Update vendor

42 files changed, 4189 insertions, 2891 deletions

diff --git a/vendor/golang.org/x/crypto/acme/types.go b/vendor/golang.org/x/crypto/acme/types.go
index e751bf52..eaae4529 100644
--- a/vendor/golang.org/x/crypto/acme/types.go
+++ b/vendor/golang.org/x/crypto/acme/types.go
@@ -57,6 +57,32 @@ var (
 	ErrNoAccount = errors.New("acme: account does not exist")
 )
 
+// A Subproblem describes an ACME subproblem as reported in an Error.
+type Subproblem struct {
+	// Type is a URI reference that identifies the problem type,
+	// typically in a "urn:acme:error:xxx" form.
+	Type string
+	// Detail is a human-readable explanation specific to this occurrence of the problem.
+	Detail string
+	// Instance indicates a URL that the client should direct a human user to visit
+	// in order for instructions on how to agree to the updated Terms of Service.
+	// In such an event CA sets StatusCode to 403, Type to
+	// "urn:ietf:params:acme:error:userActionRequired", and adds a Link header with relation
+	// "terms-of-service" containing the latest TOS URL.
+	Instance string
+	// Identifier may contain the ACME identifier that the error is for.
+	Identifier *AuthzID
+}
+
+func (sp Subproblem) String() string {
+	str := fmt.Sprintf("%s: ", sp.Type)
+	if sp.Identifier != nil {
+		str += fmt.Sprintf("[%s: %s] ", sp.Identifier.Type, sp.Identifier.Value)
+	}
+	str += sp.Detail
+	return str
+}
+
 // Error is an ACME error, defined in Problem Details for HTTP APIs doc
 // http://tools.ietf.org/html/draft-ietf-appsawg-http-problem.
 type Error struct {
@@ -76,10 +102,21 @@ type Error struct {
 	// Header is the original server error response headers.
 	// It may be nil.
 	Header http.Header
+	// Subproblems may contain more detailed information about the individual problems
+	// that caused the error. This field is only sent by RFC 8555 compatible ACME
+	// servers. Defined in RFC 8555 Section 6.7.1.
+	Subproblems []Subproblem
 }
 
 func (e *Error) Error() string {
-	return fmt.Sprintf("%d %s: %s", e.StatusCode, e.ProblemType, e.Detail)
+	str := fmt.Sprintf("%d %s: %s", e.StatusCode, e.ProblemType, e.Detail)
+	if len(e.Subproblems) > 0 {
+		str += fmt.Sprintf("; subproblems:")
+		for _, sp := range e.Subproblems {
+			str += fmt.Sprintf("\n\t%s", sp)
+		}
+	}
+	return str
 }
 
 // AuthorizationError indicates that an authorization for an identifier
@@ -533,20 +570,23 @@ func (c *wireChallenge) challenge() *Challenge {
 // wireError is a subset of fields of the Problem Details object
 // as described in https://tools.ietf.org/html/rfc7807#section-3.1.
 type wireError struct {
-	Status   int
-	Type     string
-	Detail   string
-	Instance string
+	Status      int
+	Type        string
+	Detail      string
+	Instance    string
+	Subproblems []Subproblem
 }
 
 func (e *wireError) error(h http.Header) *Error {
-	return &Error{
+	err := &Error{
 		StatusCode:  e.Status,
 		ProblemType: e.Type,
 		Detail:      e.Detail,
 		Instance:    e.Instance,
 		Header:      h,
+		Subproblems: e.Subproblems,
 	}
+	return err
 }
 
