preparing to build tox from sources

7 files changed, 2007 insertions, 0 deletions

diff --git a/libs/libsodium/src/include/sodium/private/common.h b/libs/libsodium/src/include/sodium/private/common.h
new file mode 100644
index 0000000000..954d02ccd7
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/common.h
@@ -0,0 +1,246 @@
+#ifndef common_H
+#define common_H 1
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define COMPILER_ASSERT(X) (void) sizeof(char[(X) ? 1 : -1])
+
+#ifdef HAVE_TI_MODE
+# if defined(__SIZEOF_INT128__)
+typedef unsigned __int128 uint128_t;
+# else
+typedef unsigned uint128_t __attribute__((mode(TI)));
+# endif
+#endif
+
+#define ROTL32(X, B) rotl32((X), (B))
+static inline uint32_t
+rotl32(const uint32_t x, const int b)
+{
+    return (x << b) | (x >> (32 - b));
+}
+
+#define ROTL64(X, B) rotl64((X), (B))
+static inline uint64_t
+rotl64(const uint64_t x, const int b)
+{
+    return (x << b) | (x >> (64 - b));
+}
+
+#define ROTR32(X, B) rotr32((X), (B))
+static inline uint32_t
+rotr32(const uint32_t x, const int b)
+{
+    return (x >> b) | (x << (32 - b));
+}
+
+#define ROTR64(X, B) rotr64((X), (B))
+static inline uint64_t
+rotr64(const uint64_t x, const int b)
+{
+    return (x >> b) | (x << (64 - b));
+}
+
+#define LOAD64_LE(SRC) load64_le(SRC)
+static inline uint64_t
+load64_le(const uint8_t src[8])
+{
+#ifdef NATIVE_LITTLE_ENDIAN
+    uint64_t w;
+    memcpy(&w, src, sizeof w);
+    return w;
+#else
+    uint64_t w = (uint64_t) src[0];
+    w |= (uint64_t) src[1] <<  8;
+    w |= (uint64_t) src[2] << 16;
+    w |= (uint64_t) src[3] << 24;
+    w |= (uint64_t) src[4] << 32;
+    w |= (uint64_t) src[5] << 40;
+    w |= (uint64_t) src[6] << 48;
+    w |= (uint64_t) src[7] << 56;
+    return w;
+#endif
+}
+
+#define STORE64_LE(DST, W) store64_le((DST), (W))
+static inline void
+store64_le(uint8_t dst[8], uint64_t w)
+{
+#ifdef NATIVE_LITTLE_ENDIAN
+    memcpy(dst, &w, sizeof w);
+#else
+    dst[0] = (uint8_t) w; w >>= 8;
+    dst[1] = (uint8_t) w; w >>= 8;
+    dst[2] = (uint8_t) w; w >>= 8;
+    dst[3] = (uint8_t) w; w >>= 8;
+    dst[4] = (uint8_t) w; w >>= 8;
+    dst[5] = (uint8_t) w; w >>= 8;
+    dst[6] = (uint8_t) w; w >>= 8;
+    dst[7] = (uint8_t) w;
+#endif
+}
+
+#define LOAD32_LE(SRC) load32_le(SRC)
+static inline uint32_t
+load32_le(const uint8_t src[4])
+{
+#ifdef NATIVE_LITTLE_ENDIAN
+    uint32_t w;
+    memcpy(&w, src, sizeof w);
+    return w;
+#else
+    uint32_t w = (uint32_t) src[0];
+    w |= (uint32_t) src[1] <<  8;
+    w |= (uint32_t) src[2] << 16;
+    w |= (uint32_t) src[3] << 24;
+    return w;
+#endif
+}
+
+#define STORE32_LE(DST, W) store32_le((DST), (W))
+static inline void
+store32_le(uint8_t dst[4], uint32_t w)
+{
+#ifdef NATIVE_LITTLE_ENDIAN
+    memcpy(dst, &w, sizeof w);
+#else
+    dst[0] = (uint8_t) w; w >>= 8;
+    dst[1] = (uint8_t) w; w >>= 8;
+    dst[2] = (uint8_t) w; w >>= 8;
+    dst[3] = (uint8_t) w;
+#endif
+}
+
+/* ----- */
+
+#define LOAD64_BE(SRC) load64_be(SRC)
+static inline uint64_t
+load64_be(const uint8_t src[8])
+{
+#ifdef NATIVE_BIG_ENDIAN
+    uint64_t w;
+    memcpy(&w, src, sizeof w);
+    return w;
+#else
+    uint64_t w = (uint64_t) src[7];
+    w |= (uint64_t) src[6] <<  8;
+    w |= (uint64_t) src[5] << 16;
+    w |= (uint64_t) src[4] << 24;
+    w |= (uint64_t) src[3] << 32;
+    w |= (uint64_t) src[2] << 40;
+    w |= (uint64_t) src[1] << 48;
+    w |= (uint64_t) src[0] << 56;
+    return w;
+#endif
+}
+
+#define STORE64_BE(DST, W) store64_be((DST), (W))
+static inline void
+store64_be(uint8_t dst[8], uint64_t w)
+{
+#ifdef NATIVE_BIG_ENDIAN
+    memcpy(dst, &w, sizeof w);
+#else
+    dst[7] = (uint8_t) w; w >>= 8;
+    dst[6] = (uint8_t) w; w >>= 8;
+    dst[5] = (uint8_t) w; w >>= 8;
+    dst[4] = (uint8_t) w; w >>= 8;
+    dst[3] = (uint8_t) w; w >>= 8;
+    dst[2] = (uint8_t) w; w >>= 8;
+    dst[1] = (uint8_t) w; w >>= 8;
+    dst[0] = (uint8_t) w;
+#endif
+}
+
+#define LOAD32_BE(SRC) load32_be(SRC)
+static inline uint32_t
+load32_be(const uint8_t src[4])
+{
+#ifdef NATIVE_BIG_ENDIAN
+    uint32_t w;
+    memcpy(&w, src, sizeof w);
+    return w;
+#else
+    uint32_t w = (uint32_t) src[3];
+    w |= (uint32_t) src[2] <<  8;
+    w |= (uint32_t) src[1] << 16;
+    w |= (uint32_t) src[0] << 24;
+    return w;
+#endif
+}
+
+#define STORE32_BE(DST, W) store32_be((DST), (W))
+static inline void
+store32_be(uint8_t dst[4], uint32_t w)
+{
+#ifdef NATIVE_BIG_ENDIAN
+    memcpy(dst, &w, sizeof w);
+#else
+    dst[3] = (uint8_t) w; w >>= 8;
+    dst[2] = (uint8_t) w; w >>= 8;
+    dst[1] = (uint8_t) w; w >>= 8;
+    dst[0] = (uint8_t) w;
+#endif
+}
+
+#define XOR_BUF(OUT, IN, N) xor_buf((OUT), (IN), (N))
+static inline void
+xor_buf(unsigned char *out, const unsigned char *in, size_t n)
+{
+    size_t i;
+
+    for (i = 0; i < n; i++) {
+        out[i] ^= in[i];
+    }
+}
+
+#ifndef __GNUC__
+# ifdef __attribute__
+#  undef __attribute__
+# endif
+# define __attribute__(a)
+#endif
+
+#ifndef CRYPTO_ALIGN
+# if defined(__INTEL_COMPILER) || defined(_MSC_VER)
+#  define CRYPTO_ALIGN(x) __declspec(align(x))
+# else
+#  define CRYPTO_ALIGN(x) __attribute__ ((aligned(x)))
+# endif
+#endif
+
+#if defined(_MSC_VER) && \
+    (defined(_M_X64) || defined(_M_AMD64) || defined(_M_IX86))
+
+# include <intrin.h>
+
+# define HAVE_INTRIN_H    1
+# define HAVE_MMINTRIN_H  1
+# define HAVE_EMMINTRIN_H 1
+# define HAVE_PMMINTRIN_H 1
+# define HAVE_TMMINTRIN_H 1
+# define HAVE_SMMINTRIN_H 1
+# define HAVE_AVXINTRIN_H 1
+# if _MSC_VER >= 1600
+#  define HAVE_WMMINTRIN_H 1
+# endif
+# if _MSC_VER >= 1700 && defined(_M_X64)
+#  define HAVE_AVX2INTRIN_H 1
+# endif
+#elif defined(HAVE_INTRIN_H)
+# include <intrin.h>
+#endif
+
+#ifdef HAVE_LIBCTGRIND
+extern void ct_poison  (const void *, size_t);
+extern void ct_unpoison(const void *, size_t);
+# define POISON(X, L)   ct_poison((X), (L))
+# define UNPOISON(X, L) ct_unpoison((X), (L))
+#else
+# define POISON(X, L)   (void) 0
+# define UNPOISON(X, L) (void) 0
+#endif
+
+#endif
diff --git a/libs/libsodium/src/include/sodium/private/ed25519_ref10.h b/libs/libsodium/src/include/sodium/private/ed25519_ref10.h
new file mode 100644
index 0000000000..42fcd98139
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/ed25519_ref10.h
@@ -0,0 +1,125 @@
+#ifndef ed25519_ref10_H
+#define ed25519_ref10_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+/*
+ fe means field element.
+ Here the field is \Z/(2^255-19).
+ */
+
+#ifdef HAVE_TI_MODE
+typedef uint64_t fe25519[5];
+#else
+typedef int32_t fe25519[10];
+#endif
+
+void fe25519_invert(fe25519 out, const fe25519 z);
+void fe25519_frombytes(fe25519 h, const unsigned char *s);
+void fe25519_tobytes(unsigned char *s, const fe25519 h);
+
+#ifdef HAVE_TI_MODE
+# include "ed25519_ref10_fe_51.h"
+#else
+# include "ed25519_ref10_fe_25_5.h"
+#endif
+
+
+/*
+ ge means group element.
+
+ Here the group is the set of pairs (x,y) of field elements
+ satisfying -x^2 + y^2 = 1 + d x^2y^2
+ where d = -121665/121666.
