integer.cpp

lots Elliptic curve cryptography codes. Use Visual c++ to compile

			: "%esi", "%edi", "memory", "cc" \
		);
	#define TopPrologue		MulPrologue
	#define TopEpilogue	\
			AS1(	pop		ebx) \
			".att_syntax prefix;" \
			: \
			: "d" (s_maskLow16), "c" (C), "a" (A), "D" (B), "S" (L) \
			: "memory", "cc" \
		);
#else
	#define AddPrologue \
		__asm	push edi \
		__asm	push esi \
		__asm	mov		eax, [esp+12] \
		__asm	mov		edi, [esp+16]
	#define AddEpilogue \
		__asm	pop esi \
		__asm	pop edi \
		__asm	ret 8
#if _MSC_VER < 1300
	#define SaveEBX		__asm push ebx
	#define RestoreEBX	__asm pop ebx
#else
	#define SaveEBX
	#define RestoreEBX
#endif
	#define SquPrologue					\
		AS2(	mov		eax, A)			\
		AS2(	mov		ecx, C)			\
		SaveEBX							\
		AS2(	lea		ebx, s_maskLow16)
	#define MulPrologue					\
		AS2(	mov		eax, A)			\
		AS2(	mov		edi, B)			\
		AS2(	mov		ecx, C)			\
		SaveEBX							\
		AS2(	lea		ebx, s_maskLow16)
	#define TopPrologue					\
		AS2(	mov		eax, A)			\
		AS2(	mov		edi, B)			\
		AS2(	mov		ecx, C)			\
		AS2(	mov		esi, L)			\
		SaveEBX							\
		AS2(	lea		ebx, s_maskLow16)
	#define SquEpilogue		RestoreEBX
	#define MulEpilogue		RestoreEBX
	#define TopEpilogue		RestoreEBX
#endif

#ifdef CRYPTOPP_X64_MASM_AVAILABLE
extern "C" {
int Baseline_Add(size_t N, word *C, const word *A, const word *B);
int Baseline_Sub(size_t N, word *C, const word *A, const word *B);
}
#elif defined(CRYPTOPP_X64_ASM_AVAILABLE) && defined(__GNUC__) && defined(CRYPTOPP_WORD128_AVAILABLE)
int Baseline_Add(size_t N, word *C, const word *A, const word *B)
{
	word result;
	__asm__ __volatile__
	(
	".intel_syntax;"
	AS1(	neg		%1)
	ASJ(	jz,		1, f)
	AS2(	mov		%0,[%3+8*%1])
	AS2(	add		%0,[%4+8*%1])
	AS2(	mov		[%2+8*%1],%0)
	ASL(0)
	AS2(	mov		%0,[%3+8*%1+8])
	AS2(	adc		%0,[%4+8*%1+8])
	AS2(	mov		[%2+8*%1+8],%0)
	AS2(	lea		%1,[%1+2])
	ASJ(	jrcxz,	1, f)
	AS2(	mov		%0,[%3+8*%1])
	AS2(	adc		%0,[%4+8*%1])
	AS2(	mov		[%2+8*%1],%0)
	ASJ(	jmp,	0, b)
	ASL(1)
	AS2(	mov		%0, 0)
	AS2(	adc		%0, %0)
	".att_syntax;"
	: "=&r" (result), "+c" (N)
	: "r" (C+N), "r" (A+N), "r" (B+N)
	: "memory", "cc"
	);
	return (int)result;
}

int Baseline_Sub(size_t N, word *C, const word *A, const word *B)
{
	word result;
	__asm__ __volatile__
	(
	".intel_syntax;"
	AS1(	neg		%1)
	ASJ(	jz,		1, f)
	AS2(	mov		%0,[%3+8*%1])
	AS2(	sub		%0,[%4+8*%1])
	AS2(	mov		[%2+8*%1],%0)
	ASL(0)
	AS2(	mov		%0,[%3+8*%1+8])
	AS2(	sbb		%0,[%4+8*%1+8])
	AS2(	mov		[%2+8*%1+8],%0)
	AS2(	lea		%1,[%1+2])
	ASJ(	jrcxz,	1, f)
	AS2(	mov		%0,[%3+8*%1])
	AS2(	sbb		%0,[%4+8*%1])
	AS2(	mov		[%2+8*%1],%0)
	ASJ(	jmp,	0, b)
	ASL(1)
	AS2(	mov		%0, 0)
	AS2(	adc		%0, %0)
	".att_syntax;"
	: "=&r" (result), "+c" (N)
	: "r" (C+N), "r" (A+N), "r" (B+N)
	: "memory", "cc"
	);
	return (int)result;
}
#elif defined(CRYPTOPP_X86_ASM_AVAILABLE) && CRYPTOPP_BOOL_X86
CRYPTOPP_NAKED int CRYPTOPP_FASTCALL Baseline_Add(size_t N, word *C, const word *A, const word *B)
{
	AddPrologue

