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

Add CRYPTOPP_INLINE to SIMON and SPECk for debugging

pull/739/head
Jeffrey Walton 2018-11-13 00:02:34 -05:00
parent 4da4f7ea2a
commit 27b61d8752
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
5 changed files with 68 additions and 62 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
cryptlib.vcxproj.filters
View File

@ -989,7 +989,5 @@
<CustomBuild Include="adhoc.cpp.proto"> <CustomBuild Include="adhoc.cpp.proto">
<Filter>Miscellaneous</Filter> <Filter>Miscellaneous</Filter>
</CustomBuild> </CustomBuild>
<CustomBuild Include="x64dll.asm" />
<CustomBuild Include="x64masm.asm" />
</ItemGroup> </ItemGroup>
</Project> </Project>

64
simon128_simd.cpp
View File

@ -46,6 +46,14 @@
# include "ppc_simd.h" # include "ppc_simd.h"
#endif #endif
#ifndef CRYPTOPP_INLINE
# if defined(CRYPTOPP_DEBUG)
# define CRYPTOPP_INLINE static
# else
# define CRYPTOPP_INLINE inline
# endif
#endif
// Squash MS LNK4221 and libtool warnings // Squash MS LNK4221 and libtool warnings
extern const char SIMON128_SIMD_FNAME[] = __FILE__; extern const char SIMON128_SIMD_FNAME[] = __FILE__;
@ -63,7 +71,7 @@ using CryptoPP::vec_swap; // SunCC
#if (CRYPTOPP_ARM_NEON_AVAILABLE) #if (CRYPTOPP_ARM_NEON_AVAILABLE)
template <class T> template <class T>
inline T UnpackHigh64(const T& a, const T& b) CRYPTOPP_INLINE T UnpackHigh64(const T& a, const T& b)
{ {
const uint64x1_t x(vget_high_u64((uint64x2_t)a)); const uint64x1_t x(vget_high_u64((uint64x2_t)a));
const uint64x1_t y(vget_high_u64((uint64x2_t)b)); const uint64x1_t y(vget_high_u64((uint64x2_t)b));
@ -71,7 +79,7 @@ inline T UnpackHigh64(const T& a, const T& b)
} }
template <class T> template <class T>
inline T UnpackLow64(const T& a, const T& b) CRYPTOPP_INLINE T UnpackLow64(const T& a, const T& b)
{ {
const uint64x1_t x(vget_low_u64((uint64x2_t)a)); const uint64x1_t x(vget_low_u64((uint64x2_t)a));
const uint64x1_t y(vget_low_u64((uint64x2_t)b)); const uint64x1_t y(vget_low_u64((uint64x2_t)b));
@ -79,7 +87,7 @@ inline T UnpackLow64(const T& a, const T& b)
} }
template <unsigned int R> template <unsigned int R>
inline uint64x2_t RotateLeft64(const uint64x2_t& val) CRYPTOPP_INLINE uint64x2_t RotateLeft64(const uint64x2_t& val)
{ {
const uint64x2_t a(vshlq_n_u64(val, R)); const uint64x2_t a(vshlq_n_u64(val, R));
const uint64x2_t b(vshrq_n_u64(val, 64 - R)); const uint64x2_t b(vshrq_n_u64(val, 64 - R));
@ -87,7 +95,7 @@ inline uint64x2_t RotateLeft64(const uint64x2_t& val)
} }
template <unsigned int R> template <unsigned int R>
inline uint64x2_t RotateRight64(const uint64x2_t& val) CRYPTOPP_INLINE uint64x2_t RotateRight64(const uint64x2_t& val)
{ {
const uint64x2_t a(vshlq_n_u64(val, 64 - R)); const uint64x2_t a(vshlq_n_u64(val, 64 - R));
const uint64x2_t b(vshrq_n_u64(val, R)); const uint64x2_t b(vshrq_n_u64(val, R));
@ -97,7 +105,7 @@ inline uint64x2_t RotateRight64(const uint64x2_t& val)
#if defined(__aarch32__) || defined(__aarch64__) #if defined(__aarch32__) || defined(__aarch64__)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val) CRYPTOPP_INLINE uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8_t maskb[16] = { 14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7 }; const uint8_t maskb[16] = { 14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7 };