+
+ Representations:
+ ge25519_p2 (projective): (X:Y:Z) satisfying x=X/Z, y=Y/Z
+ ge25519_p3 (extended): (X:Y:Z:T) satisfying x=X/Z, y=Y/Z, XY=ZT
+ ge25519_p1p1 (completed): ((X:Z),(Y:T)) satisfying x=X/Z, y=Y/T
+ ge25519_precomp (Duif): (y+x,y-x,2dxy)
+ */
+
+typedef struct {
+    fe25519 X;
+    fe25519 Y;
+    fe25519 Z;
+} ge25519_p2;
+
+typedef struct {
+    fe25519 X;
+    fe25519 Y;
+    fe25519 Z;
+    fe25519 T;
+} ge25519_p3;
+
+typedef struct {
+    fe25519 X;
+    fe25519 Y;
+    fe25519 Z;
+    fe25519 T;
+} ge25519_p1p1;
+
+typedef struct {
+    fe25519 yplusx;
+    fe25519 yminusx;
+    fe25519 xy2d;
+} ge25519_precomp;
+
+typedef struct {
+    fe25519 YplusX;
+    fe25519 YminusX;
+    fe25519 Z;
+    fe25519 T2d;
+} ge25519_cached;
+
+void ge25519_tobytes(unsigned char *s, const ge25519_p2 *h);
+
+void ge25519_p3_tobytes(unsigned char *s, const ge25519_p3 *h);
+
+int ge25519_frombytes(ge25519_p3 *h, const unsigned char *s);
+
+int ge25519_frombytes_negate_vartime(ge25519_p3 *h, const unsigned char *s);
+
+void ge25519_p3_to_cached(ge25519_cached *r, const ge25519_p3 *p);
+
+void ge25519_p1p1_to_p2(ge25519_p2 *r, const ge25519_p1p1 *p);
+
+void ge25519_p1p1_to_p3(ge25519_p3 *r, const ge25519_p1p1 *p);
+
+void ge25519_add(ge25519_p1p1 *r, const ge25519_p3 *p, const ge25519_cached *q);
+
+void ge25519_sub(ge25519_p1p1 *r, const ge25519_p3 *p, const ge25519_cached *q);
+
+void ge25519_scalarmult_base(ge25519_p3 *h, const unsigned char *a);
+
+void ge25519_double_scalarmult_vartime(ge25519_p2 *r, const unsigned char *a,
+                                       const ge25519_p3 *A,
+                                       const unsigned char *b);
+
+void ge25519_scalarmult(ge25519_p3 *h, const unsigned char *a,
+                        const ge25519_p3 *p);
+
+int ge25519_is_canonical(const unsigned char *s);
+
+int ge25519_is_on_curve(const ge25519_p3 *p);
+
+int ge25519_is_on_main_subgroup(const ge25519_p3 *p);
+
+int ge25519_has_small_order(const unsigned char s[32]);
+
+void ge25519_from_uniform(unsigned char s[32], const unsigned char r[32]);
+
+/*
+ The set of scalars is \Z/l
+ where l = 2^252 + 27742317777372353535851937790883648493.
+ */
+
+void sc25519_reduce(unsigned char *s);
+
+void sc25519_muladd(unsigned char *s, const unsigned char *a,
+                    const unsigned char *b, const unsigned char *c);
+
+int sc25519_is_canonical(const unsigned char *s);
+
+#endif
diff --git a/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_25_5.h b/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_25_5.h
new file mode 100644
index 0000000000..5a0b6148b7
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_25_5.h
@@ -0,0 +1,1050 @@
+#include <string.h>
+
+#include "private/common.h"
+#include "utils.h"
+
+/*
+ h = 0
+ */
+
+static inline void
+fe25519_0(fe25519 h)
+{
+    memset(&h[0], 0, 10 * sizeof h[0]);
+}
+
+/*
+ h = 1
+ */
+
+static inline void
+fe25519_1(fe25519 h)
+{
+    h[0] = 1;
+    h[1] = 0;
+    memset(&h[2], 0, 8 * sizeof h[0]);
+}
+
+/*
+ h = f + g
+ Can overlap h with f or g.
+ *
+ Preconditions:
+ |f| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ |g| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+ */
+
+static inline void
+fe25519_add(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    int32_t h0 = f[0] + g[0];
+    int32_t h1 = f[1] + g[1];
+    int32_t h2 = f[2] + g[2];
+    int32_t h3 = f[3] + g[3];
+    int32_t h4 = f[4] + g[4];
+    int32_t h5 = f[5] + g[5];
+    int32_t h6 = f[6] + g[6];
+    int32_t h7 = f[7] + g[7];
+    int32_t h8 = f[8] + g[8];
+    int32_t h9 = f[9] + g[9];
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+    h[5] = h5;
+    h[6] = h6;
+    h[7] = h7;
+    h[8] = h8;
+    h[9] = h9;
+}
+
+/*
+ h = f - g
+ Can overlap h with f or g.
+ *
+ Preconditions:
+ |f| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ |g| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+ */
+
+static void
+fe25519_sub(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    int32_t h0 = f[0] - g[0];
+    int32_t h1 = f[1] - g[1];
+    int32_t h2 = f[2] - g[2];
+    int32_t h3 = f[3] - g[3];
+    int32_t h4 = f[4] - g[4];
+    int32_t h5 = f[5] - g[5];
+    int32_t h6 = f[6] - g[6];
+    int32_t h7 = f[7] - g[7];
+    int32_t h8 = f[8] - g[8];
+    int32_t h9 = f[9] - g[9];
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+    h[5] = h5;
+    h[6] = h6;
+    h[7] = h7;
+    h[8] = h8;
+    h[9] = h9;
+}
+
+/*
+ h = -f
+ *
+ Preconditions:
+ |f| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
+ */
+
+static inline void
+fe25519_neg(fe25519 h, const fe25519 f)
+{
+    int32_t h0 = -f[0];
+    int32_t h1 = -f[1];
+    int32_t h2 = -f[2];
+    int32_t h3 = -f[3];
+    int32_t h4 = -f[4];
+    int32_t h5 = -f[5];
+    int32_t h6 = -f[6];
+    int32_t h7 = -f[7];
+    int32_t h8 = -f[8];
+    int32_t h9 = -f[9];
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+    h[5] = h5;
+    h[6] = h6;
+    h[7] = h7;
+    h[8] = h8;
+    h[9] = h9;
+}
+
+/*
+ Replace (f,g) with (g,g) if b == 1;
+ replace (f,g) with (f,g) if b == 0.
+ *
+ Preconditions: b in {0,1}.
+ */
+
+static void
+fe25519_cmov(fe25519 f, const fe25519 g, unsigned int b)
+{
+    const uint32_t mask = (uint32_t) (-(int32_t) b);
+
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    int32_t x0 = f0 ^ g[0];
+    int32_t x1 = f1 ^ g[1];
+    int32_t x2 = f2 ^ g[2];
+    int32_t x3 = f3 ^ g[3];
+    int32_t x4 = f4 ^ g[4];
+    int32_t x5 = f5 ^ g[5];
+    int32_t x6 = f6 ^ g[6];
+    int32_t x7 = f7 ^ g[7];
+    int32_t x8 = f8 ^ g[8];
+    int32_t x9 = f9 ^ g[9];
+
+    x0 &= mask;
+    x1 &= mask;
+    x2 &= mask;
+    x3 &= mask;
+    x4 &= mask;
+    x5 &= mask;
+    x6 &= mask;
+    x7 &= mask;
+    x8 &= mask;
+    x9 &= mask;
+
+    f[0] = f0 ^ x0;
+    f[1] = f1 ^ x1;
+    f[2] = f2 ^ x2;
+    f[3] = f3 ^ x3;
+    f[4] = f4 ^ x4;
+    f[5] = f5 ^ x5;
+    f[6] = f6 ^ x6;
+    f[7] = f7 ^ x7;
+    f[8] = f8 ^ x8;
+    f[9] = f9 ^ x9;
+}
+
+static void
+fe25519_cswap(fe25519 f, fe25519 g, unsigned int b)
+{
+    const uint32_t mask = (uint32_t) (-(int64_t) b);
+
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    int32_t g0 = g[0];
+    int32_t g1 = g[1];
+    int32_t g2 = g[2];
+    int32_t g3 = g[3];
+    int32_t g4 = g[4];
+    int32_t g5 = g[5];
+    int32_t g6 = g[6];
+    int32_t g7 = g[7];
+    int32_t g8 = g[8];
+    int32_t g9 = g[9];
+
+    int32_t x0 = f0 ^ g0;
+    int32_t x1 = f1 ^ g1;
+    int32_t x2 = f2 ^ g2;
+    int32_t x3 = f3 ^ g3;
+    int32_t x4 = f4 ^ g4;
+    int32_t x5 = f5 ^ g5;
+    int32_t x6 = f6 ^ g6;
+    int32_t x7 = f7 ^ g7;
+    int32_t x8 = f8 ^ g8;
+    int32_t x9 = f9 ^ g9;
+
+    x0 &= mask;
+    x1 &= mask;
+    x2 &= mask;
+    x3 &= mask;
+    x4 &= mask;
+    x5 &= mask;
+    x6 &= mask;
+    x7 &= mask;
+    x8 &= mask;
+    x9 &= mask;
+
+    f[0] = f0 ^ x0;
+    f[1] = f1 ^ x1;
+    f[2] = f2 ^ x2;
+    f[3] = f3 ^ x3;
+    f[4] = f4 ^ x4;
+    f[5] = f5 ^ x5;
+    f[6] = f6 ^ x6;
+    f[7] = f7 ^ x7;
+    f[8] = f8 ^ x8;
+    f[9] = f9 ^ x9;
+
+    g[0] = g0 ^ x0;
+    g[1] = g1 ^ x1;
+    g[2] = g2 ^ x2;
+    g[3] = g3 ^ x3;
+    g[4] = g4 ^ x4;
+    g[5] = g5 ^ x5;
+    g[6] = g6 ^ x6;
+    g[7] = g7 ^ x7;
+    g[8] = g8 ^ x8;
+    g[9] = g9 ^ x9;
+}
+
+/*
+ h = f
+ */
+
+static inline void
+fe25519_copy(fe25519 h, const fe25519 f)
+{
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    h[0] = f0;
+    h[1] = f1;
+    h[2] = f2;
+    h[3] = f3;
+    h[4] = f4;
+    h[5] = f5;
+    h[6] = f6;
+    h[7] = f7;
+    h[8] = f8;
+    h[9] = f9;
+}
+
+/*
+ return 1 if f is in {1,3,5,...,q-2}
+ return 0 if f is in {0,2,4,...,q-1}
+
+ Preconditions:
+ |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+ */
+
+static inline int
+fe25519_isnegative(const fe25519 f)
+{
+    unsigned char s[32];
+
+    fe25519_tobytes(s, f);
+
+    return s[0] & 1;
+}
+
+/*
+ return 1 if f == 0
+ return 0 if f != 0
+
+ Preconditions:
+ |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
+ */
+
+static inline int
+fe25519_iszero(const fe25519 f)
+{
+    unsigned char s[32];
+
+    fe25519_tobytes(s, f);
+
+    return sodium_is_zero(s, 32);
+}
+
+/*
+ h = f * g
+ Can overlap h with f or g.
