Update vendor of containers/buildah

Changes since 2022-05-31:
  - add --omit-history option (buildah PR 4028)

Signed-off-by: Ed Santiago <santiago@redhat.com>
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>

29 files changed, 917 insertions, 851 deletions

diff --git a/vendor/github.com/klauspost/compress/README.md b/vendor/github.com/klauspost/compress/README.md
index c3ec9d8a7..5c3c2a258 100644
--- a/vendor/github.com/klauspost/compress/README.md
+++ b/vendor/github.com/klauspost/compress/README.md
@@ -17,6 +17,24 @@ This package provides various compression algorithms.
 

 # changelog

 

+* May 25, 2022 (v1.15.5)

+	* s2: Add concurrent stream decompression https://github.com/klauspost/compress/pull/602

+	* s2: Fix final emit oob read crash on amd64 https://github.com/klauspost/compress/pull/601

+	* huff0: asm implementation of Decompress1X by @WojciechMula https://github.com/klauspost/compress/pull/596

+	* zstd: Use 1 less goroutine for stream decoding https://github.com/klauspost/compress/pull/588

+	* zstd: Copy literal in 16 byte blocks when possible https://github.com/klauspost/compress/pull/592

+	* zstd: Speed up when WithDecoderLowmem(false) https://github.com/klauspost/compress/pull/599

+	* zstd: faster next state update in BMI2 version of decode by @WojciechMula in https://github.com/klauspost/compress/pull/593

+	* huff0: Do not check max size when reading table. https://github.com/klauspost/compress/pull/586

+	* flate: Inplace hashing for level 7-9 by @klauspost in https://github.com/klauspost/compress/pull/590

+

+

+* May 11, 2022 (v1.15.4)

+	* huff0: decompress directly into output by @WojciechMula in [#577](https://github.com/klauspost/compress/pull/577)

+	* inflate: Keep dict on stack [#581](https://github.com/klauspost/compress/pull/581)

+	* zstd: Faster decoding memcopy in asm [#583](https://github.com/klauspost/compress/pull/583)

+	* zstd: Fix ignored crc [#580](https://github.com/klauspost/compress/pull/580)

+

 * May 5, 2022 (v1.15.3)

 	* zstd: Allow to ignore checksum checking by @WojciechMula [#572](https://github.com/klauspost/compress/pull/572)

 	* s2: Fix incorrect seek for io.SeekEnd in [#575](https://github.com/klauspost/compress/pull/575)

@@ -77,6 +95,9 @@ While the release has been extensively tested, it is recommended to testing when
 	* zstd: add arm64 xxhash assembly in [#464](https://github.com/klauspost/compress/pull/464)

 	* Add garbled for binaries for s2 in [#445](https://github.com/klauspost/compress/pull/445)

 

+<details>

+	<summary>See changes to v1.13.x</summary>

+	

 * Aug 30, 2021 (v1.13.5)

 	* gz/zlib/flate: Alias stdlib errors [#425](https://github.com/klauspost/compress/pull/425)

 	* s2: Add block support to commandline tools [#413](https://github.com/klauspost/compress/pull/413)

@@ -105,6 +126,8 @@ While the release has been extensively tested, it is recommended to testing when
 	* Added [gzhttp](https://github.com/klauspost/compress/tree/master/gzhttp#gzip-handler) which allows wrapping HTTP servers and clients with GZIP compressors.

 	* zstd: Detect short invalid signatures [#382](https://github.com/klauspost/compress/pull/382)

 	* zstd: Spawn decoder goroutine only if needed. [#380](https://github.com/klauspost/compress/pull/380)

+</details>

+

 

 <details>

 	<summary>See changes to v1.12.x</summary>

diff --git a/vendor/github.com/klauspost/compress/flate/deflate.go b/vendor/github.com/klauspost/compress/flate/deflate.go
index bffa2f332..f8435998e 100644
--- a/vendor/github.com/klauspost/compress/flate/deflate.go
+++ b/vendor/github.com/klauspost/compress/flate/deflate.go
@@ -84,24 +84,23 @@ type advancedState struct {
 	length         int
 	offset         int
 	maxInsertIndex int
+	chainHead      int
+	hashOffset     int
 
-	// Input hash chains
-	// hashHead[hashValue] contains the largest inputIndex with the specified hash value
-	// If hashHead[hashValue] is within the current window, then
-	// hashPrev[hashHead[hashValue] & windowMask] contains the previous index
-	// with the same hash value.
-	chainHead  int
-	hashHead   [hashSize]uint32
-	hashPrev   [windowSize]uint32
-	hashOffset int
+	ii uint16 // position of last match, intended to overflow to reset.
 
 	// input window: unprocessed data is window[index:windowEnd]
 	index          int
 	estBitsPerByte int
 	hashMatch      [maxMatchLength + minMatchLength]uint32
 
-	hash uint32
-	ii   uint16 // position of last match, intended to overflow to reset.
+	// Input hash chains
+	// hashHead[hashValue] contains the largest inputIndex with the specified hash value
+	// If hashHead[hashValue] is within the current window, then
+	// hashPrev[hashHead[hashValue] & windowMask] contains the previous index
+	// with the same hash value.
+	hashHead [hashSize]uint32
+	hashPrev [windowSize]uint32
 }
 
 type compressor struct {
@@ -259,7 +258,6 @@ func (d *compressor) fillWindow(b []byte) {
 			// Set the head of the hash chain to us.
 			s.hashHead[newH] = uint32(di + s.hashOffset)
 		}
-		s.hash = newH
 	}
 	// Update window information.
 	d.windowEnd += n
@@ -403,7 +401,6 @@ func (d *compressor) initDeflate() {
 	s.hashOffset = 1
 	s.length = minMatchLength - 1
 	s.offset = 0
-	s.hash = 0
 	s.chainHead = -1
 }
 
@@ -432,9 +429,6 @@ func (d *compressor) deflateLazy() {
 	}
 
 	s.maxInsertIndex = d.windowEnd - (minMatchLength - 1)
-	if s.index < s.maxInsertIndex {
-		s.hash = hash4(d.window[s.index:])
-	}
 
 	for {
 		if sanity && s.index > d.windowEnd {
@@ -466,11 +460,11 @@ func (d *compressor) deflateLazy() {
 		}
 		if s.index < s.maxInsertIndex {
 			// Update the hash
-			s.hash = hash4(d.window[s.index:])
-			ch := s.hashHead[s.hash&hashMask]
+			hash := hash4(d.window[s.index:])
+			ch := s.hashHead[hash]
 			s.chainHead = int(ch)
 			s.hashPrev[s.index&windowMask] = ch
-			s.hashHead[s.hash&hashMask] = uint32(s.index + s.hashOffset)
+			s.hashHead[hash] = uint32(s.index + s.hashOffset)
 		}
 		prevLength := s.length
 		prevOffset := s.offset
@@ -503,7 +497,7 @@ func (d *compressor) deflateLazy() {
 				end += prevIndex
 				idx := prevIndex + prevLength - (4 - checkOff)
 				h := hash4(d.window[idx:])
-				ch2 := int(s.hashHead[h&hashMask]) - s.hashOffset - prevLength + (4 - checkOff)
+				ch2 := int(s.hashHead[h]) - s.hashOffset - prevLength + (4 - checkOff)
 				if ch2 > minIndex {
 					length := matchLen(d.window[prevIndex:end], d.window[ch2:])
 					// It seems like a pure length metric is best.
@@ -547,7 +541,6 @@ func (d *compressor) deflateLazy() {
 					// Set the head of the hash chain to us.
 					s.hashHead[newH] = uint32(di + s.hashOffset)
 				}
-				s.hash = newH
 			}
 
 			s.index = newIndex
@@ -793,7 +786,6 @@ func (d *compressor) reset(w io.Writer) {
 		d.tokens.Reset()
 		s.length = minMatchLength - 1
 		s.offset = 0
-		s.hash = 0
 		s.ii = 0
 		s.maxInsertIndex = 0
 	}
diff --git a/vendor/github.com/klauspost/compress/flate/fast_encoder.go b/vendor/github.com/klauspost/compress/flate/fast_encoder.go
index d55ea2a77..f781aaa62 100644
--- a/vendor/github.com/klauspost/compress/flate/fast_encoder.go
+++ b/vendor/github.com/klauspost/compress/flate/fast_encoder.go
@@ -117,7 +117,7 @@ func (e *fastGen) addBlock(src []byte) int32 {
 // hash4 returns the hash of u to fit in a hash table with h bits.
 // Preferably h should be a constant and should always be <32.
 func hash4u(u uint32, h uint8) uint32 {
-	return (u * prime4bytes) >> ((32 - h) & reg8SizeMask32)
+	return (u * prime4bytes) >> (32 - h)
 }
 
 type tableEntryPrev struct {
diff --git a/vendor/github.com/klauspost/compress/huff0/bitreader.go b/vendor/github.com/klauspost/compress/huff0/bitreader.go
index 451160edd..504a7be9d 100644
--- a/vendor/github.com/klauspost/compress/huff0/bitreader.go
+++ b/vendor/github.com/klauspost/compress/huff0/bitreader.go
@@ -165,11 +165,6 @@ func (b *bitReaderShifted) peekBitsFast(n uint8) uint16 {
 	return uint16(b.value >> ((64 - n) & 63))
 }
 
-// peekTopBits(n) is equvialent to peekBitFast(64 - n)
-func (b *bitReaderShifted) peekTopBits(n uint8) uint16 {
-	return uint16(b.value >> n)
-}
-
 func (b *bitReaderShifted) advance(n uint8) {
 	b.bitsRead += n
 	b.value <<= n & 63
@@ -220,11 +215,6 @@ func (b *bitReaderShifted) fill() {
 	}
 }
 
-// finished returns true if all bits have been read from the bit stream.
-func (b *bitReaderShifted) finished() bool {
-	return b.off == 0 && b.bitsRead >= 64
-}
-
 func (b *bitReaderShifted) remaining() uint {
 	return b.off*8 + uint(64-b.bitsRead)
 }
diff --git a/vendor/github.com/klauspost/compress/huff0/bitwriter.go b/vendor/github.com/klauspost/compress/huff0/bitwriter.go
index 6bce4e87d..ec71f7a34 100644
--- a/vendor/github.com/klauspost/compress/huff0/bitwriter.go
+++ b/vendor/github.com/klauspost/compress/huff0/bitwriter.go
@@ -5,8 +5,6 @@
 
 package huff0
 
-import "fmt"
-
 // bitWriter will write bits.
 // First bit will be LSB of the first byte of output.
 type bitWriter struct {
@@ -23,14 +21,6 @@ var bitMask16 = [32]uint16{
 	0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF,
 	0xFFFF, 0xFFFF} /* up to 16 bits */
 
-// addBits16NC will add up to 16 bits.
-// It will not check if there is space for them,
-// so the caller must ensure that it has flushed recently.
-func (b *bitWriter) addBits16NC(value uint16, bits uint8) {
-	b.bitContainer |= uint64(value&bitMask16[bits&31]) << (b.nBits & 63)
-	b.nBits += bits
-}
-
 // addBits16Clean will add up to 16 bits. value may not contain more set bits than indicated.
 // It will not check if there is space for them, so the caller must ensure that it has flushed recently.
 func (b *bitWriter) addBits16Clean(value uint16, bits uint8) {
@@ -70,104 +60,6 @@ func (b *bitWriter) encTwoSymbols(ct cTable, av, bv byte) {
 	b.nBits += encA.nBits + encB.nBits
 }
 
