source/vendor/github.com/cloudflare/circl/internal/sha3/sha3.go

+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// spongeDirection indicates the direction bytes are flowing through the sponge.
+type spongeDirection int
+
+const (
+	// spongeAbsorbing indicates that the sponge is absorbing input.
+	spongeAbsorbing spongeDirection = iota
+	// spongeSqueezing indicates that the sponge is being squeezed.
+	spongeSqueezing
+)
+
+const (
+	// maxRate is the maximum size of the internal buffer. SHAKE-256
+	// currently needs the largest buffer.
+	maxRate = 168
+)
+
+func (d *State) buf() []byte {
+	return d.storage.asBytes()[d.bufo:d.bufe]
+}
+
+type State struct {
+	// Generic sponge components.
+	a    [25]uint64 // main state of the hash
+	rate int        // the number of bytes of state to use
+
+	bufo int // offset of buffer in storage
+	bufe int // end of buffer in storage
+
+	// dsbyte contains the "domain separation" bits and the first bit of
+	// the padding. Sections 6.1 and 6.2 of [1] separate the outputs of the
+	// SHA-3 and SHAKE functions by appending bitstrings to the message.
+	// Using a little-endian bit-ordering convention, these are "01" for SHA-3
+	// and "1111" for SHAKE, or 00000010b and 00001111b, respectively. Then the
+	// padding rule from section 5.1 is applied to pad the message to a multiple
+	// of the rate, which involves adding a "1" bit, zero or more "0" bits, and
+	// a final "1" bit. We merge the first "1" bit from the padding into dsbyte,
+	// giving 00000110b (0x06) and 00011111b (0x1f).
+	// [1] http://csrc.nist.gov/publications/drafts/fips-202/fips_202_draft.pdf
+	//     "Draft FIPS 202: SHA-3 Standard: Permutation-Based Hash and
+	//      Extendable-Output Functions (May 2014)"
+	dsbyte byte
+
+	storage storageBuf
+
+	// Specific to SHA-3 and SHAKE.
+	outputLen int             // the default output size in bytes
+	state     spongeDirection // whether the sponge is absorbing or squeezing
+	turbo     bool            // Whether we're using 12 rounds instead of 24
+}
+
+// BlockSize returns the rate of sponge underlying this hash function.
+func (d *State) BlockSize() int { return d.rate }
+
+// Size returns the output size of the hash function in bytes.
+func (d *State) Size() int { return d.outputLen }
+
+// Reset clears the internal state by zeroing the sponge state and
+// the byte buffer, and setting Sponge.state to absorbing.
+func (d *State) Reset() {
+	// Zero the permutation's state.
+	for i := range d.a {
+		d.a[i] = 0
+	}
+	d.state = spongeAbsorbing
+	d.bufo = 0
+	d.bufe = 0
+}
+
+func (d *State) clone() *State {
+	ret := *d
+	return &ret
+}
+
+// permute applies the KeccakF-1600 permutation. It handles
+// any input-output buffering.
+func (d *State) permute() {
+	switch d.state {
+	case spongeAbsorbing:
+		// If we're absorbing, we need to xor the input into the state
+		// before applying the permutation.
+		xorIn(d, d.buf())
+		d.bufe = 0
+		d.bufo = 0
+		KeccakF1600(&d.a, d.turbo)
+	case spongeSqueezing:
+		// If we're squeezing, we need to apply the permutation before
+		// copying more output.
+		KeccakF1600(&d.a, d.turbo)
+		d.bufe = d.rate
+		d.bufo = 0
+		copyOut(d, d.buf())
+	}
+}
+
+// pads appends the domain separation bits in dsbyte, applies
+// the multi-bitrate 10..1 padding rule, and permutes the state.
+func (d *State) padAndPermute(dsbyte byte) {
+	// Pad with this instance's domain-separator bits. We know that there's
+	// at least one byte of space in d.buf() because, if it were full,
+	// permute would have been called to empty it. dsbyte also contains the
+	// first one bit for the padding. See the comment in the state struct.
+	zerosStart := d.bufe + 1
+	d.bufe = d.rate
+	buf := d.buf()
+	buf[zerosStart-1] = dsbyte
+	for i := zerosStart; i < d.rate; i++ {
+		buf[i] = 0
+	}
+	// This adds the final one bit for the padding. Because of the way that
+	// bits are numbered from the LSB upwards, the final bit is the MSB of
+	// the last byte.
+	buf[d.rate-1] ^= 0x80
+	// Apply the permutation
+	d.permute()
+	d.state = spongeSqueezing
+	d.bufe = d.rate
+	copyOut(d, buf)
+}
+
+// Write absorbs more data into the hash's state. It produces an error
+// if more data is written to the ShakeHash after writing
+func (d *State) Write(p []byte) (written int, err error) {
+	if d.state != spongeAbsorbing {
+		panic("sha3: write to sponge after read")
+	}
+	written = len(p)
+
+	for len(p) > 0 {
+		bufl := d.bufe - d.bufo
+		if bufl == 0 && len(p) >= d.rate {
+			// The fast path; absorb a full "rate" bytes of input and apply the permutation.
+			xorIn(d, p[:d.rate])
+			p = p[d.rate:]
+			KeccakF1600(&d.a, d.turbo)
+		} else {
+			// The slow path; buffer the input until we can fill the sponge, and then xor it in.
+			todo := d.rate - bufl
+			if todo > len(p) {
+				todo = len(p)
+			}
+			d.bufe += todo
+			buf := d.buf()
+			copy(buf[bufl:], p[:todo])
+			p = p[todo:]
+
+			// If the sponge is full, apply the permutation.
+			if d.bufe == d.rate {
+				d.permute()
+			}
+		}
+	}
+
+	return written, nil
+}
+
+// Read squeezes an arbitrary number of bytes from the sponge.
+func (d *State) Read(out []byte) (n int, err error) {
+	// If we're still absorbing, pad and apply the permutation.
+	if d.state == spongeAbsorbing {
+		d.padAndPermute(d.dsbyte)
+	}
+
+	n = len(out)
+
+	// Now, do the squeezing.
+	for len(out) > 0 {
+		buf := d.buf()
+		n := copy(out, buf)
+		d.bufo += n
+		out = out[n:]
+
+		// Apply the permutation if we've squeezed the sponge dry.
+		if d.bufo == d.bufe {
+			d.permute()
+		}
+	}
+
+	return
+}
+
+// Sum applies padding to the hash state and then squeezes out the desired
+// number of output bytes.
+func (d *State) Sum(in []byte) []byte {
+	// Make a copy of the original hash so that caller can keep writing
+	// and summing.
+	dup := d.clone()
+	hash := make([]byte, dup.outputLen)
+	_, _ = dup.Read(hash)
+	return append(in, hash...)
+}
+
+func (d *State) IsAbsorbing() bool {
+	return d.state == spongeAbsorbing
+}

source/vendor/github.com/cloudflare/circl/internal/sha3/sha3_s390x.s

+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build !gccgo,!appengine
+
+#include "textflag.h"
+
+// func kimd(function code, chain *[200]byte, src []byte)
+TEXT ·kimd(SB), NOFRAME|NOSPLIT, $0-40
+	MOVD function+0(FP), R0
+	MOVD chain+8(FP), R1
+	LMG  src+16(FP), R2, R3 // R2=base, R3=len
+
+continue:
+	WORD $0xB93E0002 // KIMD --, R2
+	BVS  continue    // continue if interrupted
+	MOVD $0, R0      // reset R0 for pre-go1.8 compilers
+	RET
+
+// func klmd(function code, chain *[200]byte, dst, src []byte)
+TEXT ·klmd(SB), NOFRAME|NOSPLIT, $0-64
+	// TODO: SHAKE support
+	MOVD function+0(FP), R0
+	MOVD chain+8(FP), R1
+	LMG  dst+16(FP), R2, R3 // R2=base, R3=len
+	LMG  src+40(FP), R4, R5 // R4=base, R5=len
+
+continue:
+	WORD $0xB93F0024 // KLMD R2, R4
+	BVS  continue    // continue if interrupted
+	MOVD $0, R0      // reset R0 for pre-go1.8 compilers
+	RET

source/vendor/github.com/cloudflare/circl/internal/sha3/shake.go

+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// This file defines the ShakeHash interface, and provides
+// functions for creating SHAKE and cSHAKE instances, as well as utility
+// functions for hashing bytes to arbitrary-length output.
+//
+//
+// SHAKE implementation is based on FIPS PUB 202 [1]
+// cSHAKE implementations is based on NIST SP 800-185 [2]
+//
+// [1] https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf
+// [2] https://doi.org/10.6028/NIST.SP.800-185
+
+import (
+	"io"
+)
+
+// ShakeHash defines the interface to hash functions that
+// support arbitrary-length output.
+type ShakeHash interface {
+	// Write absorbs more data into the hash's state. It panics if input is
+	// written to it after output has been read from it.
+	io.Writer
+
+	// Read reads more output from the hash; reading affects the hash's
+	// state. (ShakeHash.Read is thus very different from Hash.Sum)
+	// It never returns an error.
+	io.Reader
+
+	// Clone returns a copy of the ShakeHash in its current state.
+	Clone() ShakeHash
+
+	// Reset resets the ShakeHash to its initial state.
+	Reset()
+}
+
+// Consts for configuring initial SHA-3 state
+const (
+	dsbyteShake = 0x1f
+	rate128     = 168
+	rate256     = 136
+)
+
+// Clone returns copy of SHAKE context within its current state.
+func (d *State) Clone() ShakeHash {
+	return d.clone()
+}
+
+// NewShake128 creates a new SHAKE128 variable-output-length ShakeHash.
+// Its generic security strength is 128 bits against all attacks if at
+// least 32 bytes of its output are used.
+func NewShake128() State {
+	return State{rate: rate128, dsbyte: dsbyteShake}
+}
+
+// NewTurboShake128 creates a new TurboSHAKE128 variable-output-length ShakeHash.
+// Its generic security strength is 128 bits against all attacks if at
+// least 32 bytes of its output are used.
+// D is the domain separation byte and must be between 0x01 and 0x7f inclusive.
+func NewTurboShake128(D byte) State {
+	if D == 0 || D > 0x7f {
+		panic("turboshake: D out of range")
+	}
+	return State{rate: rate128, dsbyte: D, turbo: true}
+}
+
+// NewShake256 creates a new SHAKE256 variable-output-length ShakeHash.
+// Its generic security strength is 256 bits against all attacks if
+// at least 64 bytes of its output are used.
+func NewShake256() State {
+	return State{rate: rate256, dsbyte: dsbyteShake}
+}
+
+// NewTurboShake256 creates a new TurboSHAKE256 variable-output-length ShakeHash.
+// Its generic security strength is 256 bits against all attacks if
+// at least 64 bytes of its output are used.
+// D is the domain separation byte and must be between 0x01 and 0x7f inclusive.
+func NewTurboShake256(D byte) State {
+	if D == 0 || D > 0x7f {
+		panic("turboshake: D out of range")
+	}
+	return State{rate: rate256, dsbyte: D, turbo: true}
+}
+
+// ShakeSum128 writes an arbitrary-length digest of data into hash.
+func ShakeSum128(hash, data []byte) {
+	h := NewShake128()
+	_, _ = h.Write(data)
+	_, _ = h.Read(hash)
+}
+
+// ShakeSum256 writes an arbitrary-length digest of data into hash.
+func ShakeSum256(hash, data []byte) {
+	h := NewShake256()
+	_, _ = h.Write(data)
+	_, _ = h.Read(hash)
+}
+
+// TurboShakeSum128 writes an arbitrary-length digest of data into hash.
+func TurboShakeSum128(hash, data []byte, D byte) {
+	h := NewTurboShake128(D)
+	_, _ = h.Write(data)
+	_, _ = h.Read(hash)
+}
+
+// TurboShakeSum256 writes an arbitrary-length digest of data into hash.
+func TurboShakeSum256(hash, data []byte, D byte) {
+	h := NewTurboShake256(D)
+	_, _ = h.Write(data)
+	_, _ = h.Read(hash)
+}
+
+func (d *State) SwitchDS(D byte) {
+	d.dsbyte = D
+}

source/vendor/github.com/cloudflare/circl/internal/sha3/xor.go

+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (!amd64 && !386 && !ppc64le) || appengine
+// +build !amd64,!386,!ppc64le appengine
+
+package sha3
+
+// A storageBuf is an aligned array of maxRate bytes.
+type storageBuf [maxRate]byte
+
+func (b *storageBuf) asBytes() *[maxRate]byte {
+	return (*[maxRate]byte)(b)
+}

source/vendor/github.com/cloudflare/circl/internal/sha3/xor_generic.go

+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (!amd64 || appengine) && (!386 || appengine) && (!ppc64le || appengine)
+// +build !amd64 appengine
+// +build !386 appengine
+// +build !ppc64le appengine
+
+package sha3
+
+import "encoding/binary"
+
+// xorIn xors the bytes in buf into the state; it
+// makes no non-portable assumptions about memory layout
+// or alignment.
+func xorIn(d *State, buf []byte) {
+	n := len(buf) / 8
+
+	for i := 0; i < n; i++ {
+		a := binary.LittleEndian.Uint64(buf)
+		d.a[i] ^= a
+		buf = buf[8:]
+	}
+}
+
+// copyOut copies ulint64s to a byte buffer.
+func copyOut(d *State, b []byte) {
+	for i := 0; len(b) >= 8; i++ {
+		binary.LittleEndian.PutUint64(b, d.a[i])
+		b = b[8:]
+	}
+}

source/vendor/github.com/cloudflare/circl/internal/sha3/xor_unaligned.go

+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build (amd64 || 386 || ppc64le) && !appengine
+// +build amd64 386 ppc64le
+// +build !appengine
+
+package sha3
+
+import "unsafe"
+
+// A storageBuf is an aligned array of maxRate bytes.
+type storageBuf [maxRate / 8]uint64
+
+func (b *storageBuf) asBytes() *[maxRate]byte {
+	return (*[maxRate]byte)(unsafe.Pointer(b))
+}
+
+// xorInuses unaligned reads and writes to update d.a to contain d.a
+// XOR buf.
+func xorIn(d *State, buf []byte) {
+	n := len(buf)
+	bw := (*[maxRate / 8]uint64)(unsafe.Pointer(&buf[0]))[: n/8 : n/8]
+	if n >= 72 {
+		d.a[0] ^= bw[0]
+		d.a[1] ^= bw[1]
+		d.a[2] ^= bw[2]
+		d.a[3] ^= bw[3]
+		d.a[4] ^= bw[4]
+		d.a[5] ^= bw[5]
+		d.a[6] ^= bw[6]
+		d.a[7] ^= bw[7]
+		d.a[8] ^= bw[8]
+	}
+	if n >= 104 {
+		d.a[9] ^= bw[9]
+		d.a[10] ^= bw[10]
+		d.a[11] ^= bw[11]
+		d.a[12] ^= bw[12]
+	}
+	if n >= 136 {
+		d.a[13] ^= bw[13]
+		d.a[14] ^= bw[14]
+		d.a[15] ^= bw[15]
+		d.a[16] ^= bw[16]
+	}
+	if n >= 144 {
+		d.a[17] ^= bw[17]
+	}
+	if n >= 168 {
+		d.a[18] ^= bw[18]
+		d.a[19] ^= bw[19]
+		d.a[20] ^= bw[20]
+	}
+}
+
+func copyOut(d *State, buf []byte) {
+	ab := (*[maxRate]uint8)(unsafe.Pointer(&d.a[0]))
+	copy(buf, ab[:])
+}

