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
312 lines
11 KiB
C
312 lines
11 KiB
C
/*
|
|
This file is for Benes network related functions
|
|
*/
|
|
|
|
#include "benes.h"
|
|
|
|
#include "params.h"
|
|
#include "transpose.h"
|
|
#include "util.h"
|
|
|
|
static void layer_x(vec128 *data, vec128 *bits) {
|
|
int i;
|
|
vec128 v0, v1;
|
|
vec128 d;
|
|
|
|
for (i = 0; i < 64; i += 2) {
|
|
v0 = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_low(data[i + 0], data[i + 1]);
|
|
v1 = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_high(data[i + 0], data[i + 1]);
|
|
|
|
d = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(v0, v1);
|
|
d = PQCLEAN_MCELIECE460896F_AVX_vec128_and(d, *bits++);
|
|
v0 = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(v0, d);
|
|
v1 = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(v1, d);
|
|
|
|
data[i + 0] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_low(v0, v1);
|
|
data[i + 1] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_high(v0, v1);
|
|
}
|
|
}
|
|
|
|
static void layer_0(vec128 *bs, vec128 *cond) {
|
|
int x;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 2) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x ], bs[ x + 1 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, *cond++);
|
|
bs[ x ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x ], diff);
|
|
bs[ x + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 1 ], diff);
|
|
}
|
|
}
|
|
|
|
static void layer_1(vec128 *bs, vec128 *cond) {
|
|
int x;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 4) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 0 ], bs[ x + 2 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[0]);
|
|
bs[ x + 0 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 0 ], diff);
|
|
bs[ x + 2 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 2 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 1 ], bs[ x + 3 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[1]);
|
|
bs[ x + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 1 ], diff);
|
|
bs[ x + 3 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 3 ], diff);
|
|
|
|
cond += 2;
|
|
}
|
|
}
|
|
|
|
static void layer_2(vec128 *bs, vec128 *cond) {
|
|
int x;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 8) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 0 ], bs[ x + 4 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[0]);
|
|
bs[ x + 0 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 0 ], diff);
|
|
bs[ x + 4 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 4 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 1 ], bs[ x + 5 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[1]);
|
|
bs[ x + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 1 ], diff);
|
|
bs[ x + 5 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 5 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 2 ], bs[ x + 6 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[2]);
|
|
bs[ x + 2 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 2 ], diff);
|
|
bs[ x + 6 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 6 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 3 ], bs[ x + 7 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[3]);
|
|
bs[ x + 3 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 3 ], diff);
|
|
bs[ x + 7 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ x + 7 ], diff);
|
|
|
|
cond += 4;
|
|
}
|
|
}
|
|
|
|
static void layer_3(vec128 *bs, vec128 *cond) {
|
|
int x, s;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 16) {
|
|
for (s = x; s < x + 8; s += 4) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], bs[ s + 8 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[0]);
|
|
bs[ s + 0 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], diff);
|
|
bs[ s + 8 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 8 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], bs[ s + 9 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[1]);
|
|
bs[ s + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], diff);
|
|
bs[ s + 9 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 9 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], bs[ s + 10 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[2]);
|
|
bs[ s + 2 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], diff);
|
|
bs[ s + 10 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 10 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], bs[ s + 11 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[3]);
|
|
bs[ s + 3 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], diff);
|
|
bs[ s + 11 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 11 ], diff);
|
|
|
|
cond += 4;
|
|
}
|
|
}
|
|
}
|
|
|
|
static void layer_4(vec128 *bs, vec128 *cond) {
|
|
int x, s;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 32) {
|
|
for (s = x; s < x + 16; s += 4) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], bs[ s + 16 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[0]);
|
|
bs[ s + 0 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], diff);
|
|
bs[ s + 16 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 16 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], bs[ s + 17 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[1]);
|
|
bs[ s + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], diff);
|
|
bs[ s + 17 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 17 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], bs[ s + 18 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[2]);
