@pulse-compute/crypto 1.0.0-beta.4 → 1.0.0-beta.6

package/guests/es256-rustcrypto/source/bearssl/src/int/i31_moddiv.c

@@ -0,0 +1,488 @@
+/*
+ * Copyright (c) 2018 Thomas Pornin <pornin@bolet.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining 
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be 
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "inner.h"
+
+/*
+ * In this file, we handle big integers with a custom format, i.e.
+ * without the usual one-word header. Value is split into 31-bit words,
+ * each stored in a 32-bit slot (top bit is zero) in little-endian
+ * order. The length (in words) is provided explicitly. In some cases,
+ * the value can be negative (using two's complement representation). In
+ * some cases, the top word is allowed to have a 32th bit.
+ */
+
+/*
+ * Negate big integer conditionally. The value consists of 'len' words,
+ * with 31 bits in each word (the top bit of each word should be 0,
+ * except possibly for the last word). If 'ctl' is 1, the negation is
+ * computed; otherwise, if 'ctl' is 0, then the value is unchanged.
+ */
+static void
+cond_negate(uint32_t *a, size_t len, uint32_t ctl)
+{
+	size_t k;
+	uint32_t cc, xm;
+
+	cc = ctl;
+	xm = -ctl >> 1;
+	for (k = 0; k < len; k ++) {
+		uint32_t aw;
+
+		aw = a[k];
+		aw = (aw ^ xm) + cc;
+		a[k] = aw & 0x7FFFFFFF;
+		cc = aw >> 31;
+	}
+}
+
+/*
+ * Finish modular reduction. Rules on input parameters:
+ *
+ *   if neg = 1, then -m <= a < 0
+ *   if neg = 0, then 0 <= a < 2*m
+ *
+ * If neg = 0, then the top word of a[] may use 32 bits.
+ *
+ * Also, modulus m must be odd.
+ */
+static void
+finish_mod(uint32_t *a, size_t len, const uint32_t *m, uint32_t neg)
+{
+	size_t k;
+	uint32_t cc, xm, ym;
+
+	/*
+	 * First pass: compare a (assumed nonnegative) with m.
+	 * Note that if the final word uses the top extra bit, then
+	 * subtracting m must yield a value less than 2^31, since we
+	 * assumed that a < 2*m.
+	 */
+	cc = 0;
+	for (k = 0; k < len; k ++) {
+		uint32_t aw, mw;
+
+		aw = a[k];
+		mw = m[k];
+		cc = (aw - mw - cc) >> 31;
+	}
+
+	/*
+	 * At this point:
+	 *   if neg = 1, then we must add m (regardless of cc)
+	 *   if neg = 0 and cc = 0, then we must subtract m
+	 *   if neg = 0 and cc = 1, then we must do nothing
+	 */
+	xm = -neg >> 1;
+	ym = -(neg | (1 - cc));
+	cc = neg;
+	for (k = 0; k < len; k ++) {
+		uint32_t aw, mw;
+
+		aw = a[k];
+		mw = (m[k] ^ xm) & ym;
+		aw = aw - mw - cc;
+		a[k] = aw & 0x7FFFFFFF;
+		cc = aw >> 31;
+	}
+}
+
+/*
+ * Compute:
+ *   a <- (a*pa+b*pb)/(2^31)
+ *   b <- (a*qa+b*qb)/(2^31)
+ * The division is assumed to be exact (i.e. the low word is dropped).
+ * If the final a is negative, then it is negated. Similarly for b.
+ * Returned value is the combination of two bits:
+ *   bit 0: 1 if a had to be negated, 0 otherwise
+ *   bit 1: 1 if b had to be negated, 0 otherwise
+ *
+ * Factors pa, pb, qa and qb must be at most 2^31 in absolute value.
+ * Source integers a and b must be nonnegative; top word is not allowed
+ * to contain an extra 32th bit.
