mirror of
https://github.com/henrydcase/pqc.git
synced 2024-11-30 03:11:43 +00:00
ac2c20045c
* Add McEliece reference implementations * Add Vec implementations of McEliece * Add sse implementations * Add AVX2 implementations * Get rid of stuff not supported by Mac ABI * restrict to two cores * Ditch .data files * Remove .hidden from all .S files * speed up duplicate consistency tests by batching * make cpuinfo more robust * Hope to stabilize macos cpuinfo without ccache * Revert "Hope to stabilize macos cpuinfo without ccache" This reverts commit 6129c3cabe1abbc8b956bc87e902a698e32bf322. * Just hardcode what's available at travis * Fixed-size types in api.h * namespace all header files in mceliece * Ditch operations.h * Get rid of static inline functions * fixup! Ditch operations.h
181 lines
4.7 KiB
C
181 lines
4.7 KiB
C
/*
|
|
This file is for Benes network related functions
|
|
*/
|
|
|
|
#include "benes.h"
|
|
#include "params.h"
|
|
#include "transpose.h"
|
|
#include "util.h"
|
|
|
|
/* middle layers of the benes network */
|
|
static void layer_in(uint64_t data[2][64], uint64_t *bits, int lgs) {
|
|
int i, j, s;
|
|
|
|
uint64_t d;
|
|
|
|
s = 1 << lgs;
|
|
|
|
for (i = 0; i < 64; i += s * 2) {
|
|
for (j = i; j < i + s; j++) {
|
|
|
|
d = (data[0][j + 0] ^ data[0][j + s]);
|
|
d &= (*bits++);
|
|
data[0][j + 0] ^= d;
|
|
data[0][j + s] ^= d;
|
|
|
|
d = (data[1][j + 0] ^ data[1][j + s]);
|
|
d &= (*bits++);
|
|
data[1][j + 0] ^= d;
|
|
data[1][j + s] ^= d;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* first and last layers of the benes network */
|
|
static void layer_ex(uint64_t *data, uint64_t *bits, int lgs) {
|
|
int i, j, s;
|
|
|
|
uint64_t d;
|
|
|
|
s = 1 << lgs;
|
|
|
|
for (i = 0; i < 128; i += s * 2) {
|
|
for (j = i; j < i + s; j++) {
|
|
|
|
d = (data[j + 0] ^ data[j + s]);
|
|
d &= (*bits++);
|
|
data[j + 0] ^= d;
|
|
data[j + s] ^= d;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* input: r, sequence of bits to be permuted */
|
|
/* bits, condition bits of the Benes network */
|
|
/* rev, 0 for normal application; !0 for inverse */
|
|
/* output: r, permuted bits */
|
|
void PQCLEAN_MCELIECE460896F_CLEAN_apply_benes(unsigned char *r, const unsigned char *bits, int rev) {
|
|
int i, iter, inc;
|
|
|
|
unsigned char *r_ptr = r;
|
|
const unsigned char *bits_ptr;
|
|
|
|
uint64_t r_int_v[2][64];
|
|
uint64_t r_int_h[2][64];
|
|
uint64_t b_int_v[64];
|
|
uint64_t b_int_h[64];
|
|
|
|
//
|
|
|
|
if (rev) {
|
|
bits_ptr = bits + 12288;
|
|
inc = -1024;
|
|
} else {
|
|
bits_ptr = bits;
|
|
inc = 0;
|
|
}
|
|
|
|
for (i = 0; i < 64; i++) {
|
|
r_int_v[0][i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(r_ptr + i * 16 + 0);
|
|
r_int_v[1][i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(r_ptr + i * 16 + 8);
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_h[0], r_int_v[0]);
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_h[1], r_int_v[1]);
|
|
|
|
for (iter = 0; iter <= 6; iter++) {
|
|
for (i = 0; i < 64; i++) {
|
|
b_int_v[i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(bits_ptr);
|
|
bits_ptr += 8;
|
|
}
|
|
|
|
bits_ptr += inc;
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(b_int_h, b_int_v);
|
|
|
|
layer_ex(r_int_h[0], b_int_h, iter);
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_v[0], r_int_h[0]);
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_v[1], r_int_h[1]);
|
|
|
|
for (iter = 0; iter <= 5; iter++) {
|
|
for (i = 0; i < 64; i++) {
|
|
b_int_v[i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(bits_ptr);
|
|
bits_ptr += 8;
|
|
}
|
|
bits_ptr += inc;
|
|
|
|
layer_in(r_int_v, b_int_v, iter);
|
|
}
|
|
|
|
for (iter = 4; iter >= 0; iter--) {
|
|
for (i = 0; i < 64; i++) {
|
|
b_int_v[i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(bits_ptr);
|
|
bits_ptr += 8;
|
|
}
|
|
bits_ptr += inc;
|
|
|
|
layer_in(r_int_v, b_int_v, iter);
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_h[0], r_int_v[0]);
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_h[1], r_int_v[1]);
|
|
|
|
for (iter = 6; iter >= 0; iter--) {
|
|
for (i = 0; i < 64; i++) {
|
|
b_int_v[i] = PQCLEAN_MCELIECE460896F_CLEAN_load8(bits_ptr);
|
|
bits_ptr += 8;
|
|
}
|
|
|
|
bits_ptr += inc;
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(b_int_h, b_int_v);
|
|
|
|
layer_ex(r_int_h[0], b_int_h, iter);
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_v[0], r_int_h[0]);
|
|
PQCLEAN_MCELIECE460896F_CLEAN_transpose_64x64(r_int_v[1], r_int_h[1]);
|
|
|
|
for (i = 0; i < 64; i++) {
|
|
PQCLEAN_MCELIECE460896F_CLEAN_store8(r_ptr + i * 16 + 0, r_int_v[0][i]);
|
|
PQCLEAN_MCELIECE460896F_CLEAN_store8(r_ptr + i * 16 + 8, r_int_v[1][i]);
|
|
}
|
|
}
|
|
|
|
/* input: condition bits c */
|
|
/* output: support s */
|
|
void PQCLEAN_MCELIECE460896F_CLEAN_support_gen(gf *s, const unsigned char *c) {
|
|
gf a;
|
|
int i, j;
|
|
unsigned char L[ GFBITS ][ (1 << GFBITS) / 8 ];
|
|
|
|
for (i = 0; i < GFBITS; i++) {
|
|
for (j = 0; j < (1 << GFBITS) / 8; j++) {
|
|
L[i][j] = 0;
|
|
}
|
|
}
|
|
|
|
for (i = 0; i < (1 << GFBITS); i++) {
|
|
a = PQCLEAN_MCELIECE460896F_CLEAN_bitrev((gf) i);
|
|
|
|
for (j = 0; j < GFBITS; j++) {
|
|
L[j][ i / 8 ] |= ((a >> j) & 1) << (i % 8);
|
|
}
|
|
}
|
|
|
|
for (j = 0; j < GFBITS; j++) {
|
|
PQCLEAN_MCELIECE460896F_CLEAN_apply_benes(L[j], c, 0);
|
|
}
|
|
|
|
for (i = 0; i < SYS_N; i++) {
|
|
s[i] = 0;
|
|
for (j = GFBITS - 1; j >= 0; j--) {
|
|
s[i] <<= 1;
|
|
s[i] |= (L[j][i / 8] >> (i % 8)) & 1;
|
|
}
|
|
}
|
|
}
|
|
|