-// addBits16ZeroNC will add up to 16 bits.
-// It will not check if there is space for them,
-// so the caller must ensure that it has flushed recently.
-// This is fastest if bits can be zero.
-func (b *bitWriter) addBits16ZeroNC(value uint16, bits uint8) {
-	if bits == 0 {
-		return
-	}
-	value <<= (16 - bits) & 15
-	value >>= (16 - bits) & 15
-	b.bitContainer |= uint64(value) << (b.nBits & 63)
-	b.nBits += bits
-}
-
-// flush will flush all pending full bytes.
-// There will be at least 56 bits available for writing when this has been called.
-// Using flush32 is faster, but leaves less space for writing.
-func (b *bitWriter) flush() {
-	v := b.nBits >> 3
-	switch v {
-	case 0:
-		return
-	case 1:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-		)
-		b.bitContainer >>= 1 << 3
-	case 2:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-		)
-		b.bitContainer >>= 2 << 3
-	case 3:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-		)
-		b.bitContainer >>= 3 << 3
-	case 4:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-		)
-		b.bitContainer >>= 4 << 3
-	case 5:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-		)
-		b.bitContainer >>= 5 << 3
-	case 6:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-		)
-		b.bitContainer >>= 6 << 3
-	case 7:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-			byte(b.bitContainer>>48),
-		)
-		b.bitContainer >>= 7 << 3
-	case 8:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-			byte(b.bitContainer>>48),
-			byte(b.bitContainer>>56),
-		)
-		b.bitContainer = 0
-		b.nBits = 0
-		return
-	default:
-		panic(fmt.Errorf("bits (%d) > 64", b.nBits))
-	}
-	b.nBits &= 7
-}
-
 // flush32 will flush out, so there are at least 32 bits available for writing.
 func (b *bitWriter) flush32() {
 	if b.nBits < 32 {
@@ -201,10 +93,3 @@ func (b *bitWriter) close() error {
 	b.flushAlign()
 	return nil
 }
-
-// reset and continue writing by appending to out.
-func (b *bitWriter) reset(out []byte) {
-	b.bitContainer = 0
-	b.nBits = 0
-	b.out = out
-}
diff --git a/vendor/github.com/klauspost/compress/huff0/bytereader.go b/vendor/github.com/klauspost/compress/huff0/bytereader.go
index 50bcdf6ea..4dcab8d23 100644
--- a/vendor/github.com/klauspost/compress/huff0/bytereader.go
+++ b/vendor/github.com/klauspost/compress/huff0/bytereader.go
@@ -20,11 +20,6 @@ func (b *byteReader) init(in []byte) {
 	b.off = 0
 }
 
-// advance the stream b n bytes.
-func (b *byteReader) advance(n uint) {
-	b.off += int(n)
-}
-
 // Int32 returns a little endian int32 starting at current offset.
 func (b byteReader) Int32() int32 {
 	v3 := int32(b.b[b.off+3])
@@ -43,11 +38,6 @@ func (b byteReader) Uint32() uint32 {
 	return (v3 << 24) | (v2 << 16) | (v1 << 8) | v0
 }
 
-// unread returns the unread portion of the input.
-func (b byteReader) unread() []byte {
-	return b.b[b.off:]
-}
-
 // remain will return the number of bytes remaining.
 func (b byteReader) remain() int {
 	return len(b.b) - b.off
diff --git a/vendor/github.com/klauspost/compress/huff0/compress.go b/vendor/github.com/klauspost/compress/huff0/compress.go
index bc95ac623..4d14542fa 100644
--- a/vendor/github.com/klauspost/compress/huff0/compress.go
+++ b/vendor/github.com/klauspost/compress/huff0/compress.go
@@ -404,6 +404,7 @@ func (s *Scratch) canUseTable(c cTable) bool {
 	return true
 }
 
+//lint:ignore U1000 used for debugging
 func (s *Scratch) validateTable(c cTable) bool {
 	if len(c) < int(s.symbolLen) {
 		return false
diff --git a/vendor/github.com/klauspost/compress/huff0/decompress.go b/vendor/github.com/klauspost/compress/huff0/decompress.go
index 04f652995..c0c48bd70 100644
--- a/vendor/github.com/klauspost/compress/huff0/decompress.go
+++ b/vendor/github.com/klauspost/compress/huff0/decompress.go
@@ -11,7 +11,6 @@ import (
 
 type dTable struct {
 	single []dEntrySingle
-	double []dEntryDouble
 }
 
 // single-symbols decoding
@@ -19,13 +18,6 @@ type dEntrySingle struct {
 	entry uint16
 }
 
-// double-symbols decoding
-type dEntryDouble struct {
-	seq   [4]byte
-	nBits uint8
-	len   uint8
-}
-
 // Uses special code for all tables that are < 8 bits.
 const use8BitTables = true
 
@@ -35,7 +27,7 @@ const use8BitTables = true
 // If no Scratch is provided a new one is allocated.
 // The returned Scratch can be used for encoding or decoding input using this table.
 func ReadTable(in []byte, s *Scratch) (s2 *Scratch, remain []byte, err error) {
-	s, err = s.prepare(in)
+	s, err = s.prepare(nil)
 	if err != nil {
 		return s, nil, err
 	}
@@ -236,108 +228,6 @@ func (d *Decoder) buffer() *[4][256]byte {
 	return &[4][256]byte{}
 }
 
-// Decompress1X will decompress a 1X encoded stream.
-// The cap of the output buffer will be the maximum decompressed size.
-// The length of the supplied input must match the end of a block exactly.
-func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
-	if len(d.dt.single) == 0 {
-		return nil, errors.New("no table loaded")
-	}
-	if use8BitTables && d.actualTableLog <= 8 {
-		return d.decompress1X8Bit(dst, src)
-	}
-	var br bitReaderShifted
-	err := br.init(src)
-	if err != nil {
-		return dst, err
-	}
-	maxDecodedSize := cap(dst)
-	dst = dst[:0]
-
-	// Avoid bounds check by always having full sized table.
-	const tlSize = 1 << tableLogMax
-	const tlMask = tlSize - 1
-	dt := d.dt.single[:tlSize]
-
-	// Use temp table to avoid bound checks/append penalty.
-	bufs := d.buffer()
-	buf := &bufs[0]
-	var off uint8
-
-	for br.off >= 8 {
-		br.fillFast()
-		v := dt[br.peekBitsFast(d.actualTableLog)&tlMask]
-		br.advance(uint8(v.entry))
-		buf[off+0] = uint8(v.entry >> 8)
-
-		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
-		br.advance(uint8(v.entry))
-		buf[off+1] = uint8(v.entry >> 8)
-
-		// Refill
-		br.fillFast()
-
-		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
-		br.advance(uint8(v.entry))
-		buf[off+2] = uint8(v.entry >> 8)
-
-		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
-		br.advance(uint8(v.entry))
-		buf[off+3] = uint8(v.entry >> 8)
-
-		off += 4
-		if off == 0 {
-			if len(dst)+256 > maxDecodedSize {
-				br.close()
-				d.bufs.Put(bufs)
-				return nil, ErrMaxDecodedSizeExceeded
-			}
-			dst = append(dst, buf[:]...)
-		}
-	}
-
-	if len(dst)+int(off) > maxDecodedSize {
-		d.bufs.Put(bufs)
-		br.close()
-		return nil, ErrMaxDecodedSizeExceeded
-	}
-	dst = append(dst, buf[:off]...)
-
-	// br < 8, so uint8 is fine
-	bitsLeft := uint8(br.off)*8 + 64 - br.bitsRead
-	for bitsLeft > 0 {
-		br.fill()
-		if false && br.bitsRead >= 32 {
-			if br.off >= 4 {
-				v := br.in[br.off-4:]
-				v = v[:4]
-				low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
-				br.value = (br.value << 32) | uint64(low)
-				br.bitsRead -= 32
-				br.off -= 4
-			} else {
-				for br.off > 0 {
-					br.value = (br.value << 8) | uint64(br.in[br.off-1])
-					br.bitsRead -= 8
-					br.off--
-				}
-			}
-		}
-		if len(dst) >= maxDecodedSize {
-			d.bufs.Put(bufs)
-			br.close()
-			return nil, ErrMaxDecodedSizeExceeded
-		}
-		v := d.dt.single[br.peekBitsFast(d.actualTableLog)&tlMask]
-		nBits := uint8(v.entry)
-		br.advance(nBits)
-		bitsLeft -= nBits
-		dst = append(dst, uint8(v.entry>>8))
-	}
-	d.bufs.Put(bufs)
-	return dst, br.close()
-}
-
 // decompress1X8Bit will decompress a 1X encoded stream with tablelog <= 8.
 // The cap of the output buffer will be the maximum decompressed size.
 // The length of the supplied input must match the end of a block exactly.
@@ -995,7 +885,6 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 
 	const shift = 56
 	const tlSize = 1 << 8
-	const tlMask = tlSize - 1
 	single := d.dt.single[:tlSize]
 
 	// Use temp table to avoid bound checks/append penalty.
diff --git a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go
index 3415e5da2..671e630a8 100644
--- a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go
+++ b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go
@@ -2,12 +2,14 @@
 // +build amd64,!appengine,!noasm,gc
 
 // This file contains the specialisation of Decoder.Decompress4X
-// that uses an asm implementation of its main loop.
+// and Decoder.Decompress1X that use an asm implementation of thir main loops.
 package huff0
 
 import (
 	"errors"
 	"fmt"
+
+	"github.com/klauspost/compress/internal/cpuinfo"
 )
 
