SPH_C32(0x01685f3d), SPH_C32(0x05a17cf4),
SPH_C32(0xbd09caca), SPH_C32(0xf4272b28),
SPH_C32(0x144ae5cc), SPH_C32(0xfaa7ae2b),
SPH_C32(0x2e48f1c1), SPH_C32(0xb923c704)
};
#if SPH_LUFFA_PARALLEL
static const sph_u64 RCW010[8] = {
SPH_C64(0xb6de10ed303994a6), SPH_C64(0x70f47aaec0e65299),
SPH_C64(0x0707a3d46cc33a12), SPH_C64(0x1c1e8f51dc56983e),
SPH_C64(0x707a3d451e00108f), SPH_C64(0xaeb285627800423d),
SPH_C64(0xbaca15898f5b7882), SPH_C64(0x40a46f3e96e1db12)
};
static const sph_u64 RCW014[8] = {
SPH_C64(0x01685f3de0337818), SPH_C64(0x05a17cf4441ba90d),
SPH_C64(0xbd09caca7f34d442), SPH_C64(0xf4272b289389217f),
SPH_C64(0x144ae5cce5a8bce6), SPH_C64(0xfaa7ae2b5274baf4),
SPH_C64(0x2e48f1c126889ba7), SPH_C64(0xb923c7049a226e9d)
};
#endif
static const sph_u32 RC20[8] = {
SPH_C32(0xfc20d9d2), SPH_C32(0x34552e25),
SPH_C32(0x7ad8818f), SPH_C32(0x8438764a),
SPH_C32(0xbb6de032), SPH_C32(0xedb780c8),
SPH_C32(0xd9847356), SPH_C32(0xa2c78434)
};
static const sph_u32 RC24[8] = {
SPH_C32(0xe25e72c1), SPH_C32(0xe623bb72),
SPH_C32(0x5c58a4a4), SPH_C32(0x1e38e2e7),
SPH_C32(0x78e38b9d), SPH_C32(0x27586719),
SPH_C32(0x36eda57f), SPH_C32(0x703aace7)
};
static const sph_u32 RC30[8] = {
SPH_C32(0xb213afa5), SPH_C32(0xc84ebe95),
SPH_C32(0x4e608a22), SPH_C32(0x56d858fe),
SPH_C32(0x343b138f), SPH_C32(0xd0ec4e3d),
SPH_C32(0x2ceb4882), SPH_C32(0xb3ad2208)
};
static const sph_u32 RC34[8] = {
SPH_C32(0xe028c9bf), SPH_C32(0x44756f91),
SPH_C32(0x7e8fce32), SPH_C32(0x956548be),
SPH_C32(0xfe191be2), SPH_C32(0x3cb226e5),
SPH_C32(0x5944a28e), SPH_C32(0xa1c4c355)
};
#if SPH_LUFFA_PARALLEL
static const sph_u64 RCW230[8] = {
SPH_C64(0xb213afa5fc20d9d2), SPH_C64(0xc84ebe9534552e25),
SPH_C64(0x4e608a227ad8818f), SPH_C64(0x56d858fe8438764a),
SPH_C64(0x343b138fbb6de032), SPH_C64(0xd0ec4e3dedb780c8),
SPH_C64(0x2ceb4882d9847356), SPH_C64(0xb3ad2208a2c78434)
};
static const sph_u64 RCW234[8] = {
SPH_C64(0xe028c9bfe25e72c1), SPH_C64(0x44756f91e623bb72),
SPH_C64(0x7e8fce325c58a4a4), SPH_C64(0x956548be1e38e2e7),
SPH_C64(0xfe191be278e38b9d), SPH_C64(0x3cb226e527586719),
SPH_C64(0x5944a28e36eda57f), SPH_C64(0xa1c4c355703aace7)
};
#endif
static const sph_u32 RC40[8] = {
SPH_C32(0xf0d2e9e3), SPH_C32(0xac11d7fa),
SPH_C32(0x1bcb66f2), SPH_C32(0x6f2d9bc9),
SPH_C32(0x78602649), SPH_C32(0x8edae952),
SPH_C32(0x3b6ba548), SPH_C32(0xedae9520)
};
static const sph_u32 RC44[8] = {
SPH_C32(0x5090d577), SPH_C32(0x2d1925ab),
SPH_C32(0xb46496ac), SPH_C32(0xd1925ab0),
SPH_C32(0x29131ab6), SPH_C32(0x0fc053c3),
SPH_C32(0x3f014f0c), SPH_C32(0xfc053c31)
};
#define DECL_TMP8(w) \
sph_u32 w ## 0, w ## 1, w ## 2, w ## 3, w ## 4, w ## 5, w ## 6, w ## 7;
#define M2(d, s) do { \
sph_u32 tmp = s ## 7; \
d ## 7 = s ## 6; \
d ## 6 = s ## 5; \
d ## 5 = s ## 4; \
d ## 4 = s ## 3 ^ tmp; \
d ## 3 = s ## 2 ^ tmp; \
d ## 2 = s ## 1; \
d ## 1 = s ## 0 ^ tmp; \
d ## 0 = tmp; \
} while (0)
#define XOR(d, s1, s2) do { \
d ## 0 = s1 ## 0 ^ s2 ## 0; \
d ## 1 = s1 ## 1 ^ s2 ## 1; \
d ## 2 = s1 ## 2 ^ s2 ## 2; \
d ## 3 = s1 ## 3 ^ s2 ## 3; \
d ## 4 = s1 ## 4 ^ s2 ## 4; \
d ## 5 = s1 ## 5 ^ s2 ## 5; \
d ## 6 = s1 ## 6 ^ s2 ## 6; \
d ## 7 = s1 ## 7 ^ s2 ## 7; \
} while (0)
#if SPH_LUFFA_PARALLEL
#define SUB_CRUMB_GEN(a0, a1, a2, a3, width) do { \
sph_u ## width tmp; \
tmp = (a0); \
(a0) |= (a1); \
(a2) ^= (a3); \
(a1) = SPH_T ## width(~(a1)); \
(a0) ^= (a3); \
(a3) &= tmp; \
(a1) ^= (a3); \
(a3) ^= (a2); \
(a2) &= (a0); \
(a0) = SPH_T ## width(~(a0)); \
(a2) ^= (a1); \
(a1) |= (a3); \
tmp ^= (a1); \
(a3) ^= (a2); \
(a2) &= (a1); \
(a1) ^= (a0); \
(a0) = tmp; \
} while (0)
#define SUB_CRUMB(a0, a1, a2, a3) SUB_CRUMB_GEN(a0, a1, a2, a3, 32)
#define SUB_CRUMBW(a0, a1, a2, a3) SUB_CRUMB_GEN(a0, a1, a2, a3, 64)
#if 0
#define ROL32W(x, n) SPH_T64( \
(((x) << (n)) \
& ~((SPH_C64(0xFFFFFFFF) >> (32 - (n))) << 32)) \
| (((x) >> (32 - (n))) \
& ~((SPH_C64(0xFFFFFFFF) >> (n)) << (n))))