	// now: eax = A, edi = B, edx = C, ecx = N
	AS2(	lea		eax, [eax+4*ecx])
	AS2(	lea		edi, [edi+4*ecx])
	AS2(	lea		edx, [edx+4*ecx])

	AS1(	neg		ecx)				// ecx is negative index
	AS2(	test	ecx, 2)				// this clears carry flag
	ASJ(	jz,		0, f)
	AS2(	sub		ecx, 2)
	ASJ(	jmp,	1, f)

	ASL(0)
	ASJ(	jecxz,	2, f)				// loop until ecx overflows and becomes zero
	AS2(	mov		esi,[eax+4*ecx])
	AS2(	adc		esi,[edi+4*ecx])
	AS2(	mov		[edx+4*ecx],esi)
	AS2(	mov		esi,[eax+4*ecx+4])
	AS2(	adc		esi,[edi+4*ecx+4])
	AS2(	mov		[edx+4*ecx+4],esi)
	ASL(1)
	AS2(	mov		esi,[eax+4*ecx+8])
	AS2(	adc		esi,[edi+4*ecx+8])
	AS2(	mov		[edx+4*ecx+8],esi)
	AS2(	mov		esi,[eax+4*ecx+12])
	AS2(	adc		esi,[edi+4*ecx+12])
	AS2(	mov		[edx+4*ecx+12],esi)

	AS2(	lea		ecx,[ecx+4])		// advance index, avoid inc which causes slowdown on Intel Core 2
	ASJ(	jmp,	0, b)

	ASL(2)
	AS2(	mov		eax, 0)
	AS1(	setc	al)					// store carry into eax (return result register)

	AddEpilogue
}

CRYPTOPP_NAKED int CRYPTOPP_FASTCALL Baseline_Sub(size_t N, word *C, const word *A, const word *B)
{
	AddPrologue

	// now: eax = A, edi = B, edx = C, ecx = N
	AS2(	lea		eax, [eax+4*ecx])
	AS2(	lea		edi, [edi+4*ecx])
	AS2(	lea		edx, [edx+4*ecx])

	AS1(	neg		ecx)				// ecx is negative index
	AS2(	test	ecx, 2)				// this clears carry flag
	ASJ(	jz,		0, f)
	AS2(	sub		ecx, 2)
	ASJ(	jmp,	1, f)

	ASL(0)
	ASJ(	jecxz,	2, f)				// loop until ecx overflows and becomes zero
	AS2(	mov		esi,[eax+4*ecx])
	AS2(	sbb		esi,[edi+4*ecx])
	AS2(	mov		[edx+4*ecx],esi)
	AS2(	mov		esi,[eax+4*ecx+4])
	AS2(	sbb		esi,[edi+4*ecx+4])
	AS2(	mov		[edx+4*ecx+4],esi)
	ASL(1)
	AS2(	mov		esi,[eax+4*ecx+8])
	AS2(	sbb		esi,[edi+4*ecx+8])
	AS2(	mov		[edx+4*ecx+8],esi)
	AS2(	mov		esi,[eax+4*ecx+12])
	AS2(	sbb		esi,[edi+4*ecx+12])
	AS2(	mov		[edx+4*ecx+12],esi)

	AS2(	lea		ecx,[ecx+4])		// advance index, avoid inc which causes slowdown on Intel Core 2
	ASJ(	jmp,	0, b)

	ASL(2)
	AS2(	mov		eax, 0)
	AS1(	setc	al)					// store carry into eax (return result register)

	AddEpilogue
}

#if CRYPTOPP_INTEGER_SSE2
CRYPTOPP_NAKED int CRYPTOPP_FASTCALL SSE2_Add(size_t N, word *C, const word *A, const word *B)
{
	AddPrologue

