MM-40676: Bump dependencies (#19525)

https://mattermost.atlassian.net/browse/MM-40676 ```release-note NONE ```
2022-02-09 23:05:16 +05:30
--- a/vendor/github.com/klauspost/compress/.goreleaser.yml
+++ b/vendor/github.com/klauspost/compress/.goreleaser.yml
@@ -3,6 +3,7 @@
 before:
  hooks:
    - ./gen.sh
+    - go install mvdan.cc/garble@latest

 builds:
  -
@@ -31,6 +32,7 @@ builds:
      - mips64le
    goarm:
      - 7
+    gobinary: garble
  -
    id: "s2d"
    binary: s2d
@@ -57,6 +59,7 @@ builds:
      - mips64le
    goarm:
      - 7
+    gobinary: garble
  -
    id: "s2sx"
    binary: s2sx
@@ -84,6 +87,7 @@ builds:
      - mips64le
    goarm:
      - 7
+    gobinary: garble

 archives:
  -
--- a/vendor/github.com/klauspost/compress/README.md
+++ b/vendor/github.com/klauspost/compress/README.md
@@ -17,6 +17,13 @@ This package provides various compression algorithms.

 # changelog

+* Jan 11, 2022 (v1.14.1)
+	* s2: Add stream index in [#462](https://github.com/klauspost/compress/pull/462)
+	* flate: Speed and efficiency improvements in [#439](https://github.com/klauspost/compress/pull/439) [#461](https://github.com/klauspost/compress/pull/461) [#455](https://github.com/klauspost/compress/pull/455) [#452](https://github.com/klauspost/compress/pull/452) [#458](https://github.com/klauspost/compress/pull/458)
+	* zstd: Performance improvement in [#420]( https://github.com/klauspost/compress/pull/420) [#456](https://github.com/klauspost/compress/pull/456) [#437](https://github.com/klauspost/compress/pull/437) [#467](https://github.com/klauspost/compress/pull/467) [#468](https://github.com/klauspost/compress/pull/468)
+	* 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)
+
 * 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)
@@ -432,6 +439,13 @@ For more information see my blog post on [Fast Linear Time Compression](http://b

 This is implemented on Go 1.7 as "Huffman Only" mode, though not exposed for gzip.

+# Other packages
+
+Here are other packages of good quality and pure Go (no cgo wrappers or autoconverted code):
+
+* [github.com/pierrec/lz4](https://github.com/pierrec/lz4) - strong multithreaded LZ4 compression.
+* [github.com/cosnicolaou/pbzip2](https://github.com/cosnicolaou/pbzip2) - multithreaded bzip2 decompression.
+* [github.com/dsnet/compress](https://github.com/dsnet/compress) - brotli decompression, bzip2 writer.

 # license

--- a/vendor/github.com/klauspost/compress/flate/deflate.go
+++ b/vendor/github.com/klauspost/compress/flate/deflate.go
@@ -6,6 +6,7 @@
 package flate

 import (
+	"encoding/binary"
 	"fmt"
 	"io"
 	"math"
@@ -37,15 +38,17 @@ const (
 	maxMatchLength   = 258 // The longest match for the compressor
 	minOffsetSize    = 1   // The shortest offset that makes any sense

-	// The maximum number of tokens we put into a single flat block, just too
-	// stop things from getting too large.
-	maxFlateBlockTokens = 1 << 14
+	// The maximum number of tokens we will encode at the time.
+	// Smaller sizes usually creates less optimal blocks.
+	// Bigger can make context switching slow.
+	// We use this for levels 7-9, so we make it big.
+	maxFlateBlockTokens = 1 << 15
 	maxStoreBlockSize   = 65535
 	hashBits            = 17 // After 17 performance degrades
 	hashSize            = 1 << hashBits
 	hashMask            = (1 << hashBits) - 1
 	hashShift           = (hashBits + minMatchLength - 1) / minMatchLength
-	maxHashOffset       = 1 << 24
+	maxHashOffset       = 1 << 28

 	skipNever = math.MaxInt32

@@ -70,9 +73,9 @@ var levels = []compressionLevel{
 	{0, 0, 0, 0, 0, 6},
 	// Levels 7-9 use increasingly more lazy matching
 	// and increasingly stringent conditions for "good enough".
-	{8, 8, 24, 16, skipNever, 7},
-	{10, 16, 24, 64, skipNever, 8},
-	{32, 258, 258, 4096, skipNever, 9},
+	{8, 12, 16, 24, skipNever, 7},
+	{16, 30, 40, 64, skipNever, 8},
+	{32, 258, 258, 1024, skipNever, 9},
 }

 // advancedState contains state for the advanced levels, with bigger hash tables, etc.
@@ -93,8 +96,9 @@ type advancedState struct {
 	hashOffset int

 	// input window: unprocessed data is window[index:windowEnd]
-	index     int
-	hashMatch [maxMatchLength + minMatchLength]uint32
+	index          int
+	estBitsPerByte int
+	hashMatch      [maxMatchLength + minMatchLength]uint32

 	hash uint32
 	ii   uint16 // position of last match, intended to overflow to reset.
@@ -103,6 +107,7 @@ type advancedState struct {
 type compressor struct {
 	compressionLevel

+	h *huffmanEncoder
 	w *huffmanBitWriter

 	// compression algorithm
@@ -170,7 +175,8 @@ func (d *compressor) writeBlock(tok *tokens, index int, eof bool) error {
 			window = d.window[d.blockStart:index]
 		}
 		d.blockStart = index
-		d.w.writeBlock(tok, eof, window)
+		//d.w.writeBlock(tok, eof, window)
+		d.w.writeBlockDynamic(tok, eof, window, d.sync)
 		return d.w.err
 	}
 	return nil
@@ -263,7 +269,7 @@ func (d *compressor) fillWindow(b []byte) {
 // Try to find a match starting at index whose length is greater than prevSize.
 // We only look at chainCount possibilities before giving up.
 // pos = s.index, prevHead = s.chainHead-s.hashOffset, prevLength=minMatchLength-1, lookahead
-func (d *compressor) findMatch(pos int, prevHead int, prevLength int, lookahead int) (length, offset int, ok bool) {
+func (d *compressor) findMatch(pos int, prevHead int, lookahead int) (length, offset int, ok bool) {
 	minMatchLook := maxMatchLength
 	if lookahead < minMatchLook {
 		minMatchLook = lookahead
@@ -279,36 +285,75 @@ func (d *compressor) findMatch(pos int, prevHead int, prevLength int, lookahead