source/vendor/github.com/cloudflare/circl/math/fp25519/fp.go

+// Package fp25519 provides prime field arithmetic over GF(2^255-19).
+package fp25519
+
+import (
+	"errors"
+
+	"github.com/cloudflare/circl/internal/conv"
+)
+
+// Size in bytes of an element.
+const Size = 32
+
+// Elt is a prime field element.
+type Elt [Size]byte
+
+func (e Elt) String() string { return conv.BytesLe2Hex(e[:]) }
+
+// p is the prime modulus 2^255-19.
+var p = Elt{
+	0xed, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x7f,
+}
+
+// P returns the prime modulus 2^255-19.
+func P() Elt { return p }
+
+// ToBytes stores in b the little-endian byte representation of x.
+func ToBytes(b []byte, x *Elt) error {
+	if len(b) != Size {
+		return errors.New("wrong size")
+	}
+	Modp(x)
+	copy(b, x[:])
+	return nil
+}
+
+// IsZero returns true if x is equal to 0.
+func IsZero(x *Elt) bool { Modp(x); return *x == Elt{} }
+
+// SetOne assigns x=1.
+func SetOne(x *Elt) { *x = Elt{}; x[0] = 1 }
+
+// Neg calculates z = -x.
+func Neg(z, x *Elt) { Sub(z, &p, x) }
+
+// InvSqrt calculates z = sqrt(x/y) iff x/y is a quadratic-residue, which is
+// indicated by returning isQR = true. Otherwise, when x/y is a quadratic
+// non-residue, z will have an undetermined value and isQR = false.
+func InvSqrt(z, x, y *Elt) (isQR bool) {
+	sqrtMinusOne := &Elt{
+		0xb0, 0xa0, 0x0e, 0x4a, 0x27, 0x1b, 0xee, 0xc4,
+		0x78, 0xe4, 0x2f, 0xad, 0x06, 0x18, 0x43, 0x2f,
+		0xa7, 0xd7, 0xfb, 0x3d, 0x99, 0x00, 0x4d, 0x2b,
+		0x0b, 0xdf, 0xc1, 0x4f, 0x80, 0x24, 0x83, 0x2b,
+	}
+	t0, t1, t2, t3 := &Elt{}, &Elt{}, &Elt{}, &Elt{}
+
+	Mul(t0, x, y)   // t0 = u*v
+	Sqr(t1, y)      // t1 = v^2
+	Mul(t2, t0, t1) // t2 = u*v^3
+	Sqr(t0, t1)     // t0 = v^4
+	Mul(t1, t0, t2) // t1 = u*v^7
+
+	var Tab [4]*Elt
+	Tab[0] = &Elt{}
+	Tab[1] = &Elt{}
+	Tab[2] = t3
+	Tab[3] = t1
+
+	*Tab[0] = *t1
+	Sqr(Tab[0], Tab[0])
+	Sqr(Tab[1], Tab[0])
+	Sqr(Tab[1], Tab[1])
+	Mul(Tab[1], Tab[1], Tab[3])
+	Mul(Tab[0], Tab[0], Tab[1])
+	Sqr(Tab[0], Tab[0])
+	Mul(Tab[0], Tab[0], Tab[1])
+	Sqr(Tab[1], Tab[0])
+	for i := 0; i < 4; i++ {
+		Sqr(Tab[1], Tab[1])
+	}
+	Mul(Tab[1], Tab[1], Tab[0])
+	Sqr(Tab[2], Tab[1])
+	for i := 0; i < 4; i++ {
+		Sqr(Tab[2], Tab[2])
+	}
+	Mul(Tab[2], Tab[2], Tab[0])
+	Sqr(Tab[1], Tab[2])
+	for i := 0; i < 14; i++ {
+		Sqr(Tab[1], Tab[1])
+	}
+	Mul(Tab[1], Tab[1], Tab[2])
+	Sqr(Tab[2], Tab[1])
+	for i := 0; i < 29; i++ {
+		Sqr(Tab[2], Tab[2])
+	}
+	Mul(Tab[2], Tab[2], Tab[1])
+	Sqr(Tab[1], Tab[2])
+	for i := 0; i < 59; i++ {
+		Sqr(Tab[1], Tab[1])
+	}
+	Mul(Tab[1], Tab[1], Tab[2])
+	for i := 0; i < 5; i++ {
+		Sqr(Tab[1], Tab[1])
+	}
+	Mul(Tab[1], Tab[1], Tab[0])
+	Sqr(Tab[2], Tab[1])
+	for i := 0; i < 124; i++ {
+		Sqr(Tab[2], Tab[2])
+	}
+	Mul(Tab[2], Tab[2], Tab[1])
+	Sqr(Tab[2], Tab[2])
+	Sqr(Tab[2], Tab[2])
+	Mul(Tab[2], Tab[2], Tab[3])
+
+	Mul(z, t3, t2) // z = xy^(p+3)/8 = xy^3*(xy^7)^(p-5)/8
+	// Checking whether y z^2 == x
+	Sqr(t0, z)     // t0 = z^2
+	Mul(t0, t0, y) // t0 = yz^2
+	Sub(t1, t0, x) // t1 = t0-u
+	Add(t2, t0, x) // t2 = t0+u
+	if IsZero(t1) {
+		return true
+	} else if IsZero(t2) {
+		Mul(z, z, sqrtMinusOne) // z = z*sqrt(-1)
+		return true
+	} else {
+		return false
+	}
+}
+
+// Inv calculates z = 1/x mod p.
+func Inv(z, x *Elt) {
+	x0, x1, x2 := &Elt{}, &Elt{}, &Elt{}
+	Sqr(x1, x)
+	Sqr(x0, x1)
+	Sqr(x0, x0)
+	Mul(x0, x0, x)
+	Mul(z, x0, x1)
+	Sqr(x1, z)
+	Mul(x0, x0, x1)
+	Sqr(x1, x0)
+	for i := 0; i < 4; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x0, x0, x1)
+	Sqr(x1, x0)
+	for i := 0; i < 9; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, x0)
+	Sqr(x2, x1)
+	for i := 0; i < 19; i++ {
+		Sqr(x2, x2)
+	}
+	Mul(x2, x2, x1)
+	for i := 0; i < 10; i++ {
+		Sqr(x2, x2)
+	}
+	Mul(x2, x2, x0)
+	Sqr(x0, x2)
+	for i := 0; i < 49; i++ {
+		Sqr(x0, x0)
+	}
+	Mul(x0, x0, x2)
+	Sqr(x1, x0)
+	for i := 0; i < 99; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, x0)
+	for i := 0; i < 50; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, x2)
+	for i := 0; i < 5; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(z, z, x1)
+}
+
+// Cmov assigns y to x if n is 1.
+func Cmov(x, y *Elt, n uint) { cmov(x, y, n) }
+
+// Cswap interchanges x and y if n is 1.
+func Cswap(x, y *Elt, n uint) { cswap(x, y, n) }
+
+// Add calculates z = x+y mod p.
+func Add(z, x, y *Elt) { add(z, x, y) }
+
+// Sub calculates z = x-y mod p.
+func Sub(z, x, y *Elt) { sub(z, x, y) }
+
+// AddSub calculates (x,y) = (x+y mod p, x-y mod p).
+func AddSub(x, y *Elt) { addsub(x, y) }
+
+// Mul calculates z = x*y mod p.
+func Mul(z, x, y *Elt) { mul(z, x, y) }
+
+// Sqr calculates z = x^2 mod p.
+func Sqr(z, x *Elt) { sqr(z, x) }
+
+// Modp ensures that z is between [0,p-1].
+func Modp(z *Elt) { modp(z) }

source/vendor/github.com/cloudflare/circl/math/fp25519/fp_amd64.go

+//go:build amd64 && !purego
+// +build amd64,!purego
+
+package fp25519
+
+import (
+	"golang.org/x/sys/cpu"
+)
+
+var hasBmi2Adx = cpu.X86.HasBMI2 && cpu.X86.HasADX
+
+var _ = hasBmi2Adx
+
+func cmov(x, y *Elt, n uint)  { cmovAmd64(x, y, n) }
+func cswap(x, y *Elt, n uint) { cswapAmd64(x, y, n) }
+func add(z, x, y *Elt)        { addAmd64(z, x, y) }
+func sub(z, x, y *Elt)        { subAmd64(z, x, y) }
+func addsub(x, y *Elt)        { addsubAmd64(x, y) }
+func mul(z, x, y *Elt)        { mulAmd64(z, x, y) }
+func sqr(z, x *Elt)           { sqrAmd64(z, x) }
+func modp(z *Elt)             { modpAmd64(z) }
+
+//go:noescape
+func cmovAmd64(x, y *Elt, n uint)
+
+//go:noescape
+func cswapAmd64(x, y *Elt, n uint)
+
+//go:noescape
+func addAmd64(z, x, y *Elt)
+
+//go:noescape
+func subAmd64(z, x, y *Elt)
+
+//go:noescape
+func addsubAmd64(x, y *Elt)
+
+//go:noescape
+func mulAmd64(z, x, y *Elt)
+
+//go:noescape
+func sqrAmd64(z, x *Elt)
+
+//go:noescape
+func modpAmd64(z *Elt)