+ */
+static uint32_t
+co_reduce(uint32_t *a, uint32_t *b, size_t len,
+	int64_t pa, int64_t pb, int64_t qa, int64_t qb)
+{
+	size_t k;
+	int64_t cca, ccb;
+	uint32_t nega, negb;
+
+	cca = 0;
+	ccb = 0;
+	for (k = 0; k < len; k ++) {
+		uint32_t wa, wb;
+		uint64_t za, zb;
+		uint64_t tta, ttb;
+
+		/*
+		 * Since:
+		 *   |pa| <= 2^31
+		 *   |pb| <= 2^31
+		 *   0 <= wa <= 2^31 - 1
+		 *   0 <= wb <= 2^31 - 1
+		 *   |cca| <= 2^32 - 1
+		 * Then:
+		 *   |za| <= (2^31-1)*(2^32) + (2^32-1) = 2^63 - 1
+		 *
+		 * Thus, the new value of cca is such that |cca| <= 2^32 - 1.
+		 * The same applies to ccb.
+		 */
+		wa = a[k];
+		wb = b[k];
+		za = wa * (uint64_t)pa + wb * (uint64_t)pb + (uint64_t)cca;
+		zb = wa * (uint64_t)qa + wb * (uint64_t)qb + (uint64_t)ccb;
+		if (k > 0) {
+			a[k - 1] = za & 0x7FFFFFFF;
+			b[k - 1] = zb & 0x7FFFFFFF;
+		}
+
+		/*
+		 * For the new values of cca and ccb, we need a signed
+		 * right-shift; since, in C, right-shifting a signed
+		 * negative value is implementation-defined, we use a
+		 * custom portable sign extension expression.
+		 */
+#define M   ((uint64_t)1 << 32)
+		tta = za >> 31;
+		ttb = zb >> 31;
+		tta = (tta ^ M) - M;
+		ttb = (ttb ^ M) - M;
+		cca = *(int64_t *)&tta;
+		ccb = *(int64_t *)&ttb;
+#undef M
+	}
+	a[len - 1] = (uint32_t)cca;
+	b[len - 1] = (uint32_t)ccb;
+
+	nega = (uint32_t)((uint64_t)cca >> 63);
+	negb = (uint32_t)((uint64_t)ccb >> 63);
+	cond_negate(a, len, nega);
+	cond_negate(b, len, negb);
+	return nega | (negb << 1);
+}
+
+/*
+ * Compute:
+ *   a <- (a*pa+b*pb)/(2^31) mod m
+ *   b <- (a*qa+b*qb)/(2^31) mod m
+ *
+ * m0i is equal to -1/m[0] mod 2^31.
+ *
+ * Factors pa, pb, qa and qb must be at most 2^31 in absolute value.
+ * Source integers a and b must be nonnegative; top word is not allowed
+ * to contain an extra 32th bit.
+ */
+static void
+co_reduce_mod(uint32_t *a, uint32_t *b, size_t len,
+	int64_t pa, int64_t pb, int64_t qa, int64_t qb,
+	const uint32_t *m, uint32_t m0i)
+{
+	size_t k;
+	int64_t cca, ccb;
+	uint32_t fa, fb;
+
+	cca = 0;
+	ccb = 0;
+	fa = ((a[0] * (uint32_t)pa + b[0] * (uint32_t)pb) * m0i) & 0x7FFFFFFF;
+	fb = ((a[0] * (uint32_t)qa + b[0] * (uint32_t)qb) * m0i) & 0x7FFFFFFF;
+	for (k = 0; k < len; k ++) {
+		uint32_t wa, wb;
+		uint64_t za, zb;
+		uint64_t tta, ttb;
+
+		/*
+		 * In this loop, carries 'cca' and 'ccb' always fit on
+		 * 33 bits (in absolute value).
