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

Fix failed compile on Ubuntu with -msse2 

Also see https://github.com/noloader/cryptopp-cmake/issues/36
pull/730/head
Jeffrey Walton 2018-10-26 02:43:35 -04:00
parent 99c65bdb35
commit c992fe98a9
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
1 changed files with 12 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
chacha-simd.cpp
View File

@ -9,7 +9,7 @@
// SSE2 implementation based on Botan's chacha_sse2.cpp. Many thanks // SSE2 implementation based on Botan's chacha_sse2.cpp. Many thanks
// to Jack Lloyd and the Botan team for allowing us to use it. // to Jack Lloyd and the Botan team for allowing us to use it.
// //
// NEON and Power7 is upcoming. // Power8 is upcoming.
#include "pch.h" #include "pch.h"
#include "config.h" #include "config.h"
@ -17,16 +17,16 @@
#include "chacha.h" #include "chacha.h"
#include "misc.h" #include "misc.h"
#if (CRYPTOPP_SSE2_INTRIN_AVAILABLE || CRYPTOPP_SSE2_ASM_AVAILABLE) #if defined(__SSE2__) || defined(_MSC_VER)
# include <xmmintrin.h> # include <xmmintrin.h>
# include <emmintrin.h> # include <emmintrin.h>
#endif #endif
#if (CRYPTOPP_SSSE3_INTRIN_AVAILABLE || CRYPTOPP_SSSE3_ASM_AVAILABLE) #if defined(__SSSE3__) || defined(_MSC_VER)
# include <tmmintrin.h> # include <tmmintrin.h>
#endif #endif
#ifdef __XOP__ #if defined(__XOP__)
# include <ammintrin.h> # include <ammintrin.h>
#endif #endif
@ -128,29 +128,31 @@ inline __m128i RotateLeft(const __m128i val)
#endif #endif
} }
#if defined(__SSSE3__)
template <> template <>
inline __m128i RotateLeft<8>(const __m128i val) inline __m128i RotateLeft<8>(const __m128i val)
{ {
#ifdef __XOP__ #if defined(__XOP__)
return _mm_roti_epi32(val, 8); return _mm_roti_epi32(val, 8);
#else #elif defined(__SSSE3__)
const __m128i mask = _mm_set_epi8(14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3); const __m128i mask = _mm_set_epi8(14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3);
return _mm_shuffle_epi8(val, mask); return _mm_shuffle_epi8(val, mask);
#else
return _mm_or_si128(_mm_slli_epi32(val, 8), _mm_srli_epi32(val, 32-8));
#endif #endif
} }
template <> template <>
inline __m128i RotateLeft<16>(const __m128i val) inline __m128i RotateLeft<16>(const __m128i val)
{ {
#ifdef __XOP__ #if defined(__XOP__)
return _mm_roti_epi32(val, 16); return _mm_roti_epi32(val, 16);
#else #elif defined(__SSSE3__)
const __m128i mask = _mm_set_epi8(13,12,15,14, 9,8,11,10, 5,4,7,6, 1,0,3,2); const __m128i mask = _mm_set_epi8(13,12,15,14, 9,8,11,10, 5,4,7,6, 1,0,3,2);
return _mm_shuffle_epi8(val, mask); return _mm_shuffle_epi8(val, mask);
#else
return _mm_or_si128(_mm_slli_epi32(val, 16), _mm_srli_epi32(val, 32-16));
#endif #endif
} }
#endif // SSE3
#endif // CRYPTOPP_SSE2_INTRIN_AVAILABLE || CRYPTOPP_SSE2_ASM_AVAILABLE #endif // CRYPTOPP_SSE2_INTRIN_AVAILABLE || CRYPTOPP_SSE2_ASM_AVAILABLE