source/vendor/github.com/cloudflare/circl/math/fp25519/fp_amd64.h

+// This code was imported from https://github.com/armfazh/rfc7748_precomputed
+
+// CHECK_BMI2ADX triggers bmi2adx if supported,
+// otherwise it fallbacks to legacy code.
+#define CHECK_BMI2ADX(label, legacy, bmi2adx) \
+    CMPB ·hasBmi2Adx(SB), $0  \
+    JE label                  \
+    bmi2adx                   \
+    RET                       \
+    label:                    \
+    legacy                    \
+    RET
+
+// cselect is a conditional move
+// if b=1: it copies y into x;
+// if b=0: x remains with the same value;
+// if b<> 0,1: undefined.
+// Uses: AX, DX, FLAGS
+// Instr: x86_64, cmov
+#define cselect(x,y,b) \
+    TESTQ b, b \
+    MOVQ  0+x, AX; MOVQ  0+y, DX; CMOVQNE DX, AX; MOVQ AX,  0+x; \
+    MOVQ  8+x, AX; MOVQ  8+y, DX; CMOVQNE DX, AX; MOVQ AX,  8+x; \
+    MOVQ 16+x, AX; MOVQ 16+y, DX; CMOVQNE DX, AX; MOVQ AX, 16+x; \
+    MOVQ 24+x, AX; MOVQ 24+y, DX; CMOVQNE DX, AX; MOVQ AX, 24+x;
+
+// cswap is a conditional swap
+// if b=1: x,y <- y,x;
+// if b=0: x,y remain with the same values;
+// if b<> 0,1: undefined.
+// Uses: AX, DX, R8, FLAGS
+// Instr: x86_64, cmov
+#define cswap(x,y,b) \
+    TESTQ b, b \
+    MOVQ  0+x, AX; MOVQ AX, R8; MOVQ  0+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX,  0+x; MOVQ DX,  0+y; \
+    MOVQ  8+x, AX; MOVQ AX, R8; MOVQ  8+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX,  8+x; MOVQ DX,  8+y; \
+    MOVQ 16+x, AX; MOVQ AX, R8; MOVQ 16+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 16+x; MOVQ DX, 16+y; \
+    MOVQ 24+x, AX; MOVQ AX, R8; MOVQ 24+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 24+x; MOVQ DX, 24+y;
+
+// additionLeg adds x and y and stores in z
+// Uses: AX, DX, R8-R11, FLAGS
+// Instr: x86_64, cmov
+#define additionLeg(z,x,y) \
+    MOVL $38, AX; \
+    MOVL  $0, DX; \
+    MOVQ  0+x,  R8;  ADDQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  ADCQ  8+y,  R9; \
+    MOVQ 16+x, R10;  ADCQ 16+y, R10; \
+    MOVQ 24+x, R11;  ADCQ 24+y, R11; \
+    CMOVQCS AX, DX;    \
+    ADDQ DX,  R8; \
+    ADCQ $0,  R9;  MOVQ  R9,  8+z; \
+    ADCQ $0, R10;  MOVQ R10, 16+z; \
+    ADCQ $0, R11;  MOVQ R11, 24+z; \
+    MOVL $0,  DX; \
+    CMOVQCS AX, DX; \
+    ADDQ DX,  R8;  MOVQ  R8,  0+z;
+
+// additionAdx adds x and y and stores in z
+// Uses: AX, DX, R8-R11, FLAGS
+// Instr: x86_64, cmov, adx
+#define additionAdx(z,x,y) \
+    MOVL $38, AX; \
+    XORL  DX, DX; \
+    MOVQ  0+x,  R8;  ADCXQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  ADCXQ  8+y,  R9; \
+    MOVQ 16+x, R10;  ADCXQ 16+y, R10; \
+    MOVQ 24+x, R11;  ADCXQ 24+y, R11; \
+    CMOVQCS AX, DX ; \
+    XORL  AX,  AX; \
+    ADCXQ DX,  R8; \
+    ADCXQ AX,  R9;  MOVQ  R9,  8+z; \
+    ADCXQ AX, R10;  MOVQ R10, 16+z; \
+    ADCXQ AX, R11;  MOVQ R11, 24+z; \
+    MOVL $38,  DX; \
+    CMOVQCS DX, AX; \
+    ADDQ  AX,  R8;  MOVQ  R8,  0+z;
+
+// subtraction subtracts y from x and stores in z
+// Uses: AX, DX, R8-R11, FLAGS
+// Instr: x86_64, cmov
+#define subtraction(z,x,y) \
+    MOVL   $38,  AX; \
+    MOVQ  0+x,  R8;  SUBQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  SBBQ  8+y,  R9; \
+    MOVQ 16+x, R10;  SBBQ 16+y, R10; \
+    MOVQ 24+x, R11;  SBBQ 24+y, R11; \
+    MOVL $0, DX; \
+    CMOVQCS AX, DX; \
+    SUBQ  DX,  R8; \
+    SBBQ  $0,  R9;  MOVQ  R9,  8+z; \
+    SBBQ  $0, R10;  MOVQ R10, 16+z; \
+    SBBQ  $0, R11;  MOVQ R11, 24+z; \
+    MOVL  $0,  DX; \
+    CMOVQCS AX, DX; \
+    SUBQ  DX,  R8;  MOVQ  R8,  0+z;
+
+// integerMulAdx multiplies x and y and stores in z
+// Uses: AX, DX, R8-R15, FLAGS
+// Instr: x86_64, bmi2, adx
+#define integerMulAdx(z,x,y) \
+    MOVL    $0,R15; \
+    MOVQ   0+y, DX;       XORL  AX,  AX; \
+    MULXQ  0+x, AX,  R8;  MOVQ  AX, 0+z; \
+    MULXQ  8+x, AX,  R9;  ADCXQ AX,  R8; \
+    MULXQ 16+x, AX, R10;  ADCXQ AX,  R9; \
+    MULXQ 24+x, AX, R11;  ADCXQ AX, R10; \
+    MOVL $0, AX;;;;;;;;;  ADCXQ AX, R11; \
+    MOVQ   8+y, DX;       XORL   AX,  AX; \
+    MULXQ  0+x, AX, R12;  ADCXQ  R8,  AX;  MOVQ  AX,  8+z; \
+    MULXQ  8+x, AX, R13;  ADCXQ  R9, R12;  ADOXQ AX, R12; \
+    MULXQ 16+x, AX, R14;  ADCXQ R10, R13;  ADOXQ AX, R13; \
+    MULXQ 24+x, AX, R15;  ADCXQ R11, R14;  ADOXQ AX, R14; \
+    MOVL $0, AX;;;;;;;;;  ADCXQ  AX, R15;  ADOXQ AX, R15; \
+    MOVQ  16+y, DX;       XORL   AX,  AX; \
+    MULXQ  0+x, AX,  R8;  ADCXQ R12,  AX;  MOVQ  AX, 16+z; \
+    MULXQ  8+x, AX,  R9;  ADCXQ R13,  R8;  ADOXQ AX,  R8; \
+    MULXQ 16+x, AX, R10;  ADCXQ R14,  R9;  ADOXQ AX,  R9; \
+    MULXQ 24+x, AX, R11;  ADCXQ R15, R10;  ADOXQ AX, R10; \
+    MOVL $0, AX;;;;;;;;;  ADCXQ  AX, R11;  ADOXQ AX, R11; \
+    MOVQ  24+y, DX;       XORL   AX,  AX; \
+    MULXQ  0+x, AX, R12;  ADCXQ  R8,  AX;  MOVQ  AX, 24+z; \
+    MULXQ  8+x, AX, R13;  ADCXQ  R9, R12;  ADOXQ AX, R12;  MOVQ R12, 32+z; \
+    MULXQ 16+x, AX, R14;  ADCXQ R10, R13;  ADOXQ AX, R13;  MOVQ R13, 40+z; \
+    MULXQ 24+x, AX, R15;  ADCXQ R11, R14;  ADOXQ AX, R14;  MOVQ R14, 48+z; \
+    MOVL $0, AX;;;;;;;;;  ADCXQ  AX, R15;  ADOXQ AX, R15;  MOVQ R15, 56+z;
+
+// integerMulLeg multiplies x and y and stores in z
+// Uses: AX, DX, R8-R15, FLAGS
+// Instr: x86_64
+#define integerMulLeg(z,x,y) \
+    MOVQ  0+y, R8; \
+    MOVQ  0+x, AX; MULQ R8; MOVQ AX, 0+z; MOVQ DX, R15; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX, R13; MOVQ DX, R10; \
+    MOVQ 16+x, AX; MULQ R8; MOVQ AX, R14; MOVQ DX, R11; \
+    MOVQ 24+x, AX; MULQ R8; \
+    ADDQ R13, R15; \
+    ADCQ R14, R10;  MOVQ R10, 16+z; \
+    ADCQ  AX, R11;  MOVQ R11, 24+z; \
+    ADCQ  $0,  DX;  MOVQ  DX, 32+z; \
+    MOVQ  8+y, R8; \
+    MOVQ  0+x, AX; MULQ R8; MOVQ AX, R12; MOVQ DX,  R9; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX, R13; MOVQ DX, R10; \
+    MOVQ 16+x, AX; MULQ R8; MOVQ AX, R14; MOVQ DX, R11; \
+    MOVQ 24+x, AX; MULQ R8; \
+    ADDQ R12, R15; MOVQ R15,  8+z; \
+    ADCQ R13,  R9; \
+    ADCQ R14, R10; \
+    ADCQ  AX, R11; \
+    ADCQ  $0,  DX; \
+    ADCQ 16+z,  R9;  MOVQ  R9,  R15; \
+    ADCQ 24+z, R10;  MOVQ R10, 24+z; \
+    ADCQ 32+z, R11;  MOVQ R11, 32+z; \
+    ADCQ   $0,  DX;  MOVQ  DX, 40+z; \
+    MOVQ 16+y, R8; \
+    MOVQ  0+x, AX; MULQ R8; MOVQ AX, R12; MOVQ DX,  R9; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX, R13; MOVQ DX, R10; \
+    MOVQ 16+x, AX; MULQ R8; MOVQ AX, R14; MOVQ DX, R11; \
+    MOVQ 24+x, AX; MULQ R8; \
+    ADDQ R12, R15;  MOVQ R15, 16+z; \
+    ADCQ R13,  R9; \
+    ADCQ R14, R10; \
+    ADCQ  AX, R11; \
+    ADCQ  $0,  DX; \
+    ADCQ 24+z,  R9;  MOVQ  R9,  R15; \
+    ADCQ 32+z, R10;  MOVQ R10, 32+z; \
+    ADCQ 40+z, R11;  MOVQ R11, 40+z; \
+    ADCQ   $0,  DX;  MOVQ  DX, 48+z; \
+    MOVQ 24+y, R8; \
+    MOVQ  0+x, AX; MULQ R8; MOVQ AX, R12; MOVQ DX,  R9; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX, R13; MOVQ DX, R10; \
+    MOVQ 16+x, AX; MULQ R8; MOVQ AX, R14; MOVQ DX, R11; \
+    MOVQ 24+x, AX; MULQ R8; \
+    ADDQ R12, R15; MOVQ R15, 24+z; \
+    ADCQ R13,  R9; \
+    ADCQ R14, R10; \
+    ADCQ  AX, R11; \
+    ADCQ  $0,  DX; \
+    ADCQ 32+z,  R9;  MOVQ  R9, 32+z; \
+    ADCQ 40+z, R10;  MOVQ R10, 40+z; \
+    ADCQ 48+z, R11;  MOVQ R11, 48+z; \
+    ADCQ   $0,  DX;  MOVQ  DX, 56+z;
+
+// integerSqrLeg squares x and stores in z
+// Uses: AX, CX, DX, R8-R15, FLAGS
+// Instr: x86_64
+#define integerSqrLeg(z,x) \
+    MOVQ  0+x, R8; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX,  R9; MOVQ DX, R10; /* A[0]*A[1] */ \
+    MOVQ 16+x, AX; MULQ R8; MOVQ AX, R14; MOVQ DX, R11; /* A[0]*A[2] */ \
+    MOVQ 24+x, AX; MULQ R8; MOVQ AX, R15; MOVQ DX, R12; /* A[0]*A[3] */ \
+    MOVQ 24+x, R8; \
+    MOVQ  8+x, AX; MULQ R8; MOVQ AX,  CX; MOVQ DX, R13; /* A[3]*A[1] */ \
+    MOVQ 16+x, AX; MULQ R8; /* A[3]*A[2] */ \
+    \
+    ADDQ R14, R10;\
+    ADCQ R15, R11; MOVL $0, R15;\
+    ADCQ  CX, R12;\
+    ADCQ  AX, R13;\
+    ADCQ  $0,  DX; MOVQ DX, R14;\
+    MOVQ 8+x, AX; MULQ 16+x;\
+    \
+    ADDQ AX, R11;\
+    ADCQ DX, R12;\
+    ADCQ $0, R13;\
+    ADCQ $0, R14;\
+    ADCQ $0, R15;\
+    \
+    SHLQ $1, R14, R15; MOVQ R15, 56+z;\
+    SHLQ $1, R13, R14; MOVQ R14, 48+z;\
+    SHLQ $1, R12, R13; MOVQ R13, 40+z;\
+    SHLQ $1, R11, R12; MOVQ R12, 32+z;\
+    SHLQ $1, R10, R11; MOVQ R11, 24+z;\
+    SHLQ $1,  R9, R10; MOVQ R10, 16+z;\
+    SHLQ $1,  R9;      MOVQ  R9,  8+z;\
+    \
+    MOVQ  0+x,AX; MULQ AX; MOVQ AX, 0+z; MOVQ DX,  R9;\
+    MOVQ  8+x,AX; MULQ AX; MOVQ AX, R10; MOVQ DX, R11;\
+    MOVQ 16+x,AX; MULQ AX; MOVQ AX, R12; MOVQ DX, R13;\
+    MOVQ 24+x,AX; MULQ AX; MOVQ AX, R14; MOVQ DX, R15;\
+    \
+    ADDQ  8+z,  R9; MOVQ  R9,  8+z;\
+    ADCQ 16+z, R10; MOVQ R10, 16+z;\
+    ADCQ 24+z, R11; MOVQ R11, 24+z;\
+    ADCQ 32+z, R12; MOVQ R12, 32+z;\
+    ADCQ 40+z, R13; MOVQ R13, 40+z;\
+    ADCQ 48+z, R14; MOVQ R14, 48+z;\
+    ADCQ 56+z, R15; MOVQ R15, 56+z;
+
+// integerSqrAdx squares x and stores in z
+// Uses: AX, CX, DX, R8-R15, FLAGS
+// Instr: x86_64, bmi2, adx
+#define integerSqrAdx(z,x) \
+    MOVQ   0+x,  DX; /* A[0] */ \
+    MULXQ  8+x,  R8, R14; /* A[1]*A[0] */  XORL  R15, R15; \
+    MULXQ 16+x,  R9, R10; /* A[2]*A[0] */  ADCXQ R14,  R9; \
+    MULXQ 24+x,  AX,  CX; /* A[3]*A[0] */  ADCXQ  AX, R10; \
+    MOVQ  24+x,  DX; /* A[3] */ \
+    MULXQ  8+x, R11, R12; /* A[1]*A[3] */  ADCXQ  CX, R11; \
+    MULXQ 16+x,  AX, R13; /* A[2]*A[3] */  ADCXQ  AX, R12; \
+    MOVQ   8+x,  DX; /* A[1] */            ADCXQ R15, R13; \
+    MULXQ 16+x,  AX,  CX; /* A[2]*A[1] */  MOVL   $0, R14; \
+    ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;  ADCXQ R15, R14; \
+    XORL  R15, R15; \
+    ADOXQ  AX, R10;  ADCXQ  R8,  R8; \
+    ADOXQ  CX, R11;  ADCXQ  R9,  R9; \
+    ADOXQ R15, R12;  ADCXQ R10, R10; \
+    ADOXQ R15, R13;  ADCXQ R11, R11; \
+    ADOXQ R15, R14;  ADCXQ R12, R12; \
+    ;;;;;;;;;;;;;;;  ADCXQ R13, R13; \
+    ;;;;;;;;;;;;;;;  ADCXQ R14, R14; \
+    MOVQ  0+x, DX;  MULXQ DX, AX, CX; /* A[0]^2 */ \
+    ;;;;;;;;;;;;;;;  MOVQ  AX,  0+z; \
+    ADDQ CX,  R8;    MOVQ  R8,  8+z; \
+    MOVQ  8+x, DX;  MULXQ DX, AX, CX; /* A[1]^2 */ \
+    ADCQ AX,  R9;    MOVQ  R9, 16+z; \
+    ADCQ CX, R10;    MOVQ R10, 24+z; \
+    MOVQ 16+x, DX;  MULXQ DX, AX, CX; /* A[2]^2 */ \
+    ADCQ AX, R11;    MOVQ R11, 32+z; \
+    ADCQ CX, R12;    MOVQ R12, 40+z; \
+    MOVQ 24+x, DX;  MULXQ DX, AX, CX; /* A[3]^2 */ \
+    ADCQ AX, R13;    MOVQ R13, 48+z; \
+    ADCQ CX, R14;    MOVQ R14, 56+z;
+
+// reduceFromDouble finds z congruent to x modulo p such that 0<z<2^256
+// Uses: AX, DX, R8-R13, FLAGS
+// Instr: x86_64
+#define reduceFromDoubleLeg(z,x) \
+    /* 2*C = 38 = 2^256 */ \
+    MOVL $38, AX; MULQ 32+x; MOVQ AX,  R8; MOVQ DX,  R9; /* C*C[4] */ \
+    MOVL $38, AX; MULQ 40+x; MOVQ AX, R12; MOVQ DX, R10; /* C*C[5] */ \
+    MOVL $38, AX; MULQ 48+x; MOVQ AX, R13; MOVQ DX, R11; /* C*C[6] */ \
+    MOVL $38, AX; MULQ 56+x; /* C*C[7] */ \
+    ADDQ R12,  R9; \
+    ADCQ R13, R10; \
+    ADCQ  AX, R11; \
+    ADCQ  $0,  DX; \
+    ADDQ  0+x,  R8; \
+    ADCQ  8+x,  R9; \
+    ADCQ 16+x, R10; \
+    ADCQ 24+x, R11; \
+    ADCQ    $0, DX; \
+    MOVL $38, AX; \
+    IMULQ AX, DX; /* C*C[4], CF=0, OF=0 */ \
+    ADDQ DX,  R8; \
+    ADCQ $0,  R9; MOVQ  R9,  8+z; \
+    ADCQ $0, R10; MOVQ R10, 16+z; \
+    ADCQ $0, R11; MOVQ R11, 24+z; \
+    MOVL $0,  DX; \
+    CMOVQCS AX, DX; \
+    ADDQ DX,  R8; MOVQ  R8,  0+z;
+
+// reduceFromDoubleAdx finds z congruent to x modulo p such that 0<z<2^256
+// Uses: AX, DX, R8-R13, FLAGS
+// Instr: x86_64, bmi2, adx
+#define reduceFromDoubleAdx(z,x) \
+    MOVL    $38,  DX; /* 2*C = 38 = 2^256 */ \
+    MULXQ 32+x,  R8, R10; /* C*C[4] */  XORL AX, AX;     ADOXQ  0+x,  R8; \
+    MULXQ 40+x,  R9, R11; /* C*C[5] */  ADCXQ R10,  R9;  ADOXQ  8+x,  R9; \
+    MULXQ 48+x, R10, R13; /* C*C[6] */  ADCXQ R11, R10;  ADOXQ 16+x, R10; \
+    MULXQ 56+x, R11, R12; /* C*C[7] */  ADCXQ R13, R11;  ADOXQ 24+x, R11; \
+    ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;  ADCXQ  AX, R12;  ADOXQ   AX, R12; \
+    IMULQ  DX, R12; /* C*C[4], CF=0, OF=0 */ \
+    ADCXQ R12, R8; \
+    ADCXQ AX,  R9; MOVQ  R9,  8+z; \
+    ADCXQ AX, R10; MOVQ R10, 16+z; \
+    ADCXQ AX, R11; MOVQ R11, 24+z; \
+    MOVL  $0, R12; \
+    CMOVQCS DX, R12; \
+    ADDQ R12,  R8; MOVQ  R8,  0+z;
+
+// addSub calculates two operations: x,y = x+y,x-y
+// Uses: AX, DX, R8-R15, FLAGS
+#define addSub(x,y) \
+    MOVL $38, AX; \
+    XORL  DX, DX; \
+    MOVQ  0+x,  R8;  MOVQ  R8, R12;  ADDQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  MOVQ  R9, R13;  ADCQ  8+y,  R9; \
+    MOVQ 16+x, R10;  MOVQ R10, R14;  ADCQ 16+y, R10; \
+    MOVQ 24+x, R11;  MOVQ R11, R15;  ADCQ 24+y, R11; \
+    CMOVQCS AX, DX; \
+    XORL AX,  AX; \
+    ADDQ DX,  R8; \
+    ADCQ $0,  R9; \
+    ADCQ $0, R10; \
+    ADCQ $0, R11; \
+    MOVL $38, DX; \
+    CMOVQCS DX, AX; \
+    ADDQ  AX, R8; \
+    MOVL $38, AX; \
+    SUBQ  0+y, R12; \
+    SBBQ  8+y, R13; \
+    SBBQ 16+y, R14; \
+    SBBQ 24+y, R15; \
+    MOVL $0, DX; \
+    CMOVQCS AX, DX; \
+    SUBQ DX, R12; \
+    SBBQ $0, R13; \
+    SBBQ $0, R14; \
+    SBBQ $0, R15; \
+    MOVL $0,  DX; \
+    CMOVQCS AX, DX; \
+    SUBQ DX, R12; \
+    MOVQ  R8,  0+x; \
+    MOVQ  R9,  8+x; \
+    MOVQ R10, 16+x; \
+    MOVQ R11, 24+x; \
+    MOVQ R12,  0+y; \
+    MOVQ R13,  8+y; \
+    MOVQ R14, 16+y; \
+    MOVQ R15, 24+y;

source/vendor/github.com/cloudflare/circl/math/fp25519/fp_amd64.s

+//go:build amd64 && !purego
+// +build amd64,!purego
+
+#include "textflag.h"
+#include "fp_amd64.h"
+
+// func cmovAmd64(x, y *Elt, n uint)
+TEXT ·cmovAmd64(SB),NOSPLIT,$0-24
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    MOVQ n+16(FP), BX
+    cselect(0(DI),0(SI),BX)
+    RET
+
+// func cswapAmd64(x, y *Elt, n uint)
+TEXT ·cswapAmd64(SB),NOSPLIT,$0-24
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    MOVQ n+16(FP), BX
+    cswap(0(DI),0(SI),BX)
+    RET
+
+// func subAmd64(z, x, y *Elt)
+TEXT ·subAmd64(SB),NOSPLIT,$0-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    subtraction(0(DI),0(SI),0(BX))
+    RET
+
+// func addsubAmd64(x, y *Elt)
+TEXT ·addsubAmd64(SB),NOSPLIT,$0-16
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    addSub(0(DI),0(SI))
+    RET
+
+#define addLegacy \
+    additionLeg(0(DI),0(SI),0(BX))
+#define addBmi2Adx \
+    additionAdx(0(DI),0(SI),0(BX))
+
+#define mulLegacy \
+    integerMulLeg(0(SP),0(SI),0(BX)) \
+    reduceFromDoubleLeg(0(DI),0(SP))
+#define mulBmi2Adx \
+    integerMulAdx(0(SP),0(SI),0(BX)) \
+    reduceFromDoubleAdx(0(DI),0(SP))
+
+#define sqrLegacy \
+    integerSqrLeg(0(SP),0(SI)) \
+    reduceFromDoubleLeg(0(DI),0(SP))
+#define sqrBmi2Adx \
+    integerSqrAdx(0(SP),0(SI)) \
+    reduceFromDoubleAdx(0(DI),0(SP))
+
+// func addAmd64(z, x, y *Elt)
+TEXT ·addAmd64(SB),NOSPLIT,$0-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    CHECK_BMI2ADX(LADD, addLegacy, addBmi2Adx)
+
+// func mulAmd64(z, x, y *Elt)
+TEXT ·mulAmd64(SB),NOSPLIT,$64-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    CHECK_BMI2ADX(LMUL, mulLegacy, mulBmi2Adx)
+
+// func sqrAmd64(z, x *Elt)
+TEXT ·sqrAmd64(SB),NOSPLIT,$64-16
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    CHECK_BMI2ADX(LSQR, sqrLegacy, sqrBmi2Adx)
+
+// func modpAmd64(z *Elt)
+TEXT ·modpAmd64(SB),NOSPLIT,$0-8
+    MOVQ z+0(FP), DI
+
+    MOVQ   (DI),  R8
+    MOVQ  8(DI),  R9
+    MOVQ 16(DI), R10
+    MOVQ 24(DI), R11
+
+    MOVL $19, AX
+    MOVL $38, CX
+
+    BTRQ $63, R11 // PUT BIT 255 IN CARRY FLAG AND CLEAR
+    CMOVLCC AX, CX // C[255] ? 38 : 19
+
+    // ADD EITHER 19 OR 38 TO C
+    ADDQ CX,  R8
+    ADCQ $0,  R9
+    ADCQ $0, R10
+    ADCQ $0, R11
+
+    // TEST FOR BIT 255 AGAIN; ONLY TRIGGERED ON OVERFLOW MODULO 2^255-19
+    MOVL     $0,  CX
+    CMOVLPL  AX,  CX // C[255] ? 0 : 19
+    BTRQ    $63, R11 // CLEAR BIT 255
+
+    // SUBTRACT 19 IF NECESSARY
+    SUBQ CX,  R8
+    MOVQ  R8,   (DI)
+    SBBQ $0,  R9
+    MOVQ  R9,  8(DI)
+    SBBQ $0, R10
+    MOVQ R10, 16(DI)
+    SBBQ $0, R11
+    MOVQ R11, 24(DI)
+    RET