 // CertOption is an optional argument type for the TLS ChallengeCert methods for
diff --git a/vendor/golang.org/x/crypto/argon2/argon2.go b/vendor/golang.org/x/crypto/argon2/argon2.go
new file mode 100644
index 00000000..b423feae
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/argon2.go
@@ -0,0 +1,285 @@
+// 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.
+
+// Package argon2 implements the key derivation function Argon2.
+// Argon2 was selected as the winner of the Password Hashing Competition and can
+// be used to derive cryptographic keys from passwords.
+//
+// For a detailed specification of Argon2 see [1].
+//
+// If you aren't sure which function you need, use Argon2id (IDKey) and
+// the parameter recommendations for your scenario.
+//
+//
+// Argon2i
+//
+// Argon2i (implemented by Key) is the side-channel resistant version of Argon2.
+// It uses data-independent memory access, which is preferred for password
+// hashing and password-based key derivation. Argon2i requires more passes over
+// memory than Argon2id to protect from trade-off attacks. The recommended
+// parameters (taken from [2]) for non-interactive operations are time=3 and to
+// use the maximum available memory.
+//
+//
+// Argon2id
+//
+// Argon2id (implemented by IDKey) is a hybrid version of Argon2 combining
+// Argon2i and Argon2d. It uses data-independent memory access for the first
+// half of the first iteration over the memory and data-dependent memory access
+// for the rest. Argon2id is side-channel resistant and provides better brute-
+// force cost savings due to time-memory tradeoffs than Argon2i. The recommended
+// parameters for non-interactive operations (taken from [2]) are time=1 and to
+// use the maximum available memory.
+//
+// [1] https://github.com/P-H-C/phc-winner-argon2/blob/master/argon2-specs.pdf
+// [2] https://tools.ietf.org/html/draft-irtf-cfrg-argon2-03#section-9.3
+package argon2
+
+import (
+	"encoding/binary"
+	"sync"
+
+	"golang.org/x/crypto/blake2b"
+)
+
+// The Argon2 version implemented by this package.
+const Version = 0x13
+
+const (
+	argon2d = iota
+	argon2i
+	argon2id
+)
+
+// Key derives a key from the password, salt, and cost parameters using Argon2i
+// returning a byte slice of length keyLen that can be used as cryptographic
+// key. The CPU cost and parallelism degree must be greater than zero.
+//
+// For example, you can get a derived key for e.g. AES-256 (which needs a
+// 32-byte key) by doing:
+//
+//      key := argon2.Key([]byte("some password"), salt, 3, 32*1024, 4, 32)
+//
+// The draft RFC recommends[2] time=3, and memory=32*1024 is a sensible number.
+// If using that amount of memory (32 MB) is not possible in some contexts then
+// the time parameter can be increased to compensate.
+//
+// The time parameter specifies the number of passes over the memory and the
+// memory parameter specifies the size of the memory in KiB. For example
+// memory=32*1024 sets the memory cost to ~32 MB. The number of threads can be
+// adjusted to the number of available CPUs. The cost parameters should be
+// increased as memory latency and CPU parallelism increases. Remember to get a
+// good random salt.
+func Key(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
+	return deriveKey(argon2i, password, salt, nil, nil, time, memory, threads, keyLen)
+}
+
+// IDKey derives a key from the password, salt, and cost parameters using
+// Argon2id returning a byte slice of length keyLen that can be used as
+// cryptographic key. The CPU cost and parallelism degree must be greater than
+// zero.
+//
+// For example, you can get a derived key for e.g. AES-256 (which needs a
+// 32-byte key) by doing:
+//
+//      key := argon2.IDKey([]byte("some password"), salt, 1, 64*1024, 4, 32)
+//
+// The draft RFC recommends[2] time=1, and memory=64*1024 is a sensible number.
+// If using that amount of memory (64 MB) is not possible in some contexts then
+// the time parameter can be increased to compensate.
+//
+// The time parameter specifies the number of passes over the memory and the
+// memory parameter specifies the size of the memory in KiB. For example
+// memory=64*1024 sets the memory cost to ~64 MB. The number of threads can be
+// adjusted to the numbers of available CPUs. The cost parameters should be
+// increased as memory latency and CPU parallelism increases. Remember to get a
+// good random salt.
+func IDKey(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
+	return deriveKey(argon2id, password, salt, nil, nil, time, memory, threads, keyLen)
+}
+
+func deriveKey(mode int, password, salt, secret, data []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
+	if time < 1 {
+		panic("argon2: number of rounds too small")
+	}
+	if threads < 1 {
+		panic("argon2: parallelism degree too low")
+	}
+	h0 := initHash(password, salt, secret, data, time, memory, uint32(threads), keyLen, mode)
+
+	memory = memory / (syncPoints * uint32(threads)) * (syncPoints * uint32(threads))
+	if memory < 2*syncPoints*uint32(threads) {
+		memory = 2 * syncPoints * uint32(threads)
+	}
+	B := initBlocks(&h0, memory, uint32(threads))
+	processBlocks(B, time, memory, uint32(threads), mode)
+	return extractKey(B, memory, uint32(threads), keyLen)
+}
+
+const (
+	blockLength = 128
+	syncPoints  = 4
+)
+
+type block [blockLength]uint64
+
+func initHash(password, salt, key, data []byte, time, memory, threads, keyLen uint32, mode int) [blake2b.Size + 8]byte {
+	var (
+		h0     [blake2b.Size + 8]byte
+		params [24]byte
+		tmp    [4]byte
+	)
+
+	b2, _ := blake2b.New512(nil)
+	binary.LittleEndian.PutUint32(params[0:4], threads)
+	binary.LittleEndian.PutUint32(params[4:8], keyLen)
+	binary.LittleEndian.PutUint32(params[8:12], memory)
+	binary.LittleEndian.PutUint32(params[12:16], time)
+	binary.LittleEndian.PutUint32(params[16:20], uint32(Version))
+	binary.LittleEndian.PutUint32(params[20:24], uint32(mode))
+	b2.Write(params[:])
+	binary.LittleEndian.PutUint32(tmp[:], uint32(len(password)))
+	b2.Write(tmp[:])
+	b2.Write(password)
+	binary.LittleEndian.PutUint32(tmp[:], uint32(len(salt)))
+	b2.Write(tmp[:])
+	b2.Write(salt)
+	binary.LittleEndian.PutUint32(tmp[:], uint32(len(key)))
+	b2.Write(tmp[:])
+	b2.Write(key)
+	binary.LittleEndian.PutUint32(tmp[:], uint32(len(data)))
+	b2.Write(tmp[:])
+	b2.Write(data)
+	b2.Sum(h0[:0])
+	return h0
+}
+
+func initBlocks(h0 *[blake2b.Size + 8]byte, memory, threads uint32) []block {
+	var block0 [1024]byte
+	B := make([]block, memory)
+	for lane := uint32(0); lane < threads; lane++ {
+		j := lane * (memory / threads)
+		binary.LittleEndian.PutUint32(h0[blake2b.Size+4:], lane)
+
+		binary.LittleEndian.PutUint32(h0[blake2b.Size:], 0)
+		blake2bHash(block0[:], h0[:])
+		for i := range B[j+0] {
+			B[j+0][i] = binary.LittleEndian.Uint64(block0[i*8:])
+		}
+
+		binary.LittleEndian.PutUint32(h0[blake2b.Size:], 1)
+		blake2bHash(block0[:], h0[:])
+		for i := range B[j+1] {
+			B[j+1][i] = binary.LittleEndian.Uint64(block0[i*8:])
+		}
+	}
+	return B
+}
+
+func processBlocks(B []block, time, memory, threads uint32, mode int) {
+	lanes := memory / threads
+	segments := lanes / syncPoints
+
+	processSegment := func(n, slice, lane uint32, wg *sync.WaitGroup) {
+		var addresses, in, zero block
+		if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
+			in[0] = uint64(n)
+			in[1] = uint64(lane)
+			in[2] = uint64(slice)
+			in[3] = uint64(memory)
+			in[4] = uint64(time)
+			in[5] = uint64(mode)
+		}
+
+		index := uint32(0)
+		if n == 0 && slice == 0 {
+			index = 2 // we have already generated the first two blocks
+			if mode == argon2i || mode == argon2id {
+				in[6]++
+				processBlock(&addresses, &in, &zero)
+				processBlock(&addresses, &addresses, &zero)
+			}
+		}
+
+		offset := lane*lanes + slice*segments + index
+		var random uint64
+		for index < segments {
+			prev := offset - 1
+			if index == 0 && slice == 0 {
+				prev += lanes // last block in lane
+			}
+			if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
+				if index%blockLength == 0 {
+					in[6]++
+					processBlock(&addresses, &in, &zero)
+					processBlock(&addresses, &addresses, &zero)
+				}
+				random = addresses[index%blockLength]
+			} else {
+				random = B[prev][0]
+			}
+			newOffset := indexAlpha(random, lanes, segments, threads, n, slice, lane, index)
+			processBlockXOR(&B[offset], &B[prev], &B[newOffset])
+			index, offset = index+1, offset+1
+		}
+		wg.Done()
+	}
+
+	for n := uint32(0); n < time; n++ {
+		for slice := uint32(0); slice < syncPoints; slice++ {
+			var wg sync.WaitGroup
+			for lane := uint32(0); lane < threads; lane++ {
+				wg.Add(1)
+				go processSegment(n, slice, lane, &wg)
+			}
+			wg.Wait()
+		}
+	}
+
+}
+
+func extractKey(B []block, memory, threads, keyLen uint32) []byte {
+	lanes := memory / threads
+	for lane := uint32(0); lane < threads-1; lane++ {
+		for i, v := range B[(lane*lanes)+lanes-1] {
+			B[memory-1][i] ^= v
+		}
+	}
+
+	var block [1024]byte
+	for i, v := range B[memory-1] {
+		binary.LittleEndian.PutUint64(block[i*8:], v)
+	}
+	key := make([]byte, keyLen)
+	blake2bHash(key, block[:])
+	return key
+}
+
+func indexAlpha(rand uint64, lanes, segments, threads, n, slice, lane, index uint32) uint32 {
+	refLane := uint32(rand>>32) % threads
+	if n == 0 && slice == 0 {
+		refLane = lane
+	}
+	m, s := 3*segments, ((slice+1)%syncPoints)*segments
+	if lane == refLane {
+		m += index
+	}
+	if n == 0 {
+		m, s = slice*segments, 0
+		if slice == 0 || lane == refLane {
+			m += index
+		}
+	}
+	if index == 0 || lane == refLane {
+		m--
+	}
+	return phi(rand, uint64(m), uint64(s), refLane, lanes)
+}
+
+func phi(rand, m, s uint64, lane, lanes uint32) uint32 {
+	p := rand & 0xFFFFFFFF
+	p = (p * p) >> 32
+	p = (p * m) >> 32
+	return lane*lanes + uint32((s+m-(p+1))%uint64(lanes))
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blake2b.go b/vendor/golang.org/x/crypto/argon2/blake2b.go
new file mode 100644
index 00000000..10f46948
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blake2b.go
@@ -0,0 +1,53 @@
+// 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.
+
+package argon2
+
+import (
+	"encoding/binary"
+	"hash"
+
+	"golang.org/x/crypto/blake2b"
+)
+
+// blake2bHash computes an arbitrary long hash value of in
+// and writes the hash to out.
+func blake2bHash(out []byte, in []byte) {
+	var b2 hash.Hash
+	if n := len(out); n < blake2b.Size {
+		b2, _ = blake2b.New(n, nil)
+	} else {
+		b2, _ = blake2b.New512(nil)
+	}
+
+	var buffer [blake2b.Size]byte
+	binary.LittleEndian.PutUint32(buffer[:4], uint32(len(out)))
+	b2.Write(buffer[:4])
+	b2.Write(in)
+
+	if len(out) <= blake2b.Size {
+		b2.Sum(out[:0])
+		return
+	}
+
+	outLen := len(out)
+	b2.Sum(buffer[:0])
+	b2.Reset()
+	copy(out, buffer[:32])
+	out = out[32:]
+	for len(out) > blake2b.Size {
+		b2.Write(buffer[:])
+		b2.Sum(buffer[:0])
+		copy(out, buffer[:32])
+		out = out[32:]
+		b2.Reset()
+	}
+
+	if outLen%blake2b.Size > 0 { // outLen > 64
+		r := ((outLen + 31) / 32) - 2 // ⌈τ /32⌉-2
+		b2, _ = blake2b.New(outLen-32*r, nil)
+	}
+	b2.Write(buffer[:])
+	b2.Sum(out[:0])
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_amd64.go b/vendor/golang.org/x/crypto/argon2/blamka_amd64.go
new file mode 100644
index 00000000..a014ac92
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_amd64.go
@@ -0,0 +1,61 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+package argon2
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func mixBlocksSSE2(out, a, b, c *block)
+
+//go:noescape
+func xorBlocksSSE2(out, a, b, c *block)
+
+//go:noescape
+func blamkaSSE4(b *block)
+
+func processBlockSSE(out, in1, in2 *block, xor bool) {
+	var t block
+	mixBlocksSSE2(&t, in1, in2, &t)
+	if useSSE4 {
+		blamkaSSE4(&t)
+	} else {
+		for i := 0; i < blockLength; i += 16 {
+			blamkaGeneric(
+				&t[i+0], &t[i+1], &t[i+2], &t[i+3],
+				&t[i+4], &t[i+5], &t[i+6], &t[i+7],
+				&t[i+8], &t[i+9], &t[i+10], &t[i+11],
+				&t[i+12], &t[i+13], &t[i+14], &t[i+15],
+			)
+		}
+		for i := 0; i < blockLength/8; i += 2 {
+			blamkaGeneric(
+				&t[i], &t[i+1], &t[16+i], &t[16+i+1],
+				&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
+				&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
+				&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
+			)
+		}
+	}
+	if xor {
+		xorBlocksSSE2(out, in1, in2, &t)
+	} else {
+		mixBlocksSSE2(out, in1, in2, &t)
+	}
+}
+
+func processBlock(out, in1, in2 *block) {
+	processBlockSSE(out, in1, in2, false)
+}
+
+func processBlockXOR(out, in1, in2 *block) {
+	processBlockSSE(out, in1, in2, true)
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_amd64.s b/vendor/golang.org/x/crypto/argon2/blamka_amd64.s
new file mode 100644
index 00000000..b2cc0515
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_amd64.s
@@ -0,0 +1,244 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
+
+#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v6, t1; \
+	PUNPCKLQDQ v6, t2; \
+	PUNPCKHQDQ v7, v6; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ v7, t2; \
+	MOVO       t1, v7; \
+	MOVO       v2, t1; \
+	PUNPCKHQDQ t2, v7; \
+	PUNPCKLQDQ v3, t2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v3
+
+#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v2, t1; \
+	PUNPCKLQDQ v2, t2; \
+	PUNPCKHQDQ v3, v2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ v3, t2; \
+	MOVO       t1, v3; \
+	MOVO       v6, t1; \
+	PUNPCKHQDQ t2, v3; \
+	PUNPCKLQDQ v7, t2; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v7
+
+#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, t0, c40, c48) \
+	MOVO    v0, t0;        \
+	PMULULQ v2, t0;        \
+	PADDQ   v2, v0;        \
+	PADDQ   t0, v0;        \
+	PADDQ   t0, v0;        \
+	PXOR    v0, v6;        \
+	PSHUFD  $0xB1, v6, v6; \
+	MOVO    v4, t0;        \
+	PMULULQ v6, t0;        \
+	PADDQ   v6, v4;        \
+	PADDQ   t0, v4;        \
+	PADDQ   t0, v4;        \
+	PXOR    v4, v2;        \
+	PSHUFB  c40, v2;       \
+	MOVO    v0, t0;        \
+	PMULULQ v2, t0;        \
+	PADDQ   v2, v0;        \
+	PADDQ   t0, v0;        \
+	PADDQ   t0, v0;        \
+	PXOR    v0, v6;        \
+	PSHUFB  c48, v6;       \
+	MOVO    v4, t0;        \
+	PMULULQ v6, t0;        \
+	PADDQ   v6, v4;        \
+	PADDQ   t0, v4;        \
+	PADDQ   t0, v4;        \
+	PXOR    v4, v2;        \
+	MOVO    v2, t0;        \
+	PADDQ   v2, t0;        \
+	PSRLQ   $63, v2;       \
+	PXOR    t0, v2;        \
+	MOVO    v1, t0;        \
+	PMULULQ v3, t0;        \
+	PADDQ   v3, v1;        \
+	PADDQ   t0, v1;        \
+	PADDQ   t0, v1;        \
+	PXOR    v1, v7;        \
+	PSHUFD  $0xB1, v7, v7; \
+	MOVO    v5, t0;        \
+	PMULULQ v7, t0;        \
+	PADDQ   v7, v5;        \
+	PADDQ   t0, v5;        \
+	PADDQ   t0, v5;        \
+	PXOR    v5, v3;        \
+	PSHUFB  c40, v3;       \
+	MOVO    v1, t0;        \
+	PMULULQ v3, t0;        \
+	PADDQ   v3, v1;        \
+	PADDQ   t0, v1;        \
+	PADDQ   t0, v1;        \
+	PXOR    v1, v7;        \
+	PSHUFB  c48, v7;       \
+	MOVO    v5, t0;        \
+	PMULULQ v7, t0;        \
+	PADDQ   v7, v5;        \
+	PADDQ   t0, v5;        \
+	PADDQ   t0, v5;        \
+	PXOR    v5, v3;        \
+	MOVO    v3, t0;        \
+	PADDQ   v3, t0;        \
+	PSRLQ   $63, v3;       \
+	PXOR    t0, v3
+
+#define LOAD_MSG_0(block, off) \
+	MOVOU 8*(off+0)(block), X0;  \
+	MOVOU 8*(off+2)(block), X1;  \
+	MOVOU 8*(off+4)(block), X2;  \
+	MOVOU 8*(off+6)(block), X3;  \
+	MOVOU 8*(off+8)(block), X4;  \
+	MOVOU 8*(off+10)(block), X5; \
+	MOVOU 8*(off+12)(block), X6; \
+	MOVOU 8*(off+14)(block), X7
+
+#define STORE_MSG_0(block, off) \
+	MOVOU X0, 8*(off+0)(block);  \
+	MOVOU X1, 8*(off+2)(block);  \
+	MOVOU X2, 8*(off+4)(block);  \
+	MOVOU X3, 8*(off+6)(block);  \
+	MOVOU X4, 8*(off+8)(block);  \
+	MOVOU X5, 8*(off+10)(block); \
+	MOVOU X6, 8*(off+12)(block); \
+	MOVOU X7, 8*(off+14)(block)
+
+#define LOAD_MSG_1(block, off) \
+	MOVOU 8*off+0*8(block), X0;  \
+	MOVOU 8*off+16*8(block), X1; \
+	MOVOU 8*off+32*8(block), X2; \
+	MOVOU 8*off+48*8(block), X3; \
+	MOVOU 8*off+64*8(block), X4; \
+	MOVOU 8*off+80*8(block), X5; \
+	MOVOU 8*off+96*8(block), X6; \
+	MOVOU 8*off+112*8(block), X7
+
+#define STORE_MSG_1(block, off) \
+	MOVOU X0, 8*off+0*8(block);  \
+	MOVOU X1, 8*off+16*8(block); \
+	MOVOU X2, 8*off+32*8(block); \
+	MOVOU X3, 8*off+48*8(block); \
+	MOVOU X4, 8*off+64*8(block); \
+	MOVOU X5, 8*off+80*8(block); \
+	MOVOU X6, 8*off+96*8(block); \
+	MOVOU X7, 8*off+112*8(block)
+
+#define BLAMKA_ROUND_0(block, off, t0, t1, c40, c48) \
+	LOAD_MSG_0(block, off);                                   \
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
+	SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1);                  \
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1);              \
+	STORE_MSG_0(block, off)
+
+#define BLAMKA_ROUND_1(block, off, t0, t1, c40, c48) \
+	LOAD_MSG_1(block, off);                                   \
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
+	SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1);                  \
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1);              \
+	STORE_MSG_1(block, off)
+
+// func blamkaSSE4(b *block)
+TEXT ·blamkaSSE4(SB), 4, $0-8
+	MOVQ b+0(FP), AX
+
+	MOVOU ·c40<>(SB), X10
+	MOVOU ·c48<>(SB), X11
+
+	BLAMKA_ROUND_0(AX, 0, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 16, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 32, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 48, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 64, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 80, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 96, X8, X9, X10, X11)
+	BLAMKA_ROUND_0(AX, 112, X8, X9, X10, X11)
+
+	BLAMKA_ROUND_1(AX, 0, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 2, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 4, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 6, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 8, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 10, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 12, X8, X9, X10, X11)
+	BLAMKA_ROUND_1(AX, 14, X8, X9, X10, X11)
+	RET
+
+// func mixBlocksSSE2(out, a, b, c *block)
+TEXT ·mixBlocksSSE2(SB), 4, $0-32
+	MOVQ out+0(FP), DX
+	MOVQ a+8(FP), AX
+	MOVQ b+16(FP), BX
+	MOVQ a+24(FP), CX
+	MOVQ $128, BP
+
+loop:
+	MOVOU 0(AX), X0
+	MOVOU 0(BX), X1
+	MOVOU 0(CX), X2
+	PXOR  X1, X0
+	PXOR  X2, X0
+	MOVOU X0, 0(DX)
+	ADDQ  $16, AX
+	ADDQ  $16, BX
+	ADDQ  $16, CX
+	ADDQ  $16, DX
+	SUBQ  $2, BP
+	JA    loop
+	RET
+
+// func xorBlocksSSE2(out, a, b, c *block)
+TEXT ·xorBlocksSSE2(SB), 4, $0-32
+	MOVQ out+0(FP), DX
+	MOVQ a+8(FP), AX
+	MOVQ b+16(FP), BX
+	MOVQ a+24(FP), CX
+	MOVQ $128, BP
+
+loop:
+	MOVOU 0(AX), X0
+	MOVOU 0(BX), X1
+	MOVOU 0(CX), X2
+	MOVOU 0(DX), X3
+	PXOR  X1, X0
+	PXOR  X2, X0
+	PXOR  X3, X0
+	MOVOU X0, 0(DX)
+	ADDQ  $16, AX
+	ADDQ  $16, BX
+	ADDQ  $16, CX
+	ADDQ  $16, DX
+	SUBQ  $2, BP
+	JA    loop
+	RET
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_generic.go b/vendor/golang.org/x/crypto/argon2/blamka_generic.go
new file mode 100644
index 00000000..a481b224
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_generic.go
@@ -0,0 +1,163 @@
+// 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.
+
+package argon2
+
+var useSSE4 bool
+
+func processBlockGeneric(out, in1, in2 *block, xor bool) {
+	var t block
+	for i := range t {
+		t[i] = in1[i] ^ in2[i]
+	}
+	for i := 0; i < blockLength; i += 16 {
+		blamkaGeneric(
+			&t[i+0], &t[i+1], &t[i+2], &t[i+3],
+			&t[i+4], &t[i+5], &t[i+6], &t[i+7],
+			&t[i+8], &t[i+9], &t[i+10], &t[i+11],
+			&t[i+12], &t[i+13], &t[i+14], &t[i+15],
+		)
+	}
+	for i := 0; i < blockLength/8; i += 2 {
+		blamkaGeneric(
+			&t[i], &t[i+1], &t[16+i], &t[16+i+1],
+			&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
+			&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
+			&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
+		)
+	}
+	if xor {
+		for i := range t {
+			out[i] ^= in1[i] ^ in2[i] ^ t[i]
+		}
+	} else {
+		for i := range t {
+			out[i] = in1[i] ^ in2[i] ^ t[i]
+		}
+	}
+}
+
+func blamkaGeneric(t00, t01, t02, t03, t04, t05, t06, t07, t08, t09, t10, t11, t12, t13, t14, t15 *uint64) {
+	v00, v01, v02, v03 := *t00, *t01, *t02, *t03
+	v04, v05, v06, v07 := *t04, *t05, *t06, *t07
+	v08, v09, v10, v11 := *t08, *t09, *t10, *t11
+	v12, v13, v14, v15 := *t12, *t13, *t14, *t15
+
+	v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
+	v12 ^= v00
+	v12 = v12>>32 | v12<<32
+	v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
+	v04 ^= v08
+	v04 = v04>>24 | v04<<40
+
+	v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
+	v12 ^= v00
+	v12 = v12>>16 | v12<<48
+	v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
+	v04 ^= v08
+	v04 = v04>>63 | v04<<1
+
+	v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
+	v13 ^= v01
+	v13 = v13>>32 | v13<<32
+	v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
+	v05 ^= v09
+	v05 = v05>>24 | v05<<40
+
+	v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
+	v13 ^= v01
+	v13 = v13>>16 | v13<<48
+	v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
+	v05 ^= v09
+	v05 = v05>>63 | v05<<1
+
+	v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
+	v14 ^= v02
+	v14 = v14>>32 | v14<<32
+	v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
+	v06 ^= v10
+	v06 = v06>>24 | v06<<40
+
+	v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
+	v14 ^= v02
+	v14 = v14>>16 | v14<<48
+	v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
+	v06 ^= v10
+	v06 = v06>>63 | v06<<1
+
+	v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
+	v15 ^= v03
+	v15 = v15>>32 | v15<<32
+	v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
+	v07 ^= v11
+	v07 = v07>>24 | v07<<40
+
+	v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
+	v15 ^= v03
+	v15 = v15>>16 | v15<<48
+	v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
+	v07 ^= v11
+	v07 = v07>>63 | v07<<1
+
+	v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
+	v15 ^= v00
+	v15 = v15>>32 | v15<<32
+	v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
+	v05 ^= v10
+	v05 = v05>>24 | v05<<40
+
+	v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
+	v15 ^= v00
+	v15 = v15>>16 | v15<<48
+	v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
+	v05 ^= v10
+	v05 = v05>>63 | v05<<1
+
+	v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
+	v12 ^= v01
+	v12 = v12>>32 | v12<<32
+	v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
+	v06 ^= v11
+	v06 = v06>>24 | v06<<40
+
+	v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
+	v12 ^= v01
+	v12 = v12>>16 | v12<<48
+	v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
+	v06 ^= v11
+	v06 = v06>>63 | v06<<1
+
+	v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
+	v13 ^= v02
+	v13 = v13>>32 | v13<<32
+	v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
+	v07 ^= v08
+	v07 = v07>>24 | v07<<40
+
+	v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
+	v13 ^= v02
+	v13 = v13>>16 | v13<<48
+	v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
+	v07 ^= v08
+	v07 = v07>>63 | v07<<1
+
+	v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
+	v14 ^= v03
+	v14 = v14>>32 | v14<<32
+	v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
+	v04 ^= v09
+	v04 = v04>>24 | v04<<40
+
+	v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
+	v14 ^= v03
+	v14 = v14>>16 | v14<<48
+	v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
+	v04 ^= v09
+	v04 = v04>>63 | v04<<1
+
+	*t00, *t01, *t02, *t03 = v00, v01, v02, v03
+	*t04, *t05, *t06, *t07 = v04, v05, v06, v07
+	*t08, *t09, *t10, *t11 = v08, v09, v10, v11
+	*t12, *t13, *t14, *t15 = v12, v13, v14, v15
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_ref.go b/vendor/golang.org/x/crypto/argon2/blamka_ref.go
new file mode 100644
index 00000000..167c59d2
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_ref.go
@@ -0,0 +1,16 @@
+// 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.
+
+//go:build !amd64 || purego || !gc
+// +build !amd64 purego !gc
+
+package argon2
+
+func processBlock(out, in1, in2 *block) {
+	processBlockGeneric(out, in1, in2, false)
+}
+
+func processBlockXOR(out, in1, in2 *block) {
+	processBlockGeneric(out, in1, in2, true)
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2b.go b/vendor/golang.org/x/crypto/blake2b/blake2b.go
new file mode 100644
index 00000000..d2e98d42
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2b.go
@@ -0,0 +1,291 @@
+// Copyright 2016 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 blake2b implements the BLAKE2b hash algorithm defined by RFC 7693
+// and the extendable output function (XOF) BLAKE2Xb.
+//
+// BLAKE2b is optimized for 64-bit platforms—including NEON-enabled ARMs—and
+// produces digests of any size between 1 and 64 bytes.
+// For a detailed specification of BLAKE2b see https://blake2.net/blake2.pdf
+// and for BLAKE2Xb see https://blake2.net/blake2x.pdf
+//
+// If you aren't sure which function you need, use BLAKE2b (Sum512 or New512).
+// If you need a secret-key MAC (message authentication code), use the New512
+// function with a non-nil key.
+//
+// BLAKE2X is a construction to compute hash values larger than 64 bytes. It
+// can produce hash values between 0 and 4 GiB.
+package blake2b
+
+import (
+	"encoding/binary"
+	"errors"
+	"hash"
+)
+
+const (
+	// The blocksize of BLAKE2b in bytes.
+	BlockSize = 128
+	// The hash size of BLAKE2b-512 in bytes.
+	Size = 64
+	// The hash size of BLAKE2b-384 in bytes.
+	Size384 = 48
+	// The hash size of BLAKE2b-256 in bytes.
+	Size256 = 32
+)
+
+var (
+	useAVX2 bool
+	useAVX  bool
+	useSSE4 bool
+)
+
+var (
+	errKeySize  = errors.New("blake2b: invalid key size")
+	errHashSize = errors.New("blake2b: invalid hash size")
+)
+
+var iv = [8]uint64{
+	0x6a09e667f3bcc908, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
+	0x510e527fade682d1, 0x9b05688c2b3e6c1f, 0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
+}
+
+// Sum512 returns the BLAKE2b-512 checksum of the data.
+func Sum512(data []byte) [Size]byte {
+	var sum [Size]byte
+	checkSum(&sum, Size, data)
+	return sum
+}
+
+// Sum384 returns the BLAKE2b-384 checksum of the data.
+func Sum384(data []byte) [Size384]byte {
+	var sum [Size]byte
+	var sum384 [Size384]byte
+	checkSum(&sum, Size384, data)
+	copy(sum384[:], sum[:Size384])
+	return sum384
+}
+
+// Sum256 returns the BLAKE2b-256 checksum of the data.
+func Sum256(data []byte) [Size256]byte {
+	var sum [Size]byte
+	var sum256 [Size256]byte
+	checkSum(&sum, Size256, data)
+	copy(sum256[:], sum[:Size256])
+	return sum256
+}
+
+// New512 returns a new hash.Hash computing the BLAKE2b-512 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New512(key []byte) (hash.Hash, error) { return newDigest(Size, key) }
+
+// New384 returns a new hash.Hash computing the BLAKE2b-384 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New384(key []byte) (hash.Hash, error) { return newDigest(Size384, key) }
+
+// New256 returns a new hash.Hash computing the BLAKE2b-256 checksum. A non-nil
+// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+func New256(key []byte) (hash.Hash, error) { return newDigest(Size256, key) }
+
+// New returns a new hash.Hash computing the BLAKE2b checksum with a custom length.
+// A non-nil key turns the hash into a MAC. The key must be between zero and 64 bytes long.
+// The hash size can be a value between 1 and 64 but it is highly recommended to use
+// values equal or greater than:
+// - 32 if BLAKE2b is used as a hash function (The key is zero bytes long).
+// - 16 if BLAKE2b is used as a MAC function (The key is at least 16 bytes long).
+// When the key is nil, the returned hash.Hash implements BinaryMarshaler
+// and BinaryUnmarshaler for state (de)serialization as documented by hash.Hash.
+func New(size int, key []byte) (hash.Hash, error) { return newDigest(size, key) }
+
+func newDigest(hashSize int, key []byte) (*digest, error) {
+	if hashSize < 1 || hashSize > Size {
+		return nil, errHashSize
+	}
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	d := &digest{
+		size:   hashSize,
+		keyLen: len(key),
+	}
+	copy(d.key[:], key)
+	d.Reset()
+	return d, nil
+}
+
+func checkSum(sum *[Size]byte, hashSize int, data []byte) {
+	h := iv
+	h[0] ^= uint64(hashSize) | (1 << 16) | (1 << 24)
+	var c [2]uint64
+
+	if length := len(data); length > BlockSize {
+		n := length &^ (BlockSize - 1)
+		if length == n {
+			n -= BlockSize
+		}
+		hashBlocks(&h, &c, 0, data[:n])
+		data = data[n:]
+	}
+
+	var block [BlockSize]byte
+	offset := copy(block[:], data)
+	remaining := uint64(BlockSize - offset)
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
+
+	for i, v := range h[:(hashSize+7)/8] {
+		binary.LittleEndian.PutUint64(sum[8*i:], v)
+	}
+}
+
+type digest struct {
+	h      [8]uint64
+	c      [2]uint64
+	size   int
+	block  [BlockSize]byte
+	offset int
+
+	key    [BlockSize]byte
+	keyLen int
+}
+
+const (
+	magic         = "b2b"
+	marshaledSize = len(magic) + 8*8 + 2*8 + 1 + BlockSize + 1
+)
+
+func (d *digest) MarshalBinary() ([]byte, error) {
+	if d.keyLen != 0 {
+		return nil, errors.New("crypto/blake2b: cannot marshal MACs")
+	}
+	b := make([]byte, 0, marshaledSize)
+	b = append(b, magic...)
+	for i := 0; i < 8; i++ {
+		b = appendUint64(b, d.h[i])
+	}
+	b = appendUint64(b, d.c[0])
+	b = appendUint64(b, d.c[1])
+	// Maximum value for size is 64
+	b = append(b, byte(d.size))
+	b = append(b, d.block[:]...)
+	b = append(b, byte(d.offset))
+	return b, nil
+}
+
+func (d *digest) UnmarshalBinary(b []byte) error {
+	if len(b) < len(magic) || string(b[:len(magic)]) != magic {
+		return errors.New("crypto/blake2b: invalid hash state identifier")
+	}
+	if len(b) != marshaledSize {
+		return errors.New("crypto/blake2b: invalid hash state size")
+	}
+	b = b[len(magic):]
+	for i := 0; i < 8; i++ {
+		b, d.h[i] = consumeUint64(b)
+	}
+	b, d.c[0] = consumeUint64(b)
+	b, d.c[1] = consumeUint64(b)
+	d.size = int(b[0])
+	b = b[1:]
+	copy(d.block[:], b[:BlockSize])
+	b = b[BlockSize:]
+	d.offset = int(b[0])
+	return nil
+}
+
+func (d *digest) BlockSize() int { return BlockSize }
+
+func (d *digest) Size() int { return d.size }
+
+func (d *digest) Reset() {
+	d.h = iv
+	d.h[0] ^= uint64(d.size) | (uint64(d.keyLen) << 8) | (1 << 16) | (1 << 24)
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	if d.keyLen > 0 {
+		d.block = d.key
+		d.offset = BlockSize
+	}
+}
+
+func (d *digest) Write(p []byte) (n int, err error) {
+	n = len(p)
+
+	if d.offset > 0 {
+		remaining := BlockSize - d.offset
+		if n <= remaining {
+			d.offset += copy(d.block[d.offset:], p)
+			return
+		}
+		copy(d.block[d.offset:], p[:remaining])
+		hashBlocks(&d.h, &d.c, 0, d.block[:])
+		d.offset = 0
+		p = p[remaining:]
+	}
+
+	if length := len(p); length > BlockSize {
+		nn := length &^ (BlockSize - 1)
+		if length == nn {
+			nn -= BlockSize
+		}
+		hashBlocks(&d.h, &d.c, 0, p[:nn])
+		p = p[nn:]
+	}
+
+	if len(p) > 0 {
+		d.offset += copy(d.block[:], p)
+	}
+
+	return
+}
+
+func (d *digest) Sum(sum []byte) []byte {
+	var hash [Size]byte
+	d.finalize(&hash)
+	return append(sum, hash[:d.size]...)
+}
+
+func (d *digest) finalize(hash *[Size]byte) {
+	var block [BlockSize]byte
+	copy(block[:], d.block[:d.offset])
+	remaining := uint64(BlockSize - d.offset)
+
+	c := d.c
+	if c[0] < remaining {
+		c[1]--
+	}
+	c[0] -= remaining
+
+	h := d.h
+	hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
+
+	for i, v := range h {
+		binary.LittleEndian.PutUint64(hash[8*i:], v)
+	}
+}
+
+func appendUint64(b []byte, x uint64) []byte {
+	var a [8]byte
+	binary.BigEndian.PutUint64(a[:], x)
+	return append(b, a[:]...)
+}
+
+func appendUint32(b []byte, x uint32) []byte {
+	var a [4]byte
+	binary.BigEndian.PutUint32(a[:], x)
+	return append(b, a[:]...)
+}
+
+func consumeUint64(b []byte) ([]byte, uint64) {
+	x := binary.BigEndian.Uint64(b)
+	return b[8:], x
+}
+
+func consumeUint32(b []byte) ([]byte, uint32) {
+	x := binary.BigEndian.Uint32(b)
+	return b[4:], x
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.go b/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.go
new file mode 100644
index 00000000..56bfaaa1
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.go
@@ -0,0 +1,38 @@
+// Copyright 2016 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 go1.7 && amd64 && gc && !purego
+// +build go1.7,amd64,gc,!purego
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useAVX2 = cpu.X86.HasAVX2
+	useAVX = cpu.X86.HasAVX
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	switch {
+	case useAVX2:
+		hashBlocksAVX2(h, c, flag, blocks)
+	case useAVX:
+		hashBlocksAVX(h, c, flag, blocks)
+	case useSSE4:
+		hashBlocksSSE4(h, c, flag, blocks)
+	default:
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.s b/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.s
new file mode 100644
index 00000000..4b9daa18
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.s
@@ -0,0 +1,745 @@
+// Copyright 2016 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 go1.7 && amd64 && gc && !purego
+// +build go1.7,amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA ·AVX2_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX2_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+DATA ·AVX2_iv0<>+0x10(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX2_iv0<>+0x18(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX2_iv0<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_iv1<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX2_iv1<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+DATA ·AVX2_iv1<>+0x10(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX2_iv1<>+0x18(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX2_iv1<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+DATA ·AVX2_c40<>+0x10(SB)/8, $0x0201000706050403
+DATA ·AVX2_c40<>+0x18(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX2_c40<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX2_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+DATA ·AVX2_c48<>+0x10(SB)/8, $0x0100070605040302
+DATA ·AVX2_c48<>+0x18(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX2_c48<>(SB), (NOPTR+RODATA), $32
+
+DATA ·AVX_iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·AVX_iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·AVX_iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·AVX_iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·AVX_iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·AVX_iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·AVX_iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·AVX_iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·AVX_iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·AVX_c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·AVX_c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·AVX_c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·AVX_c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·AVX_c48<>(SB), (NOPTR+RODATA), $16
+
+#define VPERMQ_0x39_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x39
+#define VPERMQ_0x93_Y1_Y1 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xc9; BYTE $0x93
+#define VPERMQ_0x4E_Y2_Y2 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xd2; BYTE $0x4e
+#define VPERMQ_0x93_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x93
+#define VPERMQ_0x39_Y3_Y3 BYTE $0xc4; BYTE $0xe3; BYTE $0xfd; BYTE $0x00; BYTE $0xdb; BYTE $0x39
+
+#define ROUND_AVX2(m0, m1, m2, m3, t, c40, c48) \
+	VPADDQ  m0, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m1, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y1_Y1;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y3_Y3;    \
+	VPADDQ  m2, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFD $-79, Y3, Y3; \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPSHUFB c40, Y1, Y1;  \
+	VPADDQ  m3, Y0, Y0;   \
+	VPADDQ  Y1, Y0, Y0;   \
+	VPXOR   Y0, Y3, Y3;   \
+	VPSHUFB c48, Y3, Y3;  \
+	VPADDQ  Y3, Y2, Y2;   \
+	VPXOR   Y2, Y1, Y1;   \
+	VPADDQ  Y1, Y1, t;    \
+	VPSRLQ  $63, Y1, Y1;  \
+	VPXOR   t, Y1, Y1;    \
+	VPERMQ_0x39_Y3_Y3;    \
+	VPERMQ_0x4E_Y2_Y2;    \
+	VPERMQ_0x93_Y1_Y1
+
+#define VMOVQ_SI_X11_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x1E
+#define VMOVQ_SI_X12_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x26
+#define VMOVQ_SI_X13_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x2E
