Merge branch 'dev'

include/mcl/fp_tower.hpp

@@ -125,8 +125,6 @@ public:
 	static void addPre(FpDblT& z, const FpDblT& x, const FpDblT& y) { Fp::op_.fpDbl_addPre(z.v_, x.v_, y.v_); }
 	static void subPre(FpDblT& z, const FpDblT& x, const FpDblT& y) { Fp::op_.fpDbl_subPre(z.v_, x.v_, y.v_); }
 #endif
-	static void mulPreC(FpDblT& xy, const Fp& x, const Fp& y) { Fp::op_.fpDbl_mulPre(xy.v_, x.v_, y.v_); }
-	static void sqrPreC(FpDblT& xx, const Fp& x) { Fp::op_.fpDbl_sqrPre(xx.v_, x.v_); }
 	/*
 		mul(z, x, y) = mulPre(xy, x, y) + mod(z, xy)
 	*/
@@ -152,16 +150,8 @@ public:
 		subPre = fp::func_ptr_cast<void (*)(FpDblT&, const FpDblT&, const FpDblT&)>(op.fpDbl_subPre);
 		if (subPre == 0) subPre = subPreC;
 #endif
-		if (op.fpDbl_mulPreA_) {
+		mulPre = fp::func_ptr_cast<void (*)(FpDblT&, const Fp&, const Fp&)>(op.fpDbl_mulPre);
-			mulPre = fp::func_ptr_cast<void (*)(FpDblT&, const Fp&, const Fp&)>(op.fpDbl_mulPreA_);
+		sqrPre = fp::func_ptr_cast<void (*)(FpDblT&, const Fp&)>(op.fpDbl_sqrPre);
-		} else {
-			mulPre = mulPreC;
-		}
-		if (op.fpDbl_sqrPreA_) {
-			sqrPre = fp::func_ptr_cast<void (*)(FpDblT&, const Fp&)>(op.fpDbl_sqrPreA_);
-		} else {
-			sqrPre = sqrPreC;
-		}
 	}
 	void operator+=(const FpDblT& x) { add(*this, *this, x); }
 	void operator-=(const FpDblT& x) { sub(*this, *this, x); }
@@ -690,9 +680,9 @@ struct Fp2DblT {
 			mulPre = fp::func_ptr_cast<void (*)(Fp2DblT&, const Fp2&, const Fp2&)>(op.fp2Dbl_mulPreA_);
 		} else {
 			if (op.isFullBit) {
-				mulPre = fp2Dbl_mulPreW<true>;
+				mulPre = fp2Dbl_mulPreTW<true>;
 			} else {
-				mulPre = fp2Dbl_mulPreW<false>;
+				mulPre = fp2Dbl_mulPreTW<false>;
 			}
 		}
 		if (op.fp2Dbl_sqrPreA_) {
@@ -710,7 +700,7 @@ struct Fp2DblT {
 		@note mod of NIST_P192 is fast
 	*/
 	template<bool isFullBit>
-	static void fp2Dbl_mulPreW(Fp2DblT& z, const Fp2& x, const Fp2& y)
+	static void fp2Dbl_mulPreTW(Fp2DblT& z, const Fp2& x, const Fp2& y)
 	{
 		const Fp& a = x.a;
 		const Fp& b = x.b;

include/mcl/op.hpp

@@ -221,8 +221,6 @@ struct Op {
 	void2u fp2_sqrA_;
 	void3u fpDbl_addA_;
 	void3u fpDbl_subA_;
-	void3u fpDbl_mulPreA_;
-	void2u fpDbl_sqrPreA_;
 	void2u fpDbl_modA_;
 	void3u fp2Dbl_mulPreA_;
 	void2u fp2Dbl_sqrPreA_;
@@ -308,8 +306,6 @@ struct Op {
 		fp2_sqrA_ = 0;
 		fpDbl_addA_ = 0;
 		fpDbl_subA_ = 0;
-		fpDbl_mulPreA_ = 0;
-		fpDbl_sqrPreA_ = 0;
 		fpDbl_modA_ = 0;
 		fp2Dbl_mulPreA_ = 0;
 		fp2Dbl_sqrPreA_ = 0;

src/fp.cpp

@@ -426,6 +426,7 @@ void setWasmOp(Op& op)
 	op.fp_mul = mcl::mulMontT<N>;
 	op.fp_sqr = mcl::sqrMontT<N>;
 	op.fpDbl_mulPre = mulT<N>;
+//	op.fpDbl_sqrPre = sqrT<N>;
 	op.fpDbl_mod = modT<N>;
 }
 #endif

src/fp_generator.hpp

@@ -421,12 +421,12 @@ private:
 		setFuncInfo(prof_, suf, "Dbl_subPre", op.fpDbl_subPre, getCurr());
 
