Merge branch 'dev'

.github/workflows/main.yml

@@ -1,5 +1,8 @@
 name: test
-on: [push]
+on:
+  push:
+    branches:
+      - '*'
 
 jobs:
   build:
@@ -8,8 +11,8 @@ jobs:
     steps:
     - uses: actions/checkout@v4
     - run: lscpu
-    - run: sudo apt update
-    - run: sudo apt install valgrind nasm
+    - run: apt -y update
+    - run: apt -y install valgrind nasm
     - run: make bin/pairing.exe -j4
     - run: valgrind bin/pairing.exe
     - run: make clean

Makefile

@@ -28,7 +28,7 @@ ifeq ($(MCL_STATIC_CODE),1)
 endif
 ifeq ($(CPU),x86-64)
   MCL_USE_XBYAK?=1
-  TEST_SRC+=mont_fp_test.cpp sq_test.cpp
+  TEST_SRC+=mont_fp_test.cpp #sq_test.cpp
   ifeq ($(MCL_USE_XBYAK),1)
     TEST_SRC+=fp_generator_test.cpp
   endif

include/mcl/fp.hpp

@@ -207,15 +207,7 @@ class FpT : public fp::Serializable<FpT<tag, maxBitSize>,
 	}
 	static inline bool squareRoot(FpT& y, const FpT& x)
 	{
-		if (isMont()) return op_.sq.get(y, x);
-		mpz_class mx, my;
-		bool b = false;
-		x.getMpz(&b, mx);
-		if (!b) return false;
-		b = op_.sq.get(my, mx);
-		if (!b) return false;
-		y.setMpz(&b, my);
-		return b;
+		return op_.sq.get(y, x);
 	}
 	FpT() {}
 	FpT(const FpT& x)

include/mcl/gmp_util.hpp

@@ -784,6 +784,61 @@ class SquareRoot {
 		}
 		return false;
 	}
+	/*
+		solve x^2 = a in Fp
+	*/
+	template<class Fp>
+	bool getCandidate(Fp& x, const Fp& a) const
+	{
+		assert(Fp::getOp().mp == p);
+		if (a.isZero() || a.isOne()) {
+			x = a;
+			return true;
+		}
+		if (r == 1) {
+			// (p + 1) / 4 = (q + 1) / 2
+			Fp::pow(x, a, q_add_1_div_2);
+			return true;
+		}
+		Fp c, d;
+		{
+			bool b;
+			c.setMpz(&b, s);
+			assert(b);
+		}
+		int e = r;
+		Fp::pow(d, a, q);
+		Fp::pow(x, a, q_add_1_div_2); // destroy a if &x == &a
+		Fp dd;
+		Fp b;
+		while (!d.isOne()) {
+			int i = 1;
+			Fp::sqr(dd, d);
+			while (!dd.isOne()) {
+				Fp::sqr(dd, dd);
+				i++;
+				if (i >= e) return false;
+			}
+			assert(e > i);
+			int t = e - i - 1;
+			const int tMax = 30; // int32_t max
+			if (t < tMax) {
+				b = 1 << t;
+			} else {
+				b = 1 << tMax;
+				t -= tMax;
+				for (int j = 0; j < t; j++) {
+					b += b;
+				}
+			}
+			Fp::pow(b, c, b);
+			x *= b;
+			Fp::sqr(c, b);
+			d *= c;
+			e = i;
+		}
+		return true;
+	}
 public:
 	SquareRoot() { clear(); }
 	bool isPrecomputed() const { return isPrecomputed_; }
@@ -838,100 +893,18 @@ class SquareRoot {
 		q_add_1_div_2 = (q + 1) / 2;
 		*pb = true;
 	}
-	/*
-		solve x^2 = a mod p
-	*/
-	bool get(mpz_class& x, const mpz_class& a) const
-	{
-		if (!isPrime) {
-			return false;
-		}
-		if (a == 0) {
-			x = 0;
-			return true;
-		}
-		if (gmp::legendre(a, p) < 0) return false;
-		if (r == 1) {
-			// (p + 1) / 4 = (q + 1) / 2
-			gmp::powMod(x, a, q_add_1_div_2, p);
-			return true;
-		}
-		mpz_class c = s, d;
-		int e = r;
-		gmp::powMod(d, a, q, p);
-		gmp::powMod(x, a, q_add_1_div_2, p); // destroy a if &x == &a
-		mpz_class dd;
-		mpz_class b;
-		while (d != 1) {
-			int i = 1;
-			dd = d * d; dd %= p;
-			while (dd != 1) {
-				dd *= dd; dd %= p;
-				i++;
-			}
-			b = 1;
-			b <<= e - i - 1;
-			gmp::powMod(b, c, b, p);
-			x *= b; x %= p;
-			c = b * b; c %= p;
-			d *= c; d %= p;
-			e = i;
-		}
-		return true;
-	}
-	/*
-		solve x^2 = a in Fp
-	*/
-	template<class Fp>
-	bool get(Fp& x, const Fp& a) const
+	template<class T>
+	bool get(T& x, const T& a) const
 	{
-		assert(Fp::getOp().mp == p);
-		if (a == 0) {
-			x = 0;
-			return true;
-		}
-		{
-			bool b;
-			mpz_class aa;
-			a.getMpz(&b, aa);
-			assert(b);
-			if (gmp::legendre(aa, p) < 0) return false;
-		}
-		if (r == 1) {
-			// (p + 1) / 4 = (q + 1) / 2
-			Fp::pow(x, a, q_add_1_div_2);
-			return true;
-		}
-		Fp c, d;
-		{
-			bool b;
-			c.setMpz(&b, s);
-			assert(b);
-		}
-		int e = r;
-		Fp::pow(d, a, q);
-		Fp::pow(x, a, q_add_1_div_2); // destroy a if &x == &a
-		Fp dd;
-		Fp b;
-		while (!d.isOne()) {
-			int i = 1;
-			Fp::sqr(dd, d);
-			while (!dd.isOne()) {
-				dd *= dd;
-				i++;
-			}
-			b = 1;
-//			b <<= e - i - 1;
-			for (int j = 0; j < e - i - 1; j++) {
-				b += b;
+		T t, t2;
+		if (getCandidate(t, a)) {
+			T::sqr(t2, t);
+			if (t2 == a) {
+				x = t;
+				return true;
 			}
-			Fp::pow(b, c, b);
-			x *= b;
-			Fp::sqr(c, b);
-			d *= c;
-			e = i;
 		}
-		return true;
+		return false;
 	}
 	bool operator==(const SquareRoot& rhs) const
 	{