+#define VMOVQ_SI_X14_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x36
+#define VMOVQ_SI_X15_0 BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x3E
+
+#define VMOVQ_SI_X11(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x5E; BYTE $n
+#define VMOVQ_SI_X12(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x66; BYTE $n
+#define VMOVQ_SI_X13(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x6E; BYTE $n
+#define VMOVQ_SI_X14(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x76; BYTE $n
+#define VMOVQ_SI_X15(n) BYTE $0xC5; BYTE $0x7A; BYTE $0x7E; BYTE $0x7E; BYTE $n
+
+#define VPINSRQ_1_SI_X11_0 BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x1E; BYTE $0x01
+#define VPINSRQ_1_SI_X12_0 BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x26; BYTE $0x01
+#define VPINSRQ_1_SI_X13_0 BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x2E; BYTE $0x01
+#define VPINSRQ_1_SI_X14_0 BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x36; BYTE $0x01
+#define VPINSRQ_1_SI_X15_0 BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x3E; BYTE $0x01
+
+#define VPINSRQ_1_SI_X11(n) BYTE $0xC4; BYTE $0x63; BYTE $0xA1; BYTE $0x22; BYTE $0x5E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X12(n) BYTE $0xC4; BYTE $0x63; BYTE $0x99; BYTE $0x22; BYTE $0x66; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X13(n) BYTE $0xC4; BYTE $0x63; BYTE $0x91; BYTE $0x22; BYTE $0x6E; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X14(n) BYTE $0xC4; BYTE $0x63; BYTE $0x89; BYTE $0x22; BYTE $0x76; BYTE $n; BYTE $0x01
+#define VPINSRQ_1_SI_X15(n) BYTE $0xC4; BYTE $0x63; BYTE $0x81; BYTE $0x22; BYTE $0x7E; BYTE $n; BYTE $0x01
+
+#define VMOVQ_R8_X15 BYTE $0xC4; BYTE $0x41; BYTE $0xF9; BYTE $0x6E; BYTE $0xF8
+#define VPINSRQ_1_R9_X15 BYTE $0xC4; BYTE $0x43; BYTE $0x81; BYTE $0x22; BYTE $0xF9; BYTE $0x01
+
+// load msg: Y12 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y12(i0, i1, i2, i3) \
+	VMOVQ_SI_X12(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X12(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y12, Y12
+
+// load msg: Y13 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y13(i0, i1, i2, i3) \
+	VMOVQ_SI_X13(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X13(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y13, Y13
+
+// load msg: Y14 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y14(i0, i1, i2, i3) \
+	VMOVQ_SI_X14(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X14(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y14, Y14
+
+// load msg: Y15 = (i0, i1, i2, i3)
+// i0, i1, i2, i3 must not be 0
+#define LOAD_MSG_AVX2_Y15(i0, i1, i2, i3) \
+	VMOVQ_SI_X15(i0*8);           \
+	VMOVQ_SI_X11(i2*8);           \
+	VPINSRQ_1_SI_X15(i1*8);       \
+	VPINSRQ_1_SI_X11(i3*8);       \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15() \
+	VMOVQ_SI_X12_0;                   \
+	VMOVQ_SI_X11(4*8);                \
+	VPINSRQ_1_SI_X12(2*8);            \
+	VPINSRQ_1_SI_X11(6*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(1, 3, 5, 7);    \
+	LOAD_MSG_AVX2_Y14(8, 10, 12, 14); \
+	LOAD_MSG_AVX2_Y15(9, 11, 13, 15)
+
+#define LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3() \
+	LOAD_MSG_AVX2_Y12(14, 4, 9, 13); \
+	LOAD_MSG_AVX2_Y13(10, 8, 15, 6); \
+	VMOVQ_SI_X11(11*8);              \
+	VPSHUFD     $0x4E, 0*8(SI), X14; \
+	VPINSRQ_1_SI_X11(5*8);           \
+	VINSERTI128 $1, X11, Y14, Y14;   \
+	LOAD_MSG_AVX2_Y15(12, 2, 7, 3)
+
+#define LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4() \
+	VMOVQ_SI_X11(5*8);              \
+	VMOVDQU     11*8(SI), X12;      \
+	VPINSRQ_1_SI_X11(15*8);         \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	VMOVQ_SI_X13(8*8);              \
+	VMOVQ_SI_X11(2*8);              \
+	VPINSRQ_1_SI_X13_0;             \
+	VPINSRQ_1_SI_X11(13*8);         \
+	VINSERTI128 $1, X11, Y13, Y13;  \
+	LOAD_MSG_AVX2_Y14(10, 3, 7, 9); \
+	LOAD_MSG_AVX2_Y15(14, 6, 1, 4)
+
+#define LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8() \
+	LOAD_MSG_AVX2_Y12(7, 3, 13, 11); \
+	LOAD_MSG_AVX2_Y13(9, 1, 12, 14); \
+	LOAD_MSG_AVX2_Y14(2, 5, 4, 15);  \
+	VMOVQ_SI_X15(6*8);               \
+	VMOVQ_SI_X11_0;                  \
+	VPINSRQ_1_SI_X15(10*8);          \
+	VPINSRQ_1_SI_X11(8*8);           \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13() \
+	LOAD_MSG_AVX2_Y12(9, 5, 2, 10);  \
+	VMOVQ_SI_X13_0;                  \
+	VMOVQ_SI_X11(4*8);               \
+	VPINSRQ_1_SI_X13(7*8);           \
+	VPINSRQ_1_SI_X11(15*8);          \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(14, 11, 6, 3); \
+	LOAD_MSG_AVX2_Y15(1, 12, 8, 13)
+
+#define LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9() \
+	VMOVQ_SI_X12(2*8);                \
+	VMOVQ_SI_X11_0;                   \
+	VPINSRQ_1_SI_X12(6*8);            \
+	VPINSRQ_1_SI_X11(8*8);            \
+	VINSERTI128 $1, X11, Y12, Y12;    \
+	LOAD_MSG_AVX2_Y13(12, 10, 11, 3); \
+	LOAD_MSG_AVX2_Y14(4, 7, 15, 1);   \
+	LOAD_MSG_AVX2_Y15(13, 5, 14, 9)
+
+#define LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11() \
+	LOAD_MSG_AVX2_Y12(12, 1, 14, 4);  \
+	LOAD_MSG_AVX2_Y13(5, 15, 13, 10); \
+	VMOVQ_SI_X14_0;                   \
+	VPSHUFD     $0x4E, 8*8(SI), X11;  \
+	VPINSRQ_1_SI_X14(6*8);            \
+	VINSERTI128 $1, X11, Y14, Y14;    \
+	LOAD_MSG_AVX2_Y15(7, 3, 2, 11)
+
+#define LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10() \
+	LOAD_MSG_AVX2_Y12(13, 7, 12, 3); \
+	LOAD_MSG_AVX2_Y13(11, 14, 1, 9); \
+	LOAD_MSG_AVX2_Y14(5, 15, 8, 2);  \
+	VMOVQ_SI_X15_0;                  \
+	VMOVQ_SI_X11(6*8);               \
+	VPINSRQ_1_SI_X15(4*8);           \
+	VPINSRQ_1_SI_X11(10*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5() \
+	VMOVQ_SI_X12(6*8);              \
+	VMOVQ_SI_X11(11*8);             \
+	VPINSRQ_1_SI_X12(14*8);         \
+	VPINSRQ_1_SI_X11_0;             \
+	VINSERTI128 $1, X11, Y12, Y12;  \
+	LOAD_MSG_AVX2_Y13(15, 9, 3, 8); \
+	VMOVQ_SI_X11(1*8);              \
+	VMOVDQU     12*8(SI), X14;      \
+	VPINSRQ_1_SI_X11(10*8);         \
+	VINSERTI128 $1, X11, Y14, Y14;  \
+	VMOVQ_SI_X15(2*8);              \
+	VMOVDQU     4*8(SI), X11;       \
+	VPINSRQ_1_SI_X15(7*8);          \
+	VINSERTI128 $1, X11, Y15, Y15
+
+#define LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0() \
+	LOAD_MSG_AVX2_Y12(10, 8, 7, 1);  \
+	VMOVQ_SI_X13(2*8);               \
+	VPSHUFD     $0x4E, 5*8(SI), X11; \
+	VPINSRQ_1_SI_X13(4*8);           \
+	VINSERTI128 $1, X11, Y13, Y13;   \
+	LOAD_MSG_AVX2_Y14(15, 9, 3, 13); \
+	VMOVQ_SI_X15(11*8);              \
+	VMOVQ_SI_X11(12*8);              \
+	VPINSRQ_1_SI_X15(14*8);          \
+	VPINSRQ_1_SI_X11_0;              \
+	VINSERTI128 $1, X11, Y15, Y15
+
+// func hashBlocksAVX2(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX2(SB), 4, $320-48 // frame size = 288 + 32 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, DX
+	ADDQ $31, DX
+	ANDQ $~31, DX
+
+	MOVQ CX, 16(DX)
+	XORQ CX, CX
+	MOVQ CX, 24(DX)
+
+	VMOVDQU ·AVX2_c40<>(SB), Y4
+	VMOVDQU ·AVX2_c48<>(SB), Y5
+
+	VMOVDQU 0(AX), Y8
+	VMOVDQU 32(AX), Y9
+	VMOVDQU ·AVX2_iv0<>(SB), Y6
+	VMOVDQU ·AVX2_iv1<>(SB), Y7
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+	MOVQ R9, 8(DX)
+
+loop:
+	ADDQ $128, R8
+	MOVQ R8, 0(DX)
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+	MOVQ R9, 8(DX)
+
+noinc:
+	VMOVDQA Y8, Y0
+	VMOVDQA Y9, Y1
+	VMOVDQA Y6, Y2
+	VPXOR   0(DX), Y7, Y3
+
+	LOAD_MSG_AVX2_0_2_4_6_1_3_5_7_8_10_12_14_9_11_13_15()
+	VMOVDQA Y12, 32(DX)
+	VMOVDQA Y13, 64(DX)
+	VMOVDQA Y14, 96(DX)
+	VMOVDQA Y15, 128(DX)
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_14_4_9_13_10_8_15_6_1_0_11_5_12_2_7_3()
+	VMOVDQA Y12, 160(DX)
+	VMOVDQA Y13, 192(DX)
+	VMOVDQA Y14, 224(DX)
+	VMOVDQA Y15, 256(DX)
+
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_11_12_5_15_8_0_2_13_10_3_7_9_14_6_1_4()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_7_3_13_11_9_1_12_14_2_5_4_15_6_10_0_8()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_9_5_2_10_0_7_4_15_14_11_6_3_1_12_8_13()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_2_6_0_8_12_10_11_3_4_7_15_1_13_5_14_9()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_12_1_14_4_5_15_13_10_0_6_9_8_7_3_2_11()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_13_7_12_3_11_14_1_9_5_15_8_2_0_4_6_10()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_6_14_11_0_15_9_3_8_12_13_1_10_2_7_4_5()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+	LOAD_MSG_AVX2_10_8_7_1_2_4_6_5_15_9_3_13_11_14_12_0()
+	ROUND_AVX2(Y12, Y13, Y14, Y15, Y10, Y4, Y5)
+
+	ROUND_AVX2(32(DX), 64(DX), 96(DX), 128(DX), Y10, Y4, Y5)
+	ROUND_AVX2(160(DX), 192(DX), 224(DX), 256(DX), Y10, Y4, Y5)
+
+	VPXOR Y0, Y8, Y8
+	VPXOR Y1, Y9, Y9
+	VPXOR Y2, Y8, Y8
+	VPXOR Y3, Y9, Y9
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	VMOVDQU Y8, 0(AX)
+	VMOVDQU Y9, 32(AX)
+	VZEROUPPER
+
+	RET
+
+#define VPUNPCKLQDQ_X2_X2_X15 BYTE $0xC5; BYTE $0x69; BYTE $0x6C; BYTE $0xFA
+#define VPUNPCKLQDQ_X3_X3_X15 BYTE $0xC5; BYTE $0x61; BYTE $0x6C; BYTE $0xFB
+#define VPUNPCKLQDQ_X7_X7_X15 BYTE $0xC5; BYTE $0x41; BYTE $0x6C; BYTE $0xFF
+#define VPUNPCKLQDQ_X13_X13_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x11; BYTE $0x6C; BYTE $0xFD
+#define VPUNPCKLQDQ_X14_X14_X15 BYTE $0xC4; BYTE $0x41; BYTE $0x09; BYTE $0x6C; BYTE $0xFE
+
+#define VPUNPCKHQDQ_X15_X2_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x69; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X3_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X6_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x49; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X7_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xFF
+#define VPUNPCKHQDQ_X15_X3_X2 BYTE $0xC4; BYTE $0xC1; BYTE $0x61; BYTE $0x6D; BYTE $0xD7
+#define VPUNPCKHQDQ_X15_X7_X6 BYTE $0xC4; BYTE $0xC1; BYTE $0x41; BYTE $0x6D; BYTE $0xF7
+#define VPUNPCKHQDQ_X15_X13_X3 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xDF
+#define VPUNPCKHQDQ_X15_X13_X7 BYTE $0xC4; BYTE $0xC1; BYTE $0x11; BYTE $0x6D; BYTE $0xFF
+
+#define SHUFFLE_AVX() \
+	VMOVDQA X6, X13;         \
+	VMOVDQA X2, X14;         \
+	VMOVDQA X4, X6;          \
+	VPUNPCKLQDQ_X13_X13_X15; \
+	VMOVDQA X5, X4;          \
+	VMOVDQA X6, X5;          \
+	VPUNPCKHQDQ_X15_X7_X6;   \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X13_X7;  \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VPUNPCKHQDQ_X15_X2_X2;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X3_X3;   \
+
+#define SHUFFLE_AVX_INV() \
+	VMOVDQA X2, X13;         \
+	VMOVDQA X4, X14;         \
+	VPUNPCKLQDQ_X2_X2_X15;   \
+	VMOVDQA X5, X4;          \
+	VPUNPCKHQDQ_X15_X3_X2;   \
+	VMOVDQA X14, X5;         \
+	VPUNPCKLQDQ_X3_X3_X15;   \
+	VMOVDQA X6, X14;         \
+	VPUNPCKHQDQ_X15_X13_X3;  \
+	VPUNPCKLQDQ_X7_X7_X15;   \
+	VPUNPCKHQDQ_X15_X6_X6;   \
+	VPUNPCKLQDQ_X14_X14_X15; \
+	VPUNPCKHQDQ_X15_X7_X7;   \
+
+#define HALF_ROUND_AVX(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	VPADDQ  m0, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m1, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFD $-79, v6, v6; \
+	VPSHUFD $-79, v7, v7; \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPSHUFB c40, v2, v2;  \
+	VPSHUFB c40, v3, v3;  \
+	VPADDQ  m2, v0, v0;   \
+	VPADDQ  v2, v0, v0;   \
+	VPADDQ  m3, v1, v1;   \
+	VPADDQ  v3, v1, v1;   \
+	VPXOR   v0, v6, v6;   \
+	VPXOR   v1, v7, v7;   \
+	VPSHUFB c48, v6, v6;  \
+	VPSHUFB c48, v7, v7;  \
+	VPADDQ  v6, v4, v4;   \
+	VPADDQ  v7, v5, v5;   \
+	VPXOR   v4, v2, v2;   \
+	VPXOR   v5, v3, v3;   \
+	VPADDQ  v2, v2, t0;   \
+	VPSRLQ  $63, v2, v2;  \
+	VPXOR   t0, v2, v2;   \
+	VPADDQ  v3, v3, t0;   \
+	VPSRLQ  $63, v3, v3;  \
+	VPXOR   t0, v3, v3
+
+// load msg: X12 = (i0, i1), X13 = (i2, i3), X14 = (i4, i5), X15 = (i6, i7)
+// i0, i1, i2, i3, i4, i5, i6, i7 must not be 0
+#define LOAD_MSG_AVX(i0, i1, i2, i3, i4, i5, i6, i7) \
+	VMOVQ_SI_X12(i0*8);     \
+	VMOVQ_SI_X13(i2*8);     \
+	VMOVQ_SI_X14(i4*8);     \
+	VMOVQ_SI_X15(i6*8);     \
+	VPINSRQ_1_SI_X12(i1*8); \
+	VPINSRQ_1_SI_X13(i3*8); \
+	VPINSRQ_1_SI_X14(i5*8); \
+	VPINSRQ_1_SI_X15(i7*8)
+
+// load msg: X12 = (0, 2), X13 = (4, 6), X14 = (1, 3), X15 = (5, 7)
+#define LOAD_MSG_AVX_0_2_4_6_1_3_5_7() \
+	VMOVQ_SI_X12_0;        \
+	VMOVQ_SI_X13(4*8);     \
+	VMOVQ_SI_X14(1*8);     \
+	VMOVQ_SI_X15(5*8);     \
+	VPINSRQ_1_SI_X12(2*8); \
+	VPINSRQ_1_SI_X13(6*8); \
+	VPINSRQ_1_SI_X14(3*8); \
+	VPINSRQ_1_SI_X15(7*8)
+
+// load msg: X12 = (1, 0), X13 = (11, 5), X14 = (12, 2), X15 = (7, 3)
+#define LOAD_MSG_AVX_1_0_11_5_12_2_7_3() \
+	VPSHUFD $0x4E, 0*8(SI), X12; \
+	VMOVQ_SI_X13(11*8);          \
+	VMOVQ_SI_X14(12*8);          \
+	VMOVQ_SI_X15(7*8);           \
+	VPINSRQ_1_SI_X13(5*8);       \
+	VPINSRQ_1_SI_X14(2*8);       \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (11, 12), X13 = (5, 15), X14 = (8, 0), X15 = (2, 13)
+#define LOAD_MSG_AVX_11_12_5_15_8_0_2_13() \
+	VMOVDQU 11*8(SI), X12;  \
+	VMOVQ_SI_X13(5*8);      \
+	VMOVQ_SI_X14(8*8);      \
+	VMOVQ_SI_X15(2*8);      \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14_0;     \
+	VPINSRQ_1_SI_X15(13*8)
+
+// load msg: X12 = (2, 5), X13 = (4, 15), X14 = (6, 10), X15 = (0, 8)
+#define LOAD_MSG_AVX_2_5_4_15_6_10_0_8() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13(4*8);      \
+	VMOVQ_SI_X14(6*8);      \
+	VMOVQ_SI_X15_0;         \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(15*8); \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (9, 5), X13 = (2, 10), X14 = (0, 7), X15 = (4, 15)
+#define LOAD_MSG_AVX_9_5_2_10_0_7_4_15() \
+	VMOVQ_SI_X12(9*8);      \
+	VMOVQ_SI_X13(2*8);      \
+	VMOVQ_SI_X14_0;         \
+	VMOVQ_SI_X15(4*8);      \
+	VPINSRQ_1_SI_X12(5*8);  \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VPINSRQ_1_SI_X15(15*8)
+
+// load msg: X12 = (2, 6), X13 = (0, 8), X14 = (12, 10), X15 = (11, 3)
+#define LOAD_MSG_AVX_2_6_0_8_12_10_11_3() \
+	VMOVQ_SI_X12(2*8);      \
+	VMOVQ_SI_X13_0;         \
+	VMOVQ_SI_X14(12*8);     \
+	VMOVQ_SI_X15(11*8);     \
+	VPINSRQ_1_SI_X12(6*8);  \
+	VPINSRQ_1_SI_X13(8*8);  \
+	VPINSRQ_1_SI_X14(10*8); \
+	VPINSRQ_1_SI_X15(3*8)
+
+// load msg: X12 = (0, 6), X13 = (9, 8), X14 = (7, 3), X15 = (2, 11)
+#define LOAD_MSG_AVX_0_6_9_8_7_3_2_11() \
+	MOVQ    0*8(SI), X12;        \
+	VPSHUFD $0x4E, 8*8(SI), X13; \
+	MOVQ    7*8(SI), X14;        \
+	MOVQ    2*8(SI), X15;        \
+	VPINSRQ_1_SI_X12(6*8);       \
+	VPINSRQ_1_SI_X14(3*8);       \
+	VPINSRQ_1_SI_X15(11*8)
+
+// load msg: X12 = (6, 14), X13 = (11, 0), X14 = (15, 9), X15 = (3, 8)
+#define LOAD_MSG_AVX_6_14_11_0_15_9_3_8() \
+	MOVQ 6*8(SI), X12;      \
+	MOVQ 11*8(SI), X13;     \
+	MOVQ 15*8(SI), X14;     \
+	MOVQ 3*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(14*8); \
+	VPINSRQ_1_SI_X13_0;     \
+	VPINSRQ_1_SI_X14(9*8);  \
+	VPINSRQ_1_SI_X15(8*8)
+
+// load msg: X12 = (5, 15), X13 = (8, 2), X14 = (0, 4), X15 = (6, 10)
+#define LOAD_MSG_AVX_5_15_8_2_0_4_6_10() \
+	MOVQ 5*8(SI), X12;      \
+	MOVQ 8*8(SI), X13;      \
+	MOVQ 0*8(SI), X14;      \
+	MOVQ 6*8(SI), X15;      \
+	VPINSRQ_1_SI_X12(15*8); \
+	VPINSRQ_1_SI_X13(2*8);  \
+	VPINSRQ_1_SI_X14(4*8);  \
+	VPINSRQ_1_SI_X15(10*8)
+
+// load msg: X12 = (12, 13), X13 = (1, 10), X14 = (2, 7), X15 = (4, 5)
+#define LOAD_MSG_AVX_12_13_1_10_2_7_4_5() \
+	VMOVDQU 12*8(SI), X12;  \
+	MOVQ    1*8(SI), X13;   \
+	MOVQ    2*8(SI), X14;   \
+	VPINSRQ_1_SI_X13(10*8); \
+	VPINSRQ_1_SI_X14(7*8);  \
+	VMOVDQU 4*8(SI), X15
+
+// load msg: X12 = (15, 9), X13 = (3, 13), X14 = (11, 14), X15 = (12, 0)
+#define LOAD_MSG_AVX_15_9_3_13_11_14_12_0() \
+	MOVQ 15*8(SI), X12;     \
+	MOVQ 3*8(SI), X13;      \
+	MOVQ 11*8(SI), X14;     \
+	MOVQ 12*8(SI), X15;     \
+	VPINSRQ_1_SI_X12(9*8);  \
+	VPINSRQ_1_SI_X13(13*8); \
+	VPINSRQ_1_SI_X14(14*8); \
+	VPINSRQ_1_SI_X15_0
+
+// func hashBlocksAVX(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksAVX(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, R10
+	ADDQ $15, R10
+	ANDQ $~15, R10
+
+	VMOVDQU ·AVX_c40<>(SB), X0
+	VMOVDQU ·AVX_c48<>(SB), X1
+	VMOVDQA X0, X8
+	VMOVDQA X1, X9
+
+	VMOVDQU ·AVX_iv3<>(SB), X0
+	VMOVDQA X0, 0(R10)
+	XORQ    CX, 0(R10)          // 0(R10) = ·AVX_iv3 ^ (CX || 0)
+
+	VMOVDQU 0(AX), X10
+	VMOVDQU 16(AX), X11
+	VMOVDQU 32(AX), X2
+	VMOVDQU 48(AX), X3
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	VMOVQ_R8_X15
+	VPINSRQ_1_R9_X15
+
+	VMOVDQA X10, X0
+	VMOVDQA X11, X1
+	VMOVDQU ·AVX_iv0<>(SB), X4
+	VMOVDQU ·AVX_iv1<>(SB), X5
+	VMOVDQU ·AVX_iv2<>(SB), X6
+
+	VPXOR   X15, X6, X6
+	VMOVDQA 0(R10), X7
+
+	LOAD_MSG_AVX_0_2_4_6_1_3_5_7()
+	VMOVDQA X12, 16(R10)
+	VMOVDQA X13, 32(R10)
+	VMOVDQA X14, 48(R10)
+	VMOVDQA X15, 64(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(8, 10, 12, 14, 9, 11, 13, 15)
+	VMOVDQA X12, 80(R10)
+	VMOVDQA X13, 96(R10)
+	VMOVDQA X14, 112(R10)
+	VMOVDQA X15, 128(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(14, 4, 9, 13, 10, 8, 15, 6)
+	VMOVDQA X12, 144(R10)
+	VMOVDQA X13, 160(R10)
+	VMOVDQA X14, 176(R10)
+	VMOVDQA X15, 192(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_1_0_11_5_12_2_7_3()
+	VMOVDQA X12, 208(R10)
+	VMOVDQA X13, 224(R10)
+	VMOVDQA X14, 240(R10)
+	VMOVDQA X15, 256(R10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_11_12_5_15_8_0_2_13()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_2_5_4_15_6_10_0_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_9_5_2_10_0_7_4_15()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_2_6_0_8_12_10_11_3()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX(4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_0_6_9_8_7_3_2_11()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_5_15_8_2_0_4_6_10()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX_6_14_11_0_15_9_3_8()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_12_13_1_10_2_7_4_5()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	LOAD_MSG_AVX(10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX()
+	LOAD_MSG_AVX_15_9_3_13_11_14_12_0()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, X12, X13, X14, X15, X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X15, X8, X9)
+	SHUFFLE_AVX()
+	HALF_ROUND_AVX(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X15, X8, X9)
+	SHUFFLE_AVX_INV()
+
+	VMOVDQU 32(AX), X14
+	VMOVDQU 48(AX), X15
+	VPXOR   X0, X10, X10
+	VPXOR   X1, X11, X11
+	VPXOR   X2, X14, X14
+	VPXOR   X3, X15, X15
+	VPXOR   X4, X10, X10
+	VPXOR   X5, X11, X11
+	VPXOR   X6, X14, X2
+	VPXOR   X7, X15, X3
+	VMOVDQU X2, 32(AX)
+	VMOVDQU X3, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	VMOVDQU X10, 0(AX)
+	VMOVDQU X11, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+	VZEROUPPER
+
+	RET
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.go b/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.go
new file mode 100644
index 00000000..5fa1b328
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.go
@@ -0,0 +1,25 @@
+// Copyright 2016 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 !go1.7 && amd64 && gc && !purego
+// +build !go1.7,amd64,gc,!purego
+
+package blake2b
+
+import "golang.org/x/sys/cpu"
+
+func init() {
+	useSSE4 = cpu.X86.HasSSE41
+}
+
+//go:noescape
+func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	if useSSE4 {
+		hashBlocksSSE4(h, c, flag, blocks)
+	} else {
+		hashBlocksGeneric(h, c, flag, blocks)
+	}
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.s b/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.s
new file mode 100644
index 00000000..ae75eb9a
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2b_amd64.s
@@ -0,0 +1,279 @@
+// Copyright 2016 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 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+DATA ·iv0<>+0x00(SB)/8, $0x6a09e667f3bcc908
+DATA ·iv0<>+0x08(SB)/8, $0xbb67ae8584caa73b
+GLOBL ·iv0<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv1<>+0x00(SB)/8, $0x3c6ef372fe94f82b
+DATA ·iv1<>+0x08(SB)/8, $0xa54ff53a5f1d36f1
+GLOBL ·iv1<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv2<>+0x00(SB)/8, $0x510e527fade682d1
+DATA ·iv2<>+0x08(SB)/8, $0x9b05688c2b3e6c1f
+GLOBL ·iv2<>(SB), (NOPTR+RODATA), $16
+
+DATA ·iv3<>+0x00(SB)/8, $0x1f83d9abfb41bd6b
+DATA ·iv3<>+0x08(SB)/8, $0x5be0cd19137e2179
+GLOBL ·iv3<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
+DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
+GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
+
+DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
+DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
+GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
+
+#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v6, t1; \
+	PUNPCKLQDQ v6, t2; \
+	PUNPCKHQDQ v7, v6; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ v7, t2; \
+	MOVO       t1, v7; \
+	MOVO       v2, t1; \
+	PUNPCKHQDQ t2, v7; \
+	PUNPCKLQDQ v3, t2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v3
+
+#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
+	MOVO       v4, t1; \
+	MOVO       v5, v4; \
+	MOVO       t1, v5; \
+	MOVO       v2, t1; \
+	PUNPCKLQDQ v2, t2; \
+	PUNPCKHQDQ v3, v2; \
+	PUNPCKHQDQ t2, v2; \
+	PUNPCKLQDQ v3, t2; \
+	MOVO       t1, v3; \
+	MOVO       v6, t1; \
+	PUNPCKHQDQ t2, v3; \
+	PUNPCKLQDQ v7, t2; \
+	PUNPCKHQDQ t2, v6; \
+	PUNPCKLQDQ t1, t2; \
+	PUNPCKHQDQ t2, v7
+
+#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, m0, m1, m2, m3, t0, c40, c48) \
+	PADDQ  m0, v0;        \
+	PADDQ  m1, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFD $0xB1, v6, v6; \
+	PSHUFD $0xB1, v7, v7; \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	PSHUFB c40, v2;       \
+	PSHUFB c40, v3;       \
+	PADDQ  m2, v0;        \
+	PADDQ  m3, v1;        \
+	PADDQ  v2, v0;        \
+	PADDQ  v3, v1;        \
+	PXOR   v0, v6;        \
+	PXOR   v1, v7;        \
+	PSHUFB c48, v6;       \
+	PSHUFB c48, v7;       \
+	PADDQ  v6, v4;        \
+	PADDQ  v7, v5;        \
+	PXOR   v4, v2;        \
+	PXOR   v5, v3;        \
+	MOVOU  v2, t0;        \
+	PADDQ  v2, t0;        \
+	PSRLQ  $63, v2;       \
+	PXOR   t0, v2;        \
+	MOVOU  v3, t0;        \
+	PADDQ  v3, t0;        \
+	PSRLQ  $63, v3;       \
+	PXOR   t0, v3
+
+#define LOAD_MSG(m0, m1, m2, m3, src, i0, i1, i2, i3, i4, i5, i6, i7) \
+	MOVQ   i0*8(src), m0;     \
+	PINSRQ $1, i1*8(src), m0; \
+	MOVQ   i2*8(src), m1;     \
+	PINSRQ $1, i3*8(src), m1; \
+	MOVQ   i4*8(src), m2;     \
+	PINSRQ $1, i5*8(src), m2; \
+	MOVQ   i6*8(src), m3;     \
+	PINSRQ $1, i7*8(src), m3
+
+// func hashBlocksSSE4(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte)
+TEXT ·hashBlocksSSE4(SB), 4, $288-48 // frame size = 272 + 16 byte alignment
+	MOVQ h+0(FP), AX
+	MOVQ c+8(FP), BX
+	MOVQ flag+16(FP), CX
+	MOVQ blocks_base+24(FP), SI
+	MOVQ blocks_len+32(FP), DI
+
+	MOVQ SP, R10
+	ADDQ $15, R10
+	ANDQ $~15, R10
+
+	MOVOU ·iv3<>(SB), X0
+	MOVO  X0, 0(R10)
+	XORQ  CX, 0(R10)     // 0(R10) = ·iv3 ^ (CX || 0)
+
+	MOVOU ·c40<>(SB), X13
+	MOVOU ·c48<>(SB), X14
+
+	MOVOU 0(AX), X12
+	MOVOU 16(AX), X15
+
+	MOVQ 0(BX), R8
+	MOVQ 8(BX), R9
+
+loop:
+	ADDQ $128, R8
+	CMPQ R8, $128
+	JGE  noinc
+	INCQ R9
+
+noinc:
+	MOVQ R8, X8
+	PINSRQ $1, R9, X8
+
+	MOVO X12, X0
+	MOVO X15, X1
+	MOVOU 32(AX), X2
+	MOVOU 48(AX), X3
+	MOVOU ·iv0<>(SB), X4
+	MOVOU ·iv1<>(SB), X5
+	MOVOU ·iv2<>(SB), X6
+
+	PXOR X8, X6
+	MOVO 0(R10), X7
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 2, 4, 6, 1, 3, 5, 7)
+	MOVO X8, 16(R10)
+	MOVO X9, 32(R10)
+	MOVO X10, 48(R10)
+	MOVO X11, 64(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 8, 10, 12, 14, 9, 11, 13, 15)
+	MOVO X8, 80(R10)
+	MOVO X9, 96(R10)
+	MOVO X10, 112(R10)
+	MOVO X11, 128(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 4, 9, 13, 10, 8, 15, 6)
+	MOVO X8, 144(R10)
+	MOVO X9, 160(R10)
+	MOVO X10, 176(R10)
+	MOVO X11, 192(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 1, 0, 11, 5, 12, 2, 7, 3)
+	MOVO X8, 208(R10)
+	MOVO X9, 224(R10)
+	MOVO X10, 240(R10)
+	MOVO X11, 256(R10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 11, 12, 5, 15, 8, 0, 2, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 3, 7, 9, 14, 6, 1, 4)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 7, 3, 13, 11, 9, 1, 12, 14)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 5, 4, 15, 6, 10, 0, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 9, 5, 2, 10, 0, 7, 4, 15)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 14, 11, 6, 3, 1, 12, 8, 13)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 2, 6, 0, 8, 12, 10, 11, 3)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 4, 7, 15, 1, 13, 5, 14, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 1, 14, 4, 5, 15, 13, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 0, 6, 9, 8, 7, 3, 2, 11)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 13, 7, 12, 3, 11, 14, 1, 9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 5, 15, 8, 2, 0, 4, 6, 10)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 6, 14, 11, 0, 15, 9, 3, 8)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 12, 13, 1, 10, 2, 7, 4, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	LOAD_MSG(X8, X9, X10, X11, SI, 10, 8, 7, 1, 2, 4, 6, 5)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	LOAD_MSG(X8, X9, X10, X11, SI, 15, 9, 3, 13, 11, 14, 12, 0)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 16(R10), 32(R10), 48(R10), 64(R10), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 80(R10), 96(R10), 112(R10), 128(R10), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 144(R10), 160(R10), 176(R10), 192(R10), X11, X13, X14)
+	SHUFFLE(X2, X3, X4, X5, X6, X7, X8, X9)
+	HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, 208(R10), 224(R10), 240(R10), 256(R10), X11, X13, X14)
+	SHUFFLE_INV(X2, X3, X4, X5, X6, X7, X8, X9)
+
+	MOVOU 32(AX), X10
+	MOVOU 48(AX), X11
+	PXOR  X0, X12
+	PXOR  X1, X15
+	PXOR  X2, X10
+	PXOR  X3, X11
+	PXOR  X4, X12
+	PXOR  X5, X15
+	PXOR  X6, X10
+	PXOR  X7, X11
+	MOVOU X10, 32(AX)
+	MOVOU X11, 48(AX)
+
+	LEAQ 128(SI), SI
+	SUBQ $128, DI
+	JNE  loop
+
+	MOVOU X12, 0(AX)
+	MOVOU X15, 16(AX)
+
+	MOVQ R8, 0(BX)
+	MOVQ R9, 8(BX)
+
+	RET
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2b_generic.go b/vendor/golang.org/x/crypto/blake2b/blake2b_generic.go
new file mode 100644
index 00000000..3168a8aa
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2b_generic.go
@@ -0,0 +1,182 @@
+// Copyright 2016 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 blake2b
+
+import (
+	"encoding/binary"
+	"math/bits"
+)
+
+// the precomputed values for BLAKE2b
+// there are 12 16-byte arrays - one for each round
+// the entries are calculated from the sigma constants.
+var precomputed = [12][16]byte{
+	{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15},
+	{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3},
+	{11, 12, 5, 15, 8, 0, 2, 13, 10, 3, 7, 9, 14, 6, 1, 4},
+	{7, 3, 13, 11, 9, 1, 12, 14, 2, 5, 4, 15, 6, 10, 0, 8},
+	{9, 5, 2, 10, 0, 7, 4, 15, 14, 11, 6, 3, 1, 12, 8, 13},
+	{2, 6, 0, 8, 12, 10, 11, 3, 4, 7, 15, 1, 13, 5, 14, 9},
+	{12, 1, 14, 4, 5, 15, 13, 10, 0, 6, 9, 8, 7, 3, 2, 11},
+	{13, 7, 12, 3, 11, 14, 1, 9, 5, 15, 8, 2, 0, 4, 6, 10},
+	{6, 14, 11, 0, 15, 9, 3, 8, 12, 13, 1, 10, 2, 7, 4, 5},
+	{10, 8, 7, 1, 2, 4, 6, 5, 15, 9, 3, 13, 11, 14, 12, 0},
+	{0, 2, 4, 6, 1, 3, 5, 7, 8, 10, 12, 14, 9, 11, 13, 15}, // equal to the first
+	{14, 4, 9, 13, 10, 8, 15, 6, 1, 0, 11, 5, 12, 2, 7, 3}, // equal to the second
+}
+
+func hashBlocksGeneric(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	var m [16]uint64
+	c0, c1 := c[0], c[1]
+
+	for i := 0; i < len(blocks); {
+		c0 += BlockSize
+		if c0 < BlockSize {
+			c1++
+		}
+
+		v0, v1, v2, v3, v4, v5, v6, v7 := h[0], h[1], h[2], h[3], h[4], h[5], h[6], h[7]
+		v8, v9, v10, v11, v12, v13, v14, v15 := iv[0], iv[1], iv[2], iv[3], iv[4], iv[5], iv[6], iv[7]
+		v12 ^= c0
+		v13 ^= c1
+		v14 ^= flag
+
+		for j := range m {
+			m[j] = binary.LittleEndian.Uint64(blocks[i:])
+			i += 8
+		}
+
+		for j := range precomputed {
+			s := &(precomputed[j])
+
+			v0 += m[s[0]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft64(v12, -32)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft64(v4, -24)
+			v1 += m[s[1]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft64(v13, -32)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft64(v5, -24)
+			v2 += m[s[2]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft64(v14, -32)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft64(v6, -24)
+			v3 += m[s[3]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft64(v15, -32)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft64(v7, -24)
+
+			v0 += m[s[4]]
+			v0 += v4
+			v12 ^= v0
+			v12 = bits.RotateLeft64(v12, -16)
+			v8 += v12
+			v4 ^= v8
+			v4 = bits.RotateLeft64(v4, -63)
+			v1 += m[s[5]]
+			v1 += v5
+			v13 ^= v1
+			v13 = bits.RotateLeft64(v13, -16)
+			v9 += v13
+			v5 ^= v9
+			v5 = bits.RotateLeft64(v5, -63)
+			v2 += m[s[6]]
+			v2 += v6
+			v14 ^= v2
+			v14 = bits.RotateLeft64(v14, -16)
+			v10 += v14
+			v6 ^= v10
+			v6 = bits.RotateLeft64(v6, -63)
+			v3 += m[s[7]]
+			v3 += v7
+			v15 ^= v3
+			v15 = bits.RotateLeft64(v15, -16)
+			v11 += v15
+			v7 ^= v11
+			v7 = bits.RotateLeft64(v7, -63)
+
+			v0 += m[s[8]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft64(v15, -32)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft64(v5, -24)
+			v1 += m[s[9]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft64(v12, -32)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft64(v6, -24)
+			v2 += m[s[10]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft64(v13, -32)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft64(v7, -24)
+			v3 += m[s[11]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft64(v14, -32)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft64(v4, -24)
+
+			v0 += m[s[12]]
+			v0 += v5
+			v15 ^= v0
+			v15 = bits.RotateLeft64(v15, -16)
+			v10 += v15
+			v5 ^= v10
+			v5 = bits.RotateLeft64(v5, -63)
+			v1 += m[s[13]]
+			v1 += v6
+			v12 ^= v1
+			v12 = bits.RotateLeft64(v12, -16)
+			v11 += v12
+			v6 ^= v11
+			v6 = bits.RotateLeft64(v6, -63)
+			v2 += m[s[14]]
+			v2 += v7
+			v13 ^= v2
+			v13 = bits.RotateLeft64(v13, -16)
+			v8 += v13
+			v7 ^= v8
+			v7 = bits.RotateLeft64(v7, -63)
+			v3 += m[s[15]]
+			v3 += v4
+			v14 ^= v3
+			v14 = bits.RotateLeft64(v14, -16)
+			v9 += v14
+			v4 ^= v9
+			v4 = bits.RotateLeft64(v4, -63)
+
+		}
+
+		h[0] ^= v0 ^ v8
+		h[1] ^= v1 ^ v9
+		h[2] ^= v2 ^ v10
+		h[3] ^= v3 ^ v11
+		h[4] ^= v4 ^ v12
+		h[5] ^= v5 ^ v13
+		h[6] ^= v6 ^ v14
+		h[7] ^= v7 ^ v15
+	}
+	c[0], c[1] = c0, c1
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2b_ref.go b/vendor/golang.org/x/crypto/blake2b/blake2b_ref.go
new file mode 100644
index 00000000..b0137cdf
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2b_ref.go
@@ -0,0 +1,12 @@
+// Copyright 2016 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 || purego || !gc
+// +build !amd64 purego !gc
+
+package blake2b
+
+func hashBlocks(h *[8]uint64, c *[2]uint64, flag uint64, blocks []byte) {
+	hashBlocksGeneric(h, c, flag, blocks)
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/blake2x.go b/vendor/golang.org/x/crypto/blake2b/blake2x.go
new file mode 100644
index 00000000..52c414db
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/blake2x.go
@@ -0,0 +1,177 @@
+// 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.
+
+package blake2b
+
+import (
+	"encoding/binary"
+	"errors"
+	"io"
+)
+
+// XOF defines the interface to hash functions that
+// support arbitrary-length output.
+type XOF interface {
+	// Write absorbs more data into the hash's state. It panics if called
+	// after Read.
+	io.Writer
+
+	// Read reads more output from the hash. It returns io.EOF if the limit
+	// has been reached.
+	io.Reader
+
+	// Clone returns a copy of the XOF in its current state.
+	Clone() XOF
+
+	// Reset resets the XOF to its initial state.
+	Reset()
+}
+
+// OutputLengthUnknown can be used as the size argument to NewXOF to indicate
+// the length of the output is not known in advance.
+const OutputLengthUnknown = 0
+
+// magicUnknownOutputLength is a magic value for the output size that indicates
+// an unknown number of output bytes.
+const magicUnknownOutputLength = (1 << 32) - 1
+
+// maxOutputLength is the absolute maximum number of bytes to produce when the
+// number of output bytes is unknown.
+const maxOutputLength = (1 << 32) * 64
+
+// NewXOF creates a new variable-output-length hash. The hash either produce a
+// known number of bytes (1 <= size < 2**32-1), or an unknown number of bytes
+// (size == OutputLengthUnknown). In the latter case, an absolute limit of
+// 256GiB applies.
+//
+// A non-nil key turns the hash into a MAC. The key must between
+// zero and 32 bytes long.
+func NewXOF(size uint32, key []byte) (XOF, error) {
+	if len(key) > Size {
+		return nil, errKeySize
+	}
+	if size == magicUnknownOutputLength {
+		// 2^32-1 indicates an unknown number of bytes and thus isn't a
+		// valid length.
+		return nil, errors.New("blake2b: XOF length too large")
+	}
+	if size == OutputLengthUnknown {
+		size = magicUnknownOutputLength
+	}
+	x := &xof{
+		d: digest{
+			size:   Size,
+			keyLen: len(key),
+		},
+		length: size,
+	}
+	copy(x.d.key[:], key)
+	x.Reset()
+	return x, nil
+}
+
+type xof struct {
+	d                digest
+	length           uint32
+	remaining        uint64
+	cfg, root, block [Size]byte
+	offset           int
+	nodeOffset       uint32
+	readMode         bool
+}
+
+func (x *xof) Write(p []byte) (n int, err error) {
+	if x.readMode {
+		panic("blake2b: write to XOF after read")
+	}
+	return x.d.Write(p)
+}
+
+func (x *xof) Clone() XOF {
+	clone := *x
+	return &clone
+}
+
+func (x *xof) Reset() {
+	x.cfg[0] = byte(Size)
+	binary.LittleEndian.PutUint32(x.cfg[4:], uint32(Size)) // leaf length
+	binary.LittleEndian.PutUint32(x.cfg[12:], x.length)    // XOF length
+	x.cfg[17] = byte(Size)                                 // inner hash size
+
+	x.d.Reset()
+	x.d.h[1] ^= uint64(x.length) << 32
+
+	x.remaining = uint64(x.length)
+	if x.remaining == magicUnknownOutputLength {
+		x.remaining = maxOutputLength
+	}
+	x.offset, x.nodeOffset = 0, 0
+	x.readMode = false
+}
+
+func (x *xof) Read(p []byte) (n int, err error) {
+	if !x.readMode {
+		x.d.finalize(&x.root)
+		x.readMode = true
+	}
+
+	if x.remaining == 0 {
+		return 0, io.EOF
+	}
+
+	n = len(p)
+	if uint64(n) > x.remaining {
+		n = int(x.remaining)
+		p = p[:n]
+	}
+
+	if x.offset > 0 {
+		blockRemaining := Size - x.offset
+		if n < blockRemaining {
+			x.offset += copy(p, x.block[x.offset:])
+			x.remaining -= uint64(n)
+			return
+		}
+		copy(p, x.block[x.offset:])
+		p = p[blockRemaining:]
+		x.offset = 0
+		x.remaining -= uint64(blockRemaining)
+	}
+
+	for len(p) >= Size {
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		copy(p, x.block[:])
+		p = p[Size:]
+		x.remaining -= uint64(Size)
+	}
+
+	if todo := len(p); todo > 0 {
+		if x.remaining < uint64(Size) {
+			x.cfg[0] = byte(x.remaining)
+		}
+		binary.LittleEndian.PutUint32(x.cfg[8:], x.nodeOffset)
+		x.nodeOffset++
+
+		x.d.initConfig(&x.cfg)
+		x.d.Write(x.root[:])
+		x.d.finalize(&x.block)
+
+		x.offset = copy(p, x.block[:todo])
+		x.remaining -= uint64(todo)
+	}
+	return
+}
+
+func (d *digest) initConfig(cfg *[Size]byte) {
+	d.offset, d.c[0], d.c[1] = 0, 0, 0
+	for i := range d.h {
+		d.h[i] = iv[i] ^ binary.LittleEndian.Uint64(cfg[i*8:])
+	}
+}
diff --git a/vendor/golang.org/x/crypto/blake2b/register.go b/vendor/golang.org/x/crypto/blake2b/register.go
new file mode 100644
index 00000000..9d863396
--- /dev/null
+++ b/vendor/golang.org/x/crypto/blake2b/register.go
@@ -0,0 +1,33 @@
+// 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.
+
+//go:build go1.9
+// +build go1.9
+
+package blake2b
+
+import (
+	"crypto"
+	"hash"
+)
+
+func init() {
+	newHash256 := func() hash.Hash {
+		h, _ := New256(nil)
+		return h
+	}
+	newHash384 := func() hash.Hash {
+		h, _ := New384(nil)
+		return h
+	}
+
+	newHash512 := func() hash.Hash {
+		h, _ := New512(nil)
+		return h
+	}
+
+	crypto.RegisterHash(crypto.BLAKE2b_256, newHash256)
+	crypto.RegisterHash(crypto.BLAKE2b_384, newHash384)
+	crypto.RegisterHash(crypto.BLAKE2b_512, newHash512)
+}
diff --git a/vendor/golang.org/x/crypto/chacha20/chacha_arm64.s b/vendor/golang.org/x/crypto/chacha20/chacha_arm64.s
index 8fb49a13..63cae9e6 100644
--- a/vendor/golang.org/x/crypto/chacha20/chacha_arm64.s
+++ b/vendor/golang.org/x/crypto/chacha20/chacha_arm64.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build go1.11 && gc && !purego
 // +build go1.11,gc,!purego
 