 	// If we've got a match that's good enough, only look in 1/4 the chain.
 	tries := d.chain
-	length = prevLength
-	if length >= d.good {
-		tries >>= 2
-	}
+	length = minMatchLength - 1

 	wEnd := win[pos+length]
 	wPos := win[pos:]
 	minIndex := pos - windowSize
+	if minIndex < 0 {
+		minIndex = 0
+	}
+	offset = 0

+	cGain := 0
+	if d.chain < 100 {
+		for i := prevHead; tries > 0; tries-- {
+			if wEnd == win[i+length] {
+				n := matchLen(win[i:i+minMatchLook], wPos)
+				if n > length {
+					length = n
+					offset = pos - i
+					ok = true
+					if n >= nice {
+						// The match is good enough that we don't try to find a better one.
+						break
+					}
+					wEnd = win[pos+n]
+				}
+			}
+			if i <= minIndex {
+				// hashPrev[i & windowMask] has already been overwritten, so stop now.
+				break
+			}
+			i = int(d.state.hashPrev[i&windowMask]) - d.state.hashOffset
+			if i < minIndex {
+				break
+			}
+		}
+		return
+	}
+
+	// Some like it higher (CSV), some like it lower (JSON)
+	const baseCost = 6
+	// Base is 4 bytes at with an additional cost.
+	// Matches must be better than this.
 	for i := prevHead; tries > 0; tries-- {
 		if wEnd == win[i+length] {
 			n := matchLen(win[i:i+minMatchLook], wPos)
+			if n > length {
+				// Calculate gain. Estimate
+				newGain := d.h.bitLengthRaw(wPos[:n]) - int(offsetExtraBits[offsetCode(uint32(pos-i))]) - baseCost - int(lengthExtraBits[lengthCodes[(n-3)&255]])

-			if n > length && (n > minMatchLength || pos-i <= 4096) {
-				length = n
-				offset = pos - i
-				ok = true
-				if n >= nice {
-					// The match is good enough that we don't try to find a better one.
-					break
+				//fmt.Println(n, "gain:", newGain, "prev:", cGain, "raw:", d.h.bitLengthRaw(wPos[:n]))
+				if newGain > cGain {
+					length = n
+					offset = pos - i
+					cGain = newGain
+					ok = true
+					if n >= nice {
+						// The match is good enough that we don't try to find a better one.
+						break
+					}
+					wEnd = win[pos+n]
 				}
-				wEnd = win[pos+n]
 			}
 		}
-		if i == minIndex {
+		if i <= minIndex {
 			// hashPrev[i & windowMask] has already been overwritten, so stop now.
 			break
 		}
 		i = int(d.state.hashPrev[i&windowMask]) - d.state.hashOffset
-		if i < minIndex || i < 0 {
+		if i < minIndex {
 			break
 		}
 	}
@@ -327,8 +372,7 @@ func (d *compressor) writeStoredBlock(buf []byte) error {
 // of the supplied slice.
 // The caller must ensure that len(b) >= 4.
 func hash4(b []byte) uint32 {
-	b = b[:4]
-	return hash4u(uint32(b[3])|uint32(b[2])<<8|uint32(b[1])<<16|uint32(b[0])<<24, hashBits)
+	return hash4u(binary.LittleEndian.Uint32(b), hashBits)
 }

 // bulkHash4 will compute hashes using the same
@@ -337,11 +381,12 @@ func bulkHash4(b []byte, dst []uint32) {
 	if len(b) < 4 {
 		return
 	}
-	hb := uint32(b[3]) | uint32(b[2])<<8 | uint32(b[1])<<16 | uint32(b[0])<<24
+	hb := binary.LittleEndian.Uint32(b)
+
 	dst[0] = hash4u(hb, hashBits)
 	end := len(b) - 4 + 1
 	for i := 1; i < end; i++ {
-		hb = (hb << 8) | uint32(b[i+3])
+		hb = (hb >> 8) | uint32(b[i+3])<<24
 		dst[i] = hash4u(hb, hashBits)
 	}
 }
@@ -374,10 +419,21 @@ func (d *compressor) deflateLazy() {
 	if d.windowEnd-s.index < minMatchLength+maxMatchLength && !d.sync {
 		return
 	}
+	if d.windowEnd != s.index && d.chain > 100 {
+		// Get literal huffman coder.
+		if d.h == nil {
+			d.h = newHuffmanEncoder(maxFlateBlockTokens)
+		}
+		var tmp [256]uint16
+		for _, v := range d.window[s.index:d.windowEnd] {
+			tmp[v]++
+		}
+		d.h.generate(tmp[:], 15)
+	}

 	s.maxInsertIndex = d.windowEnd - (minMatchLength - 1)
 	if s.index < s.maxInsertIndex {
-		s.hash = hash4(d.window[s.index : s.index+minMatchLength])
+		s.hash = hash4(d.window[s.index:])
 	}

 	for {
@@ -410,7 +466,7 @@ func (d *compressor) deflateLazy() {
 		}
 		if s.index < s.maxInsertIndex {
 			// Update the hash
-			s.hash = hash4(d.window[s.index : s.index+minMatchLength])
+			s.hash = hash4(d.window[s.index:])
 			ch := s.hashHead[s.hash&hashMask]
 			s.chainHead = int(ch)
 			s.hashPrev[s.index&windowMask] = ch
@@ -426,12 +482,37 @@ func (d *compressor) deflateLazy() {
 		}