+		 */
+		wa = a[k];
+		wb = b[k];
+		za = wa * (uint64_t)pa + wb * (uint64_t)pb
+			+ m[k] * (uint64_t)fa + (uint64_t)cca;
+		zb = wa * (uint64_t)qa + wb * (uint64_t)qb
+			+ m[k] * (uint64_t)fb + (uint64_t)ccb;
+		if (k > 0) {
+			a[k - 1] = (uint32_t)za & 0x7FFFFFFF;
+			b[k - 1] = (uint32_t)zb & 0x7FFFFFFF;
+		}
+
+#define M   ((uint64_t)1 << 32)
+		tta = za >> 31;
+		ttb = zb >> 31;
+		tta = (tta ^ M) - M;
+		ttb = (ttb ^ M) - M;
+		cca = *(int64_t *)&tta;
+		ccb = *(int64_t *)&ttb;
+#undef M
+	}
+	a[len - 1] = (uint32_t)cca;
+	b[len - 1] = (uint32_t)ccb;
+
+	/*
+	 * At this point:
+	 *   -m <= a < 2*m
+	 *   -m <= b < 2*m
+	 * (this is a case of Montgomery reduction)
+	 * The top word of 'a' and 'b' may have a 32-th bit set.
+	 * We may have to add or subtract the modulus.
+	 */
+	finish_mod(a, len, m, (uint32_t)((uint64_t)cca >> 63));
+	finish_mod(b, len, m, (uint32_t)((uint64_t)ccb >> 63));
+}
+
+/* see inner.h */
+uint32_t
+br_i31_moddiv(uint32_t *x, const uint32_t *y, const uint32_t *m, uint32_t m0i,
+	uint32_t *t)
+{
+	/*
+	 * Algorithm is an extended binary GCD. We maintain four values
+	 * a, b, u and v, with the following invariants:
+	 *
+	 *   a * x = y * u mod m
+	 *   b * x = y * v mod m
+	 *
+	 * Starting values are:
+	 *
+	 *   a = y
+	 *   b = m
+	 *   u = x
+	 *   v = 0
+	 *
+	 * The formal definition of the algorithm is a sequence of steps:
+	 *
+	 *   - If a is even, then a <- a/2 and u <- u/2 mod m.
+	 *   - Otherwise, if b is even, then b <- b/2 and v <- v/2 mod m.
+	 *   - Otherwise, if a > b, then a <- (a-b)/2 and u <- (u-v)/2 mod m.
+	 *   - Otherwise, b <- (b-a)/2 and v <- (v-u)/2 mod m.
+	 *
+	 * Algorithm stops when a = b. At that point, they both are equal
+	 * to GCD(y,m); the modular division succeeds if that value is 1.
+	 * The result of the modular division is then u (or v: both are
+	 * equal at that point).
+	 *
+	 * Each step makes either a or b shrink by at least one bit; hence,
+	 * if m has bit length k bits, then 2k-2 steps are sufficient.
+	 *
+	 *
+	 * Though complexity is quadratic in the size of m, the bit-by-bit
+	 * processing is not very efficient. We can speed up processing by
+	 * remarking that the decisions are taken based only on observation
+	 * of the top and low bits of a and b.
+	 *
+	 * In the loop below, at each iteration, we use the two top words
+	 * of a and b, and the low words of a and b, to compute reduction
+	 * parameters pa, pb, qa and qb such that the new values for a
+	 * and b are:
+	 *
+	 *   a' = (a*pa + b*pb) / (2^31)
+	 *   b' = (a*qa + b*qb) / (2^31)
+	 *
+	 * the division being exact.
+	 *
+	 * Since the choices are based on the top words, they may be slightly
+	 * off, requiring an optional correction: if a' < 0, then we replace
+	 * pa with -pa, and pb with -pb. The total length of a and b is
+	 * thus reduced by at least 30 bits at each iteration.
+	 *
+	 * The stopping conditions are still the same, though: when a
+	 * and b become equal, they must be both odd (since m is odd,
+	 * the GCD cannot be even), therefore the next operation is a
+	 * subtraction, and one of the values becomes 0. At that point,
+	 * nothing else happens, i.e. one value is stuck at 0, and the
+	 * other one is the GCD.
+	 */
+	size_t len, k;
+	uint32_t *a, *b, *u, *v;
+	uint32_t num, r;
+
+	len = (m[0] + 31) >> 5;
+	a = t;
+	b = a + len;
+	u = x + 1;
+	v = b + len;
+	memcpy(a, y + 1, len * sizeof *y);
+	memcpy(b, m + 1, len * sizeof *m);
+	memset(v, 0, len * sizeof *v);
+
+	/*
+	 * Loop below ensures that a and b are reduced by some bits each,
+	 * for a total of at least 30 bits.