source/vendor/github.com/cloudflare/circl/math/fp25519/fp_generic.go

+package fp25519
+
+import (
+	"encoding/binary"
+	"math/bits"
+)
+
+func cmovGeneric(x, y *Elt, n uint) {
+	m := -uint64(n & 0x1)
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+
+	x0 = (x0 &^ m) | (y0 & m)
+	x1 = (x1 &^ m) | (y1 & m)
+	x2 = (x2 &^ m) | (y2 & m)
+	x3 = (x3 &^ m) | (y3 & m)
+
+	binary.LittleEndian.PutUint64(x[0*8:1*8], x0)
+	binary.LittleEndian.PutUint64(x[1*8:2*8], x1)
+	binary.LittleEndian.PutUint64(x[2*8:3*8], x2)
+	binary.LittleEndian.PutUint64(x[3*8:4*8], x3)
+}
+
+func cswapGeneric(x, y *Elt, n uint) {
+	m := -uint64(n & 0x1)
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+
+	t0 := m & (x0 ^ y0)
+	t1 := m & (x1 ^ y1)
+	t2 := m & (x2 ^ y2)
+	t3 := m & (x3 ^ y3)
+	x0 ^= t0
+	x1 ^= t1
+	x2 ^= t2
+	x3 ^= t3
+	y0 ^= t0
+	y1 ^= t1
+	y2 ^= t2
+	y3 ^= t3
+
+	binary.LittleEndian.PutUint64(x[0*8:1*8], x0)
+	binary.LittleEndian.PutUint64(x[1*8:2*8], x1)
+	binary.LittleEndian.PutUint64(x[2*8:3*8], x2)
+	binary.LittleEndian.PutUint64(x[3*8:4*8], x3)
+
+	binary.LittleEndian.PutUint64(y[0*8:1*8], y0)
+	binary.LittleEndian.PutUint64(y[1*8:2*8], y1)
+	binary.LittleEndian.PutUint64(y[2*8:3*8], y2)
+	binary.LittleEndian.PutUint64(y[3*8:4*8], y3)
+}
+
+func addGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+
+	z0, c0 := bits.Add64(x0, y0, 0)
+	z1, c1 := bits.Add64(x1, y1, c0)
+	z2, c2 := bits.Add64(x2, y2, c1)
+	z3, c3 := bits.Add64(x3, y3, c2)
+
+	z0, c0 = bits.Add64(z0, (-c3)&38, 0)
+	z1, c1 = bits.Add64(z1, 0, c0)
+	z2, c2 = bits.Add64(z2, 0, c1)
+	z3, c3 = bits.Add64(z3, 0, c2)
+	z0, _ = bits.Add64(z0, (-c3)&38, 0)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], z0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], z1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], z2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], z3)
+}
+
+func subGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+
+	z0, c0 := bits.Sub64(x0, y0, 0)
+	z1, c1 := bits.Sub64(x1, y1, c0)
+	z2, c2 := bits.Sub64(x2, y2, c1)
+	z3, c3 := bits.Sub64(x3, y3, c2)
+
+	z0, c0 = bits.Sub64(z0, (-c3)&38, 0)
+	z1, c1 = bits.Sub64(z1, 0, c0)
+	z2, c2 = bits.Sub64(z2, 0, c1)
+	z3, c3 = bits.Sub64(z3, 0, c2)
+	z0, _ = bits.Sub64(z0, (-c3)&38, 0)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], z0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], z1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], z2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], z3)
+}
+
+func addsubGeneric(x, y *Elt) {
+	z := &Elt{}
+	addGeneric(z, x, y)
+	subGeneric(y, x, y)
+	*x = *z
+}
+
+func mulGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+
+	yi := y0
+	h0, l0 := bits.Mul64(x0, yi)
+	h1, l1 := bits.Mul64(x1, yi)
+	h2, l2 := bits.Mul64(x2, yi)
+	h3, l3 := bits.Mul64(x3, yi)
+
+	z0 := l0
+	a0, c0 := bits.Add64(h0, l1, 0)
+	a1, c1 := bits.Add64(h1, l2, c0)
+	a2, c2 := bits.Add64(h2, l3, c1)
+	a3, _ := bits.Add64(h3, 0, c2)
+
+	yi = y1
+	h0, l0 = bits.Mul64(x0, yi)
+	h1, l1 = bits.Mul64(x1, yi)
+	h2, l2 = bits.Mul64(x2, yi)
+	h3, l3 = bits.Mul64(x3, yi)
+
+	z1, c0 := bits.Add64(a0, l0, 0)
+	h0, c1 = bits.Add64(h0, l1, c0)
+	h1, c2 = bits.Add64(h1, l2, c1)
+	h2, c3 := bits.Add64(h2, l3, c2)
+	h3, _ = bits.Add64(h3, 0, c3)
+
+	a0, c0 = bits.Add64(a1, h0, 0)
+	a1, c1 = bits.Add64(a2, h1, c0)
+	a2, c2 = bits.Add64(a3, h2, c1)
+	a3, _ = bits.Add64(0, h3, c2)
+
+	yi = y2
+	h0, l0 = bits.Mul64(x0, yi)
+	h1, l1 = bits.Mul64(x1, yi)
+	h2, l2 = bits.Mul64(x2, yi)
+	h3, l3 = bits.Mul64(x3, yi)
+
+	z2, c0 := bits.Add64(a0, l0, 0)
+	h0, c1 = bits.Add64(h0, l1, c0)
+	h1, c2 = bits.Add64(h1, l2, c1)
+	h2, c3 = bits.Add64(h2, l3, c2)
+	h3, _ = bits.Add64(h3, 0, c3)
+
+	a0, c0 = bits.Add64(a1, h0, 0)
+	a1, c1 = bits.Add64(a2, h1, c0)
+	a2, c2 = bits.Add64(a3, h2, c1)
+	a3, _ = bits.Add64(0, h3, c2)
+
+	yi = y3
+	h0, l0 = bits.Mul64(x0, yi)
+	h1, l1 = bits.Mul64(x1, yi)
+	h2, l2 = bits.Mul64(x2, yi)
+	h3, l3 = bits.Mul64(x3, yi)
+
+	z3, c0 := bits.Add64(a0, l0, 0)
+	h0, c1 = bits.Add64(h0, l1, c0)
+	h1, c2 = bits.Add64(h1, l2, c1)
+	h2, c3 = bits.Add64(h2, l3, c2)
+	h3, _ = bits.Add64(h3, 0, c3)
+
+	z4, c0 := bits.Add64(a1, h0, 0)
+	z5, c1 := bits.Add64(a2, h1, c0)
+	z6, c2 := bits.Add64(a3, h2, c1)
+	z7, _ := bits.Add64(0, h3, c2)
+
+	red64(z, z0, z1, z2, z3, z4, z5, z6, z7)
+}
+
+func sqrGeneric(z, x *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	h0, a0 := bits.Mul64(x0, x1)
+	h1, l1 := bits.Mul64(x0, x2)
+	h2, l2 := bits.Mul64(x0, x3)
+	h3, l3 := bits.Mul64(x3, x1)
+	h4, l4 := bits.Mul64(x3, x2)
+	h, l := bits.Mul64(x1, x2)
+
+	a1, c0 := bits.Add64(l1, h0, 0)
+	a2, c1 := bits.Add64(l2, h1, c0)
+	a3, c2 := bits.Add64(l3, h2, c1)
+	a4, c3 := bits.Add64(l4, h3, c2)
+	a5, _ := bits.Add64(h4, 0, c3)
+
+	a2, c0 = bits.Add64(a2, l, 0)
+	a3, c1 = bits.Add64(a3, h, c0)
+	a4, c2 = bits.Add64(a4, 0, c1)
+	a5, c3 = bits.Add64(a5, 0, c2)
+	a6, _ := bits.Add64(0, 0, c3)
+
+	a0, c0 = bits.Add64(a0, a0, 0)
+	a1, c1 = bits.Add64(a1, a1, c0)
+	a2, c2 = bits.Add64(a2, a2, c1)
+	a3, c3 = bits.Add64(a3, a3, c2)
+	a4, c4 := bits.Add64(a4, a4, c3)
+	a5, c5 := bits.Add64(a5, a5, c4)
+	a6, _ = bits.Add64(a6, a6, c5)
+
+	b1, b0 := bits.Mul64(x0, x0)
+	b3, b2 := bits.Mul64(x1, x1)
+	b5, b4 := bits.Mul64(x2, x2)
+	b7, b6 := bits.Mul64(x3, x3)
+
+	b1, c0 = bits.Add64(b1, a0, 0)
+	b2, c1 = bits.Add64(b2, a1, c0)
+	b3, c2 = bits.Add64(b3, a2, c1)
+	b4, c3 = bits.Add64(b4, a3, c2)
+	b5, c4 = bits.Add64(b5, a4, c3)
+	b6, c5 = bits.Add64(b6, a5, c4)
+	b7, _ = bits.Add64(b7, a6, c5)
+
+	red64(z, b0, b1, b2, b3, b4, b5, b6, b7)
+}
+
+func modpGeneric(x *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+
+	// CX = C[255] ? 38 : 19
+	cx := uint64(19) << (x3 >> 63)
+	// PUT BIT 255 IN CARRY FLAG AND CLEAR
+	x3 &^= 1 << 63
+
+	x0, c0 := bits.Add64(x0, cx, 0)
+	x1, c1 := bits.Add64(x1, 0, c0)
+	x2, c2 := bits.Add64(x2, 0, c1)
+	x3, _ = bits.Add64(x3, 0, c2)
+
+	// TEST FOR BIT 255 AGAIN; ONLY TRIGGERED ON OVERFLOW MODULO 2^255-19
+	// cx = C[255] ? 0 : 19
+	cx = uint64(19) &^ (-(x3 >> 63))
+	// CLEAR BIT 255
+	x3 &^= 1 << 63
+
+	x0, c0 = bits.Sub64(x0, cx, 0)
+	x1, c1 = bits.Sub64(x1, 0, c0)
+	x2, c2 = bits.Sub64(x2, 0, c1)
+	x3, _ = bits.Sub64(x3, 0, c2)
+
+	binary.LittleEndian.PutUint64(x[0*8:1*8], x0)
+	binary.LittleEndian.PutUint64(x[1*8:2*8], x1)
+	binary.LittleEndian.PutUint64(x[2*8:3*8], x2)
+	binary.LittleEndian.PutUint64(x[3*8:4*8], x3)
+}
+
+func red64(z *Elt, x0, x1, x2, x3, x4, x5, x6, x7 uint64) {
+	h0, l0 := bits.Mul64(x4, 38)
+	h1, l1 := bits.Mul64(x5, 38)
+	h2, l2 := bits.Mul64(x6, 38)
+	h3, l3 := bits.Mul64(x7, 38)
+
+	l1, c0 := bits.Add64(h0, l1, 0)
+	l2, c1 := bits.Add64(h1, l2, c0)
+	l3, c2 := bits.Add64(h2, l3, c1)
+	l4, _ := bits.Add64(h3, 0, c2)
+
+	l0, c0 = bits.Add64(l0, x0, 0)
+	l1, c1 = bits.Add64(l1, x1, c0)
+	l2, c2 = bits.Add64(l2, x2, c1)
+	l3, c3 := bits.Add64(l3, x3, c2)
+	l4, _ = bits.Add64(l4, 0, c3)
+
+	_, l4 = bits.Mul64(l4, 38)
+	l0, c0 = bits.Add64(l0, l4, 0)
+	z1, c1 := bits.Add64(l1, 0, c0)
+	z2, c2 := bits.Add64(l2, 0, c1)
+	z3, c3 := bits.Add64(l3, 0, c2)
+	z0, _ := bits.Add64(l0, (-c3)&38, 0)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], z0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], z1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], z2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], z3)
+}

source/vendor/github.com/cloudflare/circl/math/fp25519/fp_noasm.go

+//go:build !amd64 || purego
+// +build !amd64 purego
+
+package fp25519
+
+func cmov(x, y *Elt, n uint)  { cmovGeneric(x, y, n) }
+func cswap(x, y *Elt, n uint) { cswapGeneric(x, y, n) }
+func add(z, x, y *Elt)        { addGeneric(z, x, y) }
+func sub(z, x, y *Elt)        { subGeneric(z, x, y) }
+func addsub(x, y *Elt)        { addsubGeneric(x, y) }
+func mul(z, x, y *Elt)        { mulGeneric(z, x, y) }
+func sqr(z, x *Elt)           { sqrGeneric(z, x) }
+func modp(z *Elt)             { modpGeneric(z) }

source/vendor/github.com/cloudflare/circl/math/fp448/fp.go

+// Package fp448 provides prime field arithmetic over GF(2^448-2^224-1).
+package fp448
+
+import (
+	"errors"
+
+	"github.com/cloudflare/circl/internal/conv"
+)
+
+// Size in bytes of an element.
+const Size = 56
+
+// Elt is a prime field element.
+type Elt [Size]byte
+
+func (e Elt) String() string { return conv.BytesLe2Hex(e[:]) }
+
+// p is the prime modulus 2^448-2^224-1.
+var p = Elt{
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xfe, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+}
+
+// P returns the prime modulus 2^448-2^224-1.
+func P() Elt { return p }
+
+// ToBytes stores in b the little-endian byte representation of x.
+func ToBytes(b []byte, x *Elt) error {
+	if len(b) != Size {
+		return errors.New("wrong size")
+	}
+	Modp(x)
+	copy(b, x[:])
+	return nil
+}
+
+// IsZero returns true if x is equal to 0.
+func IsZero(x *Elt) bool { Modp(x); return *x == Elt{} }
+
+// IsOne returns true if x is equal to 1.
+func IsOne(x *Elt) bool { Modp(x); return *x == Elt{1} }
+
+// SetOne assigns x=1.
+func SetOne(x *Elt) { *x = Elt{1} }
+
+// One returns the 1 element.
+func One() (x Elt) { x = Elt{1}; return }
+
+// Neg calculates z = -x.
+func Neg(z, x *Elt) { Sub(z, &p, x) }
+
+// Modp ensures that z is between [0,p-1].
+func Modp(z *Elt) { Sub(z, z, &p) }
+
+// InvSqrt calculates z = sqrt(x/y) iff x/y is a quadratic-residue. If so,
+// isQR = true; otherwise, isQR = false, since x/y is a quadratic non-residue,
+// and z = sqrt(-x/y).
+func InvSqrt(z, x, y *Elt) (isQR bool) {
+	// First note that x^(2(k+1)) = x^(p-1)/2 * x = legendre(x) * x
+	// so that's x if x is a quadratic residue and -x otherwise.
+	// Next, y^(6k+3) = y^(4k+2) * y^(2k+1) = y^(p-1) * y^((p-1)/2) = legendre(y).
+	// So the z we compute satisfies z^2 y = x^(2(k+1)) y^(6k+3) = legendre(x)*legendre(y).
+	// Thus if x and y are quadratic residues, then z is indeed sqrt(x/y).
+	t0, t1 := &Elt{}, &Elt{}
+	Mul(t0, x, y)         // x*y
+	Sqr(t1, y)            // y^2
+	Mul(t1, t0, t1)       // x*y^3
+	powPminus3div4(z, t1) // (x*y^3)^k
+	Mul(z, z, t0)         // z = x*y*(x*y^3)^k = x^(k+1) * y^(3k+1)
+
+	// Check if x/y is a quadratic residue
+	Sqr(t0, z)     // z^2
+	Mul(t0, t0, y) // y*z^2
+	Sub(t0, t0, x) // y*z^2-x
+	return IsZero(t0)
+}
+
+// Inv calculates z = 1/x mod p.
+func Inv(z, x *Elt) {
+	// Calculates z = x^(4k+1) = x^(p-3+1) = x^(p-2) = x^-1, where k = (p-3)/4.
+	t := &Elt{}
+	powPminus3div4(t, x) // t = x^k
+	Sqr(t, t)            // t = x^2k
+	Sqr(t, t)            // t = x^4k
+	Mul(z, t, x)         // z = x^(4k+1)
+}
+
+// powPminus3div4 calculates z = x^k mod p, where k = (p-3)/4.
+func powPminus3div4(z, x *Elt) {
+	x0, x1 := &Elt{}, &Elt{}
+	Sqr(z, x)
+	Mul(z, z, x)
+	Sqr(x0, z)
+	Mul(x0, x0, x)
+	Sqr(z, x0)
+	Sqr(z, z)
+	Sqr(z, z)
+	Mul(z, z, x0)
+	Sqr(x1, z)
+	for i := 0; i < 5; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, z)
+	Sqr(z, x1)
+	for i := 0; i < 11; i++ {
+		Sqr(z, z)
+	}
+	Mul(z, z, x1)
+	Sqr(z, z)
+	Sqr(z, z)
+	Sqr(z, z)
+	Mul(z, z, x0)
+	Sqr(x1, z)
+	for i := 0; i < 26; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, z)
+	Sqr(z, x1)
+	for i := 0; i < 53; i++ {
+		Sqr(z, z)
+	}
+	Mul(z, z, x1)
+	Sqr(z, z)
+	Sqr(z, z)
+	Sqr(z, z)
+	Mul(z, z, x0)
+	Sqr(x1, z)
+	for i := 0; i < 110; i++ {
+		Sqr(x1, x1)
+	}
+	Mul(x1, x1, z)
+	Sqr(z, x1)
+	Mul(z, z, x)
+	for i := 0; i < 223; i++ {
+		Sqr(z, z)
+	}
+	Mul(z, z, x1)
+}
+
+// Cmov assigns y to x if n is 1.
+func Cmov(x, y *Elt, n uint) { cmov(x, y, n) }
+
+// Cswap interchanges x and y if n is 1.
+func Cswap(x, y *Elt, n uint) { cswap(x, y, n) }
+
+// Add calculates z = x+y mod p.
+func Add(z, x, y *Elt) { add(z, x, y) }
+
+// Sub calculates z = x-y mod p.
+func Sub(z, x, y *Elt) { sub(z, x, y) }
+
+// AddSub calculates (x,y) = (x+y mod p, x-y mod p).
+func AddSub(x, y *Elt) { addsub(x, y) }
+
+// Mul calculates z = x*y mod p.
+func Mul(z, x, y *Elt) { mul(z, x, y) }
+
+// Sqr calculates z = x^2 mod p.
+func Sqr(z, x *Elt) { sqr(z, x) }