	// now: eax = A, edi = B, edx = C, ecx = N
	AS2(	lea		eax, [eax+4*ecx])
	AS2(	lea		edi, [edi+4*ecx])
	AS2(	lea		edx, [edx+4*ecx])

	AS1(	neg		ecx)				// ecx is negative index
	AS2(	pxor    mm2, mm2)
	ASJ(	jz,		2, f)
	AS2(	test	ecx, 2)				// this clears carry flag
	ASJ(	jz,		0, f)
	AS2(	sub		ecx, 2)
	ASJ(	jmp,	1, f)

	ASL(0)
	AS2(	movd     mm0, DWORD PTR [eax+4*ecx])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx])
	AS2(	paddq    mm0, mm1)
	AS2(	paddq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx], mm2)
	AS2(	psrlq    mm2, 32)

	AS2(	movd     mm0, DWORD PTR [eax+4*ecx+4])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+4])
	AS2(	paddq    mm0, mm1)
	AS2(	paddq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx+4], mm2)
	AS2(	psrlq    mm2, 32)

	ASL(1)
	AS2(	movd     mm0, DWORD PTR [eax+4*ecx+8])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+8])
	AS2(	paddq    mm0, mm1)
	AS2(	paddq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx+8], mm2)
	AS2(	psrlq    mm2, 32)

	AS2(	movd     mm0, DWORD PTR [eax+4*ecx+12])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+12])
	AS2(	paddq    mm0, mm1)
	AS2(	paddq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx+12], mm2)
	AS2(	psrlq    mm2, 32)

	AS2(	add		ecx, 4)
	ASJ(	jnz,	0, b)

	ASL(2)
	AS2(	movd	eax, mm2)
	AS1(	emms)

	AddEpilogue
}
CRYPTOPP_NAKED int CRYPTOPP_FASTCALL SSE2_Sub(size_t N, word *C, const word *A, const word *B)
{
	AddPrologue

	// now: eax = A, edi = B, edx = C, ecx = N
	AS2(	lea		eax, [eax+4*ecx])
	AS2(	lea		edi, [edi+4*ecx])
	AS2(	lea		edx, [edx+4*ecx])

	AS1(	neg		ecx)				// ecx is negative index
	AS2(	pxor    mm2, mm2)
	ASJ(	jz,		2, f)
	AS2(	test	ecx, 2)				// this clears carry flag
	ASJ(	jz,		0, f)
	AS2(	sub		ecx, 2)
	ASJ(	jmp,	1, f)

	ASL(0)
	AS2(	movd     mm0, DWORD PTR [eax+4*ecx])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx])
	AS2(	psubq    mm0, mm1)
	AS2(	psubq	 mm0, mm2)
	AS2(	movd	 DWORD PTR [edx+4*ecx], mm0)
	AS2(	psrlq    mm0, 63)

	AS2(	movd     mm2, DWORD PTR [eax+4*ecx+4])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+4])
	AS2(	psubq    mm2, mm1)
	AS2(	psubq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx+4], mm2)
	AS2(	psrlq    mm2, 63)

	ASL(1)
	AS2(	movd     mm0, DWORD PTR [eax+4*ecx+8])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+8])
	AS2(	psubq    mm0, mm1)
	AS2(	psubq	 mm0, mm2)
	AS2(	movd	 DWORD PTR [edx+4*ecx+8], mm0)
	AS2(	psrlq    mm0, 63)

	AS2(	movd     mm2, DWORD PTR [eax+4*ecx+12])
	AS2(	movd     mm1, DWORD PTR [edi+4*ecx+12])
	AS2(	psubq    mm2, mm1)
	AS2(	psubq	 mm2, mm0)
	AS2(	movd	 DWORD PTR [edx+4*ecx+12], mm2)
	AS2(	psrlq    mm2, 63)

	AS2(	add		ecx, 4)
	ASJ(	jnz,	0, b)

	ASL(2)
	AS2(	movd	eax, mm2)
	AS1(	emms)

	AddEpilogue
}
#endif	// #if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE
#else
int CRYPTOPP_FASTCALL Baseline_Add(size_t N, word *C, const word *A, const word *B)
{
	assert (N%2 == 0);

	Declare2Words(u);
	AssignWord(u, 0);
	for (size_t i=0; i<N; i+=2)
	{
		AddWithCarry(u, A[i], B[i]);
		C[i] = LowWord(u);
		AddWithCarry(u, A[i+1], B[i+1]);
		C[i+1] = LowWord(u);
	}
	return int(GetCarry(u));
}

int CRYPTOPP_FASTCALL Baseline_Sub(size_t N, word *C, const word *A, const word *B)
{
	assert (N%2 == 0);

