From ac0e9d9813e3f1da294505b5689d13b7d1d95429 Mon Sep 17 00:00:00 2001
From: MITSUNARI Shigeo <herumi@nifty.com>
Date: Thu, 1 Nov 2018 21:48:44 +0900
Subject: [PATCH] enable Fp::sqr for bls12

---
 src/fp.cpp           | 13 ++++++-------
 src/fp_generator.hpp | 26 +++++++++++++++++---------
 2 files changed, 23 insertions(+), 16 deletions(-)

diff --git a/src/fp.cpp b/src/fp.cpp
index 8db755d..473b26c 100644
--- a/src/fp.cpp
+++ b/src/fp.cpp
@@ -499,18 +499,17 @@ bool Op::init(const mpz_class& _p, size_t maxBitSize, Mode mode, size_t mclMaxBi
 		fpDbl_mod = &mcl::vint::mcl_fpDbl_mod_SECP256K1;
 	}
 #endif
-	if (!fp::initForMont(*this, p, mode)) return false;
-	{
-		bool b;
-		sq.set(&b, mp);
-		if (!b) return false;
-	}
 	if (N * UnitBitSize <= 256) {
 		hash = sha256;
 	} else {
 		hash = sha512;
 	}
-	return true;
+	{
+		bool b;
+		sq.set(&b, mp);
+		if (!b) return false;
+	}
+	return fp::initForMont(*this, p, mode);
 }
 
 void copyUnitToByteAsLE(uint8_t *dst, const Unit *src, size_t byteSize)
diff --git a/src/fp_generator.hpp b/src/fp_generator.hpp
index 8f55607..0749e44 100644
--- a/src/fp_generator.hpp
+++ b/src/fp_generator.hpp
@@ -836,13 +836,19 @@ private:
 			gen_montMul4();
 			return func;
 		}
-		if (pn_ == 6 && useMulx_ && useAdx_) {
+		if (pn_ == 6 && !isFullBit_ && useMulx_ && useAdx_) {
 //			gen_montMul6(p_, rp_);
-			StackFrame sf(this, 3, 10 | UseRDX, (1 + 12) * 8);
-			mov(ptr[rsp + 12 * 8], gp0);
+			if (mulPreL.getAddress() == 0 || fpDbl_modL.getAddress() == 0) return 0;
+			StackFrame sf(this, 3, 10 | UseRDX, 12 * 8);
+			/*
+				use xm3
+				rsp
+				[0, ..12 * 8) ; mul(x, y)
+			*/
+			movq(xm3, gp0);
 			mov(gp0, rsp);
 			call(mulPreL); // gp0, x, y
-			mov(gp0, ptr[rsp + 12 * 8]);
+			movq(gp0, xm3);
 			mov(gp1, rsp);
 			call(fpDbl_modL);
 			return func;
@@ -1154,7 +1160,7 @@ private:
 			ret();
 			return func;
 		}
-		if (pn_ == 6 && !isFullBit_ && useAdx_) {
+		if (pn_ == 6 && !isFullBit_ && useMulx_ && useAdx_) {
 			StackFrame sf(this, 3, 10 | UseRDX, 0, false);
 			call(fpDbl_modL);
 			sf.close();
@@ -1203,20 +1209,22 @@ private:
 #endif
 			return func;
 		}
-		if (pn_ == 6 && useMulx_ && useAdx_) {
-			StackFrame sf(this, 3, 10 | UseRDX, (1 + 12 + 6) * 8);
+		if (pn_ == 6 && !isFullBit_ && useMulx_ && useAdx_) {
+			if (fpDbl_modL.getAddress() == 0) return 0;
+			StackFrame sf(this, 3, 10 | UseRDX, (12 + 6) * 8);
 			/*
+				use xm3
 				rsp
-				[(12 + 6 * 8] ; gp0
 				[6 * 8, (12 + 6) * 8) ; sqrPre(x, x)
 				[0..6 * 8) ; stack for sqrPre6
 			*/
-			mov(ptr[rsp + (12 + 6) * 8], gp0);
+			movq(xm3, gp0);
 			Pack t = sf.t;
 			t.append(sf.p[2]);
 			// sqrPre6 uses 6 * 8 bytes stack
 			sqrPre6(rsp + 6 * 8, sf.p[1], t);
 			mov(gp0, ptr[rsp + (12 + 6) * 8]);
+			movq(gp0, xm3);
 			lea(gp1, ptr[rsp + 6 * 8]);
 			call(fpDbl_modL);
 			return func;