#define MIX_WORDW(u, v) do { \
(v) ^= (u); \
(u) = ROL32W((u), 2) ^ (v); \
(v) = ROL32W((v), 14) ^ (u); \
(u) = ROL32W((u), 10) ^ (v); \
(v) = ROL32W((v), 1); \
} while (0)
#endif
#define MIX_WORDW(u, v) do { \
sph_u32 ul, uh, vl, vh; \
(v) ^= (u); \
ul = SPH_T32((sph_u32)(u)); \
uh = SPH_T32((sph_u32)((u) >> 32)); \
vl = SPH_T32((sph_u32)(v)); \
vh = SPH_T32((sph_u32)((v) >> 32)); \
ul = SPH_ROTL32(ul, 2) ^ vl; \
vl = SPH_ROTL32(vl, 14) ^ ul; \
ul = SPH_ROTL32(ul, 10) ^ vl; \
vl = SPH_ROTL32(vl, 1); \
uh = SPH_ROTL32(uh, 2) ^ vh; \
vh = SPH_ROTL32(vh, 14) ^ uh; \
uh = SPH_ROTL32(uh, 10) ^ vh; \
vh = SPH_ROTL32(vh, 1); \
(u) = (sph_u64)ul | ((sph_u64)uh << 32); \
(v) = (sph_u64)vl | ((sph_u64)vh << 32); \
} while (0)
#else
#define SUB_CRUMB(a0, a1, a2, a3) do { \
sph_u32 tmp; \
tmp = (a0); \
(a0) |= (a1); \
(a2) ^= (a3); \
(a1) = SPH_T32(~(a1)); \
(a0) ^= (a3); \
(a3) &= tmp; \
(a1) ^= (a3); \
(a3) ^= (a2); \
(a2) &= (a0); \
(a0) = SPH_T32(~(a0)); \
(a2) ^= (a1); \
(a1) |= (a3); \
tmp ^= (a1); \
(a3) ^= (a2); \
(a2) &= (a1); \
(a1) ^= (a0); \
(a0) = tmp; \
} while (0)
#endif
#define MIX_WORD(u, v) do { \
(v) ^= (u); \
(u) = SPH_ROTL32((u), 2) ^ (v); \
(v) = SPH_ROTL32((v), 14) ^ (u); \
(u) = SPH_ROTL32((u), 10) ^ (v); \
(v) = SPH_ROTL32((v), 1); \
} while (0)
#define DECL_STATE3 \
sph_u32 V00, V01, V02, V03, V04, V05, V06, V07; \
sph_u32 V10, V11, V12, V13, V14, V15, V16, V17; \
sph_u32 V20, V21, V22, V23, V24, V25, V26, V27;
#define READ_STATE3(state) do { \
V00 = (state)->V[0][0]; \
V01 = (state)->V[0][1]; \
V02 = (state)->V[0][2]; \
V03 = (state)->V[0][3]; \
V04 = (state)->V[0][4]; \
V05 = (state)->V[0][5]; \
V06 = (state)->V[0][6]; \
V07 = (state)->V[0][7]; \
V10 = (state)->V[1][0]; \
V11 = (state)->V[1][1]; \
V12 = (state)->V[1][2]; \
V13 = (state)->V[1][3]; \
V14 = (state)->V[1][4]; \
V15 = (state)->V[1][5]; \
V16 = (state)->V[1][6]; \
V17 = (state)->V[1][7]; \
V20 = (state)->V[2][0]; \
V21 = (state)->V[2][1]; \
V22 = (state)->V[2][2]; \
V23 = (state)->V[2][3]; \
V24 = (state)->V[2][4]; \
V25 = (state)->V[2][5]; \
V26 = (state)->V[2][6]; \
V27 = (state)->V[2][7]; \
} while (0)
#define WRITE_STATE3(state) do { \
(state)->V[0][0] = V00; \
(state)->V[0][1] = V01; \
(state)->V[0][2] = V02; \
(state)->V[0][3] = V03; \
(state)->V[0][4] = V04; \
(state)->V[0][5] = V05; \
(state)->V[0][6] = V06; \
(state)->V[0][7] = V07; \
(state)->V[1][0] = V10; \
(state)->V[1][1] = V11; \
(state)->V[1][2] = V12; \
(state)->V[1][3] = V13; \
(state)->V[1][4] = V14; \
(state)->V[1][5] = V15; \
(state)->V[1][6] = V16; \
(state)->V[1][7] = V17; \
(state)->V[2][0] = V20; \
(state)->V[2][1] = V21; \
(state)->V[2][2] = V22; \
(state)->V[2][3] = V23; \
(state)->V[2][4] = V24; \
(state)->V[2][5] = V25; \
(state)->V[2][6] = V26; \
(state)->V[2][7] = V27; \
} while (0)
#define MI3 do { \
DECL_TMP8(M) \
DECL_TMP8(a) \
M0 = sph_dec32be_aligned(buf + 0); \
M1 = sph_dec32be_aligned(buf + 4); \
M2 = sph_dec32be_aligned(buf + 8); \
M3 = sph_dec32be_aligned(buf + 12); \
M4 = sph_dec32be_aligned(buf + 16); \
M5 = sph_dec32be_aligned(buf + 20); \
M6 = sph_dec32be_aligned(buf + 24); \
M7 = sph_dec32be_aligned(buf + 28); \
XOR(a, V0, V1); \
XOR(a, a, V2); \
M2(a, a); \
XOR(V0, a, V0); \
XOR(V0, M, V0); \
M2(M, M); \
XOR(V1, a, V1); \
XOR(V1, M, V1); \
M2(M, M); \
XOR(V2, a, V2); \
XOR(V2, M, V2); \
} while (0)
#define TWEAK3 do { \
V14 = SPH_ROTL32(V14, 1); \
V15 = SPH_ROTL32(V15, 1); \
V16 = SPH_ROTL32(V16, 1); \
V17 = SPH_ROTL32(V17, 1); \
V24 = SPH_ROTL32(V24, 2); \
V25 = SPH_ROTL32(V25, 2); \
V26 = SPH_ROTL32(V26, 2); \
V27 = SPH_ROTL32(V27, 2); \
} while (0)
#if SPH_LUFFA_PARALLEL
#define P3 do { \
int r; \
sph_u64 W0, W1, W2, W3, W4, W5, W6, W7; \
TWEAK3; \
W0 = (sph_u64)V00 | ((sph_u64)V10 << 32); \
W1 = (sph_u64)V01 | ((sph_u64)V11 << 32); \
W2 = (sph_u64)V02 | ((sph_u64)V12 << 32); \
