kris/pqc
1
1
Fork 0
程式碼 問題 合併請求 版本發布 Wiki 動態

ms compiler changes for fft.c

瀏覽代碼
This commit is contained in:
John M. Schanck 2020-09-14 13:04:38 -04:00 提交者 Kris Kwiatkowski
父節點 262e64168f
當前提交 2f05de259d
共有 12 個檔案被更改，包括 96 行新增 和 105 行删除
顯示統計資料 下載 Patch 檔 下載差異檔 展開所有檔案 折疊所有檔案

15
crypto_kem/hqc-128/avx2/fft.c
查看文件

@ -31,7 +31,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,7 +134,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -202,7 +202,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC128_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC128_AVX2_gf_mul(beta_m_pow, f[i]);
@ -224,7 +224,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -300,7 +300,7 @@ void PQCLEAN_HQC128_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -329,15 +329,14 @@ void PQCLEAN_HQC128_AVX2_fft_retrieve_bch_error_poly(uint64_t *error, const uint
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

28
crypto_kem/hqc-128/clean/fft.c
查看文件

@ -34,7 +34,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,10 +134,9 @@ static void radix_t_big(uint16_t *f, const uint16_t *f0, const uint16_t *f1, uin
uint16_t Q[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t R[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t n;
size_t i;
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q0, f0 + n, 2 * n);
memcpy(Q1, f1 + n, 2 * n);
memcpy(R0, f0, 2 * n);
@ -187,7 +186,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 1
if (m_f == 1) {
f[0] = 0;
x = 1 << m;
x = (size_t) (1 << m);
for (i = 0; i < x; ++i) {
f[0] ^= w[i];
}
@ -221,7 +220,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case
// Step 5: Compute f0 from u and f1 from v
f1[1] = 0;
@ -252,7 +251,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 2: compute f from g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC128_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC128_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -297,7 +296,7 @@ void PQCLEAN_HQC128_CLEAN_fft_t(uint16_t *f, const uint16_t *w, size_t f_coeffs)
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
u[0] = w[0] ^ w[k];
v[0] = w[k];
for (i = 1; i < k; ++i) {
@ -396,7 +395,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -464,7 +463,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC128_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC128_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -486,7 +485,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -562,7 +561,7 @@ void PQCLEAN_HQC128_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -637,15 +636,14 @@ void PQCLEAN_HQC128_CLEAN_fft_retrieve_bch_error_poly(uint64_t *error, const uin
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

15
crypto_kem/hqc-192/avx2/fft.c
查看文件

@ -31,7 +31,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,7 +134,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -202,7 +202,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC192_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC192_AVX2_gf_mul(beta_m_pow, f[i]);
@ -224,7 +224,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -300,7 +300,7 @@ void PQCLEAN_HQC192_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -329,15 +329,14 @@ void PQCLEAN_HQC192_AVX2_fft_retrieve_bch_error_poly(uint64_t *error, const uint
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

28
crypto_kem/hqc-192/clean/fft.c
查看文件

@ -34,7 +34,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,10 +134,9 @@ static void radix_t_big(uint16_t *f, const uint16_t *f0, const uint16_t *f1, uin
uint16_t Q[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t R[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t n;
size_t i;
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q0, f0 + n, 2 * n);
memcpy(Q1, f1 + n, 2 * n);
memcpy(R0, f0, 2 * n);
@ -187,7 +186,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 1
if (m_f == 1) {
f[0] = 0;
x = 1 << m;
x = (size_t) (1 << m);
for (i = 0; i < x; ++i) {
f[0] ^= w[i];
}
@ -221,7 +220,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case
// Step 5: Compute f0 from u and f1 from v
f1[1] = 0;
@ -252,7 +251,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 2: compute f from g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC192_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC192_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -297,7 +296,7 @@ void PQCLEAN_HQC192_CLEAN_fft_t(uint16_t *f, const uint16_t *w, size_t f_coeffs)
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
u[0] = w[0] ^ w[k];
v[0] = w[k];
for (i = 1; i < k; ++i) {
@ -396,7 +395,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -464,7 +463,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC192_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC192_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -486,7 +485,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -562,7 +561,7 @@ void PQCLEAN_HQC192_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -637,15 +636,14 @@ void PQCLEAN_HQC192_CLEAN_fft_retrieve_bch_error_poly(uint64_t *error, const uin
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

15
crypto_kem/hqc-256/avx2/fft.c
查看文件

@ -31,7 +31,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,7 +134,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -202,7 +202,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC256_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC256_AVX2_gf_mul(beta_m_pow, f[i]);
@ -224,7 +224,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -300,7 +300,7 @@ void PQCLEAN_HQC256_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -329,15 +329,14 @@ void PQCLEAN_HQC256_AVX2_fft_retrieve_bch_error_poly(uint64_t *error, const uint
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