@ -113,7 +121,7 @@ inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline uint64x2_t RotateRight64<8>(const uint64x2_t& val) CRYPTOPP_INLINE uint64x2_t RotateRight64<8>(const uint64x2_t& val)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8_t maskb[16] = { 8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1 }; const uint8_t maskb[16] = { 8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1 };
@ -128,13 +136,13 @@ inline uint64x2_t RotateRight64<8>(const uint64x2_t& val)
} }
#endif #endif
inline uint64x2_t SIMON128_f(const uint64x2_t& val) CRYPTOPP_INLINE uint64x2_t SIMON128_f(const uint64x2_t& val)
{ {
return veorq_u64(RotateLeft64<2>(val), return veorq_u64(RotateLeft64<2>(val),
vandq_u64(RotateLeft64<1>(val), RotateLeft64<8>(val))); vandq_u64(RotateLeft64<1>(val), RotateLeft64<8>(val)));
} }
inline void SIMON128_Enc_Block(uint64x2_t &block0, uint64x2_t &block1, CRYPTOPP_INLINE void SIMON128_Enc_Block(uint64x2_t &block0, uint64x2_t &block1,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
// [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ... // [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ...
@ -163,7 +171,7 @@ inline void SIMON128_Enc_Block(uint64x2_t &block0, uint64x2_t &block1,
block1 = UnpackHigh64(y1, x1); block1 = UnpackHigh64(y1, x1);
} }
inline void SIMON128_Enc_6_Blocks(uint64x2_t &block0, uint64x2_t &block1, CRYPTOPP_INLINE void SIMON128_Enc_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5, uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
@ -207,7 +215,7 @@ inline void SIMON128_Enc_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
block5 = UnpackHigh64(y3, x3); block5 = UnpackHigh64(y3, x3);
} }
inline void SIMON128_Dec_Block(uint64x2_t &block0, uint64x2_t &block1, CRYPTOPP_INLINE void SIMON128_Dec_Block(uint64x2_t &block0, uint64x2_t &block1,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
// [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ... // [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ...
@ -237,7 +245,7 @@ inline void SIMON128_Dec_Block(uint64x2_t &block0, uint64x2_t &block1,
block1 = UnpackHigh64(y1, x1); block1 = UnpackHigh64(y1, x1);
} }
inline void SIMON128_Dec_6_Blocks(uint64x2_t &block0, uint64x2_t &block1, CRYPTOPP_INLINE void SIMON128_Dec_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5, uint64x2_t &block2, uint64x2_t &block3, uint64x2_t &block4, uint64x2_t &block5,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
@ -304,7 +312,7 @@ inline void SIMON128_Dec_6_Blocks(uint64x2_t &block0, uint64x2_t &block1,
# define CONST_DOUBLE_CAST(x) ((const double *)(const void *)(x)) # define CONST_DOUBLE_CAST(x) ((const double *)(const void *)(x))
#endif #endif
inline void Swap128(__m128i& a,__m128i& b) CRYPTOPP_INLINE void Swap128(__m128i& a,__m128i& b)
{ {
#if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120) #if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120)
// __m128i is an unsigned long long[2], and support for swapping it was not added until C++11. // __m128i is an unsigned long long[2], and support for swapping it was not added until C++11.
