pqcrypto/crypto_kem/ledakemlt12/clean/gf2x_arith.c

#include "gf2x_arith.h"
#include <string.h>  // memset(...)
#include <assert.h>

/* allows the second operand to be shorter than the first */
/* the result should be as large as the first operand*/
static inline void gf2x_add_asymm(const size_t nr, DIGIT Res[],
                                  const size_t na, const DIGIT A[],
                                  const size_t nb, const DIGIT B[]) {
    assert(nr >= na && na >= nb);
    size_t i;
    size_t delta = na - nb;
    for (i = 0; i < delta; i++) {
        Res[i] = A[i];
    }
    for (i = 0; i < nb; i++) {
        Res[i + delta] = A[i + delta] ^ B[i];
    }
}

/* PRE: MAX ALLOWED ROTATION AMOUNT : DIGIT_SIZE_b */
void PQCLEAN_LEDAKEMLT12_CLEAN_right_bit_shift_n(const int length, DIGIT in[], const int amount) {
    assert(amount < DIGIT_SIZE_b);
    if ( amount == 0 ) {
        return;
    }
    int j;
    DIGIT mask;
    mask = ((DIGIT)0x01 << amount) - 1;
    for (j = length - 1; j > 0 ; j--) {
        in[j] >>= amount;
        in[j] |=  (in[j - 1] & mask) << (DIGIT_SIZE_b - amount);
    }
    in[j] >>= amount;
}

/* PRE: MAX ALLOWED ROTATION AMOUNT : DIGIT_SIZE_b */
void PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(const int length, DIGIT in[], const int amount) {
    assert(amount < DIGIT_SIZE_b);
    if ( amount == 0 ) {
        return;
    }
    int j;
    DIGIT mask;
    mask = ~(((DIGIT)0x01 << (DIGIT_SIZE_b - amount)) - 1);
    for (j = 0 ; j < length - 1 ; j++) {
        in[j] <<= amount;
        in[j] |=  (in[j + 1] & mask) >> (DIGIT_SIZE_b - amount);
    }
    in[j] <<= amount;
}


static void PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_comb(const int nr, DIGIT Res[],
        const int na, const DIGIT A[],
        const int nb, const DIGIT B[]) {
    int i, j, k;
    DIGIT u, h;

    memset(Res, 0x00, nr * sizeof(DIGIT));

    for (k = DIGIT_SIZE_b - 1; k > 0; k--) {
        for (i = na - 1; i >= 0; i--)
            if ( A[i] & (((DIGIT)0x1) << k) )
                for (j = nb - 1; j >= 0; j--) {
                    Res[i + j + 1] ^= B[j];
                }


        u = Res[na + nb - 1];
        Res[na + nb - 1] = u << 0x1;
        for (j = 1; j < na + nb; ++j) {
            h = u >> (DIGIT_SIZE_b - 1);
            u = Res[na + nb - 1 - j];
            Res[na + nb - 1 - j] = h ^ (u << 0x1);
        }
    }
    for (i = na - 1; i >= 0; i--)
        if ( A[i] & ((DIGIT)0x1) )
            for (j = nb - 1; j >= 0; j--) {
                Res[i + j + 1] ^= B[j];
            }
}

static inline void gf2x_exact_div_x_plus_one(const int na, DIGIT A[]) {
    DIGIT t = 0;
    for (int i = na - 1; i >= 0; i--) {

        t ^= A[i];

        for (int j = 1; j <= DIGIT_SIZE_b / 2; j = j * 2) {
            t ^= t << (unsigned) j;
        }
        A[i] = t;
        t >>= DIGIT_SIZE_b - 1;
    }
}

#define MIN_KAR_DIGITS 20

static void PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(const int nr, DIGIT Res[],
        const int na, const DIGIT A[],
        const int nb, const DIGIT B[]) {

    if (na < MIN_KAR_DIGITS || nb < MIN_KAR_DIGITS) {
        /* fall back to schoolbook */
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_comb(nr, Res, na, A, nb, B);
        return;
    }

    if (na % 2 == 0) {
        unsigned bih = na / 2;
        DIGIT middle[2 * bih], sumA[bih], sumB[bih];
        gf2x_add(sumA, A, A + bih, bih);
        gf2x_add(sumB, B, B + bih, bih);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * bih, middle,
                                               bih, sumA,
                                               bih, sumB);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * bih, Res + 2 * bih,
                                               bih, A + bih,
                                               bih, B + bih);
        gf2x_add(middle, middle, Res + 2 * bih, 2 * bih);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * bih, Res,
                                               bih, A,
                                               bih, B);
        gf2x_add(middle, middle, Res, 2 * bih);
        gf2x_add(Res + bih, Res + bih, middle, 2 * bih);
    } else {
        unsigned bih = na / 2 + 1;
        DIGIT middle[2 * bih], sumA[bih], sumB[bih];
        gf2x_add_asymm(bih,  sumA,
                       bih,  A + bih - 1,
                       bih - 1, A);
        gf2x_add_asymm(bih,  sumB,
                       bih,  B + bih - 1,
                       bih - 1, B);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * bih, middle,
                                               bih, sumA,
                                               bih, sumB);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * bih, Res + 2 * (bih - 1),
                                               bih, A + bih - 1,
                                               bih, B + bih - 1);
        gf2x_add(middle, middle, Res + 2 * (bih - 1), 2 * bih);
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(2 * (bih - 1), Res,
                                               (bih - 1), A,
                                               (bih - 1), B);
        gf2x_add_asymm(2 * bih, middle,
                       2 * bih, middle,
                       2 * (bih - 1), Res);
        gf2x_add(Res + bih - 2, Res + bih - 2, middle, 2 * bih);
    }
}