|
|
bs[ s + 2 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], diff);
|
|
bs[ s + 18 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 18 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], bs[ s + 19 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[3]);
|
|
bs[ s + 3 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], diff);
|
|
bs[ s + 19 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 19 ], diff);
|
|
|
|
cond += 4;
|
|
}
|
|
}
|
|
}
|
|
|
|
static void layer_5(vec128 *bs, vec128 *cond) {
|
|
int x, s;
|
|
vec128 diff;
|
|
|
|
for (x = 0; x < (1 << 6); x += 64) {
|
|
for (s = x; s < x + 32; s += 4) {
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], bs[ s + 32 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[0]);
|
|
bs[ s + 0 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 0 ], diff);
|
|
bs[ s + 32 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 32 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], bs[ s + 33 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[1]);
|
|
bs[ s + 1 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 1 ], diff);
|
|
bs[ s + 33 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 33 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], bs[ s + 34 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[2]);
|
|
bs[ s + 2 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 2 ], diff);
|
|
bs[ s + 34 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 34 ], diff);
|
|
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], bs[ s + 35 ]);
|
|
diff = PQCLEAN_MCELIECE460896F_AVX_vec128_and(diff, cond[3]);
|
|
bs[ s + 3 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 3 ], diff);
|
|
bs[ s + 35 ] = PQCLEAN_MCELIECE460896F_AVX_vec128_xor(bs[ s + 35 ], diff);
|
|
|
|
cond += 4;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* input: bits, control bits as array of bytes */
|
|
/* output: bits_int, control bits as array of 128-bit vectors */
|
|
void PQCLEAN_MCELIECE460896F_AVX_load_bits(vec128 bits_int[][32], const unsigned char *bits) {
|
|
int i, j;
|
|
const unsigned char *ptr = bits;
|
|
|
|
vec128 buf[64];
|
|
|
|
for (i = 0; i <= 5; i += 2) {
|
|
for (j = 0; j < 64; j++) {
|
|
buf[j] = PQCLEAN_MCELIECE460896F_AVX_vec128_set2x(PQCLEAN_MCELIECE460896F_AVX_load8(ptr), PQCLEAN_MCELIECE460896F_AVX_load8(ptr + 512));
|
|
ptr += 8;
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( buf );
|
|
|
|
for (j = 0; j < 32; j++) {
|
|
bits_int[i + 0][j] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_low(buf[j], buf[j + 32]);
|
|
bits_int[i + 1][j] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_high(buf[j], buf[j + 32]);
|
|
}
|
|
|
|
ptr += 512;
|
|
}
|
|
|
|
for (i = 6; i <= 18; i++) {
|
|
for (j = 0; j < 32; j++) {
|
|
bits_int[i][j] = PQCLEAN_MCELIECE460896F_AVX_load16(ptr);
|
|
ptr += 16;
|
|
}
|
|
}
|
|
|
|
for (i = 19; i < 25; i += 2) {
|
|
for (j = 0; j < 64; j++) {
|
|
buf[j] = PQCLEAN_MCELIECE460896F_AVX_vec128_set2x(PQCLEAN_MCELIECE460896F_AVX_load8(ptr), PQCLEAN_MCELIECE460896F_AVX_load8(ptr + 512));
|
|
ptr += 8;
|
|
}
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( buf );
|
|
|
|
for (j = 0; j < 32; j++) {
|
|
bits_int[i + 0][j] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_low(buf[j], buf[j + 32]);
|
|
bits_int[i + 1][j] = PQCLEAN_MCELIECE460896F_AVX_vec128_unpack_high(buf[j], buf[j + 32]);
|
|
}
|
|
|
|
ptr += 512;
|
|
}
|
|
}
|
|
|
|
/* input: r, sequence of bits to be permuted */
|
|
/* b, control bits as array of 128-bit vectors */
|
|
/* rev, 0 for normal application; !0 for inverse */
|
|
/* output: r, permuted bits */
|
|
void PQCLEAN_MCELIECE460896F_AVX_benes(vec128 *r, vec128 b[][32], int rev) {
|
|
int inc;
|
|
|
|
vec128 *b_ptr;
|
|
|
|
if (rev == 0) {
|
|
inc = 32;
|
|
b_ptr = b[ 0];
|
|
} else {
|
|
inc = -32;
|
|
b_ptr = b[24];
|
|
}
|
|
|
|
//
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( r );
|
|
|
|
layer_0(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_1(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_2(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_3(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_4(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_5(r, b_ptr);
|
|
b_ptr += inc;
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( r );
|
|
|
|
layer_x(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_0(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_1(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_2(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_3(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_4(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_5(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_4(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_3(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_2(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_1(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_0(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_x(r, b_ptr);
|
|
b_ptr += inc;
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( r );
|
|
|
|
layer_5(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_4(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_3(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_2(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_1(r, b_ptr);
|
|
b_ptr += inc;
|
|
layer_0(r, b_ptr);
|
|
//b_ptr += inc;
|
|
|
|
PQCLEAN_MCELIECE460896F_AVX_transpose_64x128_sp( r );
|
|
}
|
|
|