 		if s.chainHead-s.hashOffset >= minIndex && lookahead > prevLength && prevLength < d.lazy {
-			if newLength, newOffset, ok := d.findMatch(s.index, s.chainHead-s.hashOffset, minMatchLength-1, lookahead); ok {
+			if newLength, newOffset, ok := d.findMatch(s.index, s.chainHead-s.hashOffset, lookahead); ok {
 				s.length = newLength
 				s.offset = newOffset
 			}
 		}
+
 		if prevLength >= minMatchLength && s.length <= prevLength {
+			// Check for better match at end...
+			//
+			// checkOff must be >=2 since we otherwise risk checking s.index
+			// Offset of 2 seems to yield best results.
+			const checkOff = 2
+			prevIndex := s.index - 1
+			if prevIndex+prevLength+checkOff < s.maxInsertIndex {
+				end := lookahead
+				if lookahead > maxMatchLength {
+					end = maxMatchLength
+				}
+				end += prevIndex
+				idx := prevIndex + prevLength - (4 - checkOff)
+				h := hash4(d.window[idx:])
+				ch2 := int(s.hashHead[h&hashMask]) - 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.
+					if length > prevLength {
+						prevLength = length
+						prevOffset = prevIndex - ch2
+					}
+				}
+			}
 			// There was a match at the previous step, and the current match is
 			// not better. Output the previous match.
 			d.tokens.AddMatch(uint32(prevLength-3), uint32(prevOffset-minOffsetSize))
@@ -479,6 +560,7 @@ func (d *compressor) deflateLazy() {
 				}
 				d.tokens.Reset()
 			}
+			s.ii = 0
 		} else {
 			// Reset, if we got a match this run.
 			if s.length >= minMatchLength {
@@ -498,13 +580,12 @@ func (d *compressor) deflateLazy() {

 				// If we have a long run of no matches, skip additional bytes
 				// Resets when s.ii overflows after 64KB.
-				if s.ii > 31 {
-					n := int(s.ii >> 5)
+				if n := int(s.ii) - d.chain; n > 0 {
+					n = 1 + int(n>>6)
 					for j := 0; j < n; j++ {
 						if s.index >= d.windowEnd-1 {
 							break
 						}
-
 						d.tokens.AddLiteral(d.window[s.index-1])
 						if d.tokens.n == maxFlateBlockTokens {
 							if d.err = d.writeBlock(&d.tokens, s.index, false); d.err != nil {
@@ -512,6 +593,14 @@ func (d *compressor) deflateLazy() {
 							}
 							d.tokens.Reset()
 						}
+						// Index...
+						if s.index < s.maxInsertIndex {
+							h := hash4(d.window[s.index:])
+							ch := s.hashHead[h]
+							s.chainHead = int(ch)
+							s.hashPrev[s.index&windowMask] = ch
+							s.hashHead[h] = uint32(s.index + s.hashOffset)
+						}
 						s.index++
 					}
 					// Flush last byte
@@ -611,7 +700,9 @@ func (d *compressor) write(b []byte) (n int, err error) {
 	}
 	n = len(b)
 	for len(b) > 0 {
-		d.step(d)
+		if d.windowEnd == len(d.window) || d.sync {
+			d.step(d)
+		}
 		b = b[d.fill(d, b):]
 		if d.err != nil {
 			return 0, d.err
@@ -652,13 +743,13 @@ func (d *compressor) init(w io.Writer, level int) (err error) {
 		level = 5
 		fallthrough
 	case level >= 1 && level <= 6:
-		d.w.logNewTablePenalty = 8
+		d.w.logNewTablePenalty = 7
 		d.fast = newFastEnc(level)
 		d.window = make([]byte, maxStoreBlockSize)
 		d.fill = (*compressor).fillBlock
 		d.step = (*compressor).storeFast
 	case 7 <= level && level <= 9:
-		d.w.logNewTablePenalty = 10
+		d.w.logNewTablePenalty = 8
 		d.state = &advancedState{}
 		d.compressionLevel = levels[level]
 		d.initDeflate()
--- a/vendor/github.com/klauspost/compress/flate/fast_encoder.go
+++ b/vendor/github.com/klauspost/compress/flate/fast_encoder.go
@@ -213,11 +213,9 @@ func (e *fastGen) Reset() {
 // matchLen returns the maximum length.
 // 'a' must be the shortest of the two.
 func matchLen(a, b []byte) int {
-	b = b[:len(a)]
 	var checked int

 	for len(a) >= 8 {
-		b = b[:len(a)]
 		if diff := binary.LittleEndian.Uint64(a) ^ binary.LittleEndian.Uint64(b); diff != 0 {
 			return checked + (bits.TrailingZeros64(diff) >> 3)
 		}
--- a/vendor/github.com/klauspost/compress/flate/huffman_bit_writer.go
+++ b/vendor/github.com/klauspost/compress/flate/huffman_bit_writer.go
@@ -52,18 +52,18 @@ var lengthBase = [32]uint8{
 }

 // offset code word extra bits.
-var offsetExtraBits = [64]int8{
+var offsetExtraBits = [32]int8{
 	0, 0, 0, 0, 1, 1, 2, 2, 3, 3,
 	4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
 	9, 9, 10, 10, 11, 11, 12, 12, 13, 13,
 	/* extended window */
-	14, 14, 15, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20,
+	14, 14,
 }

 var offsetCombined = [32]uint32{}

 func init() {
-	var offsetBase = [64]uint32{
+	var offsetBase = [32]uint32{
 		/* normal deflate */
 		0x000000, 0x000001, 0x000002, 0x000003, 0x000004,
 		0x000006, 0x000008, 0x00000c, 0x000010, 0x000018,
@@ -73,9 +73,7 @@ func init() {
 		0x001800, 0x002000, 0x003000, 0x004000, 0x006000,

 		/* extended window */
-		0x008000, 0x00c000, 0x010000, 0x018000, 0x020000,
-		0x030000, 0x040000, 0x060000, 0x080000, 0x0c0000,
-		0x100000, 0x180000, 0x200000, 0x300000,
+		0x008000, 0x00c000,
 	}

 	for i := range offsetCombined[:] {
@@ -155,37 +153,33 @@ func (w *huffmanBitWriter) reset(writer io.Writer) {
 	w.lastHuffMan = false
 }

