path: root/vmac.cpp

// vmac.cpp - originally written and placed in the public domain by Wei Dai
// based on Ted Krovetz's public domain vmac.c and draft-krovetz-vmac-01.txt

#include "pch.h"
#include "config.h"

#include "vmac.h"
#include "cpu.h"
#include "argnames.h"
#include "secblock.h"

#if defined(_MSC_VER) && !CRYPTOPP_BOOL_SLOW_WORD64
#include <intrin.h>
#endif

#if defined(CRYPTOPP_DISABLE_VMAC_ASM)
# undef CRYPTOPP_X86_ASM_AVAILABLE
# undef CRYPTOPP_X32_ASM_AVAILABLE
# undef CRYPTOPP_X64_ASM_AVAILABLE
# undef CRYPTOPP_SSE2_ASM_AVAILABLE
#endif

#if CRYPTOPP_MSC_VERSION
# pragma warning(disable: 4731)
#endif

ANONYMOUS_NAMESPACE_BEGIN

#if defined(CRYPTOPP_WORD128_AVAILABLE) && !defined(CRYPTOPP_X64_ASM_AVAILABLE)
using CryptoPP::word128;
using CryptoPP::word64;
# define VMAC_BOOL_WORD128 1
#else
using CryptoPP::word64;
# define VMAC_BOOL_WORD128 0
#endif

#ifdef __BORLANDC__
#define const	// Turbo C++ 2006 workaround
#endif
const word64 p64   = W64LIT(0xfffffffffffffeff);  /* 2^64 - 257 prime  */
const word64 m62   = W64LIT(0x3fffffffffffffff);  /* 62-bit mask       */
const word64 m63   = W64LIT(0x7fffffffffffffff);  /* 63-bit mask       */
const word64 m64   = W64LIT(0xffffffffffffffff);  /* 64-bit mask       */
const word64 mpoly = W64LIT(0x1fffffff1fffffff);  /* Poly key mask     */
#ifdef __BORLANDC__
#undef const
#endif

#if VMAC_BOOL_WORD128
// workaround GCC Bug 31690: ICE with const __uint128_t and C++ front-end
# if defined(__powerpc__) && defined (CRYPTOPP_GCC_VERSION) && (CRYPTOPP_GCC_VERSION < 50300)
#  define m126				((word128(m62)<<64)|m64)
# else
const word128 m126 = (word128(m62)<<64)|m64;		 /* 126-bit mask      */
# endif
#endif

ANONYMOUS_NAMESPACE_END

NAMESPACE_BEGIN(CryptoPP)

void VMAC_Base::UncheckedSetKey(const byte *userKey, unsigned int keylength, const NameValuePairs &params)
{
	int digestLength = params.GetIntValueWithDefault(Name::DigestSize(), DefaultDigestSize());
	if (digestLength != 8 && digestLength != 16)
		throw InvalidArgument("VMAC: DigestSize must be 8 or 16");
	m_is128 = digestLength == 16;

	m_L1KeyLength = params.GetIntValueWithDefault(Name::L1KeyLength(), 128);
	if (m_L1KeyLength <= 0 || m_L1KeyLength % 128 != 0)
		throw InvalidArgument("VMAC: L1KeyLength must be a positive multiple of 128");

	AllocateBlocks();

	BlockCipher &cipher = AccessCipher();
	cipher.SetKey(userKey, keylength, params);
	const unsigned int blockSize = cipher.BlockSize();
	const unsigned int blockSizeInWords = blockSize / sizeof(word64);
	SecBlock<word64, AllocatorWithCleanup<word64, true> > out(blockSizeInWords);
	AlignedSecByteBlock in;
	in.CleanNew(blockSize);
	size_t i;

	/* Fill nh key */
	in[0] = 0x80;
	cipher.AdvancedProcessBlocks(in, NULLPTR, (byte *)m_nhKey(), m_nhKeySize()*sizeof(word64), cipher.BT_InBlockIsCounter);
	ConditionalByteReverse<word64>(BIG_ENDIAN_ORDER, m_nhKey(), m_nhKey(), m_nhKeySize()*sizeof(word64));

	/* Fill poly key */
	in[0] = 0xC0;
	in[15] = 0;
	for (i = 0; i <= (size_t)m_is128; i++)
	{
		cipher.ProcessBlock(in, out.BytePtr());
		m_polyState()[i*4+2] = GetWord<word64>(true, BIG_ENDIAN_ORDER, out.BytePtr()) & mpoly;
		m_polyState()[i*4+3]  = GetWord<word64>(true, BIG_ENDIAN_ORDER, out.BytePtr()+8) & mpoly;
		in[15]++;
	}

