diff --git a/include/mcl/fp.hpp b/include/mcl/fp.hpp
index f53c7a6..b85b691 100644
--- a/include/mcl/fp.hpp
+++ b/include/mcl/fp.hpp
@@ -165,6 +165,8 @@ public:
 		if (sqr == 0) sqr = sqrC;
 		mul2 = fp::func_ptr_cast<void (*)(FpT& y, const FpT& x)>(op_.fp_mul2A_);
 		if (mul2 == 0) mul2 = mul2C;
+		mul9 = fp::func_ptr_cast<void (*)(FpT& y, const FpT& x)>(op_.fp_mul9A_);
+		if (mul9 == 0) mul9 = mul9C;
 #endif
 		*pb = true;
 	}
@@ -499,6 +501,8 @@ public:
 	static inline void sqrC(FpT& y, const FpT& x) { op_.fp_sqr(y.v_, x.v_, op_.p); }
 	static void (*mul2)(FpT& y, const FpT& x);
 	static inline void mul2C(FpT& y, const FpT& x) { op_.fp_mul2(y.v_, x.v_, op_.p); }
+	static void (*mul9)(FpT& y, const FpT& x);
+	static inline void mul9C(FpT& y, const FpT& x) { mulSmall(y, x, 9); }
 #else
 	static inline void add(FpT& z, const FpT& x, const FpT& y) { op_.fp_add(z.v_, x.v_, y.v_, op_.p); }
 	static inline void sub(FpT& z, const FpT& x, const FpT& y) { op_.fp_sub(z.v_, x.v_, y.v_, op_.p); }
@@ -506,6 +510,7 @@ public:
 	static inline void mul(FpT& z, const FpT& x, const FpT& y) { op_.fp_mul(z.v_, x.v_, y.v_, op_.p); }
 	static inline void sqr(FpT& y, const FpT& x) { op_.fp_sqr(y.v_, x.v_, op_.p); }
 	static inline void mul2(FpT& y, const FpT& x) { op_.fp_mul2(y.v_, x.v_, op_.p); }
+	static inline void mul9(FpT& y, const FpT& x) { mulSmall(y, x, 9); }
 #endif
 	static inline void addPre(FpT& z, const FpT& x, const FpT& y) { op_.fp_addPre(z.v_, x.v_, y.v_); }
 	static inline void subPre(FpT& z, const FpT& x, const FpT& y) { op_.fp_subPre(z.v_, x.v_, y.v_); }
@@ -756,6 +761,7 @@ template<class tag, size_t maxBitSize> void (*FpT<tag, maxBitSize>::neg)(FpT& y,
 template<class tag, size_t maxBitSize> void (*FpT<tag, maxBitSize>::mul)(FpT& z, const FpT& x, const FpT& y);
 template<class tag, size_t maxBitSize> void (*FpT<tag, maxBitSize>::sqr)(FpT& y, const FpT& x);
 template<class tag, size_t maxBitSize> void (*FpT<tag, maxBitSize>::mul2)(FpT& y, const FpT& x);
+template<class tag, size_t maxBitSize> void (*FpT<tag, maxBitSize>::mul9)(FpT& y, const FpT& x);
 #endif
 
 } // mcl
diff --git a/include/mcl/op.hpp b/include/mcl/op.hpp
index 2dd358c..4fa8941 100644
--- a/include/mcl/op.hpp
+++ b/include/mcl/op.hpp
@@ -216,6 +216,7 @@ struct Op {
 	void3u fp_mulA_;
 	void2u fp_sqrA_;
 	void2u fp_mul2A_;
+	void2u fp_mul9A_;
 	void3u fp2_addA_;
 	void3u fp2_subA_;
 	void2u fp2_negA_;
@@ -305,6 +306,7 @@ struct Op {
 		fp_mulA_ = 0;
 		fp_sqrA_ = 0;
 		fp_mul2A_ = 0;
+		fp_mul9A_ = 0;
 		fp2_addA_ = 0;
 		fp2_subA_ = 0;
 		fp2_negA_ = 0;
diff --git a/test/bench.hpp b/test/bench.hpp
index 8378674..11ced31 100644
--- a/test/bench.hpp
+++ b/test/bench.hpp
@@ -118,7 +118,7 @@ void testBench(const G1& P, const G2& Q)
 	CYBOZU_BENCH_C("Fp::mul2      ", C3, Fp::mul2, x, x);
 	CYBOZU_BENCH_C("Fp::mulSmall8 ", C3, Fp::mulSmall, x, x, 8);
 	CYBOZU_BENCH_C("Fp::mulUnit8  ", C3, Fp::mulUnit, x, x, 8);
-	CYBOZU_BENCH_C("Fp::mulSmall9 ", C3, Fp::mulSmall, x, x, 9);
+	CYBOZU_BENCH_C("Fp::mul9      ", C3, Fp::mul9, x, x);
 	CYBOZU_BENCH_C("Fp::mulUnit9  ", C3, Fp::mulUnit, x, x, 9);
 	CYBOZU_BENCH_C("Fp::neg       ", C3, Fp::neg, x, x);
 	CYBOZU_BENCH_C("Fp::mul       ", C3, Fp::mul, x, x, y);