+ *
+ Preconditions:
+ |f| bounded by 1.65*2^26,1.65*2^25,1.65*2^26,1.65*2^25,etc.
+ |g| bounded by 1.65*2^26,1.65*2^25,1.65*2^26,1.65*2^25,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.01*2^25,1.01*2^24,1.01*2^25,1.01*2^24,etc.
+ */
+
+/*
+ Notes on implementation strategy:
+ *
+ Using schoolbook multiplication.
+ Karatsuba would save a little in some cost models.
+ *
+ Most multiplications by 2 and 19 are 32-bit precomputations;
+ cheaper than 64-bit postcomputations.
+ *
+ There is one remaining multiplication by 19 in the carry chain;
+ one *19 precomputation can be merged into this,
+ but the resulting data flow is considerably less clean.
+ *
+ There are 12 carries below.
+ 10 of them are 2-way parallelizable and vectorizable.
+ Can get away with 11 carries, but then data flow is much deeper.
+ *
+ With tighter constraints on inputs can squeeze carries into int32.
+ */
+
+static void
+fe25519_mul(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    int32_t g0 = g[0];
+    int32_t g1 = g[1];
+    int32_t g2 = g[2];
+    int32_t g3 = g[3];
+    int32_t g4 = g[4];
+    int32_t g5 = g[5];
+    int32_t g6 = g[6];
+    int32_t g7 = g[7];
+    int32_t g8 = g[8];
+    int32_t g9 = g[9];
+
+    int32_t g1_19 = 19 * g1; /* 1.959375*2^29 */
+    int32_t g2_19 = 19 * g2; /* 1.959375*2^30; still ok */
+    int32_t g3_19 = 19 * g3;
+    int32_t g4_19 = 19 * g4;
+    int32_t g5_19 = 19 * g5;
+    int32_t g6_19 = 19 * g6;
+    int32_t g7_19 = 19 * g7;
+    int32_t g8_19 = 19 * g8;
+    int32_t g9_19 = 19 * g9;
+    int32_t f1_2  = 2 * f1;
+    int32_t f3_2  = 2 * f3;
+    int32_t f5_2  = 2 * f5;
+    int32_t f7_2  = 2 * f7;
+    int32_t f9_2  = 2 * f9;
+
+    int64_t f0g0    = f0 * (int64_t) g0;
+    int64_t f0g1    = f0 * (int64_t) g1;
+    int64_t f0g2    = f0 * (int64_t) g2;
+    int64_t f0g3    = f0 * (int64_t) g3;
+    int64_t f0g4    = f0 * (int64_t) g4;
+    int64_t f0g5    = f0 * (int64_t) g5;
+    int64_t f0g6    = f0 * (int64_t) g6;
+    int64_t f0g7    = f0 * (int64_t) g7;
+    int64_t f0g8    = f0 * (int64_t) g8;
+    int64_t f0g9    = f0 * (int64_t) g9;
+    int64_t f1g0    = f1 * (int64_t) g0;
+    int64_t f1g1_2  = f1_2 * (int64_t) g1;
+    int64_t f1g2    = f1 * (int64_t) g2;
+    int64_t f1g3_2  = f1_2 * (int64_t) g3;
+    int64_t f1g4    = f1 * (int64_t) g4;
+    int64_t f1g5_2  = f1_2 * (int64_t) g5;
+    int64_t f1g6    = f1 * (int64_t) g6;
+    int64_t f1g7_2  = f1_2 * (int64_t) g7;
+    int64_t f1g8    = f1 * (int64_t) g8;
+    int64_t f1g9_38 = f1_2 * (int64_t) g9_19;
+    int64_t f2g0    = f2 * (int64_t) g0;
+    int64_t f2g1    = f2 * (int64_t) g1;
+    int64_t f2g2    = f2 * (int64_t) g2;
+    int64_t f2g3    = f2 * (int64_t) g3;
+    int64_t f2g4    = f2 * (int64_t) g4;
+    int64_t f2g5    = f2 * (int64_t) g5;
+    int64_t f2g6    = f2 * (int64_t) g6;
+    int64_t f2g7    = f2 * (int64_t) g7;
+    int64_t f2g8_19 = f2 * (int64_t) g8_19;
+    int64_t f2g9_19 = f2 * (int64_t) g9_19;
+    int64_t f3g0    = f3 * (int64_t) g0;
+    int64_t f3g1_2  = f3_2 * (int64_t) g1;
+    int64_t f3g2    = f3 * (int64_t) g2;
+    int64_t f3g3_2  = f3_2 * (int64_t) g3;
+    int64_t f3g4    = f3 * (int64_t) g4;
+    int64_t f3g5_2  = f3_2 * (int64_t) g5;
+    int64_t f3g6    = f3 * (int64_t) g6;
+    int64_t f3g7_38 = f3_2 * (int64_t) g7_19;
+    int64_t f3g8_19 = f3 * (int64_t) g8_19;
+    int64_t f3g9_38 = f3_2 * (int64_t) g9_19;
+    int64_t f4g0    = f4 * (int64_t) g0;
+    int64_t f4g1    = f4 * (int64_t) g1;
+    int64_t f4g2    = f4 * (int64_t) g2;
+    int64_t f4g3    = f4 * (int64_t) g3;
+    int64_t f4g4    = f4 * (int64_t) g4;
+    int64_t f4g5    = f4 * (int64_t) g5;
+    int64_t f4g6_19 = f4 * (int64_t) g6_19;
+    int64_t f4g7_19 = f4 * (int64_t) g7_19;
+    int64_t f4g8_19 = f4 * (int64_t) g8_19;
+    int64_t f4g9_19 = f4 * (int64_t) g9_19;
+    int64_t f5g0    = f5 * (int64_t) g0;
+    int64_t f5g1_2  = f5_2 * (int64_t) g1;
+    int64_t f5g2    = f5 * (int64_t) g2;
+    int64_t f5g3_2  = f5_2 * (int64_t) g3;
+    int64_t f5g4    = f5 * (int64_t) g4;
+    int64_t f5g5_38 = f5_2 * (int64_t) g5_19;
+    int64_t f5g6_19 = f5 * (int64_t) g6_19;
+    int64_t f5g7_38 = f5_2 * (int64_t) g7_19;
+    int64_t f5g8_19 = f5 * (int64_t) g8_19;
+    int64_t f5g9_38 = f5_2 * (int64_t) g9_19;
+    int64_t f6g0    = f6 * (int64_t) g0;
+    int64_t f6g1    = f6 * (int64_t) g1;
+    int64_t f6g2    = f6 * (int64_t) g2;
+    int64_t f6g3    = f6 * (int64_t) g3;
+    int64_t f6g4_19 = f6 * (int64_t) g4_19;
+    int64_t f6g5_19 = f6 * (int64_t) g5_19;
+    int64_t f6g6_19 = f6 * (int64_t) g6_19;
+    int64_t f6g7_19 = f6 * (int64_t) g7_19;
+    int64_t f6g8_19 = f6 * (int64_t) g8_19;
+    int64_t f6g9_19 = f6 * (int64_t) g9_19;
+    int64_t f7g0    = f7 * (int64_t) g0;
+    int64_t f7g1_2  = f7_2 * (int64_t) g1;
+    int64_t f7g2    = f7 * (int64_t) g2;
+    int64_t f7g3_38 = f7_2 * (int64_t) g3_19;
+    int64_t f7g4_19 = f7 * (int64_t) g4_19;
+    int64_t f7g5_38 = f7_2 * (int64_t) g5_19;
+    int64_t f7g6_19 = f7 * (int64_t) g6_19;
+    int64_t f7g7_38 = f7_2 * (int64_t) g7_19;
+    int64_t f7g8_19 = f7 * (int64_t) g8_19;
+    int64_t f7g9_38 = f7_2 * (int64_t) g9_19;
