yuzu/mbedtls
1
0
Fork 0
mirror of https://github.com/yuzu-emu/mbedtls.git synced 2024-12-23 11:05:29 +00:00
Code Issues Packages Projects Releases Wiki Activity

Make mod_p{224,256,384] a bit faster

Browse source 
Speedup is roughly 25%, giving a 6% speedup on ecp_mul() for these curves.
This commit is contained in:
Manuel Pégourié-Gonnard 2013-10-23 20:17:00 +02:00
parent c04c530a98
commit 5779cbe582
2 changed files with 19 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
include/polarssl/bignum.h
View file

@ -128,6 +128,7 @@ typedef uint32_t t_udbl;
#define POLARSSL_HAVE_UDBL
#else
#if ( defined(_MSC_VER) && defined(_M_AMD64) )
#define POLARSSL_HAVE_INT64
typedef int64_t t_sint;
typedef uint64_t t_uint;
#else
@ -137,6 +138,7 @@ typedef uint32_t t_udbl;
defined(__ia64__) || defined(__alpha__) || \
(defined(__sparc__) && defined(__arch64__)) || \
defined(__s390x__) ) )
#define POLARSSL_HAVE_INT64
typedef int64_t t_sint;
typedef uint64_t t_uint;
typedef unsigned int t_udbl __attribute__((mode(TI)));

24
library/ecp.c
View file

@ -663,22 +663,32 @@ static inline void sub32( uint32_t *dst, uint32_t src, signed char *carry )
if( c < 0 ) fix_negative( N, c, bits );
/*
* If the result is negative, we get it in the form c * 2^192 + N,
* with c negative and N positive (the c >= 0 case is handled by LAST).
* If the result is negative, we get it in the form
* c * 2^(bits + 32) + N, with c negative and N positive shorter than 'bits'
*/
static inline int fix_negative( mpi *N, signed char c, size_t bits )
{
int ret;
mpi C;
t_uint Cp[ 384 / 8 / sizeof( t_uint) + 1 ];
mpi_init( &C );
/* C = - c * 2^(bits + 32) */
C.s = 1;
C.n = bits / 8 / sizeof( t_uint ) + 1;
C.p = Cp;
memset( Cp, 0, C.n * sizeof( t_uint ) );
#if defined(POLARSSL_HAVE_INT64)
if( bits == 224 )
Cp[ C.n - 1 ] = ((t_uint) -c) << 32;
else
#endif
Cp[ C.n - 1 ] = (t_uint) -c;
MPI_CHK( mpi_lset( &C, c ) );
MPI_CHK( mpi_shift_l( &C, bits ) );
MPI_CHK( mpi_add_mpi( N, N, &C ) );
/* N = - ( C - N ) */
MPI_CHK( mpi_sub_abs( N, &C, N ) );
N->s = -1;
cleanup:
mpi_free( &C );
return( ret );
}