-rw-r--r-- 3964 libmceliece-20240513/crypto_kem/348864/avx/fft.c raw
/*
This file is for implementing the Gao-Mateer FFT, see
http://www.math.clemson.edu/~sgao/papers/GM10.pdf
For the implementation strategy, see
https://eprint.iacr.org/2017/793.pdf
*/
// 20240508 djb: include vec256_gf.h
// 20221230 djb: split these arrays into separate .c files
// 20221230 djb: rename powers array as fft_powers
// 20221230 djb: rename consts array as fft_consts
// 20221230 djb: rename s array as fft_scalars
// 20221230 djb: add linker lines
// linker define fft
// linker use vec_mul_asm
// linker use vec256_mul_asm
// linker use fft_scalars fft_consts fft_powers
#include "fft.h"
#include "fft_scalars.h"
#include "fft_consts.h"
#include "fft_powers.h"
#include "vec256_gf.h"
#include "vec.h"
/* input: in, polynomial in bitsliced form */
/* output: in, result of applying the radix conversions on in */
static void radix_conversions(uint64_t *in)
{
int i, j, k;
const uint64_t mask[5][2] =
{
{0x8888888888888888, 0x4444444444444444},
{0xC0C0C0C0C0C0C0C0, 0x3030303030303030},
{0xF000F000F000F000, 0x0F000F000F000F00},
{0xFF000000FF000000, 0x00FF000000FF0000},
{0xFFFF000000000000, 0x0000FFFF00000000}
};
//
for (j = 0; j <= 4; j++)
{
for (i = 0; i < GFBITS; i++)
for (k = 4; k >= j; k--)
{
in[i] ^= (in[i] & mask[k][0]) >> (1 << k);
in[i] ^= (in[i] & mask[k][1]) >> (1 << k);
}
vec_mul(in, in, fft_scalars[j]); // scaling
}
}
/* input: in, result of applying the radix conversions to the input polynomial */
/* output: out, evaluation results (by applying the FFT butterflies) */
static void butterflies(vec256 out[][ GFBITS ], uint64_t *in)
{
int i, j, k, s, b;
uint64_t t0, t1, t2, t3;
uint64_t consts_ptr = 0;
const unsigned char reversal[64] =
{
0, 32, 16, 48, 8, 40, 24, 56,
4, 36, 20, 52, 12, 44, 28, 60,
2, 34, 18, 50, 10, 42, 26, 58,
6, 38, 22, 54, 14, 46, 30, 62,
1, 33, 17, 49, 9, 41, 25, 57,
5, 37, 21, 53, 13, 45, 29, 61,
3, 35, 19, 51, 11, 43, 27, 59,
7, 39, 23, 55, 15, 47, 31, 63
};
// boradcast
vec256 tmp256[ GFBITS ];
vec256 x[ GFBITS ], y[ GFBITS ];
for (j = 0; j < 64; j+=8)
for (i = 0; i < GFBITS; i++)
{
t0 = (in[i] >> reversal[j+0]) & 1; t0 = -t0;
t1 = (in[i] >> reversal[j+2]) & 1; t1 = -t1;
t2 = (in[i] >> reversal[j+4]) & 1; t2 = -t2;
t3 = (in[i] >> reversal[j+6]) & 1; t3 = -t3;
out[j/4+0][i] = vec256_set4x(t0, t1, t2, t3);
t0 = (in[i] >> reversal[j+1]) & 1; t0 = -t0;
t1 = (in[i] >> reversal[j+3]) & 1; t1 = -t1;
t2 = (in[i] >> reversal[j+5]) & 1; t2 = -t2;
t3 = (in[i] >> reversal[j+7]) & 1; t3 = -t3;
out[j/4+1][i] = vec256_set4x(t0, t1, t2, t3);
}
//
for (i = 0; i < 16; i+=2)
{
vec256_mul(tmp256, out[i+1], fft_consts[ 0 ]);
for (b = 0; b < GFBITS; b++) out[i+0][b] ^= tmp256[b];
for (b = 0; b < GFBITS; b++) out[i+1][b] ^= out[i+0][b];
}
for (i = 0; i < 16; i+=2)
{
for (b = 0; b < GFBITS; b++) x[b] = vec256_unpack_low_2x(out[i+0][b], out[i+1][b]);
for (b = 0; b < GFBITS; b++) y[b] = vec256_unpack_high_2x(out[i+0][b], out[i+1][b]);
vec256_mul(tmp256, y, fft_consts[ 1 ]);
for (b = 0; b < GFBITS; b++) x[b] ^= tmp256[b];
for (b = 0; b < GFBITS; b++) y[b] ^= x[b];
for (b = 0; b < GFBITS; b++) out[i+0][b] = vec256_unpack_low(x[b], y[b]);
for (b = 0; b < GFBITS; b++) out[i+1][b] = vec256_unpack_high(x[b], y[b]);
}
consts_ptr = 2;
for (i = 0; i <= 3; i++)
{
s = 1 << i;
for (j = 0; j < 16; j += 2*s)
{
for (k = j; k < j+s; k++)
{
vec256_mul(tmp256, out[k+s], fft_consts[ consts_ptr + (k-j) ]);
for (b = 0; b < GFBITS; b++) out[k][b] ^= tmp256[b];
for (b = 0; b < GFBITS; b++) out[k+s][b] ^= out[k][b];
}
}
consts_ptr += s;
}
// adding the part contributed by x^64
for (i = 0; i < 16; i++)
for (b = 0; b < GFBITS; b++)
out[i][b] = vec256_xor(out[i][b], fft_powers[i][b]);
}
void fft(vec256 out[][ GFBITS ], uint64_t *in)
{
radix_conversions(in);
butterflies(out, in);
}