	Declare2Words(u);
	AssignWord(u, 0);
	for (size_t i=0; i<N; i+=2)
	{
		SubtractWithBorrow(u, A[i], B[i]);
		C[i] = LowWord(u);
		SubtractWithBorrow(u, A[i+1], B[i+1]);
		C[i+1] = LowWord(u);
	}
	return int(GetBorrow(u));
}
#endif

static word LinearMultiply(word *C, const word *A, word B, size_t N)
{
	word carry=0;
	for(unsigned i=0; i<N; i++)
	{
		Declare2Words(p);
		MultiplyWords(p, A[i], B);
		Acc2WordsBy1(p, carry);
		C[i] = LowWord(p);
		carry = HighWord(p);
	}
	return carry;
}

#ifndef CRYPTOPP_DOXYGEN_PROCESSING

#define Mul_2 \
	Mul_Begin(2) \
	Mul_SaveAcc(0, 0, 1) Mul_Acc(1, 0) \
	Mul_End(1, 1)

#define Mul_4 \
	Mul_Begin(4) \
	Mul_SaveAcc(0, 0, 1) Mul_Acc(1, 0) \
	Mul_SaveAcc(1, 0, 2) Mul_Acc(1, 1) Mul_Acc(2, 0)  \
	Mul_SaveAcc(2, 0, 3) Mul_Acc(1, 2) Mul_Acc(2, 1) Mul_Acc(3, 0)  \
	Mul_SaveAcc(3, 1, 3) Mul_Acc(2, 2) Mul_Acc(3, 1)  \
	Mul_SaveAcc(4, 2, 3) Mul_Acc(3, 2) \
	Mul_End(5, 3)

#define Mul_8 \
	Mul_Begin(8) \
	Mul_SaveAcc(0, 0, 1) Mul_Acc(1, 0) \
	Mul_SaveAcc(1, 0, 2) Mul_Acc(1, 1) Mul_Acc(2, 0)  \
	Mul_SaveAcc(2, 0, 3) Mul_Acc(1, 2) Mul_Acc(2, 1) Mul_Acc(3, 0)  \
	Mul_SaveAcc(3, 0, 4) Mul_Acc(1, 3) Mul_Acc(2, 2) Mul_Acc(3, 1) Mul_Acc(4, 0) \
	Mul_SaveAcc(4, 0, 5) Mul_Acc(1, 4) Mul_Acc(2, 3) Mul_Acc(3, 2) Mul_Acc(4, 1) Mul_Acc(5, 0) \
	Mul_SaveAcc(5, 0, 6) Mul_Acc(1, 5) Mul_Acc(2, 4) Mul_Acc(3, 3) Mul_Acc(4, 2) Mul_Acc(5, 1) Mul_Acc(6, 0) \
	Mul_SaveAcc(6, 0, 7) Mul_Acc(1, 6) Mul_Acc(2, 5) Mul_Acc(3, 4) Mul_Acc(4, 3) Mul_Acc(5, 2) Mul_Acc(6, 1) Mul_Acc(7, 0) \
	Mul_SaveAcc(7, 1, 7) Mul_Acc(2, 6) Mul_Acc(3, 5) Mul_Acc(4, 4) Mul_Acc(5, 3) Mul_Acc(6, 2) Mul_Acc(7, 1) \
	Mul_SaveAcc(8, 2, 7) Mul_Acc(3, 6) Mul_Acc(4, 5) Mul_Acc(5, 4) Mul_Acc(6, 3) Mul_Acc(7, 2) \
	Mul_SaveAcc(9, 3, 7) Mul_Acc(4, 6) Mul_Acc(5, 5) Mul_Acc(6, 4) Mul_Acc(7, 3) \
	Mul_SaveAcc(10, 4, 7) Mul_Acc(5, 6) Mul_Acc(6, 5) Mul_Acc(7, 4) \
	Mul_SaveAcc(11, 5, 7) Mul_Acc(6, 6) Mul_Acc(7, 5) \
	Mul_SaveAcc(12, 6, 7) Mul_Acc(7, 6) \
	Mul_End(13, 7)