@ -316,7 +324,7 @@ inline void Swap128(__m128i& a,__m128i& b)
} }
template <unsigned int R> template <unsigned int R>
inline __m128i RotateLeft64(const __m128i& val) CRYPTOPP_INLINE __m128i RotateLeft64(const __m128i& val)
{ {
#if defined(CRYPTOPP_AVX512_ROTATE) #if defined(CRYPTOPP_AVX512_ROTATE)
return _mm_rol_epi64(val, R); return _mm_rol_epi64(val, R);
@ -329,7 +337,7 @@ inline __m128i RotateLeft64(const __m128i& val)
} }
template <unsigned int R> template <unsigned int R>
inline __m128i RotateRight64(const __m128i& val) CRYPTOPP_INLINE __m128i RotateRight64(const __m128i& val)
{ {
#if defined(CRYPTOPP_AVX512_ROTATE) #if defined(CRYPTOPP_AVX512_ROTATE)
return _mm_ror_epi64(val, R); return _mm_ror_epi64(val, R);
@ -343,7 +351,7 @@ inline __m128i RotateRight64(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline __m128i RotateLeft64<8>(const __m128i& val) __m128i RotateLeft64<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi64(val, 8); return _mm_roti_epi64(val, 8);
@ -355,7 +363,7 @@ inline __m128i RotateLeft64<8>(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline __m128i RotateRight64<8>(const __m128i& val) __m128i RotateRight64<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi64(val, 64-8); return _mm_roti_epi64(val, 64-8);
@ -365,13 +373,13 @@ inline __m128i RotateRight64<8>(const __m128i& val)
#endif #endif
} }
inline __m128i SIMON128_f(const __m128i& v) CRYPTOPP_INLINE __m128i SIMON128_f(const __m128i& v)
{ {
return _mm_xor_si128(RotateLeft64<2>(v), return _mm_xor_si128(RotateLeft64<2>(v),
_mm_and_si128(RotateLeft64<1>(v), RotateLeft64<8>(v))); _mm_and_si128(RotateLeft64<1>(v), RotateLeft64<8>(v)));
} }
inline void SIMON128_Enc_Block(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON128_Enc_Block(__m128i &block0, __m128i &block1,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
// [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ... // [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ...
@ -403,7 +411,7 @@ inline void SIMON128_Enc_Block(__m128i &block0, __m128i &block1,
block1 = _mm_unpackhi_epi64(y1, x1); block1 = _mm_unpackhi_epi64(y1, x1);
} }
inline void SIMON128_Enc_6_Blocks(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON128_Enc_6_Blocks(__m128i &block0, __m128i &block1,
__m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5, __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
@ -449,7 +457,7 @@ inline void SIMON128_Enc_6_Blocks(__m128i &block0, __m128i &block1,
block5 = _mm_unpackhi_epi64(y3, x3); block5 = _mm_unpackhi_epi64(y3, x3);
} }
inline void SIMON128_Dec_Block(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON128_Dec_Block(__m128i &block0, __m128i &block1,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
// [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ... // [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ...
@ -482,7 +490,7 @@ inline void SIMON128_Dec_Block(__m128i &block0, __m128i &block1,
block1 = _mm_unpackhi_epi64(y1, x1); block1 = _mm_unpackhi_epi64(y1, x1);
} }
inline void SIMON128_Dec_6_Blocks(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON128_Dec_6_Blocks(__m128i &block0, __m128i &block1,
__m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5, __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
const word64 *subkeys, unsigned int rounds) const word64 *subkeys, unsigned int rounds)
{ {
@ -545,7 +553,7 @@ using CryptoPP::VectorXor;
// Rotate left by bit count // Rotate left by bit count
template<unsigned int C> template<unsigned int C>
inline uint64x2_p RotateLeft64(const uint64x2_p val) CRYPTOPP_INLINE uint64x2_p RotateLeft64(const uint64x2_p val)
{ {
const uint64x2_p m = {C, C}; const uint64x2_p m = {C, C};
return vec_rl(val, m); return vec_rl(val, m);
@ -553,19 +561,19 @@ inline uint64x2_p RotateLeft64(const uint64x2_p val)
// Rotate right by bit count // Rotate right by bit count