 #include "textflag.h"
diff --git a/vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.s b/vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.s
index 3dad4b2f..5c0fed26 100644
--- a/vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.s
+++ b/vendor/golang.org/x/crypto/chacha20/chacha_ppc64le.s
@@ -19,6 +19,7 @@
 // The differences in this and the original implementation are
 // due to the calling conventions and initialization of constants.
 
+//go:build gc && !purego
 // +build gc,!purego
 
 #include "textflag.h"
diff --git a/vendor/golang.org/x/crypto/chacha20/chacha_s390x.s b/vendor/golang.org/x/crypto/chacha20/chacha_s390x.s
index 81816118..f3ef5a01 100644
--- a/vendor/golang.org/x/crypto/chacha20/chacha_s390x.s
+++ b/vendor/golang.org/x/crypto/chacha20/chacha_s390x.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build gc && !purego
 // +build gc,!purego
 
 #include "go_asm.h"
diff --git a/vendor/golang.org/x/crypto/curve25519/curve25519.go b/vendor/golang.org/x/crypto/curve25519/curve25519.go
index 4b9a655d..cda3fdd3 100644
--- a/vendor/golang.org/x/crypto/curve25519/curve25519.go
+++ b/vendor/golang.org/x/crypto/curve25519/curve25519.go
@@ -10,6 +10,8 @@ package curve25519 // import "golang.org/x/crypto/curve25519"
 import (
 	"crypto/subtle"
 	"fmt"
+
+	"golang.org/x/crypto/curve25519/internal/field"
 )
 
 // ScalarMult sets dst to the product scalar * point.
@@ -18,7 +20,55 @@ import (
 // zeroes, irrespective of the scalar. Instead, use the X25519 function, which
 // will return an error.
 func ScalarMult(dst, scalar, point *[32]byte) {
-	scalarMult(dst, scalar, point)
+	var e [32]byte
+
+	copy(e[:], scalar[:])
+	e[0] &= 248
+	e[31] &= 127
+	e[31] |= 64
+
+	var x1, x2, z2, x3, z3, tmp0, tmp1 field.Element
+	x1.SetBytes(point[:])
+	x2.One()
+	x3.Set(&x1)
+	z3.One()
+
+	swap := 0
+	for pos := 254; pos >= 0; pos-- {
+		b := e[pos/8] >> uint(pos&7)
+		b &= 1
+		swap ^= int(b)
+		x2.Swap(&x3, swap)
+		z2.Swap(&z3, swap)
+		swap = int(b)
+
+		tmp0.Subtract(&x3, &z3)
+		tmp1.Subtract(&x2, &z2)
+		x2.Add(&x2, &z2)
+		z2.Add(&x3, &z3)
+		z3.Multiply(&tmp0, &x2)
+		z2.Multiply(&z2, &tmp1)
+		tmp0.Square(&tmp1)
+		tmp1.Square(&x2)
+		x3.Add(&z3, &z2)
+		z2.Subtract(&z3, &z2)
+		x2.Multiply(&tmp1, &tmp0)
+		tmp1.Subtract(&tmp1, &tmp0)
+		z2.Square(&z2)
+
+		z3.Mult32(&tmp1, 121666)
+		x3.Square(&x3)
+		tmp0.Add(&tmp0, &z3)
+		z3.Multiply(&x1, &z2)
+		z2.Multiply(&tmp1, &tmp0)
+	}
+
+	x2.Swap(&x3, swap)
+	z2.Swap(&z3, swap)
+
+	z2.Invert(&z2)
+	x2.Multiply(&x2, &z2)
+	copy(dst[:], x2.Bytes())
 }
 