W3 = (sph_u64)V03 | ((sph_u64)V13 << 32); \
W4 = (sph_u64)V04 | ((sph_u64)V14 << 32); \
W5 = (sph_u64)V05 | ((sph_u64)V15 << 32); \
W6 = (sph_u64)V06 | ((sph_u64)V16 << 32); \
W7 = (sph_u64)V07 | ((sph_u64)V17 << 32); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMBW(W0, W1, W2, W3); \
SUB_CRUMBW(W5, W6, W7, W4); \
MIX_WORDW(W0, W4); \
MIX_WORDW(W1, W5); \
MIX_WORDW(W2, W6); \
MIX_WORDW(W3, W7); \
W0 ^= RCW010[r]; \
W4 ^= RCW014[r]; \
} \
V00 = SPH_T32((sph_u32)W0); \
V10 = SPH_T32((sph_u32)(W0 >> 32)); \
V01 = SPH_T32((sph_u32)W1); \
V11 = SPH_T32((sph_u32)(W1 >> 32)); \
V02 = SPH_T32((sph_u32)W2); \
V12 = SPH_T32((sph_u32)(W2 >> 32)); \
V03 = SPH_T32((sph_u32)W3); \
V13 = SPH_T32((sph_u32)(W3 >> 32)); \
V04 = SPH_T32((sph_u32)W4); \
V14 = SPH_T32((sph_u32)(W4 >> 32)); \
V05 = SPH_T32((sph_u32)W5); \
V15 = SPH_T32((sph_u32)(W5 >> 32)); \
V06 = SPH_T32((sph_u32)W6); \
V16 = SPH_T32((sph_u32)(W6 >> 32)); \
V07 = SPH_T32((sph_u32)W7); \
V17 = SPH_T32((sph_u32)(W7 >> 32)); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V20, V21, V22, V23); \
SUB_CRUMB(V25, V26, V27, V24); \
MIX_WORD(V20, V24); \
MIX_WORD(V21, V25); \
MIX_WORD(V22, V26); \
MIX_WORD(V23, V27); \
V20 ^= RC20[r]; \
V24 ^= RC24[r]; \
} \
} while (0)
#else
#define P3 do { \
int r; \
TWEAK3; \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V00, V01, V02, V03); \
SUB_CRUMB(V05, V06, V07, V04); \
MIX_WORD(V00, V04); \
MIX_WORD(V01, V05); \
MIX_WORD(V02, V06); \
MIX_WORD(V03, V07); \
V00 ^= RC00[r]; \
V04 ^= RC04[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V10, V11, V12, V13); \
SUB_CRUMB(V15, V16, V17, V14); \
MIX_WORD(V10, V14); \
MIX_WORD(V11, V15); \
MIX_WORD(V12, V16); \
MIX_WORD(V13, V17); \
V10 ^= RC10[r]; \
V14 ^= RC14[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V20, V21, V22, V23); \
SUB_CRUMB(V25, V26, V27, V24); \
MIX_WORD(V20, V24); \
MIX_WORD(V21, V25); \
MIX_WORD(V22, V26); \
MIX_WORD(V23, V27); \
V20 ^= RC20[r]; \
V24 ^= RC24[r]; \
} \
} while (0)
#endif
#define DECL_STATE4 \
sph_u32 V00, V01, V02, V03, V04, V05, V06, V07; \
sph_u32 V10, V11, V12, V13, V14, V15, V16, V17; \
sph_u32 V20, V21, V22, V23, V24, V25, V26, V27; \
sph_u32 V30, V31, V32, V33, V34, V35, V36, V37;
#define READ_STATE4(state) do { \
V00 = (state)->V[0][0]; \
V01 = (state)->V[0][1]; \
V02 = (state)->V[0][2]; \
V03 = (state)->V[0][3]; \
V04 = (state)->V[0][4]; \
V05 = (state)->V[0][5]; \
V06 = (state)->V[0][6]; \
V07 = (state)->V[0][7]; \
V10 = (state)->V[1][0]; \
V11 = (state)->V[1][1]; \
V12 = (state)->V[1][2]; \
V13 = (state)->V[1][3]; \
V14 = (state)->V[1][4]; \
V15 = (state)->V[1][5]; \
V16 = (state)->V[1][6]; \
V17 = (state)->V[1][7]; \
V20 = (state)->V[2][0]; \
V21 = (state)->V[2][1]; \
V22 = (state)->V[2][2]; \
V23 = (state)->V[2][3]; \
V24 = (state)->V[2][4]; \
V25 = (state)->V[2][5]; \
V26 = (state)->V[2][6]; \
V27 = (state)->V[2][7]; \
V30 = (state)->V[3][0]; \
V31 = (state)->V[3][1]; \
V32 = (state)->V[3][2]; \
V33 = (state)->V[3][3]; \
V34 = (state)->V[3][4]; \
V35 = (state)->V[3][5]; \
V36 = (state)->V[3][6]; \
V37 = (state)->V[3][7]; \
} while (0)
#define WRITE_STATE4(state) do { \
(state)->V[0][0] = V00; \
(state)->V[0][1] = V01; \
(state)->V[0][2] = V02; \
(state)->V[0][3] = V03; \
(state)->V[0][4] = V04; \
(state)->V[0][5] = V05; \
(state)->V[0][6] = V06; \
(state)->V[0][7] = V07; \
(state)->V[1][0] = V10; \
(state)->V[1][1] = V11; \
(state)->V[1][2] = V12; \
(state)->V[1][3] = V13; \
(state)->V[1][4] = V14; \
(state)->V[1][5] = V15; \
(state)->V[1][6] = V16; \
(state)->V[1][7] = V17; \
(state)->V[2][0] = V20; \
(state)->V[2][1] = V21; \
(state)->V[2][2] = V22; \
(state)->V[2][3] = V23; \
(state)->V[2][4] = V24; \
(state)->V[2][5] = V25; \
(state)->V[2][6] = V26; \
(state)->V[2][7] = V27; \
(state)->V[3][0] = V30; \
(state)->V[3][1] = V31; \
(state)->V[3][2] = V32; \