 // decompress4x_main_loop_x86 is an x86 assembler implementation
@@ -146,3 +148,81 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 	}
 	return dst, nil
 }
+
+// decompress4x_main_loop_x86 is an x86 assembler implementation
+// of Decompress1X when tablelog > 8.
+//go:noescape
+func decompress1x_main_loop_amd64(ctx *decompress1xContext)
+
+// decompress4x_main_loop_x86 is an x86 with BMI2 assembler implementation
+// of Decompress1X when tablelog > 8.
+//go:noescape
+func decompress1x_main_loop_bmi2(ctx *decompress1xContext)
+
+type decompress1xContext struct {
+	pbr      *bitReaderShifted
+	peekBits uint8
+	out      *byte
+	outCap   int
+	tbl      *dEntrySingle
+	decoded  int
+}
+
+// Error reported by asm implementations
+const error_max_decoded_size_exeeded = -1
+
+// Decompress1X will decompress a 1X encoded stream.
+// The cap of the output buffer will be the maximum decompressed size.
+// The length of the supplied input must match the end of a block exactly.
+func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
+	if len(d.dt.single) == 0 {
+		return nil, errors.New("no table loaded")
+	}
+	var br bitReaderShifted
+	err := br.init(src)
+	if err != nil {
+		return dst, err
+	}
+	maxDecodedSize := cap(dst)
+	dst = dst[:maxDecodedSize]
+
+	const tlSize = 1 << tableLogMax
+	const tlMask = tlSize - 1
+
+	if maxDecodedSize >= 4 {
+		ctx := decompress1xContext{
+			pbr:      &br,
+			out:      &dst[0],
+			outCap:   maxDecodedSize,
+			peekBits: uint8((64 - d.actualTableLog) & 63), // see: bitReaderShifted.peekBitsFast()
+			tbl:      &d.dt.single[0],
+		}
+
+		if cpuinfo.HasBMI2() {
+			decompress1x_main_loop_bmi2(&ctx)
+		} else {
+			decompress1x_main_loop_amd64(&ctx)
+		}
+		if ctx.decoded == error_max_decoded_size_exeeded {
+			return nil, ErrMaxDecodedSizeExceeded
+		}
+
+		dst = dst[:ctx.decoded]
+	}
+
+	// br < 8, so uint8 is fine
+	bitsLeft := uint8(br.off)*8 + 64 - br.bitsRead
+	for bitsLeft > 0 {
+		br.fill()
+		if len(dst) >= maxDecodedSize {
+			br.close()
+			return nil, ErrMaxDecodedSizeExceeded
+		}
+		v := d.dt.single[br.peekBitsFast(d.actualTableLog)&tlMask]
+		nBits := uint8(v.entry)
+		br.advance(nBits)
+		bitsLeft -= nBits
+		dst = append(dst, uint8(v.entry>>8))
+	}
+	return dst, br.close()
+}
diff --git a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.s b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.s
index 06287f568..6c65c6e2b 100644
--- a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.s
+++ b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.s
@@ -660,3 +660,206 @@ skip_fill1003:
 	SHLQ  $0x02, DX
 	MOVQ  DX, 64(AX)
 	RET
+
+// func decompress1x_main_loop_amd64(ctx *decompress1xContext)
+TEXT ·decompress1x_main_loop_amd64(SB), $0-8
+	MOVQ    ctx+0(FP), CX
+	MOVQ    16(CX), DX
+	MOVQ    24(CX), BX
+	CMPQ    BX, $0x04
+	JB      error_max_decoded_size_exeeded
+	LEAQ    (DX)(BX*1), BX
+	MOVQ    (CX), SI
+	MOVQ    (SI), R8
+	MOVQ    24(SI), R9
+	MOVQ    32(SI), R10
+	MOVBQZX 40(SI), R11
+	MOVQ    32(CX), SI
+	MOVBQZX 8(CX), DI
+	JMP     loop_condition
+
+main_loop:
+	// Check if we have room for 4 bytes in the output buffer
+	LEAQ 4(DX), CX
+	CMPQ CX, BX
+	JGE  error_max_decoded_size_exeeded
+
+	// Decode 4 values
+	CMPQ R11, $0x20
+	JL   bitReader_fillFast_1_end
+	SUBQ $0x20, R11
+	SUBQ $0x04, R9
+	MOVL (R8)(R9*1), R12
+	MOVQ R11, CX
+	SHLQ CL, R12
+	ORQ  R12, R10
+
+bitReader_fillFast_1_end:
+	MOVQ    DI, CX
+	MOVQ    R10, R12
+	SHRQ    CL, R12
+	MOVW    (SI)(R12*2), CX
+	MOVB    CH, AL
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLQ    CL, R10
+	MOVQ    DI, CX
+	MOVQ    R10, R12
+	SHRQ    CL, R12
+	MOVW    (SI)(R12*2), CX
+	MOVB    CH, AH
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLQ    CL, R10
+	BSWAPL  AX
+	CMPQ    R11, $0x20
+	JL      bitReader_fillFast_2_end
+	SUBQ    $0x20, R11
+	SUBQ    $0x04, R9
+	MOVL    (R8)(R9*1), R12
+	MOVQ    R11, CX
+	SHLQ    CL, R12
+	ORQ     R12, R10
+
+bitReader_fillFast_2_end:
+	MOVQ    DI, CX
+	MOVQ    R10, R12
+	SHRQ    CL, R12
+	MOVW    (SI)(R12*2), CX
+	MOVB    CH, AH
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLQ    CL, R10
+	MOVQ    DI, CX
+	MOVQ    R10, R12
+	SHRQ    CL, R12
+	MOVW    (SI)(R12*2), CX
+	MOVB    CH, AL
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLQ    CL, R10
+	BSWAPL  AX
+
+	// Store the decoded values
+	MOVL AX, (DX)
+	ADDQ $0x04, DX
+
+loop_condition:
+	CMPQ R9, $0x08
+	JGE  main_loop
+
+	// Update ctx structure
+	MOVQ ctx+0(FP), AX
+	MOVQ DX, CX
+	MOVQ 16(AX), DX
+	SUBQ DX, CX
+	MOVQ CX, 40(AX)
+	MOVQ (AX), AX
+	MOVQ R9, 24(AX)
+	MOVQ R10, 32(AX)
+	MOVB R11, 40(AX)
+	RET
+
+	// Report error
+error_max_decoded_size_exeeded:
+	MOVQ ctx+0(FP), AX
+	MOVQ $-1, CX
+	MOVQ CX, 40(AX)
+	RET
+
+// func decompress1x_main_loop_bmi2(ctx *decompress1xContext)
+// Requires: BMI2
+TEXT ·decompress1x_main_loop_bmi2(SB), $0-8
+	MOVQ    ctx+0(FP), CX
+	MOVQ    16(CX), DX
+	MOVQ    24(CX), BX
+	CMPQ    BX, $0x04
+	JB      error_max_decoded_size_exeeded
+	LEAQ    (DX)(BX*1), BX
+	MOVQ    (CX), SI
+	MOVQ    (SI), R8
+	MOVQ    24(SI), R9
+	MOVQ    32(SI), R10
+	MOVBQZX 40(SI), R11
+	MOVQ    32(CX), SI
+	MOVBQZX 8(CX), DI
+	JMP     loop_condition
+
+main_loop:
+	// Check if we have room for 4 bytes in the output buffer
+	LEAQ 4(DX), CX
+	CMPQ CX, BX
+	JGE  error_max_decoded_size_exeeded
+
+	// Decode 4 values
+	CMPQ  R11, $0x20
+	JL    bitReader_fillFast_1_end
+	SUBQ  $0x20, R11
+	SUBQ  $0x04, R9
+	MOVL  (R8)(R9*1), CX
+	SHLXQ R11, CX, CX
+	ORQ   CX, R10
+
+bitReader_fillFast_1_end:
+	SHRXQ   DI, R10, CX
+	MOVW    (SI)(CX*2), CX
+	MOVB    CH, AL
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLXQ   CX, R10, R10
+	SHRXQ   DI, R10, CX
+	MOVW    (SI)(CX*2), CX
+	MOVB    CH, AH
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLXQ   CX, R10, R10
+	BSWAPL  AX
+	CMPQ    R11, $0x20
+	JL      bitReader_fillFast_2_end
+	SUBQ    $0x20, R11
+	SUBQ    $0x04, R9
+	MOVL    (R8)(R9*1), CX
+	SHLXQ   R11, CX, CX
+	ORQ     CX, R10
+
+bitReader_fillFast_2_end:
+	SHRXQ   DI, R10, CX
+	MOVW    (SI)(CX*2), CX
+	MOVB    CH, AH
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLXQ   CX, R10, R10
+	SHRXQ   DI, R10, CX
+	MOVW    (SI)(CX*2), CX
+	MOVB    CH, AL
+	MOVBQZX CL, CX
+	ADDQ    CX, R11
+	SHLXQ   CX, R10, R10
+	BSWAPL  AX
+
+	// Store the decoded values
+	MOVL AX, (DX)
+	ADDQ $0x04, DX
+
+loop_condition:
+	CMPQ R9, $0x08
+	JGE  main_loop
+
+	// Update ctx structure
+	MOVQ ctx+0(FP), AX
+	MOVQ DX, CX
+	MOVQ 16(AX), DX
+	SUBQ DX, CX
+	MOVQ CX, 40(AX)
+	MOVQ (AX), AX
+	MOVQ R9, 24(AX)
+	MOVQ R10, 32(AX)
+	MOVB R11, 40(AX)
+	RET
+
+	// Report error
+error_max_decoded_size_exeeded:
+	MOVQ ctx+0(FP), AX
+	MOVQ $-1, CX
+	MOVQ CX, 40(AX)
+	RET
diff --git a/vendor/github.com/klauspost/compress/huff0/decompress_generic.go b/vendor/github.com/klauspost/compress/huff0/decompress_generic.go
index 126b4d68a..4f6f37cb2 100644
--- a/vendor/github.com/klauspost/compress/huff0/decompress_generic.go
+++ b/vendor/github.com/klauspost/compress/huff0/decompress_generic.go
@@ -191,3 +191,105 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 	}
 	return dst, nil
 }
+
+// Decompress1X will decompress a 1X encoded stream.
+// The cap of the output buffer will be the maximum decompressed size.
+// The length of the supplied input must match the end of a block exactly.
+func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
+	if len(d.dt.single) == 0 {
+		return nil, errors.New("no table loaded")
+	}
+	if use8BitTables && d.actualTableLog <= 8 {
+		return d.decompress1X8Bit(dst, src)
+	}
+	var br bitReaderShifted
+	err := br.init(src)
+	if err != nil {
+		return dst, err
+	}
+	maxDecodedSize := cap(dst)
+	dst = dst[:0]
+
+	// Avoid bounds check by always having full sized table.
+	const tlSize = 1 << tableLogMax
+	const tlMask = tlSize - 1
+	dt := d.dt.single[:tlSize]
+
+	// Use temp table to avoid bound checks/append penalty.
+	bufs := d.buffer()
+	buf := &bufs[0]
+	var off uint8
+
+	for br.off >= 8 {
+		br.fillFast()
+		v := dt[br.peekBitsFast(d.actualTableLog)&tlMask]
+		br.advance(uint8(v.entry))
+		buf[off+0] = uint8(v.entry >> 8)
+
+		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
+		br.advance(uint8(v.entry))
+		buf[off+1] = uint8(v.entry >> 8)
+
+		// Refill
+		br.fillFast()
+
+		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
+		br.advance(uint8(v.entry))
+		buf[off+2] = uint8(v.entry >> 8)
+
+		v = dt[br.peekBitsFast(d.actualTableLog)&tlMask]
+		br.advance(uint8(v.entry))
+		buf[off+3] = uint8(v.entry >> 8)
+
+		off += 4
+		if off == 0 {
+			if len(dst)+256 > maxDecodedSize {
+				br.close()
+				d.bufs.Put(bufs)
+				return nil, ErrMaxDecodedSizeExceeded
+			}
+			dst = append(dst, buf[:]...)
+		}
+	}
+
+	if len(dst)+int(off) > maxDecodedSize {
+		d.bufs.Put(bufs)
+		br.close()
+		return nil, ErrMaxDecodedSizeExceeded
+	}
+	dst = append(dst, buf[:off]...)
+
+	// br < 8, so uint8 is fine
+	bitsLeft := uint8(br.off)*8 + 64 - br.bitsRead
+	for bitsLeft > 0 {
+		br.fill()
+		if false && br.bitsRead >= 32 {
+			if br.off >= 4 {
+				v := br.in[br.off-4:]
+				v = v[:4]
+				low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
+				br.value = (br.value << 32) | uint64(low)
+				br.bitsRead -= 32
+				br.off -= 4
+			} else {
+				for br.off > 0 {
+					br.value = (br.value << 8) | uint64(br.in[br.off-1])
+					br.bitsRead -= 8
+					br.off--
+				}
+			}
+		}
+		if len(dst) >= maxDecodedSize {
+			d.bufs.Put(bufs)
+			br.close()
+			return nil, ErrMaxDecodedSizeExceeded
+		}
+		v := d.dt.single[br.peekBitsFast(d.actualTableLog)&tlMask]
+		nBits := uint8(v.entry)
+		br.advance(nBits)
+		bitsLeft -= nBits
+		dst = append(dst, uint8(v.entry>>8))
+	}
+	d.bufs.Put(bufs)
+	return dst, br.close()
+}
diff --git a/vendor/github.com/klauspost/compress/zstd/bitreader.go b/vendor/github.com/klauspost/compress/zstd/bitreader.go
index d7cd15ba2..97299d499 100644
--- a/vendor/github.com/klauspost/compress/zstd/bitreader.go
+++ b/vendor/github.com/klauspost/compress/zstd/bitreader.go
@@ -63,13 +63,6 @@ func (b *bitReader) get32BitsFast(n uint8) uint32 {
 	return v
 }
 