 // ScalarBaseMult sets dst to the product scalar * base where base is the
diff --git a/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.go b/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.go
deleted file mode 100644
index 84858480..00000000
--- a/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.go
+++ /dev/null
@@ -1,241 +0,0 @@
-// Copyright 2012 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 && gc && !purego
-// +build amd64,gc,!purego
-
-package curve25519
-
-// These functions are implemented in the .s files. The names of the functions
-// in the rest of the file are also taken from the SUPERCOP sources to help
-// people following along.
-
-//go:noescape
-
-func cswap(inout *[5]uint64, v uint64)
-
-//go:noescape
-
-func ladderstep(inout *[5][5]uint64)
-
-//go:noescape
-
-func freeze(inout *[5]uint64)
-
-//go:noescape
-
-func mul(dest, a, b *[5]uint64)
-
-//go:noescape
-
-func square(out, in *[5]uint64)
-
-// mladder uses a Montgomery ladder to calculate (xr/zr) *= s.
-func mladder(xr, zr *[5]uint64, s *[32]byte) {
-	var work [5][5]uint64
-
-	work[0] = *xr
-	setint(&work[1], 1)
-	setint(&work[2], 0)
-	work[3] = *xr
-	setint(&work[4], 1)
-
-	j := uint(6)
-	var prevbit byte
-
-	for i := 31; i >= 0; i-- {
-		for j < 8 {
-			bit := ((*s)[i] >> j) & 1
-			swap := bit ^ prevbit
-			prevbit = bit
-			cswap(&work[1], uint64(swap))
-			ladderstep(&work)
-			j--
-		}
-		j = 7
-	}
-
-	*xr = work[1]
-	*zr = work[2]
-}
-
-func scalarMult(out, in, base *[32]byte) {
-	var e [32]byte
-	copy(e[:], (*in)[:])
-	e[0] &= 248
-	e[31] &= 127
-	e[31] |= 64
-
-	var t, z [5]uint64
-	unpack(&t, base)
-	mladder(&t, &z, &e)
-	invert(&z, &z)
-	mul(&t, &t, &z)
-	pack(out, &t)
-}
-
-func setint(r *[5]uint64, v uint64) {
-	r[0] = v
-	r[1] = 0
-	r[2] = 0
-	r[3] = 0
-	r[4] = 0
-}
-
-// unpack sets r = x where r consists of 5, 51-bit limbs in little-endian
-// order.
-func unpack(r *[5]uint64, x *[32]byte) {
-	r[0] = uint64(x[0]) |
-		uint64(x[1])<<8 |
-		uint64(x[2])<<16 |
-		uint64(x[3])<<24 |
-		uint64(x[4])<<32 |
-		uint64(x[5])<<40 |
-		uint64(x[6]&7)<<48
-
-	r[1] = uint64(x[6])>>3 |
-		uint64(x[7])<<5 |
-		uint64(x[8])<<13 |
-		uint64(x[9])<<21 |
-		uint64(x[10])<<29 |
-		uint64(x[11])<<37 |
-		uint64(x[12]&63)<<45
-
-	r[2] = uint64(x[12])>>6 |
-		uint64(x[13])<<2 |
-		uint64(x[14])<<10 |
-		uint64(x[15])<<18 |
-		uint64(x[16])<<26 |
-		uint64(x[17])<<34 |
-		uint64(x[18])<<42 |
-		uint64(x[19]&1)<<50
-
-	r[3] = uint64(x[19])>>1 |
-		uint64(x[20])<<7 |
-		uint64(x[21])<<15 |
-		uint64(x[22])<<23 |
-		uint64(x[23])<<31 |
-		uint64(x[24])<<39 |
-		uint64(x[25]&15)<<47
-
-	r[4] = uint64(x[25])>>4 |
-		uint64(x[26])<<4 |
-		uint64(x[27])<<12 |
-		uint64(x[28])<<20 |
-		uint64(x[29])<<28 |
-		uint64(x[30])<<36 |
-		uint64(x[31]&127)<<44
-}
-
-// pack sets out = x where out is the usual, little-endian form of the 5,
-// 51-bit limbs in x.
-func pack(out *[32]byte, x *[5]uint64) {
-	t := *x
-	freeze(&t)
-
-	out[0] = byte(t[0])
-	out[1] = byte(t[0] >> 8)
-	out[2] = byte(t[0] >> 16)
-	out[3] = byte(t[0] >> 24)
-	out[4] = byte(t[0] >> 32)
-	out[5] = byte(t[0] >> 40)
-	out[6] = byte(t[0] >> 48)
-
-	out[6] ^= byte(t[1]<<3) & 0xf8
-	out[7] = byte(t[1] >> 5)
-	out[8] = byte(t[1] >> 13)
-	out[9] = byte(t[1] >> 21)
-	out[10] = byte(t[1] >> 29)
-	out[11] = byte(t[1] >> 37)
-	out[12] = byte(t[1] >> 45)
-
-	out[12] ^= byte(t[2]<<6) & 0xc0
-	out[13] = byte(t[2] >> 2)
-	out[14] = byte(t[2] >> 10)
-	out[15] = byte(t[2] >> 18)
-	out[16] = byte(t[2] >> 26)
-	out[17] = byte(t[2] >> 34)
-	out[18] = byte(t[2] >> 42)
-	out[19] = byte(t[2] >> 50)
-
-	out[19] ^= byte(t[3]<<1) & 0xfe
-	out[20] = byte(t[3] >> 7)
-	out[21] = byte(t[3] >> 15)
-	out[22] = byte(t[3] >> 23)
-	out[23] = byte(t[3] >> 31)
-	out[24] = byte(t[3] >> 39)
-	out[25] = byte(t[3] >> 47)
-
-	out[25] ^= byte(t[4]<<4) & 0xf0
-	out[26] = byte(t[4] >> 4)
-	out[27] = byte(t[4] >> 12)
-	out[28] = byte(t[4] >> 20)
-	out[29] = byte(t[4] >> 28)
-	out[30] = byte(t[4] >> 36)
-	out[31] = byte(t[4] >> 44)
-}
-
-// invert calculates r = x^-1 mod p using Fermat's little theorem.
-func invert(r *[5]uint64, x *[5]uint64) {
-	var z2, z9, z11, z2_5_0, z2_10_0, z2_20_0, z2_50_0, z2_100_0, t [5]uint64
-
-	square(&z2, x)        /* 2 */
-	square(&t, &z2)       /* 4 */
-	square(&t, &t)        /* 8 */
-	mul(&z9, &t, x)       /* 9 */
-	mul(&z11, &z9, &z2)   /* 11 */
-	square(&t, &z11)      /* 22 */
-	mul(&z2_5_0, &t, &z9) /* 2^5 - 2^0 = 31 */
-
-	square(&t, &z2_5_0)      /* 2^6 - 2^1 */
-	for i := 1; i < 5; i++ { /* 2^20 - 2^10 */
-		square(&t, &t)
-	}
-	mul(&z2_10_0, &t, &z2_5_0) /* 2^10 - 2^0 */
-
-	square(&t, &z2_10_0)      /* 2^11 - 2^1 */
-	for i := 1; i < 10; i++ { /* 2^20 - 2^10 */
-		square(&t, &t)
-	}
-	mul(&z2_20_0, &t, &z2_10_0) /* 2^20 - 2^0 */
-
-	square(&t, &z2_20_0)      /* 2^21 - 2^1 */
-	for i := 1; i < 20; i++ { /* 2^40 - 2^20 */
-		square(&t, &t)
-	}
-	mul(&t, &t, &z2_20_0) /* 2^40 - 2^0 */
-
-	square(&t, &t)            /* 2^41 - 2^1 */
-	for i := 1; i < 10; i++ { /* 2^50 - 2^10 */
-		square(&t, &t)
-	}
-	mul(&z2_50_0, &t, &z2_10_0) /* 2^50 - 2^0 */
-
-	square(&t, &z2_50_0)      /* 2^51 - 2^1 */
-	for i := 1; i < 50; i++ { /* 2^100 - 2^50 */
-		square(&t, &t)
-	}
-	mul(&z2_100_0, &t, &z2_50_0) /* 2^100 - 2^0 */
-
-	square(&t, &z2_100_0)      /* 2^101 - 2^1 */
-	for i := 1; i < 100; i++ { /* 2^200 - 2^100 */
-		square(&t, &t)
-	}
-	mul(&t, &t, &z2_100_0) /* 2^200 - 2^0 */
-
-	square(&t, &t)            /* 2^201 - 2^1 */
-	for i := 1; i < 50; i++ { /* 2^250 - 2^50 */
-		square(&t, &t)
-	}
-	mul(&t, &t, &z2_50_0) /* 2^250 - 2^0 */
-
-	square(&t, &t) /* 2^251 - 2^1 */
-	square(&t, &t) /* 2^252 - 2^2 */
-	square(&t, &t) /* 2^253 - 2^3 */
-
-	square(&t, &t) /* 2^254 - 2^4 */
-
-	square(&t, &t)   /* 2^255 - 2^5 */
-	mul(r, &t, &z11) /* 2^255 - 21 */
-}
diff --git a/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.s b/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.s
deleted file mode 100644
index 6c533809..00000000
--- a/vendor/golang.org/x/crypto/curve25519/curve25519_amd64.s
+++ /dev/null
@@ -1,1793 +0,0 @@
-// Copyright 2012 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.
-
-// This code was translated into a form compatible with 6a from the public
-// domain sources in SUPERCOP: https://bench.cr.yp.to/supercop.html
-
-// +build amd64,gc,!purego
-
-#define REDMASK51     0x0007FFFFFFFFFFFF
-
-// These constants cannot be encoded in non-MOVQ immediates.
-// We access them directly from memory instead.
-
-DATA ·_121666_213(SB)/8, $996687872
-GLOBL ·_121666_213(SB), 8, $8
-
-DATA ·_2P0(SB)/8, $0xFFFFFFFFFFFDA
-GLOBL ·_2P0(SB), 8, $8
-
-DATA ·_2P1234(SB)/8, $0xFFFFFFFFFFFFE
-GLOBL ·_2P1234(SB), 8, $8
-
-// func freeze(inout *[5]uint64)
-TEXT ·freeze(SB),7,$0-8
-	MOVQ inout+0(FP), DI
-
-	MOVQ 0(DI),SI
-	MOVQ 8(DI),DX
-	MOVQ 16(DI),CX
-	MOVQ 24(DI),R8
-	MOVQ 32(DI),R9
-	MOVQ $REDMASK51,AX
-	MOVQ AX,R10
-	SUBQ $18,R10
-	MOVQ $3,R11
-REDUCELOOP:
-	MOVQ SI,R12
-	SHRQ $51,R12
-	ANDQ AX,SI
-	ADDQ R12,DX
-	MOVQ DX,R12
-	SHRQ $51,R12
-	ANDQ AX,DX
-	ADDQ R12,CX
-	MOVQ CX,R12
-	SHRQ $51,R12
-	ANDQ AX,CX
-	ADDQ R12,R8
-	MOVQ R8,R12
-	SHRQ $51,R12
-	ANDQ AX,R8
-	ADDQ R12,R9
-	MOVQ R9,R12
-	SHRQ $51,R12
-	ANDQ AX,R9
-	IMUL3Q $19,R12,R12
-	ADDQ R12,SI
-	SUBQ $1,R11
-	JA REDUCELOOP
-	MOVQ $1,R12
-	CMPQ R10,SI
-	CMOVQLT R11,R12
-	CMPQ AX,DX
-	CMOVQNE R11,R12
-	CMPQ AX,CX
-	CMOVQNE R11,R12
-	CMPQ AX,R8
-	CMOVQNE R11,R12
-	CMPQ AX,R9
-	CMOVQNE R11,R12
-	NEGQ R12
-	ANDQ R12,AX
-	ANDQ R12,R10
-	SUBQ R10,SI
-	SUBQ AX,DX
-	SUBQ AX,CX
-	SUBQ AX,R8
-	SUBQ AX,R9
-	MOVQ SI,0(DI)
-	MOVQ DX,8(DI)
-	MOVQ CX,16(DI)
-	MOVQ R8,24(DI)
-	MOVQ R9,32(DI)
-	RET
-
-// func ladderstep(inout *[5][5]uint64)
-TEXT ·ladderstep(SB),0,$296-8
-	MOVQ inout+0(FP),DI
-
-	MOVQ 40(DI),SI
-	MOVQ 48(DI),DX
-	MOVQ 56(DI),CX
-	MOVQ 64(DI),R8
-	MOVQ 72(DI),R9
-	MOVQ SI,AX
-	MOVQ DX,R10
-	MOVQ CX,R11
-	MOVQ R8,R12
-	MOVQ R9,R13
-	ADDQ ·_2P0(SB),AX
-	ADDQ ·_2P1234(SB),R10
-	ADDQ ·_2P1234(SB),R11
-	ADDQ ·_2P1234(SB),R12
-	ADDQ ·_2P1234(SB),R13
-	ADDQ 80(DI),SI
-	ADDQ 88(DI),DX
-	ADDQ 96(DI),CX
-	ADDQ 104(DI),R8
-	ADDQ 112(DI),R9
-	SUBQ 80(DI),AX
-	SUBQ 88(DI),R10
-	SUBQ 96(DI),R11
-	SUBQ 104(DI),R12
-	SUBQ 112(DI),R13
-	MOVQ SI,0(SP)
-	MOVQ DX,8(SP)
-	MOVQ CX,16(SP)
-	MOVQ R8,24(SP)
-	MOVQ R9,32(SP)
-	MOVQ AX,40(SP)
-	MOVQ R10,48(SP)
-	MOVQ R11,56(SP)
-	MOVQ R12,64(SP)
-	MOVQ R13,72(SP)
-	MOVQ 40(SP),AX
-	MULQ 40(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 40(SP),AX
-	SHLQ $1,AX
-	MULQ 48(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 40(SP),AX
-	SHLQ $1,AX
-	MULQ 56(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 40(SP),AX
-	SHLQ $1,AX
-	MULQ 64(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 40(SP),AX
-	SHLQ $1,AX
-	MULQ 72(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 48(SP),AX
-	MULQ 48(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 48(SP),AX
-	SHLQ $1,AX
-	MULQ 56(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 48(SP),AX
-	SHLQ $1,AX
-	MULQ 64(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 48(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 56(SP),AX
-	MULQ 56(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 56(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 64(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 56(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 64(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 64(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 64(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 72(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	ANDQ DX,SI
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ADDQ R10,CX
-	ANDQ DX,R8
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ADDQ R12,CX
-	ANDQ DX,R9
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ADDQ R14,CX
-	ANDQ DX,AX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,80(SP)
-	MOVQ R8,88(SP)
-	MOVQ R9,96(SP)
-	MOVQ AX,104(SP)
-	MOVQ R10,112(SP)
-	MOVQ 0(SP),AX
-	MULQ 0(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 0(SP),AX
-	SHLQ $1,AX
-	MULQ 8(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 0(SP),AX
-	SHLQ $1,AX
-	MULQ 16(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 0(SP),AX
-	SHLQ $1,AX
-	MULQ 24(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 0(SP),AX
-	SHLQ $1,AX
-	MULQ 32(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 8(SP),AX
-	MULQ 8(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	SHLQ $1,AX
-	MULQ 16(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 8(SP),AX
-	SHLQ $1,AX
-	MULQ 24(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 16(SP),AX
-	MULQ 16(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 16(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 24(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 16(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 24(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 24(SP),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 32(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	ANDQ DX,SI
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ADDQ R10,CX
-	ANDQ DX,R8
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ADDQ R12,CX
-	ANDQ DX,R9
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ADDQ R14,CX
-	ANDQ DX,AX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,120(SP)
-	MOVQ R8,128(SP)
-	MOVQ R9,136(SP)
-	MOVQ AX,144(SP)
-	MOVQ R10,152(SP)
-	MOVQ SI,SI
-	MOVQ R8,DX
-	MOVQ R9,CX
-	MOVQ AX,R8
-	MOVQ R10,R9
-	ADDQ ·_2P0(SB),SI
-	ADDQ ·_2P1234(SB),DX
-	ADDQ ·_2P1234(SB),CX
-	ADDQ ·_2P1234(SB),R8
-	ADDQ ·_2P1234(SB),R9
-	SUBQ 80(SP),SI
-	SUBQ 88(SP),DX
-	SUBQ 96(SP),CX
-	SUBQ 104(SP),R8
-	SUBQ 112(SP),R9
-	MOVQ SI,160(SP)
-	MOVQ DX,168(SP)
-	MOVQ CX,176(SP)
-	MOVQ R8,184(SP)
-	MOVQ R9,192(SP)
-	MOVQ 120(DI),SI
-	MOVQ 128(DI),DX
-	MOVQ 136(DI),CX
-	MOVQ 144(DI),R8
-	MOVQ 152(DI),R9
-	MOVQ SI,AX
-	MOVQ DX,R10
-	MOVQ CX,R11
-	MOVQ R8,R12
-	MOVQ R9,R13
-	ADDQ ·_2P0(SB),AX
-	ADDQ ·_2P1234(SB),R10
-	ADDQ ·_2P1234(SB),R11
-	ADDQ ·_2P1234(SB),R12
-	ADDQ ·_2P1234(SB),R13
-	ADDQ 160(DI),SI
-	ADDQ 168(DI),DX
-	ADDQ 176(DI),CX
-	ADDQ 184(DI),R8
-	ADDQ 192(DI),R9
-	SUBQ 160(DI),AX
-	SUBQ 168(DI),R10
-	SUBQ 176(DI),R11
-	SUBQ 184(DI),R12
-	SUBQ 192(DI),R13
-	MOVQ SI,200(SP)
-	MOVQ DX,208(SP)
-	MOVQ CX,216(SP)
-	MOVQ R8,224(SP)
-	MOVQ R9,232(SP)
-	MOVQ AX,240(SP)
-	MOVQ R10,248(SP)
-	MOVQ R11,256(SP)
-	MOVQ R12,264(SP)
-	MOVQ R13,272(SP)
-	MOVQ 224(SP),SI
-	IMUL3Q $19,SI,AX
-	MOVQ AX,280(SP)
-	MULQ 56(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 232(SP),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,288(SP)
-	MULQ 48(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 200(SP),AX
-	MULQ 40(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 200(SP),AX
-	MULQ 48(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 200(SP),AX
-	MULQ 56(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 200(SP),AX
-	MULQ 64(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 200(SP),AX
-	MULQ 72(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 208(SP),AX
-	MULQ 40(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 208(SP),AX
-	MULQ 48(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 208(SP),AX
-	MULQ 56(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 208(SP),AX
-	MULQ 64(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 208(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 216(SP),AX
-	MULQ 40(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 216(SP),AX
-	MULQ 48(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 216(SP),AX
-	MULQ 56(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 216(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 64(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 216(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 72(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 224(SP),AX
-	MULQ 40(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 224(SP),AX
-	MULQ 48(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 280(SP),AX
-	MULQ 64(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 280(SP),AX
-	MULQ 72(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 232(SP),AX
-	MULQ 40(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 288(SP),AX
-	MULQ 56(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 288(SP),AX
-	MULQ 64(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 288(SP),AX
-	MULQ 72(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ANDQ DX,SI
-	ADDQ R10,CX
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ANDQ DX,R8
-	ADDQ R12,CX
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ANDQ DX,R9
-	ADDQ R14,CX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	ANDQ DX,AX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,40(SP)
-	MOVQ R8,48(SP)
-	MOVQ R9,56(SP)
-	MOVQ AX,64(SP)
-	MOVQ R10,72(SP)
-	MOVQ 264(SP),SI
-	IMUL3Q $19,SI,AX
-	MOVQ AX,200(SP)
-	MULQ 16(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 272(SP),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,208(SP)
-	MULQ 8(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 240(SP),AX
-	MULQ 0(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 240(SP),AX
-	MULQ 8(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 240(SP),AX
-	MULQ 16(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 240(SP),AX
-	MULQ 24(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 240(SP),AX
-	MULQ 32(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 248(SP),AX
-	MULQ 0(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 248(SP),AX
-	MULQ 8(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 248(SP),AX
-	MULQ 16(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 248(SP),AX
-	MULQ 24(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 248(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 256(SP),AX
-	MULQ 0(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 256(SP),AX
-	MULQ 8(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 256(SP),AX
-	MULQ 16(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 256(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 256(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 264(SP),AX
-	MULQ 0(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 264(SP),AX
-	MULQ 8(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 200(SP),AX
-	MULQ 24(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 200(SP),AX
-	MULQ 32(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 272(SP),AX
-	MULQ 0(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 208(SP),AX
-	MULQ 16(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 208(SP),AX
-	MULQ 24(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 208(SP),AX
-	MULQ 32(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ANDQ DX,SI
-	ADDQ R10,CX
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ANDQ DX,R8
-	ADDQ R12,CX
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ANDQ DX,R9
-	ADDQ R14,CX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	ANDQ DX,AX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,DX
-	MOVQ R8,CX
-	MOVQ R9,R11
-	MOVQ AX,R12
-	MOVQ R10,R13
-	ADDQ ·_2P0(SB),DX
-	ADDQ ·_2P1234(SB),CX
-	ADDQ ·_2P1234(SB),R11
-	ADDQ ·_2P1234(SB),R12
-	ADDQ ·_2P1234(SB),R13
-	ADDQ 40(SP),SI
-	ADDQ 48(SP),R8
-	ADDQ 56(SP),R9
-	ADDQ 64(SP),AX
-	ADDQ 72(SP),R10
-	SUBQ 40(SP),DX
-	SUBQ 48(SP),CX
-	SUBQ 56(SP),R11
-	SUBQ 64(SP),R12
-	SUBQ 72(SP),R13
-	MOVQ SI,120(DI)
-	MOVQ R8,128(DI)
-	MOVQ R9,136(DI)
-	MOVQ AX,144(DI)
-	MOVQ R10,152(DI)
-	MOVQ DX,160(DI)
-	MOVQ CX,168(DI)
-	MOVQ R11,176(DI)
-	MOVQ R12,184(DI)
-	MOVQ R13,192(DI)
-	MOVQ 120(DI),AX
-	MULQ 120(DI)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 120(DI),AX
-	SHLQ $1,AX
-	MULQ 128(DI)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 120(DI),AX
-	SHLQ $1,AX
-	MULQ 136(DI)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 120(DI),AX
-	SHLQ $1,AX
-	MULQ 144(DI)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 120(DI),AX
-	SHLQ $1,AX
-	MULQ 152(DI)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 128(DI),AX
-	MULQ 128(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 128(DI),AX
-	SHLQ $1,AX
-	MULQ 136(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 128(DI),AX
-	SHLQ $1,AX
-	MULQ 144(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 128(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 152(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 136(DI),AX
-	MULQ 136(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 136(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 144(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 136(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 152(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 144(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 144(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 144(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 152(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 152(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 152(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	ANDQ DX,SI
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ADDQ R10,CX
-	ANDQ DX,R8
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ADDQ R12,CX
-	ANDQ DX,R9
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ADDQ R14,CX
-	ANDQ DX,AX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,120(DI)
-	MOVQ R8,128(DI)
-	MOVQ R9,136(DI)
-	MOVQ AX,144(DI)
-	MOVQ R10,152(DI)
-	MOVQ 160(DI),AX
-	MULQ 160(DI)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 160(DI),AX
-	SHLQ $1,AX
-	MULQ 168(DI)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 160(DI),AX
-	SHLQ $1,AX
-	MULQ 176(DI)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 160(DI),AX
-	SHLQ $1,AX
-	MULQ 184(DI)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 160(DI),AX
-	SHLQ $1,AX
-	MULQ 192(DI)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 168(DI),AX
-	MULQ 168(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 168(DI),AX
-	SHLQ $1,AX
-	MULQ 176(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 168(DI),AX
-	SHLQ $1,AX
-	MULQ 184(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 168(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 192(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 176(DI),AX
-	MULQ 176(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 176(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 184(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 176(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 192(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 184(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 184(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 184(DI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 192(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 192(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 192(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	ANDQ DX,SI
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ADDQ R10,CX
-	ANDQ DX,R8
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ADDQ R12,CX
-	ANDQ DX,R9
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ADDQ R14,CX
-	ANDQ DX,AX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,160(DI)
-	MOVQ R8,168(DI)
-	MOVQ R9,176(DI)
-	MOVQ AX,184(DI)
-	MOVQ R10,192(DI)
-	MOVQ 184(DI),SI
-	IMUL3Q $19,SI,AX
-	MOVQ AX,0(SP)
-	MULQ 16(DI)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 192(DI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,8(SP)
-	MULQ 8(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 160(DI),AX
-	MULQ 0(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 160(DI),AX
-	MULQ 8(DI)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 160(DI),AX
-	MULQ 16(DI)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 160(DI),AX
-	MULQ 24(DI)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 160(DI),AX
-	MULQ 32(DI)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 168(DI),AX
-	MULQ 0(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 168(DI),AX
-	MULQ 8(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 168(DI),AX
-	MULQ 16(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 168(DI),AX
-	MULQ 24(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 168(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 176(DI),AX
-	MULQ 0(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 176(DI),AX
-	MULQ 8(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 176(DI),AX
-	MULQ 16(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 176(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(DI)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 176(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 184(DI),AX
-	MULQ 0(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 184(DI),AX
-	MULQ 8(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 0(SP),AX
-	MULQ 24(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SP),AX
-	MULQ 32(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 192(DI),AX
-	MULQ 0(DI)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SP),AX
-	MULQ 16(DI)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 8(SP),AX
-	MULQ 24(DI)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	MULQ 32(DI)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ANDQ DX,SI
-	ADDQ R10,CX
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ANDQ DX,R8
-	ADDQ R12,CX
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ANDQ DX,R9
-	ADDQ R14,CX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	ANDQ DX,AX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,160(DI)
-	MOVQ R8,168(DI)
-	MOVQ R9,176(DI)
-	MOVQ AX,184(DI)
-	MOVQ R10,192(DI)
-	MOVQ 144(SP),SI
-	IMUL3Q $19,SI,AX
-	MOVQ AX,0(SP)
-	MULQ 96(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 152(SP),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,8(SP)
-	MULQ 88(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 120(SP),AX
-	MULQ 80(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 120(SP),AX
-	MULQ 88(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 120(SP),AX
-	MULQ 96(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 120(SP),AX
-	MULQ 104(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 120(SP),AX
-	MULQ 112(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 128(SP),AX
-	MULQ 80(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 128(SP),AX
-	MULQ 88(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 128(SP),AX
-	MULQ 96(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 128(SP),AX
-	MULQ 104(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 128(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 112(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 136(SP),AX
-	MULQ 80(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 136(SP),AX
-	MULQ 88(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 136(SP),AX
-	MULQ 96(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 136(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 104(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 136(SP),DX
-	IMUL3Q $19,DX,AX
-	MULQ 112(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 144(SP),AX
-	MULQ 80(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 144(SP),AX
-	MULQ 88(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 0(SP),AX
-	MULQ 104(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SP),AX
-	MULQ 112(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 152(SP),AX
-	MULQ 80(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SP),AX
-	MULQ 96(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 8(SP),AX
-	MULQ 104(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	MULQ 112(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ANDQ DX,SI
-	ADDQ R10,CX
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ANDQ DX,R8
-	ADDQ R12,CX
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ANDQ DX,R9
-	ADDQ R14,CX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	ANDQ DX,AX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,40(DI)
-	MOVQ R8,48(DI)
-	MOVQ R9,56(DI)
-	MOVQ AX,64(DI)
-	MOVQ R10,72(DI)
-	MOVQ 160(SP),AX
-	MULQ ·_121666_213(SB)
-	SHRQ $13,AX
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 168(SP),AX
-	MULQ ·_121666_213(SB)
-	SHRQ $13,AX
-	ADDQ AX,CX
-	MOVQ DX,R8
-	MOVQ 176(SP),AX
-	MULQ ·_121666_213(SB)
-	SHRQ $13,AX
-	ADDQ AX,R8
-	MOVQ DX,R9
-	MOVQ 184(SP),AX
-	MULQ ·_121666_213(SB)
-	SHRQ $13,AX
-	ADDQ AX,R9
-	MOVQ DX,R10
-	MOVQ 192(SP),AX
-	MULQ ·_121666_213(SB)
-	SHRQ $13,AX
-	ADDQ AX,R10
-	IMUL3Q $19,DX,DX
-	ADDQ DX,SI
-	ADDQ 80(SP),SI
-	ADDQ 88(SP),CX
-	ADDQ 96(SP),R8
-	ADDQ 104(SP),R9
-	ADDQ 112(SP),R10
-	MOVQ SI,80(DI)
-	MOVQ CX,88(DI)
-	MOVQ R8,96(DI)
-	MOVQ R9,104(DI)
-	MOVQ R10,112(DI)
-	MOVQ 104(DI),SI
-	IMUL3Q $19,SI,AX
-	MOVQ AX,0(SP)
-	MULQ 176(SP)
-	MOVQ AX,SI
-	MOVQ DX,CX
-	MOVQ 112(DI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,8(SP)
-	MULQ 168(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 80(DI),AX
-	MULQ 160(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 80(DI),AX
-	MULQ 168(SP)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 80(DI),AX
-	MULQ 176(SP)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 80(DI),AX
-	MULQ 184(SP)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 80(DI),AX
-	MULQ 192(SP)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 88(DI),AX
-	MULQ 160(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 88(DI),AX
-	MULQ 168(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 88(DI),AX
-	MULQ 176(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 88(DI),AX
-	MULQ 184(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 88(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 192(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 96(DI),AX
-	MULQ 160(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 96(DI),AX
-	MULQ 168(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 96(DI),AX
-	MULQ 176(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 96(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 184(SP)
-	ADDQ AX,SI
-	ADCQ DX,CX
-	MOVQ 96(DI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 192(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 104(DI),AX
-	MULQ 160(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 104(DI),AX
-	MULQ 168(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 0(SP),AX
-	MULQ 184(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SP),AX
-	MULQ 192(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 112(DI),AX
-	MULQ 160(SP)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SP),AX
-	MULQ 176(SP)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 8(SP),AX
-	MULQ 184(SP)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	MULQ 192(SP)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ $REDMASK51,DX
-	SHLQ $13,SI,CX
-	ANDQ DX,SI
-	SHLQ $13,R8,R9
-	ANDQ DX,R8
-	ADDQ CX,R8
-	SHLQ $13,R10,R11
-	ANDQ DX,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ DX,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ DX,R14
-	ADDQ R13,R14
-	IMUL3Q $19,R15,CX
-	ADDQ CX,SI
-	MOVQ SI,CX
-	SHRQ $51,CX
-	ADDQ R8,CX
-	MOVQ CX,R8
-	SHRQ $51,CX
-	ANDQ DX,SI
-	ADDQ R10,CX
-	MOVQ CX,R9
-	SHRQ $51,CX
-	ANDQ DX,R8
-	ADDQ R12,CX
-	MOVQ CX,AX
-	SHRQ $51,CX
-	ANDQ DX,R9
-	ADDQ R14,CX
-	MOVQ CX,R10
-	SHRQ $51,CX
-	ANDQ DX,AX
-	IMUL3Q $19,CX,CX
-	ADDQ CX,SI
-	ANDQ DX,R10
-	MOVQ SI,80(DI)
-	MOVQ R8,88(DI)
-	MOVQ R9,96(DI)
-	MOVQ AX,104(DI)
-	MOVQ R10,112(DI)
-	RET
-
-// func cswap(inout *[4][5]uint64, v uint64)
-TEXT ·cswap(SB),7,$0
-	MOVQ inout+0(FP),DI
-	MOVQ v+8(FP),SI
-
-	SUBQ $1, SI
-	NOTQ SI
-	MOVQ SI, X15
-	PSHUFD $0x44, X15, X15
-
-	MOVOU 0(DI), X0
-	MOVOU 16(DI), X2
-	MOVOU 32(DI), X4
-	MOVOU 48(DI), X6
-	MOVOU 64(DI), X8
-	MOVOU 80(DI), X1
-	MOVOU 96(DI), X3
-	MOVOU 112(DI), X5
-	MOVOU 128(DI), X7
-	MOVOU 144(DI), X9
-
-	MOVO X1, X10
-	MOVO X3, X11
-	MOVO X5, X12
-	MOVO X7, X13
-	MOVO X9, X14
-
-	PXOR X0, X10
-	PXOR X2, X11
-	PXOR X4, X12
-	PXOR X6, X13
-	PXOR X8, X14
-	PAND X15, X10
-	PAND X15, X11
-	PAND X15, X12
-	PAND X15, X13
-	PAND X15, X14
-	PXOR X10, X0
-	PXOR X10, X1
-	PXOR X11, X2
-	PXOR X11, X3
-	PXOR X12, X4
-	PXOR X12, X5
-	PXOR X13, X6
-	PXOR X13, X7
-	PXOR X14, X8
-	PXOR X14, X9
-
-	MOVOU X0, 0(DI)
-	MOVOU X2, 16(DI)
-	MOVOU X4, 32(DI)
-	MOVOU X6, 48(DI)
-	MOVOU X8, 64(DI)
-	MOVOU X1, 80(DI)
-	MOVOU X3, 96(DI)
-	MOVOU X5, 112(DI)
-	MOVOU X7, 128(DI)
-	MOVOU X9, 144(DI)
-	RET
-
-// func mul(dest, a, b *[5]uint64)
-TEXT ·mul(SB),0,$16-24
-	MOVQ dest+0(FP), DI