include/mcl/op.hpp

@@ -29,7 +29,7 @@
 
 namespace mcl {
 
-static const int version = 0x210; /* 0xABC = A.BC */
+static const int version = 0x211; /* 0xABC = A.BC */
 
 /*
 	specifies available string format mode for X::setIoMode()

include/mcl/vint.hpp

@@ -696,6 +696,9 @@ class Vint {
 	// logical left shift (copy sign)
 	static void shl(Vint& y, const Vint& x, size_t shiftBit)
 	{
+if (shiftBit > MCL_MAX_BIT_SIZE*2) {
+	printf("shiftBit=%zd\n", shiftBit);
+}
 		assert(shiftBit <= MCL_MAX_BIT_SIZE * 2); // many be too big
 		size_t xn = x.size();
 		size_t yn = xn + (shiftBit + UnitBitSize - 1) / UnitBitSize;

src/bint_impl.hpp

@@ -46,50 +46,18 @@ struct UnrollMulLowT<1, 1> {
 } // impl
 
 #if MCL_BINT_ASM != 1
-#ifdef MCL_WASM32
-inline uint64_t load8byte(const uint32_t *x)
-{
-	return x[0] | (uint64_t(x[1]) << 32);
-}
-inline void store8byte(uint32_t *x, uint64_t v)
-{
-	x[0] = uint32_t(v);
-	x[1] = uint32_t(v >> 32);
-}
-#endif
 template<size_t N>
 Unit addT(Unit *z, const Unit *x, const Unit *y)
 {
-#if defined(MCL_WASM32) && MCL_SIZEOF_UNIT == 4
+#ifdef MCL_WASM32
 	// wasm32 supports 64-bit add
-#if 1
 	uint64_t c = 0;
 	for (size_t i = 0; i < N; i++) {
 		uint64_t v = uint64_t(x[i]) + y[i] + c;
 		z[i] = uint32_t(v);
 		c = v >> 32;
 	}
 	return uint32_t(c);
-#else
-	uint32_t c = 0;
-	for (size_t i = 0; i < N - 1; i += 2) {
-		uint64_t xc = load8byte(x + i) + c;
-		c = xc < c;
-		uint64_t yi = load8byte(y + i);
-		xc += yi;
-		c += xc < yi;
-		store8byte(z + i, xc);
-	}
-	if (N & 1) {
-		uint32_t xc = x[N - 1] + c;
-		c = xc < c;
-		uint32_t yi = y[N - 1];
-		xc += yi;
-		c += xc < yi;
-		z[N - 1] = xc;
-	}
-	return c;
-#endif
 #else
 	Unit c = 0;
 	for (size_t i = 0; i < N; i++) {
@@ -107,36 +75,15 @@ Unit addT(Unit *z, const Unit *x, const Unit *y)
 template<size_t N, typename T>
 Unit subT(Unit *z, const T *x, const Unit *y)
 {
-#if defined(MCL_WASM32) && MCL_SIZEOF_UNIT == 4
+#ifdef MCL_WASM32
 	// wasm32 supports 64-bit sub
-#if 1
 	uint64_t c = 0;
 	for (size_t i = 0; i < N; i++) {
 		uint64_t v = uint64_t(x[i]) - y[i] - c;
 		z[i] = uint32_t(v);
 		c = v >> 63;
 	}
 	return c;
-#else
-	uint32_t c = 0;
-	for (size_t i = 0; i < N - 1; i += 2) {
-		uint64_t yi = load8byte(y + i);
-		yi += c;
-		c = yi < c;
-		uint64_t xi = load8byte(x + i);
-		c += xi < yi;
-		store8byte(z + i, xi - yi);
-	}
-	if (N & 1) {
-		uint32_t yi = y[N - 1];
-		yi += c;
-		c = yi < c;
-		uint32_t xi = x[N - 1];
-		c += xi < yi;
-		z[N - 1] = xi - yi;
-	}
-	return c;
-#endif
 #else
 	Unit c = 0;
 	for (size_t i = 0; i < N; i++) {
@@ -203,7 +150,7 @@ Unit mulUnitT(T *z, const Unit *x, Unit y)
 template<size_t N, typename T>
 Unit mulUnitAddT(T *z, const Unit *x, Unit y)
 {
-#if defined(MCL_WASM32) && MCL_SIZEOF_UNIT == 4
+#ifdef MCL_WASM32
 	uint64_t y_ = y;
 	uint64_t v = z[0] + x[0] * y_;
 	z[0] = uint32_t(v);