#define Mul_16 \
	Mul_Begin(16) \
	Mul_SaveAcc(0, 0, 1) Mul_Acc(1, 0) \
	Mul_SaveAcc(1, 0, 2) Mul_Acc(1, 1) Mul_Acc(2, 0) \
	Mul_SaveAcc(2, 0, 3) Mul_Acc(1, 2) Mul_Acc(2, 1) Mul_Acc(3, 0) \
	Mul_SaveAcc(3, 0, 4) Mul_Acc(1, 3) Mul_Acc(2, 2) Mul_Acc(3, 1) Mul_Acc(4, 0) \
	Mul_SaveAcc(4, 0, 5) Mul_Acc(1, 4) Mul_Acc(2, 3) Mul_Acc(3, 2) Mul_Acc(4, 1) Mul_Acc(5, 0) \
	Mul_SaveAcc(5, 0, 6) Mul_Acc(1, 5) Mul_Acc(2, 4) Mul_Acc(3, 3) Mul_Acc(4, 2) Mul_Acc(5, 1) Mul_Acc(6, 0) \
	Mul_SaveAcc(6, 0, 7) Mul_Acc(1, 6) Mul_Acc(2, 5) Mul_Acc(3, 4) Mul_Acc(4, 3) Mul_Acc(5, 2) Mul_Acc(6, 1) Mul_Acc(7, 0) \
	Mul_SaveAcc(7, 0, 8) Mul_Acc(1, 7) Mul_Acc(2, 6) Mul_Acc(3, 5) Mul_Acc(4, 4) Mul_Acc(5, 3) Mul_Acc(6, 2) Mul_Acc(7, 1) Mul_Acc(8, 0) \
	Mul_SaveAcc(8, 0, 9) Mul_Acc(1, 8) Mul_Acc(2, 7) Mul_Acc(3, 6) Mul_Acc(4, 5) Mul_Acc(5, 4) Mul_Acc(6, 3) Mul_Acc(7, 2) Mul_Acc(8, 1) Mul_Acc(9, 0) \
	Mul_SaveAcc(9, 0, 10) Mul_Acc(1, 9) Mul_Acc(2, 8) Mul_Acc(3, 7) Mul_Acc(4, 6) Mul_Acc(5, 5) Mul_Acc(6, 4) Mul_Acc(7, 3) Mul_Acc(8, 2) Mul_Acc(9, 1) Mul_Acc(10, 0) \
	Mul_SaveAcc(10, 0, 11) Mul_Acc(1, 10) Mul_Acc(2, 9) Mul_Acc(3, 8) Mul_Acc(4, 7) Mul_Acc(5, 6) Mul_Acc(6, 5) Mul_Acc(7, 4) Mul_Acc(8, 3) Mul_Acc(9, 2) Mul_Acc(10, 1) Mul_Acc(11, 0) \
	Mul_SaveAcc(11, 0, 12) Mul_Acc(1, 11) Mul_Acc(2, 10) Mul_Acc(3, 9) Mul_Acc(4, 8) Mul_Acc(5, 7) Mul_Acc(6, 6) Mul_Acc(7, 5) Mul_Acc(8, 4) Mul_Acc(9, 3) Mul_Acc(10, 2) Mul_Acc(11, 1) Mul_Acc(12, 0) \
	Mul_SaveAcc(12, 0, 13) Mul_Acc(1, 12) Mul_Acc(2, 11) Mul_Acc(3, 10) Mul_Acc(4, 9) Mul_Acc(5, 8) Mul_Acc(6, 7) Mul_Acc(7, 6) Mul_Acc(8, 5) Mul_Acc(9, 4) Mul_Acc(10, 3) Mul_Acc(11, 2) Mul_Acc(12, 1) Mul_Acc(13, 0) \
	Mul_SaveAcc(13, 0, 14) Mul_Acc(1, 13) Mul_Acc(2, 12) Mul_Acc(3, 11) Mul_Acc(4, 10) Mul_Acc(5, 9) Mul_Acc(6, 8) Mul_Acc(7, 7) Mul_Acc(8, 6) Mul_Acc(9, 5) Mul_Acc(10, 4) Mul_Acc(11, 3) Mul_Acc(12, 2) Mul_Acc(13, 1) Mul_Acc(14, 0) \
	Mul_SaveAcc(14, 0, 15) Mul_Acc(1, 14) Mul_Acc(2, 13) Mul_Acc(3, 12) Mul_Acc(4, 11) Mul_Acc(5, 10) Mul_Acc(6, 9) Mul_Acc(7, 8) Mul_Acc(8, 7) Mul_Acc(9, 6) Mul_Acc(10, 5) Mul_Acc(11, 4) Mul_Acc(12, 3) Mul_Acc(13, 2) Mul_Acc(14, 1) Mul_Acc(15, 0) \