#define MIN_TOOM_DIGITS 35

void PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(const int nr, DIGIT Res[],
        const int na, const DIGIT A[],
        const int nb, const DIGIT B[]) {

    if (na < MIN_TOOM_DIGITS || nb < MIN_TOOM_DIGITS) {
        /* fall back to Karatsuba */
        PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_Kar(nr, Res, na, A, nb, B);
        return;
    }

    unsigned int bih; //number of limbs for each part.
    if (na % 3 == 0) {
        bih = na / 3;
    } else {
        bih = na / 3 + 1;
    }

    DIGIT u2[bih], u1[bih], u0[bih];

    unsigned int leading_slack = (3 - (na) % 3) % 3;
//     printf("leading slack %d",leading_slack);
    unsigned int i;
    for (i = 0; i < leading_slack ; i++) {
        u2[i] = 0;
    }
    for (; i < bih; ++i) {
        u2[i] = A[i - leading_slack];
    }
    /* note: only u2 needs to be a copy, refactor */
    for (; i < 2 * bih; ++i) {
        u1[i - bih] = A[i - leading_slack];
    }
    for (; i < 3 * bih; ++i) {
        u0[i - 2 * bih] = A[i - leading_slack];
    }

    DIGIT v2[bih], v1[bih], v0[bih]; /* partitioned inputs */
    /* note: only v2 needs to be a copy, refactor */
    for (i = 0; i < leading_slack ; i++) {
        v2[i] = 0;
    }
    for (; i < bih; ++i) {
        v2[i] = B[i - leading_slack];
    }
    /* note , only v2 needs to be a copy */
    for (; i < 2 * bih; ++i) {
        v1[i - bih] = B[i - leading_slack];
    }
    for (; i < 3 * bih; ++i) {
        v0[i - 2 * bih] = B[i - leading_slack];
    }

    DIGIT sum_u[bih]; /*bih digit wide*/
    gf2x_add(sum_u, u0, u1, bih);
    gf2x_add(sum_u, sum_u, u2, bih);

    DIGIT sum_v[bih]; /*bih digit wide*/
    gf2x_add(sum_v, v0, v1, bih);
    gf2x_add(sum_v, sum_v, v2, bih);


    DIGIT w1[2 * bih];
    PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(2 * bih, w1,
                                           bih, sum_u,
                                           bih, sum_v);


    DIGIT u2_x2[bih + 1];
    u2_x2[0] = 0;
    memcpy(u2_x2 + 1, u2, bih * DIGIT_SIZE_B);
    PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(bih + 1, u2_x2, 2);

    DIGIT u1_x[bih + 1];
    u1_x[0] = 0;
    memcpy(u1_x + 1, u1, bih * DIGIT_SIZE_B);
    PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(bih + 1, u1_x, 1);

    DIGIT u1_x1_u2_x2[bih + 1];
    gf2x_add(u1_x1_u2_x2, u1_x, u2_x2, bih + 1);

    DIGIT temp_u_components[bih + 1];
    gf2x_add_asymm(bih + 1, temp_u_components,
                   bih + 1, u1_x1_u2_x2,
                   bih, sum_u);

    DIGIT v2_x2[bih + 1];
    v2_x2[0] = 0;
    memcpy(v2_x2 + 1, v2, bih * DIGIT_SIZE_B);
    PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(bih + 1, v2_x2, 2);

    DIGIT v1_x[bih + 1];
    v1_x[0] = 0;
    memcpy(v1_x + 1, v1, bih * DIGIT_SIZE_B);
    PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(bih + 1, v1_x, 1);

    DIGIT v1_x1_v2_x2[bih + 1];
    gf2x_add(v1_x1_v2_x2, v1_x, v2_x2, bih + 1);

    DIGIT temp_v_components[bih + 1];
    gf2x_add_asymm(bih + 1, temp_v_components,
                   bih + 1, v1_x1_v2_x2,
                   bih, sum_v);