-func (w *huffmanBitWriter) canReuse(t *tokens) (offsets, lits bool) {
-	offsets, lits = true, true
+func (w *huffmanBitWriter) canReuse(t *tokens) (ok bool) {
 	a := t.offHist[:offsetCodeCount]
-	b := w.offsetFreq[:len(a)]
-	for i := range a {
-		if b[i] == 0 && a[i] != 0 {
-			offsets = false
-			break
+	b := w.offsetEncoding.codes
+	b = b[:len(a)]
+	for i, v := range a {
+		if v != 0 && b[i].len == 0 {
+			return false
 		}
 	}

 	a = t.extraHist[:literalCount-256]
-	b = w.literalFreq[256:literalCount]
+	b = w.literalEncoding.codes[256:literalCount]
 	b = b[:len(a)]
-	for i := range a {
-		if b[i] == 0 && a[i] != 0 {
-			lits = false
-			break
+	for i, v := range a {
+		if v != 0 && b[i].len == 0 {
+			return false
 		}
 	}
-	if lits {
-		a = t.litHist[:]
-		b = w.literalFreq[:len(a)]
-		for i := range a {
-			if b[i] == 0 && a[i] != 0 {
-				lits = false
-				break
-			}
+
+	a = t.litHist[:256]
+	b = w.literalEncoding.codes[:len(a)]
+	for i, v := range a {
+		if v != 0 && b[i].len == 0 {
+			return false
 		}
 	}
-	return
+	return true
 }

 func (w *huffmanBitWriter) flush() {
@@ -222,7 +216,7 @@ func (w *huffmanBitWriter) write(b []byte) {
 }

 func (w *huffmanBitWriter) writeBits(b int32, nb uint16) {
-	w.bits |= uint64(b) << w.nbits
+	w.bits |= uint64(b) << (w.nbits & 63)
 	w.nbits += nb
 	if w.nbits >= 48 {
 		w.writeOutBits()
@@ -423,7 +417,7 @@ func (w *huffmanBitWriter) storedSize(in []byte) (int, bool) {

 func (w *huffmanBitWriter) writeCode(c hcode) {
 	// The function does not get inlined if we "& 63" the shift.
-	w.bits |= uint64(c.code) << w.nbits
+	w.bits |= uint64(c.code) << (w.nbits & 63)
 	w.nbits += c.len
 	if w.nbits >= 48 {
 		w.writeOutBits()
@@ -566,7 +560,7 @@ func (w *huffmanBitWriter) writeBlock(tokens *tokens, eof bool, input []byte) {
 		w.lastHeader = 0
 	}
 	numLiterals, numOffsets := w.indexTokens(tokens, false)
-	w.generate(tokens)
+	w.generate()
 	var extraBits int
 	storedSize, storable := w.storedSize(input)
 	if storable {
@@ -595,7 +589,7 @@ func (w *huffmanBitWriter) writeBlock(tokens *tokens, eof bool, input []byte) {
 	}

 	// Stored bytes?
-	if storable && storedSize < size {
+	if storable && storedSize <= size {
 		w.writeStoredHeader(len(input), eof)
 		w.writeBytes(input)
 		return
@@ -634,22 +628,39 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 		w.lastHeader = 0
 		w.lastHuffMan = false
 	}
-	if !sync {
-		tokens.Fill()
+
+	// fillReuse enables filling of empty values.
+	// This will make encodings always reusable without testing.
+	// However, this does not appear to benefit on most cases.
+	const fillReuse = false
+
+	// Check if we can reuse...
+	if !fillReuse && w.lastHeader > 0 && !w.canReuse(tokens) {
+		w.writeCode(w.literalEncoding.codes[endBlockMarker])
+		w.lastHeader = 0
 	}
+
 	numLiterals, numOffsets := w.indexTokens(tokens, !sync)
+	extraBits := 0
+	ssize, storable := w.storedSize(input)
+
+	const usePrefs = true
+	if storable || w.lastHeader > 0 {
+		extraBits = w.extraBitSize()
+	}

 	var size int
+
 	// Check if we should reuse.
 	if w.lastHeader > 0 {
 		// Estimate size for using a new table.
 		// Use the previous header size as the best estimate.
 		newSize := w.lastHeader + tokens.EstimatedBits()
-		newSize += newSize >> w.logNewTablePenalty
+		newSize += int(w.literalEncoding.codes[endBlockMarker].len) + newSize>>w.logNewTablePenalty

 		// The estimated size is calculated as an optimal table.
 		// We add a penalty to make it more realistic and re-use a bit more.
-		reuseSize := w.dynamicReuseSize(w.literalEncoding, w.offsetEncoding) + w.extraBitSize()
+		reuseSize := w.dynamicReuseSize(w.literalEncoding, w.offsetEncoding) + extraBits

 		// Check if a new table is better.
 		if newSize < reuseSize {
@@ -660,35 +671,79 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 		} else {
 			size = reuseSize
 		}
+
+		if preSize := w.fixedSize(extraBits) + 7; usePrefs && preSize < size {
+			// Check if we get a reasonable size decrease.
+			if storable && ssize <= size {
+				w.writeStoredHeader(len(input), eof)
+				w.writeBytes(input)
+				return
+			}
+			w.writeFixedHeader(eof)
+			if !sync {
+				tokens.AddEOB()
+			}
+			w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
+			return
+		}
 		// Check if we get a reasonable size decrease.
-		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
+		if storable && ssize <= size {
 			w.writeStoredHeader(len(input), eof)
 			w.writeBytes(input)
-			w.lastHeader = 0
 			return
 		}
 	}

 	// We want a new block/table
 	if w.lastHeader == 0 {
-		w.generate(tokens)
+		if fillReuse && !sync {
+			w.fillTokens()
+			numLiterals, numOffsets = maxNumLit, maxNumDist
+		} else {
+			w.literalFreq[endBlockMarker] = 1
+		}
+
+		w.generate()
 		// Generate codegen and codegenFrequencies, which indicates how to encode
 		// the literalEncoding and the offsetEncoding.
 		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
 		w.codegenEncoding.generate(w.codegenFreq[:], 7)
+
 		var numCodegens int
-		size, numCodegens = w.dynamicSize(w.literalEncoding, w.offsetEncoding, w.extraBitSize())
-		// Store bytes, if we don't get a reasonable improvement.
-		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
+		if fillReuse && !sync {
+			// Reindex for accurate size...
+			w.indexTokens(tokens, true)
+		}
+		size, numCodegens = w.dynamicSize(w.literalEncoding, w.offsetEncoding, extraBits)
+
+		// Store predefined, if we don't get a reasonable improvement.
+		if preSize := w.fixedSize(extraBits); usePrefs && preSize <= size {
+			// Store bytes, if we don't get an improvement.
+			if storable && ssize <= preSize {
+				w.writeStoredHeader(len(input), eof)
+				w.writeBytes(input)
+				return
+			}
+			w.writeFixedHeader(eof)
+			if !sync {
+				tokens.AddEOB()
+			}
+			w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
+			return
+		}
+
+		if storable && ssize <= size {
+			// Store bytes, if we don't get an improvement.
 			w.writeStoredHeader(len(input), eof)
 			w.writeBytes(input)
-			w.lastHeader = 0
 			return
 		}