source/vendor/github.com/cloudflare/circl/math/fp448/fp_amd64.go

+//go:build amd64 && !purego
+// +build amd64,!purego
+
+package fp448
+
+import (
+	"golang.org/x/sys/cpu"
+)
+
+var hasBmi2Adx = cpu.X86.HasBMI2 && cpu.X86.HasADX
+
+var _ = hasBmi2Adx
+
+func cmov(x, y *Elt, n uint)  { cmovAmd64(x, y, n) }
+func cswap(x, y *Elt, n uint) { cswapAmd64(x, y, n) }
+func add(z, x, y *Elt)        { addAmd64(z, x, y) }
+func sub(z, x, y *Elt)        { subAmd64(z, x, y) }
+func addsub(x, y *Elt)        { addsubAmd64(x, y) }
+func mul(z, x, y *Elt)        { mulAmd64(z, x, y) }
+func sqr(z, x *Elt)           { sqrAmd64(z, x) }
+
+/* Functions defined in fp_amd64.s */
+
+//go:noescape
+func cmovAmd64(x, y *Elt, n uint)
+
+//go:noescape
+func cswapAmd64(x, y *Elt, n uint)
+
+//go:noescape
+func addAmd64(z, x, y *Elt)
+
+//go:noescape
+func subAmd64(z, x, y *Elt)
+
+//go:noescape
+func addsubAmd64(x, y *Elt)
+
+//go:noescape
+func mulAmd64(z, x, y *Elt)
+
+//go:noescape
+func sqrAmd64(z, x *Elt)