-	MOVQ a+8(FP), SI
-	MOVQ b+16(FP), DX
-
-	MOVQ DX,CX
-	MOVQ 24(SI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,0(SP)
-	MULQ 16(CX)
-	MOVQ AX,R8
-	MOVQ DX,R9
-	MOVQ 32(SI),DX
-	IMUL3Q $19,DX,AX
-	MOVQ AX,8(SP)
-	MULQ 8(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 0(SI),AX
-	MULQ 8(CX)
-	MOVQ AX,R10
-	MOVQ DX,R11
-	MOVQ 0(SI),AX
-	MULQ 16(CX)
-	MOVQ AX,R12
-	MOVQ DX,R13
-	MOVQ 0(SI),AX
-	MULQ 24(CX)
-	MOVQ AX,R14
-	MOVQ DX,R15
-	MOVQ 0(SI),AX
-	MULQ 32(CX)
-	MOVQ AX,BX
-	MOVQ DX,BP
-	MOVQ 8(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 8(SI),AX
-	MULQ 16(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 8(SI),AX
-	MULQ 24(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 8(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 16(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 16(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 16(SI),AX
-	MULQ 16(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 16(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(CX)
-	ADDQ AX,R8
-	ADCQ DX,R9
-	MOVQ 16(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 24(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ 24(SI),AX
-	MULQ 8(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 0(SP),AX
-	MULQ 24(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 0(SP),AX
-	MULQ 32(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 32(SI),AX
-	MULQ 0(CX)
-	ADDQ AX,BX
-	ADCQ DX,BP
-	MOVQ 8(SP),AX
-	MULQ 16(CX)
-	ADDQ AX,R10
-	ADCQ DX,R11
-	MOVQ 8(SP),AX
-	MULQ 24(CX)
-	ADDQ AX,R12
-	ADCQ DX,R13
-	MOVQ 8(SP),AX
-	MULQ 32(CX)
-	ADDQ AX,R14
-	ADCQ DX,R15
-	MOVQ $REDMASK51,SI
-	SHLQ $13,R8,R9
-	ANDQ SI,R8
-	SHLQ $13,R10,R11
-	ANDQ SI,R10
-	ADDQ R9,R10
-	SHLQ $13,R12,R13
-	ANDQ SI,R12
-	ADDQ R11,R12
-	SHLQ $13,R14,R15
-	ANDQ SI,R14
-	ADDQ R13,R14
-	SHLQ $13,BX,BP
-	ANDQ SI,BX
-	ADDQ R15,BX
-	IMUL3Q $19,BP,DX
-	ADDQ DX,R8
-	MOVQ R8,DX
-	SHRQ $51,DX
-	ADDQ R10,DX
-	MOVQ DX,CX
-	SHRQ $51,DX
-	ANDQ SI,R8
-	ADDQ R12,DX
-	MOVQ DX,R9
-	SHRQ $51,DX
-	ANDQ SI,CX
-	ADDQ R14,DX
-	MOVQ DX,AX
-	SHRQ $51,DX
-	ANDQ SI,R9
-	ADDQ BX,DX
-	MOVQ DX,R10
-	SHRQ $51,DX
-	ANDQ SI,AX
-	IMUL3Q $19,DX,DX
-	ADDQ DX,R8
-	ANDQ SI,R10
-	MOVQ R8,0(DI)
-	MOVQ CX,8(DI)
-	MOVQ R9,16(DI)
-	MOVQ AX,24(DI)
-	MOVQ R10,32(DI)
-	RET
-
-// func square(out, in *[5]uint64)
-TEXT ·square(SB),7,$0-16
-	MOVQ out+0(FP), DI
-	MOVQ in+8(FP), SI
-
-	MOVQ 0(SI),AX
-	MULQ 0(SI)
-	MOVQ AX,CX
-	MOVQ DX,R8
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 8(SI)
-	MOVQ AX,R9
-	MOVQ DX,R10
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 16(SI)
-	MOVQ AX,R11
-	MOVQ DX,R12
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 24(SI)
-	MOVQ AX,R13
-	MOVQ DX,R14
-	MOVQ 0(SI),AX
-	SHLQ $1,AX
-	MULQ 32(SI)
-	MOVQ AX,R15
-	MOVQ DX,BX
-	MOVQ 8(SI),AX
-	MULQ 8(SI)
-	ADDQ AX,R11
-	ADCQ DX,R12
-	MOVQ 8(SI),AX
-	SHLQ $1,AX
-	MULQ 16(SI)
-	ADDQ AX,R13
-	ADCQ DX,R14
-	MOVQ 8(SI),AX
-	SHLQ $1,AX
-	MULQ 24(SI)
-	ADDQ AX,R15
-	ADCQ DX,BX
-	MOVQ 8(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,CX
-	ADCQ DX,R8
-	MOVQ 16(SI),AX
-	MULQ 16(SI)
-	ADDQ AX,R15
-	ADCQ DX,BX
-	MOVQ 16(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 24(SI)
-	ADDQ AX,CX
-	ADCQ DX,R8
-	MOVQ 16(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R9
-	ADCQ DX,R10
-	MOVQ 24(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 24(SI)
-	ADDQ AX,R9
-	ADCQ DX,R10
-	MOVQ 24(SI),DX
-	IMUL3Q $38,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R11
-	ADCQ DX,R12
-	MOVQ 32(SI),DX
-	IMUL3Q $19,DX,AX
-	MULQ 32(SI)
-	ADDQ AX,R13
-	ADCQ DX,R14
-	MOVQ $REDMASK51,SI
-	SHLQ $13,CX,R8
-	ANDQ SI,CX
-	SHLQ $13,R9,R10
-	ANDQ SI,R9
-	ADDQ R8,R9
-	SHLQ $13,R11,R12
-	ANDQ SI,R11
-	ADDQ R10,R11
-	SHLQ $13,R13,R14
-	ANDQ SI,R13
-	ADDQ R12,R13
-	SHLQ $13,R15,BX
-	ANDQ SI,R15
-	ADDQ R14,R15
-	IMUL3Q $19,BX,DX
-	ADDQ DX,CX
-	MOVQ CX,DX
-	SHRQ $51,DX
-	ADDQ R9,DX
-	ANDQ SI,CX
-	MOVQ DX,R8
-	SHRQ $51,DX
-	ADDQ R11,DX
-	ANDQ SI,R8
-	MOVQ DX,R9
-	SHRQ $51,DX
-	ADDQ R13,DX
-	ANDQ SI,R9
-	MOVQ DX,AX
-	SHRQ $51,DX
-	ADDQ R15,DX
-	ANDQ SI,AX
-	MOVQ DX,R10
-	SHRQ $51,DX
-	IMUL3Q $19,DX,DX
-	ADDQ DX,CX
-	ANDQ SI,R10
-	MOVQ CX,0(DI)
-	MOVQ R8,8(DI)
-	MOVQ R9,16(DI)
-	MOVQ AX,24(DI)
-	MOVQ R10,32(DI)
-	RET
diff --git a/vendor/golang.org/x/crypto/curve25519/curve25519_generic.go b/vendor/golang.org/x/crypto/curve25519/curve25519_generic.go
deleted file mode 100644
index c43b13fc..00000000
--- a/vendor/golang.org/x/crypto/curve25519/curve25519_generic.go
+++ /dev/null
@@ -1,828 +0,0 @@
-// Copyright 2013 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 curve25519
-
-import "encoding/binary"
-
-// This code is a port of the public domain, "ref10" implementation of
-// curve25519 from SUPERCOP 20130419 by D. J. Bernstein.
-
-// fieldElement represents an element of the field GF(2^255 - 19). An element
-// t, entries t[0]...t[9], represents the integer t[0]+2^26 t[1]+2^51 t[2]+2^77
-// t[3]+2^102 t[4]+...+2^230 t[9]. Bounds on each t[i] vary depending on
-// context.
-type fieldElement [10]int32
-
-func feZero(fe *fieldElement) {
-	for i := range fe {
-		fe[i] = 0
-	}
-}
-
-func feOne(fe *fieldElement) {
-	feZero(fe)
-	fe[0] = 1
-}
-
-func feAdd(dst, a, b *fieldElement) {
-	for i := range dst {
-		dst[i] = a[i] + b[i]
-	}
-}
-
-func feSub(dst, a, b *fieldElement) {
-	for i := range dst {
-		dst[i] = a[i] - b[i]
-	}
-}
-
-func feCopy(dst, src *fieldElement) {
-	for i := range dst {
-		dst[i] = src[i]
-	}
-}
-
-// feCSwap replaces (f,g) with (g,f) if b == 1; replaces (f,g) with (f,g) if b == 0.
-//
-// Preconditions: b in {0,1}.
-func feCSwap(f, g *fieldElement, b int32) {
-	b = -b
-	for i := range f {
-		t := b & (f[i] ^ g[i])
-		f[i] ^= t
-		g[i] ^= t
-	}
-}
-
-// load3 reads a 24-bit, little-endian value from in.
-func load3(in []byte) int64 {
-	var r int64
-	r = int64(in[0])
-	r |= int64(in[1]) << 8
-	r |= int64(in[2]) << 16
-	return r
-}
-
-// load4 reads a 32-bit, little-endian value from in.
-func load4(in []byte) int64 {
-	return int64(binary.LittleEndian.Uint32(in))
-}
-
-func feFromBytes(dst *fieldElement, src *[32]byte) {
-	h0 := load4(src[:])
-	h1 := load3(src[4:]) << 6
-	h2 := load3(src[7:]) << 5
-	h3 := load3(src[10:]) << 3
-	h4 := load3(src[13:]) << 2
-	h5 := load4(src[16:])
-	h6 := load3(src[20:]) << 7
-	h7 := load3(src[23:]) << 5
-	h8 := load3(src[26:]) << 4
-	h9 := (load3(src[29:]) & 0x7fffff) << 2
-
-	var carry [10]int64
-	carry[9] = (h9 + 1<<24) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	carry[1] = (h1 + 1<<24) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[3] = (h3 + 1<<24) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[5] = (h5 + 1<<24) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	carry[7] = (h7 + 1<<24) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[0] = (h0 + 1<<25) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[2] = (h2 + 1<<25) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[4] = (h4 + 1<<25) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[6] = (h6 + 1<<25) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	carry[8] = (h8 + 1<<25) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	dst[0] = int32(h0)
-	dst[1] = int32(h1)
-	dst[2] = int32(h2)
-	dst[3] = int32(h3)
-	dst[4] = int32(h4)
-	dst[5] = int32(h5)
-	dst[6] = int32(h6)
-	dst[7] = int32(h7)
-	dst[8] = int32(h8)
-	dst[9] = int32(h9)
-}
-
-// feToBytes marshals h to s.
-// Preconditions:
-//   |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-//
-// Write p=2^255-19; q=floor(h/p).
-// Basic claim: q = floor(2^(-255)(h + 19 2^(-25)h9 + 2^(-1))).
-//
-// Proof:
-//   Have |h|<=p so |q|<=1 so |19^2 2^(-255) q|<1/4.
-//   Also have |h-2^230 h9|<2^230 so |19 2^(-255)(h-2^230 h9)|<1/4.
-//
-//   Write y=2^(-1)-19^2 2^(-255)q-19 2^(-255)(h-2^230 h9).
-//   Then 0<y<1.
-//
-//   Write r=h-pq.
-//   Have 0<=r<=p-1=2^255-20.
-//   Thus 0<=r+19(2^-255)r<r+19(2^-255)2^255<=2^255-1.
-//
-//   Write x=r+19(2^-255)r+y.
-//   Then 0<x<2^255 so floor(2^(-255)x) = 0 so floor(q+2^(-255)x) = q.
-//
-//   Have q+2^(-255)x = 2^(-255)(h + 19 2^(-25) h9 + 2^(-1))
-//   so floor(2^(-255)(h + 19 2^(-25) h9 + 2^(-1))) = q.
-func feToBytes(s *[32]byte, h *fieldElement) {
-	var carry [10]int32
-
-	q := (19*h[9] + (1 << 24)) >> 25
-	q = (h[0] + q) >> 26
-	q = (h[1] + q) >> 25
-	q = (h[2] + q) >> 26
-	q = (h[3] + q) >> 25
-	q = (h[4] + q) >> 26
-	q = (h[5] + q) >> 25
-	q = (h[6] + q) >> 26
-	q = (h[7] + q) >> 25
-	q = (h[8] + q) >> 26
-	q = (h[9] + q) >> 25
-
-	// Goal: Output h-(2^255-19)q, which is between 0 and 2^255-20.
-	h[0] += 19 * q
-	// Goal: Output h-2^255 q, which is between 0 and 2^255-20.
-
-	carry[0] = h[0] >> 26
-	h[1] += carry[0]
-	h[0] -= carry[0] << 26
-	carry[1] = h[1] >> 25
-	h[2] += carry[1]
-	h[1] -= carry[1] << 25
-	carry[2] = h[2] >> 26
-	h[3] += carry[2]
-	h[2] -= carry[2] << 26
-	carry[3] = h[3] >> 25
-	h[4] += carry[3]
-	h[3] -= carry[3] << 25
-	carry[4] = h[4] >> 26
-	h[5] += carry[4]
-	h[4] -= carry[4] << 26
-	carry[5] = h[5] >> 25
-	h[6] += carry[5]
-	h[5] -= carry[5] << 25
-	carry[6] = h[6] >> 26
-	h[7] += carry[6]
-	h[6] -= carry[6] << 26
-	carry[7] = h[7] >> 25
-	h[8] += carry[7]
-	h[7] -= carry[7] << 25
-	carry[8] = h[8] >> 26
-	h[9] += carry[8]
-	h[8] -= carry[8] << 26
-	carry[9] = h[9] >> 25
-	h[9] -= carry[9] << 25
-	// h10 = carry9
-
-	// Goal: Output h[0]+...+2^255 h10-2^255 q, which is between 0 and 2^255-20.
-	// Have h[0]+...+2^230 h[9] between 0 and 2^255-1;
-	// evidently 2^255 h10-2^255 q = 0.
-	// Goal: Output h[0]+...+2^230 h[9].
-
-	s[0] = byte(h[0] >> 0)
-	s[1] = byte(h[0] >> 8)
-	s[2] = byte(h[0] >> 16)
-	s[3] = byte((h[0] >> 24) | (h[1] << 2))
-	s[4] = byte(h[1] >> 6)
-	s[5] = byte(h[1] >> 14)
-	s[6] = byte((h[1] >> 22) | (h[2] << 3))
-	s[7] = byte(h[2] >> 5)
-	s[8] = byte(h[2] >> 13)
-	s[9] = byte((h[2] >> 21) | (h[3] << 5))
-	s[10] = byte(h[3] >> 3)
-	s[11] = byte(h[3] >> 11)
-	s[12] = byte((h[3] >> 19) | (h[4] << 6))
-	s[13] = byte(h[4] >> 2)
-	s[14] = byte(h[4] >> 10)
-	s[15] = byte(h[4] >> 18)
-	s[16] = byte(h[5] >> 0)
-	s[17] = byte(h[5] >> 8)
-	s[18] = byte(h[5] >> 16)
-	s[19] = byte((h[5] >> 24) | (h[6] << 1))
-	s[20] = byte(h[6] >> 7)
-	s[21] = byte(h[6] >> 15)
-	s[22] = byte((h[6] >> 23) | (h[7] << 3))
-	s[23] = byte(h[7] >> 5)
-	s[24] = byte(h[7] >> 13)
-	s[25] = byte((h[7] >> 21) | (h[8] << 4))
-	s[26] = byte(h[8] >> 4)
-	s[27] = byte(h[8] >> 12)
-	s[28] = byte((h[8] >> 20) | (h[9] << 6))
-	s[29] = byte(h[9] >> 2)
-	s[30] = byte(h[9] >> 10)
-	s[31] = byte(h[9] >> 18)
-}
-
-// feMul calculates h = f * g
-// Can overlap h with f or g.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//    |g| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-//
-// Notes on implementation strategy:
-//
-// Using schoolbook multiplication.
-// Karatsuba would save a little in some cost models.
-//
-// Most multiplications by 2 and 19 are 32-bit precomputations;
-// cheaper than 64-bit postcomputations.
-//
-// There is one remaining multiplication by 19 in the carry chain;
-// one *19 precomputation can be merged into this,
-// but the resulting data flow is considerably less clean.
-//
-// There are 12 carries below.
-// 10 of them are 2-way parallelizable and vectorizable.
-// Can get away with 11 carries, but then data flow is much deeper.
-//
-// With tighter constraints on inputs can squeeze carries into int32.
-func feMul(h, f, g *fieldElement) {
-	f0 := f[0]
-	f1 := f[1]
-	f2 := f[2]
-	f3 := f[3]
-	f4 := f[4]
-	f5 := f[5]
-	f6 := f[6]
-	f7 := f[7]
-	f8 := f[8]
-	f9 := f[9]
-	g0 := g[0]
-	g1 := g[1]
-	g2 := g[2]
-	g3 := g[3]
-	g4 := g[4]
-	g5 := g[5]
-	g6 := g[6]
-	g7 := g[7]
-	g8 := g[8]
-	g9 := g[9]
-	g1_19 := 19 * g1 // 1.4*2^29
-	g2_19 := 19 * g2 // 1.4*2^30; still ok
-	g3_19 := 19 * g3
-	g4_19 := 19 * g4
-	g5_19 := 19 * g5
-	g6_19 := 19 * g6
-	g7_19 := 19 * g7
-	g8_19 := 19 * g8
-	g9_19 := 19 * g9
-	f1_2 := 2 * f1
-	f3_2 := 2 * f3
-	f5_2 := 2 * f5
-	f7_2 := 2 * f7
-	f9_2 := 2 * f9
-	f0g0 := int64(f0) * int64(g0)
-	f0g1 := int64(f0) * int64(g1)
-	f0g2 := int64(f0) * int64(g2)
-	f0g3 := int64(f0) * int64(g3)
-	f0g4 := int64(f0) * int64(g4)
-	f0g5 := int64(f0) * int64(g5)
-	f0g6 := int64(f0) * int64(g6)
-	f0g7 := int64(f0) * int64(g7)
-	f0g8 := int64(f0) * int64(g8)
-	f0g9 := int64(f0) * int64(g9)
-	f1g0 := int64(f1) * int64(g0)
-	f1g1_2 := int64(f1_2) * int64(g1)
-	f1g2 := int64(f1) * int64(g2)
-	f1g3_2 := int64(f1_2) * int64(g3)
-	f1g4 := int64(f1) * int64(g4)
-	f1g5_2 := int64(f1_2) * int64(g5)
-	f1g6 := int64(f1) * int64(g6)
-	f1g7_2 := int64(f1_2) * int64(g7)
-	f1g8 := int64(f1) * int64(g8)
-	f1g9_38 := int64(f1_2) * int64(g9_19)
-	f2g0 := int64(f2) * int64(g0)
-	f2g1 := int64(f2) * int64(g1)
-	f2g2 := int64(f2) * int64(g2)
-	f2g3 := int64(f2) * int64(g3)
-	f2g4 := int64(f2) * int64(g4)
-	f2g5 := int64(f2) * int64(g5)
-	f2g6 := int64(f2) * int64(g6)
-	f2g7 := int64(f2) * int64(g7)
-	f2g8_19 := int64(f2) * int64(g8_19)
-	f2g9_19 := int64(f2) * int64(g9_19)
-	f3g0 := int64(f3) * int64(g0)
-	f3g1_2 := int64(f3_2) * int64(g1)
-	f3g2 := int64(f3) * int64(g2)
-	f3g3_2 := int64(f3_2) * int64(g3)
-	f3g4 := int64(f3) * int64(g4)
-	f3g5_2 := int64(f3_2) * int64(g5)
-	f3g6 := int64(f3) * int64(g6)
-	f3g7_38 := int64(f3_2) * int64(g7_19)
-	f3g8_19 := int64(f3) * int64(g8_19)
-	f3g9_38 := int64(f3_2) * int64(g9_19)
-	f4g0 := int64(f4) * int64(g0)
-	f4g1 := int64(f4) * int64(g1)
-	f4g2 := int64(f4) * int64(g2)
-	f4g3 := int64(f4) * int64(g3)
-	f4g4 := int64(f4) * int64(g4)
-	f4g5 := int64(f4) * int64(g5)
-	f4g6_19 := int64(f4) * int64(g6_19)
-	f4g7_19 := int64(f4) * int64(g7_19)
-	f4g8_19 := int64(f4) * int64(g8_19)
-	f4g9_19 := int64(f4) * int64(g9_19)
-	f5g0 := int64(f5) * int64(g0)
-	f5g1_2 := int64(f5_2) * int64(g1)
-	f5g2 := int64(f5) * int64(g2)
-	f5g3_2 := int64(f5_2) * int64(g3)
-	f5g4 := int64(f5) * int64(g4)
-	f5g5_38 := int64(f5_2) * int64(g5_19)
-	f5g6_19 := int64(f5) * int64(g6_19)
-	f5g7_38 := int64(f5_2) * int64(g7_19)
-	f5g8_19 := int64(f5) * int64(g8_19)
-	f5g9_38 := int64(f5_2) * int64(g9_19)
-	f6g0 := int64(f6) * int64(g0)
-	f6g1 := int64(f6) * int64(g1)
-	f6g2 := int64(f6) * int64(g2)
-	f6g3 := int64(f6) * int64(g3)
-	f6g4_19 := int64(f6) * int64(g4_19)
-	f6g5_19 := int64(f6) * int64(g5_19)
-	f6g6_19 := int64(f6) * int64(g6_19)
-	f6g7_19 := int64(f6) * int64(g7_19)
-	f6g8_19 := int64(f6) * int64(g8_19)
-	f6g9_19 := int64(f6) * int64(g9_19)
-	f7g0 := int64(f7) * int64(g0)
-	f7g1_2 := int64(f7_2) * int64(g1)
-	f7g2 := int64(f7) * int64(g2)
-	f7g3_38 := int64(f7_2) * int64(g3_19)
-	f7g4_19 := int64(f7) * int64(g4_19)
-	f7g5_38 := int64(f7_2) * int64(g5_19)
-	f7g6_19 := int64(f7) * int64(g6_19)
-	f7g7_38 := int64(f7_2) * int64(g7_19)
-	f7g8_19 := int64(f7) * int64(g8_19)
-	f7g9_38 := int64(f7_2) * int64(g9_19)
-	f8g0 := int64(f8) * int64(g0)
-	f8g1 := int64(f8) * int64(g1)
-	f8g2_19 := int64(f8) * int64(g2_19)
-	f8g3_19 := int64(f8) * int64(g3_19)
-	f8g4_19 := int64(f8) * int64(g4_19)
-	f8g5_19 := int64(f8) * int64(g5_19)
-	f8g6_19 := int64(f8) * int64(g6_19)
-	f8g7_19 := int64(f8) * int64(g7_19)
-	f8g8_19 := int64(f8) * int64(g8_19)
-	f8g9_19 := int64(f8) * int64(g9_19)
-	f9g0 := int64(f9) * int64(g0)
-	f9g1_38 := int64(f9_2) * int64(g1_19)
-	f9g2_19 := int64(f9) * int64(g2_19)
-	f9g3_38 := int64(f9_2) * int64(g3_19)
-	f9g4_19 := int64(f9) * int64(g4_19)
-	f9g5_38 := int64(f9_2) * int64(g5_19)
-	f9g6_19 := int64(f9) * int64(g6_19)
-	f9g7_38 := int64(f9_2) * int64(g7_19)
-	f9g8_19 := int64(f9) * int64(g8_19)
-	f9g9_38 := int64(f9_2) * int64(g9_19)
-	h0 := f0g0 + f1g9_38 + f2g8_19 + f3g7_38 + f4g6_19 + f5g5_38 + f6g4_19 + f7g3_38 + f8g2_19 + f9g1_38
-	h1 := f0g1 + f1g0 + f2g9_19 + f3g8_19 + f4g7_19 + f5g6_19 + f6g5_19 + f7g4_19 + f8g3_19 + f9g2_19
-	h2 := f0g2 + f1g1_2 + f2g0 + f3g9_38 + f4g8_19 + f5g7_38 + f6g6_19 + f7g5_38 + f8g4_19 + f9g3_38
-	h3 := f0g3 + f1g2 + f2g1 + f3g0 + f4g9_19 + f5g8_19 + f6g7_19 + f7g6_19 + f8g5_19 + f9g4_19
-	h4 := f0g4 + f1g3_2 + f2g2 + f3g1_2 + f4g0 + f5g9_38 + f6g8_19 + f7g7_38 + f8g6_19 + f9g5_38
-	h5 := f0g5 + f1g4 + f2g3 + f3g2 + f4g1 + f5g0 + f6g9_19 + f7g8_19 + f8g7_19 + f9g6_19
-	h6 := f0g6 + f1g5_2 + f2g4 + f3g3_2 + f4g2 + f5g1_2 + f6g0 + f7g9_38 + f8g8_19 + f9g7_38
-	h7 := f0g7 + f1g6 + f2g5 + f3g4 + f4g3 + f5g2 + f6g1 + f7g0 + f8g9_19 + f9g8_19
-	h8 := f0g8 + f1g7_2 + f2g6 + f3g5_2 + f4g4 + f5g3_2 + f6g2 + f7g1_2 + f8g0 + f9g9_38
-	h9 := f0g9 + f1g8 + f2g7 + f3g6 + f4g5 + f5g4 + f6g3 + f7g2 + f8g1 + f9g0
-	var carry [10]int64
-
-	// |h0| <= (1.1*1.1*2^52*(1+19+19+19+19)+1.1*1.1*2^50*(38+38+38+38+38))
-	//   i.e. |h0| <= 1.2*2^59; narrower ranges for h2, h4, h6, h8
-	// |h1| <= (1.1*1.1*2^51*(1+1+19+19+19+19+19+19+19+19))
-	//   i.e. |h1| <= 1.5*2^58; narrower ranges for h3, h5, h7, h9
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	// |h0| <= 2^25
-	// |h4| <= 2^25
-	// |h1| <= 1.51*2^58
-	// |h5| <= 1.51*2^58
-
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	// |h1| <= 2^24; from now on fits into int32
-	// |h5| <= 2^24; from now on fits into int32
-	// |h2| <= 1.21*2^59
-	// |h6| <= 1.21*2^59
-
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	// |h2| <= 2^25; from now on fits into int32 unchanged
-	// |h6| <= 2^25; from now on fits into int32 unchanged
-	// |h3| <= 1.51*2^58
-	// |h7| <= 1.51*2^58
-
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-	// |h3| <= 2^24; from now on fits into int32 unchanged
-	// |h7| <= 2^24; from now on fits into int32 unchanged
-	// |h4| <= 1.52*2^33
-	// |h8| <= 1.52*2^33
-
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-	// |h4| <= 2^25; from now on fits into int32 unchanged
-	// |h8| <= 2^25; from now on fits into int32 unchanged
-	// |h5| <= 1.01*2^24
-	// |h9| <= 1.51*2^58
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	// |h9| <= 2^24; from now on fits into int32 unchanged
-	// |h0| <= 1.8*2^37
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	// |h0| <= 2^25; from now on fits into int32 unchanged
-	// |h1| <= 1.01*2^24
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feSquare calculates h = f*f. Can overlap h with f.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-func feSquare(h, f *fieldElement) {
-	f0 := f[0]
-	f1 := f[1]
-	f2 := f[2]
-	f3 := f[3]
-	f4 := f[4]
-	f5 := f[5]
-	f6 := f[6]
-	f7 := f[7]
-	f8 := f[8]
-	f9 := f[9]
-	f0_2 := 2 * f0
-	f1_2 := 2 * f1
-	f2_2 := 2 * f2
-	f3_2 := 2 * f3
-	f4_2 := 2 * f4
-	f5_2 := 2 * f5
-	f6_2 := 2 * f6
-	f7_2 := 2 * f7
-	f5_38 := 38 * f5 // 1.31*2^30
-	f6_19 := 19 * f6 // 1.31*2^30
-	f7_38 := 38 * f7 // 1.31*2^30
-	f8_19 := 19 * f8 // 1.31*2^30
-	f9_38 := 38 * f9 // 1.31*2^30
-	f0f0 := int64(f0) * int64(f0)
-	f0f1_2 := int64(f0_2) * int64(f1)
-	f0f2_2 := int64(f0_2) * int64(f2)
-	f0f3_2 := int64(f0_2) * int64(f3)
-	f0f4_2 := int64(f0_2) * int64(f4)
-	f0f5_2 := int64(f0_2) * int64(f5)
-	f0f6_2 := int64(f0_2) * int64(f6)
-	f0f7_2 := int64(f0_2) * int64(f7)
-	f0f8_2 := int64(f0_2) * int64(f8)
-	f0f9_2 := int64(f0_2) * int64(f9)
-	f1f1_2 := int64(f1_2) * int64(f1)
-	f1f2_2 := int64(f1_2) * int64(f2)
-	f1f3_4 := int64(f1_2) * int64(f3_2)
-	f1f4_2 := int64(f1_2) * int64(f4)
-	f1f5_4 := int64(f1_2) * int64(f5_2)
-	f1f6_2 := int64(f1_2) * int64(f6)
-	f1f7_4 := int64(f1_2) * int64(f7_2)
-	f1f8_2 := int64(f1_2) * int64(f8)
-	f1f9_76 := int64(f1_2) * int64(f9_38)
-	f2f2 := int64(f2) * int64(f2)
-	f2f3_2 := int64(f2_2) * int64(f3)
-	f2f4_2 := int64(f2_2) * int64(f4)
-	f2f5_2 := int64(f2_2) * int64(f5)
-	f2f6_2 := int64(f2_2) * int64(f6)
-	f2f7_2 := int64(f2_2) * int64(f7)
-	f2f8_38 := int64(f2_2) * int64(f8_19)
-	f2f9_38 := int64(f2) * int64(f9_38)
-	f3f3_2 := int64(f3_2) * int64(f3)
-	f3f4_2 := int64(f3_2) * int64(f4)
-	f3f5_4 := int64(f3_2) * int64(f5_2)
-	f3f6_2 := int64(f3_2) * int64(f6)
-	f3f7_76 := int64(f3_2) * int64(f7_38)
-	f3f8_38 := int64(f3_2) * int64(f8_19)
-	f3f9_76 := int64(f3_2) * int64(f9_38)
-	f4f4 := int64(f4) * int64(f4)
-	f4f5_2 := int64(f4_2) * int64(f5)
-	f4f6_38 := int64(f4_2) * int64(f6_19)
-	f4f7_38 := int64(f4) * int64(f7_38)
-	f4f8_38 := int64(f4_2) * int64(f8_19)
-	f4f9_38 := int64(f4) * int64(f9_38)
-	f5f5_38 := int64(f5) * int64(f5_38)
-	f5f6_38 := int64(f5_2) * int64(f6_19)
-	f5f7_76 := int64(f5_2) * int64(f7_38)
-	f5f8_38 := int64(f5_2) * int64(f8_19)
-	f5f9_76 := int64(f5_2) * int64(f9_38)
-	f6f6_19 := int64(f6) * int64(f6_19)
-	f6f7_38 := int64(f6) * int64(f7_38)
-	f6f8_38 := int64(f6_2) * int64(f8_19)
-	f6f9_38 := int64(f6) * int64(f9_38)
-	f7f7_38 := int64(f7) * int64(f7_38)
-	f7f8_38 := int64(f7_2) * int64(f8_19)
-	f7f9_76 := int64(f7_2) * int64(f9_38)
-	f8f8_19 := int64(f8) * int64(f8_19)
-	f8f9_38 := int64(f8) * int64(f9_38)
-	f9f9_38 := int64(f9) * int64(f9_38)
-	h0 := f0f0 + f1f9_76 + f2f8_38 + f3f7_76 + f4f6_38 + f5f5_38
-	h1 := f0f1_2 + f2f9_38 + f3f8_38 + f4f7_38 + f5f6_38
-	h2 := f0f2_2 + f1f1_2 + f3f9_76 + f4f8_38 + f5f7_76 + f6f6_19
-	h3 := f0f3_2 + f1f2_2 + f4f9_38 + f5f8_38 + f6f7_38
-	h4 := f0f4_2 + f1f3_4 + f2f2 + f5f9_76 + f6f8_38 + f7f7_38
-	h5 := f0f5_2 + f1f4_2 + f2f3_2 + f6f9_38 + f7f8_38
-	h6 := f0f6_2 + f1f5_4 + f2f4_2 + f3f3_2 + f7f9_76 + f8f8_19
-	h7 := f0f7_2 + f1f6_2 + f2f5_2 + f3f4_2 + f8f9_38
-	h8 := f0f8_2 + f1f7_4 + f2f6_2 + f3f5_4 + f4f4 + f9f9_38
-	h9 := f0f9_2 + f1f8_2 + f2f7_2 + f3f6_2 + f4f5_2
-	var carry [10]int64
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feMul121666 calculates h = f * 121666. Can overlap h with f.
-//
-// Preconditions:
-//    |f| bounded by 1.1*2^26,1.1*2^25,1.1*2^26,1.1*2^25,etc.
-//
-// Postconditions:
-//    |h| bounded by 1.1*2^25,1.1*2^24,1.1*2^25,1.1*2^24,etc.
-func feMul121666(h, f *fieldElement) {
-	h0 := int64(f[0]) * 121666
-	h1 := int64(f[1]) * 121666
-	h2 := int64(f[2]) * 121666
-	h3 := int64(f[3]) * 121666
-	h4 := int64(f[4]) * 121666
-	h5 := int64(f[5]) * 121666
-	h6 := int64(f[6]) * 121666
-	h7 := int64(f[7]) * 121666
-	h8 := int64(f[8]) * 121666
-	h9 := int64(f[9]) * 121666
-	var carry [10]int64
-
-	carry[9] = (h9 + (1 << 24)) >> 25
-	h0 += carry[9] * 19
-	h9 -= carry[9] << 25
-	carry[1] = (h1 + (1 << 24)) >> 25
-	h2 += carry[1]
-	h1 -= carry[1] << 25
-	carry[3] = (h3 + (1 << 24)) >> 25
-	h4 += carry[3]
-	h3 -= carry[3] << 25
-	carry[5] = (h5 + (1 << 24)) >> 25
-	h6 += carry[5]
-	h5 -= carry[5] << 25
-	carry[7] = (h7 + (1 << 24)) >> 25
-	h8 += carry[7]
-	h7 -= carry[7] << 25
-
-	carry[0] = (h0 + (1 << 25)) >> 26
-	h1 += carry[0]
-	h0 -= carry[0] << 26
-	carry[2] = (h2 + (1 << 25)) >> 26
-	h3 += carry[2]
-	h2 -= carry[2] << 26
-	carry[4] = (h4 + (1 << 25)) >> 26
-	h5 += carry[4]
-	h4 -= carry[4] << 26
-	carry[6] = (h6 + (1 << 25)) >> 26
-	h7 += carry[6]
-	h6 -= carry[6] << 26
-	carry[8] = (h8 + (1 << 25)) >> 26
-	h9 += carry[8]
-	h8 -= carry[8] << 26
-
-	h[0] = int32(h0)
-	h[1] = int32(h1)
-	h[2] = int32(h2)
-	h[3] = int32(h3)
-	h[4] = int32(h4)
-	h[5] = int32(h5)
-	h[6] = int32(h6)
-	h[7] = int32(h7)
-	h[8] = int32(h8)
-	h[9] = int32(h9)
-}
-
-// feInvert sets out = z^-1.
-func feInvert(out, z *fieldElement) {
-	var t0, t1, t2, t3 fieldElement
-	var i int
-
-	feSquare(&t0, z)
-	for i = 1; i < 1; i++ {
-		feSquare(&t0, &t0)
-	}
-	feSquare(&t1, &t0)
-	for i = 1; i < 2; i++ {
-		feSquare(&t1, &t1)
-	}
-	feMul(&t1, z, &t1)
-	feMul(&t0, &t0, &t1)
-	feSquare(&t2, &t0)
-	for i = 1; i < 1; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t1, &t2)
-	feSquare(&t2, &t1)
-	for i = 1; i < 5; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t2, &t1)
-	for i = 1; i < 10; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t2, &t2, &t1)
-	feSquare(&t3, &t2)
-	for i = 1; i < 20; i++ {
-		feSquare(&t3, &t3)
-	}
-	feMul(&t2, &t3, &t2)
-	feSquare(&t2, &t2)
-	for i = 1; i < 10; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t2, &t1)
-	for i = 1; i < 50; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t2, &t2, &t1)
-	feSquare(&t3, &t2)
-	for i = 1; i < 100; i++ {
-		feSquare(&t3, &t3)
-	}
-	feMul(&t2, &t3, &t2)
-	feSquare(&t2, &t2)
-	for i = 1; i < 50; i++ {
-		feSquare(&t2, &t2)
-	}
-	feMul(&t1, &t2, &t1)
-	feSquare(&t1, &t1)
-	for i = 1; i < 5; i++ {
-		feSquare(&t1, &t1)
-	}
-	feMul(out, &t1, &t0)
-}
-
-func scalarMultGeneric(out, in, base *[32]byte) {
-	var e [32]byte
-
-	copy(e[:], in[:])
-	e[0] &= 248
-	e[31] &= 127
-	e[31] |= 64
-
-	var x1, x2, z2, x3, z3, tmp0, tmp1 fieldElement
-	feFromBytes(&x1, base)
-	feOne(&x2)
-	feCopy(&x3, &x1)
-	feOne(&z3)
-
-	swap := int32(0)
-	for pos := 254; pos >= 0; pos-- {
-		b := e[pos/8] >> uint(pos&7)
-		b &= 1
-		swap ^= int32(b)
-		feCSwap(&x2, &x3, swap)
-		feCSwap(&z2, &z3, swap)
-		swap = int32(b)
-
-		feSub(&tmp0, &x3, &z3)
-		feSub(&tmp1, &x2, &z2)
-		feAdd(&x2, &x2, &z2)
-		feAdd(&z2, &x3, &z3)
-		feMul(&z3, &tmp0, &x2)
-		feMul(&z2, &z2, &tmp1)
-		feSquare(&tmp0, &tmp1)
-		feSquare(&tmp1, &x2)
-		feAdd(&x3, &z3, &z2)
-		feSub(&z2, &z3, &z2)
-		feMul(&x2, &tmp1, &tmp0)
-		feSub(&tmp1, &tmp1, &tmp0)
-		feSquare(&z2, &z2)
-		feMul121666(&z3, &tmp1)
-		feSquare(&x3, &x3)
-		feAdd(&tmp0, &tmp0, &z3)
-		feMul(&z3, &x1, &z2)
-		feMul(&z2, &tmp1, &tmp0)
-	}
-
-	feCSwap(&x2, &x3, swap)
-	feCSwap(&z2, &z3, swap)
-
-	feInvert(&z2, &z2)
-	feMul(&x2, &x2, &z2)
-	feToBytes(out, &x2)
-}
diff --git a/vendor/golang.org/x/crypto/curve25519/curve25519_noasm.go b/vendor/golang.org/x/crypto/curve25519/curve25519_noasm.go
deleted file mode 100644
index 259728af..00000000
--- a/vendor/golang.org/x/crypto/curve25519/curve25519_noasm.go
+++ /dev/null
@@ -1,12 +0,0 @@
-// Copyright 2019 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 || !gc || purego
-// +build !amd64 !gc purego
-
-package curve25519
-
-func scalarMult(out, in, base *[32]byte) {
-	scalarMultGeneric(out, in, base)
-}
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/README b/vendor/golang.org/x/crypto/curve25519/internal/field/README
new file mode 100644
index 00000000..e25bca7d
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/README
@@ -0,0 +1,7 @@
+This package is kept in sync with crypto/ed25519/internal/edwards25519/field in
+the standard library.
+
+If there are any changes in the standard library that need to be synced to this
+package, run sync.sh. It will not overwrite any local changes made since the
+previous sync, so it's ok to land changes in this package first, and then sync
+to the standard library later.
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe.go
new file mode 100644
index 00000000..ca841ad9
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe.go
@@ -0,0 +1,416 @@
+// Copyright (c) 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.
+
+// Package field implements fast arithmetic modulo 2^255-19.
+package field
+
+import (
+	"crypto/subtle"
+	"encoding/binary"
+	"math/bits"
+)
+
+// Element represents an element of the field GF(2^255-19). Note that this
+// is not a cryptographically secure group, and should only be used to interact
+// with edwards25519.Point coordinates.
+//
+// This type works similarly to math/big.Int, and all arguments and receivers
+// are allowed to alias.
+//
+// The zero value is a valid zero element.
+type Element struct {
+	// An element t represents the integer
+	//     t.l0 + t.l1*2^51 + t.l2*2^102 + t.l3*2^153 + t.l4*2^204
+	//
+	// Between operations, all limbs are expected to be lower than 2^52.
+	l0 uint64
+	l1 uint64
+	l2 uint64
+	l3 uint64
+	l4 uint64
+}
+
+const maskLow51Bits uint64 = (1 << 51) - 1
+
+var feZero = &Element{0, 0, 0, 0, 0}
+
+// Zero sets v = 0, and returns v.
+func (v *Element) Zero() *Element {
+	*v = *feZero
+	return v
+}
+
+var feOne = &Element{1, 0, 0, 0, 0}
+
+// One sets v = 1, and returns v.
+func (v *Element) One() *Element {
+	*v = *feOne
+	return v
+}
+
+// reduce reduces v modulo 2^255 - 19 and returns it.
+func (v *Element) reduce() *Element {
+	v.carryPropagate()
+
+	// After the light reduction we now have a field element representation
+	// v < 2^255 + 2^13 * 19, but need v < 2^255 - 19.
+
+	// If v >= 2^255 - 19, then v + 19 >= 2^255, which would overflow 2^255 - 1,
+	// generating a carry. That is, c will be 0 if v < 2^255 - 19, and 1 otherwise.
+	c := (v.l0 + 19) >> 51
+	c = (v.l1 + c) >> 51
+	c = (v.l2 + c) >> 51
+	c = (v.l3 + c) >> 51
+	c = (v.l4 + c) >> 51
+
+	// If v < 2^255 - 19 and c = 0, this will be a no-op. Otherwise, it's
+	// effectively applying the reduction identity to the carry.
+	v.l0 += 19 * c
+
+	v.l1 += v.l0 >> 51
+	v.l0 = v.l0 & maskLow51Bits
+	v.l2 += v.l1 >> 51
+	v.l1 = v.l1 & maskLow51Bits
+	v.l3 += v.l2 >> 51
+	v.l2 = v.l2 & maskLow51Bits
+	v.l4 += v.l3 >> 51
+	v.l3 = v.l3 & maskLow51Bits
+	// no additional carry
+	v.l4 = v.l4 & maskLow51Bits
+
+	return v
+}
+
+// Add sets v = a + b, and returns v.
+func (v *Element) Add(a, b *Element) *Element {
+	v.l0 = a.l0 + b.l0
+	v.l1 = a.l1 + b.l1
+	v.l2 = a.l2 + b.l2
+	v.l3 = a.l3 + b.l3
+	v.l4 = a.l4 + b.l4
+	// Using the generic implementation here is actually faster than the
+	// assembly. Probably because the body of this function is so simple that
+	// the compiler can figure out better optimizations by inlining the carry
+	// propagation. TODO
+	return v.carryPropagateGeneric()
+}
+
+// Subtract sets v = a - b, and returns v.
+func (v *Element) Subtract(a, b *Element) *Element {
+	// We first add 2 * p, to guarantee the subtraction won't underflow, and
+	// then subtract b (which can be up to 2^255 + 2^13 * 19).
+	v.l0 = (a.l0 + 0xFFFFFFFFFFFDA) - b.l0
+	v.l1 = (a.l1 + 0xFFFFFFFFFFFFE) - b.l1
+	v.l2 = (a.l2 + 0xFFFFFFFFFFFFE) - b.l2
+	v.l3 = (a.l3 + 0xFFFFFFFFFFFFE) - b.l3
+	v.l4 = (a.l4 + 0xFFFFFFFFFFFFE) - b.l4
+	return v.carryPropagate()
+}
+
+// Negate sets v = -a, and returns v.
+func (v *Element) Negate(a *Element) *Element {
+	return v.Subtract(feZero, a)
+}
+
+// Invert sets v = 1/z mod p, and returns v.
+//
+// If z == 0, Invert returns v = 0.
+func (v *Element) Invert(z *Element) *Element {
+	// Inversion is implemented as exponentiation with exponent p − 2. It uses the
+	// same sequence of 255 squarings and 11 multiplications as [Curve25519].
+	var z2, z9, z11, z2_5_0, z2_10_0, z2_20_0, z2_50_0, z2_100_0, t Element
+
+	z2.Square(z)             // 2
+	t.Square(&z2)            // 4
+	t.Square(&t)             // 8
+	z9.Multiply(&t, z)       // 9
+	z11.Multiply(&z9, &z2)   // 11
+	t.Square(&z11)           // 22
+	z2_5_0.Multiply(&t, &z9) // 31 = 2^5 - 2^0
+
+	t.Square(&z2_5_0) // 2^6 - 2^1
+	for i := 0; i < 4; i++ {
+		t.Square(&t) // 2^10 - 2^5
+	}
+	z2_10_0.Multiply(&t, &z2_5_0) // 2^10 - 2^0
+
+	t.Square(&z2_10_0) // 2^11 - 2^1
+	for i := 0; i < 9; i++ {
+		t.Square(&t) // 2^20 - 2^10
+	}
+	z2_20_0.Multiply(&t, &z2_10_0) // 2^20 - 2^0
+
+	t.Square(&z2_20_0) // 2^21 - 2^1
