eden-emu/mbedtls
15
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #7784 from daverodgman/aesce-unroll

Browse source
This commit is contained in:
Dave Rodgman 2023-07-04 18:41:13 +01:00 • committed by GitHub
commit c8d81ad54d
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 156 additions and 58 deletions
Download patch file Download diff file Expand all files Collapse all files

37
library/aes.c
View file

@ -1077,23 +1077,6 @@ int mbedtls_aes_crypt_ecb(mbedtls_aes_context *ctx,
#if defined(MBEDTLS_CIPHER_MODE_CBC)
#if defined(__ARM_NEON) && defined(__aarch64__)
/* Avoid using the NEON implementation of mbedtls_xor. Because of the dependency on
* the result for the next block in CBC, and the cost of transferring that data from
* NEON registers, it is faster to use the following on aarch64.
* For 32-bit arm, NEON should be faster. */
#define CBC_XOR_16(r, a, b) do { \
mbedtls_put_unaligned_uint64(r, \
mbedtls_get_unaligned_uint64(a) ^ \
mbedtls_get_unaligned_uint64(b)); \
mbedtls_put_unaligned_uint64(r + 8, \
mbedtls_get_unaligned_uint64(a + 8) ^ \
mbedtls_get_unaligned_uint64(b + 8)); \
} while (0)
#else
#define CBC_XOR_16(r, a, b) mbedtls_xor(r, a, b, 16)
#endif
/*
* AES-CBC buffer encryption/decryption
*/
@ -1136,7 +1119,10 @@ int mbedtls_aes_crypt_cbc(mbedtls_aes_context *ctx,
if (ret != 0) {
goto exit;
}
CBC_XOR_16(output, output, iv);
/* Avoid using the NEON implementation of mbedtls_xor. Because of the dependency on
* the result for the next block in CBC, and the cost of transferring that data from
* NEON registers, NEON is slower on aarch64. */
mbedtls_xor_no_simd(output, output, iv, 16);
memcpy(iv, temp, 16);
@ -1146,7 +1132,7 @@ int mbedtls_aes_crypt_cbc(mbedtls_aes_context *ctx,
}
} else {
while (length > 0) {
CBC_XOR_16(output, input, ivp);
mbedtls_xor_no_simd(output, input, ivp, 16);
ret = mbedtls_aes_crypt_ecb(ctx, mode, output, output);
if (ret != 0) {
@ -1179,8 +1165,11 @@ typedef unsigned char mbedtls_be128[16];
* for machine endianness and hence works correctly on both big and little
* endian machines.
*/
static void mbedtls_gf128mul_x_ble(unsigned char r[16],
const unsigned char x[16])
#if defined(MBEDTLS_AESCE_C) || defined(MBEDTLS_AESNI_C)
MBEDTLS_OPTIMIZE_FOR_PERFORMANCE
#endif
static inline void mbedtls_gf128mul_x_ble(unsigned char r[16],
const unsigned char x[16])
{
uint64_t a, b, ra, rb;
@ -1196,7 +1185,13 @@ static void mbedtls_gf128mul_x_ble(unsigned char r[16],
/*
* AES-XTS buffer encryption/decryption
*
* Use of MBEDTLS_OPTIMIZE_FOR_PERFORMANCE here and for mbedtls_gf128mul_x_ble()
* is a 3x performance improvement for gcc -Os, if we have hardware AES support.
*/
#if defined(MBEDTLS_AESCE_C) || defined(MBEDTLS_AESNI_C)
MBEDTLS_OPTIMIZE_FOR_PERFORMANCE
#endif
int mbedtls_aes_crypt_xts(mbedtls_aes_xts_context *ctx,
int mode,
size_t length,