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

Add Octet suffix for vec_sld 

We need to make room for packed shifts and rotates
pull/748/head
Jeffrey Walton 2018-11-14 23:32:26 -05:00
parent d656545875
commit 3c7bdf1a26
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
3 changed files with 14 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
gcm_simd.cpp
View File

@ -753,7 +753,7 @@ uint64x2_p GCM_Reduce_VMULL(uint64x2_p c0, uint64x2_p c1, uint64x2_p c2, uint64x
{ {
const uint64x2_p m1 = {1,1}, m63 = {63,63}; const uint64x2_p m1 = {1,1}, m63 = {63,63};
c1 = VectorXor(c1, VectorShiftRight<8>(c0)); c1 = VectorXor(c1, VectorShiftRightOctet<8>(c0));
c1 = VectorXor(c1, VMULL_10LE(c0, r)); c1 = VectorXor(c1, VMULL_10LE(c0, r));
c0 = VectorXor(c1, VectorShiftLeftOctet<8>(c0)); c0 = VectorXor(c1, VectorShiftLeftOctet<8>(c0));
c0 = VMULL_00LE(vec_sl(c0, m1), r); c0 = VMULL_00LE(vec_sl(c0, m1), r);

14
ppc_simd.h
View File

@ -215,22 +215,22 @@ inline T VectorShiftLeftOctet(const T& vec)
/// \tparam T vector type /// \tparam T vector type
/// \param vec the vector /// \param vec the vector
/// \returns vector /// \returns vector
/// \details VectorShiftRight() returns a new vector after shifting the /// \details VectorShiftRightOctet() returns a new vector after shifting the
/// concatenation of the zero vector and the source vector by the specified /// concatenation of the zero vector and the source vector by the specified
/// number of bytes. The return vector is the same type as vec. /// number of bytes. The return vector is the same type as vec.
/// \details On big endian machines VectorShiftRight() is <tt>vec_sld(a, z, /// \details On big endian machines VectorShiftRightOctet() is <tt>vec_sld(a, z,
/// c)</tt>. On little endian machines VectorShiftRight() is translated to /// c)</tt>. On little endian machines VectorShiftRightOctet() is translated to
/// <tt>vec_sld(z, a, 16-c)</tt>. You should always call the function as /// <tt>vec_sld(z, a, 16-c)</tt>. You should always call the function as
/// if on a big endian machine as shown below. /// if on a big endian machine as shown below.
/// <pre> /// <pre>
/// uint8x16_p r1 = VectorLoad(ptr); /// uint8x16_p r1 = VectorLoad(ptr);
/// uint8x16_p r5 = VectorShiftRight<12>(r1); /// uint8x16_p r5 = VectorShiftRightOctet<12>(r1);
/// </pre> /// </pre>
/// \sa <A HREF="https://stackoverflow.com/q/46341923/608639">Is vec_sld /// \sa <A HREF="https://stackoverflow.com/q/46341923/608639">Is vec_sld
/// endian sensitive?</A> on Stack Overflow /// endian sensitive?</A> on Stack Overflow
/// \since Crypto++ 6.0 /// \since Crypto++ 6.0
template <unsigned int C, class T> template <unsigned int C, class T>
inline T VectorShiftRight(const T& vec) inline T VectorShiftRightOctet(const T& vec)
{ {
const T zero = {0}; const T zero = {0};
if (C >= 16) if (C >= 16)
@ -322,7 +322,7 @@ inline T VectorGetLow(const T& val)
//const T zero = {0}; //const T zero = {0};
//const uint8x16_p mask = {16,16,16,16, 16,16,16,16, 8,9,10,11, 12,13,14,15 }; //const uint8x16_p mask = {16,16,16,16, 16,16,16,16, 8,9,10,11, 12,13,14,15 };
//return (T)vec_perm(zero, val, mask); //return (T)vec_perm(zero, val, mask);
return VectorShiftRight<8>(VectorShiftLeftOctet<8>(val)); return VectorShiftRightOctet<8>(VectorShiftLeftOctet<8>(val));
} }
/// \brief Extract a dword from a vector /// \brief Extract a dword from a vector
@ -339,7 +339,7 @@ inline T VectorGetHigh(const T& val)
//const T zero = {0}; //const T zero = {0};
//const uint8x16_p mask = {16,16,16,16, 16,16,16,16, 0,1,2,3, 4,5,6,7 }; //const uint8x16_p mask = {16,16,16,16, 16,16,16,16, 0,1,2,3, 4,5,6,7 };
//return (T)vec_perm(zero, val, mask); //return (T)vec_perm(zero, val, mask);
return VectorShiftRight<8>(val); return VectorShiftRightOctet<8>(val);
} }
/// \brief Compare two vectors /// \brief Compare two vectors

