yuzu-emu/mbedtls
1
0
Fork 0
mirror of https://github.com/yuzu-emu/mbedtls.git synced 2024-11-23 03:45:39 +01:00
Code Issues Packages Projects Releases Wiki Activity

aria: optimise byte perms on Intel

Browse Source 
(A similar commit for Arm follows.)

Use specific instructions for moving bytes around in a word. This speeds
things up, and as a side-effect, slightly lowers code size.

ARIA_P3 (aka reverse byte order) is now 1 instruction on x86, which speeds up
key schedule. (Clang 3.8 finds this but GCC 5.4 doesn't.)

I couldn't find an Intel equivalent of ARM's ret16 (aka ARIA_P1), so I made it
two instructions, which is still much better than the code generated with
the previous mask-shift-or definition, and speeds up en/decryption. (Neither
Clang 3.8 nor GCC 5.4 find this.)

Before:
O	aria.o	ins
s	7976	43,865
2	10520	37,631
3	13040	28,146

After:
O	aria.o	ins
s	7768	33,497
2	9816	28,268
3	11432	20,829

For measurement method, see previous commit:
"aria: turn macro into static inline function"
This commit is contained in:
Manuel Pégourié-Gonnard 2018-02-26 16:08:40 +01:00
parent cac5008b17
commit fb0e4f0d1a
1 changed files with 28 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
library/aria.c
View File

@ -81,13 +81,26 @@ static void mbedtls_zeroize( void *v, size_t n ) {
* modify byte order: ( A B C D ) -> ( B A D C ), i.e. swap pairs of bytes * modify byte order: ( A B C D ) -> ( B A D C ), i.e. swap pairs of bytes
* *
* This is submatrix P1 in [1] Appendix B.1 * This is submatrix P1 in [1] Appendix B.1
*
* Common compilers fail to translate this to minimal number of instructions,
* so let's provide asm versions for common platforms with C fallback.
*/ */
#if defined(MBEDTLS_HAVE_ASM) && defined(__GNUC__)
#if defined(__i386__) || defined(__amd64__) || defined( __x86_64__)
/* I couldn't find an Intel equivalent of ret16, so two instructions */
#define ARIA_P1(x) ARIA_P2( ARIA_P3( x ) )
#endif
#endif /* MBEDTLS_HAVE_ASM && GNUC */
#if !defined(ARIA_P1)
#define ARIA_P1(x) ((((x) >> 8) & 0x00FF00FF) ^ (((x) & 0x00FF00FF) << 8)) #define ARIA_P1(x) ((((x) >> 8) & 0x00FF00FF) ^ (((x) & 0x00FF00FF) << 8))
#endif
/* /*
* modify byte order: ( A B C D ) -> ( C D A B ), i.e. rotate by 16 bits * modify byte order: ( A B C D ) -> ( C D A B ), i.e. rotate by 16 bits
* *
* This is submatrix P2 in [1] Appendix B.1 * This is submatrix P2 in [1] Appendix B.1
*
* Common compilers will translate this to a single instruction.
*/ */
#define ARIA_P2(x) (((x) >> 16) ^ ((x) << 16)) #define ARIA_P2(x) (((x) >> 16) ^ ((x) << 16))
@ -95,8 +108,23 @@ static void mbedtls_zeroize( void *v, size_t n ) {
* modify byte order: ( A B C D ) -> ( D C B A ), i.e. change endianness * modify byte order: ( A B C D ) -> ( D C B A ), i.e. change endianness
* *
* This is submatrix P3 in [1] Appendix B.1 * This is submatrix P3 in [1] Appendix B.1
*
* Some compilers fail to translate this to a single instruction,
* so let's provide asm versions for common platforms with C fallback.
*/ */
#if defined(MBEDTLS_HAVE_ASM) && defined(__GNUC__)
#if defined(__i386__) || defined(__amd64__) || defined( __x86_64__)
static inline uint32_t aria_p3( uint32_t x )
{
asm( "bswap %0" : "=r" (x) : "0" (x) );
return( x );
}
#define ARIA_P3 aria_p3
#endif
#endif /* MBEDTLS_HAVE_ASM && GNUC */
#if !defined(ARIA_P3)
#define ARIA_P3(x) ARIA_P2( ARIA_P1 ( x ) ) #define ARIA_P3(x) ARIA_P2( ARIA_P1 ( x ) )
#endif
/* /*
* ARIA Affine Transform * ARIA Affine Transform