-func (b *bitReader) get16BitsFast(n uint8) uint16 {
-	const regMask = 64 - 1
-	v := uint16((b.value << (b.bitsRead & regMask)) >> ((regMask + 1 - n) & regMask))
-	b.bitsRead += n
-	return v
-}
-
 // fillFast() will make sure at least 32 bits are available.
 // There must be at least 4 bytes available.
 func (b *bitReader) fillFast() {
diff --git a/vendor/github.com/klauspost/compress/zstd/bitwriter.go b/vendor/github.com/klauspost/compress/zstd/bitwriter.go
index b36618285..78b3c61be 100644
--- a/vendor/github.com/klauspost/compress/zstd/bitwriter.go
+++ b/vendor/github.com/klauspost/compress/zstd/bitwriter.go
@@ -5,8 +5,6 @@
 
 package zstd
 
-import "fmt"
-
 // bitWriter will write bits.
 // First bit will be LSB of the first byte of output.
 type bitWriter struct {
@@ -73,80 +71,6 @@ func (b *bitWriter) addBits16Clean(value uint16, bits uint8) {
 	b.nBits += bits
 }
 
-// flush will flush all pending full bytes.
-// There will be at least 56 bits available for writing when this has been called.
-// Using flush32 is faster, but leaves less space for writing.
-func (b *bitWriter) flush() {
-	v := b.nBits >> 3
-	switch v {
-	case 0:
-	case 1:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-		)
-	case 2:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-		)
-	case 3:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-		)
-	case 4:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-		)
-	case 5:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-		)
-	case 6:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-		)
-	case 7:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-			byte(b.bitContainer>>48),
-		)
-	case 8:
-		b.out = append(b.out,
-			byte(b.bitContainer),
-			byte(b.bitContainer>>8),
-			byte(b.bitContainer>>16),
-			byte(b.bitContainer>>24),
-			byte(b.bitContainer>>32),
-			byte(b.bitContainer>>40),
-			byte(b.bitContainer>>48),
-			byte(b.bitContainer>>56),
-		)
-	default:
-		panic(fmt.Errorf("bits (%d) > 64", b.nBits))
-	}
-	b.bitContainer >>= v << 3
-	b.nBits &= 7
-}
-
 // flush32 will flush out, so there are at least 32 bits available for writing.
 func (b *bitWriter) flush32() {
 	if b.nBits < 32 {
diff --git a/vendor/github.com/klauspost/compress/zstd/blockdec.go b/vendor/github.com/klauspost/compress/zstd/blockdec.go
index b2bca3301..7eed729be 100644
--- a/vendor/github.com/klauspost/compress/zstd/blockdec.go
+++ b/vendor/github.com/klauspost/compress/zstd/blockdec.go
@@ -49,11 +49,8 @@ const (
 	// Maximum possible block size (all Raw+Uncompressed).
 	maxBlockSize = (1 << 21) - 1
 
-	// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#literals_section_header
-	maxCompressedLiteralSize = 1 << 18
-	maxRLELiteralSize        = 1 << 20
-	maxMatchLen              = 131074
-	maxSequences             = 0x7f00 + 0xffff
+	maxMatchLen  = 131074
+	maxSequences = 0x7f00 + 0xffff
 
 	// We support slightly less than the reference decoder to be able to
 	// use ints on 32 bit archs.
@@ -105,7 +102,6 @@ type blockDec struct {
 
 	// Block is RLE, this is the size.
 	RLESize uint32
-	tmp     [4]byte
 
 	Type blockType
 
@@ -368,14 +364,9 @@ func (b *blockDec) decodeLiterals(in []byte, hist *history) (remain []byte, err
 		}
 		if cap(b.literalBuf) < litRegenSize {
 			if b.lowMem {
-				b.literalBuf = make([]byte, litRegenSize)
+				b.literalBuf = make([]byte, litRegenSize, litRegenSize+compressedBlockOverAlloc)
 			} else {
-				if litRegenSize > maxCompressedLiteralSize {
-					// Exceptional
-					b.literalBuf = make([]byte, litRegenSize)
-				} else {
-					b.literalBuf = make([]byte, litRegenSize, maxCompressedLiteralSize)
-				}
+				b.literalBuf = make([]byte, litRegenSize, maxCompressedBlockSize+compressedBlockOverAlloc)
 			}
 		}
 		literals = b.literalBuf[:litRegenSize]
@@ -405,14 +396,14 @@ func (b *blockDec) decodeLiterals(in []byte, hist *history) (remain []byte, err
 		// Ensure we have space to store it.
 		if cap(b.literalBuf) < litRegenSize {
 			if b.lowMem {
-				b.literalBuf = make([]byte, 0, litRegenSize)
+				b.literalBuf = make([]byte, 0, litRegenSize+compressedBlockOverAlloc)
 			} else {
-				b.literalBuf = make([]byte, 0, maxCompressedLiteralSize)
+				b.literalBuf = make([]byte, 0, maxCompressedBlockSize+compressedBlockOverAlloc)
 			}
 		}
 		var err error
 		// Use our out buffer.
-		huff.MaxDecodedSize = maxCompressedBlockSize
+		huff.MaxDecodedSize = litRegenSize
 		if fourStreams {
 			literals, err = huff.Decoder().Decompress4X(b.literalBuf[:0:litRegenSize], literals)
 		} else {
@@ -437,9 +428,9 @@ func (b *blockDec) decodeLiterals(in []byte, hist *history) (remain []byte, err
 		// Ensure we have space to store it.
 		if cap(b.literalBuf) < litRegenSize {
 			if b.lowMem {
-				b.literalBuf = make([]byte, 0, litRegenSize)
+				b.literalBuf = make([]byte, 0, litRegenSize+compressedBlockOverAlloc)
 			} else {
-				b.literalBuf = make([]byte, 0, maxCompressedBlockSize)
+				b.literalBuf = make([]byte, 0, maxCompressedBlockSize+compressedBlockOverAlloc)
 			}
 		}
 		huff := hist.huffTree
@@ -456,7 +447,7 @@ func (b *blockDec) decodeLiterals(in []byte, hist *history) (remain []byte, err
 			return in, err
 		}
 		hist.huffTree = huff
-		huff.MaxDecodedSize = maxCompressedBlockSize
+		huff.MaxDecodedSize = litRegenSize
 		// Use our out buffer.
 		if fourStreams {
 			literals, err = huff.Decoder().Decompress4X(b.literalBuf[:0:litRegenSize], literals)
@@ -471,6 +462,8 @@ func (b *blockDec) decodeLiterals(in []byte, hist *history) (remain []byte, err
 		if len(literals) != litRegenSize {
 			return in, fmt.Errorf("literal output size mismatch want %d, got %d", litRegenSize, len(literals))
 		}
+		// Re-cap to get extra size.
+		literals = b.literalBuf[:len(literals)]
 		if debugDecoder {
 			printf("Decompressed %d literals into %d bytes\n", litCompSize, litRegenSize)
 		}
diff --git a/vendor/github.com/klauspost/compress/zstd/bytebuf.go b/vendor/github.com/klauspost/compress/zstd/bytebuf.go
index b80191e4b..4493baa75 100644
--- a/vendor/github.com/klauspost/compress/zstd/bytebuf.go
+++ b/vendor/github.com/klauspost/compress/zstd/bytebuf.go
@@ -52,10 +52,6 @@ func (b *byteBuf) readBig(n int, dst []byte) ([]byte, error) {
 	return r, nil
 }
 
-func (b *byteBuf) remain() []byte {
-	return *b
-}
-
 func (b *byteBuf) readByte() (byte, error) {
 	bb := *b
 	if len(bb) < 1 {
diff --git a/vendor/github.com/klauspost/compress/zstd/bytereader.go b/vendor/github.com/klauspost/compress/zstd/bytereader.go
index 2c4fca17f..0e59a242d 100644
--- a/vendor/github.com/klauspost/compress/zstd/bytereader.go
+++ b/vendor/github.com/klauspost/compress/zstd/bytereader.go
@@ -13,12 +13,6 @@ type byteReader struct {
 	off int
 }
 
-// init will initialize the reader and set the input.
-func (b *byteReader) init(in []byte) {
-	b.b = in
-	b.off = 0
-}
-
 // advance the stream b n bytes.
 func (b *byteReader) advance(n uint) {
 	b.off += int(n)
diff --git a/vendor/github.com/klauspost/compress/zstd/decoder.go b/vendor/github.com/klauspost/compress/zstd/decoder.go
index 36119f385..286c8f9d7 100644
--- a/vendor/github.com/klauspost/compress/zstd/decoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/decoder.go
@@ -637,60 +637,18 @@ func (d *Decoder) startSyncDecoder(r io.Reader) error {
 
 // Create Decoder:
 // ASYNC:
-// Spawn 4 go routines.
-// 0: Read frames and decode blocks.
-// 1: Decode block and literals. Receives hufftree and seqdecs, returns seqdecs and huff tree.
-// 2: Wait for recentOffsets if needed. Decode sequences, send recentOffsets.
-// 3: Wait for stream history, execute sequences, send stream history.
+// Spawn 3 go routines.
+// 0: Read frames and decode block literals.
+// 1: Decode sequences.
+// 2: Execute sequences, send to output.
 func (d *Decoder) startStreamDecoder(ctx context.Context, r io.Reader, output chan decodeOutput) {
 	defer d.streamWg.Done()
 	br := readerWrapper{r: r}
 
-	var seqPrepare = make(chan *blockDec, d.o.concurrent)
 	var seqDecode = make(chan *blockDec, d.o.concurrent)
 	var seqExecute = make(chan *blockDec, d.o.concurrent)
 
-	// Async 1: Prepare blocks...
-	go func() {
-		var hist history
-		var hasErr bool
-		for block := range seqPrepare {
-			if hasErr {
-				if block != nil {
-					seqDecode <- block
-				}
-				continue
-			}
-			if block.async.newHist != nil {
-				if debugDecoder {
-					println("Async 1: new history")
-				}
-				hist.reset()
-				if block.async.newHist.dict != nil {
-					hist.setDict(block.async.newHist.dict)
-				}
-			}
-			if block.err != nil || block.Type != blockTypeCompressed {
-				hasErr = block.err != nil
-				seqDecode <- block
-				continue
-			}
-
-			remain, err := block.decodeLiterals(block.data, &hist)
-			block.err = err
-			hasErr = block.err != nil
-			if err == nil {
-				block.async.literals = hist.decoders.literals
-				block.async.seqData = remain
-			} else if debugDecoder {
-				println("decodeLiterals error:", err)
-			}
-			seqDecode <- block
-		}
-		close(seqDecode)
-	}()
-
-	// Async 2: Decode sequences...
+	// Async 1: Decode sequences...
 	go func() {
 		var hist history
 		var hasErr bool
@@ -704,7 +662,7 @@ func (d *Decoder) startStreamDecoder(ctx context.Context, r io.Reader, output ch
 			}
 			if block.async.newHist != nil {
 				if debugDecoder {
-					println("Async 2: new history, recent:", block.async.newHist.recentOffsets)
+					println("Async 1: new history, recent:", block.async.newHist.recentOffsets)
 				}
 				hist.decoders = block.async.newHist.decoders
 				hist.recentOffsets = block.async.newHist.recentOffsets
@@ -758,7 +716,7 @@ func (d *Decoder) startStreamDecoder(ctx context.Context, r io.Reader, output ch
 			}
 			if block.async.newHist != nil {
 				if debugDecoder {
-					println("Async 3: new history")
+					println("Async 2: new history")
 				}
 				hist.windowSize = block.async.newHist.windowSize
 				hist.allocFrameBuffer = block.async.newHist.allocFrameBuffer
@@ -845,6 +803,33 @@ func (d *Decoder) startStreamDecoder(ctx context.Context, r io.Reader, output ch
 