	/* Fill ip key */
	in[0] = 0xE0;
	in[15] = 0;
	word64 *l3Key = m_l3Key();
	CRYPTOPP_ASSERT(IsAlignedOn(l3Key,GetAlignmentOf<word64>()));

	for (i = 0; i <= (size_t)m_is128; i++)
		do
		{
			cipher.ProcessBlock(in, out.BytePtr());
			l3Key[i*2+0] = GetWord<word64>(true, BIG_ENDIAN_ORDER, out.BytePtr());
			l3Key[i*2+1] = GetWord<word64>(true, BIG_ENDIAN_ORDER, out.BytePtr()+8);
			in[15]++;
		} while ((l3Key[i*2+0] >= p64) || (l3Key[i*2+1] >= p64));

	m_padCached = false;
	size_t nonceLength;
	const byte *nonce = GetIVAndThrowIfInvalid(params, nonceLength);
	Resynchronize(nonce, (int)nonceLength);
}

void VMAC_Base::GetNextIV(RandomNumberGenerator &rng, byte *IV)
{
	SimpleKeyingInterface::GetNextIV(rng, IV);
	IV[0] &= 0x7f;
}

void VMAC_Base::Resynchronize(const byte *nonce, int len)
{
	size_t length = ThrowIfInvalidIVLength(len);
	size_t s = IVSize();
	byte *storedNonce = m_nonce();

	if (m_is128)
	{
		std::memset(storedNonce, 0, s-length);
		std::memcpy(storedNonce+s-length, nonce, length);
		AccessCipher().ProcessBlock(storedNonce, m_pad());
	}
	else
	{
		if (m_padCached && (storedNonce[s-1] | 1) == (nonce[length-1] | 1))
		{
			m_padCached = VerifyBufsEqual(storedNonce+s-length, nonce, length-1);
			for (size_t i=0; m_padCached && i<s-length; i++)
				m_padCached = (storedNonce[i] == 0);
		}
		if (!m_padCached)
		{
			std::memset(storedNonce, 0, s-length);
			std::memcpy(storedNonce+s-length, nonce, length-1);
			storedNonce[s-1] = nonce[length-1] & 0xfe;
			AccessCipher().ProcessBlock(storedNonce, m_pad());
			m_padCached = true;
		}
		storedNonce[s-1] = nonce[length-1];
	}
	m_isFirstBlock = true;
	Restart();
}

void VMAC_Base::HashEndianCorrectedBlock(const word64 *data)
{
	CRYPTOPP_UNUSED(data);
	CRYPTOPP_ASSERT(false);
	throw NotImplemented("VMAC: HashEndianCorrectedBlock is not implemented");
}

unsigned int VMAC_Base::OptimalDataAlignment() const
{
	return
#if CRYPTOPP_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)
		HasSSE2() ? 16 :
#endif
		GetCipher().OptimalDataAlignment();
}

#if CRYPTOPP_SSE2_ASM_AVAILABLE && CRYPTOPP_BOOL_X86
#if CRYPTOPP_MSC_VERSION
# pragma warning(disable: 4731)	// frame pointer register 'ebp' modified by inline assembly code
#endif

CRYPTOPP_NOINLINE
void VMAC_Base::VHASH_Update_SSE2(const word64 *data, size_t blocksRemainingInWord64, int tagPart)
{
	const word64 *nhK = m_nhKey();
	word64 *polyS = (word64*)(void*)m_polyState();
	word32 L1KeyLength = m_L1KeyLength;

	// These are used in the ASM, but some analysis services miss it.
	CRYPTOPP_UNUSED(data); CRYPTOPP_UNUSED(tagPart);
	CRYPTOPP_UNUSED(L1KeyLength);
	CRYPTOPP_UNUSED(blocksRemainingInWord64);

	// This inline ASM is tricky, and down right difficult on 32-bit when
	// PIC is in effect. The ASM uses all the general purpose registers
	// and all the XMM registers on 32-bit machines. When PIC is in effect
	// on a 32-bit machine, GCC uses EBX as a base register for PLT. Saving
	// EBX with 'mov %%ebx, %0' and restoring EBX with 'mov %0, %%ebx'
	// causes GCC to generate 'mov -0x40(%ebx), %ebx' for the restore. That
	// obviously won't work because EBX is no longer valid. We can push and
	// pop EBX, but that breaks the stack-based references. Attempting to
	// sidestep with clobber lists results in "error: ‘asm’ operand has
	// impossible constraints". Eventually, we found we could save EBX to
	// ESP-20, which is one word below our stack in the frame.
#ifdef __GNUC__
	__asm__ __volatile__
	(
# if CRYPTOPP_BOOL_X86
	// Hack. Save EBX for PIC. Do NOT 'push EBX' here.
	// GCC issues 'mov ESP+8, EBX' to load L1KeyLength.
	// A push breaks the reference to L1KeyLength.
	AS2(	mov 	%%ebx, -20(%%esp))
# endif
	// L1KeyLength into EBX.
	// GCC generates 'mov ESP+8, EBX'.
	AS2(	mov 	%0, %%ebx)
	INTEL_NOPREFIX
#else
	#if defined(__INTEL_COMPILER)
	char isFirstBlock = m_isFirstBlock;
	AS2(	mov 	ebx, [L1KeyLength])
	AS2(	mov 	dl, [isFirstBlock])
	#else
	AS2(	mov 	ecx, this)
	AS2(	mov 	ebx, [ecx+m_L1KeyLength])
	AS2(	mov 	dl, [ecx+m_isFirstBlock])
	#endif
	AS2(	mov 	eax, tagPart)
	AS2(	shl 	eax, 4)
	AS2(	mov 	edi, nhK)
	AS2(	add 	edi, eax)
	AS2(	add 	eax, eax)
	AS2(	add 	eax, polyS)