template<unsigned int C> template<unsigned int C>
inline uint64x2_p RotateRight64(const uint64x2_p val) CRYPTOPP_INLINE uint64x2_p RotateRight64(const uint64x2_p val)
{ {
const uint64x2_p m = {64-C, 64-C}; const uint64x2_p m = {64-C, 64-C};
return vec_rl(val, m); return vec_rl(val, m);
} }
inline uint64x2_p SIMON128_f(const uint64x2_p val) CRYPTOPP_INLINE uint64x2_p SIMON128_f(const uint64x2_p val)
{ {
return VectorXor(RotateLeft64<2>(val), return VectorXor(RotateLeft64<2>(val),
VectorAnd(RotateLeft64<1>(val), RotateLeft64<8>(val))); VectorAnd(RotateLeft64<1>(val), RotateLeft64<8>(val)));
} }
inline void SIMON128_Enc_Block(uint32x4_p &block, const word64 *subkeys, unsigned int rounds) CRYPTOPP_INLINE void SIMON128_Enc_Block(uint32x4_p &block, const word64 *subkeys, unsigned int rounds)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m1 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8}; const uint8x16_p m1 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8};
@ -607,7 +615,7 @@ inline void SIMON128_Enc_Block(uint32x4_p &block, const word64 *subkeys, unsigne
block = (uint32x4_p)vec_perm(x1, y1, m3); block = (uint32x4_p)vec_perm(x1, y1, m3);
} }
inline void SIMON128_Dec_Block(uint32x4_p &block, const word64 *subkeys, unsigned int rounds) CRYPTOPP_INLINE void SIMON128_Dec_Block(uint32x4_p &block, const word64 *subkeys, unsigned int rounds)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m1 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8}; const uint8x16_p m1 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8};
@ -650,7 +658,7 @@ inline void SIMON128_Dec_Block(uint32x4_p &block, const word64 *subkeys, unsigne
block = (uint32x4_p)vec_perm(x1, y1, m3); block = (uint32x4_p)vec_perm(x1, y1, m3);
} }
inline void SIMON128_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON128_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4, uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4,
uint32x4_p &block5, const word64 *subkeys, unsigned int rounds) uint32x4_p &block5, const word64 *subkeys, unsigned int rounds)
{ {
@ -709,7 +717,7 @@ inline void SIMON128_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
block5 = (uint32x4_p)vec_perm(x3, y3, m4); block5 = (uint32x4_p)vec_perm(x3, y3, m4);
} }
inline void SIMON128_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON128_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4, uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4,
uint32x4_p &block5, const word64 *subkeys, unsigned int rounds) uint32x4_p &block5, const word64 *subkeys, unsigned int rounds)
{ {

56
simon64_simd.cpp
View File

@ -75,7 +75,7 @@ using CryptoPP::vec_swap; // SunCC
#if (CRYPTOPP_ARM_NEON_AVAILABLE) #if (CRYPTOPP_ARM_NEON_AVAILABLE)
template <class T> template <class T>
inline T UnpackHigh32(const T& a, const T& b) CRYPTOPP_INLINE T UnpackHigh32(const T& a, const T& b)
{ {
const uint32x2_t x(vget_high_u32((uint32x4_t)a)); const uint32x2_t x(vget_high_u32((uint32x4_t)a));
const uint32x2_t y(vget_high_u32((uint32x4_t)b)); const uint32x2_t y(vget_high_u32((uint32x4_t)b));
@ -84,7 +84,7 @@ inline T UnpackHigh32(const T& a, const T& b)
} }
template <class T> template <class T>
inline T UnpackLow32(const T& a, const T& b) CRYPTOPP_INLINE T UnpackLow32(const T& a, const T& b)
{ {
const uint32x2_t x(vget_low_u32((uint32x4_t)a)); const uint32x2_t x(vget_low_u32((uint32x4_t)a));
const uint32x2_t y(vget_low_u32((uint32x4_t)b)); const uint32x2_t y(vget_low_u32((uint32x4_t)b));
@ -93,7 +93,7 @@ inline T UnpackLow32(const T& a, const T& b)
} }
template <unsigned int R> template <unsigned int R>
inline uint32x4_t RotateLeft32(const uint32x4_t& val) CRYPTOPP_INLINE uint32x4_t RotateLeft32(const uint32x4_t& val)
{ {
const uint32x4_t a(vshlq_n_u32(val, R)); const uint32x4_t a(vshlq_n_u32(val, R));
const uint32x4_t b(vshrq_n_u32(val, 32 - R)); const uint32x4_t b(vshrq_n_u32(val, 32 - R));