+	for i := 0; i < 19; i++ {
+		t.Square(&t) // 2^40 - 2^20
+	}
+	t.Multiply(&t, &z2_20_0) // 2^40 - 2^0
+
+	t.Square(&t) // 2^41 - 2^1
+	for i := 0; i < 9; i++ {
+		t.Square(&t) // 2^50 - 2^10
+	}
+	z2_50_0.Multiply(&t, &z2_10_0) // 2^50 - 2^0
+
+	t.Square(&z2_50_0) // 2^51 - 2^1
+	for i := 0; i < 49; i++ {
+		t.Square(&t) // 2^100 - 2^50
+	}
+	z2_100_0.Multiply(&t, &z2_50_0) // 2^100 - 2^0
+
+	t.Square(&z2_100_0) // 2^101 - 2^1
+	for i := 0; i < 99; i++ {
+		t.Square(&t) // 2^200 - 2^100
+	}
+	t.Multiply(&t, &z2_100_0) // 2^200 - 2^0
+
+	t.Square(&t) // 2^201 - 2^1
+	for i := 0; i < 49; i++ {
+		t.Square(&t) // 2^250 - 2^50
+	}
+	t.Multiply(&t, &z2_50_0) // 2^250 - 2^0
+
+	t.Square(&t) // 2^251 - 2^1
+	t.Square(&t) // 2^252 - 2^2
+	t.Square(&t) // 2^253 - 2^3
+	t.Square(&t) // 2^254 - 2^4
+	t.Square(&t) // 2^255 - 2^5
+
+	return v.Multiply(&t, &z11) // 2^255 - 21
+}
+
+// Set sets v = a, and returns v.
+func (v *Element) Set(a *Element) *Element {
+	*v = *a
+	return v
+}
+
+// SetBytes sets v to x, which must be a 32-byte little-endian encoding.
+//
+// Consistent with RFC 7748, the most significant bit (the high bit of the
+// last byte) is ignored, and non-canonical values (2^255-19 through 2^255-1)
+// are accepted. Note that this is laxer than specified by RFC 8032.
+func (v *Element) SetBytes(x []byte) *Element {
+	if len(x) != 32 {
+		panic("edwards25519: invalid field element input size")
+	}
+
+	// Bits 0:51 (bytes 0:8, bits 0:64, shift 0, mask 51).
+	v.l0 = binary.LittleEndian.Uint64(x[0:8])
+	v.l0 &= maskLow51Bits
+	// Bits 51:102 (bytes 6:14, bits 48:112, shift 3, mask 51).
+	v.l1 = binary.LittleEndian.Uint64(x[6:14]) >> 3
+	v.l1 &= maskLow51Bits
+	// Bits 102:153 (bytes 12:20, bits 96:160, shift 6, mask 51).
+	v.l2 = binary.LittleEndian.Uint64(x[12:20]) >> 6
+	v.l2 &= maskLow51Bits
+	// Bits 153:204 (bytes 19:27, bits 152:216, shift 1, mask 51).
+	v.l3 = binary.LittleEndian.Uint64(x[19:27]) >> 1
+	v.l3 &= maskLow51Bits
+	// Bits 204:251 (bytes 24:32, bits 192:256, shift 12, mask 51).
+	// Note: not bytes 25:33, shift 4, to avoid overread.
+	v.l4 = binary.LittleEndian.Uint64(x[24:32]) >> 12
+	v.l4 &= maskLow51Bits
+
+	return v
+}
+
+// Bytes returns the canonical 32-byte little-endian encoding of v.
+func (v *Element) Bytes() []byte {
+	// This function is outlined to make the allocations inline in the caller
+	// rather than happen on the heap.
+	var out [32]byte
+	return v.bytes(&out)
+}
+
+func (v *Element) bytes(out *[32]byte) []byte {
+	t := *v
+	t.reduce()
+
+	var buf [8]byte
+	for i, l := range [5]uint64{t.l0, t.l1, t.l2, t.l3, t.l4} {
+		bitsOffset := i * 51
+		binary.LittleEndian.PutUint64(buf[:], l<<uint(bitsOffset%8))
+		for i, bb := range buf {
+			off := bitsOffset/8 + i
+			if off >= len(out) {
+				break
+			}
+			out[off] |= bb
+		}
+	}
+
+	return out[:]
+}
+
+// Equal returns 1 if v and u are equal, and 0 otherwise.
+func (v *Element) Equal(u *Element) int {
+	sa, sv := u.Bytes(), v.Bytes()
+	return subtle.ConstantTimeCompare(sa, sv)
+}
+
+// mask64Bits returns 0xffffffff if cond is 1, and 0 otherwise.
+func mask64Bits(cond int) uint64 { return ^(uint64(cond) - 1) }
+
+// Select sets v to a if cond == 1, and to b if cond == 0.
+func (v *Element) Select(a, b *Element, cond int) *Element {
+	m := mask64Bits(cond)
+	v.l0 = (m & a.l0) | (^m & b.l0)
+	v.l1 = (m & a.l1) | (^m & b.l1)
+	v.l2 = (m & a.l2) | (^m & b.l2)
+	v.l3 = (m & a.l3) | (^m & b.l3)
+	v.l4 = (m & a.l4) | (^m & b.l4)
+	return v
+}
+
+// Swap swaps v and u if cond == 1 or leaves them unchanged if cond == 0, and returns v.
+func (v *Element) Swap(u *Element, cond int) {
+	m := mask64Bits(cond)
+	t := m & (v.l0 ^ u.l0)
+	v.l0 ^= t
+	u.l0 ^= t
+	t = m & (v.l1 ^ u.l1)
+	v.l1 ^= t
+	u.l1 ^= t
+	t = m & (v.l2 ^ u.l2)
+	v.l2 ^= t
+	u.l2 ^= t
+	t = m & (v.l3 ^ u.l3)
+	v.l3 ^= t
+	u.l3 ^= t
+	t = m & (v.l4 ^ u.l4)
+	v.l4 ^= t
+	u.l4 ^= t
+}
+
+// IsNegative returns 1 if v is negative, and 0 otherwise.
+func (v *Element) IsNegative() int {
+	return int(v.Bytes()[0] & 1)
+}
+
+// Absolute sets v to |u|, and returns v.
+func (v *Element) Absolute(u *Element) *Element {
+	return v.Select(new(Element).Negate(u), u, u.IsNegative())
+}
+
+// Multiply sets v = x * y, and returns v.
+func (v *Element) Multiply(x, y *Element) *Element {
+	feMul(v, x, y)
+	return v
+}
+
+// Square sets v = x * x, and returns v.
+func (v *Element) Square(x *Element) *Element {
+	feSquare(v, x)
+	return v
+}
+
+// Mult32 sets v = x * y, and returns v.
+func (v *Element) Mult32(x *Element, y uint32) *Element {
+	x0lo, x0hi := mul51(x.l0, y)
+	x1lo, x1hi := mul51(x.l1, y)
+	x2lo, x2hi := mul51(x.l2, y)
+	x3lo, x3hi := mul51(x.l3, y)
+	x4lo, x4hi := mul51(x.l4, y)
+	v.l0 = x0lo + 19*x4hi // carried over per the reduction identity
+	v.l1 = x1lo + x0hi
+	v.l2 = x2lo + x1hi
+	v.l3 = x3lo + x2hi
+	v.l4 = x4lo + x3hi
+	// The hi portions are going to be only 32 bits, plus any previous excess,
+	// so we can skip the carry propagation.
+	return v
+}
+
+// mul51 returns lo + hi * 2⁵¹ = a * b.
+func mul51(a uint64, b uint32) (lo uint64, hi uint64) {
+	mh, ml := bits.Mul64(a, uint64(b))
+	lo = ml & maskLow51Bits
+	hi = (mh << 13) | (ml >> 51)
+	return
+}
+
+// Pow22523 set v = x^((p-5)/8), and returns v. (p-5)/8 is 2^252-3.
+func (v *Element) Pow22523(x *Element) *Element {
+	var t0, t1, t2 Element
+
+	t0.Square(x)             // x^2
+	t1.Square(&t0)           // x^4
+	t1.Square(&t1)           // x^8
+	t1.Multiply(x, &t1)      // x^9
+	t0.Multiply(&t0, &t1)    // x^11
+	t0.Square(&t0)           // x^22
+	t0.Multiply(&t1, &t0)    // x^31
+	t1.Square(&t0)           // x^62
+	for i := 1; i < 5; i++ { // x^992
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // x^1023 -> 1023 = 2^10 - 1
+	t1.Square(&t0)            // 2^11 - 2
+	for i := 1; i < 10; i++ { // 2^20 - 2^10
+		t1.Square(&t1)
+	}
+	t1.Multiply(&t1, &t0)     // 2^20 - 1
+	t2.Square(&t1)            // 2^21 - 2
+	for i := 1; i < 20; i++ { // 2^40 - 2^20
+		t2.Square(&t2)
+	}
+	t1.Multiply(&t2, &t1)     // 2^40 - 1
+	t1.Square(&t1)            // 2^41 - 2
+	for i := 1; i < 10; i++ { // 2^50 - 2^10
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // 2^50 - 1
+	t1.Square(&t0)            // 2^51 - 2
+	for i := 1; i < 50; i++ { // 2^100 - 2^50
+		t1.Square(&t1)
+	}
+	t1.Multiply(&t1, &t0)      // 2^100 - 1
+	t2.Square(&t1)             // 2^101 - 2
+	for i := 1; i < 100; i++ { // 2^200 - 2^100
+		t2.Square(&t2)
+	}
+	t1.Multiply(&t2, &t1)     // 2^200 - 1
+	t1.Square(&t1)            // 2^201 - 2
+	for i := 1; i < 50; i++ { // 2^250 - 2^50
+		t1.Square(&t1)
+	}
+	t0.Multiply(&t1, &t0)     // 2^250 - 1
+	t0.Square(&t0)            // 2^251 - 2
+	t0.Square(&t0)            // 2^252 - 4
+	return v.Multiply(&t0, x) // 2^252 - 3 -> x^(2^252-3)
+}
+
+// sqrtM1 is 2^((p-1)/4), which squared is equal to -1 by Euler's Criterion.
+var sqrtM1 = &Element{1718705420411056, 234908883556509,
+	2233514472574048, 2117202627021982, 765476049583133}
+
+// SqrtRatio sets r to the non-negative square root of the ratio of u and v.
+//
+// If u/v is square, SqrtRatio returns r and 1. If u/v is not square, SqrtRatio
+// sets r according to Section 4.3 of draft-irtf-cfrg-ristretto255-decaf448-00,
+// and returns r and 0.
+func (r *Element) SqrtRatio(u, v *Element) (rr *Element, wasSquare int) {
+	var a, b Element
+
+	// r = (u * v3) * (u * v7)^((p-5)/8)
+	v2 := a.Square(v)
+	uv3 := b.Multiply(u, b.Multiply(v2, v))
+	uv7 := a.Multiply(uv3, a.Square(v2))
+	r.Multiply(uv3, r.Pow22523(uv7))
+
+	check := a.Multiply(v, a.Square(r)) // check = v * r^2
+
+	uNeg := b.Negate(u)
+	correctSignSqrt := check.Equal(u)
+	flippedSignSqrt := check.Equal(uNeg)
+	flippedSignSqrtI := check.Equal(uNeg.Multiply(uNeg, sqrtM1))
+
+	rPrime := b.Multiply(r, sqrtM1) // r_prime = SQRT_M1 * r
+	// r = CT_SELECT(r_prime IF flipped_sign_sqrt | flipped_sign_sqrt_i ELSE r)
+	r.Select(rPrime, r, flippedSignSqrt|flippedSignSqrtI)
+
+	r.Absolute(r) // Choose the nonnegative square root.
+	return r, correctSignSqrt | flippedSignSqrt
+}
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go
new file mode 100644
index 00000000..44dc8e8c
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.go
@@ -0,0 +1,13 @@
+// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT.
+
+// +build amd64,gc,!purego
+
+package field
+
+// feMul sets out = a * b. It works like feMulGeneric.
+//go:noescape
+func feMul(out *Element, a *Element, b *Element)
+
+// feSquare sets out = a * a. It works like feSquareGeneric.
+//go:noescape
+func feSquare(out *Element, a *Element)
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.s b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.s
new file mode 100644
index 00000000..293f013c
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64.s
@@ -0,0 +1,379 @@
+// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#include "textflag.h"
+
+// func feMul(out *Element, a *Element, b *Element)
+TEXT ·feMul(SB), NOSPLIT, $0-24
+	MOVQ a+8(FP), CX
+	MOVQ b+16(FP), BX
+
+	// r0 = a0×b0
+	MOVQ (CX), AX
+	MULQ (BX)
+	MOVQ AX, DI
+	MOVQ DX, SI
+
+	// r0 += 19×a1×b4
+	MOVQ   8(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a2×b3
+	MOVQ   16(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a3×b2
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   16(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r0 += 19×a4×b1
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   8(BX)
+	ADDQ   AX, DI
+	ADCQ   DX, SI
+
+	// r1 = a0×b1
+	MOVQ (CX), AX
+	MULQ 8(BX)
+	MOVQ AX, R9
+	MOVQ DX, R8
+
+	// r1 += a1×b0
+	MOVQ 8(CX), AX
+	MULQ (BX)
+	ADDQ AX, R9
+	ADCQ DX, R8
+
+	// r1 += 19×a2×b4
+	MOVQ   16(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r1 += 19×a3×b3
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r1 += 19×a4×b2
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   16(BX)
+	ADDQ   AX, R9
+	ADCQ   DX, R8
+
+	// r2 = a0×b2
+	MOVQ (CX), AX
+	MULQ 16(BX)
+	MOVQ AX, R11
+	MOVQ DX, R10
+
+	// r2 += a1×b1
+	MOVQ 8(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R11
+	ADCQ DX, R10
+
+	// r2 += a2×b0
+	MOVQ 16(CX), AX
+	MULQ (BX)
+	ADDQ AX, R11
+	ADCQ DX, R10
+
+	// r2 += 19×a3×b4
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R11
+	ADCQ   DX, R10
+
+	// r2 += 19×a4×b3
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(BX)
+	ADDQ   AX, R11
+	ADCQ   DX, R10
+
+	// r3 = a0×b3
+	MOVQ (CX), AX
+	MULQ 24(BX)
+	MOVQ AX, R13
+	MOVQ DX, R12
+
+	// r3 += a1×b2
+	MOVQ 8(CX), AX
+	MULQ 16(BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += a2×b1
+	MOVQ 16(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += a3×b0
+	MOVQ 24(CX), AX
+	MULQ (BX)
+	ADDQ AX, R13
+	ADCQ DX, R12
+
+	// r3 += 19×a4×b4
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(BX)
+	ADDQ   AX, R13
+	ADCQ   DX, R12
+
+	// r4 = a0×b4
+	MOVQ (CX), AX
+	MULQ 32(BX)
+	MOVQ AX, R15
+	MOVQ DX, R14
+
+	// r4 += a1×b3
+	MOVQ 8(CX), AX
+	MULQ 24(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a2×b2
+	MOVQ 16(CX), AX
+	MULQ 16(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a3×b1
+	MOVQ 24(CX), AX
+	MULQ 8(BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// r4 += a4×b0
+	MOVQ 32(CX), AX
+	MULQ (BX)
+	ADDQ AX, R15
+	ADCQ DX, R14
+
+	// First reduction chain
+	MOVQ   $0x0007ffffffffffff, AX
+	SHLQ   $0x0d, DI, SI
+	SHLQ   $0x0d, R9, R8
+	SHLQ   $0x0d, R11, R10
+	SHLQ   $0x0d, R13, R12
+	SHLQ   $0x0d, R15, R14
+	ANDQ   AX, DI
+	IMUL3Q $0x13, R14, R14
+	ADDQ   R14, DI
+	ANDQ   AX, R9
+	ADDQ   SI, R9
+	ANDQ   AX, R11
+	ADDQ   R8, R11
+	ANDQ   AX, R13
+	ADDQ   R10, R13
+	ANDQ   AX, R15
+	ADDQ   R12, R15
+
+	// Second reduction chain (carryPropagate)
+	MOVQ   DI, SI
+	SHRQ   $0x33, SI
+	MOVQ   R9, R8
+	SHRQ   $0x33, R8
+	MOVQ   R11, R10
+	SHRQ   $0x33, R10
+	MOVQ   R13, R12
+	SHRQ   $0x33, R12
+	MOVQ   R15, R14
+	SHRQ   $0x33, R14
+	ANDQ   AX, DI
+	IMUL3Q $0x13, R14, R14
+	ADDQ   R14, DI
+	ANDQ   AX, R9
+	ADDQ   SI, R9
+	ANDQ   AX, R11
+	ADDQ   R8, R11
+	ANDQ   AX, R13
+	ADDQ   R10, R13
+	ANDQ   AX, R15
+	ADDQ   R12, R15
+
+	// Store output
+	MOVQ out+0(FP), AX
+	MOVQ DI, (AX)
+	MOVQ R9, 8(AX)
+	MOVQ R11, 16(AX)
+	MOVQ R13, 24(AX)
+	MOVQ R15, 32(AX)
+	RET
+
+// func feSquare(out *Element, a *Element)
+TEXT ·feSquare(SB), NOSPLIT, $0-16
+	MOVQ a+8(FP), CX
+
+	// r0 = l0×l0
+	MOVQ (CX), AX
+	MULQ (CX)
+	MOVQ AX, SI
+	MOVQ DX, BX
+
+	// r0 += 38×l1×l4
+	MOVQ   8(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, SI
+	ADCQ   DX, BX
+
+	// r0 += 38×l2×l3
+	MOVQ   16(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, SI
+	ADCQ   DX, BX
+
+	// r1 = 2×l0×l1
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 8(CX)
+	MOVQ AX, R8
+	MOVQ DX, DI
+
+	// r1 += 38×l2×l4
+	MOVQ   16(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R8
+	ADCQ   DX, DI
+
+	// r1 += 19×l3×l3
+	MOVQ   24(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, R8
+	ADCQ   DX, DI
+
+	// r2 = 2×l0×l2
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 16(CX)
+	MOVQ AX, R10
+	MOVQ DX, R9
+
+	// r2 += l1×l1
+	MOVQ 8(CX), AX
+	MULQ 8(CX)
+	ADDQ AX, R10
+	ADCQ DX, R9
+
+	// r2 += 38×l3×l4
+	MOVQ   24(CX), AX
+	IMUL3Q $0x26, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R10
+	ADCQ   DX, R9
+
+	// r3 = 2×l0×l3
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 24(CX)
+	MOVQ AX, R12
+	MOVQ DX, R11
+
+	// r3 += 2×l1×l2
+	MOVQ   8(CX), AX
+	IMUL3Q $0x02, AX, AX
+	MULQ   16(CX)
+	ADDQ   AX, R12
+	ADCQ   DX, R11
+
+	// r3 += 19×l4×l4
+	MOVQ   32(CX), AX
+	IMUL3Q $0x13, AX, AX
+	MULQ   32(CX)
+	ADDQ   AX, R12
+	ADCQ   DX, R11
+
+	// r4 = 2×l0×l4
+	MOVQ (CX), AX
+	SHLQ $0x01, AX
+	MULQ 32(CX)
+	MOVQ AX, R14
+	MOVQ DX, R13
+
+	// r4 += 2×l1×l3
+	MOVQ   8(CX), AX
+	IMUL3Q $0x02, AX, AX
+	MULQ   24(CX)
+	ADDQ   AX, R14
+	ADCQ   DX, R13
+
+	// r4 += l2×l2
+	MOVQ 16(CX), AX
+	MULQ 16(CX)
+	ADDQ AX, R14
+	ADCQ DX, R13
+
+	// First reduction chain
+	MOVQ   $0x0007ffffffffffff, AX
+	SHLQ   $0x0d, SI, BX
+	SHLQ   $0x0d, R8, DI
+	SHLQ   $0x0d, R10, R9
+	SHLQ   $0x0d, R12, R11
+	SHLQ   $0x0d, R14, R13
+	ANDQ   AX, SI
+	IMUL3Q $0x13, R13, R13
+	ADDQ   R13, SI
+	ANDQ   AX, R8
+	ADDQ   BX, R8
+	ANDQ   AX, R10
+	ADDQ   DI, R10
+	ANDQ   AX, R12
+	ADDQ   R9, R12
+	ANDQ   AX, R14
+	ADDQ   R11, R14
+
+	// Second reduction chain (carryPropagate)
+	MOVQ   SI, BX
+	SHRQ   $0x33, BX
+	MOVQ   R8, DI
+	SHRQ   $0x33, DI
+	MOVQ   R10, R9
+	SHRQ   $0x33, R9
+	MOVQ   R12, R11
+	SHRQ   $0x33, R11
+	MOVQ   R14, R13
+	SHRQ   $0x33, R13
+	ANDQ   AX, SI
+	IMUL3Q $0x13, R13, R13
+	ADDQ   R13, SI
+	ANDQ   AX, R8
+	ADDQ   BX, R8
+	ANDQ   AX, R10
+	ADDQ   DI, R10
+	ANDQ   AX, R12
+	ADDQ   R9, R12
+	ANDQ   AX, R14
+	ADDQ   R11, R14
+
+	// Store output
+	MOVQ out+0(FP), AX
+	MOVQ SI, (AX)
+	MOVQ R8, 8(AX)
+	MOVQ R10, 16(AX)
+	MOVQ R12, 24(AX)
+	MOVQ R14, 32(AX)
+	RET
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go
new file mode 100644
index 00000000..ddb6c9b8
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_amd64_noasm.go
@@ -0,0 +1,12 @@
+// Copyright (c) 2019 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 || !gc || purego
+// +build !amd64 !gc purego
+
+package field
+
+func feMul(v, x, y *Element) { feMulGeneric(v, x, y) }
+
+func feSquare(v, x *Element) { feSquareGeneric(v, x) }
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go
new file mode 100644
index 00000000..af459ef5
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.go
@@ -0,0 +1,16 @@
+// Copyright (c) 2020 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 arm64 && gc && !purego
+// +build arm64,gc,!purego
+
+package field
+
+//go:noescape
+func carryPropagate(v *Element)
+
+func (v *Element) carryPropagate() *Element {
+	carryPropagate(v)
+	return v
+}
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.s b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.s
new file mode 100644
index 00000000..5c91e458
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64.s
@@ -0,0 +1,43 @@
+// Copyright (c) 2020 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 arm64 && gc && !purego
+// +build arm64,gc,!purego
+
+#include "textflag.h"
+
+// carryPropagate works exactly like carryPropagateGeneric and uses the
+// same AND, ADD, and LSR+MADD instructions emitted by the compiler, but
+// avoids loading R0-R4 twice and uses LDP and STP.
+//
+// See https://golang.org/issues/43145 for the main compiler issue.
+//
+// func carryPropagate(v *Element)
+TEXT ·carryPropagate(SB),NOFRAME|NOSPLIT,$0-8
+	MOVD v+0(FP), R20
+
+	LDP 0(R20), (R0, R1)
+	LDP 16(R20), (R2, R3)
+	MOVD 32(R20), R4
+
+	AND $0x7ffffffffffff, R0, R10
+	AND $0x7ffffffffffff, R1, R11
+	AND $0x7ffffffffffff, R2, R12
+	AND $0x7ffffffffffff, R3, R13
+	AND $0x7ffffffffffff, R4, R14
+
+	ADD R0>>51, R11, R11
+	ADD R1>>51, R12, R12
+	ADD R2>>51, R13, R13
+	ADD R3>>51, R14, R14
+	// R4>>51 * 19 + R10 -> R10
+	LSR $51, R4, R21
+	MOVD $19, R22
+	MADD R22, R10, R21, R10
+
+	STP (R10, R11), 0(R20)
+	STP (R12, R13), 16(R20)
+	MOVD R14, 32(R20)
+
+	RET
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go
new file mode 100644
index 00000000..234a5b2e
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_arm64_noasm.go
@@ -0,0 +1,12 @@
+// Copyright (c) 2021 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 !arm64 || !gc || purego
+// +build !arm64 !gc purego
+
+package field
+
+func (v *Element) carryPropagate() *Element {
+	return v.carryPropagateGeneric()
+}
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/fe_generic.go b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_generic.go
new file mode 100644
index 00000000..7b5b78cb
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/fe_generic.go
@@ -0,0 +1,264 @@
+// Copyright (c) 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.
+
+package field
+
+import "math/bits"
+
+// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
+// bits.Mul64 and bits.Add64 intrinsics.
+type uint128 struct {
+	lo, hi uint64
+}
+
+// mul64 returns a * b.
+func mul64(a, b uint64) uint128 {
+	hi, lo := bits.Mul64(a, b)
+	return uint128{lo, hi}
+}
+
+// addMul64 returns v + a * b.
+func addMul64(v uint128, a, b uint64) uint128 {
+	hi, lo := bits.Mul64(a, b)
+	lo, c := bits.Add64(lo, v.lo, 0)
+	hi, _ = bits.Add64(hi, v.hi, c)
+	return uint128{lo, hi}
+}
+
+// shiftRightBy51 returns a >> 51. a is assumed to be at most 115 bits.
+func shiftRightBy51(a uint128) uint64 {
+	return (a.hi << (64 - 51)) | (a.lo >> 51)
+}
+
+func feMulGeneric(v, a, b *Element) {
+	a0 := a.l0
+	a1 := a.l1
+	a2 := a.l2
+	a3 := a.l3
+	a4 := a.l4
+
+	b0 := b.l0
+	b1 := b.l1
+	b2 := b.l2
+	b3 := b.l3
+	b4 := b.l4
+
+	// Limb multiplication works like pen-and-paper columnar multiplication, but
+	// with 51-bit limbs instead of digits.
+	//
+	//                          a4   a3   a2   a1   a0  x
+	//                          b4   b3   b2   b1   b0  =
+	//                         ------------------------
+	//                        a4b0 a3b0 a2b0 a1b0 a0b0  +
+	//                   a4b1 a3b1 a2b1 a1b1 a0b1       +
+	//              a4b2 a3b2 a2b2 a1b2 a0b2            +
+	//         a4b3 a3b3 a2b3 a1b3 a0b3                 +
+	//    a4b4 a3b4 a2b4 a1b4 a0b4                      =
+	//   ----------------------------------------------
+	//      r8   r7   r6   r5   r4   r3   r2   r1   r0
+	//
+	// We can then use the reduction identity (a * 2²⁵⁵ + b = a * 19 + b) to
+	// reduce the limbs that would overflow 255 bits. r5 * 2²⁵⁵ becomes 19 * r5,
+	// r6 * 2³⁰⁶ becomes 19 * r6 * 2⁵¹, etc.
+	//
+	// Reduction can be carried out simultaneously to multiplication. For
+	// example, we do not compute r5: whenever the result of a multiplication
+	// belongs to r5, like a1b4, we multiply it by 19 and add the result to r0.
+	//
+	//            a4b0    a3b0    a2b0    a1b0    a0b0  +
+	//            a3b1    a2b1    a1b1    a0b1 19×a4b1  +
+	//            a2b2    a1b2    a0b2 19×a4b2 19×a3b2  +
+	//            a1b3    a0b3 19×a4b3 19×a3b3 19×a2b3  +
+	//            a0b4 19×a4b4 19×a3b4 19×a2b4 19×a1b4  =
+	//           --------------------------------------
+	//              r4      r3      r2      r1      r0
+	//
+	// Finally we add up the columns into wide, overlapping limbs.
+
+	a1_19 := a1 * 19
+	a2_19 := a2 * 19
+	a3_19 := a3 * 19
+	a4_19 := a4 * 19
+
+	// r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1)
+	r0 := mul64(a0, b0)
+	r0 = addMul64(r0, a1_19, b4)
+	r0 = addMul64(r0, a2_19, b3)
+	r0 = addMul64(r0, a3_19, b2)
+	r0 = addMul64(r0, a4_19, b1)
+
+	// r1 = a0×b1 + a1×b0 + 19×(a2×b4 + a3×b3 + a4×b2)
+	r1 := mul64(a0, b1)
+	r1 = addMul64(r1, a1, b0)
+	r1 = addMul64(r1, a2_19, b4)
+	r1 = addMul64(r1, a3_19, b3)
+	r1 = addMul64(r1, a4_19, b2)
+
+	// r2 = a0×b2 + a1×b1 + a2×b0 + 19×(a3×b4 + a4×b3)
+	r2 := mul64(a0, b2)
+	r2 = addMul64(r2, a1, b1)
+	r2 = addMul64(r2, a2, b0)
+	r2 = addMul64(r2, a3_19, b4)
+	r2 = addMul64(r2, a4_19, b3)
+
+	// r3 = a0×b3 + a1×b2 + a2×b1 + a3×b0 + 19×a4×b4
+	r3 := mul64(a0, b3)
+	r3 = addMul64(r3, a1, b2)
+	r3 = addMul64(r3, a2, b1)
+	r3 = addMul64(r3, a3, b0)
+	r3 = addMul64(r3, a4_19, b4)
+
+	// r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0
+	r4 := mul64(a0, b4)
+	r4 = addMul64(r4, a1, b3)
+	r4 = addMul64(r4, a2, b2)
+	r4 = addMul64(r4, a3, b1)
+	r4 = addMul64(r4, a4, b0)
+
+	// After the multiplication, we need to reduce (carry) the five coefficients
+	// to obtain a result with limbs that are at most slightly larger than 2⁵¹,
+	// to respect the Element invariant.
+	//
+	// Overall, the reduction works the same as carryPropagate, except with
+	// wider inputs: we take the carry for each coefficient by shifting it right
+	// by 51, and add it to the limb above it. The top carry is multiplied by 19
+	// according to the reduction identity and added to the lowest limb.
+	//
+	// The largest coefficient (r0) will be at most 111 bits, which guarantees
+	// that all carries are at most 111 - 51 = 60 bits, which fits in a uint64.
+	//
+	//     r0 = a0×b0 + 19×(a1×b4 + a2×b3 + a3×b2 + a4×b1)
+	//     r0 < 2⁵²×2⁵² + 19×(2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵² + 2⁵²×2⁵²)
+	//     r0 < (1 + 19 × 4) × 2⁵² × 2⁵²
+	//     r0 < 2⁷ × 2⁵² × 2⁵²
+	//     r0 < 2¹¹¹
+	//
+	// Moreover, the top coefficient (r4) is at most 107 bits, so c4 is at most
+	// 56 bits, and c4 * 19 is at most 61 bits, which again fits in a uint64 and
+	// allows us to easily apply the reduction identity.
+	//
+	//     r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0
+	//     r4 < 5 × 2⁵² × 2⁵²
+	//     r4 < 2¹⁰⁷
+	//
+
+	c0 := shiftRightBy51(r0)
+	c1 := shiftRightBy51(r1)
+	c2 := shiftRightBy51(r2)
+	c3 := shiftRightBy51(r3)
+	c4 := shiftRightBy51(r4)
+
+	rr0 := r0.lo&maskLow51Bits + c4*19
+	rr1 := r1.lo&maskLow51Bits + c0
+	rr2 := r2.lo&maskLow51Bits + c1
+	rr3 := r3.lo&maskLow51Bits + c2
+	rr4 := r4.lo&maskLow51Bits + c3
+
+	// Now all coefficients fit into 64-bit registers but are still too large to
+	// be passed around as a Element. We therefore do one last carry chain,
+	// where the carries will be small enough to fit in the wiggle room above 2⁵¹.
+	*v = Element{rr0, rr1, rr2, rr3, rr4}
+	v.carryPropagate()
+}
+
+func feSquareGeneric(v, a *Element) {
+	l0 := a.l0
+	l1 := a.l1
+	l2 := a.l2
+	l3 := a.l3
+	l4 := a.l4
+
+	// Squaring works precisely like multiplication above, but thanks to its
+	// symmetry we get to group a few terms together.
+	//
+	//                          l4   l3   l2   l1   l0  x
+	//                          l4   l3   l2   l1   l0  =
+	//                         ------------------------
+	//                        l4l0 l3l0 l2l0 l1l0 l0l0  +
+	//                   l4l1 l3l1 l2l1 l1l1 l0l1       +
+	//              l4l2 l3l2 l2l2 l1l2 l0l2            +
+	//         l4l3 l3l3 l2l3 l1l3 l0l3                 +
+	//    l4l4 l3l4 l2l4 l1l4 l0l4                      =
+	//   ----------------------------------------------
+	//      r8   r7   r6   r5   r4   r3   r2   r1   r0
+	//
+	//            l4l0    l3l0    l2l0    l1l0    l0l0  +
+	//            l3l1    l2l1    l1l1    l0l1 19×l4l1  +
+	//            l2l2    l1l2    l0l2 19×l4l2 19×l3l2  +
+	//            l1l3    l0l3 19×l4l3 19×l3l3 19×l2l3  +
+	//            l0l4 19×l4l4 19×l3l4 19×l2l4 19×l1l4  =
+	//           --------------------------------------
+	//              r4      r3      r2      r1      r0
+	//
+	// With precomputed 2×, 19×, and 2×19× terms, we can compute each limb with
+	// only three Mul64 and four Add64, instead of five and eight.
+
+	l0_2 := l0 * 2
+	l1_2 := l1 * 2
+
+	l1_38 := l1 * 38
+	l2_38 := l2 * 38
+	l3_38 := l3 * 38
+
+	l3_19 := l3 * 19
+	l4_19 := l4 * 19
+
+	// r0 = l0×l0 + 19×(l1×l4 + l2×l3 + l3×l2 + l4×l1) = l0×l0 + 19×2×(l1×l4 + l2×l3)
+	r0 := mul64(l0, l0)
+	r0 = addMul64(r0, l1_38, l4)
+	r0 = addMul64(r0, l2_38, l3)
+
+	// r1 = l0×l1 + l1×l0 + 19×(l2×l4 + l3×l3 + l4×l2) = 2×l0×l1 + 19×2×l2×l4 + 19×l3×l3
+	r1 := mul64(l0_2, l1)
+	r1 = addMul64(r1, l2_38, l4)
+	r1 = addMul64(r1, l3_19, l3)
+
+	// r2 = l0×l2 + l1×l1 + l2×l0 + 19×(l3×l4 + l4×l3) = 2×l0×l2 + l1×l1 + 19×2×l3×l4
+	r2 := mul64(l0_2, l2)
+	r2 = addMul64(r2, l1, l1)
+	r2 = addMul64(r2, l3_38, l4)
+
+	// r3 = l0×l3 + l1×l2 + l2×l1 + l3×l0 + 19×l4×l4 = 2×l0×l3 + 2×l1×l2 + 19×l4×l4
+	r3 := mul64(l0_2, l3)
+	r3 = addMul64(r3, l1_2, l2)
+	r3 = addMul64(r3, l4_19, l4)
+
+	// r4 = l0×l4 + l1×l3 + l2×l2 + l3×l1 + l4×l0 = 2×l0×l4 + 2×l1×l3 + l2×l2
+	r4 := mul64(l0_2, l4)
+	r4 = addMul64(r4, l1_2, l3)
+	r4 = addMul64(r4, l2, l2)
+
+	c0 := shiftRightBy51(r0)
+	c1 := shiftRightBy51(r1)
+	c2 := shiftRightBy51(r2)
+	c3 := shiftRightBy51(r3)
+	c4 := shiftRightBy51(r4)
+
+	rr0 := r0.lo&maskLow51Bits + c4*19
+	rr1 := r1.lo&maskLow51Bits + c0
+	rr2 := r2.lo&maskLow51Bits + c1
+	rr3 := r3.lo&maskLow51Bits + c2
+	rr4 := r4.lo&maskLow51Bits + c3
+
+	*v = Element{rr0, rr1, rr2, rr3, rr4}
+	v.carryPropagate()
+}
+
+// carryPropagate brings the limbs below 52 bits by applying the reduction
+// identity (a * 2²⁵⁵ + b = a * 19 + b) to the l4 carry. TODO inline
+func (v *Element) carryPropagateGeneric() *Element {
+	c0 := v.l0 >> 51
+	c1 := v.l1 >> 51
+	c2 := v.l2 >> 51
+	c3 := v.l3 >> 51
+	c4 := v.l4 >> 51
+
+	v.l0 = v.l0&maskLow51Bits + c4*19
+	v.l1 = v.l1&maskLow51Bits + c0
+	v.l2 = v.l2&maskLow51Bits + c1
+	v.l3 = v.l3&maskLow51Bits + c2
+	v.l4 = v.l4&maskLow51Bits + c3
+
+	return v
+}
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/sync.checkpoint b/vendor/golang.org/x/crypto/curve25519/internal/field/sync.checkpoint
new file mode 100644
index 00000000..e3685f95
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/sync.checkpoint
@@ -0,0 +1 @@
+b0c49ae9f59d233526f8934262c5bbbe14d4358d
diff --git a/vendor/golang.org/x/crypto/curve25519/internal/field/sync.sh b/vendor/golang.org/x/crypto/curve25519/internal/field/sync.sh
new file mode 100644
index 00000000..1ba22a8b
--- /dev/null
+++ b/vendor/golang.org/x/crypto/curve25519/internal/field/sync.sh
@@ -0,0 +1,19 @@
+#! /bin/bash
+set -euo pipefail
+
+cd "$(git rev-parse --show-toplevel)"
+
+STD_PATH=src/crypto/ed25519/internal/edwards25519/field
+LOCAL_PATH=curve25519/internal/field
+LAST_SYNC_REF=$(cat $LOCAL_PATH/sync.checkpoint)
+
+git fetch https://go.googlesource.com/go master
+
+if git diff --quiet $LAST_SYNC_REF:$STD_PATH FETCH_HEAD:$STD_PATH; then
+    echo "No changes."
+else
+    NEW_REF=$(git rev-parse FETCH_HEAD | tee $LOCAL_PATH/sync.checkpoint)
+    echo "Applying changes from $LAST_SYNC_REF to $NEW_REF..."
+    git diff $LAST_SYNC_REF:$STD_PATH FETCH_HEAD:$STD_PATH | \
+        git apply -3 --directory=$LOCAL_PATH
+fi
diff --git a/vendor/golang.org/x/crypto/pbkdf2/pbkdf2.go b/vendor/golang.org/x/crypto/pbkdf2/pbkdf2.go
new file mode 100644
index 00000000..593f6530
--- /dev/null
+++ b/vendor/golang.org/x/crypto/pbkdf2/pbkdf2.go
@@ -0,0 +1,77 @@
+// Copyright 2012 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 pbkdf2 implements the key derivation function PBKDF2 as defined in RFC
+2898 / PKCS #5 v2.0.
+
+A key derivation function is useful when encrypting data based on a password
+or any other not-fully-random data. It uses a pseudorandom function to derive
+a secure encryption key based on the password.
+
+While v2.0 of the standard defines only one pseudorandom function to use,
+HMAC-SHA1, the drafted v2.1 specification allows use of all five FIPS Approved
+Hash Functions SHA-1, SHA-224, SHA-256, SHA-384 and SHA-512 for HMAC. To
+choose, you can pass the `New` functions from the different SHA packages to
+pbkdf2.Key.
+*/
+package pbkdf2 // import "golang.org/x/crypto/pbkdf2"
+
+import (
+	"crypto/hmac"
+	"hash"
+)
+
+// Key derives a key from the password, salt and iteration count, returning a
+// []byte of length keylen that can be used as cryptographic key. The key is
+// derived based on the method described as PBKDF2 with the HMAC variant using
+// the supplied hash function.
+//
+// For example, to use a HMAC-SHA-1 based PBKDF2 key derivation function, you
+// can get a derived key for e.g. AES-256 (which needs a 32-byte key) by
+// doing:
+//
+// 	dk := pbkdf2.Key([]byte("some password"), salt, 4096, 32, sha1.New)
+//
+// Remember to get a good random salt. At least 8 bytes is recommended by the
+// RFC.
+//
+// Using a higher iteration count will increase the cost of an exhaustive
+// search but will also make derivation proportionally slower.
+func Key(password, salt []byte, iter, keyLen int, h func() hash.Hash) []byte {
+	prf := hmac.New(h, password)
+	hashLen := prf.Size()
+	numBlocks := (keyLen + hashLen - 1) / hashLen
+
+	var buf [4]byte
+	dk := make([]byte, 0, numBlocks*hashLen)
+	U := make([]byte, hashLen)
+	for block := 1; block <= numBlocks; block++ {
+		// N.B.: || means concatenation, ^ means XOR
+		// for each block T_i = U_1 ^ U_2 ^ ... ^ U_iter
+		// U_1 = PRF(password, salt || uint(i))
+		prf.Reset()
+		prf.Write(salt)
+		buf[0] = byte(block >> 24)
+		buf[1] = byte(block >> 16)
+		buf[2] = byte(block >> 8)
+		buf[3] = byte(block)
+		prf.Write(buf[:4])
+		dk = prf.Sum(dk)
+		T := dk[len(dk)-hashLen:]
+		copy(U, T)
+
+		// U_n = PRF(password, U_(n-1))
+		for n := 2; n <= iter; n++ {
+			prf.Reset()
+			prf.Write(U)
+			U = U[:0]
+			U = prf.Sum(U)
+			for x := range U {
+				T[x] ^= U[x]
+			}
+		}
+	}
+	return dk[:keyLen]
+}
diff --git a/vendor/golang.org/x/crypto/poly1305/sum_amd64.s b/vendor/golang.org/x/crypto/poly1305/sum_amd64.s
index 2cb03731..1d74f0f8 100644
--- a/vendor/golang.org/x/crypto/poly1305/sum_amd64.s
+++ b/vendor/golang.org/x/crypto/poly1305/sum_amd64.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build gc && !purego
 // +build gc,!purego
 