 		// Write Huffman table.
 		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
-		w.lastHeader, _ = w.headerSize()
+		if !sync {
+			w.lastHeader, _ = w.headerSize()
+		}
 		w.lastHuffMan = false
 	}

@@ -699,6 +754,19 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 	w.writeTokens(tokens.Slice(), w.literalEncoding.codes, w.offsetEncoding.codes)
 }

+func (w *huffmanBitWriter) fillTokens() {
+	for i, v := range w.literalFreq[:literalCount] {
+		if v == 0 {
+			w.literalFreq[i] = 1
+		}
+	}
+	for i, v := range w.offsetFreq[:offsetCodeCount] {
+		if v == 0 {
+			w.offsetFreq[i] = 1
+		}
+	}
+}
+
 // indexTokens indexes a slice of tokens, and updates
 // literalFreq and offsetFreq, and generates literalEncoding
 // and offsetEncoding.
@@ -733,7 +801,7 @@ func (w *huffmanBitWriter) indexTokens(t *tokens, filled bool) (numLiterals, num
 	return
 }

-func (w *huffmanBitWriter) generate(t *tokens) {
+func (w *huffmanBitWriter) generate() {
 	w.literalEncoding.generate(w.literalFreq[:literalCount], 15)
 	w.offsetEncoding.generate(w.offsetFreq[:offsetCodeCount], 15)
 }
@@ -768,7 +836,7 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		if t < matchType {
 			//w.writeCode(lits[t.literal()])
 			c := lits[t.literal()]
-			bits |= uint64(c.code) << nbits
+			bits |= uint64(c.code) << (nbits & 63)
 			nbits += c.len
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
@@ -796,7 +864,7 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		} else {
 			// inlined
 			c := lengths[lengthCode&31]
-			bits |= uint64(c.code) << nbits
+			bits |= uint64(c.code) << (nbits & 63)
 			nbits += c.len
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
@@ -819,7 +887,7 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		if extraLengthBits > 0 {
 			//w.writeBits(extraLength, extraLengthBits)
 			extraLength := int32(length - lengthBase[lengthCode&31])
-			bits |= uint64(extraLength) << nbits
+			bits |= uint64(extraLength) << (nbits & 63)
 			nbits += extraLengthBits
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
@@ -846,7 +914,7 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		} else {
 			// inlined
 			c := offs[offsetCode]
-			bits |= uint64(c.code) << nbits
+			bits |= uint64(c.code) << (nbits & 63)
 			nbits += c.len
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
@@ -867,7 +935,7 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		offsetComb := offsetCombined[offsetCode]
 		if offsetComb > 1<<16 {
 			//w.writeBits(extraOffset, extraOffsetBits)
-			bits |= uint64(offset&matchOffsetOnlyMask-(offsetComb&0xffff)) << nbits
+			bits |= uint64(offset-(offsetComb&0xffff)) << (nbits & 63)
 			nbits += uint16(offsetComb >> 16)
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
@@ -996,10 +1064,41 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 	encoding := w.literalEncoding.codes[:256]
 	// Go 1.16 LOVES having these on stack. At least 1.5x the speed.
 	bits, nbits, nbytes := w.bits, w.nbits, w.nbytes
+
+	// Unroll, write 3 codes/loop.
+	// Fastest number of unrolls.
+	for len(input) > 3 {
+		// We must have at least 48 bits free.
+		if nbits >= 8 {
+			n := nbits >> 3
+			binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
+			bits >>= (n * 8) & 63
+			nbits -= n * 8
+			nbytes += uint8(n)
+		}
+		if nbytes >= bufferFlushSize {
+			if w.err != nil {
+				nbytes = 0
+				return
+			}
+			_, w.err = w.writer.Write(w.bytes[:nbytes])
+			nbytes = 0
+		}
+		a, b := encoding[input[0]], encoding[input[1]]
+		bits |= uint64(a.code) << (nbits & 63)
+		bits |= uint64(b.code) << ((nbits + a.len) & 63)
+		c := encoding[input[2]]
+		nbits += b.len + a.len
+		bits |= uint64(c.code) << (nbits & 63)
+		nbits += c.len
+		input = input[3:]
+	}
+
+	// Remaining...
 	for _, t := range input {
 		// Bitwriting inlined, ~30% speedup
 		c := encoding[t]
-		bits |= uint64(c.code) << nbits
+		bits |= uint64(c.code) << (nbits & 63)
 		nbits += c.len
 		if debugDeflate {
 			count += int(c.len)
--- a/vendor/github.com/klauspost/compress/flate/huffman_code.go
+++ b/vendor/github.com/klauspost/compress/flate/huffman_code.go
@@ -129,9 +129,7 @@ func (h *huffmanEncoder) bitLength(freq []uint16) int {
 func (h *huffmanEncoder) bitLengthRaw(b []byte) int {
 	var total int
 	for _, f := range b {
-		if f != 0 {
-			total += int(h.codes[f].len)
-		}
+		total += int(h.codes[f].len)
 	}
 	return total
 }
--- a/vendor/github.com/klauspost/compress/flate/inflate.go
+++ b/vendor/github.com/klauspost/compress/flate/inflate.go
@@ -328,11 +328,17 @@ func (f *decompressor) nextBlock() {
 	switch typ {
 	case 0:
 		f.dataBlock()
+		if debugDecode {
+			fmt.Println("stored block")
+		}
 	case 1:
 		// compressed, fixed Huffman tables
 		f.hl = &fixedHuffmanDecoder
 		f.hd = nil
 		f.huffmanBlockDecoder()()
+		if debugDecode {
+			fmt.Println("predefinied huffman block")
+		}
 	case 2:
 		// compressed, dynamic Huffman tables
 		if f.err = f.readHuffman(); f.err != nil {
@@ -341,6 +347,9 @@ func (f *decompressor) nextBlock() {
 		f.hl = &f.h1
 		f.hd = &f.h2
 		f.huffmanBlockDecoder()()
+		if debugDecode {
+			fmt.Println("dynamic huffman block")
+		}
 	default:
 		// 3 is reserved.
 		if debugDecode {
--- a/vendor/github.com/klauspost/compress/flate/token.go
+++ b/vendor/github.com/klauspost/compress/flate/token.go
@@ -129,11 +129,11 @@ var offsetCodes14 = [256]uint32{
 type token uint32

 type tokens struct {
-	nLits     int
 	extraHist [32]uint16  // codes 256->maxnumlit
 	offHist   [32]uint16  // offset codes
 	litHist   [256]uint16 // codes 0->255
-	n         uint16      // Must be able to contain maxStoreBlockSize
+	nFilled   int
+	n         uint16 // Must be able to contain maxStoreBlockSize
 	tokens    [maxStoreBlockSize + 1]token
 }