	AS2(	mov 	esi, data)
	AS2(	mov 	ecx, blocksRemainingInWord64)
#endif

	AS2(	shr 	ebx, 3)
	AS_PUSH_IF86(	bp)
	AS2(	sub 	esp, 12)
	ASL(4)
	AS2(	mov 	ebp, ebx)
	AS2(	cmp 	ecx, ebx)
	AS2(	cmovl	ebp, ecx)
	AS2(	sub 	ecx, ebp)
	AS2(	lea 	ebp, [edi+8*ebp])	// end of nhK
	AS2(	movq	mm6, [esi])
	AS2(	paddq	mm6, [edi])
	AS2(	movq	mm5, [esi+8])
	AS2(	paddq	mm5, [edi+8])
	AS2(	add 	esi, 16)
	AS2(	add 	edi, 16)
	AS2(	movq	mm4, mm6)
	ASS(	pshufw	mm2, mm6, 1, 0, 3, 2)
	AS2(	pmuludq	mm6, mm5)
	ASS(	pshufw	mm3, mm5, 1, 0, 3, 2)
	AS2(	pmuludq	mm5, mm2)
	AS2(	pmuludq	mm2, mm3)
	AS2(	pmuludq	mm3, mm4)
	AS2(	pxor	mm7, mm7)
	AS2(	movd	[esp], mm6)
	AS2(	psrlq	mm6, 32)
	AS2(	movd	[esp+4], mm5)
	AS2(	psrlq	mm5, 32)
	AS2(	cmp 	edi, ebp)
	ASJ(	je,  	1, f)
	ASL(0)
	AS2(	movq	mm0, [esi])
	AS2(	paddq	mm0, [edi])
	AS2(	movq	mm1, [esi+8])
	AS2(	paddq	mm1, [edi+8])
	AS2(	add 	esi, 16)
	AS2(	add 	edi, 16)
	AS2(	movq	mm4, mm0)
	AS2(	paddq	mm5, mm2)
	ASS(	pshufw	mm2, mm0, 1, 0, 3, 2)
	AS2(	pmuludq	mm0, mm1)
	AS2(	movd	[esp+8], mm3)
	AS2(	psrlq	mm3, 32)
	AS2(	paddq	mm5, mm3)
	ASS(	pshufw	mm3, mm1, 1, 0, 3, 2)
	AS2(	pmuludq	mm1, mm2)
	AS2(	pmuludq	mm2, mm3)
	AS2(	pmuludq	mm3, mm4)
	AS2(	movd	mm4, [esp])
	AS2(	paddq	mm7, mm4)
	AS2(	movd	mm4, [esp+4])
	AS2(	paddq	mm6, mm4)
	AS2(	movd	mm4, [esp+8])
	AS2(	paddq	mm6, mm4)
	AS2(	movd	[esp], mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	paddq	mm6, mm0)
	AS2(	movd	[esp+4], mm1)
	AS2(	psrlq	mm1, 32)
	AS2(	paddq	mm5, mm1)
	AS2(	cmp 	edi, ebp)
	ASJ(	jne,	0, b)
	ASL(1)
	AS2(	paddq	mm5, mm2)
	AS2(	movd	[esp+8], mm3)
	AS2(	psrlq	mm3, 32)
	AS2(	paddq	mm5, mm3)
	AS2(	movd	mm4, [esp])
	AS2(	paddq	mm7, mm4)
	AS2(	movd	mm4, [esp+4])
	AS2(	paddq	mm6, mm4)
	AS2(	movd	mm4, [esp+8])
	AS2(	paddq	mm6, mm4)
	AS2(	lea 	ebp, [8*ebx])
	AS2(	sub 	edi, ebp)		// reset edi to start of nhK

	AS2(	movd	[esp], mm7)
	AS2(	psrlq	mm7, 32)
	AS2(	paddq	mm6, mm7)
	AS2(	movd	[esp+4], mm6)
	AS2(	psrlq	mm6, 32)
	AS2(	paddq	mm5, mm6)
	AS2(	psllq	mm5, 2)
	AS2(	psrlq	mm5, 2)