 #include "textflag.h"
diff --git a/vendor/golang.org/x/crypto/poly1305/sum_ppc64le.s b/vendor/golang.org/x/crypto/poly1305/sum_ppc64le.s
index 5cd7494b..58422aad 100644
--- a/vendor/golang.org/x/crypto/poly1305/sum_ppc64le.s
+++ b/vendor/golang.org/x/crypto/poly1305/sum_ppc64le.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build gc && !purego
 // +build gc,!purego
 
 #include "textflag.h"
@@ -82,7 +83,7 @@ multiply:
 	BGE loop
 
 bytes_between_0_and_15:
-	CMP  $0, R5
+	CMP  R5, $0
 	BEQ  done
 	MOVD $0, R16 // h0
 	MOVD $0, R17 // h1
@@ -122,7 +123,7 @@ just1:
 	// Exactly 8
 	MOVD (R4), R16
 
-	CMP $0, R17
+	CMP R17, $0
 
 	// Check if we've already set R17; if not
 	// set 1 to indicate end of msg.
@@ -151,7 +152,7 @@ less4:
 	ADD   $2, R4
 
 less2:
-	CMP   $0, R5
+	CMP   R5, $0
 	BEQ   insert1
 	MOVBZ (R4), R21
 	SLD   R22, R21, R21
@@ -166,12 +167,12 @@ insert1:
 
 carry:
 	// Add new values to h0, h1, h2
-	ADDC R16, R8
-	ADDE R17, R9
-	ADDE $0, R10
-	MOVD $16, R5
-	ADD  R5, R4
-	BR   multiply
+	ADDC  R16, R8
+	ADDE  R17, R9
+	ADDZE R10, R10
+	MOVD  $16, R5
+	ADD   R5, R4
+	BR    multiply
 
 done:
 	// Save h0, h1, h2 in state
diff --git a/vendor/golang.org/x/crypto/poly1305/sum_s390x.s b/vendor/golang.org/x/crypto/poly1305/sum_s390x.s
index bdd882c6..69c64f84 100644
--- a/vendor/golang.org/x/crypto/poly1305/sum_s390x.s
+++ b/vendor/golang.org/x/crypto/poly1305/sum_s390x.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build gc && !purego
 // +build gc,!purego
 
 #include "textflag.h"
diff --git a/vendor/golang.org/x/crypto/salsa20/salsa/salsa20_amd64.s b/vendor/golang.org/x/crypto/salsa20/salsa/salsa20_amd64.s
index f97efc67..c0892772 100644
--- a/vendor/golang.org/x/crypto/salsa20/salsa/salsa20_amd64.s
+++ b/vendor/golang.org/x/crypto/salsa20/salsa/salsa20_amd64.s
@@ -2,6 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build amd64 && !purego && gc
 // +build amd64,!purego,gc
 
 // This code was translated into a form compatible with 6a from the public
diff --git a/vendor/golang.org/x/crypto/scrypt/scrypt.go b/vendor/golang.org/x/crypto/scrypt/scrypt.go
new file mode 100644
index 00000000..bbe4494c
--- /dev/null
+++ b/vendor/golang.org/x/crypto/scrypt/scrypt.go
@@ -0,0 +1,212 @@
+// Copyright 2012 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 scrypt implements the scrypt key derivation function as defined in
+// Colin Percival's paper "Stronger Key Derivation via Sequential Memory-Hard
+// Functions" (https://www.tarsnap.com/scrypt/scrypt.pdf).
+package scrypt // import "golang.org/x/crypto/scrypt"
+
+import (
+	"crypto/sha256"
+	"encoding/binary"
+	"errors"
+	"math/bits"
+
+	"golang.org/x/crypto/pbkdf2"
+)
+
+const maxInt = int(^uint(0) >> 1)
+
+// blockCopy copies n numbers from src into dst.
+func blockCopy(dst, src []uint32, n int) {
+	copy(dst, src[:n])
+}
+
+// blockXOR XORs numbers from dst with n numbers from src.
+func blockXOR(dst, src []uint32, n int) {
+	for i, v := range src[:n] {
+		dst[i] ^= v
+	}
+}
+
+// salsaXOR applies Salsa20/8 to the XOR of 16 numbers from tmp and in,
+// and puts the result into both tmp and out.
+func salsaXOR(tmp *[16]uint32, in, out []uint32) {
+	w0 := tmp[0] ^ in[0]
+	w1 := tmp[1] ^ in[1]
+	w2 := tmp[2] ^ in[2]
+	w3 := tmp[3] ^ in[3]
+	w4 := tmp[4] ^ in[4]
+	w5 := tmp[5] ^ in[5]
+	w6 := tmp[6] ^ in[6]
+	w7 := tmp[7] ^ in[7]
+	w8 := tmp[8] ^ in[8]
+	w9 := tmp[9] ^ in[9]
+	w10 := tmp[10] ^ in[10]
+	w11 := tmp[11] ^ in[11]
+	w12 := tmp[12] ^ in[12]
+	w13 := tmp[13] ^ in[13]
+	w14 := tmp[14] ^ in[14]
+	w15 := tmp[15] ^ in[15]
+
+	x0, x1, x2, x3, x4, x5, x6, x7, x8 := w0, w1, w2, w3, w4, w5, w6, w7, w8
+	x9, x10, x11, x12, x13, x14, x15 := w9, w10, w11, w12, w13, w14, w15
+
+	for i := 0; i < 8; i += 2 {
+		x4 ^= bits.RotateLeft32(x0+x12, 7)
+		x8 ^= bits.RotateLeft32(x4+x0, 9)
+		x12 ^= bits.RotateLeft32(x8+x4, 13)
+		x0 ^= bits.RotateLeft32(x12+x8, 18)
+
+		x9 ^= bits.RotateLeft32(x5+x1, 7)
+		x13 ^= bits.RotateLeft32(x9+x5, 9)
+		x1 ^= bits.RotateLeft32(x13+x9, 13)
+		x5 ^= bits.RotateLeft32(x1+x13, 18)
+
+		x14 ^= bits.RotateLeft32(x10+x6, 7)
+		x2 ^= bits.RotateLeft32(x14+x10, 9)
+		x6 ^= bits.RotateLeft32(x2+x14, 13)
+		x10 ^= bits.RotateLeft32(x6+x2, 18)
+
+		x3 ^= bits.RotateLeft32(x15+x11, 7)
+		x7 ^= bits.RotateLeft32(x3+x15, 9)
+		x11 ^= bits.RotateLeft32(x7+x3, 13)
+		x15 ^= bits.RotateLeft32(x11+x7, 18)
+
+		x1 ^= bits.RotateLeft32(x0+x3, 7)
+		x2 ^= bits.RotateLeft32(x1+x0, 9)
+		x3 ^= bits.RotateLeft32(x2+x1, 13)
+		x0 ^= bits.RotateLeft32(x3+x2, 18)
+
+		x6 ^= bits.RotateLeft32(x5+x4, 7)
+		x7 ^= bits.RotateLeft32(x6+x5, 9)
+		x4 ^= bits.RotateLeft32(x7+x6, 13)
+		x5 ^= bits.RotateLeft32(x4+x7, 18)
+
+		x11 ^= bits.RotateLeft32(x10+x9, 7)
+		x8 ^= bits.RotateLeft32(x11+x10, 9)
+		x9 ^= bits.RotateLeft32(x8+x11, 13)
+		x10 ^= bits.RotateLeft32(x9+x8, 18)
+
+		x12 ^= bits.RotateLeft32(x15+x14, 7)
+		x13 ^= bits.RotateLeft32(x12+x15, 9)
+		x14 ^= bits.RotateLeft32(x13+x12, 13)
+		x15 ^= bits.RotateLeft32(x14+x13, 18)
+	}
+	x0 += w0
+	x1 += w1
+	x2 += w2
+	x3 += w3
+	x4 += w4
+	x5 += w5
+	x6 += w6
+	x7 += w7
+	x8 += w8
+	x9 += w9
+	x10 += w10
+	x11 += w11
+	x12 += w12
+	x13 += w13
+	x14 += w14
+	x15 += w15
+
+	out[0], tmp[0] = x0, x0
+	out[1], tmp[1] = x1, x1
+	out[2], tmp[2] = x2, x2
+	out[3], tmp[3] = x3, x3
+	out[4], tmp[4] = x4, x4
+	out[5], tmp[5] = x5, x5
+	out[6], tmp[6] = x6, x6
+	out[7], tmp[7] = x7, x7
+	out[8], tmp[8] = x8, x8
+	out[9], tmp[9] = x9, x9
+	out[10], tmp[10] = x10, x10
+	out[11], tmp[11] = x11, x11
+	out[12], tmp[12] = x12, x12
+	out[13], tmp[13] = x13, x13
+	out[14], tmp[14] = x14, x14
+	out[15], tmp[15] = x15, x15
+}
+
+func blockMix(tmp *[16]uint32, in, out []uint32, r int) {
+	blockCopy(tmp[:], in[(2*r-1)*16:], 16)
+	for i := 0; i < 2*r; i += 2 {
+		salsaXOR(tmp, in[i*16:], out[i*8:])
+		salsaXOR(tmp, in[i*16+16:], out[i*8+r*16:])
+	}
+}
+
+func integer(b []uint32, r int) uint64 {
+	j := (2*r - 1) * 16
+	return uint64(b[j]) | uint64(b[j+1])<<32
+}
+
+func smix(b []byte, r, N int, v, xy []uint32) {
+	var tmp [16]uint32
+	R := 32 * r
+	x := xy
+	y := xy[R:]
+
+	j := 0
+	for i := 0; i < R; i++ {
+		x[i] = binary.LittleEndian.Uint32(b[j:])
+		j += 4
+	}
+	for i := 0; i < N; i += 2 {
+		blockCopy(v[i*R:], x, R)
+		blockMix(&tmp, x, y, r)
+
+		blockCopy(v[(i+1)*R:], y, R)
+		blockMix(&tmp, y, x, r)
+	}
+	for i := 0; i < N; i += 2 {
+		j := int(integer(x, r) & uint64(N-1))
+		blockXOR(x, v[j*R:], R)
+		blockMix(&tmp, x, y, r)
+
+		j = int(integer(y, r) & uint64(N-1))
+		blockXOR(y, v[j*R:], R)
+		blockMix(&tmp, y, x, r)
+	}
+	j = 0
+	for _, v := range x[:R] {
+		binary.LittleEndian.PutUint32(b[j:], v)
+		j += 4
+	}
+}
+
+// Key derives a key from the password, salt, and cost parameters, returning
+// a byte slice of length keyLen that can be used as cryptographic key.
+//
+// N is a CPU/memory cost parameter, which must be a power of two greater than 1.
+// r and p must satisfy r * p < 2³⁰. If the parameters do not satisfy the
+// limits, the function returns a nil byte slice and an error.
+//
+// For example, you can get a derived key for e.g. AES-256 (which needs a
+// 32-byte key) by doing:
+//
+//      dk, err := scrypt.Key([]byte("some password"), salt, 32768, 8, 1, 32)
+//
+// The recommended parameters for interactive logins as of 2017 are N=32768, r=8
+// and p=1. The parameters N, r, and p should be increased as memory latency and
+// CPU parallelism increases; consider setting N to the highest power of 2 you
+// can derive within 100 milliseconds. Remember to get a good random salt.
+func Key(password, salt []byte, N, r, p, keyLen int) ([]byte, error) {
+	if N <= 1 || N&(N-1) != 0 {
+		return nil, errors.New("scrypt: N must be > 1 and a power of 2")
+	}
+	if uint64(r)*uint64(p) >= 1<<30 || r > maxInt/128/p || r > maxInt/256 || N > maxInt/128/r {
+		return nil, errors.New("scrypt: parameters are too large")
+	}
+
+	xy := make([]uint32, 64*r)
+	v := make([]uint32, 32*N*r)
+	b := pbkdf2.Key(password, salt, 1, p*128*r, sha256.New)
+
+	for i := 0; i < p; i++ {
+		smix(b[i*128*r:], r, N, v, xy)
+	}
+
+	return pbkdf2.Key(password, b, 1, keyLen, sha256.New), nil
+}
diff --git a/vendor/golang.org/x/crypto/ssh/client.go b/vendor/golang.org/x/crypto/ssh/client.go
index 7b00bff1..99f68bd3 100644
--- a/vendor/golang.org/x/crypto/ssh/client.go
+++ b/vendor/golang.org/x/crypto/ssh/client.go
@@ -77,7 +77,7 @@ func NewClientConn(c net.Conn, addr string, config *ClientConfig) (Conn, <-chan
 	}
 
 	conn := &connection{
-		sshConn: sshConn{conn: c},
+		sshConn: sshConn{conn: c, user: fullConf.User},
 	}
 
 	if err := conn.clientHandshake(addr, &fullConf); err != nil {