Simplify scrypt blkcpy()/blkxor()

This commit is contained in:
Frank Denis 2022-05-11 18:12:45 +02:00
parent 60aa65f68a
commit fd4b6edc0f

View File

@ -39,69 +39,25 @@
#include "private/common.h"
static inline void
blkcpy_64(escrypt_block_t *dest, const escrypt_block_t *src)
blkcpy(uint32_t *dest, const uint32_t *src, size_t len)
{
int i;
#if (ARCH_BITS == 32)
for (i = 0; i < 16; ++i) {
dest->w[i] = src->w[i];
}
#else
for (i = 0; i < 8; ++i) {
dest->d[i] = src->d[i];
}
#endif
memcpy(dest, src, len * 64);
}
static inline void
blkxor_64(escrypt_block_t *dest, const escrypt_block_t *src)
blkxor(uint32_t *dest, const uint32_t *src, size_t len)
{
int i;
size_t i;
#if (ARCH_BITS == 32)
for (i = 0; i < 16; ++i) {
dest->w[i] ^= src->w[i];
#if ARCH_BITS == 32
for (i = 0; i < len * 16; i++) {
dest[i] ^= src[i];
}
#else
for (i = 0; i < 8; ++i) {
dest->d[i] ^= src->d[i];
}
#endif
}
static inline void
blkcpy(escrypt_block_t *dest, const escrypt_block_t *src, size_t len)
{
size_t i, L;
#if (ARCH_BITS == 32)
L = (len >> 2);
for (i = 0; i < L; ++i) {
dest->w[i] = src->w[i];
}
#else
L = (len >> 3);
for (i = 0; i < L; ++i) {
dest->d[i] = src->d[i];
}
#endif
}
static inline void
blkxor(escrypt_block_t *dest, const escrypt_block_t *src, size_t len)
{
size_t i, L;
#if (ARCH_BITS == 32)
L = (len >> 2);
for (i = 0; i < L; ++i) {
dest->w[i] ^= src->w[i];
}
#else
L = (len >> 3);
for (i = 0; i < L; ++i) {
dest->d[i] ^= src->d[i];
uint64_t *dest_ = (uint64_t *) (void *) dest;
const uint64_t *src_ = (const uint64_t *) (const void *) src;
for (i = 0; i < len * 8; i++) {
dest_[i] ^= src_[i];
}
#endif
}
@ -113,11 +69,10 @@ blkxor(escrypt_block_t *dest, const escrypt_block_t *src, size_t len)
static void
salsa20_8(uint32_t B[16])
{
escrypt_block_t X;
uint32_t *x = X.w;
size_t i;
uint32_t x[16];
size_t i;
blkcpy_64(&X, (escrypt_block_t *) B);
blkcpy(x, B, 1);
for (i = 0; i < 8; i += 2) {
#define R(a, b) (((a) << (b)) | ((a) >> (32 - (b))))
/* Operate on columns. */
@ -181,30 +136,25 @@ blockmix_salsa8(const uint32_t *Bin, uint32_t *Bout, uint32_t *X, size_t r)
size_t i;
/* 1: X <-- B_{2r - 1} */
blkcpy_64((escrypt_block_t *) X,
(const escrypt_block_t *) &Bin[(2 * r - 1) * 16]);
blkcpy(X, &Bin[(2 * r - 1) * 16], 1);
/* 2: for i = 0 to 2r - 1 do */
for (i = 0; i < 2 * r; i += 2) {
/* 3: X <-- H(X \xor B_i) */
blkxor_64((escrypt_block_t *) X,
(const escrypt_block_t *) &Bin[i * 16]);
blkxor(X, &Bin[i * 16], 1);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy_64((escrypt_block_t *) &Bout[i * 8],
(const escrypt_block_t *) X);
blkcpy(&Bout[i * 8], X, 1);
/* 3: X <-- H(X \xor B_i) */
blkxor_64((escrypt_block_t *) X,
(const escrypt_block_t *) &Bin[i * 16 + 16]);
blkxor(X, &Bin[i * 16 + 16], 1);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy_64((escrypt_block_t *) &Bout[i * 8 + r * 16],
(escrypt_block_t *) X);
blkcpy(&Bout[i * 8 + r * 16], X, 1);
}
}
@ -245,15 +195,13 @@ smix(uint8_t *B, size_t r, uint64_t N, uint32_t *V, uint32_t *XY)
/* 2: for i = 0 to N - 1 do */
for (i = 0; i < N; i += 2) {
/* 3: V_i <-- X */
blkcpy((escrypt_block_t *) &V[i * (32 * r)], (escrypt_block_t *) X,
128 * r);
blkcpy(&V[i * (32 * r)], X, 2 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(X, Y, Z, r);
/* 3: V_i <-- X */
blkcpy((escrypt_block_t *) &V[(i + 1) * (32 * r)],
(escrypt_block_t *) Y, 128 * r);
blkcpy(&V[(i + 1) * (32 * r)], Y, 2 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(Y, X, Z, r);
@ -265,16 +213,14 @@ smix(uint8_t *B, size_t r, uint64_t N, uint32_t *V, uint32_t *XY)
j = integerify(X, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor((escrypt_block_t *) X, (escrypt_block_t *) &V[j * (32 * r)],
128 * r);
blkxor(X, &V[j * (32 * r)], 2 * r);
blockmix_salsa8(X, Y, Z, r);
/* 7: j <-- Integerify(X) mod N */
j = integerify(Y, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor((escrypt_block_t *) Y, (escrypt_block_t *) &V[j * (32 * r)],
128 * r);
blkxor(Y, &V[j * (32 * r)], 2 * r);
blockmix_salsa8(Y, X, Z, r);
}
/* 10: B' <-- X */