+	 */
+	for (num = ((m[0] - (m[0] >> 5)) << 1) + 30; num >= 30; num -= 30) {
+		size_t j;
+		uint32_t c0, c1;
+		uint32_t a0, a1, b0, b1;
+		uint64_t a_hi, b_hi;
+		uint32_t a_lo, b_lo;
+		int64_t pa, pb, qa, qb;
+		int i;
+
+		/*
+		 * Extract top words of a and b. If j is the highest
+		 * index >= 1 such that a[j] != 0 or b[j] != 0, then we want
+		 * (a[j] << 31) + a[j - 1], and (b[j] << 31) + b[j - 1].
+		 * If a and b are down to one word each, then we use a[0]
+		 * and b[0].
+		 */
+		c0 = (uint32_t)-1;
+		c1 = (uint32_t)-1;
+		a0 = 0;
+		a1 = 0;
+		b0 = 0;
+		b1 = 0;
+		j = len;
+		while (j -- > 0) {
+			uint32_t aw, bw;
+
+			aw = a[j];
+			bw = b[j];
+			a0 ^= (a0 ^ aw) & c0;
+			a1 ^= (a1 ^ aw) & c1;
+			b0 ^= (b0 ^ bw) & c0;
+			b1 ^= (b1 ^ bw) & c1;
+			c1 = c0;
+			c0 &= (((aw | bw) + 0x7FFFFFFF) >> 31) - (uint32_t)1;
+		}
+
+		/*
+		 * If c1 = 0, then we grabbed two words for a and b.
+		 * If c1 != 0 but c0 = 0, then we grabbed one word. It
+		 * is not possible that c1 != 0 and c0 != 0, because that
+		 * would mean that both integers are zero.
+		 */
+		a1 |= a0 & c1;
+		a0 &= ~c1;
+		b1 |= b0 & c1;
+		b0 &= ~c1;
+		a_hi = ((uint64_t)a0 << 31) + a1;
+		b_hi = ((uint64_t)b0 << 31) + b1;
+		a_lo = a[0];
+		b_lo = b[0];
+
+		/*
+		 * Compute reduction factors:
+		 *
+		 *   a' = a*pa + b*pb
+		 *   b' = a*qa + b*qb
+		 *
+		 * such that a' and b' are both multiple of 2^31, but are
+		 * only marginally larger than a and b.
+		 */
+		pa = 1;
+		pb = 0;
+		qa = 0;
+		qb = 1;
+		for (i = 0; i < 31; i ++) {
+			/*
+			 * At each iteration:
+			 *
+			 *   a <- (a-b)/2 if: a is odd, b is odd, a_hi > b_hi
+			 *   b <- (b-a)/2 if: a is odd, b is odd, a_hi <= b_hi
+			 *   a <- a/2 if: a is even
+			 *   b <- b/2 if: a is odd, b is even
+			 *
+			 * We multiply a_lo and b_lo by 2 at each
+			 * iteration, thus a division by 2 really is a
+			 * non-multiplication by 2.
+			 */
+			uint32_t r, oa, ob, cAB, cBA, cA;
+			uint64_t rz;
+
+			/*
+			 * r = GT(a_hi, b_hi)
+			 * But the GT() function works on uint32_t operands,
+			 * so we inline a 64-bit version here.
+			 */
+			rz = b_hi - a_hi;
+			r = (uint32_t)((rz ^ ((a_hi ^ b_hi)
+				& (a_hi ^ rz))) >> 63);
+
+			/*
+			 * cAB = 1 if b must be subtracted from a
+			 * cBA = 1 if a must be subtracted from b
+			 * cA = 1 if a is divided by 2, 0 otherwise
+			 *
+			 * Rules:
+			 *
+			 *   cAB and cBA cannot be both 1.
+			 *   if a is not divided by 2, b is.
+			 */
+			oa = (a_lo >> i) & 1;
+			ob = (b_lo >> i) & 1;
+			cAB = oa & ob & r;
+			cBA = oa & ob & NOT(r);
+			cA = cAB | NOT(oa);
+
+			/*
+			 * Conditional subtractions.