 decodeStream:
 	for {
+		var hist history
+		var hasErr bool
+
+		decodeBlock := func(block *blockDec) {
+			if hasErr {
+				if block != nil {
+					seqDecode <- block
+				}
+				return
+			}
+			if block.err != nil || block.Type != blockTypeCompressed {
+				hasErr = block.err != nil
+				seqDecode <- block
+				return
+			}
+
+			remain, err := block.decodeLiterals(block.data, &hist)
+			block.err = err
+			hasErr = block.err != nil
+			if err == nil {
+				block.async.literals = hist.decoders.literals
+				block.async.seqData = remain
+			} else if debugDecoder {
+				println("decodeLiterals error:", err)
+			}
+			seqDecode <- block
+		}
 		frame := d.frame
 		if debugDecoder {
 			println("New frame...")
@@ -871,7 +856,7 @@ decodeStream:
 			case <-ctx.Done():
 			case dec := <-d.decoders:
 				dec.sendErr(err)
-				seqPrepare <- dec
+				decodeBlock(dec)
 			}
 			break decodeStream
 		}
@@ -891,6 +876,10 @@ decodeStream:
 				if debugDecoder {
 					println("Alloc History:", h.allocFrameBuffer)
 				}
+				hist.reset()
+				if h.dict != nil {
+					hist.setDict(h.dict)
+				}
 				dec.async.newHist = &h
 				dec.async.fcs = frame.FrameContentSize
 				historySent = true
@@ -917,7 +906,7 @@ decodeStream:
 			}
 			err = dec.err
 			last := dec.Last
-			seqPrepare <- dec
+			decodeBlock(dec)
 			if err != nil {
 				break decodeStream
 			}
@@ -926,7 +915,7 @@ decodeStream:
 			}
 		}
 	}
-	close(seqPrepare)
+	close(seqDecode)
 	wg.Wait()
 	d.frame.history.b = frameHistCache
 }
diff --git a/vendor/github.com/klauspost/compress/zstd/enc_better.go b/vendor/github.com/klauspost/compress/zstd/enc_better.go
index 602c05ee0..c769f6941 100644
--- a/vendor/github.com/klauspost/compress/zstd/enc_better.go
+++ b/vendor/github.com/klauspost/compress/zstd/enc_better.go
@@ -156,8 +156,8 @@ encodeLoop:
 				panic("offset0 was 0")
 			}
 
-			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			nextHashL := hashLen(cv, betterLongTableBits, betterLongLen)
+			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			candidateL := e.longTable[nextHashL]
 			candidateS := e.table[nextHashS]
 
@@ -518,8 +518,8 @@ encodeLoop:
 			}
 
 			// Store this, since we have it.
-			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			nextHashL := hashLen(cv, betterLongTableBits, betterLongLen)
+			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
@@ -674,8 +674,8 @@ encodeLoop:
 				panic("offset0 was 0")
 			}
 
-			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			nextHashL := hashLen(cv, betterLongTableBits, betterLongLen)
+			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			candidateL := e.longTable[nextHashL]
 			candidateS := e.table[nextHashS]
 
@@ -1047,8 +1047,8 @@ encodeLoop:
 			}
 
 			// Store this, since we have it.
-			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 			nextHashL := hashLen(cv, betterLongTableBits, betterLongLen)
+			nextHashS := hashLen(cv, betterShortTableBits, betterShortLen)
 
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
diff --git a/vendor/github.com/klauspost/compress/zstd/enc_dfast.go b/vendor/github.com/klauspost/compress/zstd/enc_dfast.go
index d6b310424..7ff0c64fa 100644
--- a/vendor/github.com/klauspost/compress/zstd/enc_dfast.go
+++ b/vendor/github.com/klauspost/compress/zstd/enc_dfast.go
@@ -127,8 +127,8 @@ encodeLoop:
 				panic("offset0 was 0")
 			}
 
-			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			nextHashL := hashLen(cv, dFastLongTableBits, dFastLongLen)
+			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			candidateL := e.longTable[nextHashL]
 			candidateS := e.table[nextHashS]
 
@@ -439,8 +439,8 @@ encodeLoop:
 		var t int32
 		for {
 
-			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			nextHashL := hashLen(cv, dFastLongTableBits, dFastLongLen)
+			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			candidateL := e.longTable[nextHashL]
 			candidateS := e.table[nextHashS]
 
@@ -785,8 +785,8 @@ encodeLoop:
 				panic("offset0 was 0")
 			}
 
-			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			nextHashL := hashLen(cv, dFastLongTableBits, dFastLongLen)
+			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			candidateL := e.longTable[nextHashL]
 			candidateS := e.table[nextHashS]
 
@@ -969,7 +969,7 @@ encodeLoop:
 		te0 := tableEntry{offset: index0 + e.cur, val: uint32(cv0)}
 		te1 := tableEntry{offset: index1 + e.cur, val: uint32(cv1)}
 		longHash1 := hashLen(cv0, dFastLongTableBits, dFastLongLen)
-		longHash2 := hashLen(cv0, dFastLongTableBits, dFastLongLen)
+		longHash2 := hashLen(cv1, dFastLongTableBits, dFastLongLen)
 		e.longTable[longHash1] = te0
 		e.longTable[longHash2] = te1
 		e.markLongShardDirty(longHash1)
@@ -1002,8 +1002,8 @@ encodeLoop:
 			}
 
 			// Store this, since we have it.
-			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 			nextHashL := hashLen(cv, dFastLongTableBits, dFastLongLen)
+			nextHashS := hashLen(cv, dFastShortTableBits, dFastShortLen)
 
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
diff --git a/vendor/github.com/klauspost/compress/zstd/encoder.go b/vendor/github.com/klauspost/compress/zstd/encoder.go
index dcc987a7c..e6b1d01cf 100644
--- a/vendor/github.com/klauspost/compress/zstd/encoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/encoder.go
@@ -551,7 +551,7 @@ func (e *Encoder) EncodeAll(src, dst []byte) []byte {
 	}
 
 	// If we can do everything in one block, prefer that.
-	if len(src) <= maxCompressedBlockSize {
+	if len(src) <= e.o.blockSize {
 		enc.Reset(e.o.dict, true)
 		// Slightly faster with no history and everything in one block.
 		if e.o.crc {
diff --git a/vendor/github.com/klauspost/compress/zstd/framedec.go b/vendor/github.com/klauspost/compress/zstd/framedec.go
index 3ff109cce..fa0a633f3 100644
--- a/vendor/github.com/klauspost/compress/zstd/framedec.go
+++ b/vendor/github.com/klauspost/compress/zstd/framedec.go
@@ -253,10 +253,11 @@ func (d *frameDec) reset(br byteBuffer) error {
 		return ErrWindowSizeTooSmall
 	}
 	d.history.windowSize = int(d.WindowSize)
-	if d.o.lowMem && d.history.windowSize < maxBlockSize {
+	if !d.o.lowMem || d.history.windowSize < maxBlockSize {
+		// Alloc 2x window size if not low-mem, or very small window size.
 		d.history.allocFrameBuffer = d.history.windowSize * 2
-		// TODO: Maybe use FrameContent size
 	} else {
+		// Alloc with one additional block
 		d.history.allocFrameBuffer = d.history.windowSize + maxBlockSize
 	}
 
diff --git a/vendor/github.com/klauspost/compress/zstd/fse_decoder.go b/vendor/github.com/klauspost/compress/zstd/fse_decoder.go
index fde4e6b60..23333b969 100644
--- a/vendor/github.com/klauspost/compress/zstd/fse_decoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/fse_decoder.go
@@ -229,18 +229,10 @@ func (d decSymbol) newState() uint16 {
 	return uint16(d >> 16)
 }
 
-func (d decSymbol) baseline() uint32 {
-	return uint32(d >> 32)
-}
-
 func (d decSymbol) baselineInt() int {
 	return int(d >> 32)
 }
 
-func (d *decSymbol) set(nbits, addBits uint8, newState uint16, baseline uint32) {
-	*d = decSymbol(nbits) | (decSymbol(addBits) << 8) | (decSymbol(newState) << 16) | (decSymbol(baseline) << 32)
-}
-
 func (d *decSymbol) setNBits(nBits uint8) {
 	const mask = 0xffffffffffffff00
 	*d = (*d & mask) | decSymbol(nBits)
@@ -256,11 +248,6 @@ func (d *decSymbol) setNewState(state uint16) {
 	*d = (*d & mask) | decSymbol(state)<<16
 }
 
-func (d *decSymbol) setBaseline(baseline uint32) {
-	const mask = 0xffffffff
-	*d = (*d & mask) | decSymbol(baseline)<<32
-}
-
 func (d *decSymbol) setExt(addBits uint8, baseline uint32) {
 	const mask = 0xffff00ff
 	*d = (*d & mask) | (decSymbol(addBits) << 8) | (decSymbol(baseline) << 32)
@@ -377,34 +364,7 @@ func (s *fseState) init(br *bitReader, tableLog uint8, dt []decSymbol) {
 	s.state = dt[br.getBits(tableLog)]
 }
 
-// next returns the current symbol and sets the next state.
-// At least tablelog bits must be available in the bit reader.
-func (s *fseState) next(br *bitReader) {
-	lowBits := uint16(br.getBits(s.state.nbBits()))
-	s.state = s.dt[s.state.newState()+lowBits]
-}
-
-// finished returns true if all bits have been read from the bitstream
-// and the next state would require reading bits from the input.
-func (s *fseState) finished(br *bitReader) bool {
-	return br.finished() && s.state.nbBits() > 0
-}
-
-// final returns the current state symbol without decoding the next.
-func (s *fseState) final() (int, uint8) {
-	return s.state.baselineInt(), s.state.addBits()
-}
-
 // final returns the current state symbol without decoding the next.
 func (s decSymbol) final() (int, uint8) {
 	return s.baselineInt(), s.addBits()
 }
-
-// nextFast returns the next symbol and sets the next state.
-// This can only be used if no symbols are 0 bits.
-// At least tablelog bits must be available in the bit reader.
-func (s *fseState) nextFast(br *bitReader) (uint32, uint8) {
-	lowBits := br.get16BitsFast(s.state.nbBits())
-	s.state = s.dt[s.state.newState()+lowBits]
-	return s.state.baseline(), s.state.addBits()
-}
diff --git a/vendor/github.com/klauspost/compress/zstd/fse_encoder.go b/vendor/github.com/klauspost/compress/zstd/fse_encoder.go
index 5442061b1..ab26326a8 100644
--- a/vendor/github.com/klauspost/compress/zstd/fse_encoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/fse_encoder.go
@@ -76,21 +76,6 @@ func (s *fseEncoder) HistogramFinished(maxSymbol uint8, maxCount int) {
 	s.clearCount = maxCount != 0
 }
 
-// prepare will prepare and allocate scratch tables used for both compression and decompression.
-func (s *fseEncoder) prepare() (*fseEncoder, error) {
-	if s == nil {
-		s = &fseEncoder{}
-	}
-	s.useRLE = false
-	if s.clearCount && s.maxCount == 0 {
-		for i := range s.count {
-			s.count[i] = 0
-		}
-		s.clearCount = false
-	}
-	return s, nil
-}
-
 // allocCtable will allocate tables needed for compression.
 // If existing tables a re big enough, they are simply re-used.
 func (s *fseEncoder) allocCtable() {
@@ -709,14 +694,6 @@ func (c *cState) init(bw *bitWriter, ct *cTable, first symbolTransform) {
 	c.state = c.stateTable[lu]
 }
 