+    int64_t f8g0    = f8 * (int64_t) g0;
+    int64_t f8g1    = f8 * (int64_t) g1;
+    int64_t f8g2_19 = f8 * (int64_t) g2_19;
+    int64_t f8g3_19 = f8 * (int64_t) g3_19;
+    int64_t f8g4_19 = f8 * (int64_t) g4_19;
+    int64_t f8g5_19 = f8 * (int64_t) g5_19;
+    int64_t f8g6_19 = f8 * (int64_t) g6_19;
+    int64_t f8g7_19 = f8 * (int64_t) g7_19;
+    int64_t f8g8_19 = f8 * (int64_t) g8_19;
+    int64_t f8g9_19 = f8 * (int64_t) g9_19;
+    int64_t f9g0    = f9 * (int64_t) g0;
+    int64_t f9g1_38 = f9_2 * (int64_t) g1_19;
+    int64_t f9g2_19 = f9 * (int64_t) g2_19;
+    int64_t f9g3_38 = f9_2 * (int64_t) g3_19;
+    int64_t f9g4_19 = f9 * (int64_t) g4_19;
+    int64_t f9g5_38 = f9_2 * (int64_t) g5_19;
+    int64_t f9g6_19 = f9 * (int64_t) g6_19;
+    int64_t f9g7_38 = f9_2 * (int64_t) g7_19;
+    int64_t f9g8_19 = f9 * (int64_t) g8_19;
+    int64_t f9g9_38 = f9_2 * (int64_t) g9_19;
+
+    int64_t h0 = f0g0 + f1g9_38 + f2g8_19 + f3g7_38 + f4g6_19 + f5g5_38 +
+                 f6g4_19 + f7g3_38 + f8g2_19 + f9g1_38;
+    int64_t h1 = f0g1 + f1g0 + f2g9_19 + f3g8_19 + f4g7_19 + f5g6_19 + f6g5_19 +
+                 f7g4_19 + f8g3_19 + f9g2_19;
+    int64_t h2 = f0g2 + f1g1_2 + f2g0 + f3g9_38 + f4g8_19 + f5g7_38 + f6g6_19 +
+                 f7g5_38 + f8g4_19 + f9g3_38;
+    int64_t h3 = f0g3 + f1g2 + f2g1 + f3g0 + f4g9_19 + f5g8_19 + f6g7_19 +
+                 f7g6_19 + f8g5_19 + f9g4_19;
+    int64_t h4 = f0g4 + f1g3_2 + f2g2 + f3g1_2 + f4g0 + f5g9_38 + f6g8_19 +
+                 f7g7_38 + f8g6_19 + f9g5_38;
+    int64_t h5 = f0g5 + f1g4 + f2g3 + f3g2 + f4g1 + f5g0 + f6g9_19 + f7g8_19 +
+                 f8g7_19 + f9g6_19;
+    int64_t h6 = f0g6 + f1g5_2 + f2g4 + f3g3_2 + f4g2 + f5g1_2 + f6g0 +
+                 f7g9_38 + f8g8_19 + f9g7_38;
+    int64_t h7 = f0g7 + f1g6 + f2g5 + f3g4 + f4g3 + f5g2 + f6g1 + f7g0 +
+                 f8g9_19 + f9g8_19;
+    int64_t h8 = f0g8 + f1g7_2 + f2g6 + f3g5_2 + f4g4 + f5g3_2 + f6g2 + f7g1_2 +
+                 f8g0 + f9g9_38;
+    int64_t h9 =
+        f0g9 + f1g8 + f2g7 + f3g6 + f4g5 + f5g4 + f6g3 + f7g2 + f8g1 + f9g0;
+
+    int64_t carry0;
+    int64_t carry1;
+    int64_t carry2;
+    int64_t carry3;
+    int64_t carry4;
+    int64_t carry5;
+    int64_t carry6;
+    int64_t carry7;
+    int64_t carry8;
+    int64_t carry9;
+
+    /*
+     |h0| <= (1.65*1.65*2^52*(1+19+19+19+19)+1.65*1.65*2^50*(38+38+38+38+38))
+     i.e. |h0| <= 1.4*2^60; narrower ranges for h2, h4, h6, h8
+     |h1| <= (1.65*1.65*2^51*(1+1+19+19+19+19+19+19+19+19))
+     i.e. |h1| <= 1.7*2^59; narrower ranges for h3, h5, h7, h9
+     */
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+    /* |h0| <= 2^25 */
+    /* |h4| <= 2^25 */
+    /* |h1| <= 1.71*2^59 */
+    /* |h5| <= 1.71*2^59 */
+
+    carry1 = (h1 + (int64_t)(1L << 24)) >> 25;
+    h2 += carry1;
+    h1 -= carry1 * ((uint64_t) 1L << 25);
+    carry5 = (h5 + (int64_t)(1L << 24)) >> 25;
+    h6 += carry5;
+    h5 -= carry5 * ((uint64_t) 1L << 25);
+    /* |h1| <= 2^24; from now on fits into int32 */
+    /* |h5| <= 2^24; from now on fits into int32 */
+    /* |h2| <= 1.41*2^60 */
+    /* |h6| <= 1.41*2^60 */
+
+    carry2 = (h2 + (int64_t)(1L << 25)) >> 26;
+    h3 += carry2;
+    h2 -= carry2 * ((uint64_t) 1L << 26);
+    carry6 = (h6 + (int64_t)(1L << 25)) >> 26;
+    h7 += carry6;
+    h6 -= carry6 * ((uint64_t) 1L << 26);
+    /* |h2| <= 2^25; from now on fits into int32 unchanged */
+    /* |h6| <= 2^25; from now on fits into int32 unchanged */
+    /* |h3| <= 1.71*2^59 */
+    /* |h7| <= 1.71*2^59 */
+
+    carry3 = (h3 + (int64_t)(1L << 24)) >> 25;
+    h4 += carry3;
+    h3 -= carry3 * ((uint64_t) 1L << 25);
+    carry7 = (h7 + (int64_t)(1L << 24)) >> 25;
+    h8 += carry7;
+    h7 -= carry7 * ((uint64_t) 1L << 25);
+    /* |h3| <= 2^24; from now on fits into int32 unchanged */
+    /* |h7| <= 2^24; from now on fits into int32 unchanged */
+    /* |h4| <= 1.72*2^34 */
+    /* |h8| <= 1.41*2^60 */
+
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+    carry8 = (h8 + (int64_t)(1L << 25)) >> 26;
+    h9 += carry8;
+    h8 -= carry8 * ((uint64_t) 1L << 26);
+    /* |h4| <= 2^25; from now on fits into int32 unchanged */
+    /* |h8| <= 2^25; from now on fits into int32 unchanged */
+    /* |h5| <= 1.01*2^24 */
+    /* |h9| <= 1.71*2^59 */
+
+    carry9 = (h9 + (int64_t)(1L << 24)) >> 25;
+    h0 += carry9 * 19;
+    h9 -= carry9 * ((uint64_t) 1L << 25);
+    /* |h9| <= 2^24; from now on fits into int32 unchanged */
+    /* |h0| <= 1.1*2^39 */
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+    /* |h0| <= 2^25; from now on fits into int32 unchanged */
+    /* |h1| <= 1.01*2^24 */
+
+    h[0] = (int32_t) h0;
+    h[1] = (int32_t) h1;
+    h[2] = (int32_t) h2;
+    h[3] = (int32_t) h3;
+    h[4] = (int32_t) h4;
+    h[5] = (int32_t) h5;
+    h[6] = (int32_t) h6;
+    h[7] = (int32_t) h7;
+    h[8] = (int32_t) h8;
+    h[9] = (int32_t) h9;
+}
+
+/*
+ h = f * f
+ Can overlap h with f.
+ *
+ Preconditions:
+ |f| bounded by 1.65*2^26,1.65*2^25,1.65*2^26,1.65*2^25,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.01*2^25,1.01*2^24,1.01*2^25,1.01*2^24,etc.