(state)->V[3][3] = V33; \
(state)->V[3][4] = V34; \
(state)->V[3][5] = V35; \
(state)->V[3][6] = V36; \
(state)->V[3][7] = V37; \
} while (0)
#define MI4 do { \
DECL_TMP8(M) \
DECL_TMP8(a) \
DECL_TMP8(b) \
M0 = sph_dec32be_aligned(buf + 0); \
M1 = sph_dec32be_aligned(buf + 4); \
M2 = sph_dec32be_aligned(buf + 8); \
M3 = sph_dec32be_aligned(buf + 12); \
M4 = sph_dec32be_aligned(buf + 16); \
M5 = sph_dec32be_aligned(buf + 20); \
M6 = sph_dec32be_aligned(buf + 24); \
M7 = sph_dec32be_aligned(buf + 28); \
XOR(a, V0, V1); \
XOR(b, V2, V3); \
XOR(a, a, b); \
M2(a, a); \
XOR(V0, a, V0); \
XOR(V1, a, V1); \
XOR(V2, a, V2); \
XOR(V3, a, V3); \
M2(b, V0); \
XOR(b, b, V3); \
M2(V3, V3); \
XOR(V3, V3, V2); \
M2(V2, V2); \
XOR(V2, V2, V1); \
M2(V1, V1); \
XOR(V1, V1, V0); \
XOR(V0, b, M); \
M2(M, M); \
XOR(V1, V1, M); \
M2(M, M); \
XOR(V2, V2, M); \
M2(M, M); \
XOR(V3, V3, M); \
} while (0)
#define TWEAK4 do { \
V14 = SPH_ROTL32(V14, 1); \
V15 = SPH_ROTL32(V15, 1); \
V16 = SPH_ROTL32(V16, 1); \
V17 = SPH_ROTL32(V17, 1); \
V24 = SPH_ROTL32(V24, 2); \
V25 = SPH_ROTL32(V25, 2); \
V26 = SPH_ROTL32(V26, 2); \
V27 = SPH_ROTL32(V27, 2); \
V34 = SPH_ROTL32(V34, 3); \
V35 = SPH_ROTL32(V35, 3); \
V36 = SPH_ROTL32(V36, 3); \
V37 = SPH_ROTL32(V37, 3); \
} while (0)
#if SPH_LUFFA_PARALLEL
#define P4 do { \
int r; \
sph_u64 W0, W1, W2, W3, W4, W5, W6, W7; \
TWEAK4; \
W0 = (sph_u64)V00 | ((sph_u64)V10 << 32); \
W1 = (sph_u64)V01 | ((sph_u64)V11 << 32); \
W2 = (sph_u64)V02 | ((sph_u64)V12 << 32); \
W3 = (sph_u64)V03 | ((sph_u64)V13 << 32); \
W4 = (sph_u64)V04 | ((sph_u64)V14 << 32); \
W5 = (sph_u64)V05 | ((sph_u64)V15 << 32); \
W6 = (sph_u64)V06 | ((sph_u64)V16 << 32); \
W7 = (sph_u64)V07 | ((sph_u64)V17 << 32); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMBW(W0, W1, W2, W3); \
SUB_CRUMBW(W5, W6, W7, W4); \
MIX_WORDW(W0, W4); \
MIX_WORDW(W1, W5); \
MIX_WORDW(W2, W6); \
MIX_WORDW(W3, W7); \
W0 ^= RCW010[r]; \
W4 ^= RCW014[r]; \
} \
V00 = SPH_T32((sph_u32)W0); \
V10 = SPH_T32((sph_u32)(W0 >> 32)); \
V01 = SPH_T32((sph_u32)W1); \
V11 = SPH_T32((sph_u32)(W1 >> 32)); \
V02 = SPH_T32((sph_u32)W2); \
V12 = SPH_T32((sph_u32)(W2 >> 32)); \
V03 = SPH_T32((sph_u32)W3); \
V13 = SPH_T32((sph_u32)(W3 >> 32)); \
V04 = SPH_T32((sph_u32)W4); \
V14 = SPH_T32((sph_u32)(W4 >> 32)); \
V05 = SPH_T32((sph_u32)W5); \
V15 = SPH_T32((sph_u32)(W5 >> 32)); \
V06 = SPH_T32((sph_u32)W6); \
V16 = SPH_T32((sph_u32)(W6 >> 32)); \
V07 = SPH_T32((sph_u32)W7); \
V17 = SPH_T32((sph_u32)(W7 >> 32)); \
W0 = (sph_u64)V20 | ((sph_u64)V30 << 32); \
W1 = (sph_u64)V21 | ((sph_u64)V31 << 32); \
W2 = (sph_u64)V22 | ((sph_u64)V32 << 32); \
W3 = (sph_u64)V23 | ((sph_u64)V33 << 32); \
W4 = (sph_u64)V24 | ((sph_u64)V34 << 32); \
W5 = (sph_u64)V25 | ((sph_u64)V35 << 32); \
W6 = (sph_u64)V26 | ((sph_u64)V36 << 32); \
W7 = (sph_u64)V27 | ((sph_u64)V37 << 32); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMBW(W0, W1, W2, W3); \
SUB_CRUMBW(W5, W6, W7, W4); \
MIX_WORDW(W0, W4); \
MIX_WORDW(W1, W5); \
MIX_WORDW(W2, W6); \
MIX_WORDW(W3, W7); \
W0 ^= RCW230[r]; \
W4 ^= RCW234[r]; \
} \
V20 = SPH_T32((sph_u32)W0); \
V30 = SPH_T32((sph_u32)(W0 >> 32)); \
V21 = SPH_T32((sph_u32)W1); \
V31 = SPH_T32((sph_u32)(W1 >> 32)); \
V22 = SPH_T32((sph_u32)W2); \
V32 = SPH_T32((sph_u32)(W2 >> 32)); \
V23 = SPH_T32((sph_u32)W3); \
V33 = SPH_T32((sph_u32)(W3 >> 32)); \
V24 = SPH_T32((sph_u32)W4); \
V34 = SPH_T32((sph_u32)(W4 >> 32)); \
V25 = SPH_T32((sph_u32)W5); \
V35 = SPH_T32((sph_u32)(W5 >> 32)); \
V26 = SPH_T32((sph_u32)W6); \
V36 = SPH_T32((sph_u32)(W6 >> 32)); \
V27 = SPH_T32((sph_u32)W7); \
V37 = SPH_T32((sph_u32)(W7 >> 32)); \
} while (0)
#else
#define P4 do { \
int r; \
TWEAK4; \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V00, V01, V02, V03); \
SUB_CRUMB(V05, V06, V07, V04); \
MIX_WORD(V00, V04); \
MIX_WORD(V01, V05); \
