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

remove commented code for Power8

pull/507/head
Jeffrey Walton 2017-09-13 03:59:25 -04:00
parent 6d459afa15
commit 397ccd7e49
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
1 changed files with 23 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
rijndael-simd.cpp
View File

@ -939,16 +939,18 @@ inline VectorType VectorLoad(int off, const byte src[16])
// Loads an aligned byte array, does not perform an endian conversion. // Loads an aligned byte array, does not perform an endian conversion.
// This function presumes the subkey table is correct endianess. // This function presumes the subkey table is correct endianess.
inline VectorType VectorLoadKey(const byte vec[16]) inline VectorType VectorLoadKey(const byte src[16])
{ {
return (VectorType)vec_ld(0, vec); CRYPTOPP_ASSERT(IsAlignedOn(src, 16));
return (VectorType)vec_ld(0, src);
} }
// Loads an aligned byte array, does not perform an endian conversion. // Loads an aligned byte array, does not perform an endian conversion.
// This function presumes the subkey table is correct endianess. // This function presumes the subkey table is correct endianess.
inline VectorType VectorLoadKey(int off, const byte vec[16]) inline VectorType VectorLoadKey(int off, const byte src[16])
{ {
return (VectorType)vec_ld(off, vec); CRYPTOPP_ASSERT(IsAlignedOn(src, 16));
return (VectorType)vec_ld(off, src);
} }
// Stores to a mis-aligned byte array, performs an endian conversion. // Stores to a mis-aligned byte array, performs an endian conversion.
@ -1150,7 +1152,7 @@ size_t Rijndael_AdvancedProcessBlocks_POWER8(F1 func1, F4 func4, const word32 *s
if (flags & BlockTransformation::BT_InBlockIsCounter) if (flags & BlockTransformation::BT_InBlockIsCounter)
{ {
#if defined(IS_LITTLE_ENDIAN) #if defined(IS_LITTLE_ENDIAN)
const VectorType one = {1}; const VectorType one = (VectorType)((uint64x2_p8){1,0});
#else #else
const VectorType one = (VectorType)((uint64x2_p8){0,1}); const VectorType one = (VectorType)((uint64x2_p8){0,1});
#endif #endif
@ -1162,45 +1164,35 @@ size_t Rijndael_AdvancedProcessBlocks_POWER8(F1 func1, F4 func4, const word32 *s
} }
else else
{ {
//inBlocks += inIncrement; const int inc = static_cast<int>(inIncrement);
block1 = VectorLoad(1*inIncrement, inBlocks); block1 = VectorLoad(1*inc, inBlocks);
//inBlocks += inIncrement; block2 = VectorLoad(2*inc, inBlocks);
block2 = VectorLoad(2*inIncrement, inBlocks); block3 = VectorLoad(3*inc, inBlocks);
//inBlocks += inIncrement;
block3 = VectorLoad(3*inIncrement, inBlocks);
//inBlocks += inIncrement;
inBlocks += 4*inIncrement; inBlocks += 4*inIncrement;
} }
if (flags & BlockTransformation::BT_XorInput) if (flags & BlockTransformation::BT_XorInput)
{ {
block0 = VectorXor(block0, VectorLoad(0*xorIncrement, xorBlocks)); const int inc = static_cast<int>(xorIncrement);
//xorBlocks += xorIncrement; block0 = VectorXor(block0, VectorLoad(0*inc, xorBlocks));
block1 = VectorXor(block1, VectorLoad(1*xorIncrement, xorBlocks)); block1 = VectorXor(block1, VectorLoad(1*inc, xorBlocks));
//xorBlocks += xorIncrement; block2 = VectorXor(block2, VectorLoad(2*inc, xorBlocks));
block2 = VectorXor(block2, VectorLoad(2*xorIncrement, xorBlocks)); block3 = VectorXor(block3, VectorLoad(3*inc, xorBlocks));
//xorBlocks += xorIncrement; xorBlocks += 4*inc;
block3 = VectorXor(block3, VectorLoad(3*xorIncrement, xorBlocks));
//xorBlocks += xorIncrement;
xorBlocks += 4*xorIncrement;
} }
func4(block0, block1, block2, block3, subKeys, rounds); func4(block0, block1, block2, block3, subKeys, rounds);
if (xorBlocks && !(flags & BlockTransformation::BT_XorInput)) if (xorBlocks && !(flags & BlockTransformation::BT_XorInput))
{ {
block0 = VectorXor(block0, VectorLoad(0*xorIncrement, xorBlocks)); const int inc = static_cast<int>(xorIncrement);
//xorBlocks += xorIncrement; block0 = VectorXor(block0, VectorLoad(0*inc, xorBlocks));
block1 = VectorXor(block1, VectorLoad(1*xorIncrement, xorBlocks)); block1 = VectorXor(block1, VectorLoad(1*inc, xorBlocks));
//xorBlocks += xorIncrement; block2 = VectorXor(block2, VectorLoad(2*inc, xorBlocks));
block2 = VectorXor(block2, VectorLoad(2*xorIncrement, xorBlocks)); block3 = VectorXor(block3, VectorLoad(3*inc, xorBlocks));
//xorBlocks += xorIncrement; xorBlocks += 4*inc;
block3 = VectorXor(block3, VectorLoad(3*xorIncrement, xorBlocks));
//xorBlocks += xorIncrement;
xorBlocks += 4*xorIncrement;
} }
// I can't get Store to run faster using indexed offsets
VectorStore(block0, outBlocks); VectorStore(block0, outBlocks);
outBlocks += outIncrement; outBlocks += outIncrement;
VectorStore(block1, outBlocks); VectorStore(block1, outBlocks);