#define a0 [eax+2*4]
#define a1 [eax+3*4]
#define a2 [eax+0*4]
#define a3 [eax+1*4]
#define k0 [eax+2*8+2*4]
#define k1 [eax+2*8+3*4]
#define k2 [eax+2*8+0*4]
#define k3 [eax+2*8+1*4]

	AS2(	test	dl, dl)
	ASJ(	jz,  	2, f)
	AS2(	movd	mm1, k0)
	AS2(	movd	mm0, [esp])
	AS2(	paddq	mm0, mm1)
	AS2(	movd	a0, mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	movd	mm1, k1)
	AS2(	movd	mm2, [esp+4])
	AS2(	paddq	mm1, mm2)
	AS2(	paddq	mm0, mm1)
	AS2(	movd	a1, mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	paddq	mm5, k2)
	AS2(	paddq	mm0, mm5)
	AS2(	movq	a2, mm0)
	AS2(	xor 	edx, edx)
	ASJ(	jmp,	3, f)
	ASL(2)
	AS2(	movd	mm0, a3)
	AS2(	movq	mm4, mm0)
	AS2(	pmuludq	mm0, k3)		// a3*k3
	AS2(	movd	mm1, a0)
	AS2(	pmuludq	mm1, k2)		// a0*k2
	AS2(	movd	mm2, a1)
	AS2(	movd	mm6, k1)
	AS2(	pmuludq	mm2, mm6)		// a1*k1
	AS2(	movd	mm3, a2)
	AS2(	psllq	mm0, 1)
	AS2(	paddq	mm0, mm5)
	AS2(	movq	mm5, mm3)
	AS2(	movd	mm7, k0)
	AS2(	pmuludq	mm3, mm7)		// a2*k0
	AS2(	pmuludq	mm4, mm7)		// a3*k0
	AS2(	pmuludq	mm5, mm6)		// a2*k1
	AS2(	paddq	mm0, mm1)
	AS2(	movd	mm1, a1)
	AS2(	paddq	mm4, mm5)
	AS2(	movq	mm5, mm1)
	AS2(	pmuludq	mm1, k2)		// a1*k2
	AS2(	paddq	mm0, mm2)
	AS2(	movd	mm2, a0)
	AS2(	paddq	mm0, mm3)
	AS2(	movq	mm3, mm2)
	AS2(	pmuludq	mm2, k3)		// a0*k3
	AS2(	pmuludq	mm3, mm7)		// a0*k0
	AS2(	movd	[esp+8], mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	pmuludq	mm7, mm5)		// a1*k0
	AS2(	pmuludq	mm5, k3)		// a1*k3
	AS2(	paddq	mm0, mm1)
	AS2(	movd	mm1, a2)
	AS2(	pmuludq	mm1, k2)		// a2*k2
	AS2(	paddq	mm0, mm2)
	AS2(	paddq	mm0, mm4)
	AS2(	movq	mm4, mm0)
	AS2(	movd	mm2, a3)
	AS2(	pmuludq	mm2, mm6)		// a3*k1
	AS2(	pmuludq	mm6, a0)		// a0*k1
	AS2(	psrlq	mm0, 31)
	AS2(	paddq	mm0, mm3)
	AS2(	movd	mm3, [esp])
	AS2(	paddq	mm0, mm3)
	AS2(	movd	mm3, a2)
	AS2(	pmuludq	mm3, k3)		// a2*k3
	AS2(	paddq	mm5, mm1)
	AS2(	movd	mm1, a3)
	AS2(	pmuludq	mm1, k2)		// a3*k2
	AS2(	paddq	mm5, mm2)
	AS2(	movd	mm2, [esp+4])
	AS2(	psllq	mm5, 1)
	AS2(	paddq	mm0, mm5)
	AS2(	psllq	mm4, 33)
	AS2(	movd	a0, mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	paddq	mm6, mm7)
	AS2(	movd	mm7, [esp+8])
	AS2(	paddq	mm0, mm6)
	AS2(	paddq	mm0, mm2)
	AS2(	paddq	mm3, mm1)
	AS2(	psllq	mm3, 1)
	AS2(	paddq	mm0, mm3)
	AS2(	psrlq	mm4, 1)
	AS2(	movd	a1, mm0)
	AS2(	psrlq	mm0, 32)
	AS2(	por 	mm4, mm7)
	AS2(	paddq	mm0, mm4)
	AS2(	movq	a2, mm0)