+			 */
+			a_lo -= b_lo & -cAB;
+			a_hi -= b_hi & -(uint64_t)cAB;
+			pa -= qa & -(int64_t)cAB;
+			pb -= qb & -(int64_t)cAB;
+			b_lo -= a_lo & -cBA;
+			b_hi -= a_hi & -(uint64_t)cBA;
+			qa -= pa & -(int64_t)cBA;
+			qb -= pb & -(int64_t)cBA;
+
+			/*
+			 * Shifting.
+			 */
+			a_lo += a_lo & (cA - 1);
+			pa += pa & ((int64_t)cA - 1);
+			pb += pb & ((int64_t)cA - 1);
+			a_hi ^= (a_hi ^ (a_hi >> 1)) & -(uint64_t)cA;
+			b_lo += b_lo & -cA;
+			qa += qa & -(int64_t)cA;
+			qb += qb & -(int64_t)cA;
+			b_hi ^= (b_hi ^ (b_hi >> 1)) & ((uint64_t)cA - 1);
+		}
+
+		/*
+		 * Replace a and b with new values a' and b'.
+		 */
+		r = co_reduce(a, b, len, pa, pb, qa, qb);
+		pa -= pa * ((r & 1) << 1);
+		pb -= pb * ((r & 1) << 1);
+		qa -= qa * (r & 2);
+		qb -= qb * (r & 2);
+		co_reduce_mod(u, v, len, pa, pb, qa, qb, m + 1, m0i);
+	}
+
+	/*
+	 * Now one of the arrays should be 0, and the other contains
+	 * the GCD. If a is 0, then u is 0 as well, and v contains
+	 * the division result.
+	 * Result is correct if and only if GCD is 1.
+	 */
+	r = (a[0] | b[0]) ^ 1;
+	u[0] |= v[0];
+	for (k = 1; k < len; k ++) {
+		r |= a[k] | b[k];
+		u[k] |= v[k];
+	}
+	return EQ0(r);
+}

package/guests/es256-rustcrypto/source/bearssl/src/int/i31_modpow.c

@@ -0,0 +1,65 @@
+/*
+ * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining 
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be 
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "inner.h"
+
+/* see inner.h */
+void
+br_i31_modpow(uint32_t *x,
+	const unsigned char *e, size_t elen,
+	const uint32_t *m, uint32_t m0i, uint32_t *t1, uint32_t *t2)
+{
+	size_t mlen;
+	uint32_t k;
+
+	/*
+	 * 'mlen' is the length of m[] expressed in bytes (including
+	 * the "bit length" first field).
+	 */
+	mlen = ((m[0] + 63) >> 5) * sizeof m[0];
+
+	/*
+	 * Throughout the algorithm:
+	 * -- t1[] is in Montgomery representation; it contains x, x^2,
+	 * x^4, x^8...
+	 * -- The result is accumulated, in normal representation, in
+	 * the x[] array.
+	 * -- t2[] is used as destination buffer for each multiplication.
+	 *
+	 * Note that there is no need to call br_i32_from_monty().
+	 */
+	memcpy(t1, x, mlen);
+	br_i31_to_monty(t1, m);
+	br_i31_zero(x, m[0]);
+	x[1] = 1;
+	for (k = 0; k < ((uint32_t)elen << 3); k ++) {
+		uint32_t ctl;
+
+		ctl = (e[elen - 1 - (k >> 3)] >> (k & 7)) & 1;
+		br_i31_montymul(t2, x, t1, m, m0i);
+		CCOPY(ctl, x, t2, mlen);
+		br_i31_montymul(t2, t1, t1, m, m0i);
+		memcpy(t1, t2, mlen);
+	}
+}

package/guests/es256-rustcrypto/source/bearssl/src/int/i31_modpow2.c

@@ -0,0 +1,160 @@
+/*
+ * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining 
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be 
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "inner.h"
+
+/* see inner.h */
+uint32_t
+br_i31_modpow_opt(uint32_t *x,
+	const unsigned char *e, size_t elen,
+	const uint32_t *m, uint32_t m0i, uint32_t *tmp, size_t twlen)
+{
+	size_t mlen, mwlen;
+	uint32_t *t1, *t2, *base;
+	size_t u, v;
+	uint32_t acc;
+	int acc_len, win_len;
+
+	/*
+	 * Get modulus size.