@@ -142,7 +142,7 @@ func (t *tokens) Reset() {
 		return
 	}
 	t.n = 0
-	t.nLits = 0
+	t.nFilled = 0
 	for i := range t.litHist[:] {
 		t.litHist[i] = 0
 	}
@@ -161,12 +161,12 @@ func (t *tokens) Fill() {
 	for i, v := range t.litHist[:] {
 		if v == 0 {
 			t.litHist[i] = 1
-			t.nLits++
+			t.nFilled++
 		}
 	}
 	for i, v := range t.extraHist[:literalCount-256] {
 		if v == 0 {
-			t.nLits++
+			t.nFilled++
 			t.extraHist[i] = 1
 		}
 	}
@@ -202,14 +202,12 @@ func emitLiteral(dst *tokens, lit []byte) {
 		dst.litHist[v]++
 	}
 	dst.n += uint16(len(lit))
-	dst.nLits += len(lit)
 }

 func (t *tokens) AddLiteral(lit byte) {
 	t.tokens[t.n] = token(lit)
 	t.litHist[lit]++
 	t.n++
-	t.nLits++
 }

 // from https://stackoverflow.com/a/28730362
@@ -230,8 +228,9 @@ func (t *tokens) EstimatedBits() int {
 	shannon := float32(0)
 	bits := int(0)
 	nMatches := 0
-	if t.nLits > 0 {
-		invTotal := 1.0 / float32(t.nLits)
+	total := int(t.n) + t.nFilled
+	if total > 0 {
+		invTotal := 1.0 / float32(total)
 		for _, v := range t.litHist[:] {
 			if v > 0 {
 				n := float32(v)
@@ -275,7 +274,6 @@ func (t *tokens) AddMatch(xlength uint32, xoffset uint32) {
 	}
 	oCode := offsetCode(xoffset)
 	xoffset |= oCode << 16
-	t.nLits++

 	t.extraHist[lengthCodes1[uint8(xlength)]]++
 	t.offHist[oCode]++
@@ -301,7 +299,6 @@ func (t *tokens) AddMatchLong(xlength int32, xoffset uint32) {
 		}
 		xlength -= xl
 		xl -= baseMatchLength
-		t.nLits++
 		t.extraHist[lengthCodes1[uint8(xl)]]++
 		t.offHist[oc]++
 		t.tokens[t.n] = token(matchType | uint32(xl)<<lengthShift | xoffset)
--- a/vendor/github.com/klauspost/compress/huff0/decompress.go
+++ b/vendor/github.com/klauspost/compress/huff0/decompress.go
@@ -20,7 +20,7 @@ type dEntrySingle struct {

 // double-symbols decoding
 type dEntryDouble struct {
-	seq   uint16
+	seq   [4]byte
 	nBits uint8
 	len   uint8
 }
@@ -753,23 +753,21 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			br[stream2].fillFast()

 			val := br[stream].peekBitsFast(d.actualTableLog)
-			v := single[val&tlMask]
-			br[stream].advance(uint8(v.entry))
-			buf[off+bufoff*stream] = uint8(v.entry >> 8)
-
 			val2 := br[stream2].peekBitsFast(d.actualTableLog)
+			v := single[val&tlMask]
 			v2 := single[val2&tlMask]
+			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
+			buf[off+bufoff*stream] = uint8(v.entry >> 8)
 			buf[off+bufoff*stream2] = uint8(v2.entry >> 8)

 			val = br[stream].peekBitsFast(d.actualTableLog)
-			v = single[val&tlMask]
-			br[stream].advance(uint8(v.entry))
-			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
-
 			val2 = br[stream2].peekBitsFast(d.actualTableLog)
+			v = single[val&tlMask]
 			v2 = single[val2&tlMask]
+			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
+			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
 			buf[off+bufoff*stream2+1] = uint8(v2.entry >> 8)
 		}

@@ -780,23 +778,21 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			br[stream2].fillFast()

 			val := br[stream].peekBitsFast(d.actualTableLog)
-			v := single[val&tlMask]
-			br[stream].advance(uint8(v.entry))
-			buf[off+bufoff*stream] = uint8(v.entry >> 8)
-
 			val2 := br[stream2].peekBitsFast(d.actualTableLog)
+			v := single[val&tlMask]
 			v2 := single[val2&tlMask]
+			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
+			buf[off+bufoff*stream] = uint8(v.entry >> 8)
 			buf[off+bufoff*stream2] = uint8(v2.entry >> 8)

 			val = br[stream].peekBitsFast(d.actualTableLog)
-			v = single[val&tlMask]
-			br[stream].advance(uint8(v.entry))
-			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
-
 			val2 = br[stream2].peekBitsFast(d.actualTableLog)
+			v = single[val&tlMask]
 			v2 = single[val2&tlMask]
+			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
+			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
 			buf[off+bufoff*stream2+1] = uint8(v2.entry >> 8)
 		}

@@ -914,7 +910,7 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 	out := dst
 	dstEvery := (dstSize + 3) / 4

-	shift := (8 - d.actualTableLog) & 7
+	shift := (56 + (8 - d.actualTableLog)) & 63

 	const tlSize = 1 << 8
 	single := d.dt.single[:tlSize]
@@ -935,79 +931,91 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			// Interleave 2 decodes.
 			const stream = 0
 			const stream2 = 1
-			br[stream].fillFast()
-			br[stream2].fillFast()
+			br1 := &br[stream]
+			br2 := &br[stream2]
+			br1.fillFast()
+			br2.fillFast()

-			v := single[br[stream].peekByteFast()>>shift].entry
+			v := single[uint8(br1.value>>shift)].entry
+			v2 := single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 := single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))
+			buf[off+bufoff*stream+3] = uint8(v >> 8)
 		}