source/vendor/github.com/cloudflare/circl/math/fp448/fp_amd64.h

+// This code was imported from https://github.com/armfazh/rfc7748_precomputed
+
+// CHECK_BMI2ADX triggers bmi2adx if supported,
+// otherwise it fallbacks to legacy code.
+#define CHECK_BMI2ADX(label, legacy, bmi2adx) \
+    CMPB ·hasBmi2Adx(SB), $0  \
+    JE label                  \
+    bmi2adx                   \
+    RET                       \
+    label:                    \
+    legacy                    \
+    RET
+
+// cselect is a conditional move
+// if b=1: it copies y into x;
+// if b=0: x remains with the same value;
+// if b<> 0,1: undefined.
+// Uses: AX, DX, FLAGS
+// Instr: x86_64, cmov
+#define cselect(x,y,b) \
+    TESTQ b, b \
+    MOVQ  0+x, AX; MOVQ  0+y, DX; CMOVQNE DX, AX; MOVQ AX,  0+x; \
+    MOVQ  8+x, AX; MOVQ  8+y, DX; CMOVQNE DX, AX; MOVQ AX,  8+x; \
+    MOVQ 16+x, AX; MOVQ 16+y, DX; CMOVQNE DX, AX; MOVQ AX, 16+x; \
+    MOVQ 24+x, AX; MOVQ 24+y, DX; CMOVQNE DX, AX; MOVQ AX, 24+x; \
+    MOVQ 32+x, AX; MOVQ 32+y, DX; CMOVQNE DX, AX; MOVQ AX, 32+x; \
+    MOVQ 40+x, AX; MOVQ 40+y, DX; CMOVQNE DX, AX; MOVQ AX, 40+x; \
+    MOVQ 48+x, AX; MOVQ 48+y, DX; CMOVQNE DX, AX; MOVQ AX, 48+x;
+
+// cswap is a conditional swap
+// if b=1: x,y <- y,x;
+// if b=0: x,y remain with the same values;
+// if b<> 0,1: undefined.
+// Uses: AX, DX, R8, FLAGS
+// Instr: x86_64, cmov
+#define cswap(x,y,b) \
+    TESTQ b, b \
+    MOVQ  0+x, AX; MOVQ AX, R8; MOVQ  0+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX,  0+x; MOVQ DX,  0+y; \
+    MOVQ  8+x, AX; MOVQ AX, R8; MOVQ  8+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX,  8+x; MOVQ DX,  8+y; \
+    MOVQ 16+x, AX; MOVQ AX, R8; MOVQ 16+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 16+x; MOVQ DX, 16+y; \
+    MOVQ 24+x, AX; MOVQ AX, R8; MOVQ 24+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 24+x; MOVQ DX, 24+y; \
+    MOVQ 32+x, AX; MOVQ AX, R8; MOVQ 32+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 32+x; MOVQ DX, 32+y; \
+    MOVQ 40+x, AX; MOVQ AX, R8; MOVQ 40+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 40+x; MOVQ DX, 40+y; \
+    MOVQ 48+x, AX; MOVQ AX, R8; MOVQ 48+y, DX; CMOVQNE DX, AX; CMOVQNE R8, DX; MOVQ AX, 48+x; MOVQ DX, 48+y;
+
+// additionLeg adds x and y and stores in z
+// Uses: AX, DX, R8-R14, FLAGS
+// Instr: x86_64
+#define additionLeg(z,x,y) \
+    MOVQ  0+x,  R8;  ADDQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  ADCQ  8+y,  R9; \
+    MOVQ 16+x, R10;  ADCQ 16+y, R10; \
+    MOVQ 24+x, R11;  ADCQ 24+y, R11; \
+    MOVQ 32+x, R12;  ADCQ 32+y, R12; \
+    MOVQ 40+x, R13;  ADCQ 40+y, R13; \
+    MOVQ 48+x, R14;  ADCQ 48+y, R14; \
+    MOVQ   $0,  AX;  ADCQ   $0,  AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    ADDQ AX,  R8; MOVQ  $0, AX; \
+    ADCQ $0,  R9; \
+    ADCQ $0, R10; \
+    ADCQ DX, R11; \
+    ADCQ $0, R12; \
+    ADCQ $0, R13; \
+    ADCQ $0, R14; \
+    ADCQ $0,  AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    ADDQ AX,  R8;  MOVQ  R8,  0+z; \
+    ADCQ $0,  R9;  MOVQ  R9,  8+z; \
+    ADCQ $0, R10;  MOVQ R10, 16+z; \
+    ADCQ DX, R11;  MOVQ R11, 24+z; \
+    ADCQ $0, R12;  MOVQ R12, 32+z; \
+    ADCQ $0, R13;  MOVQ R13, 40+z; \
+    ADCQ $0, R14;  MOVQ R14, 48+z;
+
+
+// additionAdx adds x and y and stores in z
+// Uses: AX, DX, R8-R15, FLAGS
+// Instr: x86_64, adx
+#define additionAdx(z,x,y) \
+    MOVL $32, R15; \
+    XORL DX, DX; \
+    MOVQ  0+x,  R8;  ADCXQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  ADCXQ  8+y,  R9; \
+    MOVQ 16+x, R10;  ADCXQ 16+y, R10; \
+    MOVQ 24+x, R11;  ADCXQ 24+y, R11; \
+    MOVQ 32+x, R12;  ADCXQ 32+y, R12; \
+    MOVQ 40+x, R13;  ADCXQ 40+y, R13; \
+    MOVQ 48+x, R14;  ADCXQ 48+y, R14; \
+    ;;;;;;;;;;;;;;;  ADCXQ   DX,  DX; \
+    XORL AX, AX; \
+    ADCXQ DX,  R8; SHLXQ R15, DX, DX; \
+    ADCXQ AX,  R9; \
+    ADCXQ AX, R10; \
+    ADCXQ DX, R11; \
+    ADCXQ AX, R12; \
+    ADCXQ AX, R13; \
+    ADCXQ AX, R14; \
+    ADCXQ AX,  AX; \
+    XORL  DX,  DX; \
+    ADCXQ AX,  R8;  MOVQ  R8,  0+z; SHLXQ R15, AX, AX; \
+    ADCXQ DX,  R9;  MOVQ  R9,  8+z; \
+    ADCXQ DX, R10;  MOVQ R10, 16+z; \
+    ADCXQ AX, R11;  MOVQ R11, 24+z; \
+    ADCXQ DX, R12;  MOVQ R12, 32+z; \
+    ADCXQ DX, R13;  MOVQ R13, 40+z; \
+    ADCXQ DX, R14;  MOVQ R14, 48+z;
+
+// subtraction subtracts y from x and stores in z
+// Uses: AX, DX, R8-R14, FLAGS
+// Instr: x86_64
+#define subtraction(z,x,y) \
+    MOVQ  0+x,  R8;  SUBQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  SBBQ  8+y,  R9; \
+    MOVQ 16+x, R10;  SBBQ 16+y, R10; \
+    MOVQ 24+x, R11;  SBBQ 24+y, R11; \
+    MOVQ 32+x, R12;  SBBQ 32+y, R12; \
+    MOVQ 40+x, R13;  SBBQ 40+y, R13; \
+    MOVQ 48+x, R14;  SBBQ 48+y, R14; \
+    MOVQ   $0,  AX;  SETCS AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    SUBQ AX,  R8; MOVQ  $0, AX; \
+    SBBQ $0,  R9; \
+    SBBQ $0, R10; \
+    SBBQ DX, R11; \
+    SBBQ $0, R12; \
+    SBBQ $0, R13; \
+    SBBQ $0, R14; \
+    SETCS AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    SUBQ AX,  R8;  MOVQ  R8,  0+z; \
+    SBBQ $0,  R9;  MOVQ  R9,  8+z; \
+    SBBQ $0, R10;  MOVQ R10, 16+z; \
+    SBBQ DX, R11;  MOVQ R11, 24+z; \
+    SBBQ $0, R12;  MOVQ R12, 32+z; \
+    SBBQ $0, R13;  MOVQ R13, 40+z; \
+    SBBQ $0, R14;  MOVQ R14, 48+z;
+
+// maddBmi2Adx multiplies x and y and accumulates in z
+// Uses: AX, DX, R15, FLAGS
+// Instr: x86_64, bmi2, adx
+#define maddBmi2Adx(z,x,y,i,r0,r1,r2,r3,r4,r5,r6) \
+    MOVQ   i+y, DX; XORL AX, AX; \
+    MULXQ  0+x, AX, R8;  ADOXQ AX, r0;  ADCXQ R8, r1; MOVQ r0,i+z; \
+    MULXQ  8+x, AX, r0;  ADOXQ AX, r1;  ADCXQ r0, r2; MOVQ $0, R8; \
+    MULXQ 16+x, AX, r0;  ADOXQ AX, r2;  ADCXQ r0, r3; \
+    MULXQ 24+x, AX, r0;  ADOXQ AX, r3;  ADCXQ r0, r4; \
+    MULXQ 32+x, AX, r0;  ADOXQ AX, r4;  ADCXQ r0, r5; \
+    MULXQ 40+x, AX, r0;  ADOXQ AX, r5;  ADCXQ r0, r6; \
+    MULXQ 48+x, AX, r0;  ADOXQ AX, r6;  ADCXQ R8, r0; \
+    ;;;;;;;;;;;;;;;;;;;  ADOXQ R8, r0;
+
+// integerMulAdx multiplies x and y and stores in z
+// Uses: AX, DX, R8-R15, FLAGS
+// Instr: x86_64, bmi2, adx
+#define integerMulAdx(z,x,y) \
+    MOVL    $0,R15; \
+    MOVQ   0+y, DX;  XORL AX, AX;  MOVQ $0, R8; \
+    MULXQ  0+x, AX,  R9;  MOVQ  AX, 0+z; \
+    MULXQ  8+x, AX, R10;  ADCXQ AX,  R9; \
+    MULXQ 16+x, AX, R11;  ADCXQ AX, R10; \
+    MULXQ 24+x, AX, R12;  ADCXQ AX, R11; \
+    MULXQ 32+x, AX, R13;  ADCXQ AX, R12; \
+    MULXQ 40+x, AX, R14;  ADCXQ AX, R13; \
+    MULXQ 48+x, AX, R15;  ADCXQ AX, R14; \
+    ;;;;;;;;;;;;;;;;;;;;  ADCXQ R8, R15; \
+    maddBmi2Adx(z,x,y, 8, R9,R10,R11,R12,R13,R14,R15) \
+    maddBmi2Adx(z,x,y,16,R10,R11,R12,R13,R14,R15, R9) \
+    maddBmi2Adx(z,x,y,24,R11,R12,R13,R14,R15, R9,R10) \
+    maddBmi2Adx(z,x,y,32,R12,R13,R14,R15, R9,R10,R11) \
+    maddBmi2Adx(z,x,y,40,R13,R14,R15, R9,R10,R11,R12) \
+    maddBmi2Adx(z,x,y,48,R14,R15, R9,R10,R11,R12,R13) \
+    MOVQ R15,  56+z; \
+    MOVQ  R9,  64+z; \
+    MOVQ R10,  72+z; \
+    MOVQ R11,  80+z; \
+    MOVQ R12,  88+z; \
+    MOVQ R13,  96+z; \
+    MOVQ R14, 104+z;
+
+// maddLegacy multiplies x and y and accumulates in z
+// Uses: AX, DX, R15, FLAGS
+// Instr: x86_64
+#define maddLegacy(z,x,y,i) \
+    MOVQ  i+y, R15; \
+    MOVQ  0+x, AX; MULQ R15; MOVQ AX,  R8; ;;;;;;;;;;;; MOVQ DX,  R9; \
+    MOVQ  8+x, AX; MULQ R15; ADDQ AX,  R9; ADCQ $0, DX; MOVQ DX, R10; \
+    MOVQ 16+x, AX; MULQ R15; ADDQ AX, R10; ADCQ $0, DX; MOVQ DX, R11; \
+    MOVQ 24+x, AX; MULQ R15; ADDQ AX, R11; ADCQ $0, DX; MOVQ DX, R12; \
+    MOVQ 32+x, AX; MULQ R15; ADDQ AX, R12; ADCQ $0, DX; MOVQ DX, R13; \
+    MOVQ 40+x, AX; MULQ R15; ADDQ AX, R13; ADCQ $0, DX; MOVQ DX, R14; \
+    MOVQ 48+x, AX; MULQ R15; ADDQ AX, R14; ADCQ $0, DX; \
+    ADDQ  0+i+z,  R8; MOVQ  R8,  0+i+z; \
+    ADCQ  8+i+z,  R9; MOVQ  R9,  8+i+z; \
+    ADCQ 16+i+z, R10; MOVQ R10, 16+i+z; \
+    ADCQ 24+i+z, R11; MOVQ R11, 24+i+z; \
+    ADCQ 32+i+z, R12; MOVQ R12, 32+i+z; \
+    ADCQ 40+i+z, R13; MOVQ R13, 40+i+z; \
+    ADCQ 48+i+z, R14; MOVQ R14, 48+i+z; \
+    ADCQ     $0,  DX; MOVQ  DX, 56+i+z;
+
+// integerMulLeg multiplies x and y and stores in z
+// Uses: AX, DX, R8-R15, FLAGS
+// Instr: x86_64
+#define integerMulLeg(z,x,y) \
+    MOVQ  0+y, R15; \
+    MOVQ  0+x, AX; MULQ R15; MOVQ AX, 0+z; ;;;;;;;;;;;; MOVQ DX,  R8; \
+    MOVQ  8+x, AX; MULQ R15; ADDQ AX,  R8; ADCQ $0, DX; MOVQ DX,  R9; MOVQ  R8,  8+z; \
+    MOVQ 16+x, AX; MULQ R15; ADDQ AX,  R9; ADCQ $0, DX; MOVQ DX, R10; MOVQ  R9, 16+z; \
+    MOVQ 24+x, AX; MULQ R15; ADDQ AX, R10; ADCQ $0, DX; MOVQ DX, R11; MOVQ R10, 24+z; \
+    MOVQ 32+x, AX; MULQ R15; ADDQ AX, R11; ADCQ $0, DX; MOVQ DX, R12; MOVQ R11, 32+z; \
+    MOVQ 40+x, AX; MULQ R15; ADDQ AX, R12; ADCQ $0, DX; MOVQ DX, R13; MOVQ R12, 40+z; \
+    MOVQ 48+x, AX; MULQ R15; ADDQ AX, R13; ADCQ $0, DX; MOVQ DX,56+z; MOVQ R13, 48+z; \
+    maddLegacy(z,x,y, 8) \
+    maddLegacy(z,x,y,16) \
+    maddLegacy(z,x,y,24) \
+    maddLegacy(z,x,y,32) \
+    maddLegacy(z,x,y,40) \
+    maddLegacy(z,x,y,48)
+
+// integerSqrLeg squares x and stores in z
+// Uses: AX, CX, DX, R8-R15, FLAGS
+// Instr: x86_64
+#define integerSqrLeg(z,x) \
+    XORL R15, R15; \
+    MOVQ  0+x, CX; \
+    MOVQ   CX, AX; MULQ CX; MOVQ AX, 0+z; MOVQ DX, R8; \
+    ADDQ   CX, CX; ADCQ $0, R15; \
+    MOVQ  8+x, AX; MULQ CX; ADDQ AX,  R8; ADCQ $0, DX; MOVQ DX,  R9; MOVQ R8, 8+z; \
+    MOVQ 16+x, AX; MULQ CX; ADDQ AX,  R9; ADCQ $0, DX; MOVQ DX, R10; \
+    MOVQ 24+x, AX; MULQ CX; ADDQ AX, R10; ADCQ $0, DX; MOVQ DX, R11; \
+    MOVQ 32+x, AX; MULQ CX; ADDQ AX, R11; ADCQ $0, DX; MOVQ DX, R12; \
+    MOVQ 40+x, AX; MULQ CX; ADDQ AX, R12; ADCQ $0, DX; MOVQ DX, R13; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX, R13; ADCQ $0, DX; MOVQ DX, R14; \
+    \
+    MOVQ  8+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ  AX, R9; ADCQ $0, DX; MOVQ R9,16+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 8+x, AX; ADDQ AX, DX; ADCQ $0, R11; MOVQ DX, R8; \
+    ADDQ  8+x, CX; ADCQ $0, R15; \
+    MOVQ 16+x, AX; MULQ CX; ADDQ AX, R10; ADCQ $0, DX; ADDQ R8, R10; ADCQ $0, DX; MOVQ DX, R8; MOVQ R10, 24+z; \
+    MOVQ 24+x, AX; MULQ CX; ADDQ AX, R11; ADCQ $0, DX; ADDQ R8, R11; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 32+x, AX; MULQ CX; ADDQ AX, R12; ADCQ $0, DX; ADDQ R8, R12; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 40+x, AX; MULQ CX; ADDQ AX, R13; ADCQ $0, DX; ADDQ R8, R13; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX, R14; ADCQ $0, DX; ADDQ R8, R14; ADCQ $0, DX; MOVQ DX, R9; \
+    \
+    MOVQ 16+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ AX, R11; ADCQ $0, DX; MOVQ R11, 32+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 16+x,AX; ADDQ AX, DX; ADCQ $0, R13; MOVQ DX, R8; \
+    ADDQ 16+x, CX; ADCQ $0, R15; \
+    MOVQ 24+x, AX; MULQ CX; ADDQ AX, R12; ADCQ $0, DX; ADDQ R8, R12; ADCQ $0, DX; MOVQ DX, R8; MOVQ R12, 40+z; \
+    MOVQ 32+x, AX; MULQ CX; ADDQ AX, R13; ADCQ $0, DX; ADDQ R8, R13; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 40+x, AX; MULQ CX; ADDQ AX, R14; ADCQ $0, DX; ADDQ R8, R14; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX,  R9; ADCQ $0, DX; ADDQ R8,  R9; ADCQ $0, DX; MOVQ DX,R10; \
+    \
+    MOVQ 24+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ AX, R13; ADCQ $0, DX; MOVQ R13, 48+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 24+x,AX; ADDQ AX, DX; ADCQ $0,  R9; MOVQ DX, R8; \
+    ADDQ 24+x, CX; ADCQ $0, R15; \
+    MOVQ 32+x, AX; MULQ CX; ADDQ AX, R14; ADCQ $0, DX; ADDQ R8, R14; ADCQ $0, DX; MOVQ DX, R8; MOVQ R14, 56+z; \
+    MOVQ 40+x, AX; MULQ CX; ADDQ AX,  R9; ADCQ $0, DX; ADDQ R8,  R9; ADCQ $0, DX; MOVQ DX, R8; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX, R10; ADCQ $0, DX; ADDQ R8, R10; ADCQ $0, DX; MOVQ DX,R11; \
+    \
+    MOVQ 32+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ AX,  R9; ADCQ $0, DX; MOVQ R9, 64+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 32+x,AX; ADDQ AX, DX; ADCQ $0, R11; MOVQ DX, R8; \
+    ADDQ 32+x, CX; ADCQ $0, R15; \
+    MOVQ 40+x, AX; MULQ CX; ADDQ AX, R10; ADCQ $0, DX; ADDQ R8, R10; ADCQ $0, DX; MOVQ DX, R8; MOVQ R10, 72+z; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX, R11; ADCQ $0, DX; ADDQ R8, R11; ADCQ $0, DX; MOVQ DX,R12; \
+    \
+    XORL R13, R13; \
+    XORL R14, R14; \
+    MOVQ 40+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ AX, R11; ADCQ $0, DX; MOVQ R11, 80+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 40+x,AX; ADDQ AX, DX; ADCQ $0, R13; MOVQ DX, R8; \
+    ADDQ 40+x, CX; ADCQ $0, R15; \
+    MOVQ 48+x, AX; MULQ CX; ADDQ AX, R12; ADCQ $0, DX; ADDQ R8, R12; ADCQ $0, DX; MOVQ DX, R8; MOVQ R12, 88+z; \
+    ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ADDQ R8, R13; ADCQ $0,R14; \
+    \
+    XORL   R9, R9; \
+    MOVQ 48+x, CX; \
+    MOVQ   CX, AX; ADDQ R15, CX; MOVQ $0, R15; ADCQ $0, R15; \
+    ;;;;;;;;;;;;;; MULQ CX; ADDQ AX, R13; ADCQ $0, DX; MOVQ R13, 96+z; \
+    MOVQ  R15, AX; NEGQ AX; ANDQ 48+x,AX; ADDQ AX, DX; ADCQ $0, R9; MOVQ DX, R8; \
+    ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ADDQ R8,R14; ADCQ $0, R9; MOVQ R14, 104+z;
+
+
+// integerSqrAdx squares x and stores in z
+// Uses: AX, CX, DX, R8-R15, FLAGS
+// Instr: x86_64, bmi2, adx
+#define integerSqrAdx(z,x) \
+    XORL R15, R15; \
+    MOVQ  0+x, DX; \
+    ;;;;;;;;;;;;;; MULXQ DX, AX, R8; MOVQ AX, 0+z; \
+    ADDQ   DX, DX; ADCQ $0, R15; CLC; \
+    MULXQ  8+x, AX,  R9; ADCXQ AX,  R8; MOVQ R8, 8+z; \
+    MULXQ 16+x, AX, R10; ADCXQ AX,  R9; MOVQ $0, R8;\
+    MULXQ 24+x, AX, R11; ADCXQ AX, R10; \
+    MULXQ 32+x, AX, R12; ADCXQ AX, R11; \
+    MULXQ 40+x, AX, R13; ADCXQ AX, R12; \
+    MULXQ 48+x, AX, R14; ADCXQ AX, R13; \
+    ;;;;;;;;;;;;;;;;;;;; ADCXQ R8, R14; \
+    \
+    MOVQ  8+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 8+x, R8; \
+    ADDQ AX,  R9; MOVQ R9, 16+z; \
+    ADCQ CX,  R8; \
+    ADCQ $0, R11; \
+    ADDQ  8+x,  DX; \
+    ADCQ   $0, R15; \
+    XORL R9, R9; ;;;;;;;;;;;;;;;;;;;;; ADOXQ R8, R10; \
+    MULXQ 16+x, AX, CX; ADCXQ AX, R10; ADOXQ CX, R11; MOVQ R10, 24+z; \
+    MULXQ 24+x, AX, CX; ADCXQ AX, R11; ADOXQ CX, R12; MOVQ  $0, R10; \
+    MULXQ 32+x, AX, CX; ADCXQ AX, R12; ADOXQ CX, R13; \
+    MULXQ 40+x, AX, CX; ADCXQ AX, R13; ADOXQ CX, R14; \
+    MULXQ 48+x, AX, CX; ADCXQ AX, R14; ADOXQ CX,  R9; \
+    ;;;;;;;;;;;;;;;;;;; ADCXQ R10, R9; \
+    \
+    MOVQ 16+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 16+x, R8; \
+    ADDQ AX, R11; MOVQ R11, 32+z; \
+    ADCQ CX,  R8; \
+    ADCQ $0, R13; \
+    ADDQ 16+x,  DX; \
+    ADCQ   $0, R15; \
+    XORL R11, R11; ;;;;;;;;;;;;;;;;;;; ADOXQ R8, R12; \
+    MULXQ 24+x, AX, CX; ADCXQ AX, R12; ADOXQ CX, R13; MOVQ R12, 40+z; \
+    MULXQ 32+x, AX, CX; ADCXQ AX, R13; ADOXQ CX, R14; MOVQ  $0, R12; \
+    MULXQ 40+x, AX, CX; ADCXQ AX, R14; ADOXQ CX,  R9; \
+    MULXQ 48+x, AX, CX; ADCXQ AX,  R9; ADOXQ CX, R10; \
+    ;;;;;;;;;;;;;;;;;;; ADCXQ R11,R10; \
+    \
+    MOVQ 24+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 24+x, R8; \
+    ADDQ AX, R13; MOVQ R13, 48+z; \
+    ADCQ CX,  R8; \
+    ADCQ $0,  R9; \
+    ADDQ 24+x,  DX; \
+    ADCQ   $0, R15; \
+    XORL R13, R13; ;;;;;;;;;;;;;;;;;;; ADOXQ R8, R14; \
+    MULXQ 32+x, AX, CX; ADCXQ AX, R14; ADOXQ CX,  R9; MOVQ R14, 56+z; \
+    MULXQ 40+x, AX, CX; ADCXQ AX,  R9; ADOXQ CX, R10; MOVQ  $0, R14; \
+    MULXQ 48+x, AX, CX; ADCXQ AX, R10; ADOXQ CX, R11; \
+    ;;;;;;;;;;;;;;;;;;; ADCXQ R12,R11; \
+    \
+    MOVQ 32+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 32+x, R8; \
+    ADDQ AX,  R9; MOVQ R9, 64+z; \
+    ADCQ CX,  R8; \
+    ADCQ $0, R11; \
+    ADDQ 32+x,  DX; \
+    ADCQ   $0, R15; \
+    XORL R9, R9; ;;;;;;;;;;;;;;;;;;;;; ADOXQ R8, R10; \
+    MULXQ 40+x, AX, CX; ADCXQ AX, R10; ADOXQ CX, R11; MOVQ R10, 72+z; \
+    MULXQ 48+x, AX, CX; ADCXQ AX, R11; ADOXQ CX, R12; \
+    ;;;;;;;;;;;;;;;;;;; ADCXQ R13,R12; \
+    \
+    MOVQ 40+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 40+x, R8; \
+    ADDQ AX, R11; MOVQ R11, 80+z; \
+    ADCQ CX,  R8; \
+    ADCQ $0, R13; \
+    ADDQ 40+x,  DX; \
+    ADCQ   $0, R15; \
+    XORL R11, R11; ;;;;;;;;;;;;;;;;;;; ADOXQ R8, R12; \
+    MULXQ 48+x, AX, CX; ADCXQ AX, R12; ADOXQ CX, R13; MOVQ R12, 88+z; \
+    ;;;;;;;;;;;;;;;;;;; ADCXQ R14,R13; \
+    \
+    MOVQ 48+x, DX; \
+    MOVQ   DX, AX; ADDQ R15, DX; MOVQ $0, R15; ADCQ  $0, R15; \
+    MULXQ AX,  AX, CX; \
+    MOVQ R15,  R8; NEGQ R8; ANDQ 48+x, R8; \
+    XORL R10, R10; ;;;;;;;;;;;;;; ADOXQ CX, R14; \
+    ;;;;;;;;;;;;;; ADCXQ AX, R13; ;;;;;;;;;;;;;; MOVQ R13, 96+z; \
+    ;;;;;;;;;;;;;; ADCXQ R8, R14; MOVQ R14, 104+z;
+
+// reduceFromDoubleLeg finds a z=x modulo p such that z<2^448 and stores in z
+// Uses: AX, R8-R15, FLAGS
+// Instr: x86_64
+#define reduceFromDoubleLeg(z,x) \
+    /* (   ,2C13,2C12,2C11,2C10|C10,C9,C8, C7) + (C6,...,C0) */ \
+    /* (r14, r13, r12, r11,     r10,r9,r8,r15) */ \
+    MOVQ 80+x,AX; MOVQ AX,R10; \
+    MOVQ $0xFFFFFFFF00000000, R8; \
+    ANDQ R8,R10; \
+    \
+    MOVQ $0,R14; \
+    MOVQ 104+x,R13; SHLQ $1,R13,R14; \
+    MOVQ  96+x,R12; SHLQ $1,R12,R13; \
+    MOVQ  88+x,R11; SHLQ $1,R11,R12; \
+    MOVQ  72+x, R9; SHLQ $1,R10,R11; \
+    MOVQ  64+x, R8; SHLQ $1,R10; \
+    MOVQ $0xFFFFFFFF,R15; ANDQ R15,AX; ORQ AX,R10; \
+    MOVQ  56+x,R15; \
+    \
+    ADDQ  0+x,R15; MOVQ R15, 0+z; MOVQ  56+x,R15; \
+    ADCQ  8+x, R8; MOVQ  R8, 8+z; MOVQ  64+x, R8; \
+    ADCQ 16+x, R9; MOVQ  R9,16+z; MOVQ  72+x, R9; \
+    ADCQ 24+x,R10; MOVQ R10,24+z; MOVQ  80+x,R10; \
+    ADCQ 32+x,R11; MOVQ R11,32+z; MOVQ  88+x,R11; \
+    ADCQ 40+x,R12; MOVQ R12,40+z; MOVQ  96+x,R12; \
+    ADCQ 48+x,R13; MOVQ R13,48+z; MOVQ 104+x,R13; \
+    ADCQ   $0,R14; \
+    /* (c10c9,c9c8,c8c7,c7c13,c13c12,c12c11,c11c10) + (c6,...,c0) */ \
+    /* (   r9,  r8, r15,  r13,   r12,   r11,   r10) */ \
+    MOVQ R10, AX; \
+    SHRQ $32,R11,R10; \
+    SHRQ $32,R12,R11; \
+    SHRQ $32,R13,R12; \
+    SHRQ $32,R15,R13; \
+    SHRQ $32, R8,R15; \
+    SHRQ $32, R9, R8; \
+    SHRQ $32, AX, R9; \
+    \
+    ADDQ  0+z,R10; \
+    ADCQ  8+z,R11; \
+    ADCQ 16+z,R12; \
+    ADCQ 24+z,R13; \
+    ADCQ 32+z,R15; \
+    ADCQ 40+z, R8; \
+    ADCQ 48+z, R9; \
+    ADCQ   $0,R14; \
+    /* ( c7) + (c6,...,c0) */ \
+    /* (r14) */ \
+    MOVQ R14, AX; SHLQ $32, AX; \
+    ADDQ R14,R10; MOVQ  $0,R14; \
+    ADCQ  $0,R11; \
+    ADCQ  $0,R12; \
+    ADCQ  AX,R13; \
+    ADCQ  $0,R15; \
+    ADCQ  $0, R8; \
+    ADCQ  $0, R9; \
+    ADCQ  $0,R14; \
+    /* ( c7) + (c6,...,c0) */ \
+    /* (r14) */ \
+    MOVQ R14, AX; SHLQ $32,AX; \
+    ADDQ R14,R10; MOVQ R10, 0+z; \
+    ADCQ  $0,R11; MOVQ R11, 8+z; \
+    ADCQ  $0,R12; MOVQ R12,16+z; \
+    ADCQ  AX,R13; MOVQ R13,24+z; \
+    ADCQ  $0,R15; MOVQ R15,32+z; \
+    ADCQ  $0, R8; MOVQ  R8,40+z; \
+    ADCQ  $0, R9; MOVQ  R9,48+z;
+
+// reduceFromDoubleAdx finds a z=x modulo p such that z<2^448 and stores in z
+// Uses: AX, R8-R15, FLAGS
+// Instr: x86_64, adx
+#define reduceFromDoubleAdx(z,x) \
+    /* (   ,2C13,2C12,2C11,2C10|C10,C9,C8, C7) + (C6,...,C0) */ \
+    /* (r14, r13, r12, r11,     r10,r9,r8,r15) */ \
+    MOVQ 80+x,AX; MOVQ AX,R10; \
+    MOVQ $0xFFFFFFFF00000000, R8; \
+    ANDQ R8,R10; \
+    \
+    MOVQ $0,R14; \
+    MOVQ 104+x,R13; SHLQ $1,R13,R14; \
+    MOVQ  96+x,R12; SHLQ $1,R12,R13; \
+    MOVQ  88+x,R11; SHLQ $1,R11,R12; \
+    MOVQ  72+x, R9; SHLQ $1,R10,R11; \
+    MOVQ  64+x, R8; SHLQ $1,R10; \
+    MOVQ $0xFFFFFFFF,R15; ANDQ R15,AX; ORQ AX,R10; \
+    MOVQ  56+x,R15; \
+    \
+    XORL AX,AX; \
+    ADCXQ  0+x,R15; MOVQ R15, 0+z; MOVQ  56+x,R15; \
+    ADCXQ  8+x, R8; MOVQ  R8, 8+z; MOVQ  64+x, R8; \
+    ADCXQ 16+x, R9; MOVQ  R9,16+z; MOVQ  72+x, R9; \
+    ADCXQ 24+x,R10; MOVQ R10,24+z; MOVQ  80+x,R10; \
+    ADCXQ 32+x,R11; MOVQ R11,32+z; MOVQ  88+x,R11; \
+    ADCXQ 40+x,R12; MOVQ R12,40+z; MOVQ  96+x,R12; \
+    ADCXQ 48+x,R13; MOVQ R13,48+z; MOVQ 104+x,R13; \
+    ADCXQ   AX,R14; \
+    /* (c10c9,c9c8,c8c7,c7c13,c13c12,c12c11,c11c10) + (c6,...,c0) */ \
+    /* (   r9,  r8, r15,  r13,   r12,   r11,   r10) */ \
+    MOVQ R10, AX; \
+    SHRQ $32,R11,R10; \
+    SHRQ $32,R12,R11; \
+    SHRQ $32,R13,R12; \
+    SHRQ $32,R15,R13; \
+    SHRQ $32, R8,R15; \
+    SHRQ $32, R9, R8; \
+    SHRQ $32, AX, R9; \
+    \
+    XORL AX,AX; \
+    ADCXQ  0+z,R10; \
+    ADCXQ  8+z,R11; \
+    ADCXQ 16+z,R12; \
+    ADCXQ 24+z,R13; \
+    ADCXQ 32+z,R15; \
+    ADCXQ 40+z, R8; \
+    ADCXQ 48+z, R9; \
+    ADCXQ   AX,R14; \
+    /* ( c7) + (c6,...,c0) */ \
+    /* (r14) */ \
+    MOVQ R14, AX; SHLQ $32, AX; \
+    CLC; \
+    ADCXQ R14,R10; MOVQ $0,R14; \
+    ADCXQ R14,R11; \
+    ADCXQ R14,R12; \
+    ADCXQ  AX,R13; \
+    ADCXQ R14,R15; \
+    ADCXQ R14, R8; \
+    ADCXQ R14, R9; \
+    ADCXQ R14,R14; \
+    /* ( c7) + (c6,...,c0) */ \
+    /* (r14) */ \
+    MOVQ R14, AX; SHLQ $32, AX; \
+    CLC; \
+    ADCXQ R14,R10; MOVQ R10, 0+z; MOVQ $0,R14; \
+    ADCXQ R14,R11; MOVQ R11, 8+z; \
+    ADCXQ R14,R12; MOVQ R12,16+z; \
+    ADCXQ  AX,R13; MOVQ R13,24+z; \
+    ADCXQ R14,R15; MOVQ R15,32+z; \
+    ADCXQ R14, R8; MOVQ  R8,40+z; \
+    ADCXQ R14, R9; MOVQ  R9,48+z;
+
+// addSub calculates two operations: x,y = x+y,x-y
+// Uses: AX, DX, R8-R15, FLAGS
+#define addSub(x,y) \
+    MOVQ  0+x,  R8;  ADDQ  0+y,  R8; \
+    MOVQ  8+x,  R9;  ADCQ  8+y,  R9; \
+    MOVQ 16+x, R10;  ADCQ 16+y, R10; \
+    MOVQ 24+x, R11;  ADCQ 24+y, R11; \
+    MOVQ 32+x, R12;  ADCQ 32+y, R12; \
+    MOVQ 40+x, R13;  ADCQ 40+y, R13; \
+    MOVQ 48+x, R14;  ADCQ 48+y, R14; \
+    MOVQ   $0,  AX;  ADCQ   $0,  AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    ADDQ AX,  R8; MOVQ  $0, AX; \
+    ADCQ $0,  R9; \
+    ADCQ $0, R10; \
+    ADCQ DX, R11; \
+    ADCQ $0, R12; \
+    ADCQ $0, R13; \
+    ADCQ $0, R14; \
+    ADCQ $0,  AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    ADDQ AX,  R8;  MOVQ  0+x,AX; MOVQ  R8,  0+x; MOVQ AX,  R8; \
+    ADCQ $0,  R9;  MOVQ  8+x,AX; MOVQ  R9,  8+x; MOVQ AX,  R9; \
+    ADCQ $0, R10;  MOVQ 16+x,AX; MOVQ R10, 16+x; MOVQ AX, R10; \
+    ADCQ DX, R11;  MOVQ 24+x,AX; MOVQ R11, 24+x; MOVQ AX, R11; \
+    ADCQ $0, R12;  MOVQ 32+x,AX; MOVQ R12, 32+x; MOVQ AX, R12; \
+    ADCQ $0, R13;  MOVQ 40+x,AX; MOVQ R13, 40+x; MOVQ AX, R13; \
+    ADCQ $0, R14;  MOVQ 48+x,AX; MOVQ R14, 48+x; MOVQ AX, R14; \
+    SUBQ  0+y,  R8; \
+    SBBQ  8+y,  R9; \
+    SBBQ 16+y, R10; \
+    SBBQ 24+y, R11; \
+    SBBQ 32+y, R12; \
+    SBBQ 40+y, R13; \
+    SBBQ 48+y, R14; \
+    MOVQ   $0,  AX;  SETCS AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    SUBQ AX,  R8; MOVQ  $0, AX; \
+    SBBQ $0,  R9; \
+    SBBQ $0, R10; \
+    SBBQ DX, R11; \
+    SBBQ $0, R12; \
+    SBBQ $0, R13; \
+    SBBQ $0, R14; \
+    SETCS AX; \
+    MOVQ AX,  DX; \
+    SHLQ $32, DX; \
+    SUBQ AX,  R8;  MOVQ  R8,  0+y; \
+    SBBQ $0,  R9;  MOVQ  R9,  8+y; \
+    SBBQ $0, R10;  MOVQ R10, 16+y; \
+    SBBQ DX, R11;  MOVQ R11, 24+y; \
+    SBBQ $0, R12;  MOVQ R12, 32+y; \
+    SBBQ $0, R13;  MOVQ R13, 40+y; \
+    SBBQ $0, R14;  MOVQ R14, 48+y;

