arne
/
cryptopp
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

fix compile on Mac OS X

pull/2/head
weidai 2007-05-04 20:46:32 +00:00
parent a740a0ff83
commit f46f0d73c5
2 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
config.h
View File

@ -196,7 +196,7 @@ NAMESPACE_END
#endif #endif
#ifndef CRYPTOPP_SECTION_ALIGN16 #ifndef CRYPTOPP_SECTION_ALIGN16
#ifdef __GNUC__ #if defined(__GNUC__) && !defined(__APPLE__)
// the alignment attribute doesn't seem to work without this section attribute when -fdata-sections is turned on // the alignment attribute doesn't seem to work without this section attribute when -fdata-sections is turned on
#define CRYPTOPP_SECTION_ALIGN16 __attribute__((section ("CryptoPP_Align16"))) #define CRYPTOPP_SECTION_ALIGN16 __attribute__((section ("CryptoPP_Align16")))
#else #else

13
salsa.cpp
View File

@ -72,6 +72,11 @@ unsigned int Salsa20_Policy::GetOptimalBlockSize() const
} }
#endif #endif
#if CRYPTOPP_BOOL_SSE2_INTRINSICS_AVAILABLE
static const __m128i s_maskLo32 = _mm_shuffle_epi32(_mm_cvtsi32_si128(-1), _MM_SHUFFLE(1, 0, 1, 0));
static const __m128i s_maskHi32 = _mm_slli_epi64(s_maskLo32, 32);
#endif
void Salsa20_Policy::OperateKeystream(KeystreamOperation operation, byte *output, const byte *input, size_t iterationCount) void Salsa20_Policy::OperateKeystream(KeystreamOperation operation, byte *output, const byte *input, size_t iterationCount)
{ {
int i; int i;
@ -238,16 +243,12 @@ void Salsa20_Policy::OperateKeystream(KeystreamOperation operation, byte *output
if (++m_state[8] == 0) if (++m_state[8] == 0)
++m_state[5]; ++m_state[5];
CRYPTOPP_ALIGN_DATA(16) static const word32 masks[8] CRYPTOPP_SECTION_ALIGN16 =
{0, 0xffffffff, 0, 0xffffffff, 0xffffffff, 0, 0xffffffff, 0};
__m128i k02 = _mm_or_si128(_mm_slli_epi64(x0, 32), _mm_srli_epi64(x3, 32)); __m128i k02 = _mm_or_si128(_mm_slli_epi64(x0, 32), _mm_srli_epi64(x3, 32));
k02 = _mm_shuffle_epi32(k02, _MM_SHUFFLE(0, 1, 2, 3)); k02 = _mm_shuffle_epi32(k02, _MM_SHUFFLE(0, 1, 2, 3));
__m128i k13 = _mm_or_si128(_mm_slli_epi64(x1, 32), _mm_srli_epi64(x0, 32)); __m128i k13 = _mm_or_si128(_mm_slli_epi64(x1, 32), _mm_srli_epi64(x0, 32));
k13 = _mm_shuffle_epi32(k13, _MM_SHUFFLE(0, 1, 2, 3)); k13 = _mm_shuffle_epi32(k13, _MM_SHUFFLE(0, 1, 2, 3));
__m128i maskLo32 = ((__m128i*)masks)[1], maskHi32 = ((__m128i*)masks)[0]; __m128i k20 = _mm_or_si128(_mm_and_si128(x2, s_maskLo32), _mm_and_si128(x1, s_maskHi32));
__m128i k20 = _mm_or_si128(_mm_and_si128(x2, maskLo32), _mm_and_si128(x1, maskHi32)); __m128i k31 = _mm_or_si128(_mm_and_si128(x3, s_maskLo32), _mm_and_si128(x2, s_maskHi32));
__m128i k31 = _mm_or_si128(_mm_and_si128(x3, maskLo32), _mm_and_si128(x2, maskHi32));
__m128i k0 = _mm_unpackhi_epi64(k02, k20); __m128i k0 = _mm_unpackhi_epi64(k02, k20);
__m128i k1 = _mm_unpackhi_epi64(k13, k31); __m128i k1 = _mm_unpackhi_epi64(k13, k31);