MIX_WORD(V02, V06); \
MIX_WORD(V03, V07); \
V00 ^= RC00[r]; \
V04 ^= RC04[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V10, V11, V12, V13); \
SUB_CRUMB(V15, V16, V17, V14); \
MIX_WORD(V10, V14); \
MIX_WORD(V11, V15); \
MIX_WORD(V12, V16); \
MIX_WORD(V13, V17); \
V10 ^= RC10[r]; \
V14 ^= RC14[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V20, V21, V22, V23); \
SUB_CRUMB(V25, V26, V27, V24); \
MIX_WORD(V20, V24); \
MIX_WORD(V21, V25); \
MIX_WORD(V22, V26); \
MIX_WORD(V23, V27); \
V20 ^= RC20[r]; \
V24 ^= RC24[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V30, V31, V32, V33); \
SUB_CRUMB(V35, V36, V37, V34); \
MIX_WORD(V30, V34); \
MIX_WORD(V31, V35); \
MIX_WORD(V32, V36); \
MIX_WORD(V33, V37); \
V30 ^= RC30[r]; \
V34 ^= RC34[r]; \
} \
} while (0)
#endif
#define DECL_STATE5 \
sph_u32 V00, V01, V02, V03, V04, V05, V06, V07; \
sph_u32 V10, V11, V12, V13, V14, V15, V16, V17; \
sph_u32 V20, V21, V22, V23, V24, V25, V26, V27; \
sph_u32 V30, V31, V32, V33, V34, V35, V36, V37; \
sph_u32 V40, V41, V42, V43, V44, V45, V46, V47;
#define READ_STATE5(state) do { \
V00 = (state)->V[0][0]; \
V01 = (state)->V[0][1]; \
V02 = (state)->V[0][2]; \
V03 = (state)->V[0][3]; \
V04 = (state)->V[0][4]; \
V05 = (state)->V[0][5]; \
V06 = (state)->V[0][6]; \
V07 = (state)->V[0][7]; \
V10 = (state)->V[1][0]; \
V11 = (state)->V[1][1]; \
V12 = (state)->V[1][2]; \
V13 = (state)->V[1][3]; \
V14 = (state)->V[1][4]; \
V15 = (state)->V[1][5]; \
V16 = (state)->V[1][6]; \
V17 = (state)->V[1][7]; \
V20 = (state)->V[2][0]; \
V21 = (state)->V[2][1]; \
V22 = (state)->V[2][2]; \
V23 = (state)->V[2][3]; \
V24 = (state)->V[2][4]; \
V25 = (state)->V[2][5]; \
V26 = (state)->V[2][6]; \
V27 = (state)->V[2][7]; \
V30 = (state)->V[3][0]; \
V31 = (state)->V[3][1]; \
V32 = (state)->V[3][2]; \
V33 = (state)->V[3][3]; \
V34 = (state)->V[3][4]; \
V35 = (state)->V[3][5]; \
V36 = (state)->V[3][6]; \
V37 = (state)->V[3][7]; \
V40 = (state)->V[4][0]; \
V41 = (state)->V[4][1]; \
V42 = (state)->V[4][2]; \
V43 = (state)->V[4][3]; \
V44 = (state)->V[4][4]; \
V45 = (state)->V[4][5]; \
V46 = (state)->V[4][6]; \
V47 = (state)->V[4][7]; \
} while (0)
#define WRITE_STATE5(state) do { \
(state)->V[0][0] = V00; \
(state)->V[0][1] = V01; \
(state)->V[0][2] = V02; \
(state)->V[0][3] = V03; \
(state)->V[0][4] = V04; \
(state)->V[0][5] = V05; \
(state)->V[0][6] = V06; \
(state)->V[0][7] = V07; \
(state)->V[1][0] = V10; \
(state)->V[1][1] = V11; \
(state)->V[1][2] = V12; \
(state)->V[1][3] = V13; \
(state)->V[1][4] = V14; \
(state)->V[1][5] = V15; \
(state)->V[1][6] = V16; \
(state)->V[1][7] = V17; \
(state)->V[2][0] = V20; \
(state)->V[2][1] = V21; \
(state)->V[2][2] = V22; \
(state)->V[2][3] = V23; \
(state)->V[2][4] = V24; \
(state)->V[2][5] = V25; \
(state)->V[2][6] = V26; \
(state)->V[2][7] = V27; \
(state)->V[3][0] = V30; \
(state)->V[3][1] = V31; \
(state)->V[3][2] = V32; \
(state)->V[3][3] = V33; \
(state)->V[3][4] = V34; \
(state)->V[3][5] = V35; \
(state)->V[3][6] = V36; \
(state)->V[3][7] = V37; \
(state)->V[4][0] = V40; \
(state)->V[4][1] = V41; \
(state)->V[4][2] = V42; \
(state)->V[4][3] = V43; \
(state)->V[4][4] = V44; \
(state)->V[4][5] = V45; \
(state)->V[4][6] = V46; \
(state)->V[4][7] = V47; \
} while (0)
#define MI5 do { \
DECL_TMP8(M) \
DECL_TMP8(a) \
DECL_TMP8(b) \
M0 = sph_dec32be_aligned(buf + 0); \
M1 = sph_dec32be_aligned(buf + 4); \
M2 = sph_dec32be_aligned(buf + 8); \
M3 = sph_dec32be_aligned(buf + 12); \
M4 = sph_dec32be_aligned(buf + 16); \
M5 = sph_dec32be_aligned(buf + 20); \
M6 = sph_dec32be_aligned(buf + 24); \
M7 = sph_dec32be_aligned(buf + 28); \
XOR(a, V0, V1); \
XOR(b, V2, V3); \
XOR(a, a, b); \
XOR(a, a, V4); \
M2(a, a); \
XOR(V0, a, V0); \
XOR(V1, a, V1); \
XOR(V2, a, V2); \
XOR(V3, a, V3); \
XOR(V4, a, V4); \
M2(b, V0); \
XOR(b, b, V1); \
M2(V1, V1); \
XOR(V1, V1, V2); \