@ -101,7 +101,7 @@ inline uint32x4_t RotateLeft32(const uint32x4_t& val)
} }
template <unsigned int R> template <unsigned int R>
inline uint32x4_t RotateRight32(const uint32x4_t& val) CRYPTOPP_INLINE uint32x4_t RotateRight32(const uint32x4_t& val)
{ {
const uint32x4_t a(vshlq_n_u32(val, 32 - R)); const uint32x4_t a(vshlq_n_u32(val, 32 - R));
const uint32x4_t b(vshrq_n_u32(val, R)); const uint32x4_t b(vshrq_n_u32(val, R));
@ -111,7 +111,7 @@ inline uint32x4_t RotateRight32(const uint32x4_t& val)
#if defined(__aarch32__) || defined(__aarch64__) #if defined(__aarch32__) || defined(__aarch64__)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val) CRYPTOPP_INLINE uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8_t maskb[16] = { 14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3 }; const uint8_t maskb[16] = { 14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3 };
@ -127,7 +127,7 @@ inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline uint32x4_t RotateRight32<8>(const uint32x4_t& val) CRYPTOPP_INLINE uint32x4_t RotateRight32<8>(const uint32x4_t& val)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
const uint8_t maskb[16] = { 12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1 }; const uint8_t maskb[16] = { 12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1 };
@ -142,13 +142,13 @@ inline uint32x4_t RotateRight32<8>(const uint32x4_t& val)
} }
#endif #endif
inline uint32x4_t SIMON64_f(const uint32x4_t& val) CRYPTOPP_INLINE uint32x4_t SIMON64_f(const uint32x4_t& val)
{ {
return veorq_u32(RotateLeft32<2>(val), return veorq_u32(RotateLeft32<2>(val),
vandq_u32(RotateLeft32<1>(val), RotateLeft32<8>(val))); vandq_u32(RotateLeft32<1>(val), RotateLeft32<8>(val)));
} }
inline void SIMON64_Enc_Block(uint32x4_t &block1, uint32x4_t &block0, CRYPTOPP_INLINE void SIMON64_Enc_Block(uint32x4_t &block1, uint32x4_t &block0,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
// [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ... // [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ...
@ -177,7 +177,7 @@ inline void SIMON64_Enc_Block(uint32x4_t &block1, uint32x4_t &block0,
block1 = UnpackHigh32(y1, x1); block1 = UnpackHigh32(y1, x1);
} }
inline void SIMON64_Dec_Block(uint32x4_t &block0, uint32x4_t &block1, CRYPTOPP_INLINE void SIMON64_Dec_Block(uint32x4_t &block0, uint32x4_t &block1,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
// [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ... // [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ...
@ -207,7 +207,7 @@ inline void SIMON64_Dec_Block(uint32x4_t &block0, uint32x4_t &block1,
block1 = UnpackHigh32(y1, x1); block1 = UnpackHigh32(y1, x1);
} }
inline void SIMON64_Enc_6_Blocks(uint32x4_t &block0, uint32x4_t &block1, CRYPTOPP_INLINE void SIMON64_Enc_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5, uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
@ -251,7 +251,7 @@ inline void SIMON64_Enc_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
block5 = UnpackHigh32(y3, x3); block5 = UnpackHigh32(y3, x3);
} }
inline void SIMON64_Dec_6_Blocks(uint32x4_t &block0, uint32x4_t &block1, CRYPTOPP_INLINE void SIMON64_Dec_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5, uint32x4_t &block2, uint32x4_t &block3, uint32x4_t &block4, uint32x4_t &block5,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
@ -302,7 +302,7 @@ inline void SIMON64_Dec_6_Blocks(uint32x4_t &block0, uint32x4_t &block1,
#if defined(CRYPTOPP_SSE41_AVAILABLE) #if defined(CRYPTOPP_SSE41_AVAILABLE)
inline void Swap128(__m128i& a,__m128i& b) CRYPTOPP_INLINE void Swap128(__m128i& a,__m128i& b)
{ {
#if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120) #if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120)
// __m128i is an unsigned long long[2], and support for swapping it was not added until C++11. // __m128i is an unsigned long long[2], and support for swapping it was not added until C++11.