+	 */
+	mwlen = (m[0] + 63) >> 5;
+	mlen = mwlen * sizeof m[0];
+	mwlen += (mwlen & 1);
+	t1 = tmp;
+	t2 = tmp + mwlen;
+
+	/*
+	 * Compute possible window size, with a maximum of 5 bits.
+	 * When the window has size 1 bit, we use a specific code
+	 * that requires only two temporaries. Otherwise, for a
+	 * window of k bits, we need 2^k+1 temporaries.
+	 */
+	if (twlen < (mwlen << 1)) {
+		return 0;
+	}
+	for (win_len = 5; win_len > 1; win_len --) {
+		if ((((uint32_t)1 << win_len) + 1) * mwlen <= twlen) {
+			break;
+		}
+	}
+
+	/*
+	 * Everything is done in Montgomery representation.
+	 */
+	br_i31_to_monty(x, m);
+
+	/*
+	 * Compute window contents. If the window has size one bit only,
+	 * then t2 is set to x; otherwise, t2[0] is left untouched, and
+	 * t2[k] is set to x^k (for k >= 1).
+	 */
+	if (win_len == 1) {
+		memcpy(t2, x, mlen);
+	} else {
+		memcpy(t2 + mwlen, x, mlen);
+		base = t2 + mwlen;
+		for (u = 2; u < ((unsigned)1 << win_len); u ++) {
+			br_i31_montymul(base + mwlen, base, x, m, m0i);
+			base += mwlen;
+		}
+	}
+
+	/*
+	 * We need to set x to 1, in Montgomery representation. This can
+	 * be done efficiently by setting the high word to 1, then doing
+	 * one word-sized shift.
+	 */
+	br_i31_zero(x, m[0]);
+	x[(m[0] + 31) >> 5] = 1;
+	br_i31_muladd_small(x, 0, m);
+
+	/*
+	 * We process bits from most to least significant. At each
+	 * loop iteration, we have acc_len bits in acc.
+	 */
+	acc = 0;
+	acc_len = 0;
+	while (acc_len > 0 || elen > 0) {
+		int i, k;
+		uint32_t bits;
+
+		/*
+		 * Get the next bits.
+		 */
+		k = win_len;
+		if (acc_len < win_len) {
+			if (elen > 0) {
+				acc = (acc << 8) | *e ++;
+				elen --;
+				acc_len += 8;
+			} else {
+				k = acc_len;
+			}
+		}
+		bits = (acc >> (acc_len - k)) & (((uint32_t)1 << k) - 1);
+		acc_len -= k;
+
+		/*
+		 * We could get exactly k bits. Compute k squarings.
+		 */
+		for (i = 0; i < k; i ++) {
+			br_i31_montymul(t1, x, x, m, m0i);
+			memcpy(x, t1, mlen);
+		}
+
+		/*
+		 * Window lookup: we want to set t2 to the window
+		 * lookup value, assuming the bits are non-zero. If
+		 * the window length is 1 bit only, then t2 is
+		 * already set; otherwise, we do a constant-time lookup.
+		 */
+		if (win_len > 1) {
+			br_i31_zero(t2, m[0]);
+			base = t2 + mwlen;
+			for (u = 1; u < ((uint32_t)1 << k); u ++) {
+				uint32_t mask;
+
+				mask = -EQ(u, bits);
+				for (v = 1; v < mwlen; v ++) {
+					t2[v] |= mask & base[v];
+				}
+				base += mwlen;
+			}
+		}
+
+		/*
+		 * Multiply with the looked-up value. We keep the
+		 * product only if the exponent bits are not all-zero.
+		 */
+		br_i31_montymul(t1, x, t2, m, m0i);
+		CCOPY(NEQ(bits, 0), x, t1, mlen);
+	}
+
+	/*
+	 * Convert back from Montgomery representation, and exit.
+	 */
+	br_i31_from_monty(x, m, m0i);
+	return 1;
+}