M2(V2, V2); \
XOR(V2, V2, V3); \
M2(V3, V3); \
XOR(V3, V3, V4); \
M2(V4, V4); \
XOR(V4, V4, V0); \
M2(V0, b); \
XOR(V0, V0, V4); \
M2(V4, V4); \
XOR(V4, V4, V3); \
M2(V3, V3); \
XOR(V3, V3, V2); \
M2(V2, V2); \
XOR(V2, V2, V1); \
M2(V1, V1); \
XOR(V1, V1, b); \
XOR(V0, V0, M); \
M2(M, M); \
XOR(V1, V1, M); \
M2(M, M); \
XOR(V2, V2, M); \
M2(M, M); \
XOR(V3, V3, M); \
M2(M, M); \
XOR(V4, V4, M); \
} while (0)
#define TWEAK5 do { \
V14 = SPH_ROTL32(V14, 1); \
V15 = SPH_ROTL32(V15, 1); \
V16 = SPH_ROTL32(V16, 1); \
V17 = SPH_ROTL32(V17, 1); \
V24 = SPH_ROTL32(V24, 2); \
V25 = SPH_ROTL32(V25, 2); \
V26 = SPH_ROTL32(V26, 2); \
V27 = SPH_ROTL32(V27, 2); \
V34 = SPH_ROTL32(V34, 3); \
V35 = SPH_ROTL32(V35, 3); \
V36 = SPH_ROTL32(V36, 3); \
V37 = SPH_ROTL32(V37, 3); \
V44 = SPH_ROTL32(V44, 4); \
V45 = SPH_ROTL32(V45, 4); \
V46 = SPH_ROTL32(V46, 4); \
V47 = SPH_ROTL32(V47, 4); \
} while (0)
#if SPH_LUFFA_PARALLEL
#define P5 do { \
int r; \
sph_u64 W0, W1, W2, W3, W4, W5, W6, W7; \
TWEAK5; \
W0 = (sph_u64)V00 | ((sph_u64)V10 << 32); \
W1 = (sph_u64)V01 | ((sph_u64)V11 << 32); \
W2 = (sph_u64)V02 | ((sph_u64)V12 << 32); \
W3 = (sph_u64)V03 | ((sph_u64)V13 << 32); \
W4 = (sph_u64)V04 | ((sph_u64)V14 << 32); \
W5 = (sph_u64)V05 | ((sph_u64)V15 << 32); \
W6 = (sph_u64)V06 | ((sph_u64)V16 << 32); \
W7 = (sph_u64)V07 | ((sph_u64)V17 << 32); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMBW(W0, W1, W2, W3); \
SUB_CRUMBW(W5, W6, W7, W4); \
MIX_WORDW(W0, W4); \
MIX_WORDW(W1, W5); \
MIX_WORDW(W2, W6); \
MIX_WORDW(W3, W7); \
W0 ^= RCW010[r]; \
W4 ^= RCW014[r]; \
} \
V00 = SPH_T32((sph_u32)W0); \
V10 = SPH_T32((sph_u32)(W0 >> 32)); \
V01 = SPH_T32((sph_u32)W1); \
V11 = SPH_T32((sph_u32)(W1 >> 32)); \
V02 = SPH_T32((sph_u32)W2); \
V12 = SPH_T32((sph_u32)(W2 >> 32)); \
V03 = SPH_T32((sph_u32)W3); \
V13 = SPH_T32((sph_u32)(W3 >> 32)); \
V04 = SPH_T32((sph_u32)W4); \
V14 = SPH_T32((sph_u32)(W4 >> 32)); \
V05 = SPH_T32((sph_u32)W5); \
V15 = SPH_T32((sph_u32)(W5 >> 32)); \
V06 = SPH_T32((sph_u32)W6); \
V16 = SPH_T32((sph_u32)(W6 >> 32)); \
V07 = SPH_T32((sph_u32)W7); \
V17 = SPH_T32((sph_u32)(W7 >> 32)); \
W0 = (sph_u64)V20 | ((sph_u64)V30 << 32); \
W1 = (sph_u64)V21 | ((sph_u64)V31 << 32); \
W2 = (sph_u64)V22 | ((sph_u64)V32 << 32); \
W3 = (sph_u64)V23 | ((sph_u64)V33 << 32); \
W4 = (sph_u64)V24 | ((sph_u64)V34 << 32); \
W5 = (sph_u64)V25 | ((sph_u64)V35 << 32); \
W6 = (sph_u64)V26 | ((sph_u64)V36 << 32); \
W7 = (sph_u64)V27 | ((sph_u64)V37 << 32); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMBW(W0, W1, W2, W3); \
SUB_CRUMBW(W5, W6, W7, W4); \
MIX_WORDW(W0, W4); \
MIX_WORDW(W1, W5); \
MIX_WORDW(W2, W6); \
MIX_WORDW(W3, W7); \
W0 ^= RCW230[r]; \
W4 ^= RCW234[r]; \
} \
V20 = SPH_T32((sph_u32)W0); \
V30 = SPH_T32((sph_u32)(W0 >> 32)); \
V21 = SPH_T32((sph_u32)W1); \
V31 = SPH_T32((sph_u32)(W1 >> 32)); \
V22 = SPH_T32((sph_u32)W2); \
V32 = SPH_T32((sph_u32)(W2 >> 32)); \
V23 = SPH_T32((sph_u32)W3); \
V33 = SPH_T32((sph_u32)(W3 >> 32)); \
V24 = SPH_T32((sph_u32)W4); \
V34 = SPH_T32((sph_u32)(W4 >> 32)); \
V25 = SPH_T32((sph_u32)W5); \
V35 = SPH_T32((sph_u32)(W5 >> 32)); \
V26 = SPH_T32((sph_u32)W6); \
V36 = SPH_T32((sph_u32)(W6 >> 32)); \
V27 = SPH_T32((sph_u32)W7); \
V37 = SPH_T32((sph_u32)(W7 >> 32)); \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V40, V41, V42, V43); \
SUB_CRUMB(V45, V46, V47, V44); \
MIX_WORD(V40, V44); \
MIX_WORD(V41, V45); \
MIX_WORD(V42, V46); \
MIX_WORD(V43, V47); \
V40 ^= RC40[r]; \
V44 ^= RC44[r]; \
} \
} while (0)
#else
#define P5 do { \
int r; \
TWEAK5; \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V00, V01, V02, V03); \
SUB_CRUMB(V05, V06, V07, V04); \
MIX_WORD(V00, V04); \
MIX_WORD(V01, V05); \
MIX_WORD(V02, V06); \
MIX_WORD(V03, V07); \
V00 ^= RC00[r]; \
V04 ^= RC04[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V10, V11, V12, V13); \
