From 32f0347498c119e243cfda6048aa1c1181b69d3e Mon Sep 17 00:00:00 2001
From: MITSUNARI Shigeo <herumi@nifty.com>
Date: Fri, 7 Oct 2016 16:31:08 +0900
Subject: [PATCH] add generic SqrMont, Mul, Sqr

---
 Makefile         |   4 --
 src/fp.cpp       |  86 +++----------------------------------
 src/fp_proto.hpp | 109 ++++++++++++++++++++++++++++++++++++++---------
 3 files changed, 97 insertions(+), 102 deletions(-)

diff --git a/Makefile b/Makefile
index 4e9cd7e..55af2cd 100644
--- a/Makefile
+++ b/Makefile
@@ -50,11 +50,9 @@ ifeq ($(USE_LOW_ASM),1)
 endif
 
 $(MCL_LIB): $(LIB_OBJ)
-	-$(MKDIR) $(@D)
 	$(AR) $@ $(LIB_OBJ)
 
 $(ASM_OBJ): $(ASM_SRC)
-	-$(MKDIR) $(@D)
 	$(PRE)$(CXX) -c $< -o $@ $(CFLAGS)
 
 $(ASM_SRC): $(LLVM_SRC)
@@ -87,11 +85,9 @@ VPATH=test sample src
 .SUFFIXES: .cpp .d .exe
 
 $(OBJ_DIR)/%.o: %.cpp
-	-$(MKDIR) $(@D)
 	$(PRE)$(CXX) $(CFLAGS) -c $< -o $@ -MMD -MP -MF $(@:.o=.d)
 
 $(EXE_DIR)/%.exe: $(OBJ_DIR)/%.o $(MCL_LIB)
-	-$(MKDIR) $(@D)
 	$(PRE)$(CXX) $< -o $@ $(MCL_LIB) $(LDFLAGS)
 
 SAMPLE_EXE=$(addprefix $(EXE_DIR)/,$(SAMPLE_SRC:.cpp=.exe))
diff --git a/src/fp.cpp b/src/fp.cpp
index 6f09bd0..7ff29fc 100644
--- a/src/fp.cpp
+++ b/src/fp.cpp
@@ -122,33 +122,14 @@ template<>const void3u MontRed<n, Ltag>::f = &mcl_fp_montRed ## n ## L; \
 template<>const void4u DblAdd<n, Ltag>::f = &mcl_fpDbl_add ## n ## L; \
 template<>const void4u DblSub<n, Ltag>::f = &mcl_fpDbl_sub ## n ## L; \
 
+// use Dbl_Mod of gmp
 template<size_t N>
-struct Mul<N, Ltag> {
-	static inline void func(Unit *z, const Unit *x, const Unit *y, const Unit *p)
-	{
-		Unit xy[N * 2];
-		MulPre<N, Ltag>::f(xy, x, y);
-		Dbl_Mod<N, Gtag>::f(z, xy, p);
-	}
-	static const void4u f;
-};
-
-template<size_t N>
-const void4u Mul<N, Ltag>::f = Mul<N, Ltag>::func;
-
-template<size_t N>
-struct Sqr<N, Ltag> {
-	static inline void func(Unit *y, const Unit *x, const Unit *p)
-	{
-		Unit xx[N * 2];
-		SqrPre<N, Ltag>::f(xx, x);
-		Dbl_Mod<N, Gtag>::f(y, xx, p);
-	}
+struct Dbl_Mod<N, Ltag> {
 	static const void3u f;
 };
 
 template<size_t N>
-const void3u Sqr<N, Ltag>::f = Sqr<N, Ltag>::func;
+const void3u Dbl_Mod<N, Ltag>::f = Dbl_Mod<N, Gtag>::f;
 