-// encode the output symbol provided and write it to the bitstream.
-func (c *cState) encode(symbolTT symbolTransform) {
-	nbBitsOut := (uint32(c.state) + symbolTT.deltaNbBits) >> 16
-	dstState := int32(c.state>>(nbBitsOut&15)) + int32(symbolTT.deltaFindState)
-	c.bw.addBits16NC(c.state, uint8(nbBitsOut))
-	c.state = c.stateTable[dstState]
-}
-
 // flush will write the tablelog to the output and flush the remaining full bytes.
 func (c *cState) flush(tableLog uint8) {
 	c.bw.flush32()
diff --git a/vendor/github.com/klauspost/compress/zstd/hash.go b/vendor/github.com/klauspost/compress/zstd/hash.go
index cf33f29a1..5d73c21eb 100644
--- a/vendor/github.com/klauspost/compress/zstd/hash.go
+++ b/vendor/github.com/klauspost/compress/zstd/hash.go
@@ -33,9 +33,3 @@ func hashLen(u uint64, length, mls uint8) uint32 {
 		return (uint32(u) * prime4bytes) >> (32 - length)
 	}
 }
-
-// hash3 returns the hash of the lower 3 bytes of u to fit in a hash table with h bits.
-// Preferably h should be a constant and should always be <32.
-func hash3(u uint32, h uint8) uint32 {
-	return ((u << (32 - 24)) * prime3bytes) >> ((32 - h) & 31)
-}
diff --git a/vendor/github.com/klauspost/compress/zstd/seqdec.go b/vendor/github.com/klauspost/compress/zstd/seqdec.go
index e80139dd9..df0447203 100644
--- a/vendor/github.com/klauspost/compress/zstd/seqdec.go
+++ b/vendor/github.com/klauspost/compress/zstd/seqdec.go
@@ -188,6 +188,7 @@ func (s *sequenceDecs) execute(seqs []seqVals, hist []byte) error {
 			}
 		}
 	}
+
 	// Add final literals
 	copy(out[t:], s.literals)
 	if debugDecoder {
@@ -203,12 +204,11 @@ func (s *sequenceDecs) execute(seqs []seqVals, hist []byte) error {
 
 // decode sequences from the stream with the provided history.
 func (s *sequenceDecs) decodeSync(hist []byte) error {
-	if true {
-		supported, err := s.decodeSyncSimple(hist)
-		if supported {
-			return err
-		}
+	supported, err := s.decodeSyncSimple(hist)
+	if supported {
+		return err
 	}
+
 	br := s.br
 	seqs := s.nSeqs
 	startSize := len(s.out)
@@ -396,6 +396,7 @@ func (s *sequenceDecs) decodeSync(hist []byte) error {
 			ofState = ofTable[ofState.newState()&maxTableMask]
 		} else {
 			bits := br.get32BitsFast(nBits)
+
 			lowBits := uint16(bits >> ((ofState.nbBits() + mlState.nbBits()) & 31))
 			llState = llTable[(llState.newState()+lowBits)&maxTableMask]
 
@@ -418,16 +419,6 @@ func (s *sequenceDecs) decodeSync(hist []byte) error {
 	return br.close()
 }
 
-// update states, at least 27 bits must be available.
-func (s *sequenceDecs) update(br *bitReader) {
-	// Max 8 bits
-	s.litLengths.state.next(br)
-	// Max 9 bits
-	s.matchLengths.state.next(br)
-	// Max 8 bits
-	s.offsets.state.next(br)
-}
-
 var bitMask [16]uint16
 
 func init() {
@@ -436,87 +427,6 @@ func init() {
 	}
 }
 
-// update states, at least 27 bits must be available.
-func (s *sequenceDecs) updateAlt(br *bitReader) {
-	// Update all 3 states at once. Approx 20% faster.
-	a, b, c := s.litLengths.state.state, s.matchLengths.state.state, s.offsets.state.state
-
-	nBits := a.nbBits() + b.nbBits() + c.nbBits()
-	if nBits == 0 {
-		s.litLengths.state.state = s.litLengths.state.dt[a.newState()]
-		s.matchLengths.state.state = s.matchLengths.state.dt[b.newState()]
-		s.offsets.state.state = s.offsets.state.dt[c.newState()]
-		return
-	}
-	bits := br.get32BitsFast(nBits)
-	lowBits := uint16(bits >> ((c.nbBits() + b.nbBits()) & 31))
-	s.litLengths.state.state = s.litLengths.state.dt[a.newState()+lowBits]
-
-	lowBits = uint16(bits >> (c.nbBits() & 31))
-	lowBits &= bitMask[b.nbBits()&15]
-	s.matchLengths.state.state = s.matchLengths.state.dt[b.newState()+lowBits]
-
-	lowBits = uint16(bits) & bitMask[c.nbBits()&15]
-	s.offsets.state.state = s.offsets.state.dt[c.newState()+lowBits]
-}
-
-// nextFast will return new states when there are at least 4 unused bytes left on the stream when done.
-func (s *sequenceDecs) nextFast(br *bitReader, llState, mlState, ofState decSymbol) (ll, mo, ml int) {
-	// Final will not read from stream.
-	ll, llB := llState.final()
-	ml, mlB := mlState.final()
-	mo, moB := ofState.final()
-
-	// extra bits are stored in reverse order.
-	br.fillFast()
-	mo += br.getBits(moB)
-	if s.maxBits > 32 {
-		br.fillFast()
-	}
-	ml += br.getBits(mlB)
-	ll += br.getBits(llB)
-
-	if moB > 1 {
-		s.prevOffset[2] = s.prevOffset[1]
-		s.prevOffset[1] = s.prevOffset[0]
-		s.prevOffset[0] = mo
-		return
-	}
-	// mo = s.adjustOffset(mo, ll, moB)
-	// Inlined for rather big speedup
-	if ll == 0 {
-		// There is an exception though, when current sequence's literals_length = 0.
-		// In this case, repeated offsets are shifted by one, so an offset_value of 1 means Repeated_Offset2,
-		// an offset_value of 2 means Repeated_Offset3, and an offset_value of 3 means Repeated_Offset1 - 1_byte.
-		mo++
-	}
-
-	if mo == 0 {
-		mo = s.prevOffset[0]
-		return
-	}
-	var temp int
-	if mo == 3 {
-		temp = s.prevOffset[0] - 1
-	} else {
-		temp = s.prevOffset[mo]
-	}
-
-	if temp == 0 {
-		// 0 is not valid; input is corrupted; force offset to 1
-		println("temp was 0")
-		temp = 1
-	}
-
-	if mo != 1 {
-		s.prevOffset[2] = s.prevOffset[1]
-	}
-	s.prevOffset[1] = s.prevOffset[0]
-	s.prevOffset[0] = temp
-	mo = temp
-	return
-}
-
 func (s *sequenceDecs) next(br *bitReader, llState, mlState, ofState decSymbol) (ll, mo, ml int) {
 	// Final will not read from stream.
 	ll, llB := llState.final()
diff --git a/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.go b/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.go
index 4676b09cc..847b322ae 100644
--- a/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.go
+++ b/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.go
@@ -62,6 +62,10 @@ func (s *sequenceDecs) decodeSyncSimple(hist []byte) (bool, error) {
 	if s.maxSyncLen > 0 && cap(s.out)-len(s.out)-compressedBlockOverAlloc < int(s.maxSyncLen) {
 		useSafe = true
 	}
+	if cap(s.literals) < len(s.literals)+compressedBlockOverAlloc {
+		useSafe = true
+	}
+
 	br := s.br
 
 	maxBlockSize := maxCompressedBlockSize
@@ -301,6 +305,10 @@ type executeAsmContext struct {
 //go:noescape
 func sequenceDecs_executeSimple_amd64(ctx *executeAsmContext) bool
 
+// Same as above, but with safe memcopies
+//go:noescape
+func sequenceDecs_executeSimple_safe_amd64(ctx *executeAsmContext) bool
+
 // executeSimple handles cases when dictionary is not used.
 func (s *sequenceDecs) executeSimple(seqs []seqVals, hist []byte) error {
 	// Ensure we have enough output size...
@@ -327,8 +335,12 @@ func (s *sequenceDecs) executeSimple(seqs []seqVals, hist []byte) error {
 		literals:    s.literals,
 		windowSize:  s.windowSize,
 	}
-
-	ok := sequenceDecs_executeSimple_amd64(&ctx)
+	var ok bool
+	if cap(s.literals) < len(s.literals)+compressedBlockOverAlloc {
+		ok = sequenceDecs_executeSimple_safe_amd64(&ctx)
+	} else {
+		ok = sequenceDecs_executeSimple_amd64(&ctx)
+	}
 	if !ok {
 		return fmt.Errorf("match offset (%d) bigger than current history (%d)",
 			seqs[ctx.seqIndex].mo, ctx.outPosition+len(hist))
diff --git a/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.s b/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.s
index 2585b2e98..212c6cac3 100644
--- a/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.s
+++ b/vendor/github.com/klauspost/compress/zstd/seqdec_amd64.s
@@ -705,60 +705,55 @@ sequenceDecs_decode_bmi2_fill_2_end:
 	MOVQ   CX, (R9)
 
 	// Fill bitreader for state updates
-	MOVQ   R13, (SP)
-	MOVQ   $0x00000808, CX
-	BEXTRQ CX, R8, R13
-	MOVQ   ctx+16(FP), CX
-	CMPQ   96(CX), $0x00
-	JZ     sequenceDecs_decode_bmi2_skip_update
-
-	// Update Literal Length State
-	MOVBQZX SI, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, SI, SI
+	MOVQ    R13, (SP)
+	MOVQ    $0x00000808, CX
+	BEXTRQ  CX, R8, R13
+	MOVQ    ctx+16(FP), CX
+	CMPQ    96(CX), $0x00
+	JZ      sequenceDecs_decode_bmi2_skip_update
+	LEAQ    (SI)(DI*1), R14
+	ADDQ    R8, R14
+	MOVBQZX R14, R14
 	LEAQ    (DX)(R14*1), CX
 	MOVQ    AX, R15
 	MOVQ    CX, DX
 	ROLQ    CL, R15
 	BZHIQ   R14, R15, R15
-	ADDQ    R15, SI
 
-	// Load ctx.llTable
+	// Update Offset State
+	BZHIQ  R8, R15, CX
+	SHRXQ  R8, R15, R15
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, R8, R8
+	ADDQ   CX, R8
+
+	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
-	MOVQ (CX), CX
-	MOVQ (CX)(SI*8), SI
+	MOVQ 48(CX), CX
+	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	MOVBQZX DI, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, DI, DI
-	LEAQ    (DX)(R14*1), CX
-	MOVQ    AX, R15
-	MOVQ    CX, DX
-	ROLQ    CL, R15
-	BZHIQ   R14, R15, R15
-	ADDQ    R15, DI
+	BZHIQ  DI, R15, CX
+	SHRXQ  DI, R15, R15
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, DI, DI
+	ADDQ   CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
 	MOVQ 24(CX), CX
 	MOVQ (CX)(DI*8), DI
 
-	// Update Offset State
-	MOVBQZX R8, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, R8, R8
-	LEAQ    (DX)(R14*1), CX
-	MOVQ    AX, R15
-	MOVQ    CX, DX
-	ROLQ    CL, R15
-	BZHIQ   R14, R15, R15
-	ADDQ    R15, R8
+	// Update Literal Length State
+	BZHIQ  SI, R15, CX
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, SI, SI
+	ADDQ   CX, SI
 
-	// Load ctx.ofTable
+	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
-	MOVQ 48(CX), CX
-	MOVQ (CX)(R8*8), R8
+	MOVQ (CX), CX
+	MOVQ (CX)(SI*8), SI
 
 sequenceDecs_decode_bmi2_skip_update:
 	// Adjust offset
@@ -965,60 +960,55 @@ sequenceDecs_decode_56_bmi2_fill_end:
 	MOVQ   CX, (R9)
 