SUB_CRUMB(V15, V16, V17, V14); \
MIX_WORD(V10, V14); \
MIX_WORD(V11, V15); \
MIX_WORD(V12, V16); \
MIX_WORD(V13, V17); \
V10 ^= RC10[r]; \
V14 ^= RC14[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V20, V21, V22, V23); \
SUB_CRUMB(V25, V26, V27, V24); \
MIX_WORD(V20, V24); \
MIX_WORD(V21, V25); \
MIX_WORD(V22, V26); \
MIX_WORD(V23, V27); \
V20 ^= RC20[r]; \
V24 ^= RC24[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V30, V31, V32, V33); \
SUB_CRUMB(V35, V36, V37, V34); \
MIX_WORD(V30, V34); \
MIX_WORD(V31, V35); \
MIX_WORD(V32, V36); \
MIX_WORD(V33, V37); \
V30 ^= RC30[r]; \
V34 ^= RC34[r]; \
} \
for (r = 0; r < 8; r ++) { \
SUB_CRUMB(V40, V41, V42, V43); \
SUB_CRUMB(V45, V46, V47, V44); \
MIX_WORD(V40, V44); \
MIX_WORD(V41, V45); \
MIX_WORD(V42, V46); \
MIX_WORD(V43, V47); \
V40 ^= RC40[r]; \
V44 ^= RC44[r]; \
} \
} while (0)
#endif
static void
luffa3(sph_luffa224_context *sc, const void *data, size_t len)
{
unsigned char *buf;
size_t ptr;
DECL_STATE3
buf = sc->buf;
ptr = sc->ptr;
if (len < (sizeof sc->buf) - ptr) {
memcpy(buf + ptr, data, len);
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE3(sc);
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - ptr;
if (clen > len)
clen = len;
memcpy(buf + ptr, data, clen);
ptr += clen;
data = (const unsigned char *)data + clen;
len -= clen;
if (ptr == sizeof sc->buf) {
MI3;
P3;
ptr = 0;
}
}
WRITE_STATE3(sc);
sc->ptr = ptr;
}
static void
luffa3_close(sph_luffa224_context *sc, unsigned ub, unsigned n,
void *dst, unsigned out_size_w32)
{
unsigned char *buf, *out;
size_t ptr;
unsigned z;
int i;
DECL_STATE3
buf = sc->buf;
ptr = sc->ptr;
z = 0x80 >> n;
buf[ptr ++] = ((ub & -z) | z) & 0xFF;
memset(buf + ptr, 0, (sizeof sc->buf) - ptr);
READ_STATE3(sc);
for (i = 0; i < 2; i ++) {
MI3;
P3;
memset(buf, 0, sizeof sc->buf);
}
out = dst;
sph_enc32be(out + 0, V00 ^ V10 ^ V20);
sph_enc32be(out + 4, V01 ^ V11 ^ V21);
sph_enc32be(out + 8, V02 ^ V12 ^ V22);
sph_enc32be(out + 12, V03 ^ V13 ^ V23);
sph_enc32be(out + 16, V04 ^ V14 ^ V24);
sph_enc32be(out + 20, V05 ^ V15 ^ V25);
sph_enc32be(out + 24, V06 ^ V16 ^ V26);
if (out_size_w32 > 7)
sph_enc32be(out + 28, V07 ^ V17 ^ V27);
}
static void
luffa4(sph_luffa384_context *sc, const void *data, size_t len)
{
unsigned char *buf;
size_t ptr;
DECL_STATE4
buf = sc->buf;
ptr = sc->ptr;
if (len < (sizeof sc->buf) - ptr) {
memcpy(buf + ptr, data, len);
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE4(sc);
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - ptr;
if (clen > len)
clen = len;
memcpy(buf + ptr, data, clen);
ptr += clen;
data = (const unsigned char *)data + clen;
len -= clen;
if (ptr == sizeof sc->buf) {
MI4;
P4;
ptr = 0;
}
}
WRITE_STATE4(sc);
sc->ptr = ptr;
}
static void
luffa4_close(sph_luffa384_context *sc, unsigned ub, unsigned n, void *dst)
{
unsigned char *buf, *out;
size_t ptr;
unsigned z;
int i;
DECL_STATE4
buf = sc->buf;
ptr = sc->ptr;
out = dst;
z = 0x80 >> n;
buf[ptr ++] = ((ub & -z) | z) & 0xFF;
memset(buf + ptr, 0, (sizeof sc->buf) - ptr);
READ_STATE4(sc);
for (i = 0; i < 3; i ++) {
MI4;
P4;
switch (i) {
case 0:
memset(buf, 0, sizeof sc->buf);
break;
case 1:
sph_enc32be(out + 0, V00 ^ V10 ^ V20 ^ V30);
sph_enc32be(out + 4, V01 ^ V11 ^ V21 ^ V31);
sph_enc32be(out + 8, V02 ^ V12 ^ V22 ^ V32);
sph_enc32be(out + 12, V03 ^ V13 ^ V23 ^ V33);
sph_enc32be(out + 16, V04 ^ V14 ^ V24 ^ V34);
sph_enc32be(out + 20, V05 ^ V15 ^ V25 ^ V35);
sph_enc32be(out + 24, V06 ^ V16 ^ V26 ^ V36);
sph_enc32be(out + 28, V07 ^ V17 ^ V27 ^ V37);
break;
case 2:
sph_enc32be(out + 32, V00 ^ V10 ^ V20 ^ V30);
sph_enc32be(out + 36, V01 ^ V11 ^ V21 ^ V31);
sph_enc32be(out + 40, V02 ^ V12 ^ V22 ^ V32);
sph_enc32be(out + 44, V03 ^ V13 ^ V23 ^ V33);
break;
}
}
}
static void
luffa5(sph_luffa512_context *sc, const void *data, size_t len)
{
unsigned char *buf;
size_t ptr;