+ */
+
+static void
+fe25519_sq(fe25519 h, const fe25519 f)
+{
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    int32_t f0_2  = 2 * f0;
+    int32_t f1_2  = 2 * f1;
+    int32_t f2_2  = 2 * f2;
+    int32_t f3_2  = 2 * f3;
+    int32_t f4_2  = 2 * f4;
+    int32_t f5_2  = 2 * f5;
+    int32_t f6_2  = 2 * f6;
+    int32_t f7_2  = 2 * f7;
+    int32_t f5_38 = 38 * f5; /* 1.959375*2^30 */
+    int32_t f6_19 = 19 * f6; /* 1.959375*2^30 */
+    int32_t f7_38 = 38 * f7; /* 1.959375*2^30 */
+    int32_t f8_19 = 19 * f8; /* 1.959375*2^30 */
+    int32_t f9_38 = 38 * f9; /* 1.959375*2^30 */
+
+    int64_t f0f0    = f0 * (int64_t) f0;
+    int64_t f0f1_2  = f0_2 * (int64_t) f1;
+    int64_t f0f2_2  = f0_2 * (int64_t) f2;
+    int64_t f0f3_2  = f0_2 * (int64_t) f3;
+    int64_t f0f4_2  = f0_2 * (int64_t) f4;
+    int64_t f0f5_2  = f0_2 * (int64_t) f5;
+    int64_t f0f6_2  = f0_2 * (int64_t) f6;
+    int64_t f0f7_2  = f0_2 * (int64_t) f7;
+    int64_t f0f8_2  = f0_2 * (int64_t) f8;
+    int64_t f0f9_2  = f0_2 * (int64_t) f9;
+    int64_t f1f1_2  = f1_2 * (int64_t) f1;
+    int64_t f1f2_2  = f1_2 * (int64_t) f2;
+    int64_t f1f3_4  = f1_2 * (int64_t) f3_2;
+    int64_t f1f4_2  = f1_2 * (int64_t) f4;
+    int64_t f1f5_4  = f1_2 * (int64_t) f5_2;
+    int64_t f1f6_2  = f1_2 * (int64_t) f6;
+    int64_t f1f7_4  = f1_2 * (int64_t) f7_2;
+    int64_t f1f8_2  = f1_2 * (int64_t) f8;
+    int64_t f1f9_76 = f1_2 * (int64_t) f9_38;
+    int64_t f2f2    = f2 * (int64_t) f2;
+    int64_t f2f3_2  = f2_2 * (int64_t) f3;
+    int64_t f2f4_2  = f2_2 * (int64_t) f4;
+    int64_t f2f5_2  = f2_2 * (int64_t) f5;
+    int64_t f2f6_2  = f2_2 * (int64_t) f6;
+    int64_t f2f7_2  = f2_2 * (int64_t) f7;
+    int64_t f2f8_38 = f2_2 * (int64_t) f8_19;
+    int64_t f2f9_38 = f2 * (int64_t) f9_38;
+    int64_t f3f3_2  = f3_2 * (int64_t) f3;
+    int64_t f3f4_2  = f3_2 * (int64_t) f4;
+    int64_t f3f5_4  = f3_2 * (int64_t) f5_2;
+    int64_t f3f6_2  = f3_2 * (int64_t) f6;
+    int64_t f3f7_76 = f3_2 * (int64_t) f7_38;
+    int64_t f3f8_38 = f3_2 * (int64_t) f8_19;
+    int64_t f3f9_76 = f3_2 * (int64_t) f9_38;
+    int64_t f4f4    = f4 * (int64_t) f4;
+    int64_t f4f5_2  = f4_2 * (int64_t) f5;
+    int64_t f4f6_38 = f4_2 * (int64_t) f6_19;
+    int64_t f4f7_38 = f4 * (int64_t) f7_38;
+    int64_t f4f8_38 = f4_2 * (int64_t) f8_19;
+    int64_t f4f9_38 = f4 * (int64_t) f9_38;
+    int64_t f5f5_38 = f5 * (int64_t) f5_38;
+    int64_t f5f6_38 = f5_2 * (int64_t) f6_19;
+    int64_t f5f7_76 = f5_2 * (int64_t) f7_38;
+    int64_t f5f8_38 = f5_2 * (int64_t) f8_19;
+    int64_t f5f9_76 = f5_2 * (int64_t) f9_38;
+    int64_t f6f6_19 = f6 * (int64_t) f6_19;
+    int64_t f6f7_38 = f6 * (int64_t) f7_38;
+    int64_t f6f8_38 = f6_2 * (int64_t) f8_19;
+    int64_t f6f9_38 = f6 * (int64_t) f9_38;
+    int64_t f7f7_38 = f7 * (int64_t) f7_38;
+    int64_t f7f8_38 = f7_2 * (int64_t) f8_19;
+    int64_t f7f9_76 = f7_2 * (int64_t) f9_38;
+    int64_t f8f8_19 = f8 * (int64_t) f8_19;
+    int64_t f8f9_38 = f8 * (int64_t) f9_38;
+    int64_t f9f9_38 = f9 * (int64_t) f9_38;
+
+    int64_t h0 = f0f0 + f1f9_76 + f2f8_38 + f3f7_76 + f4f6_38 + f5f5_38;
+    int64_t h1 = f0f1_2 + f2f9_38 + f3f8_38 + f4f7_38 + f5f6_38;
+    int64_t h2 = f0f2_2 + f1f1_2 + f3f9_76 + f4f8_38 + f5f7_76 + f6f6_19;
+    int64_t h3 = f0f3_2 + f1f2_2 + f4f9_38 + f5f8_38 + f6f7_38;
+    int64_t h4 = f0f4_2 + f1f3_4 + f2f2 + f5f9_76 + f6f8_38 + f7f7_38;
+    int64_t h5 = f0f5_2 + f1f4_2 + f2f3_2 + f6f9_38 + f7f8_38;
+    int64_t h6 = f0f6_2 + f1f5_4 + f2f4_2 + f3f3_2 + f7f9_76 + f8f8_19;
+    int64_t h7 = f0f7_2 + f1f6_2 + f2f5_2 + f3f4_2 + f8f9_38;
+    int64_t h8 = f0f8_2 + f1f7_4 + f2f6_2 + f3f5_4 + f4f4 + f9f9_38;
+    int64_t h9 = f0f9_2 + f1f8_2 + f2f7_2 + f3f6_2 + f4f5_2;
+
+    int64_t carry0;
+    int64_t carry1;
+    int64_t carry2;
+    int64_t carry3;
+    int64_t carry4;
+    int64_t carry5;
+    int64_t carry6;
+    int64_t carry7;
+    int64_t carry8;
+    int64_t carry9;
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+
+    carry1 = (h1 + (int64_t)(1L << 24)) >> 25;
+    h2 += carry1;
+    h1 -= carry1 * ((uint64_t) 1L << 25);
+    carry5 = (h5 + (int64_t)(1L << 24)) >> 25;
+    h6 += carry5;
+    h5 -= carry5 * ((uint64_t) 1L << 25);
+
+    carry2 = (h2 + (int64_t)(1L << 25)) >> 26;
+    h3 += carry2;
+    h2 -= carry2 * ((uint64_t) 1L << 26);
+    carry6 = (h6 + (int64_t)(1L << 25)) >> 26;
+    h7 += carry6;
+    h6 -= carry6 * ((uint64_t) 1L << 26);
+
+    carry3 = (h3 + (int64_t)(1L << 24)) >> 25;
+    h4 += carry3;
+    h3 -= carry3 * ((uint64_t) 1L << 25);
+    carry7 = (h7 + (int64_t)(1L << 24)) >> 25;
+    h8 += carry7;
+    h7 -= carry7 * ((uint64_t) 1L << 25);
+
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+    carry8 = (h8 + (int64_t)(1L << 25)) >> 26;
+    h9 += carry8;
+    h8 -= carry8 * ((uint64_t) 1L << 26);
+
+    carry9 = (h9 + (int64_t)(1L << 24)) >> 25;
+    h0 += carry9 * 19;
+    h9 -= carry9 * ((uint64_t) 1L << 25);
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+
+    h[0] = (int32_t) h0;
+    h[1] = (int32_t) h1;
+    h[2] = (int32_t) h2;
+    h[3] = (int32_t) h3;
+    h[4] = (int32_t) h4;
+    h[5] = (int32_t) h5;
+    h[6] = (int32_t) h6;
+    h[7] = (int32_t) h7;
+    h[8] = (int32_t) h8;
+    h[9] = (int32_t) h9;
+}
+
+/*
+ h = 2 * f * f
+ Can overlap h with f.
+ *
+ Preconditions:
+ |f| bounded by 1.65*2^26,1.65*2^25,1.65*2^26,1.65*2^25,etc.
+ *
+ Postconditions:
+ |h| bounded by 1.01*2^25,1.01*2^24,1.01*2^25,1.01*2^24,etc.
+ */
+
+static void
+fe25519_sq2(fe25519 h, const fe25519 f)
+{
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+
+    int32_t f0_2  = 2 * f0;
+    int32_t f1_2  = 2 * f1;
+    int32_t f2_2  = 2 * f2;
+    int32_t f3_2  = 2 * f3;
+    int32_t f4_2  = 2 * f4;
+    int32_t f5_2  = 2 * f5;
+    int32_t f6_2  = 2 * f6;
+    int32_t f7_2  = 2 * f7;
+    int32_t f5_38 = 38 * f5; /* 1.959375*2^30 */
+    int32_t f6_19 = 19 * f6; /* 1.959375*2^30 */
+    int32_t f7_38 = 38 * f7; /* 1.959375*2^30 */
+    int32_t f8_19 = 19 * f8; /* 1.959375*2^30 */
+    int32_t f9_38 = 38 * f9; /* 1.959375*2^30 */
+
+    int64_t f0f0    = f0 * (int64_t) f0;
+    int64_t f0f1_2  = f0_2 * (int64_t) f1;
+    int64_t f0f2_2  = f0_2 * (int64_t) f2;
+    int64_t f0f3_2  = f0_2 * (int64_t) f3;
+    int64_t f0f4_2  = f0_2 * (int64_t) f4;
+    int64_t f0f5_2  = f0_2 * (int64_t) f5;
+    int64_t f0f6_2  = f0_2 * (int64_t) f6;
+    int64_t f0f7_2  = f0_2 * (int64_t) f7;
+    int64_t f0f8_2  = f0_2 * (int64_t) f8;
+    int64_t f0f9_2  = f0_2 * (int64_t) f9;
+    int64_t f1f1_2  = f1_2 * (int64_t) f1;
+    int64_t f1f2_2  = f1_2 * (int64_t) f2;
+    int64_t f1f3_4  = f1_2 * (int64_t) f3_2;
+    int64_t f1f4_2  = f1_2 * (int64_t) f4;
+    int64_t f1f5_4  = f1_2 * (int64_t) f5_2;
+    int64_t f1f6_2  = f1_2 * (int64_t) f6;
+    int64_t f1f7_4  = f1_2 * (int64_t) f7_2;
+    int64_t f1f8_2  = f1_2 * (int64_t) f8;