@ -314,7 +314,7 @@ inline void Swap128(__m128i& a,__m128i& b)
} }
template <unsigned int R> template <unsigned int R>
inline __m128i RotateLeft32(const __m128i& val) CRYPTOPP_INLINE __m128i RotateLeft32(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, R); return _mm_roti_epi32(val, R);
@ -325,7 +325,7 @@ inline __m128i RotateLeft32(const __m128i& val)
} }
template <unsigned int R> template <unsigned int R>
inline __m128i RotateRight32(const __m128i& val) CRYPTOPP_INLINE __m128i RotateRight32(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, 32-R); return _mm_roti_epi32(val, 32-R);
@ -337,7 +337,7 @@ inline __m128i RotateRight32(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline __m128i RotateLeft32<8>(const __m128i& val) __m128i RotateLeft32<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, 8); return _mm_roti_epi32(val, 8);
@ -349,7 +349,7 @@ inline __m128i RotateLeft32<8>(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
inline __m128i RotateRight32<8>(const __m128i& val) __m128i RotateRight32<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, 32-8); return _mm_roti_epi32(val, 32-8);
@ -359,13 +359,13 @@ inline __m128i RotateRight32<8>(const __m128i& val)
#endif #endif
} }
inline __m128i SIMON64_f(const __m128i& v) CRYPTOPP_INLINE __m128i SIMON64_f(const __m128i& v)
{ {
return _mm_xor_si128(RotateLeft32<2>(v), return _mm_xor_si128(RotateLeft32<2>(v),
_mm_and_si128(RotateLeft32<1>(v), RotateLeft32<8>(v))); _mm_and_si128(RotateLeft32<1>(v), RotateLeft32<8>(v)));
} }
inline void SIMON64_Enc_Block(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON64_Enc_Block(__m128i &block0, __m128i &block1,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
// [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ... // [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ...
@ -395,7 +395,7 @@ inline void SIMON64_Enc_Block(__m128i &block0, __m128i &block1,
block1 = _mm_unpackhi_epi32(y1, x1); block1 = _mm_unpackhi_epi32(y1, x1);
} }
inline void SIMON64_Dec_Block(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON64_Dec_Block(__m128i &block0, __m128i &block1,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
// [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ... // [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ...
@ -426,7 +426,7 @@ inline void SIMON64_Dec_Block(__m128i &block0, __m128i &block1,
block1 = _mm_unpackhi_epi32(y1, x1); block1 = _mm_unpackhi_epi32(y1, x1);
} }
inline void SIMON64_Enc_6_Blocks(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON64_Enc_6_Blocks(__m128i &block0, __m128i &block1,
__m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5, __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
@ -477,7 +477,7 @@ inline void SIMON64_Enc_6_Blocks(__m128i &block0, __m128i &block1,
block5 = _mm_unpackhi_epi32(y3, x3); block5 = _mm_unpackhi_epi32(y3, x3);
} }
inline void SIMON64_Dec_6_Blocks(__m128i &block0, __m128i &block1, CRYPTOPP_INLINE void SIMON64_Dec_6_Blocks(__m128i &block0, __m128i &block1,
__m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5, __m128i &block2, __m128i &block3, __m128i &block4, __m128i &block5,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
@ -545,7 +545,7 @@ using CryptoPP::VectorLoadBE;
// Rotate left by bit count // Rotate left by bit count
template<unsigned int C> template<unsigned int C>
inline uint32x4_p RotateLeft32(const uint32x4_p val) CRYPTOPP_INLINE uint32x4_p RotateLeft32(const uint32x4_p val)
{ {
const uint32x4_p m = {C, C, C, C}; const uint32x4_p m = {C, C, C, C};
return vec_rl(val, m); return vec_rl(val, m);