DECL_STATE5
buf = sc->buf;
ptr = sc->ptr;
if (len < (sizeof sc->buf) - ptr) {
memcpy(buf + ptr, data, len);
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE5(sc);
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - ptr;
if (clen > len)
clen = len;
memcpy(buf + ptr, data, clen);
ptr += clen;
data = (const unsigned char *)data + clen;
len -= clen;
if (ptr == sizeof sc->buf) {
MI5;
P5;
ptr = 0;
}
}
WRITE_STATE5(sc);
sc->ptr = ptr;
}
static void
luffa5_close(sph_luffa512_context *sc, unsigned ub, unsigned n, void *dst)
{
unsigned char *buf, *out;
size_t ptr;
unsigned z;
int i;
DECL_STATE5
buf = sc->buf;
ptr = sc->ptr;
out = dst;
z = 0x80 >> n;
buf[ptr ++] = ((ub & -z) | z) & 0xFF;
memset(buf + ptr, 0, (sizeof sc->buf) - ptr);
READ_STATE5(sc);
for (i = 0; i < 3; i ++) {
MI5;
P5;
switch (i) {
case 0:
memset(buf, 0, sizeof sc->buf);
break;
case 1:
sph_enc32be(out + 0, V00 ^ V10 ^ V20 ^ V30 ^ V40);
sph_enc32be(out + 4, V01 ^ V11 ^ V21 ^ V31 ^ V41);
sph_enc32be(out + 8, V02 ^ V12 ^ V22 ^ V32 ^ V42);
sph_enc32be(out + 12, V03 ^ V13 ^ V23 ^ V33 ^ V43);
sph_enc32be(out + 16, V04 ^ V14 ^ V24 ^ V34 ^ V44);
sph_enc32be(out + 20, V05 ^ V15 ^ V25 ^ V35 ^ V45);
sph_enc32be(out + 24, V06 ^ V16 ^ V26 ^ V36 ^ V46);
sph_enc32be(out + 28, V07 ^ V17 ^ V27 ^ V37 ^ V47);
break;
case 2:
sph_enc32be(out + 32, V00 ^ V10 ^ V20 ^ V30 ^ V40);
sph_enc32be(out + 36, V01 ^ V11 ^ V21 ^ V31 ^ V41);
sph_enc32be(out + 40, V02 ^ V12 ^ V22 ^ V32 ^ V42);
sph_enc32be(out + 44, V03 ^ V13 ^ V23 ^ V33 ^ V43);
sph_enc32be(out + 48, V04 ^ V14 ^ V24 ^ V34 ^ V44);
sph_enc32be(out + 52, V05 ^ V15 ^ V25 ^ V35 ^ V45);
sph_enc32be(out + 56, V06 ^ V16 ^ V26 ^ V36 ^ V46);
sph_enc32be(out + 60, V07 ^ V17 ^ V27 ^ V37 ^ V47);
break;
}
}
}
/* see sph_luffa.h */
void
sph_luffa224_init(void *cc)
{
sph_luffa224_context *sc;
sc = cc;
memcpy(sc->V, V_INIT, sizeof(sc->V));
sc->ptr = 0;
}
/* see sph_luffa.h */
void
sph_luffa224(void *cc, const void *data, size_t len)
{
luffa3(cc, data, len);
}
/* see sph_luffa.h */
void
sph_luffa224_close(void *cc, void *dst)
{
sph_luffa224_addbits_and_close(cc, 0, 0, dst);
}
/* see sph_luffa.h */
void
sph_luffa224_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
luffa3_close(cc, ub, n, dst, 7);
sph_luffa224_init(cc);
}
/* see sph_luffa.h */
void
sph_luffa256_init(void *cc)
{
sph_luffa256_context *sc;
sc = cc;
memcpy(sc->V, V_INIT, sizeof(sc->V));
sc->ptr = 0;
}
/* see sph_luffa.h */
void
sph_luffa256(void *cc, const void *data, size_t len)
{
luffa3(cc, data, len);
}
/* see sph_luffa.h */
void
sph_luffa256_close(void *cc, void *dst)
{
sph_luffa256_addbits_and_close(cc, 0, 0, dst);
}
/* see sph_luffa.h */
void
sph_luffa256_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
luffa3_close(cc, ub, n, dst, 8);
sph_luffa256_init(cc);
}
/* see sph_luffa.h */
void
sph_luffa384_init(void *cc)
{
sph_luffa384_context *sc;
sc = cc;
memcpy(sc->V, V_INIT, sizeof(sc->V));
sc->ptr = 0;
}
/* see sph_luffa.h */
void
sph_luffa384(void *cc, const void *data, size_t len)
{
luffa4(cc, data, len);
}
/* see sph_luffa.h */
void
sph_luffa384_close(void *cc, void *dst)
{
sph_luffa384_addbits_and_close(cc, 0, 0, dst);
}
/* see sph_luffa.h */
void
sph_luffa384_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
luffa4_close(cc, ub, n, dst);
sph_luffa384_init(cc);
}
/* see sph_luffa.h */
void
sph_luffa512_init(void *cc)
{
sph_luffa512_context *sc;
sc = cc;
memcpy(sc->V, V_INIT, sizeof(sc->V));
sc->ptr = 0;
}
/* see sph_luffa.h */
void
sph_luffa512(void *cc, const void *data, size_t len)
{
luffa5(cc, data, len);
}
/* see sph_luffa.h */
void
sph_luffa512_close(void *cc, void *dst)
{
sph_luffa512_addbits_and_close(cc, 0, 0, dst);
}
/* see sph_luffa.h */
void
sph_luffa512_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
luffa5_close(cc, ub, n, dst);
sph_luffa512_init(cc);
}
#ifdef __cplusplus
}
#endif