pqc/crypto_kem/mceliece6688128/avx/decrypt.c
Thom Wiggers ac2c20045c Classic McEliece (#259)
* 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
2021-03-24 21:02:45 +00:00

235 lines
6.1 KiB
C

/*
This file is for Niederreiter decryption
*/
#include "decrypt.h"
#include "benes.h"
#include "bm.h"
#include "fft.h"
#include "fft_tr.h"
#include "params.h"
#include "util.h"
#include <stdio.h>
static void scaling(vec256 out[][GFBITS], vec256 inv[][GFBITS], const unsigned char *sk, vec256 *recv) {
int i, j;
vec128 sk_int[ GFBITS ];
vec256 eval[32][ GFBITS ];
vec256 tmp[ GFBITS ];
// computing inverses
PQCLEAN_MCELIECE6688128_AVX_irr_load(sk_int, sk);
PQCLEAN_MCELIECE6688128_AVX_fft(eval, sk_int);
for (i = 0; i < 32; i++) {
PQCLEAN_MCELIECE6688128_AVX_vec256_sq(eval[i], eval[i]);
}
PQCLEAN_MCELIECE6688128_AVX_vec256_copy(inv[0], eval[0]);
for (i = 1; i < 32; i++) {
PQCLEAN_MCELIECE6688128_AVX_vec256_mul(inv[i], inv[i - 1], eval[i]);
}
PQCLEAN_MCELIECE6688128_AVX_vec256_inv(tmp, inv[31]);
for (i = 30; i >= 0; i--) {
PQCLEAN_MCELIECE6688128_AVX_vec256_mul(inv[i + 1], tmp, inv[i]);
PQCLEAN_MCELIECE6688128_AVX_vec256_mul(tmp, tmp, eval[i + 1]);
}
PQCLEAN_MCELIECE6688128_AVX_vec256_copy(inv[0], tmp);
//
for (i = 0; i < 32; i++) {
for (j = 0; j < GFBITS; j++) {
out[i][j] = PQCLEAN_MCELIECE6688128_AVX_vec256_and(inv[i][j], recv[i]);
}
}
}
static void preprocess(vec128 *recv, const unsigned char *s) {
int i;
unsigned char r[ 1024 ];
for (i = 0; i < SYND_BYTES; i++) {
r[i] = s[i];
}
for (i = SYND_BYTES; i < 1024; i++) {
r[i] = 0;
}
for (i = 0; i < 64; i++) {
recv[i] = PQCLEAN_MCELIECE6688128_AVX_load16(r + i * 16);
}
}
static void postprocess(unsigned char *e, vec128 *err) {
int i;
unsigned char error8[ (1 << GFBITS) / 8 ];
uint64_t v[2];
for (i = 0; i < 64; i++) {
v[0] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(err[i], 0);
v[1] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(err[i], 1);
PQCLEAN_MCELIECE6688128_AVX_store8(error8 + i * 16 + 0, v[0]);
PQCLEAN_MCELIECE6688128_AVX_store8(error8 + i * 16 + 8, v[1]);
}
for (i = 0; i < SYS_N / 8; i++) {
e[i] = error8[i];
}
}
static void scaling_inv(vec256 out[][GFBITS], vec256 inv[][GFBITS], vec256 *recv) {
int i, j;
for (i = 0; i < 32; i++) {
for (j = 0; j < GFBITS; j++) {
out[i][j] = PQCLEAN_MCELIECE6688128_AVX_vec256_and(inv[i][j], recv[i]);
}
}
}
static uint16_t weight_check(unsigned char *e, vec128 *error) {
int i;
uint16_t w0 = 0;
uint16_t w1 = 0;
uint16_t check;
for (i = 0; i < 64; i++) {
w0 += _mm_popcnt_u64( PQCLEAN_MCELIECE6688128_AVX_vec128_extract(error[i], 0) );
w0 += _mm_popcnt_u64( PQCLEAN_MCELIECE6688128_AVX_vec128_extract(error[i], 1) );
}
for (i = 0; i < SYS_N / 8; i++) {
w1 += _mm_popcnt_u64( e[i] );
}
check = (w0 ^ SYS_T) | (w1 ^ SYS_T);
check -= 1;
check >>= 15;
return check;
}
static uint16_t synd_cmp(vec256 *s0, vec256 *s1) {
int i;
vec256 diff;