+    int64_t f1f9_76 = f1_2 * (int64_t) f9_38;
+    int64_t f2f2    = f2 * (int64_t) f2;
+    int64_t f2f3_2  = f2_2 * (int64_t) f3;
+    int64_t f2f4_2  = f2_2 * (int64_t) f4;
+    int64_t f2f5_2  = f2_2 * (int64_t) f5;
+    int64_t f2f6_2  = f2_2 * (int64_t) f6;
+    int64_t f2f7_2  = f2_2 * (int64_t) f7;
+    int64_t f2f8_38 = f2_2 * (int64_t) f8_19;
+    int64_t f2f9_38 = f2 * (int64_t) f9_38;
+    int64_t f3f3_2  = f3_2 * (int64_t) f3;
+    int64_t f3f4_2  = f3_2 * (int64_t) f4;
+    int64_t f3f5_4  = f3_2 * (int64_t) f5_2;
+    int64_t f3f6_2  = f3_2 * (int64_t) f6;
+    int64_t f3f7_76 = f3_2 * (int64_t) f7_38;
+    int64_t f3f8_38 = f3_2 * (int64_t) f8_19;
+    int64_t f3f9_76 = f3_2 * (int64_t) f9_38;
+    int64_t f4f4    = f4 * (int64_t) f4;
+    int64_t f4f5_2  = f4_2 * (int64_t) f5;
+    int64_t f4f6_38 = f4_2 * (int64_t) f6_19;
+    int64_t f4f7_38 = f4 * (int64_t) f7_38;
+    int64_t f4f8_38 = f4_2 * (int64_t) f8_19;
+    int64_t f4f9_38 = f4 * (int64_t) f9_38;
+    int64_t f5f5_38 = f5 * (int64_t) f5_38;
+    int64_t f5f6_38 = f5_2 * (int64_t) f6_19;
+    int64_t f5f7_76 = f5_2 * (int64_t) f7_38;
+    int64_t f5f8_38 = f5_2 * (int64_t) f8_19;
+    int64_t f5f9_76 = f5_2 * (int64_t) f9_38;
+    int64_t f6f6_19 = f6 * (int64_t) f6_19;
+    int64_t f6f7_38 = f6 * (int64_t) f7_38;
+    int64_t f6f8_38 = f6_2 * (int64_t) f8_19;
+    int64_t f6f9_38 = f6 * (int64_t) f9_38;
+    int64_t f7f7_38 = f7 * (int64_t) f7_38;
+    int64_t f7f8_38 = f7_2 * (int64_t) f8_19;
+    int64_t f7f9_76 = f7_2 * (int64_t) f9_38;
+    int64_t f8f8_19 = f8 * (int64_t) f8_19;
+    int64_t f8f9_38 = f8 * (int64_t) f9_38;
+    int64_t f9f9_38 = f9 * (int64_t) f9_38;
+
+    int64_t h0 = f0f0 + f1f9_76 + f2f8_38 + f3f7_76 + f4f6_38 + f5f5_38;
+    int64_t h1 = f0f1_2 + f2f9_38 + f3f8_38 + f4f7_38 + f5f6_38;
+    int64_t h2 = f0f2_2 + f1f1_2 + f3f9_76 + f4f8_38 + f5f7_76 + f6f6_19;
+    int64_t h3 = f0f3_2 + f1f2_2 + f4f9_38 + f5f8_38 + f6f7_38;
+    int64_t h4 = f0f4_2 + f1f3_4 + f2f2 + f5f9_76 + f6f8_38 + f7f7_38;
+    int64_t h5 = f0f5_2 + f1f4_2 + f2f3_2 + f6f9_38 + f7f8_38;
+    int64_t h6 = f0f6_2 + f1f5_4 + f2f4_2 + f3f3_2 + f7f9_76 + f8f8_19;
+    int64_t h7 = f0f7_2 + f1f6_2 + f2f5_2 + f3f4_2 + f8f9_38;
+    int64_t h8 = f0f8_2 + f1f7_4 + f2f6_2 + f3f5_4 + f4f4 + f9f9_38;
+    int64_t h9 = f0f9_2 + f1f8_2 + f2f7_2 + f3f6_2 + f4f5_2;
+
+    int64_t carry0;
+    int64_t carry1;
+    int64_t carry2;
+    int64_t carry3;
+    int64_t carry4;
+    int64_t carry5;
+    int64_t carry6;
+    int64_t carry7;
+    int64_t carry8;
+    int64_t carry9;
+
+    h0 += h0;
+    h1 += h1;
+    h2 += h2;
+    h3 += h3;
+    h4 += h4;
+    h5 += h5;
+    h6 += h6;
+    h7 += h7;
+    h8 += h8;
+    h9 += h9;
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+
+    carry1 = (h1 + (int64_t)(1L << 24)) >> 25;
+    h2 += carry1;
+    h1 -= carry1 * ((uint64_t) 1L << 25);
+    carry5 = (h5 + (int64_t)(1L << 24)) >> 25;
+    h6 += carry5;
+    h5 -= carry5 * ((uint64_t) 1L << 25);
+
+    carry2 = (h2 + (int64_t)(1L << 25)) >> 26;
+    h3 += carry2;
+    h2 -= carry2 * ((uint64_t) 1L << 26);
+    carry6 = (h6 + (int64_t)(1L << 25)) >> 26;
+    h7 += carry6;
+    h6 -= carry6 * ((uint64_t) 1L << 26);
+
+    carry3 = (h3 + (int64_t)(1L << 24)) >> 25;
+    h4 += carry3;
+    h3 -= carry3 * ((uint64_t) 1L << 25);
+    carry7 = (h7 + (int64_t)(1L << 24)) >> 25;
+    h8 += carry7;
+    h7 -= carry7 * ((uint64_t) 1L << 25);
+
+    carry4 = (h4 + (int64_t)(1L << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((uint64_t) 1L << 26);
+    carry8 = (h8 + (int64_t)(1L << 25)) >> 26;
+    h9 += carry8;
+    h8 -= carry8 * ((uint64_t) 1L << 26);
+
+    carry9 = (h9 + (int64_t)(1L << 24)) >> 25;
+    h0 += carry9 * 19;
+    h9 -= carry9 * ((uint64_t) 1L << 25);
+
+    carry0 = (h0 + (int64_t)(1L << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((uint64_t) 1L << 26);
+
+    h[0] = (int32_t) h0;
+    h[1] = (int32_t) h1;
+    h[2] = (int32_t) h2;
+    h[3] = (int32_t) h3;
+    h[4] = (int32_t) h4;
+    h[5] = (int32_t) h5;
+    h[6] = (int32_t) h6;
+    h[7] = (int32_t) h7;
+    h[8] = (int32_t) h8;
+    h[9] = (int32_t) h9;
+}
+
+static void
+fe25519_scalar_product(fe25519 h, const fe25519 f, uint32_t n)
+{
+    int64_t sn = (int64_t) n;
+    int32_t f0 = f[0];
+    int32_t f1 = f[1];
+    int32_t f2 = f[2];
+    int32_t f3 = f[3];
+    int32_t f4 = f[4];
+    int32_t f5 = f[5];
+    int32_t f6 = f[6];
+    int32_t f7 = f[7];
+    int32_t f8 = f[8];
+    int32_t f9 = f[9];
+    int64_t h0 = f0 * sn;
+    int64_t h1 = f1 * sn;
+    int64_t h2 = f2 * sn;
+    int64_t h3 = f3 * sn;
+    int64_t h4 = f4 * sn;
+    int64_t h5 = f5 * sn;
+    int64_t h6 = f6 * sn;
+    int64_t h7 = f7 * sn;
+    int64_t h8 = f8 * sn;
+    int64_t h9 = f9 * sn;
+    int64_t carry0, carry1, carry2, carry3, carry4, carry5, carry6, carry7,
+            carry8, carry9;
+
+    carry9 = (h9 + ((int64_t) 1 << 24)) >> 25;
+    h0 += carry9 * 19;
+    h9 -= carry9 * ((int64_t) 1 << 25);
+    carry1 = (h1 + ((int64_t) 1 << 24)) >> 25;
+    h2 += carry1;
+    h1 -= carry1 * ((int64_t) 1 << 25);
+    carry3 = (h3 + ((int64_t) 1 << 24)) >> 25;
+    h4 += carry3;
+    h3 -= carry3 * ((int64_t) 1 << 25);
+    carry5 = (h5 + ((int64_t) 1 << 24)) >> 25;
+    h6 += carry5;
+    h5 -= carry5 * ((int64_t) 1 << 25);
+    carry7 = (h7 + ((int64_t) 1 << 24)) >> 25;
+    h8 += carry7;
+    h7 -= carry7 * ((int64_t) 1 << 25);
+
+    carry0 = (h0 + ((int64_t) 1 << 25)) >> 26;
+    h1 += carry0;
+    h0 -= carry0 * ((int64_t) 1 << 26);
+    carry2 = (h2 + ((int64_t) 1 << 25)) >> 26;
+    h3 += carry2;
+    h2 -= carry2 * ((int64_t) 1 << 26);
+    carry4 = (h4 + ((int64_t) 1 << 25)) >> 26;
+    h5 += carry4;
+    h4 -= carry4 * ((int64_t) 1 << 26);
+    carry6 = (h6 + ((int64_t) 1 << 25)) >> 26;
+    h7 += carry6;
+    h6 -= carry6 * ((int64_t) 1 << 26);
+    carry8 = (h8 + ((int64_t) 1 << 25)) >> 26;
+    h9 += carry8;
+    h8 -= carry8 * ((int64_t) 1 << 26);
+
+    h[0] = (int32_t) h0;
+    h[1] = (int32_t) h1;
+    h[2] = (int32_t) h2;
+    h[3] = (int32_t) h3;
+    h[4] = (int32_t) h4;
+    h[5] = (int32_t) h5;
+    h[6] = (int32_t) h6;
+    h[7] = (int32_t) h7;
+    h[8] = (int32_t) h8;
+    h[9] = (int32_t) h9;
+}
diff --git a/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_51.h b/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_51.h
new file mode 100644
index 0000000000..3a30f30148
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/ed25519_ref10_fe_51.h
@@ -0,0 +1,518 @@
+#include <string.h>
+
+#include "private/common.h"
+#include "utils.h"
+
+/*
+ h = 0
+ */
+
+static inline void
+fe25519_0(fe25519 h)
+{
+    memset(&h[0], 0, 5 * sizeof h[0]);
+}
+
+/*
+ h = 1
+ */
+
+static inline void
+fe25519_1(fe25519 h)
+{
+    h[0] = 1;
+    memset(&h[1], 0, 4 * sizeof h[0]);
+}
+
+/*
+ h = f + g
+ Can overlap h with f or g.