28
crypto_kem/hqc-256/clean/fft.c
查看文件

@ -34,7 +34,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -134,10 +134,9 @@ static void radix_t_big(uint16_t *f, const uint16_t *f0, const uint16_t *f1, uin
uint16_t Q[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t R[1 << 2 * (PARAM_FFT_T - 2)] = {0};
uint16_t n;
size_t i;
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q0, f0 + n, 2 * n);
memcpy(Q1, f1 + n, 2 * n);
memcpy(R0, f0, 2 * n);
@ -187,7 +186,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 1
if (m_f == 1) {
f[0] = 0;
x = 1 << m;
x = (size_t) (1 << m);
for (i = 0; i < x; ++i) {
f[0] ^= w[i];
}
@ -221,7 +220,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case
// Step 5: Compute f0 from u and f1 from v
f1[1] = 0;
@ -252,7 +251,7 @@ static void fft_t_rec(uint16_t *f, const uint16_t *w, size_t f_coeffs, uint8_t m
// Step 2: compute f from g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC256_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC256_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -297,7 +296,7 @@ void PQCLEAN_HQC256_CLEAN_fft_t(uint16_t *f, const uint16_t *w, size_t f_coeffs)
* Transpose:
* u[i] = w[i] + w[k+i]
* v[i] = G[i].w[i] + (G[i]+1).w[k+i] = G[i].u[i] + w[k+i] */
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
u[0] = w[0] ^ w[k];
v[0] = w[k];
for (i = 1; i < k; ++i) {
@ -396,7 +395,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -464,7 +463,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQC256_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQC256_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -486,7 +485,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -562,7 +561,7 @@ void PQCLEAN_HQC256_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs) {
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -637,15 +636,14 @@ void PQCLEAN_HQC256_CLEAN_fft_retrieve_bch_error_poly(uint64_t *error, const uin
uint16_t gammas[PARAM_M - 1] = {0};
uint16_t gammas_sums[1 << (PARAM_M - 1)] = {0};
uint64_t bit;
uint16_t k;
size_t i, index;
size_t i, k, index;
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
index = PARAM_GF_MUL_ORDER;
bit = 1 ^ ((uint16_t) - w[k] >> 15);
error[index / 8] ^= bit << (index % 64);

12
crypto_kem/hqc-rmrs-128/avx2/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS128_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS128_AVX2_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS128_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS128_AVX2_fft_retrieve_error_poly(uint8_t *error, const uint1
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);

12
crypto_kem/hqc-rmrs-128/clean/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS128_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS128_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS128_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeff
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS128_CLEAN_fft_retrieve_error_poly(uint8_t *error, const uint
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);

12
crypto_kem/hqc-rmrs-192/avx2/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS192_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS192_AVX2_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS192_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS192_AVX2_fft_retrieve_error_poly(uint8_t *error, const uint1
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);

12
crypto_kem/hqc-rmrs-192/clean/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS192_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS192_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS192_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeff
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS192_CLEAN_fft_retrieve_error_poly(uint8_t *error, const uint
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);

12
crypto_kem/hqc-rmrs-256/avx2/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS256_AVX2_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS256_AVX2_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS256_AVX2_fft(uint16_t *w, const uint16_t *f, size_t f_coeffs
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS256_AVX2_fft_retrieve_error_poly(uint8_t *error, const uint1
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);

12
crypto_kem/hqc-rmrs-256/clean/fft.c
查看文件

@ -30,7 +30,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
static void compute_fft_betas(uint16_t *betas) {
size_t i;
for (i = 0; i < PARAM_M - 1; ++i) {
betas[i] = 1 << (PARAM_M - 1 - i);
betas[i] = (uint16_t) (1 << (PARAM_M - 1 - i));
}
}
@ -133,7 +133,7 @@ static void radix_big(uint16_t *f0, uint16_t *f1, const uint16_t *f, uint32_t m_
size_t i, n;
n = 1 << (m_f - 2);
n = (size_t) (1 << (m_f - 2));
memcpy(Q, f + 3 * n, 2 * n);
memcpy(Q + n, f + 3 * n, 2 * n);
memcpy(R, f, 4 * n);
@ -201,7 +201,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 2: compute g
if (betas[m - 1] != 1) {
beta_m_pow = 1;
x = 1 << m_f;
x = (size_t) (1 << m_f);
for (i = 1; i < x; ++i) {
beta_m_pow = PQCLEAN_HQCRMRS256_CLEAN_gf_mul(beta_m_pow, betas[m - 1]);
f[i] = PQCLEAN_HQCRMRS256_CLEAN_gf_mul(beta_m_pow, f[i]);
@ -223,7 +223,7 @@ static void fft_rec(uint16_t *w, uint16_t *f, size_t f_coeffs, uint8_t m, uint32
// Step 5
fft_rec(u, f0, (f_coeffs + 1) / 2, m - 1, m_f - 1, deltas);
k = 1 << ((m - 1) & 0xf); // &0xf is to let the compiler know that m-1 is small.
k = (size_t) (1 << ((m - 1) & 0xf)); // &0xf is to let the compiler know that m-1 is small.
if (f_coeffs <= 3) { // 3-coefficient polynomial f case: f1 is constant
w[0] = u[0];
w[k] = u[0] ^ f1[0];
@ -299,7 +299,7 @@ void PQCLEAN_HQCRMRS256_CLEAN_fft(uint16_t *w, const uint16_t *f, size_t f_coeff
fft_rec(u, f0, (f_coeffs + 1) / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
fft_rec(v, f1, f_coeffs / 2, PARAM_M - 1, PARAM_FFT - 1, deltas);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
// Step 6, 7 and error polynomial computation
memcpy(w + k, v, 2 * k);
@ -334,7 +334,7 @@ void PQCLEAN_HQCRMRS256_CLEAN_fft_retrieve_error_poly(uint8_t *error, const uint
compute_fft_betas(gammas);
compute_subset_sums(gammas_sums, gammas, PARAM_M - 1);
k = 1 << (PARAM_M - 1);
k = (size_t) (1 << (PARAM_M - 1));
error[0] ^= 1 ^ ((uint16_t) - w[0] >> 15);
error[0] ^= 1 ^ ((uint16_t) - w[k] >> 15);