 	// Fill bitreader for state updates
-	MOVQ   R13, (SP)
-	MOVQ   $0x00000808, CX
-	BEXTRQ CX, R8, R13
-	MOVQ   ctx+16(FP), CX
-	CMPQ   96(CX), $0x00
-	JZ     sequenceDecs_decode_56_bmi2_skip_update
-
-	// Update Literal Length State
-	MOVBQZX SI, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, SI, SI
+	MOVQ    R13, (SP)
+	MOVQ    $0x00000808, CX
+	BEXTRQ  CX, R8, R13
+	MOVQ    ctx+16(FP), CX
+	CMPQ    96(CX), $0x00
+	JZ      sequenceDecs_decode_56_bmi2_skip_update
+	LEAQ    (SI)(DI*1), R14
+	ADDQ    R8, R14
+	MOVBQZX R14, R14
 	LEAQ    (DX)(R14*1), CX
 	MOVQ    AX, R15
 	MOVQ    CX, DX
 	ROLQ    CL, R15
 	BZHIQ   R14, R15, R15
-	ADDQ    R15, SI
 
-	// Load ctx.llTable
+	// Update Offset State
+	BZHIQ  R8, R15, CX
+	SHRXQ  R8, R15, R15
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, R8, R8
+	ADDQ   CX, R8
+
+	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
-	MOVQ (CX), CX
-	MOVQ (CX)(SI*8), SI
+	MOVQ 48(CX), CX
+	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	MOVBQZX DI, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, DI, DI
-	LEAQ    (DX)(R14*1), CX
-	MOVQ    AX, R15
-	MOVQ    CX, DX
-	ROLQ    CL, R15
-	BZHIQ   R14, R15, R15
-	ADDQ    R15, DI
+	BZHIQ  DI, R15, CX
+	SHRXQ  DI, R15, R15
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, DI, DI
+	ADDQ   CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
 	MOVQ 24(CX), CX
 	MOVQ (CX)(DI*8), DI
 
-	// Update Offset State
-	MOVBQZX R8, R14
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, R8, R8
-	LEAQ    (DX)(R14*1), CX
-	MOVQ    AX, R15
-	MOVQ    CX, DX
-	ROLQ    CL, R15
-	BZHIQ   R14, R15, R15
-	ADDQ    R15, R8
+	// Update Literal Length State
+	BZHIQ  SI, R15, CX
+	MOVQ   $0x00001010, R14
+	BEXTRQ R14, SI, SI
+	ADDQ   CX, SI
 
-	// Load ctx.ofTable
+	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
-	MOVQ 48(CX), CX
-	MOVQ (CX)(R8*8), R8
+	MOVQ (CX), CX
+	MOVQ (CX)(SI*8), SI
 
 sequenceDecs_decode_56_bmi2_skip_update:
 	// Adjust offset
@@ -1171,6 +1161,228 @@ main_loop:
 	TESTQ R11, R11
 	JZ    check_offset
 	XORQ  R14, R14
+
+copy_1:
+	MOVUPS (SI)(R14*1), X0
+	MOVUPS X0, (BX)(R14*1)
+	ADDQ   $0x10, R14
+	CMPQ   R14, R11
+	JB     copy_1
+	ADDQ   R11, SI
+	ADDQ   R11, BX
+	ADDQ   R11, DI
+
+	// Malformed input if seq.mo > t+len(hist) || seq.mo > s.windowSize)
+check_offset:
+	LEAQ (DI)(R10*1), R11
+	CMPQ R12, R11
+	JG   error_match_off_too_big
+	CMPQ R12, R8
+	JG   error_match_off_too_big
+
+	// Copy match from history
+	MOVQ  R12, R11
+	SUBQ  DI, R11
+	JLS   copy_match
+	MOVQ  R9, R14
+	SUBQ  R11, R14
+	CMPQ  R13, R11
+	JGE   copy_all_from_history
+	XORQ  R11, R11
+	TESTQ $0x00000001, R13
+	JZ    copy_4_word
+	MOVB  (R14)(R11*1), R12
+	MOVB  R12, (BX)(R11*1)
+	ADDQ  $0x01, R11
+
+copy_4_word:
+	TESTQ $0x00000002, R13
+	JZ    copy_4_dword
+	MOVW  (R14)(R11*1), R12
+	MOVW  R12, (BX)(R11*1)
+	ADDQ  $0x02, R11
+
+copy_4_dword:
+	TESTQ $0x00000004, R13
+	JZ    copy_4_qword
+	MOVL  (R14)(R11*1), R12
+	MOVL  R12, (BX)(R11*1)
+	ADDQ  $0x04, R11
+
+copy_4_qword:
+	TESTQ $0x00000008, R13
+	JZ    copy_4_test
+	MOVQ  (R14)(R11*1), R12
+	MOVQ  R12, (BX)(R11*1)
+	ADDQ  $0x08, R11
+	JMP   copy_4_test
+
+copy_4:
+	MOVUPS (R14)(R11*1), X0
+	MOVUPS X0, (BX)(R11*1)
+	ADDQ   $0x10, R11
+
+copy_4_test:
+	CMPQ R11, R13
+	JB   copy_4
+	ADDQ R13, DI
+	ADDQ R13, BX
+	ADDQ $0x18, AX
+	INCQ DX
+	CMPQ DX, CX
+	JB   main_loop
+	JMP  loop_finished
+
+copy_all_from_history:
+	XORQ  R15, R15
+	TESTQ $0x00000001, R11
+	JZ    copy_5_word
+	MOVB  (R14)(R15*1), BP
+	MOVB  BP, (BX)(R15*1)
+	ADDQ  $0x01, R15
+
+copy_5_word:
+	TESTQ $0x00000002, R11
+	JZ    copy_5_dword
+	MOVW  (R14)(R15*1), BP
+	MOVW  BP, (BX)(R15*1)
+	ADDQ  $0x02, R15
+
+copy_5_dword:
+	TESTQ $0x00000004, R11
+	JZ    copy_5_qword
+	MOVL  (R14)(R15*1), BP
+	MOVL  BP, (BX)(R15*1)
+	ADDQ  $0x04, R15
+
+copy_5_qword:
+	TESTQ $0x00000008, R11
+	JZ    copy_5_test
+	MOVQ  (R14)(R15*1), BP
+	MOVQ  BP, (BX)(R15*1)
+	ADDQ  $0x08, R15
+	JMP   copy_5_test
+
+copy_5:
+	MOVUPS (R14)(R15*1), X0
+	MOVUPS X0, (BX)(R15*1)
+	ADDQ   $0x10, R15
+
+copy_5_test:
+	CMPQ R15, R11
+	JB   copy_5
+	ADDQ R11, BX
+	ADDQ R11, DI
+	SUBQ R11, R13
+
+	// Copy match from the current buffer
+copy_match:
+	TESTQ R13, R13
+	JZ    handle_loop
+	MOVQ  BX, R11
+	SUBQ  R12, R11
+
+	// ml <= mo
+	CMPQ R13, R12
+	JA   copy_overlapping_match
+
+	// Copy non-overlapping match
+	ADDQ R13, DI
+	MOVQ BX, R12
+	ADDQ R13, BX
+
+copy_2:
+	MOVUPS (R11), X0
+	MOVUPS X0, (R12)
+	ADDQ   $0x10, R11
+	ADDQ   $0x10, R12
+	SUBQ   $0x10, R13
+	JHI    copy_2
+	JMP    handle_loop
+
+	// Copy overlapping match
+copy_overlapping_match:
+	ADDQ R13, DI
+
+copy_slow_3:
+	MOVB (R11), R12
+	MOVB R12, (BX)
+	INCQ R11
+	INCQ BX
+	DECQ R13
+	JNZ  copy_slow_3
+
+handle_loop:
+	ADDQ $0x18, AX
+	INCQ DX
+	CMPQ DX, CX
+	JB   main_loop
+
+loop_finished:
+	// Return value
+	MOVB $0x01, ret+8(FP)
+
+	// Update the context
+	MOVQ ctx+0(FP), AX
+	MOVQ DX, 24(AX)
+	MOVQ DI, 104(AX)
+	MOVQ 80(AX), CX
+	SUBQ CX, SI
+	MOVQ SI, 112(AX)
+	RET
+
+error_match_off_too_big:
+	// Return value
+	MOVB $0x00, ret+8(FP)
+
+	// Update the context
+	MOVQ ctx+0(FP), AX
+	MOVQ DX, 24(AX)
+	MOVQ DI, 104(AX)
+	MOVQ 80(AX), CX
+	SUBQ CX, SI
+	MOVQ SI, 112(AX)
+	RET
+
+empty_seqs:
+	// Return value
+	MOVB $0x01, ret+8(FP)
+	RET
+
+// func sequenceDecs_executeSimple_safe_amd64(ctx *executeAsmContext) bool
+// Requires: SSE
+TEXT ·sequenceDecs_executeSimple_safe_amd64(SB), $8-9
+	MOVQ  ctx+0(FP), R10
+	MOVQ  8(R10), CX
+	TESTQ CX, CX
+	JZ    empty_seqs
+	MOVQ  (R10), AX
+	MOVQ  24(R10), DX
+	MOVQ  32(R10), BX
+	MOVQ  80(R10), SI
+	MOVQ  104(R10), DI
+	MOVQ  120(R10), R8
+	MOVQ  56(R10), R9
+	MOVQ  64(R10), R10
+	ADDQ  R10, R9
+
+	// seqsBase += 24 * seqIndex
+	LEAQ (DX)(DX*2), R11
+	SHLQ $0x03, R11
+	ADDQ R11, AX
+
+	// outBase += outPosition
+	ADDQ DI, BX
+
+main_loop:
+	MOVQ (AX), R11
+	MOVQ 16(AX), R12
+	MOVQ 8(AX), R13
+
+	// Copy literals
+	TESTQ R11, R11
+	JZ    check_offset
+	XORQ  R14, R14
 	TESTQ $0x00000001, R11
 	JZ    copy_1_word
 	MOVB  (SI)(R14*1), R15
@@ -1326,18 +1538,46 @@ copy_match:
 	JA   copy_overlapping_match
 
 	// Copy non-overlapping match
-	ADDQ R13, DI
-	MOVQ BX, R12
-	ADDQ R13, BX
+	ADDQ  R13, DI
+	XORQ  R12, R12
+	TESTQ $0x00000001, R13
+	JZ    copy_2_word
+	MOVB  (R11)(R12*1), R14
+	MOVB  R14, (BX)(R12*1)
+	ADDQ  $0x01, R12
+
+copy_2_word:
+	TESTQ $0x00000002, R13
+	JZ    copy_2_dword
+	MOVW  (R11)(R12*1), R14
+	MOVW  R14, (BX)(R12*1)
+	ADDQ  $0x02, R12
+
+copy_2_dword:
+	TESTQ $0x00000004, R13
+	JZ    copy_2_qword
+	MOVL  (R11)(R12*1), R14
+	MOVL  R14, (BX)(R12*1)
+	ADDQ  $0x04, R12
+
+copy_2_qword:
+	TESTQ $0x00000008, R13
+	JZ    copy_2_test
+	MOVQ  (R11)(R12*1), R14
+	MOVQ  R14, (BX)(R12*1)
+	ADDQ  $0x08, R12
+	JMP   copy_2_test
 
 copy_2:
-	MOVUPS (R11), X0
-	MOVUPS X0, (R12)
-	ADDQ   $0x10, R11
+	MOVUPS (R11)(R12*1), X0
+	MOVUPS X0, (BX)(R12*1)
 	ADDQ   $0x10, R12
-	SUBQ   $0x10, R13
-	JHI    copy_2
-	JMP    handle_loop
+
+copy_2_test:
+	CMPQ R12, R13
+	JB   copy_2
+	ADDQ R13, BX
+	JMP  handle_loop
 