+ */
+
+static inline void
+fe25519_add(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    uint64_t h0 = f[0] + g[0];
+    uint64_t h1 = f[1] + g[1];
+    uint64_t h2 = f[2] + g[2];
+    uint64_t h3 = f[3] + g[3];
+    uint64_t h4 = f[4] + g[4];
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+}
+
+/*
+ h = f - g
+ */
+
+static void
+fe25519_sub(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    const uint64_t mask = 0x7ffffffffffffULL;
+    uint64_t h0, h1, h2, h3, h4;
+
+    h0 = g[0];
+    h1 = g[1];
+    h2 = g[2];
+    h3 = g[3];
+    h4 = g[4];
+
+    h1 += h0 >> 51;
+    h0 &= mask;
+    h2 += h1 >> 51;
+    h1 &= mask;
+    h3 += h2 >> 51;
+    h2 &= mask;
+    h4 += h3 >> 51;
+    h3 &= mask;
+    h0 += 19ULL * (h4 >> 51);
+    h4 &= mask;
+
+    h0 = (f[0] + 0xfffffffffffdaULL) - h0;
+    h1 = (f[1] + 0xffffffffffffeULL) - h1;
+    h2 = (f[2] + 0xffffffffffffeULL) - h2;
+    h3 = (f[3] + 0xffffffffffffeULL) - h3;
+    h4 = (f[4] + 0xffffffffffffeULL) - h4;
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+}
+
+/*
+ h = -f
+ */
+
+static inline void
+fe25519_neg(fe25519 h, const fe25519 f)
+{
+    fe25519 zero;
+
+    fe25519_0(zero);
+    fe25519_sub(h, zero, f);
+}
+
+/*
+ Replace (f,g) with (g,g) if b == 1;
+ replace (f,g) with (f,g) if b == 0.
+ *
+ Preconditions: b in {0,1}.
+ */
+
+static void
+fe25519_cmov(fe25519 f, const fe25519 g, unsigned int b)
+{
+    const uint64_t mask = (uint64_t) (-(int64_t) b);
+
+    uint64_t f0 = f[0];
+    uint64_t f1 = f[1];
+    uint64_t f2 = f[2];
+    uint64_t f3 = f[3];
+    uint64_t f4 = f[4];
+
+    uint64_t x0 = f0 ^ g[0];
+    uint64_t x1 = f1 ^ g[1];
+    uint64_t x2 = f2 ^ g[2];
+    uint64_t x3 = f3 ^ g[3];
+    uint64_t x4 = f4 ^ g[4];
+
+    x0 &= mask;
+    x1 &= mask;
+    x2 &= mask;
+    x3 &= mask;
+    x4 &= mask;
+
+    f[0] = f0 ^ x0;
+    f[1] = f1 ^ x1;
+    f[2] = f2 ^ x2;
+    f[3] = f3 ^ x3;
+    f[4] = f4 ^ x4;
+}
+
+/*
+Replace (f,g) with (g,f) if b == 1;
+replace (f,g) with (f,g) if b == 0.
+
+Preconditions: b in {0,1}.
+*/
+
+static void
+fe25519_cswap(fe25519 f, fe25519 g, unsigned int b)
+{
+    const uint64_t mask = (uint64_t) (-(int64_t) b);
+
+    uint64_t f0 = f[0];
+    uint64_t f1 = f[1];
+    uint64_t f2 = f[2];
+    uint64_t f3 = f[3];
+    uint64_t f4 = f[4];
+
+    uint64_t g0 = g[0];
+    uint64_t g1 = g[1];
+    uint64_t g2 = g[2];
+    uint64_t g3 = g[3];
+    uint64_t g4 = g[4];
+
+    uint64_t x0 = f0 ^ g0;
+    uint64_t x1 = f1 ^ g1;
+    uint64_t x2 = f2 ^ g2;
+    uint64_t x3 = f3 ^ g3;
+    uint64_t x4 = f4 ^ g4;
+
+    x0 &= mask;
+    x1 &= mask;
+    x2 &= mask;
+    x3 &= mask;
+    x4 &= mask;
+
+    f[0] = f0 ^ x0;
+    f[1] = f1 ^ x1;
+    f[2] = f2 ^ x2;
+    f[3] = f3 ^ x3;
+    f[4] = f4 ^ x4;
+
+    g[0] = g0 ^ x0;
+    g[1] = g1 ^ x1;
+    g[2] = g2 ^ x2;
+    g[3] = g3 ^ x3;
+    g[4] = g4 ^ x4;
+}
+
+/*
+ h = f
+ */
+
+static inline void
+fe25519_copy(fe25519 h, const fe25519 f)
+{
+    uint64_t f0 = f[0];
+    uint64_t f1 = f[1];
+    uint64_t f2 = f[2];
+    uint64_t f3 = f[3];
+    uint64_t f4 = f[4];
+
+    h[0] = f0;
+    h[1] = f1;
+    h[2] = f2;
+    h[3] = f3;
+    h[4] = f4;
+}
+
+/*
+ return 1 if f is in {1,3,5,...,q-2}
+ return 0 if f is in {0,2,4,...,q-1}
+ */
+
+static inline int
+fe25519_isnegative(const fe25519 f)
+{
+    unsigned char s[32];
+
+    fe25519_tobytes(s, f);
+
+    return s[0] & 1;
+}
+
+/*
+ return 1 if f == 0
+ return 0 if f != 0
+ */
+
+static inline int
+fe25519_iszero(const fe25519 f)
+{
+    unsigned char s[32];
+
+    fe25519_tobytes(s, f);
+
+    return sodium_is_zero(s, 32);
+}
+
+/*
+ h = f * g
+ Can overlap h with f or g.
+ */
+
+static void
+fe25519_mul(fe25519 h, const fe25519 f, const fe25519 g)
+{
+    const uint64_t mask = 0x7ffffffffffffULL;
+    uint128_t r0, r1, r2, r3, r4, carry;
+    uint64_t  f0, f1, f2, f3, f4;
+    uint64_t  f1_19, f2_19, f3_19, f4_19;
+    uint64_t  g0, g1, g2, g3, g4;
+    uint64_t  r00, r01, r02, r03, r04;
+
+    f0 = f[0];
+    f1 = f[1];
+    f2 = f[2];
+    f3 = f[3];
+    f4 = f[4];
+
+    g0 = g[0];
+    g1 = g[1];
+    g2 = g[2];
+    g3 = g[3];
+    g4 = g[4];
+
+    f1_19 = 19ULL * f1;
+    f2_19 = 19ULL * f2;
+    f3_19 = 19ULL * f3;
+    f4_19 = 19ULL * f4;
+
+    r0  = ((uint128_t) f0   ) * ((uint128_t) g0);
+    r0 += ((uint128_t) f1_19) * ((uint128_t) g4);
+    r0 += ((uint128_t) f2_19) * ((uint128_t) g3);
+    r0 += ((uint128_t) f3_19) * ((uint128_t) g2);
+    r0 += ((uint128_t) f4_19) * ((uint128_t) g1);
+
+    r1  = ((uint128_t) f0   ) * ((uint128_t) g1);
+    r1 += ((uint128_t) f1   ) * ((uint128_t) g0);
+    r1 += ((uint128_t) f2_19) * ((uint128_t) g4);
+    r1 += ((uint128_t) f3_19) * ((uint128_t) g3);
+    r1 += ((uint128_t) f4_19) * ((uint128_t) g2);
+
+    r2  = ((uint128_t) f0   ) * ((uint128_t) g2);
+    r2 += ((uint128_t) f1   ) * ((uint128_t) g1);
+    r2 += ((uint128_t) f2   ) * ((uint128_t) g0);
+    r2 += ((uint128_t) f3_19) * ((uint128_t) g4);
+    r2 += ((uint128_t) f4_19) * ((uint128_t) g3);
+
+    r3  = ((uint128_t) f0   ) * ((uint128_t) g3);
+    r3 += ((uint128_t) f1   ) * ((uint128_t) g2);
+    r3 += ((uint128_t) f2   ) * ((uint128_t) g1);
+    r3 += ((uint128_t) f3   ) * ((uint128_t) g0);
+    r3 += ((uint128_t) f4_19) * ((uint128_t) g4);
+
+    r4  = ((uint128_t) f0   ) * ((uint128_t) g4);
+    r4 += ((uint128_t) f1   ) * ((uint128_t) g3);
+    r4 += ((uint128_t) f2   ) * ((uint128_t) g2);
+    r4 += ((uint128_t) f3   ) * ((uint128_t) g1);
+    r4 += ((uint128_t) f4   ) * ((uint128_t) g0);
+
+    r00    = ((uint64_t) r0) & mask;
+    carry  = r0 >> 51;
+    r1    += carry;
+    r01    = ((uint64_t) r1) & mask;
+    carry  = r1 >> 51;
+    r2    += carry;
+    r02    = ((uint64_t) r2) & mask;
+    carry  = r2 >> 51;
+    r3    += carry;
+    r03    = ((uint64_t) r3) & mask;
+    carry  = r3 >> 51;
+    r4    += carry;
+    r04    = ((uint64_t) r4) & mask;
+    carry  = r4 >> 51;
+    r00   += 19ULL * (uint64_t) carry;
+    carry  = r00 >> 51;
+    r00   &= mask;
+    r01   += (uint64_t) carry;
+    carry  = r01 >> 51;
+    r01   &= mask;
+    r02   += (uint64_t) carry;
+
+    h[0] = r00;
+    h[1] = r01;
+    h[2] = r02;
+    h[3] = r03;
+    h[4] = r04;
+}
+
+/*
+ h = f * f
+ Can overlap h with f.