#undef a0
#undef a1
#undef a2
#undef a3
#undef k0
#undef k1
#undef k2
#undef k3

	ASL(3)
	AS2(	test	ecx, ecx)
	ASJ(	jnz,	4, b)
	AS2(	add 	esp, 12)
	AS_POP_IF86(	bp)
	AS1(	emms)
#ifdef __GNUC__
	ATT_PREFIX
# if CRYPTOPP_BOOL_X86
	// Restore EBX for PIC
	AS2(	mov 	-20(%%esp), %%ebx)
# endif
		:
		: "m" (L1KeyLength), "c" (blocksRemainingInWord64), "S" (data),
		  "D" (nhK+tagPart*2), "d" (m_isFirstBlock), "a" (polyS+tagPart*4)
		: "memory", "cc"
	);
#endif
}
#endif

#if VMAC_BOOL_WORD128
	#define DeclareNH(a) word128 a=0
	#define MUL64(rh,rl,i1,i2) {word128 p = word128(i1)*(i2); rh = word64(p>>64); rl = word64(p);}
	#define AccumulateNH(a, b, c) a += word128(b)*(c)
	#define Multiply128(r, i1, i2) r = word128(word64(i1)) * word64(i2)
#else
	#if _MSC_VER >= 1400 && !defined(__INTEL_COMPILER) && (defined(_M_IX86) || defined(_M_X64) || defined(_M_IA64))
		#define MUL32(a, b) __emulu(word32(a), word32(b))
	#else
		#define MUL32(a, b) ((word64)((word32)(a)) * (word32)(b))
	#endif
	#if defined(CRYPTOPP_X64_ASM_AVAILABLE)
		#define DeclareNH(a)			word64 a##0=0, a##1=0
		#define MUL64(rh,rl,i1,i2)		asm ("mulq %3" : "=a"(rl), "=d"(rh) : "a"(i1), "g"(i2) : "cc");
		#define AccumulateNH(a, b, c)	asm ("mulq %3; addq %%rax, %0; adcq %%rdx, %1" : "+r"(a##0), "+r"(a##1) : "a"(b), "g"(c) : "%rdx", "cc");
		#define ADD128(rh,rl,ih,il)     asm ("addq %3, %1; adcq %2, %0" : "+r"(rh),"+r"(rl) : "r"(ih),"r"(il) : "cc");
	#elif defined(_MSC_VER) && !CRYPTOPP_BOOL_SLOW_WORD64
		#define DeclareNH(a) word64 a##0=0, a##1=0
		#define MUL64(rh,rl,i1,i2)   (rl) = _umul128(i1,i2,&(rh));
		#define AccumulateNH(a, b, c)	{\
			word64 ph, pl;\
			pl = _umul128(b,c,&ph);\
			a##0 += pl;\
			a##1 += ph + (a##0 < pl);}
	#else
		#define VMAC_BOOL_32BIT 1
		#define DeclareNH(a) word64 a##0=0, a##1=0, a##2=0
		#define MUL64(rh,rl,i1,i2)                                               \
			{   word64 _i1 = (i1), _i2 = (i2);                                 \
				word64 m1= MUL32(_i1,_i2>>32);                                 \
				word64 m2= MUL32(_i1>>32,_i2);                                 \
				rh         = MUL32(_i1>>32,_i2>>32);                             \
				rl         = MUL32(_i1,_i2);                                     \
				ADD128(rh,rl,(m1 >> 32),(m1 << 32));                             \
				ADD128(rh,rl,(m2 >> 32),(m2 << 32));                             \
			}
		#define AccumulateNH(a, b, c)	{\
			word64 p = MUL32(b, c);\
			a##1 += word32((p)>>32);\
			a##0 += word32(p);\
			p = MUL32((b)>>32, c);\
			a##2 += word32((p)>>32);\
			a##1 += word32(p);\
			p = MUL32((b)>>32, (c)>>32);\
			a##2 += p;\
			p = MUL32(b, (c)>>32);\
			a##1 += word32(p);\
			a##2 += word32(p>>32);}
	#endif
#endif
#ifndef VMAC_BOOL_32BIT
	#define VMAC_BOOL_32BIT 0
#endif
#ifndef ADD128
	#define ADD128(rh,rl,ih,il)                                          \
		{   word64 _il = (il);                                         \
			(rl) += (_il);                                               \
			(rh) += (ih) + ((rl) < (_il));                               \
		}
#endif

template <bool T_128BitTag>
void VMAC_Base::VHASH_Update_Template(const word64 *data, size_t blocksRemainingInWord64)
{
	CRYPTOPP_ASSERT(IsAlignedOn(m_polyState(),GetAlignmentOf<word64>()));
	CRYPTOPP_ASSERT(IsAlignedOn(m_nhKey(),GetAlignmentOf<word64>()));

