yuzu-emu/mbedtls
1
0
Fork 0
mirror of https://github.com/yuzu-emu/mbedtls.git synced 2024-11-25 16:05:43 +01:00
Code Issues Packages Projects Releases Wiki Activity

aesni: Unroll aes rounds

Browse Source
This commit is contained in:
Morph 2022-09-05 17:00:02 -04:00
parent 3ef96cc1ad
commit 37d6376968
1 changed files with 78 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

76
library/aesni.c
View File

@ -117,14 +117,90 @@ int mbedtls_aesni_crypt_ecb( mbedtls_aes_context *ctx,
if (mode == MBEDTLS_AES_ENCRYPT) if (mode == MBEDTLS_AES_ENCRYPT)
{ {
if (ctx->nr == 10) {
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
} else if (ctx->nr == 12) {
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
} else if (ctx->nr == 14) {
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
} else {
for (i = ctx->nr - 1; i; --i) for (i = ctx->nr - 1; i; --i)
a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) ); a = _mm_aesenc_si128( a, _mm_loadu_si128( rk++ ) );
}
a = _mm_aesenclast_si128( a, _mm_loadu_si128( rk ) ); a = _mm_aesenclast_si128( a, _mm_loadu_si128( rk ) );
} }
else else
{ {
if (ctx->nr == 10) {
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
} else if (ctx->nr == 12) {
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
} else if (ctx->nr == 14) {
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
} else {
for (i = ctx->nr - 1; i; --i) for (i = ctx->nr - 1; i; --i)
a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) ); a = _mm_aesdec_si128( a, _mm_loadu_si128( rk++ ) );
}
a = _mm_aesdeclast_si128( a, _mm_loadu_si128( rk ) ); a = _mm_aesdeclast_si128( a, _mm_loadu_si128( rk ) );
} }