source/vendor/github.com/cloudflare/circl/math/fp448/fp_amd64.s

+//go:build amd64 && !purego
+// +build amd64,!purego
+
+#include "textflag.h"
+#include "fp_amd64.h"
+
+// func cmovAmd64(x, y *Elt, n uint)
+TEXT ·cmovAmd64(SB),NOSPLIT,$0-24
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    MOVQ n+16(FP), BX
+    cselect(0(DI),0(SI),BX)
+    RET
+
+// func cswapAmd64(x, y *Elt, n uint)
+TEXT ·cswapAmd64(SB),NOSPLIT,$0-24
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    MOVQ n+16(FP), BX
+    cswap(0(DI),0(SI),BX)
+    RET
+
+// func subAmd64(z, x, y *Elt)
+TEXT ·subAmd64(SB),NOSPLIT,$0-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    subtraction(0(DI),0(SI),0(BX))
+    RET
+
+// func addsubAmd64(x, y *Elt)
+TEXT ·addsubAmd64(SB),NOSPLIT,$0-16
+    MOVQ x+0(FP), DI
+    MOVQ y+8(FP), SI
+    addSub(0(DI),0(SI))
+    RET
+
+#define addLegacy \
+    additionLeg(0(DI),0(SI),0(BX))
+#define addBmi2Adx \
+    additionAdx(0(DI),0(SI),0(BX))
+
+#define mulLegacy \
+    integerMulLeg(0(SP),0(SI),0(BX)) \
+    reduceFromDoubleLeg(0(DI),0(SP))
+#define mulBmi2Adx \
+    integerMulAdx(0(SP),0(SI),0(BX)) \
+    reduceFromDoubleAdx(0(DI),0(SP))
+
+#define sqrLegacy \
+    integerSqrLeg(0(SP),0(SI)) \
+    reduceFromDoubleLeg(0(DI),0(SP))
+#define sqrBmi2Adx \
+    integerSqrAdx(0(SP),0(SI)) \
+    reduceFromDoubleAdx(0(DI),0(SP))
+
+// func addAmd64(z, x, y *Elt)
+TEXT ·addAmd64(SB),NOSPLIT,$0-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    CHECK_BMI2ADX(LADD, addLegacy, addBmi2Adx)
+
+// func mulAmd64(z, x, y *Elt)
+TEXT ·mulAmd64(SB),NOSPLIT,$112-24
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    MOVQ y+16(FP), BX
+    CHECK_BMI2ADX(LMUL, mulLegacy, mulBmi2Adx)
+
+// func sqrAmd64(z, x *Elt)
+TEXT ·sqrAmd64(SB),NOSPLIT,$112-16
+    MOVQ z+0(FP), DI
+    MOVQ x+8(FP), SI
+    CHECK_BMI2ADX(LSQR, sqrLegacy, sqrBmi2Adx)