+ */
+
+static void
+fe25519_sq(fe25519 h, const fe25519 f)
+{
+    const uint64_t mask = 0x7ffffffffffffULL;
+    uint128_t r0, r1, r2, r3, r4, carry;
+    uint64_t  f0, f1, f2, f3, f4;
+    uint64_t  f0_2, f1_2, f1_38, f2_38, f3_38, f3_19, f4_19;
+    uint64_t  r00, r01, r02, r03, r04;
+
+    f0 = f[0];
+    f1 = f[1];
+    f2 = f[2];
+    f3 = f[3];
+    f4 = f[4];
+
+    f0_2 = f0 << 1;
+    f1_2 = f1 << 1;
+
+    f1_38 = 38ULL * f1;
+    f2_38 = 38ULL * f2;
+    f3_38 = 38ULL * f3;
+
+    f3_19 = 19ULL * f3;
+    f4_19 = 19ULL * f4;
+
+    r0  = ((uint128_t) f0   ) * ((uint128_t) f0);
+    r0 += ((uint128_t) f1_38) * ((uint128_t) f4);
+    r0 += ((uint128_t) f2_38) * ((uint128_t) f3);
+
+    r1  = ((uint128_t) f0_2 ) * ((uint128_t) f1);
+    r1 += ((uint128_t) f2_38) * ((uint128_t) f4);
+    r1 += ((uint128_t) f3_19) * ((uint128_t) f3);
+
+    r2  = ((uint128_t) f0_2 ) * ((uint128_t) f2);
+    r2 += ((uint128_t) f1   ) * ((uint128_t) f1);
+    r2 += ((uint128_t) f3_38) * ((uint128_t) f4);
+
+    r3  = ((uint128_t) f0_2 ) * ((uint128_t) f3);
+    r3 += ((uint128_t) f1_2 ) * ((uint128_t) f2);
+    r3 += ((uint128_t) f4_19) * ((uint128_t) f4);
+
+    r4  = ((uint128_t) f0_2 ) * ((uint128_t) f4);
+    r4 += ((uint128_t) f1_2 ) * ((uint128_t) f3);
+    r4 += ((uint128_t) f2   ) * ((uint128_t) f2);
+
+    r00    = ((uint64_t) r0) & mask;
+    carry  = r0 >> 51;
+    r1    += carry;
+    r01    = ((uint64_t) r1) & mask;
+    carry  = r1 >> 51;
+    r2    += carry;
+    r02    = ((uint64_t) r2) & mask;
+    carry  = r2 >> 51;
+    r3    += carry;
+    r03    = ((uint64_t) r3) & mask;
+    carry  = r3 >> 51;
+    r4    += carry;
+    r04    = ((uint64_t) r4) & mask;
+    carry  = r4 >> 51;
+    r00   += 19ULL * (uint64_t) carry;
+    carry  = r00 >> 51;
+    r00   &= mask;
+    r01   += (uint64_t) carry;
+    carry  = r01 >> 51;
+    r01   &= mask;
+    r02   += (uint64_t) carry;
+
+    h[0] = r00;
+    h[1] = r01;
+    h[2] = r02;
+    h[3] = r03;
+    h[4] = r04;
+}
+
+/*
+ h = 2 * f * f
+ Can overlap h with f.
+*/
+
+static void
+fe25519_sq2(fe25519 h, const fe25519 f)
+{
+    const uint64_t mask = 0x7ffffffffffffULL;
+    uint128_t r0, r1, r2, r3, r4, carry;
+    uint64_t  f0, f1, f2, f3, f4;
+    uint64_t  f0_2, f1_2, f1_38, f2_38, f3_38, f3_19, f4_19;
+    uint64_t  r00, r01, r02, r03, r04;
+
+    f0 = f[0];
+    f1 = f[1];
+    f2 = f[2];
+    f3 = f[3];
+    f4 = f[4];
+
+    f0_2 = f0 << 1;
+    f1_2 = f1 << 1;
+
+    f1_38 = 38ULL * f1;
+    f2_38 = 38ULL * f2;
+    f3_38 = 38ULL * f3;
+
+    f3_19 = 19ULL * f3;
+    f4_19 = 19ULL * f4;
+
+    r0  = ((uint128_t) f0   ) * ((uint128_t) f0);
+    r0 += ((uint128_t) f1_38) * ((uint128_t) f4);
+    r0 += ((uint128_t) f2_38) * ((uint128_t) f3);
+
+    r1  = ((uint128_t) f0_2 ) * ((uint128_t) f1);
+    r1 += ((uint128_t) f2_38) * ((uint128_t) f4);
+    r1 += ((uint128_t) f3_19) * ((uint128_t) f3);
+
+    r2  = ((uint128_t) f0_2 ) * ((uint128_t) f2);
+    r2 += ((uint128_t) f1   ) * ((uint128_t) f1);
+    r2 += ((uint128_t) f3_38) * ((uint128_t) f4);
+
+    r3  = ((uint128_t) f0_2 ) * ((uint128_t) f3);
+    r3 += ((uint128_t) f1_2 ) * ((uint128_t) f2);
+    r3 += ((uint128_t) f4_19) * ((uint128_t) f4);
+
+    r4  = ((uint128_t) f0_2 ) * ((uint128_t) f4);
+    r4 += ((uint128_t) f1_2 ) * ((uint128_t) f3);
+    r4 += ((uint128_t) f2   ) * ((uint128_t) f2);
+
+    r0 <<= 1;
+    r1 <<= 1;
+    r2 <<= 1;
+    r3 <<= 1;
+    r4 <<= 1;
+
+    r00    = ((uint64_t) r0) & mask;
+    carry  = r0 >> 51;
+    r1    += carry;
+    r01    = ((uint64_t) r1) & mask;
+    carry  = r1 >> 51;
+    r2    += carry;
+    r02    = ((uint64_t) r2) & mask;
+    carry  = r2 >> 51;
+    r3    += carry;
+    r03    = ((uint64_t) r3) & mask;
+    carry  = r3 >> 51;
+    r4    += carry;
+    r04    = ((uint64_t) r4) & mask;
+    carry  = r4 >> 51;
+    r00   += 19ULL * (uint64_t) carry;
+    carry  = r00 >> 51;
+    r00   &= mask;
+    r01   += (uint64_t) carry;
+    carry  = r01 >> 51;
+    r01   &= mask;
+    r02   += (uint64_t) carry;
+
+    h[0] = r00;
+    h[1] = r01;
+    h[2] = r02;
+    h[3] = r03;
+    h[4] = r04;
+}
+
+static void
+fe25519_scalar_product(fe25519 h, const fe25519 f, uint32_t n)
+{
+    const uint64_t mask = 0x7ffffffffffffULL;
+    uint128_t a;
+    uint128_t sn = (uint128_t) n;
+    uint64_t  h0, h1, h2, h3, h4;
+
+    a  = f[0] * sn;
+    h0 = ((uint64_t) a) & mask;
+    a  = f[1] * sn + ((uint64_t) (a >> 51));
+    h1 = ((uint64_t) a) & mask;
+    a  = f[2] * sn + ((uint64_t) (a >> 51));
+    h2 = ((uint64_t) a) & mask;
+    a  = f[3] * sn + ((uint64_t) (a >> 51));
+    h3 = ((uint64_t) a) & mask;
+    a  = f[4] * sn + ((uint64_t) (a >> 51));
+    h4 = ((uint64_t) a) & mask;
+
+    h0 += (a >> 51) * 19ULL;
+
+    h[0] = h0;
+    h[1] = h1;
+    h[2] = h2;
+    h[3] = h3;
+    h[4] = h4;
+}
diff --git a/libs/libsodium/src/include/sodium/private/implementations.h b/libs/libsodium/src/include/sodium/private/implementations.h
new file mode 100644
index 0000000000..c7237f851d
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/implementations.h
@@ -0,0 +1,11 @@
+#ifndef implementations_H
+#define implementations_H
+
+int _crypto_generichash_blake2b_pick_best_implementation(void);
+int _crypto_onetimeauth_poly1305_pick_best_implementation(void);
+int _crypto_pwhash_argon2_pick_best_implementation(void);
+int _crypto_scalarmult_curve25519_pick_best_implementation(void);
+int _crypto_stream_chacha20_pick_best_implementation(void);
+int _crypto_stream_salsa20_pick_best_implementation(void);
+
+#endif
diff --git a/libs/libsodium/src/include/sodium/private/mutex.h b/libs/libsodium/src/include/sodium/private/mutex.h
new file mode 100644
index 0000000000..322b6742b2
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/mutex.h
@@ -0,0 +1,7 @@
+#ifndef mutex_H
+#define mutex_H 1
+
+extern int sodium_crit_enter(void);
+extern int sodium_crit_leave(void);
+
+#endif
diff --git a/libs/libsodium/src/include/sodium/private/sse2_64_32.h b/libs/libsodium/src/include/sodium/private/sse2_64_32.h
new file mode 100644
index 0000000000..d0455b41b4
--- /dev/null
+++ b/libs/libsodium/src/include/sodium/private/sse2_64_32.h
@@ -0,0 +1,50 @@
+#ifndef sse2_64_32_H
+#define sse2_64_32_H 1
+
+#include "common.h"
+
+#ifdef HAVE_INTRIN_H
+# include <intrin.h>
+#endif
+
+#if defined(HAVE_EMMINTRIN_H) && \
+    !(defined(__amd64) || defined(__amd64__) || defined(__x86_64__) || \
+      defined(_M_X64) || defined(_M_AMD64))
+
+# include <emmintrin.h>
+# include <stdint.h>
+
+# ifndef _mm_set_epi64x
+#  define _mm_set_epi64x(Q0, Q1) sodium__mm_set_epi64x((Q0), (Q1))
+static inline __m128i
+sodium__mm_set_epi64x(int64_t q1, int64_t q0)
+{
+    union { int64_t as64; int32_t as32[2]; } x0, x1;
+    x0.as64 = q0; x1.as64 = q1;
+    return _mm_set_epi32(x1.as32[1], x1.as32[0], x0.as32[1], x0.as32[0]);
+}
+# endif
+
+# ifndef _mm_set1_epi64x
+#  define _mm_set1_epi64x(Q) sodium__mm_set1_epi64x(Q)
+static inline __m128i
+sodium__mm_set1_epi64x(int64_t q)
+{
+    return _mm_set_epi64x(q, q);
+}
+# endif
+
+# ifndef _mm_cvtsi64_si128
+#  define _mm_cvtsi64_si128(Q) sodium__mm_cvtsi64_si128(Q)
+static inline __m128i
+sodium__mm_cvtsi64_si128(int64_t q)
+{
+    union { int64_t as64; int32_t as32[2]; } x;
+    x.as64 = q;
+    return _mm_setr_epi32(x.as32[0], x.as32[1], 0, 0);
+}
+# endif
+
+#endif
+
+#endif