    DIGIT w3[2 * bih + 2];
    PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(2 * bih + 2, w3,
                                           bih + 1, temp_u_components,
                                           bih + 1, temp_v_components);

    gf2x_add_asymm(bih + 1, u1_x1_u2_x2,
                   bih + 1, u1_x1_u2_x2,
                   bih, u0);
    gf2x_add_asymm(bih + 1, v1_x1_v2_x2,
                   bih + 1, v1_x1_v2_x2,
                   bih, v0);

    DIGIT w2[2 * bih + 2];
    PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(2 * bih + 2, w2,
                                           bih + 1, u1_x1_u2_x2,
                                           bih + 1, v1_x1_v2_x2);

    DIGIT w4[2 * bih];
    PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(2 * bih, w4,
                                           bih, u2,
                                           bih, v2);
    DIGIT w0[2 * bih];
    PQCLEAN_LEDAKEMLT12_CLEAN_gf2x_mul_TC3(2 * bih, w0,
                                           bih, u0,
                                           bih, v0);

    // Interpolation starts
    gf2x_add(w3, w2, w3, 2 * bih + 2);
    gf2x_add_asymm(2 * bih + 2, w2,
                   2 * bih + 2, w2,
                   2 * bih, w0);
    PQCLEAN_LEDAKEMLT12_CLEAN_right_bit_shift_n(2 * bih + 2, w2, 1);
    gf2x_add(w2, w2, w3, 2 * bih + 2);

    // w2 + (w4 * x^3+1) = w2 + w4 + w4 << 3
    DIGIT w4_x3_plus_1[2 * bih + 1];
    w4_x3_plus_1[0] = 0;
    memcpy(w4_x3_plus_1 + 1, w4, 2 * bih * DIGIT_SIZE_B);
    PQCLEAN_LEDAKEMLT12_CLEAN_left_bit_shift_n(2 * bih + 1, w4_x3_plus_1, 3);
    gf2x_add_asymm(2 * bih + 2, w2,
                   2 * bih + 2, w2,
                   2 * bih, w4);
    gf2x_add_asymm(2 * bih + 2, w2,
                   2 * bih + 2, w2,
                   2 * bih + 1, w4_x3_plus_1);

    gf2x_exact_div_x_plus_one(2 * bih + 2, w2);

    gf2x_add(w1, w1, w0, 2 * bih);
    gf2x_add_asymm(2 * bih + 2, w3,
                   2 * bih + 2, w3,
                   2 * bih, w1);

    PQCLEAN_LEDAKEMLT12_CLEAN_right_bit_shift_n(2 * bih + 2, w3, 1);
    gf2x_exact_div_x_plus_one(2 * bih + 2, w3);

    gf2x_add(w1, w1, w4, 2 * bih);

    DIGIT w1_final[2 * bih + 2];
    gf2x_add_asymm(2 * bih + 2, w1_final,
                   2 * bih + 2, w2,
                   2 * bih, w1);
    gf2x_add(w2, w2, w3, 2 * bih + 2);

    // Result recombination starts here

    memset(Res, 0, nr * DIGIT_SIZE_B);
    /* optimization: topmost slack digits should be computed, and not addedd,
     * zeroization can be avoided altogether with a proper merge of the
     * results */

    int leastSignifDigitIdx = nr - 1;
    for (int i = 0; i < 2 * bih; i++) {
        Res[leastSignifDigitIdx - i] ^= w0[2 * bih   - 1 - i];
    }
    leastSignifDigitIdx -= bih;
    for (int i = 0; i < 2 * bih + 2; i++) {
        Res[leastSignifDigitIdx - i] ^= w1_final[2 * bih + 2   - 1 - i];
    }
    leastSignifDigitIdx -= bih;
    for (int i = 0; i < 2 * bih + 2; i++) {
        Res[leastSignifDigitIdx - i] ^= w2[2 * bih + 2 - 1 - i];
    }
    leastSignifDigitIdx -= bih;
    for (int i = 0; i < 2 * bih + 2 ; i++) {
        Res[leastSignifDigitIdx - i] ^= w3[2 * bih + 2 - 1 - i];
    }
    leastSignifDigitIdx -= bih;
    for (int i = 0; i < 2 * bih && (leastSignifDigitIdx - i >= 0) ; i++) {
        Res[leastSignifDigitIdx - i] ^= w4[2 * bih   - 1 - i];
    }
}
// // Unused
// static int gf2x_cmp(const unsigned lenA, const DIGIT A[],
//                     const unsigned lenB, const DIGIT B[]) {
//
//     int i;
//     unsigned lA = lenA, lB = lenB;
//     for (i = 0; i < lenA && A[i] == 0; i++) {
//         lA--;
//     }
//     for (i = 0; i < lenB && B[i] == 0; i++) {
//         lB--;
//     }
//     if (lA < lB) {
//         return -1;
//     }
//     if (lA > lB) {
//         return +1;
//     }
//     for (i = 0; i < lA; i++) {
//         if (A[i] > B[i]) {
//             return +1;
//         }
//         if (A[i] < B[i]) {
//             return -1;
//         }
//     }
//     return 0;
//
// }