 		align(16);
-		op.fpDbl_mulPreA_ = gen_fpDbl_mulPre();
+		gen_fpDbl_mulPre(op.fpDbl_mulPre);
-		setFuncInfo(prof_, suf, "Dbl_mulPre", op.fpDbl_mulPreA_, getCurr());
+		setFuncInfo(prof_, suf, "Dbl_mulPre", op.fpDbl_mulPre, getCurr());
 
 		align(16);
-		op.fpDbl_sqrPreA_ = gen_fpDbl_sqrPre();
+		gen_fpDbl_sqrPre(op.fpDbl_sqrPre);
-		setFuncInfo(prof_, suf, "Dbl_sqrPre", op.fpDbl_sqrPreA_, getCurr());
+		setFuncInfo(prof_, suf, "Dbl_sqrPre", op.fpDbl_sqrPre, getCurr());
 
 		align(16);
 		op.fp2_addA_ = gen_fp2_add();
@@ -2373,36 +2373,35 @@ private:
 		cmovc_rr(zp, keep);
 		store_mr(z, zp);
 	}
-	void2u gen_fpDbl_sqrPre()
+	void gen_fpDbl_sqrPre(void2u& f)
 	{
 		void2u func = getCurr<void2u>();
 		if (pn_ == 2 && useMulx_) {
 			StackFrame sf(this, 2, 7 | UseRDX);
 			sqrPre2(sf.p[0], sf.p[1], sf.t);
-			return func;
+			f = func;
 		}
 		if (pn_ == 3) {
 			StackFrame sf(this, 3, 10 | UseRDX);
 			Pack t = sf.t;
 			t.append(sf.p[2]);
 			sqrPre3(sf.p[0], sf.p[1], t);
-			return func;
+			f = func;
 		}
 		if (pn_ == 4 && useMulx_) {
 			StackFrame sf(this, 3, 10 | UseRDX);
 			Pack t = sf.t;
 			t.append(sf.p[2]);
 			sqrPre4(sf.p[0], sf.p[1], t);
-			return func;
+			f = func;
 		}
 		if (pn_ == 6 && useMulx_ && useAdx_) {
 			StackFrame sf(this, 3, 10 | UseRDX, 6 * 8);
 			Pack t = sf.t;
 			t.append(sf.p[2]);
 			sqrPre6(sf.p[0], sf.p[1], t);
-			return func;
+			f = func;
 		}
-		return 0;
 #if 0
 #ifdef XBYAK64_WIN
 		mov(r8, rdx);
@@ -2413,18 +2412,18 @@ private:
 		return func;
 #endif
 	}
-	void3u gen_fpDbl_mulPre()
+	void gen_fpDbl_mulPre(void3u& f)
 	{
 		void3u func = getCurr<void3u>();
 		if (pn_ == 2 && useMulx_) {
 			StackFrame sf(this, 3, 5 | UseRDX);
 			mulPre2(sf.p[0], sf.p[1], sf.p[2], sf.t);
-			return func;
+			f = func;
 		}
 		if (pn_ == 3) {
 			StackFrame sf(this, 3, 10 | UseRDX);
 			mulPre3(sf.p[0], sf.p[1], sf.p[2], sf.t);
-			return func;
+			f = func;
 		}
 		if (pn_ == 4) {
 			/*
@@ -2437,7 +2436,7 @@ private:
 		L(mulPreL); // called only from asm code
 			mulPre4(gp0, gp1, gp2, sf.t);
 			ret();
-			return func;
+			f = func;
 		}
 		if (pn_ == 6 && useAdx_) {
 			StackFrame sf(this, 3, 10 | UseRDX, 0, false);
@@ -2446,9 +2445,8 @@ private:
 		L(mulPreL); // called only from asm code
 			mulPre6(sf.t);
 			ret();
-			return func;
+			f = func;
 		}
-		return 0;
 	}
 	static inline void debug_put_inner(const uint64_t *ptr, int n)
 	{

src/fp_static_code.hpp

@@ -65,8 +65,8 @@ void setStaticCode(mcl::fp::Op& op)
 		op.fpDbl_subA_ = mclx_FpDbl_sub;
 		op.fpDbl_addPre = mclx_FpDbl_addPre;
 		op.fpDbl_subPre = mclx_FpDbl_subPre;
-		op.fpDbl_mulPreA_ = mclx_FpDbl_mulPre;
+		op.fpDbl_mulPre = mclx_FpDbl_mulPre;
-		op.fpDbl_sqrPreA_ = mclx_FpDbl_sqrPre;
+		op.fpDbl_sqrPre = mclx_FpDbl_sqrPre;
 		op.fp2_addA_ = mclx_Fp2_add;
 		op.fp2_subA_ = mclx_Fp2_sub;
 		op.fp2_negA_ = mclx_Fp2_neg;