	#define INNER_LOOP_ITERATION(j)	{\
		word64 d0 = ConditionalByteReverse(LITTLE_ENDIAN_ORDER, data[i+2*j+0]);\
		word64 d1 = ConditionalByteReverse(LITTLE_ENDIAN_ORDER, data[i+2*j+1]);\
		AccumulateNH(nhA, d0+nhK[i+2*j+0], d1+nhK[i+2*j+1]);\
		if (T_128BitTag)\
			AccumulateNH(nhB, d0+nhK[i+2*j+2], d1+nhK[i+2*j+3]);\
		}

	size_t L1KeyLengthInWord64 = m_L1KeyLength / 8;
	size_t innerLoopEnd = L1KeyLengthInWord64;
	const word64 *nhK = m_nhKey();
	word64 *polyS = (word64*)(void*)m_polyState();
	bool isFirstBlock = true;
	size_t i;

	#if !VMAC_BOOL_32BIT
		#if VMAC_BOOL_WORD128
			word128 a1=0, a2=0;
		#else
			word64 ah1=0, al1=0, ah2=0, al2=0;
		#endif
		word64 kh1, kl1, kh2, kl2;
		kh1=(polyS+0*4+2)[0]; kl1=(polyS+0*4+2)[1];
		if (T_128BitTag)
		{
			kh2=(polyS+1*4+2)[0]; kl2=(polyS+1*4+2)[1];
		}
	#endif

	do
	{
		DeclareNH(nhA);
		DeclareNH(nhB);

		i = 0;
		if (blocksRemainingInWord64 < L1KeyLengthInWord64)
		{
			if (blocksRemainingInWord64 % 8)
			{
				innerLoopEnd = blocksRemainingInWord64 % 8;
				for (; i<innerLoopEnd; i+=2)
					INNER_LOOP_ITERATION(0);
			}
			innerLoopEnd = blocksRemainingInWord64;
		}
		for (; i<innerLoopEnd; i+=8)
		{
			INNER_LOOP_ITERATION(0);
			INNER_LOOP_ITERATION(1);
			INNER_LOOP_ITERATION(2);
			INNER_LOOP_ITERATION(3);
		}
		blocksRemainingInWord64 -= innerLoopEnd;
		data += innerLoopEnd;

		#if VMAC_BOOL_32BIT
			word32 nh0[2],  nh1[2];
			word64 nh2[2];

			nh0[0] = word32(nhA0);
			nhA1 += (nhA0 >> 32);
			nh1[0] = word32(nhA1);
			nh2[0] = (nhA2 + (nhA1 >> 32)) & m62;

			if (T_128BitTag)
			{
				nh0[1] = word32(nhB0);
				nhB1 += (nhB0 >> 32);
				nh1[1] = word32(nhB1);
				nh2[1] = (nhB2 + (nhB1 >> 32)) & m62;
			}

			#define a0 (((word32 *)(polyS+i*4))[2+NativeByteOrder::ToEnum()])
			#define a1 (*(((word32 *)(polyS+i*4))+3-NativeByteOrder::ToEnum()))		// workaround for GCC 3.2
			#define a2 (((word32 *)(polyS+i*4))[0+NativeByteOrder::ToEnum()])
			#define a3 (*(((word32 *)(polyS+i*4))+1-NativeByteOrder::ToEnum()))
			#define aHi ((polyS+i*4)[0])
			#define k0 (((word32 *)(polyS+i*4+2))[2+NativeByteOrder::ToEnum()])
			#define k1 (*(((word32 *)(polyS+i*4+2))+3-NativeByteOrder::ToEnum()))
			#define k2 (((word32 *)(polyS+i*4+2))[0+NativeByteOrder::ToEnum()])
			#define k3 (*(((word32 *)(polyS+i*4+2))+1-NativeByteOrder::ToEnum()))
			#define kHi ((polyS+i*4+2)[0])

			if (isFirstBlock)
			{
				isFirstBlock = false;
				if (m_isFirstBlock)
				{
					m_isFirstBlock = false;
					for (i=0; i<=(size_t)T_128BitTag; i++)
					{
						word64 t = (word64)nh0[i] + k0;
						a0 = (word32)t;
						t = (t >> 32) + nh1[i] + k1;
						a1 = (word32)t;
						aHi = (t >> 32) + nh2[i] + kHi;
					}
					continue;
				}
			}
			for (i=0; i<=(size_t)T_128BitTag; i++)
			{
				word64 p, t;
				word32 t2;

				p = MUL32(a3, 2*k3);
				p += nh2[i];
				p += MUL32(a0, k2);
				p += MUL32(a1, k1);
				p += MUL32(a2, k0);
				t2 = (word32)p;
				p >>= 32;
				p += MUL32(a0, k3);
				p += MUL32(a1, k2);
				p += MUL32(a2, k1);
				p += MUL32(a3, k0);
				t = (word64(word32(p) & 0x7fffffff) << 32) | t2;
				p >>= 31;
				p += nh0[i];
				p += MUL32(a0, k0);
				p += MUL32(a1, 2*k3);
				p += MUL32(a2, 2*k2);
				p += MUL32(a3, 2*k1);
				t2 = (word32)p;
				p >>= 32;
				p += nh1[i];
				p += MUL32(a0, k1);
				p += MUL32(a1, k0);
				p += MUL32(a2, 2*k3);
				p += MUL32(a3, 2*k2);
				a0 = t2;
				a1 = (word32)p;
				aHi = (p >> 32) + t;
			}