diff = PQCLEAN_MCELIECE6688128_AVX_vec256_xor(s0[0], s1[0]);
for (i = 1; i < GFBITS; i++) {
diff = PQCLEAN_MCELIECE6688128_AVX_vec256_or(diff, PQCLEAN_MCELIECE6688128_AVX_vec256_xor(s0[i], s1[i]));
}
return (uint16_t)PQCLEAN_MCELIECE6688128_AVX_vec256_testz(diff);
}
static void reformat_128to256(vec256 *out, vec128 *in) {
int i;
uint64_t v[4];
for (i = 0; i < 32; i++) {
v[0] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(in[2 * i + 0], 0);
v[1] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(in[2 * i + 0], 1);
v[2] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(in[2 * i + 1], 0);
v[3] = PQCLEAN_MCELIECE6688128_AVX_vec128_extract(in[2 * i + 1], 1);
out[i] = PQCLEAN_MCELIECE6688128_AVX_vec256_set4x(v[0], v[1], v[2], v[3]);
}
}
static void reformat_256to128(vec128 *out, vec256 *in) {
int i;
uint64_t v[4];
for (i = 0; i < 32; i++) {
v[0] = PQCLEAN_MCELIECE6688128_AVX_vec256_extract(in[i], 0);
v[1] = PQCLEAN_MCELIECE6688128_AVX_vec256_extract(in[i], 1);
v[2] = PQCLEAN_MCELIECE6688128_AVX_vec256_extract(in[i], 2);
v[3] = PQCLEAN_MCELIECE6688128_AVX_vec256_extract(in[i], 3);
out[2 * i + 0] = PQCLEAN_MCELIECE6688128_AVX_vec128_set2x(v[0], v[1]);
out[2 * i + 1] = PQCLEAN_MCELIECE6688128_AVX_vec128_set2x(v[2], v[3]);
}
}
/* Niederreiter decryption with the Berlekamp decoder */
/* intput: sk, secret key */
/* c, ciphertext (syndrome) */
/* output: e, error vector */
/* return: 0 for success; 1 for failure */
int PQCLEAN_MCELIECE6688128_AVX_decrypt(unsigned char *e, const unsigned char *sk, const unsigned char *c) {
int i;
uint16_t check_synd;
uint16_t check_weight;
vec256 inv[ 32 ][ GFBITS ];
vec256 scaled[ 32 ][ GFBITS ];
vec256 eval[32][ GFBITS ];
vec128 error128[ 64 ];
vec256 error256[ 32 ];
vec256 s_priv[ GFBITS ];
vec256 s_priv_cmp[ GFBITS ];
vec128 locator[ GFBITS ];
vec128 recv128[ 64 ];
vec256 recv256[ 32 ];
vec256 allone;
vec128 bits_int[25][32];
// Berlekamp decoder
preprocess(recv128, c);
PQCLEAN_MCELIECE6688128_AVX_load_bits(bits_int, sk + IRR_BYTES);
PQCLEAN_MCELIECE6688128_AVX_benes(recv128, bits_int, 1);
reformat_128to256(recv256, recv128);
scaling(scaled, inv, sk, recv256);
PQCLEAN_MCELIECE6688128_AVX_fft_tr(s_priv, scaled);
PQCLEAN_MCELIECE6688128_AVX_bm(locator, s_priv);
PQCLEAN_MCELIECE6688128_AVX_fft(eval, locator);
// reencryption and weight check
allone = PQCLEAN_MCELIECE6688128_AVX_vec256_set1_16b(0xFFFF);
for (i = 0; i < 32; i++) {
error256[i] = PQCLEAN_MCELIECE6688128_AVX_vec256_or_reduce(eval[i]);
error256[i] = PQCLEAN_MCELIECE6688128_AVX_vec256_xor(error256[i], allone);
}
scaling_inv(scaled, inv, error256);
PQCLEAN_MCELIECE6688128_AVX_fft_tr(s_priv_cmp, scaled);
check_synd = synd_cmp(s_priv, s_priv_cmp);
//
reformat_256to128(error128, error256);
PQCLEAN_MCELIECE6688128_AVX_benes(error128, bits_int, 0);
postprocess(e, error128);
check_weight = weight_check(e, error128);
return 1 - (check_synd & check_weight);
}