12
validat1.cpp
View File

@ -1145,7 +1145,7 @@ bool TestAltivecOps()
pass2 = (VectorEqual(val, VectorShiftLeftOctet<0>(val))) && pass2; pass2 = (VectorEqual(val, VectorShiftLeftOctet<0>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
pass2 = (VectorEqual(val, VectorShiftRight<0>(val))) && pass2; pass2 = (VectorEqual(val, VectorShiftRightOctet<0>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
uint8x16_p lsh1 = {0xff,0xff,0xff,0xff, 0xff,0xff,0xff,0xff, uint8x16_p lsh1 = {0xff,0xff,0xff,0xff, 0xff,0xff,0xff,0xff,
@ -1155,7 +1155,7 @@ bool TestAltivecOps()
pass2 = (VectorEqual(lsh1, VectorShiftLeftOctet<1>(val))) && pass2; pass2 = (VectorEqual(lsh1, VectorShiftLeftOctet<1>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
pass2 = (VectorEqual(rsh1, VectorShiftRight<1>(val))) && pass2; pass2 = (VectorEqual(rsh1, VectorShiftRightOctet<1>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
uint8x16_p lsh15 = {0xff,0x00,0x00,0x00, 0x00,0x00,0x00,0x00, uint8x16_p lsh15 = {0xff,0x00,0x00,0x00, 0x00,0x00,0x00,0x00,
@ -1165,7 +1165,7 @@ bool TestAltivecOps()
pass2 = (VectorEqual(lsh15, VectorShiftLeftOctet<15>(val))) && pass2; pass2 = (VectorEqual(lsh15, VectorShiftLeftOctet<15>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
pass2 = (VectorEqual(rsh15, VectorShiftRight<15>(val))) && pass2; pass2 = (VectorEqual(rsh15, VectorShiftRightOctet<15>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
uint8x16_p lsh16 = {0x00,0x00,0x00,0x00, 0x00,0x00,0x00,0x00, uint8x16_p lsh16 = {0x00,0x00,0x00,0x00, 0x00,0x00,0x00,0x00,
@ -1175,7 +1175,7 @@ bool TestAltivecOps()
pass2 = (VectorEqual(lsh16, VectorShiftLeftOctet<16>(val))) && pass2; pass2 = (VectorEqual(lsh16, VectorShiftLeftOctet<16>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
pass2 = (VectorEqual(rsh16, VectorShiftRight<16>(val))) && pass2; pass2 = (VectorEqual(rsh16, VectorShiftRightOctet<16>(val))) && pass2;
CRYPTOPP_ASSERT(pass2); CRYPTOPP_ASSERT(pass2);
if (!pass2) if (!pass2)
@ -1199,10 +1199,10 @@ bool TestAltivecOps()
pass3 = VectorEqual(ex3, VectorGetHigh(ex1)) && pass3; pass3 = VectorEqual(ex3, VectorGetHigh(ex1)) && pass3;
CRYPTOPP_ASSERT(pass3); CRYPTOPP_ASSERT(pass3);
uint8x16_p ex4 = VectorShiftRight<8>(VectorShiftLeftOctet<8>(ex1)); uint8x16_p ex4 = VectorShiftRightOctet<8>(VectorShiftLeftOctet<8>(ex1));
pass3 = VectorEqual(ex4, VectorGetLow(ex1)) && pass3; pass3 = VectorEqual(ex4, VectorGetLow(ex1)) && pass3;
CRYPTOPP_ASSERT(pass3); CRYPTOPP_ASSERT(pass3);
uint8x16_p ex5 = VectorShiftRight<8>(ex1); uint8x16_p ex5 = VectorShiftRightOctet<8>(ex1);
pass3 = VectorEqual(ex5, VectorGetHigh(ex1)) && pass3; pass3 = VectorEqual(ex5, VectorGetHigh(ex1)) && pass3;
CRYPTOPP_ASSERT(pass3); CRYPTOPP_ASSERT(pass3);