			#undef a0
			#undef a1
			#undef a2
			#undef a3
			#undef aHi
			#undef k0
			#undef k1
			#undef k2
			#undef k3
			#undef kHi
		#else		// #if VMAC_BOOL_32BIT
			if (isFirstBlock)
			{
				isFirstBlock = false;
				if (m_isFirstBlock)
				{
					m_isFirstBlock = false;
					#if VMAC_BOOL_WORD128
						#define first_poly_step(a, kh, kl, m)	a = (m & m126) + ((word128(kh) << 64) | kl)

						first_poly_step(a1, kh1, kl1, nhA);
						if (T_128BitTag)
							first_poly_step(a2, kh2, kl2, nhB);
					#else
						#define first_poly_step(ah, al, kh, kl, mh, ml)		{\
							mh &= m62;\
							ADD128(mh, ml, kh, kl);	\
							ah = mh; al = ml;}

						first_poly_step(ah1, al1, kh1, kl1, nhA1, nhA0);
						if (T_128BitTag)
							first_poly_step(ah2, al2, kh2, kl2, nhB1, nhB0);
					#endif
					continue;
				}
				else
				{
					#if VMAC_BOOL_WORD128
						a1 = (word128((polyS+0*4)[0]) << 64) | (polyS+0*4)[1];
					#else
						ah1=(polyS+0*4)[0]; al1=(polyS+0*4)[1];
					#endif
					if (T_128BitTag)
					{
						#if VMAC_BOOL_WORD128
							a2 = (word128((polyS+1*4)[0]) << 64) | (polyS+1*4)[1];
						#else
							ah2=(polyS+1*4)[0]; al2=(polyS+1*4)[1];
						#endif
					}
				}
			}

			#if VMAC_BOOL_WORD128
				#define poly_step(a, kh, kl, m)	\
				{   word128 t1, t2, t3, t4;\
					Multiply128(t2, a>>64, kl);\
					Multiply128(t3, a, kh);\
					Multiply128(t1, a, kl);\
					Multiply128(t4, a>>64, 2*kh);\
					t2 += t3;\
					t4 += t1;\
					t2 += t4>>64;\
					a = (word128(word64(t2)&m63) << 64) | word64(t4);\
					t2 *= 2;\
					a += m & m126;\
					a += t2>>64;}

				poly_step(a1, kh1, kl1, nhA);
				if (T_128BitTag)
					poly_step(a2, kh2, kl2, nhB);
			#else
				#define poly_step(ah, al, kh, kl, mh, ml)					\
				{   word64 t1h, t1l, t2h, t2l, t3h, t3l, z=0;				\
					/* compute ab*cd, put bd into result registers */       \
					MUL64(t2h,t2l,ah,kl);                                   \
					MUL64(t3h,t3l,al,kh);                                   \
					MUL64(t1h,t1l,ah,2*kh);                                 \
					MUL64(ah,al,al,kl);                                     \
					/* add together ad + bc */                              \
					ADD128(t2h,t2l,t3h,t3l);                                \
					/* add 2 * ac to result */                              \
					ADD128(ah,al,t1h,t1l);                                  \
					/* now (ah,al), (t2l,2*t2h) need summing */             \
					/* first add the high registers, carrying into t2h */   \
					ADD128(t2h,ah,z,t2l);                                   \
					/* double t2h and add top bit of ah */                  \
					t2h += t2h + (ah >> 63);                                \
					ah &= m63;                                              \
					/* now add the low registers */                         \
					mh &= m62;												\
					ADD128(ah,al,mh,ml);                                    \
					ADD128(ah,al,z,t2h);                                    \
				}

				poly_step(ah1, al1, kh1, kl1, nhA1, nhA0);
				if (T_128BitTag)
					poly_step(ah2, al2, kh2, kl2, nhB1, nhB0);
			#endif
		#endif		// #if VMAC_BOOL_32BIT
	} while (blocksRemainingInWord64);