	Mul_SaveAcc(15, 1, 15) Mul_Acc(2, 14) Mul_Acc(3, 13) Mul_Acc(4, 12) Mul_Acc(5, 11) Mul_Acc(6, 10) Mul_Acc(7, 9) Mul_Acc(8, 8) Mul_Acc(9, 7) Mul_Acc(10, 6) Mul_Acc(11, 5) Mul_Acc(12, 4) Mul_Acc(13, 3) Mul_Acc(14, 2) Mul_Acc(15, 1) \
	Mul_SaveAcc(16, 2, 15) Mul_Acc(3, 14) Mul_Acc(4, 13) Mul_Acc(5, 12) Mul_Acc(6, 11) Mul_Acc(7, 10) Mul_Acc(8, 9) Mul_Acc(9, 8) Mul_Acc(10, 7) Mul_Acc(11, 6) Mul_Acc(12, 5) Mul_Acc(13, 4) Mul_Acc(14, 3) Mul_Acc(15, 2) \
	Mul_SaveAcc(17, 3, 15) Mul_Acc(4, 14) Mul_Acc(5, 13) Mul_Acc(6, 12) Mul_Acc(7, 11) Mul_Acc(8, 10) Mul_Acc(9, 9) Mul_Acc(10, 8) Mul_Acc(11, 7) Mul_Acc(12, 6) Mul_Acc(13, 5) Mul_Acc(14, 4) Mul_Acc(15, 3) \
	Mul_SaveAcc(18, 4, 15) Mul_Acc(5, 14) Mul_Acc(6, 13) Mul_Acc(7, 12) Mul_Acc(8, 11) Mul_Acc(9, 10) Mul_Acc(10, 9) Mul_Acc(11, 8) Mul_Acc(12, 7) Mul_Acc(13, 6) Mul_Acc(14, 5) Mul_Acc(15, 4) \
	Mul_SaveAcc(19, 5, 15) Mul_Acc(6, 14) Mul_Acc(7, 13) Mul_Acc(8, 12) Mul_Acc(9, 11) Mul_Acc(10, 10) Mul_Acc(11, 9) Mul_Acc(12, 8) Mul_Acc(13, 7) Mul_Acc(14, 6) Mul_Acc(15, 5) \
	Mul_SaveAcc(20, 6, 15) Mul_Acc(7, 14) Mul_Acc(8, 13) Mul_Acc(9, 12) Mul_Acc(10, 11) Mul_Acc(11, 10) Mul_Acc(12, 9) Mul_Acc(13, 8) Mul_Acc(14, 7) Mul_Acc(15, 6) \
	Mul_SaveAcc(21, 7, 15) Mul_Acc(8, 14) Mul_Acc(9, 13) Mul_Acc(10, 12) Mul_Acc(11, 11) Mul_Acc(12, 10) Mul_Acc(13, 9) Mul_Acc(14, 8) Mul_Acc(15, 7) \
	Mul_SaveAcc(22, 8, 15) Mul_Acc(9, 14) Mul_Acc(10, 13) Mul_Acc(11, 12) Mul_Acc(12, 11) Mul_Acc(13, 10) Mul_Acc(14, 9) Mul_Acc(15, 8) \
	Mul_SaveAcc(23, 9, 15) Mul_Acc(10, 14) Mul_Acc(11, 13) Mul_Acc(12, 12) Mul_Acc(13, 11) Mul_Acc(14, 10) Mul_Acc(15, 9) \
	Mul_SaveAcc(24, 10, 15) Mul_Acc(11, 14) Mul_Acc(12, 13) Mul_Acc(13, 12) Mul_Acc(14, 11) Mul_Acc(15, 10) \
	Mul_SaveAcc(25, 11, 15) Mul_Acc(12, 14) Mul_Acc(13, 13) Mul_Acc(14, 12) Mul_Acc(15, 11) \
	Mul_SaveAcc(26, 12, 15) Mul_Acc(13, 14) Mul_Acc(14, 13) Mul_Acc(15, 12) \
	Mul_SaveAcc(27, 13, 15) Mul_Acc(14, 14) Mul_Acc(15, 13) \
	Mul_SaveAcc(28, 14, 15) Mul_Acc(15, 14) \
	Mul_End(29, 15)