 	// Copy overlapping match
 copy_overlapping_match:
@@ -1673,45 +1913,16 @@ sequenceDecs_decodeSync_amd64_match_len_ofs_ok:
 	TESTQ AX, AX
 	JZ    check_offset
 	XORQ  R14, R14
-	TESTQ $0x00000001, AX
-	JZ    copy_1_word
-	MOVB  (R11)(R14*1), R15
-	MOVB  R15, (R10)(R14*1)
-	ADDQ  $0x01, R14
-
-copy_1_word:
-	TESTQ $0x00000002, AX
-	JZ    copy_1_dword
-	MOVW  (R11)(R14*1), R15
-	MOVW  R15, (R10)(R14*1)
-	ADDQ  $0x02, R14
-
-copy_1_dword:
-	TESTQ $0x00000004, AX
-	JZ    copy_1_qword
-	MOVL  (R11)(R14*1), R15
-	MOVL  R15, (R10)(R14*1)
-	ADDQ  $0x04, R14
-
-copy_1_qword:
-	TESTQ $0x00000008, AX
-	JZ    copy_1_test
-	MOVQ  (R11)(R14*1), R15
-	MOVQ  R15, (R10)(R14*1)
-	ADDQ  $0x08, R14
-	JMP   copy_1_test
 
 copy_1:
 	MOVUPS (R11)(R14*1), X0
 	MOVUPS X0, (R10)(R14*1)
 	ADDQ   $0x10, R14
-
-copy_1_test:
-	CMPQ R14, AX
-	JB   copy_1
-	ADDQ AX, R11
-	ADDQ AX, R10
-	ADDQ AX, R12
+	CMPQ   R14, AX
+	JB     copy_1
+	ADDQ   AX, R11
+	ADDQ   AX, R10
+	ADDQ   AX, R12
 
 	// Malformed input if seq.mo > t+len(hist) || seq.mo > s.windowSize)
 check_offset:
@@ -2044,60 +2255,55 @@ sequenceDecs_decodeSync_bmi2_fill_2_end:
 	MOVQ   CX, 24(SP)
 
 	// Fill bitreader for state updates
-	MOVQ   R12, (SP)
-	MOVQ   $0x00000808, CX
-	BEXTRQ CX, R8, R12
-	MOVQ   ctx+16(FP), CX
-	CMPQ   96(CX), $0x00
-	JZ     sequenceDecs_decodeSync_bmi2_skip_update
-
-	// Update Literal Length State
-	MOVBQZX SI, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, SI, SI
+	MOVQ    R12, (SP)
+	MOVQ    $0x00000808, CX
+	BEXTRQ  CX, R8, R12
+	MOVQ    ctx+16(FP), CX
+	CMPQ    96(CX), $0x00
+	JZ      sequenceDecs_decodeSync_bmi2_skip_update
+	LEAQ    (SI)(DI*1), R13
+	ADDQ    R8, R13
+	MOVBQZX R13, R13
 	LEAQ    (DX)(R13*1), CX
 	MOVQ    AX, R14
 	MOVQ    CX, DX
 	ROLQ    CL, R14
 	BZHIQ   R13, R14, R14
-	ADDQ    R14, SI
 
-	// Load ctx.llTable
+	// Update Offset State
+	BZHIQ  R8, R14, CX
+	SHRXQ  R8, R14, R14
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, R8, R8
+	ADDQ   CX, R8
+
+	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
-	MOVQ (CX), CX
-	MOVQ (CX)(SI*8), SI
+	MOVQ 48(CX), CX
+	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	MOVBQZX DI, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, DI, DI
-	LEAQ    (DX)(R13*1), CX
-	MOVQ    AX, R14
-	MOVQ    CX, DX
-	ROLQ    CL, R14
-	BZHIQ   R13, R14, R14
-	ADDQ    R14, DI
+	BZHIQ  DI, R14, CX
+	SHRXQ  DI, R14, R14
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, DI, DI
+	ADDQ   CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
 	MOVQ 24(CX), CX
 	MOVQ (CX)(DI*8), DI
 
-	// Update Offset State
-	MOVBQZX R8, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, R8, R8
-	LEAQ    (DX)(R13*1), CX
-	MOVQ    AX, R14
-	MOVQ    CX, DX
-	ROLQ    CL, R14
-	BZHIQ   R13, R14, R14
-	ADDQ    R14, R8
+	// Update Literal Length State
+	BZHIQ  SI, R14, CX
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, SI, SI
+	ADDQ   CX, SI
 
-	// Load ctx.ofTable
+	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
-	MOVQ 48(CX), CX
-	MOVQ (CX)(R8*8), R8
+	MOVQ (CX), CX
+	MOVQ (CX)(SI*8), SI
 
 sequenceDecs_decodeSync_bmi2_skip_update:
 	// Adjust offset
@@ -2180,45 +2386,16 @@ sequenceDecs_decodeSync_bmi2_match_len_ofs_ok:
 	TESTQ CX, CX
 	JZ    check_offset
 	XORQ  R14, R14
-	TESTQ $0x00000001, CX
-	JZ    copy_1_word
-	MOVB  (R10)(R14*1), R15
-	MOVB  R15, (R9)(R14*1)
-	ADDQ  $0x01, R14
-
-copy_1_word:
-	TESTQ $0x00000002, CX
-	JZ    copy_1_dword
-	MOVW  (R10)(R14*1), R15
-	MOVW  R15, (R9)(R14*1)
-	ADDQ  $0x02, R14
-
-copy_1_dword:
-	TESTQ $0x00000004, CX
-	JZ    copy_1_qword
-	MOVL  (R10)(R14*1), R15
-	MOVL  R15, (R9)(R14*1)
-	ADDQ  $0x04, R14
-
-copy_1_qword:
-	TESTQ $0x00000008, CX
-	JZ    copy_1_test
-	MOVQ  (R10)(R14*1), R15
-	MOVQ  R15, (R9)(R14*1)
-	ADDQ  $0x08, R14
-	JMP   copy_1_test
 
 copy_1:
 	MOVUPS (R10)(R14*1), X0
 	MOVUPS X0, (R9)(R14*1)
 	ADDQ   $0x10, R14
-
-copy_1_test:
-	CMPQ R14, CX
-	JB   copy_1
-	ADDQ CX, R10
-	ADDQ CX, R9
-	ADDQ CX, R11
+	CMPQ   R14, CX
+	JB     copy_1
+	ADDQ   CX, R10
+	ADDQ   CX, R9
+	ADDQ   CX, R11
 
 	// Malformed input if seq.mo > t+len(hist) || seq.mo > s.windowSize)
 check_offset:
@@ -3108,60 +3285,55 @@ sequenceDecs_decodeSync_safe_bmi2_fill_2_end:
 	MOVQ   CX, 24(SP)
 
 	// Fill bitreader for state updates
-	MOVQ   R12, (SP)
-	MOVQ   $0x00000808, CX
-	BEXTRQ CX, R8, R12
-	MOVQ   ctx+16(FP), CX
-	CMPQ   96(CX), $0x00
-	JZ     sequenceDecs_decodeSync_safe_bmi2_skip_update
-
-	// Update Literal Length State
-	MOVBQZX SI, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, SI, SI
+	MOVQ    R12, (SP)
+	MOVQ    $0x00000808, CX
+	BEXTRQ  CX, R8, R12
+	MOVQ    ctx+16(FP), CX
+	CMPQ    96(CX), $0x00
+	JZ      sequenceDecs_decodeSync_safe_bmi2_skip_update
+	LEAQ    (SI)(DI*1), R13
+	ADDQ    R8, R13
+	MOVBQZX R13, R13
 	LEAQ    (DX)(R13*1), CX
 	MOVQ    AX, R14
 	MOVQ    CX, DX
 	ROLQ    CL, R14
 	BZHIQ   R13, R14, R14
-	ADDQ    R14, SI
 
-	// Load ctx.llTable
+	// Update Offset State
+	BZHIQ  R8, R14, CX
+	SHRXQ  R8, R14, R14
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, R8, R8
+	ADDQ   CX, R8
+
+	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
-	MOVQ (CX), CX
-	MOVQ (CX)(SI*8), SI
+	MOVQ 48(CX), CX
+	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	MOVBQZX DI, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, DI, DI
-	LEAQ    (DX)(R13*1), CX
-	MOVQ    AX, R14
-	MOVQ    CX, DX
-	ROLQ    CL, R14
-	BZHIQ   R13, R14, R14
-	ADDQ    R14, DI
+	BZHIQ  DI, R14, CX
+	SHRXQ  DI, R14, R14
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, DI, DI
+	ADDQ   CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
 	MOVQ 24(CX), CX
 	MOVQ (CX)(DI*8), DI
 
-	// Update Offset State
-	MOVBQZX R8, R13
-	MOVQ    $0x00001010, CX
-	BEXTRQ  CX, R8, R8
-	LEAQ    (DX)(R13*1), CX
-	MOVQ    AX, R14
-	MOVQ    CX, DX
-	ROLQ    CL, R14
-	BZHIQ   R13, R14, R14
-	ADDQ    R14, R8
+	// Update Literal Length State
+	BZHIQ  SI, R14, CX
+	MOVQ   $0x00001010, R13
+	BEXTRQ R13, SI, SI
+	ADDQ   CX, SI
 
-	// Load ctx.ofTable
+	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
-	MOVQ 48(CX), CX
-	MOVQ (CX)(R8*8), R8
+	MOVQ (CX), CX
+	MOVQ (CX)(SI*8), SI
 
 sequenceDecs_decodeSync_safe_bmi2_skip_update:
 	// Adjust offset
diff --git a/vendor/github.com/klauspost/compress/zstd/zip.go b/vendor/github.com/klauspost/compress/zstd/zip.go
index b53f606a1..29c15c8c4 100644
--- a/vendor/github.com/klauspost/compress/zstd/zip.go
+++ b/vendor/github.com/klauspost/compress/zstd/zip.go
@@ -18,7 +18,14 @@ const ZipMethodWinZip = 93
 // See https://pkware.cachefly.net/webdocs/APPNOTE/APPNOTE-6.3.9.TXT
 const ZipMethodPKWare = 20
 
-var zipReaderPool sync.Pool
+// zipReaderPool is the default reader pool.
+var zipReaderPool = sync.Pool{New: func() interface{} {
+	z, err := NewReader(nil, WithDecoderLowmem(true), WithDecoderMaxWindow(128<<20), WithDecoderConcurrency(1))
+	if err != nil {
+		panic(err)
+	}
+	return z
+}}
 
 // newZipReader creates a pooled zip decompressor.
 func newZipReader(opts ...DOption) func(r io.Reader) io.ReadCloser {
diff --git a/vendor/github.com/klauspost/compress/zstd/zstd.go b/vendor/github.com/klauspost/compress/zstd/zstd.go
index c1c90b4a0..3eb3f1c82 100644
--- a/vendor/github.com/klauspost/compress/zstd/zstd.go
+++ b/vendor/github.com/klauspost/compress/zstd/zstd.go
@@ -110,17 +110,6 @@ func printf(format string, a ...interface{}) {
 	}
 }
 
-// matchLenFast does matching, but will not match the last up to 7 bytes.
-func matchLenFast(a, b []byte) int {
-	endI := len(a) & (math.MaxInt32 - 7)
-	for i := 0; i < endI; i += 8 {
-		if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-			return i + bits.TrailingZeros64(diff)>>3
-		}
-	}
-	return endI
-}
-
 // matchLen returns the maximum length.
 // a must be the shortest of the two.
 // The function also returns whether all bytes matched.