| 1 |
/* |
| 2 |
This file is for Benes network related functions |
| 3 |
|
| 4 |
For the implementation strategy, see |
| 5 |
https://eprint.iacr.org/2017/793.pdf |
| 6 |
*/ |
| 7 |
// 20221230 djb: add linker lines |
| 8 |
|
| 9 |
// linker define benes |
| 10 |
|
| 11 |
#include "util.h" |
| 12 |
#include "transpose.h" |
| 13 |
#include "params.h" |
| 14 |
#include "benes.h" |
| 15 |
|
| 16 |
/* middle layers of the benes network */ |
| 17 |
static void layer_in(uint64_t data[2][64], uint64_t * bits, int lgs) |
| 18 |
{ |
| 19 |
int i, j, s; |
| 20 |
|
| 21 |
uint64_t d; |
| 22 |
|
| 23 |
s = 1 << lgs; |
| 24 |
|
| 25 |
for (i = 0; i < 64; i += s*2) |
| 26 |
for (j = i; j < i+s; j++) |
| 27 |
{ |
| 28 |
|
| 29 |
d = (data[0][j+0] ^ data[0][j+s]); |
| 30 |
d &= (*bits++); |
| 31 |
data[0][j+0] ^= d; |
| 32 |
data[0][j+s] ^= d; |
| 33 |
|
| 34 |
d = (data[1][j+0] ^ data[1][j+s]); |
| 35 |
d &= (*bits++); |
| 36 |
data[1][j+0] ^= d; |
| 37 |
data[1][j+s] ^= d; |
| 38 |
} |
| 39 |
} |
| 40 |
|
| 41 |
/* first and last layers of the benes network */ |
| 42 |
static void layer_ex(uint64_t * data, uint64_t * bits, int lgs) |
| 43 |
{ |
| 44 |
int i, j, s; |
| 45 |
|
| 46 |
uint64_t d; |
| 47 |
|
| 48 |
s = 1 << lgs; |
| 49 |
|
| 50 |
for (i = 0; i < 128; i += s*2) |
| 51 |
for (j = i; j < i+s; j++) |
| 52 |
{ |
| 53 |
|
| 54 |
d = (data[j+0] ^ data[j+s]); |
| 55 |
d &= (*bits++); |
| 56 |
data[j+0] ^= d; |
| 57 |
data[j+s] ^= d; |
| 58 |
} |
| 59 |
} |
| 60 |
|
| 61 |
/* input: r, sequence of bits to be permuted */ |
| 62 |
/* bits, condition bits of the Benes network */ |
| 63 |
/* rev, 0 for normal application; !0 for inverse */ |
| 64 |
/* output: r, permuted bits */ |
| 65 |
void benes(vec * r, const unsigned char * bits, int rev) |
| 66 |
{ |
| 67 |
int i, iter, inc; |
| 68 |
|
| 69 |
const unsigned char *bits_ptr; |
| 70 |
|
| 71 |
uint64_t r_int_v[2][64]; |
| 72 |
uint64_t r_int_h[2][64]; |
| 73 |
uint64_t b_int_v[64]; |
| 74 |
uint64_t b_int_h[64]; |
| 75 |
|
| 76 |
// |
| 77 |
|
| 78 |
if (rev) { bits_ptr = bits + 12288; inc = -1024; } |
| 79 |
else { bits_ptr = bits; inc = 0; } |
| 80 |
|
| 81 |
for (i = 0; i < 64; i++) |
| 82 |
{ |
| 83 |
r_int_v[0][i] = r[i*2 + 0]; |
| 84 |
r_int_v[1][i] = r[i*2 + 1]; |
| 85 |
} |
| 86 |
|
| 87 |
transpose_64x64(r_int_h[0], r_int_v[0]); |
| 88 |
transpose_64x64(r_int_h[1], r_int_v[1]); |
| 89 |
|
| 90 |
for (iter = 0; iter <= 6; iter++) |
| 91 |
{ |
| 92 |
for (i = 0; i < 64; i++) |
| 93 |
{ |
| 94 |
b_int_v[i] = load8(bits_ptr); bits_ptr += 8; |
| 95 |
} |
| 96 |
|
| 97 |
bits_ptr += inc; |
| 98 |
|
| 99 |
transpose_64x64(b_int_h, b_int_v); |
| 100 |
|
| 101 |
layer_ex(r_int_h[0], b_int_h, iter); |
| 102 |
} |
| 103 |
|
| 104 |
transpose_64x64(r_int_v[0], r_int_h[0]); |
| 105 |
transpose_64x64(r_int_v[1], r_int_h[1]); |
| 106 |
|
| 107 |
for (iter = 0; iter <= 5; iter++) |
| 108 |
{ |
| 109 |
for (i = 0; i < 64; i++) { b_int_v[i] = load8(bits_ptr); bits_ptr += 8; } |
| 110 |
bits_ptr += inc; |
| 111 |
|
| 112 |
layer_in(r_int_v, b_int_v, iter); |
| 113 |
} |
| 114 |
|
| 115 |
for (iter = 4; iter >= 0; iter--) |
| 116 |
{ |
| 117 |
for (i = 0; i < 64; i++) { b_int_v[i] = load8(bits_ptr); bits_ptr += 8; } |
| 118 |
bits_ptr += inc; |
| 119 |
|
| 120 |
layer_in(r_int_v, b_int_v, iter); |
| 121 |
} |
| 122 |
|
| 123 |
transpose_64x64(r_int_h[0], r_int_v[0]); |
| 124 |
transpose_64x64(r_int_h[1], r_int_v[1]); |
| 125 |
|
| 126 |
for (iter = 6; iter >= 0; iter--) |
| 127 |
{ |
| 128 |
for (i = 0; i < 64; i++) |
| 129 |
{ |
| 130 |
b_int_v[i] = load8(bits_ptr); bits_ptr += 8; |
| 131 |
} |
| 132 |
|
| 133 |
bits_ptr += inc; |
| 134 |
|
| 135 |
transpose_64x64(b_int_h, b_int_v); |
| 136 |
|
| 137 |
layer_ex(r_int_h[0], b_int_h, iter); |
| 138 |
} |
| 139 |
|
| 140 |
transpose_64x64(r_int_v[0], r_int_h[0]); |
| 141 |
transpose_64x64(r_int_v[1], r_int_h[1]); |
| 142 |
|
| 143 |
for (i = 0; i < 64; i++) |
| 144 |
{ |
| 145 |
r[i*2+0] = r_int_v[0][i]; |
| 146 |
r[i*2+1] = r_int_v[1][i]; |
| 147 |
} |
| 148 |
} |
| 149 |
|