add addNC/subNC

include/mcl/fp.hpp

@@ -278,6 +278,8 @@ public:
 	}
 	static inline void add(FpT& z, const FpT& x, const FpT& y) { op_.fp_add(z.v_, x.v_, y.v_); }
 	static inline void sub(FpT& z, const FpT& x, const FpT& y) { op_.fp_sub(z.v_, x.v_, y.v_); }
+	static inline void addNC(FpT& z, const FpT& x, const FpT& y) { op_.fp_addNC(z.v_, x.v_, y.v_); }
+	static inline void subNC(FpT& z, const FpT& x, const FpT& y) { op_.fp_subNC(z.v_, x.v_, y.v_); }
 	static inline void mul(FpT& z, const FpT& x, const FpT& y) { op_.fp_mul(z.v_, x.v_, y.v_); }
 	static inline void inv(FpT& y, const FpT& x) { op_.fp_invOp(y.v_, x.v_, op_); }
 	static inline void neg(FpT& y, const FpT& x) { op_.fp_neg(y.v_, x.v_); }

include/mcl/fp_tower.hpp

@@ -126,6 +126,9 @@ public:
 	}
 	static inline void init(int xi_c)
 	{
+		if (Fp::maxSize > 256) {
+			throw cybozu::Exception("Fp2T:init:not support size") << Fp::maxSize;
+		}
 		xi_c_ = xi_c;
 		Fp::op_.fp2_add = fp2_addW;
 		Fp::op_.fp2_sub = fp2_subW;
@@ -259,5 +262,17 @@ private:
 
 template<class Fp> Fp Fp2T<Fp>::xi_c_;
 
+/*
+	Fp6T = Fp2[v] / (v^3 - xi)
+	xi = -u - 1
+	x = a + b v + c v^2
+*/
+template<class Fp>
+class Fp6T {
+	typedef Fp2T<Fp> Fp2;
+public:
+};
+
+
 } // mcl
 

include/mcl/op.hpp

@@ -79,6 +79,13 @@ struct Op {
 	void3u fp_add;
 	void3u fp_sub;
 	void3u fp_mul;
+
+	bool fullBit; // true if bitSize % uniSize == 0
+	/*
+		same fp_add, fp_sub if fullBit
+	*/
+	void3u fp_addNC; // assume no carry if !fullBit
+	void3u fp_subNC; // assume x > y
 	// for Montgomery
 	bool useMont;
 	int2u fp_preInv;
@@ -112,6 +119,7 @@ struct Op {
 		: N(0), bitSize(0)
 		, fp_isZero(0), fp_clear(0), fp_copy(0)
 		, fp_neg(0), fp_sqr(0), fp_add(0), fp_sub(0), fp_mul(0)
+		, fullBit(true), fp_addNC(0), fp_subNC(0)
 		, useMont(false), fp_preInv(0)
 		, rp(0), mont(0)
 		, fp_negP(0), fp_sqrPreP(0), fp_invOp(0), fp_addP(0), fp_subP(0), fp_mulPreP(0), fp_modP(0)

src/fp.cpp

@@ -122,6 +122,27 @@ struct OpeFunc {
 		}
 		Gmp::getArray(z, N, mz);
 	}
+	// z = x + y
+	static inline void fp_addNCC(Unit *z, const Unit *x, const Unit *y)
+	{
+		Unit ret[N + 2]; // not N + 1
+		mpz_t mz, mx, my;
+		set_zero(mz, ret, N + 2);
+		set_mpz_t(mx, x);
+		set_mpz_t(my, y);
+		mpz_add(mz, mx, my);
+		Gmp::getArray(z, N, mz);
+	}
+	static inline void fp_subNCC(Unit *z, const Unit *x, const Unit *y)
+	{
+		Unit ret[N + 1];
+		mpz_t mz, mx, my;
+		set_zero(mz, ret, N + 1);
+		set_mpz_t(mx, x);
+		set_mpz_t(my, y);
+		mpz_sub(mz, mx, my);
+		Gmp::getArray(z, N, mz);
+	}
 	// z[N * 2] <- x[N] * y[N]
 	static inline void fp_mulPreC(Unit *z, const Unit *x, const Unit *y)
 	{
@@ -210,6 +231,13 @@ struct OpeFunc {
 		} \
 		fp_addP = OpeFunc<n>::fp_addC; \
 		fp_subP = OpeFunc<n>::fp_subC; \
+		if (fullBit) { \
+			fp_addNC = fp_add; \
+			fp_subNC = fp_sub; \
+		} else { \
+			fp_addNC = OpeFunc<n>::fp_addNCC; \
+			fp_subNC = OpeFunc<n>::fp_subNCC; \
+		} \
 		fp_mulPreP = OpeFunc<n>::fp_mulPreC; \
 		fp_sqrPreP = OpeFunc<n>::fp_sqrPreC; \
 		fp_modP = OpeFunc<n>::fp_modC; \
@@ -284,6 +312,7 @@ void Op::init(const std::string& mstr, int base, size_t maxBitSize, Mode mode)
 	bool isMinus = fp::strToMpzArray(&bitSize, p, maxBitSize, mp, mstr, base);
 	if (isMinus) throw cybozu::Exception("Op:init:mstr is minus") << mstr;
 	if (mp == 0) throw cybozu::Exception("Op:init:mstr is zero") << mstr;
+	fullBit = (bitSize % UnitBitSize) == 0;
 
 	const size_t roundBit = (bitSize + UnitBitSize - 1) & ~(UnitBitSize - 1);
 	switch (roundBit) {

test/fp_test.cpp

@@ -239,6 +239,11 @@ CYBOZU_TEST_AUTO(ope)
 		z *= y;
 		CYBOZU_TEST_EQUAL(z, tbl[i].x);
 	}
+	Fp x(5), y(3), z;
+	Fp::addNC(z, x, y);
+	CYBOZU_TEST_EQUAL(z, Fp(8));
+	Fp::subNC(z, x, y);
+	CYBOZU_TEST_EQUAL(z, Fp(2));
 }
 
 struct tag2;

test/mont_fp_test.cpp

@@ -456,6 +456,13 @@ struct Test {
 			z *= y;
 			CYBOZU_TEST_EQUAL(z, castTo<Fp>(tbl[i].x));
 		}
+		{
+			Fp x(5), y(3), z;
+			Fp::addNC(z, x, y);
+			CYBOZU_TEST_EQUAL(z, Fp(8));
+//			Fp::subNC(z, x, y);
+//			CYBOZU_TEST_EQUAL(z, Fp(2));
+		}
 	}
 	void power()
 	{