 MCL_DEF_LLVM_FUNC(1)
 MCL_DEF_LLVM_FUNC(2)
@@ -201,41 +182,6 @@ struct OpeFunc {
 	{
 		copyArray(y, x, N);
 	}
-	// z[N] <- mont(x[N], y[N])
-	static inline void fp_mulMontC(Unit *z, const Unit *x, const Unit *y, const Unit *p)
-	{
-#if 0
-		Unit xy[N * 2];
-		MulPre<N, Gtag>::f(xy, x, y);
-		fpDbl_modMontC(z, xy, p);
-#else
-		const Unit rp = p[-1];
-		Unit buf[N * 2 + 2];
-		Unit *c = buf;
-		Mul_UnitPre<N, Gtag>::f(c, x, y[0]); // x * y[0]
-		Unit q = c[0] * rp;
-		Unit t[N + 2];
-		Mul_UnitPre<N, Gtag>::f(t, p, q); // p * q
-		t[N + 1] = 0; // always zero
-		c[N + 1] = AddNC<N + 1, Gtag>::f(c, c, t);
-		c++;
-		for (size_t i = 1; i < N; i++) {
-			Mul_UnitPre<N, Gtag>::f(t, x, y[i]);
-			c[N + 1] = AddNC<N + 1, Gtag>::f(c, c, t);
-			q = c[0] * rp;
-			Mul_UnitPre<N, Gtag>::f(t, p, q);
-			AddNC<N + 2, Gtag>::f(c, c, t);
-			c++;
-		}
-		if (c[N]) {
-			SubNC<N, Gtag>::f(z, c, p);
-		} else {
-			if (SubNC<N, Gtag>::f(z, c, p)) {
-				memcpy(z, c, N * sizeof(Unit));
-			}
-		}
-#endif
-	}
 	/*
 		z[N] <- montRed(xy[N * 2])
 		REMARK : assume p[-1] = rp
@@ -272,24 +218,6 @@ struct OpeFunc {
 		Mul_UnitPre<N, Gtag>::f(xy, x, y);
 		N1_Mod<N, Gtag>::f(z, xy, p);
 	}
-	static inline void fp_mulC(Unit *z, const Unit *x, const Unit *y, const Unit *p)
-	{
-		Unit xy[N * 2];
-		MulPre<N, Gtag>::f(xy, x, y);
-		Dbl_Mod<N, Gtag>::f(z, xy, p);
-	}
-	static inline void fp_sqrC(Unit *y, const Unit *x, const Unit *p)
-	{
-		Unit xx[N * 2];
-		SqrPre<N, Gtag>::f(xx, x);
-		Dbl_Mod<N, Gtag>::f(y, xx, p);
-	}
-	static inline void fp_sqrMontC(Unit *y, const Unit *x, const Unit *p)
-	{
-		Unit xx[N * 2];
-		SqrPre<N, Gtag>::f(xx, x);
-		fpDbl_modMontC(y, xx, p);
-	}
 	static inline void fp_invOpC(Unit *y, const Unit *x, const Op& op)
 	{
 		mpz_class my;
@@ -373,13 +301,13 @@ void setOp(Op& op, Mode mode)
 	op.fp_add = Add<N, Gtag>::f;
 	op.fp_sub = Sub<N, Gtag>::f;
 	if (op.isMont) {
-		op.fp_mul = OpeFunc<N>::fp_mulMontC;
-		op.fp_sqr = OpeFunc<N>::fp_sqrMontC;
+		op.fp_mul = Mont<N, Gtag>::f;
+		op.fp_sqr = SqrMont<N, Gtag>::f;
 		op.fp_invOp = OpeFunc<N>::fp_invMontOpC;
 		op.fpDbl_mod = OpeFunc<N>::fpDbl_modMontC;
 	} else {
-		op.fp_mul = OpeFunc<N>::fp_mulC;
-		op.fp_sqr = OpeFunc<N>::fp_sqrC;
+		op.fp_mul = Mul<N, Gtag>::f;
+		op.fp_sqr = Sqr<N, Gtag>::f;
 		op.fp_invOp = OpeFunc<N>::fp_invOpC;
 		op.fpDbl_mod = Dbl_Mod<N, Gtag>::f;
 	}
diff --git a/src/fp_proto.hpp b/src/fp_proto.hpp
index 24c5330..68cca54 100644
--- a/src/fp_proto.hpp
+++ b/src/fp_proto.hpp
@@ -27,28 +27,9 @@ template<size_t N, class Tag>struct Mul_UnitPre { static const void2uI f; };
 template<size_t N, class Tag>struct N1_Mod { static const void3u f; };
 // z[N] <- x[N * 2] % p[N]
 template<size_t N, class Tag>struct Dbl_Mod { static const void3u f; };
-// z[N] <- Montgomery(x[N], y[N], p[N])
-template<size_t N, class Tag>struct Mont { static const void4u f; };
 // z[N] <- MontRed(xy[N], p[N])
 template<size_t N, class Tag>struct MontRed { static const void3u f; };
 