	#if VMAC_BOOL_WORD128
		(polyS+0*4)[0]=word64(a1>>64); (polyS+0*4)[1]=word64(a1);
		if (T_128BitTag)
		{
			(polyS+1*4)[0]=word64(a2>>64); (polyS+1*4)[1]=word64(a2);
		}
	#elif !VMAC_BOOL_32BIT
		(polyS+0*4)[0]=ah1; (polyS+0*4)[1]=al1;
		if (T_128BitTag)
		{
			(polyS+1*4)[0]=ah2; (polyS+1*4)[1]=al2;
		}
	#endif
}

inline void VMAC_Base::VHASH_Update(const word64 *data, size_t blocksRemainingInWord64)
{
#if CRYPTOPP_SSE2_ASM_AVAILABLE && CRYPTOPP_BOOL_X86
	if (HasSSE2())
	{
		VHASH_Update_SSE2(data, blocksRemainingInWord64, 0);
		if (m_is128)
			VHASH_Update_SSE2(data, blocksRemainingInWord64, 1);
		m_isFirstBlock = false;
	}
	else
#endif
	{
		if (m_is128)
			VHASH_Update_Template<true>(data, blocksRemainingInWord64);
		else
			VHASH_Update_Template<false>(data, blocksRemainingInWord64);
	}
}

size_t VMAC_Base::HashMultipleBlocks(const word64 *data, size_t length)
{
	size_t remaining = ModPowerOf2(length, m_L1KeyLength);
	VHASH_Update(data, (length-remaining)/8);
	return remaining;
}

word64 L3Hash(const word64 *input, const word64 *l3Key, size_t len)
{
    word64 rh, rl, t, z=0;
	word64 p1 = input[0], p2 = input[1];
	word64 k1 = l3Key[0], k2 = l3Key[1];

    /* fully reduce (p1,p2)+(len,0) mod p127 */
    t = p1 >> 63;
    p1 &= m63;
    ADD128(p1, p2, len, t);
    /* At this point, (p1,p2) is at most 2^127+(len<<64) */
    t = (p1 > m63) + ((p1 == m63) & (p2 == m64));
    ADD128(p1, p2, z, t);
    p1 &= m63;

    /* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
    t = p1 + (p2 >> 32);
    t += (t >> 32);
    t += (word32)t > 0xfffffffeU;
    p1 += (t >> 32);
    p2 += (p1 << 32);

    /* compute (p1+k1)%p64 and (p2+k2)%p64 */
    p1 += k1;
    p1 += (0 - (p1 < k1)) & 257;
    p2 += k2;
    p2 += (0 - (p2 < k2)) & 257;

    /* compute (p1+k1)*(p2+k2)%p64 */
    MUL64(rh, rl, p1, p2);
    t = rh >> 56;
    ADD128(t, rl, z, rh);
    rh <<= 8;
    ADD128(t, rl, z, rh);
    t += t << 8;
    rl += t;
    rl += (0 - (rl < t)) & 257;
    rl += (0 - (rl > p64-1)) & 257;
    return rl;
}

void VMAC_Base::TruncatedFinal(byte *mac, size_t size)
{
	CRYPTOPP_ASSERT(IsAlignedOn(DataBuf(),GetAlignmentOf<word64>()));
	CRYPTOPP_ASSERT(IsAlignedOn(m_polyState(),GetAlignmentOf<word64>()));
	size_t len = ModPowerOf2(GetBitCountLo()/8, m_L1KeyLength);

	if (len)
	{
		std::memset(m_data()+len, 0, (0-len)%16);
		VHASH_Update(DataBuf(), ((len+15)/16)*2);
		len *= 8;	// convert to bits
	}
	else if (m_isFirstBlock)
	{
		// special case for empty string
		m_polyState()[0] = m_polyState()[2];
		m_polyState()[1] = m_polyState()[3];
		if (m_is128)
		{
			m_polyState()[4] = m_polyState()[6];
			m_polyState()[5] = m_polyState()[7];
		}
	}

	if (m_is128)
	{
		word64 t[2];
		t[0] = L3Hash(m_polyState(), m_l3Key(), len) + GetWord<word64>(true, BIG_ENDIAN_ORDER, m_pad());
		t[1] = L3Hash(m_polyState()+4, m_l3Key()+2, len) + GetWord<word64>(true, BIG_ENDIAN_ORDER, m_pad()+8);
		if (size == 16)
		{
			PutWord(false, BIG_ENDIAN_ORDER, mac, t[0]);
			PutWord(false, BIG_ENDIAN_ORDER, mac+8, t[1]);
		}
		else
		{
			t[0] = ConditionalByteReverse(BIG_ENDIAN_ORDER, t[0]);
			t[1] = ConditionalByteReverse(BIG_ENDIAN_ORDER, t[1]);
			std::memcpy(mac, t, size);
		}
	}
	else
	{
		word64 t = L3Hash(m_polyState(), m_l3Key(), len);
		t += GetWord<word64>(true, BIG_ENDIAN_ORDER, m_pad() + (m_nonce()[IVSize()-1]&1) * 8);
		if (size == 8)
			PutWord(false, BIG_ENDIAN_ORDER, mac, t);
		else
		{
			t = ConditionalByteReverse(BIG_ENDIAN_ORDER, t);
			std::memcpy(mac, &t, size);
		}
	}
}

NAMESPACE_END