@ -553,19 +553,19 @@ inline uint32x4_p RotateLeft32(const uint32x4_p val)
// Rotate right by bit count // Rotate right by bit count
template<unsigned int C> template<unsigned int C>
inline uint32x4_p RotateRight32(const uint32x4_p val) CRYPTOPP_INLINE uint32x4_p RotateRight32(const uint32x4_p val)
{ {
const uint32x4_p m = {32-C, 32-C, 32-C, 32-C}; const uint32x4_p m = {32-C, 32-C, 32-C, 32-C};
return vec_rl(val, m); return vec_rl(val, m);
} }
inline uint32x4_p SIMON64_f(const uint32x4_p val) CRYPTOPP_INLINE uint32x4_p SIMON64_f(const uint32x4_p val)
{ {
return VectorXor(RotateLeft32<2>(val), return VectorXor(RotateLeft32<2>(val),
VectorAnd(RotateLeft32<1>(val), RotateLeft32<8>(val))); VectorAnd(RotateLeft32<1>(val), RotateLeft32<8>(val)));
} }
inline void SIMON64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
@ -622,7 +622,7 @@ inline void SIMON64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1,
block1 = (uint32x4_p)vec_perm(x1, y1, m4); block1 = (uint32x4_p)vec_perm(x1, y1, m4);
} }
inline void SIMON64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
const word32 *subkeys, unsigned int rounds) const word32 *subkeys, unsigned int rounds)
{ {
#if (CRYPTOPP_BIG_ENDIAN) #if (CRYPTOPP_BIG_ENDIAN)
@ -680,7 +680,7 @@ inline void SIMON64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
block1 = (uint32x4_p)vec_perm(x1, y1, m4); block1 = (uint32x4_p)vec_perm(x1, y1, m4);
} }
inline void SIMON64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4, uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4,
uint32x4_p &block5, const word32 *subkeys, unsigned int rounds) uint32x4_p &block5, const word32 *subkeys, unsigned int rounds)
{ {
@ -753,7 +753,7 @@ inline void SIMON64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
block5 = (uint32x4_p)vec_perm(x3, y3, m4); block5 = (uint32x4_p)vec_perm(x3, y3, m4);
} }
inline void SIMON64_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1, CRYPTOPP_INLINE void SIMON64_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4, uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4,
uint32x4_p &block5, const word32 *subkeys, unsigned int rounds) uint32x4_p &block5, const word32 *subkeys, unsigned int rounds)
{ {

4
speck128_simd.cpp
View File

@ -313,7 +313,7 @@ CRYPTOPP_INLINE __m128i RotateRight64(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
CRYPTOPP_INLINE __m128i RotateLeft64<8>(const __m128i& val) __m128i RotateLeft64<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi64(val, 8); return _mm_roti_epi64(val, 8);
@ -325,7 +325,7 @@ CRYPTOPP_INLINE __m128i RotateLeft64<8>(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
CRYPTOPP_INLINE __m128i RotateRight64<8>(const __m128i& val) __m128i RotateRight64<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi64(val, 64-8); return _mm_roti_epi64(val, 64-8);

4
speck64_simd.cpp
View File

@ -299,7 +299,7 @@ CRYPTOPP_INLINE __m128i RotateRight32(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
CRYPTOPP_INLINE __m128i RotateLeft32<8>(const __m128i& val) __m128i RotateLeft32<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, 8); return _mm_roti_epi32(val, 8);
@ -311,7 +311,7 @@ CRYPTOPP_INLINE __m128i RotateLeft32<8>(const __m128i& val)
// Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks. // Faster than two Shifts and an Or. Thanks to Louis Wingers and Bryan Weeks.
template <> template <>
CRYPTOPP_INLINE __m128i RotateRight32<8>(const __m128i& val) __m128i RotateRight32<8>(const __m128i& val)
{ {
#if defined(__XOP__) #if defined(__XOP__)
return _mm_roti_epi32(val, 32-8); return _mm_roti_epi32(val, 32-8);