source/vendor/github.com/cloudflare/circl/math/fp448/fp_generic.go

+package fp448
+
+import (
+	"encoding/binary"
+	"math/bits"
+)
+
+func cmovGeneric(x, y *Elt, n uint) {
+	m := -uint64(n & 0x1)
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+	x4 := binary.LittleEndian.Uint64(x[4*8 : 5*8])
+	x5 := binary.LittleEndian.Uint64(x[5*8 : 6*8])
+	x6 := binary.LittleEndian.Uint64(x[6*8 : 7*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+	y4 := binary.LittleEndian.Uint64(y[4*8 : 5*8])
+	y5 := binary.LittleEndian.Uint64(y[5*8 : 6*8])
+	y6 := binary.LittleEndian.Uint64(y[6*8 : 7*8])
+
+	x0 = (x0 &^ m) | (y0 & m)
+	x1 = (x1 &^ m) | (y1 & m)
+	x2 = (x2 &^ m) | (y2 & m)
+	x3 = (x3 &^ m) | (y3 & m)
+	x4 = (x4 &^ m) | (y4 & m)
+	x5 = (x5 &^ m) | (y5 & m)
+	x6 = (x6 &^ m) | (y6 & m)
+
+	binary.LittleEndian.PutUint64(x[0*8:1*8], x0)
+	binary.LittleEndian.PutUint64(x[1*8:2*8], x1)
+	binary.LittleEndian.PutUint64(x[2*8:3*8], x2)
+	binary.LittleEndian.PutUint64(x[3*8:4*8], x3)
+	binary.LittleEndian.PutUint64(x[4*8:5*8], x4)
+	binary.LittleEndian.PutUint64(x[5*8:6*8], x5)
+	binary.LittleEndian.PutUint64(x[6*8:7*8], x6)
+}
+
+func cswapGeneric(x, y *Elt, n uint) {
+	m := -uint64(n & 0x1)
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+	x4 := binary.LittleEndian.Uint64(x[4*8 : 5*8])
+	x5 := binary.LittleEndian.Uint64(x[5*8 : 6*8])
+	x6 := binary.LittleEndian.Uint64(x[6*8 : 7*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+	y4 := binary.LittleEndian.Uint64(y[4*8 : 5*8])
+	y5 := binary.LittleEndian.Uint64(y[5*8 : 6*8])
+	y6 := binary.LittleEndian.Uint64(y[6*8 : 7*8])
+
+	t0 := m & (x0 ^ y0)
+	t1 := m & (x1 ^ y1)
+	t2 := m & (x2 ^ y2)
+	t3 := m & (x3 ^ y3)
+	t4 := m & (x4 ^ y4)
+	t5 := m & (x5 ^ y5)
+	t6 := m & (x6 ^ y6)
+	x0 ^= t0
+	x1 ^= t1
+	x2 ^= t2
+	x3 ^= t3
+	x4 ^= t4
+	x5 ^= t5
+	x6 ^= t6
+	y0 ^= t0
+	y1 ^= t1
+	y2 ^= t2
+	y3 ^= t3
+	y4 ^= t4
+	y5 ^= t5
+	y6 ^= t6
+
+	binary.LittleEndian.PutUint64(x[0*8:1*8], x0)
+	binary.LittleEndian.PutUint64(x[1*8:2*8], x1)
+	binary.LittleEndian.PutUint64(x[2*8:3*8], x2)
+	binary.LittleEndian.PutUint64(x[3*8:4*8], x3)
+	binary.LittleEndian.PutUint64(x[4*8:5*8], x4)
+	binary.LittleEndian.PutUint64(x[5*8:6*8], x5)
+	binary.LittleEndian.PutUint64(x[6*8:7*8], x6)
+
+	binary.LittleEndian.PutUint64(y[0*8:1*8], y0)
+	binary.LittleEndian.PutUint64(y[1*8:2*8], y1)
+	binary.LittleEndian.PutUint64(y[2*8:3*8], y2)
+	binary.LittleEndian.PutUint64(y[3*8:4*8], y3)
+	binary.LittleEndian.PutUint64(y[4*8:5*8], y4)
+	binary.LittleEndian.PutUint64(y[5*8:6*8], y5)
+	binary.LittleEndian.PutUint64(y[6*8:7*8], y6)
+}
+
+func addGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+	x4 := binary.LittleEndian.Uint64(x[4*8 : 5*8])
+	x5 := binary.LittleEndian.Uint64(x[5*8 : 6*8])
+	x6 := binary.LittleEndian.Uint64(x[6*8 : 7*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+	y4 := binary.LittleEndian.Uint64(y[4*8 : 5*8])
+	y5 := binary.LittleEndian.Uint64(y[5*8 : 6*8])
+	y6 := binary.LittleEndian.Uint64(y[6*8 : 7*8])
+
+	z0, c0 := bits.Add64(x0, y0, 0)
+	z1, c1 := bits.Add64(x1, y1, c0)
+	z2, c2 := bits.Add64(x2, y2, c1)
+	z3, c3 := bits.Add64(x3, y3, c2)
+	z4, c4 := bits.Add64(x4, y4, c3)
+	z5, c5 := bits.Add64(x5, y5, c4)
+	z6, z7 := bits.Add64(x6, y6, c5)
+
+	z0, c0 = bits.Add64(z0, z7, 0)
+	z1, c1 = bits.Add64(z1, 0, c0)
+	z2, c2 = bits.Add64(z2, 0, c1)
+	z3, c3 = bits.Add64(z3, z7<<32, c2)
+	z4, c4 = bits.Add64(z4, 0, c3)
+	z5, c5 = bits.Add64(z5, 0, c4)
+	z6, z7 = bits.Add64(z6, 0, c5)
+
+	z0, c0 = bits.Add64(z0, z7, 0)
+	z1, c1 = bits.Add64(z1, 0, c0)
+	z2, c2 = bits.Add64(z2, 0, c1)
+	z3, c3 = bits.Add64(z3, z7<<32, c2)
+	z4, c4 = bits.Add64(z4, 0, c3)
+	z5, c5 = bits.Add64(z5, 0, c4)
+	z6, _ = bits.Add64(z6, 0, c5)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], z0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], z1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], z2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], z3)
+	binary.LittleEndian.PutUint64(z[4*8:5*8], z4)
+	binary.LittleEndian.PutUint64(z[5*8:6*8], z5)
+	binary.LittleEndian.PutUint64(z[6*8:7*8], z6)
+}
+
+func subGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+	x4 := binary.LittleEndian.Uint64(x[4*8 : 5*8])
+	x5 := binary.LittleEndian.Uint64(x[5*8 : 6*8])
+	x6 := binary.LittleEndian.Uint64(x[6*8 : 7*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+	y4 := binary.LittleEndian.Uint64(y[4*8 : 5*8])
+	y5 := binary.LittleEndian.Uint64(y[5*8 : 6*8])
+	y6 := binary.LittleEndian.Uint64(y[6*8 : 7*8])
+
+	z0, c0 := bits.Sub64(x0, y0, 0)
+	z1, c1 := bits.Sub64(x1, y1, c0)
+	z2, c2 := bits.Sub64(x2, y2, c1)
+	z3, c3 := bits.Sub64(x3, y3, c2)
+	z4, c4 := bits.Sub64(x4, y4, c3)
+	z5, c5 := bits.Sub64(x5, y5, c4)
+	z6, z7 := bits.Sub64(x6, y6, c5)
+
+	z0, c0 = bits.Sub64(z0, z7, 0)
+	z1, c1 = bits.Sub64(z1, 0, c0)
+	z2, c2 = bits.Sub64(z2, 0, c1)
+	z3, c3 = bits.Sub64(z3, z7<<32, c2)
+	z4, c4 = bits.Sub64(z4, 0, c3)
+	z5, c5 = bits.Sub64(z5, 0, c4)
+	z6, z7 = bits.Sub64(z6, 0, c5)
+
+	z0, c0 = bits.Sub64(z0, z7, 0)
+	z1, c1 = bits.Sub64(z1, 0, c0)
+	z2, c2 = bits.Sub64(z2, 0, c1)
+	z3, c3 = bits.Sub64(z3, z7<<32, c2)
+	z4, c4 = bits.Sub64(z4, 0, c3)
+	z5, c5 = bits.Sub64(z5, 0, c4)
+	z6, _ = bits.Sub64(z6, 0, c5)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], z0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], z1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], z2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], z3)
+	binary.LittleEndian.PutUint64(z[4*8:5*8], z4)
+	binary.LittleEndian.PutUint64(z[5*8:6*8], z5)
+	binary.LittleEndian.PutUint64(z[6*8:7*8], z6)
+}
+
+func addsubGeneric(x, y *Elt) {
+	z := &Elt{}
+	addGeneric(z, x, y)
+	subGeneric(y, x, y)
+	*x = *z
+}
+
+func mulGeneric(z, x, y *Elt) {
+	x0 := binary.LittleEndian.Uint64(x[0*8 : 1*8])
+	x1 := binary.LittleEndian.Uint64(x[1*8 : 2*8])
+	x2 := binary.LittleEndian.Uint64(x[2*8 : 3*8])
+	x3 := binary.LittleEndian.Uint64(x[3*8 : 4*8])
+	x4 := binary.LittleEndian.Uint64(x[4*8 : 5*8])
+	x5 := binary.LittleEndian.Uint64(x[5*8 : 6*8])
+	x6 := binary.LittleEndian.Uint64(x[6*8 : 7*8])
+
+	y0 := binary.LittleEndian.Uint64(y[0*8 : 1*8])
+	y1 := binary.LittleEndian.Uint64(y[1*8 : 2*8])
+	y2 := binary.LittleEndian.Uint64(y[2*8 : 3*8])
+	y3 := binary.LittleEndian.Uint64(y[3*8 : 4*8])
+	y4 := binary.LittleEndian.Uint64(y[4*8 : 5*8])
+	y5 := binary.LittleEndian.Uint64(y[5*8 : 6*8])
+	y6 := binary.LittleEndian.Uint64(y[6*8 : 7*8])
+
+	yy := [7]uint64{y0, y1, y2, y3, y4, y5, y6}
+	zz := [7]uint64{}
+
+	yi := yy[0]
+	h0, l0 := bits.Mul64(x0, yi)
+	h1, l1 := bits.Mul64(x1, yi)
+	h2, l2 := bits.Mul64(x2, yi)
+	h3, l3 := bits.Mul64(x3, yi)
+	h4, l4 := bits.Mul64(x4, yi)
+	h5, l5 := bits.Mul64(x5, yi)
+	h6, l6 := bits.Mul64(x6, yi)
+
+	zz[0] = l0
+	a0, c0 := bits.Add64(h0, l1, 0)
+	a1, c1 := bits.Add64(h1, l2, c0)
+	a2, c2 := bits.Add64(h2, l3, c1)
+	a3, c3 := bits.Add64(h3, l4, c2)
+	a4, c4 := bits.Add64(h4, l5, c3)
+	a5, c5 := bits.Add64(h5, l6, c4)
+	a6, _ := bits.Add64(h6, 0, c5)
+
+	for i := 1; i < 7; i++ {
+		yi = yy[i]
+		h0, l0 = bits.Mul64(x0, yi)
+		h1, l1 = bits.Mul64(x1, yi)
+		h2, l2 = bits.Mul64(x2, yi)
+		h3, l3 = bits.Mul64(x3, yi)
+		h4, l4 = bits.Mul64(x4, yi)
+		h5, l5 = bits.Mul64(x5, yi)
+		h6, l6 = bits.Mul64(x6, yi)
+
+		zz[i], c0 = bits.Add64(a0, l0, 0)
+		a0, c1 = bits.Add64(a1, l1, c0)
+		a1, c2 = bits.Add64(a2, l2, c1)
+		a2, c3 = bits.Add64(a3, l3, c2)
+		a3, c4 = bits.Add64(a4, l4, c3)
+		a4, c5 = bits.Add64(a5, l5, c4)
+		a5, a6 = bits.Add64(a6, l6, c5)
+
+		a0, c0 = bits.Add64(a0, h0, 0)
+		a1, c1 = bits.Add64(a1, h1, c0)
+		a2, c2 = bits.Add64(a2, h2, c1)
+		a3, c3 = bits.Add64(a3, h3, c2)
+		a4, c4 = bits.Add64(a4, h4, c3)
+		a5, c5 = bits.Add64(a5, h5, c4)
+		a6, _ = bits.Add64(a6, h6, c5)
+	}
+	red64(z, &zz, &[7]uint64{a0, a1, a2, a3, a4, a5, a6})
+}
+
+func sqrGeneric(z, x *Elt) { mulGeneric(z, x, x) }
+
+func red64(z *Elt, l, h *[7]uint64) {
+	/* (2C13, 2C12, 2C11, 2C10|C10, C9, C8, C7) + (C6,...,C0) */
+	h0 := h[0]
+	h1 := h[1]
+	h2 := h[2]
+	h3 := ((h[3] & (0xFFFFFFFF << 32)) << 1) | (h[3] & 0xFFFFFFFF)
+	h4 := (h[3] >> 63) | (h[4] << 1)
+	h5 := (h[4] >> 63) | (h[5] << 1)
+	h6 := (h[5] >> 63) | (h[6] << 1)
+	h7 := (h[6] >> 63)
+
+	l0, c0 := bits.Add64(h0, l[0], 0)
+	l1, c1 := bits.Add64(h1, l[1], c0)
+	l2, c2 := bits.Add64(h2, l[2], c1)
+	l3, c3 := bits.Add64(h3, l[3], c2)
+	l4, c4 := bits.Add64(h4, l[4], c3)
+	l5, c5 := bits.Add64(h5, l[5], c4)
+	l6, c6 := bits.Add64(h6, l[6], c5)
+	l7, _ := bits.Add64(h7, 0, c6)
+
+	/* (C10C9, C9C8,C8C7,C7C13,C13C12,C12C11,C11C10) + (C6,...,C0) */
+	h0 = (h[3] >> 32) | (h[4] << 32)
+	h1 = (h[4] >> 32) | (h[5] << 32)
+	h2 = (h[5] >> 32) | (h[6] << 32)
+	h3 = (h[6] >> 32) | (h[0] << 32)
+	h4 = (h[0] >> 32) | (h[1] << 32)
+	h5 = (h[1] >> 32) | (h[2] << 32)
+	h6 = (h[2] >> 32) | (h[3] << 32)
+
+	l0, c0 = bits.Add64(l0, h0, 0)
+	l1, c1 = bits.Add64(l1, h1, c0)
+	l2, c2 = bits.Add64(l2, h2, c1)
+	l3, c3 = bits.Add64(l3, h3, c2)
+	l4, c4 = bits.Add64(l4, h4, c3)
+	l5, c5 = bits.Add64(l5, h5, c4)
+	l6, c6 = bits.Add64(l6, h6, c5)
+	l7, _ = bits.Add64(l7, 0, c6)
+
+	/* (C7) + (C6,...,C0) */
+	l0, c0 = bits.Add64(l0, l7, 0)
+	l1, c1 = bits.Add64(l1, 0, c0)
+	l2, c2 = bits.Add64(l2, 0, c1)
+	l3, c3 = bits.Add64(l3, l7<<32, c2)
+	l4, c4 = bits.Add64(l4, 0, c3)
+	l5, c5 = bits.Add64(l5, 0, c4)
+	l6, l7 = bits.Add64(l6, 0, c5)
+
+	/* (C7) + (C6,...,C0) */
+	l0, c0 = bits.Add64(l0, l7, 0)
+	l1, c1 = bits.Add64(l1, 0, c0)
+	l2, c2 = bits.Add64(l2, 0, c1)
+	l3, c3 = bits.Add64(l3, l7<<32, c2)
+	l4, c4 = bits.Add64(l4, 0, c3)
+	l5, c5 = bits.Add64(l5, 0, c4)
+	l6, _ = bits.Add64(l6, 0, c5)
+
+	binary.LittleEndian.PutUint64(z[0*8:1*8], l0)
+	binary.LittleEndian.PutUint64(z[1*8:2*8], l1)
+	binary.LittleEndian.PutUint64(z[2*8:3*8], l2)
+	binary.LittleEndian.PutUint64(z[3*8:4*8], l3)
+	binary.LittleEndian.PutUint64(z[4*8:5*8], l4)
+	binary.LittleEndian.PutUint64(z[5*8:6*8], l5)
+	binary.LittleEndian.PutUint64(z[6*8:7*8], l6)
+}

source/vendor/github.com/cloudflare/circl/math/fp448/fp_noasm.go

+//go:build !amd64 || purego
+// +build !amd64 purego
+
+package fp448
+
+func cmov(x, y *Elt, n uint)  { cmovGeneric(x, y, n) }
+func cswap(x, y *Elt, n uint) { cswapGeneric(x, y, n) }
+func add(z, x, y *Elt)        { addGeneric(z, x, y) }
+func sub(z, x, y *Elt)        { subGeneric(z, x, y) }
+func addsub(x, y *Elt)        { addsubGeneric(x, y) }
+func mul(z, x, y *Elt)        { mulGeneric(z, x, y) }
+func sqr(z, x *Elt)           { sqrGeneric(z, x) }

source/vendor/github.com/cloudflare/circl/math/fp448/fuzzer.go

+//go:build gofuzz
+// +build gofuzz
+
+// How to run the fuzzer:
+//
+//	$ go get -u github.com/dvyukov/go-fuzz/go-fuzz
+//	$ go get -u github.com/dvyukov/go-fuzz/go-fuzz-build
+//	$ go-fuzz-build -libfuzzer -func FuzzReduction -o lib.a
+//	$ clang -fsanitize=fuzzer lib.a -o fu.exe
+//	$ ./fu.exe
+package fp448
+
+import (
+	"encoding/binary"
+	"fmt"
+	"math/big"
+
+	"github.com/cloudflare/circl/internal/conv"
+)
+
+// FuzzReduction is a fuzzer target for red64 function, which reduces t
+// (112 bits) to a number t' (56 bits) congruent modulo p448.
+func FuzzReduction(data []byte) int {
+	if len(data) != 2*Size {
+		return -1
+	}
+	var got, want Elt
+	var lo, hi [7]uint64
+	a := data[:Size]
+	b := data[Size:]
+	lo[0] = binary.LittleEndian.Uint64(a[0*8 : 1*8])
+	lo[1] = binary.LittleEndian.Uint64(a[1*8 : 2*8])
+	lo[2] = binary.LittleEndian.Uint64(a[2*8 : 3*8])
+	lo[3] = binary.LittleEndian.Uint64(a[3*8 : 4*8])
+	lo[4] = binary.LittleEndian.Uint64(a[4*8 : 5*8])
+	lo[5] = binary.LittleEndian.Uint64(a[5*8 : 6*8])
+	lo[6] = binary.LittleEndian.Uint64(a[6*8 : 7*8])
+
+	hi[0] = binary.LittleEndian.Uint64(b[0*8 : 1*8])
+	hi[1] = binary.LittleEndian.Uint64(b[1*8 : 2*8])
+	hi[2] = binary.LittleEndian.Uint64(b[2*8 : 3*8])
+	hi[3] = binary.LittleEndian.Uint64(b[3*8 : 4*8])
+	hi[4] = binary.LittleEndian.Uint64(b[4*8 : 5*8])
+	hi[5] = binary.LittleEndian.Uint64(b[5*8 : 6*8])
+	hi[6] = binary.LittleEndian.Uint64(b[6*8 : 7*8])
+
+	red64(&got, &lo, &hi)
+
+	t := conv.BytesLe2BigInt(data[:2*Size])
+
+	two448 := big.NewInt(1)
+	two448.Lsh(two448, 448) // 2^448
+	mask448 := big.NewInt(1)
+	mask448.Sub(two448, mask448) // 2^448-1
+	two224plus1 := big.NewInt(1)
+	two224plus1.Lsh(two224plus1, 224)
+	two224plus1.Add(two224plus1, big.NewInt(1)) // 2^224+1
+
+	var loBig, hiBig big.Int
+	for t.Cmp(two448) >= 0 {
+		loBig.And(t, mask448)
+		hiBig.Rsh(t, 448)
+		t.Mul(&hiBig, two224plus1)
+		t.Add(t, &loBig)
+	}
+	conv.BigInt2BytesLe(want[:], t)
+
+	if got != want {
+		fmt.Printf("in:   %v\n", conv.BytesLe2BigInt(data[:2*Size]))
+		fmt.Printf("got:  %v\n", got)
+		fmt.Printf("want: %v\n", want)
+		panic("error found")
+	}
+	return 1
+}

source/vendor/github.com/cloudflare/circl/math/mlsbset/mlsbset.go

+// Package mlsbset provides a constant-time exponentiation method with precomputation.
+//
+// References: "Efficient and secure algorithms for GLV-based scalar
+// multiplication and their implementation on GLV–GLS curves" by (Faz-Hernandez et al.)
+//   - https://doi.org/10.1007/s13389-014-0085-7
+//   - https://eprint.iacr.org/2013/158
+package mlsbset
+
+import (
+	"errors"
+	"fmt"
+	"math/big"
+
+	"github.com/cloudflare/circl/internal/conv"
+)
+
+// EltG is a group element.
+type EltG interface{}
+
+// EltP is a precomputed group element.
+type EltP interface{}
+
+// Group defines the operations required by MLSBSet exponentiation method.
+type Group interface {
+	Identity() EltG                    // Returns the identity of the group.
+	Sqr(x EltG)                        // Calculates x = x^2.
+	Mul(x EltG, y EltP)                // Calculates x = x*y.
+	NewEltP() EltP                     // Returns an arbitrary precomputed element.
+	ExtendedEltP() EltP                // Returns the precomputed element x^(2^(w*d)).
+	Lookup(a EltP, v uint, s, u int32) // Sets a = s*T[v][u].
+}
+
+// Params contains the parameters of the encoding.
+type Params struct {
+	T uint // T is the maximum size (in bits) of exponents.
+	V uint // V is the number of tables.
+	W uint // W is the window size.
+	E uint // E is the number of digits per table.
+	D uint // D is the number of digits in total.
+	L uint // L is the length of the code.
+}
+
+// Encoder allows to convert integers into valid powers.
+type Encoder struct{ p Params }
+
+// New produces an encoder of the MLSBSet algorithm.
+func New(t, v, w uint) (Encoder, error) {
+	if !(t > 1 && v >= 1 && w >= 2) {
+		return Encoder{}, errors.New("t>1, v>=1, w>=2")
+	}
+	e := (t + w*v - 1) / (w * v)
+	d := e * v
+	l := d * w
+	return Encoder{Params{t, v, w, e, d, l}}, nil
+}
+
+// Encode converts an odd integer k into a valid power for exponentiation.
+func (m Encoder) Encode(k []byte) (*Power, error) {
+	if len(k) == 0 {
+		return nil, errors.New("empty slice")
+	}
+	if !(len(k) <= int(m.p.L+7)>>3) {
+		return nil, errors.New("k too big")
+	}
+	if k[0]%2 == 0 {
+		return nil, errors.New("k must be odd")
+	}
+	ap := int((m.p.L+7)/8) - len(k)
+	k = append(k, make([]byte, ap)...)
+	s := m.signs(k)
+	b := make([]int32, m.p.L-m.p.D)
+	c := conv.BytesLe2BigInt(k)
+	c.Rsh(c, m.p.D)
+	var bi big.Int
+	for i := m.p.D; i < m.p.L; i++ {
+		c0 := int32(c.Bit(0))
+		b[i-m.p.D] = s[i%m.p.D] * c0
+		bi.SetInt64(int64(b[i-m.p.D] >> 1))
+		c.Rsh(c, 1)
+		c.Sub(c, &bi)
+	}
+	carry := int(c.Int64())
+	return &Power{m, s, b, carry}, nil
+}
+
+// signs calculates the set of signs.
+func (m Encoder) signs(k []byte) []int32 {
+	s := make([]int32, m.p.D)
+	s[m.p.D-1] = 1
+	for i := uint(1); i < m.p.D; i++ {
+		ki := int32((k[i>>3] >> (i & 0x7)) & 0x1)
+		s[i-1] = 2*ki - 1
+	}
+	return s
+}
+
+// GetParams returns the complementary parameters of the encoding.
+func (m Encoder) GetParams() Params { return m.p }
+
+// tableSize returns the size of each table.
+func (m Encoder) tableSize() uint { return 1 << (m.p.W - 1) }
+
+// Elts returns the total number of elements that must be precomputed.
+func (m Encoder) Elts() uint { return m.p.V * m.tableSize() }
+
+// IsExtended returns true if the element x^(2^(wd)) must be calculated.
+func (m Encoder) IsExtended() bool { q := m.p.T / (m.p.V * m.p.W); return m.p.T == q*m.p.V*m.p.W }
+
+// Ops returns the number of squares and multiplications executed during an exponentiation.
+func (m Encoder) Ops() (S uint, M uint) {
+	S = m.p.E
+	M = m.p.E * m.p.V
+	if m.IsExtended() {
+		M++
+	}
+	return
+}
+
+func (m Encoder) String() string {
+	return fmt.Sprintf("T: %v W: %v V: %v e: %v d: %v l: %v wv|t: %v",
+		m.p.T, m.p.W, m.p.V, m.p.E, m.p.D, m.p.L, m.IsExtended())
+}