#define Squ_2 \
	Squ_Begin(2) \
	Squ_End(2)

#define Squ_4 \
	Squ_Begin(4) \
	Squ_SaveAcc(1, 0, 2) Squ_Diag(1) \
	Squ_SaveAcc(2, 0, 3) Squ_Acc(1, 2) Squ_NonDiag \
	Squ_SaveAcc(3, 1, 3) Squ_Diag(2) \
	Squ_SaveAcc(4, 2, 3) Squ_NonDiag \
	Squ_End(4)

#define Squ_8 \
	Squ_Begin(8) \
	Squ_SaveAcc(1, 0, 2) Squ_Diag(1) \
	Squ_SaveAcc(2, 0, 3) Squ_Acc(1, 2) Squ_NonDiag \
	Squ_SaveAcc(3, 0, 4) Squ_Acc(1, 3) Squ_Diag(2) \
	Squ_SaveAcc(4, 0, 5) Squ_Acc(1, 4) Squ_Acc(2, 3) Squ_NonDiag \
	Squ_SaveAcc(5, 0, 6) Squ_Acc(1, 5) Squ_Acc(2, 4) Squ_Diag(3) \
	Squ_SaveAcc(6, 0, 7) Squ_Acc(1, 6) Squ_Acc(2, 5) Squ_Acc(3, 4) Squ_NonDiag \
	Squ_SaveAcc(7, 1, 7) Squ_Acc(2, 6) Squ_Acc(3, 5) Squ_Diag(4) \
	Squ_SaveAcc(8, 2, 7) Squ_Acc(3, 6) Squ_Acc(4, 5)  Squ_NonDiag \
	Squ_SaveAcc(9, 3, 7) Squ_Acc(4, 6) Squ_Diag(5) \
	Squ_SaveAcc(10, 4, 7) Squ_Acc(5, 6) Squ_NonDiag \
	Squ_SaveAcc(11, 5, 7) Squ_Diag(6) \
	Squ_SaveAcc(12, 6, 7) Squ_NonDiag \
	Squ_End(8)

#define Squ_16 \
	Squ_Begin(16) \
	Squ_SaveAcc(1, 0, 2) Squ_Diag(1) \
	Squ_SaveAcc(2, 0, 3) Squ_Acc(1, 2) Squ_NonDiag \
	Squ_SaveAcc(3, 0, 4) Squ_Acc(1, 3) Squ_Diag(2) \
	Squ_SaveAcc(4, 0, 5) Squ_Acc(1, 4) Squ_Acc(2, 3) Squ_NonDiag \
	Squ_SaveAcc(5, 0, 6) Squ_Acc(1, 5) Squ_Acc(2, 4) Squ_Diag(3) \
	Squ_SaveAcc(6, 0, 7) Squ_Acc(1, 6) Squ_Acc(2, 5) Squ_Acc(3, 4) Squ_NonDiag \
	Squ_SaveAcc(7, 0, 8) Squ_Acc(1, 7) Squ_Acc(2, 6) Squ_Acc(3, 5) Squ_Diag(4) \
	Squ_SaveAcc(8, 0, 9) Squ_Acc(1, 8) Squ_Acc(2, 7) Squ_Acc(3, 6) Squ_Acc(4, 5) Squ_NonDiag \
	Squ_SaveAcc(9, 0, 10) Squ_Acc(1, 9) Squ_Acc(2, 8) Squ_Acc(3, 7) Squ_Acc(4, 6) Squ_Diag(5) \
	Squ_SaveAcc(10, 0, 11) Squ_Acc(1, 10) Squ_Acc(2, 9) Squ_Acc(3, 8) Squ_Acc(4, 7) Squ_Acc(5, 6) Squ_NonDiag \
	Squ_SaveAcc(11, 0, 12) Squ_Acc(1, 11) Squ_Acc(2, 10) Squ_Acc(3, 9) Squ_Acc(4, 8) Squ_Acc(5, 7) Squ_Diag(6) \
	Squ_SaveAcc(12, 0, 13) Squ_Acc(1, 12) Squ_Acc(2, 11) Squ_Acc(3, 10) Squ_Acc(4, 9) Squ_Acc(5, 8) Squ_Acc(6, 7) Squ_NonDiag \