-// z[N] <- (x[N] * y[N]) % p[N]
-template<size_t N, class Tag>struct Mul { static const void4u f; };
-// z[N] <- (x[N] ^ 2) % p[N]
-template<size_t N, class Tag>struct Sqr { static const void3u f; };
-
-// z[N] <- Montgomery(x[N], x[N], p[N])
-template<size_t N, class Tag>
-struct SqrMont {
-	static inline void func(Unit *y, const Unit *x, const Unit *p)
-	{
-		Mont<N, Tag>::f(y, x, x, p);
-	}
-	static const void3u f;
-};
-template<size_t N, class Tag>
-const void3u SqrMont<N, Tag>::f = SqrMont<N, Tag>::func;
-
 // z[N] <- (x[N] + y[N]) % p[N]
 template<size_t N, class Tag>
 struct Add {
@@ -119,6 +100,96 @@ struct DblSub {
 template<size_t N, class Tag>
 const void4u DblSub<N, Tag>::f = DblSub<N, Tag>::func;
 
+// z[N] <- Montgomery(x[N], y[N], p[N])
+template<size_t N, class Tag>
+struct Mont {
+	static inline void func(Unit *z, const Unit *x, const Unit *y, const Unit *p)
+	{
+#if 0
+		Unit xy[N * 2];
+		MulPre<N, Gtag>::f(xy, x, y);
+		fpDbl_modMontC(z, xy, p);
+#else
+		const Unit rp = p[-1];
+		Unit buf[N * 2 + 2];
+		Unit *c = buf;
+		Mul_UnitPre<N, Tag>::f(c, x, y[0]); // x * y[0]
+		Unit q = c[0] * rp;
+		Unit t[N + 2];
+		Mul_UnitPre<N, Tag>::f(t, p, q); // p * q
+		t[N + 1] = 0; // always zero
+		c[N + 1] = AddNC<N + 1, Tag>::f(c, c, t);
+		c++;
+		for (size_t i = 1; i < N; i++) {
+			Mul_UnitPre<N, Tag>::f(t, x, y[i]);
+			c[N + 1] = AddNC<N + 1, Tag>::f(c, c, t);
+			q = c[0] * rp;
+			Mul_UnitPre<N, Tag>::f(t, p, q);
+			AddNC<N + 2, Tag>::f(c, c, t);
+			c++;
+		}
+		if (c[N]) {
+			SubNC<N, Tag>::f(z, c, p);
+		} else {
+			if (SubNC<N, Tag>::f(z, c, p)) {
+				memcpy(z, c, N * sizeof(Unit));
+			}
+		}
+#endif
+	}
+	static const void4u f;
+};
+
+template<size_t N, class Tag>
+const void4u Mont<N, Tag>::f = Mont<N, Tag>::func;
+
+// z[N] <- Montgomery(x[N], x[N], p[N])
+template<size_t N, class Tag>
+struct SqrMont {
+	static inline void func(Unit *y, const Unit *x, const Unit *p)
+	{
+#if 0
+		Unit xx[N * 2];
+		SqrPre<N, Tag>::f(xx, x);
+		MontRed<N, Tag>(y, xx, p);
+#else
+		Mont<N, Tag>::f(y, x, x, p);
+#endif
+	}
+	static const void3u f;
+};
+template<size_t N, class Tag>
+const void3u SqrMont<N, Tag>::f = SqrMont<N, Tag>::func;
+
+// z[N] <- (x[N] * y[N]) % p[N]
+template<size_t N, class Tag>
+struct Mul {
+	static inline void func(Unit *z, const Unit *x, const Unit *y, const Unit *p)
+	{
+		Unit xy[N * 2];
+		MulPre<N, Tag>::f(xy, x, y);
+		Dbl_Mod<N, Tag>::f(z, xy, p);
+	}
+	static const void4u f;
+};
+template<size_t N, class Tag>
+const void4u Mul<N, Tag>::f = Mul<N, Tag>::func;
+
+// y[N] <- (x[N] * x[N]) % p[N]
+template<size_t N, class Tag>
+struct Sqr {
+	static inline void func(Unit *y, const Unit *x, const Unit *p)
+	{
+		Unit xx[N * 2];
+		SqrPre<N, Tag>::f(xx, x);
+		Dbl_Mod<N, Tag>::f(y, xx, p);
+	}
+	static const void3u f;
+};
+template<size_t N, class Tag>
+const void3u Sqr<N, Tag>::f = Sqr<N, Tag>::func;
+
+
 } } // mcl::fp
 
 #ifdef MCL_USE_LLVM