 		{
 			const stream = 2
 			const stream2 = 3
-			br[stream].fillFast()
-			br[stream2].fillFast()
+			br1 := &br[stream]
+			br2 := &br[stream2]
+			br1.fillFast()
+			br2.fillFast()

-			v := single[br[stream].peekByteFast()>>shift].entry
+			v := single[uint8(br1.value>>shift)].entry
+			v2 := single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 := single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
 			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))
+			buf[off+bufoff*stream+3] = uint8(v >> 8)
 		}

 		off += 4
@@ -1073,7 +1081,7 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			}

 			// Read value and increment offset.
-			v := single[br.peekByteFast()>>shift].entry
+			v := single[uint8(br.value>>shift)].entry
 			nBits := uint8(v)
 			br.advance(nBits)
 			bitsLeft -= int(nBits)
@@ -1121,7 +1129,7 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 	out := dst
 	dstEvery := (dstSize + 3) / 4

-	const shift = 0
+	const shift = 56
 	const tlSize = 1 << 8
 	const tlMask = tlSize - 1
 	single := d.dt.single[:tlSize]
@@ -1145,37 +1153,41 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 			br[stream].fillFast()
 			br[stream2].fillFast()

-			v := single[br[stream].peekByteFast()>>shift].entry
+			v := single[uint8(br[stream].value>>shift)].entry
+			v2 := single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 := single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))
 		}

 		{
@@ -1184,37 +1196,41 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 			br[stream].fillFast()
 			br[stream2].fillFast()

-			v := single[br[stream].peekByteFast()>>shift].entry
+			v := single[uint8(br[stream].value>>shift)].entry
+			v2 := single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 := single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))

-			v = single[br[stream].peekByteFast()>>shift].entry
+			v = single[uint8(br[stream].value>>shift)].entry
+			v2 = single[uint8(br[stream2].value>>shift)].entry
+			br[stream].bitsRead += uint8(v)
+			br[stream].value <<= v & 63
+			br[stream2].bitsRead += uint8(v2)
+			br[stream2].value <<= v2 & 63
 			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			br[stream].advance(uint8(v))
-
-			v2 = single[br[stream2].peekByteFast()>>shift].entry
 			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			br[stream2].advance(uint8(v2))
 		}

 		off += 4
@@ -1280,7 +1296,7 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 			}

 			// Read value and increment offset.
-			v := single[br.peekByteFast()>>shift].entry
+			v := single[br.peekByteFast()].entry
 			nBits := uint8(v)
 			br.advance(nBits)
 			bitsLeft -= int(nBits)
--- a/vendor/github.com/klauspost/compress/s2/README.md
+++ b/vendor/github.com/klauspost/compress/s2/README.md
@@ -20,6 +20,7 @@ This is important, so you don't have to worry about spending CPU cycles on alrea
 * Concurrent stream compression
 * Faster decompression, even for Snappy compatible content
 * Ability to quickly skip forward in compressed stream
+* Random seeking with indexes
 * Compatible with reading Snappy compressed content
 * Smaller block size overhead on incompressible blocks
 * Block concatenation
@@ -29,8 +30,8 @@ This is important, so you don't have to worry about spending CPU cycles on alrea

 ## Drawbacks over Snappy

-* Not optimized for 32 bit systems.
-* Streams use slightly more memory due to larger blocks and concurrency (configurable).
+* Not optimized for 32 bit systems
+* Streams use slightly more memory due to larger blocks and concurrency (configurable)

 # Usage

@@ -141,7 +142,7 @@ Binaries can be downloaded on the [Releases Page](https://github.com/klauspost/c

 Installing then requires Go to be installed. To install them, use:

-`go install github.com/klauspost/compress/s2/cmd/s2c && go install github.com/klauspost/compress/s2/cmd/s2d`
+`go install github.com/klauspost/compress/s2/cmd/s2c@latest && go install github.com/klauspost/compress/s2/cmd/s2d@latest`

 To build binaries to the current folder use:

@@ -176,6 +177,8 @@ Options:
    	Compress faster, but with a minor compression loss
  -help
    	Display help
+  -index
+        Add seek index (default true)    	
  -o string
        Write output to another file. Single input file only
  -pad string
@@ -217,11 +220,15 @@ Options:
    	Display help
  -o string
        Write output to another file. Single input file only
-  -q	Don't write any output to terminal, except errors
+  -offset string
+        Start at offset. Examples: 92, 64K, 256K, 1M, 4M. Requires Index
+  -q    Don't write any output to terminal, except errors
  -rm
-    	Delete source file(s) after successful decompression
+        Delete source file(s) after successful decompression
  -safe
-    	Do not overwrite output files
+        Do not overwrite output files
+  -tail string
+        Return last of compressed file. Examples: 92, 64K, 256K, 1M, 4M. Requires Index
  -verify
    	Verify files, but do not write output                                      
 ```
@@ -633,12 +640,12 @@ Compression and speed is typically a bit better `MaxEncodedLen` is also smaller
 Comparison of [`webdevdata.org-2015-01-07-subset`](https://files.klauspost.com/compress/webdevdata.org-2015-01-07-4GB-subset.7z),
 53927 files, total input size: 4,014,735,833 bytes. amd64, single goroutine used:

-| Encoder               | Size       | MB/s   | Reduction |
-|-----------------------|------------|--------|------------
-| snappy.Encode         | 1128706759 | 725.59 | 71.89%    |
-| s2.EncodeSnappy       | 1093823291 | 899.16 | 72.75%    |
-| s2.EncodeSnappyBetter | 1001158548 | 578.49 | 75.06%    |
-| s2.EncodeSnappyBest   | 944507998  | 66.00  | 76.47%    |
+| Encoder               | Size       | MB/s       | Reduction |
+|-----------------------|------------|------------|------------
+| snappy.Encode         | 1128706759 | 725.59     | 71.89%    |
+| s2.EncodeSnappy       | 1093823291 | **899.16** | 72.75%    |
+| s2.EncodeSnappyBetter | 1001158548 | 578.49     | 75.06%    |
+| s2.EncodeSnappyBest   | 944507998  | 66.00      | **76.47%**|

 ## Streams

@@ -649,11 +656,11 @@ Comparison of different streams, AMD Ryzen 3950x, 16 cores. Size and throughput:

 | File                        | snappy.NewWriter         | S2 Snappy                 | S2 Snappy, Better        | S2 Snappy, Best         |
 |-----------------------------|--------------------------|---------------------------|--------------------------|-------------------------|
-| nyc-taxi-data-10M.csv       | 1316042016 - 517.54MB/s  | 1307003093 - 8406.29MB/s  | 1174534014 - 4984.35MB/s | 1115904679 - 177.81MB/s |
-| enwik10                     | 5088294643 - 433.45MB/s  | 5175840939 - 8454.52MB/s  | 4560784526 - 4403.10MB/s | 4340299103 - 159.71MB/s |
-| 10gb.tar                    | 6056946612 - 703.25MB/s  | 6208571995 - 9035.75MB/s  | 5741646126 - 2402.08MB/s | 5548973895 - 171.17MB/s |
-| github-june-2days-2019.json | 1525176492 - 908.11MB/s  | 1476519054 - 12625.93MB/s | 1400547532 - 6163.61MB/s | 1321887137 - 200.71MB/s |
-| consensus.db.10gb           | 5412897703 - 1054.38MB/s | 5354073487 - 12634.82MB/s | 5335069899 - 2472.23MB/s | 5201000954 - 166.32MB/s |
+| nyc-taxi-data-10M.csv       | 1316042016 - 539.47MB/s  | 1307003093 - 10132.73MB/s | 1174534014 - 5002.44MB/s | 1115904679 - 177.97MB/s |
+| enwik10 (xml)               | 5088294643 - 451.13MB/s  | 5175840939 -  9440.69MB/s | 4560784526 - 4487.21MB/s | 4340299103 - 158.92MB/s |
+| 10gb.tar (mixed)            | 6056946612 - 729.73MB/s  | 6208571995 -  9978.05MB/s | 5741646126 - 4919.98MB/s | 5548973895 - 180.44MB/s |
+| github-june-2days-2019.json | 1525176492 - 933.00MB/s  | 1476519054 - 13150.12MB/s | 1400547532 - 5803.40MB/s | 1321887137 - 204.29MB/s |
+| consensus.db.10gb (db)      | 5412897703 - 1102.14MB/s | 5354073487 - 13562.91MB/s | 5335069899 - 5294.73MB/s | 5201000954 - 175.72MB/s |

 # Decompression

@@ -679,7 +686,220 @@ The 10 byte 'stream identifier' of the second stream can optionally be stripped,

 Blocks can be concatenated using the `ConcatBlocks` function.

-Snappy blocks/streams can safely be concatenated with S2 blocks and streams. 
+Snappy blocks/streams can safely be concatenated with S2 blocks and streams.
+Streams with indexes (see below) will currently not work on concatenated streams.
+
+# Stream Seek Index
+
+S2 and Snappy streams can have indexes. These indexes will allow random seeking within the compressed data.
+
+The index can either be appended to the stream as a skippable block or returned for separate storage.
+
+When the index is appended to a stream it will be skipped by regular decoders, 
+so the output remains compatible with other decoders. 
+
+## Creating an Index
+
+To automatically add an index to a stream, add `WriterAddIndex()` option to your writer.
+Then the index will be added to the stream when `Close()` is called.
+
+```
+	// Add Index to stream...
+	enc := s2.NewWriter(w, s2.WriterAddIndex())
+	io.Copy(enc, r)
+	enc.Close()
+```
+
+If you want to store the index separately, you can use `CloseIndex()` instead of the regular `Close()`.
+This will return the index. Note that `CloseIndex()` should only be called once, and you shouldn't call `Close()`.
+
+```
+	// Get index for separate storage... 
+	enc := s2.NewWriter(w)
+	io.Copy(enc, r)
+	index, err := enc.CloseIndex()
+```
+
+The `index` can then be used needing to read from the stream. 
+This means the index can be used without needing to seek to the end of the stream 
+or for manually forwarding streams. See below.
+
+Finally, an existing S2/Snappy stream can be indexed using the `s2.IndexStream(r io.Reader)` function.
+
+## Using Indexes
+
+To use indexes there is a `ReadSeeker(random bool, index []byte) (*ReadSeeker, error)` function available.
+
+Calling ReadSeeker will return an [io.ReadSeeker](https://pkg.go.dev/io#ReadSeeker) compatible version of the reader.
+
+If 'random' is specified the returned io.Seeker can be used for random seeking, otherwise only forward seeking is supported.
+Enabling random seeking requires the original input to support the [io.Seeker](https://pkg.go.dev/io#Seeker) interface.
+
+```
+	dec := s2.NewReader(r)
+	rs, err := dec.ReadSeeker(false, nil)
+	rs.Seek(wantOffset, io.SeekStart)	
+```
+
+Get a seeker to seek forward. Since no index is provided, the index is read from the stream.
+This requires that an index was added and that `r` supports the [io.Seeker](https://pkg.go.dev/io#Seeker) interface.
+
+A custom index can be specified which will be used if supplied.
+When using a custom index, it will not be read from the input stream.
+
+```
+	dec := s2.NewReader(r)
+	rs, err := dec.ReadSeeker(false, index)
+	rs.Seek(wantOffset, io.SeekStart)	
+```
+
+This will read the index from `index`. Since we specify non-random (forward only) seeking `r` does not have to be an io.Seeker
+
+```
+	dec := s2.NewReader(r)
+	rs, err := dec.ReadSeeker(true, index)
+	rs.Seek(wantOffset, io.SeekStart)	
+```
+
+Finally, since we specify that we want to do random seeking `r` must be an io.Seeker. 
+
+The returned [ReadSeeker](https://pkg.go.dev/github.com/klauspost/compress/s2#ReadSeeker) contains a shallow reference to the existing Reader,
+meaning changes performed to one is reflected in the other.
+
+To check if a stream contains an index at the end, the `(*Index).LoadStream(rs io.ReadSeeker) error` can be used.
+
+## Manually Forwarding Streams
+
+Indexes can also be read outside the decoder using the [Index](https://pkg.go.dev/github.com/klauspost/compress/s2#Index) type.
+This can be used for parsing indexes, either separate or in streams.
+
+In some cases it may not be possible to serve a seekable stream.
+This can for instance be an HTTP stream, where the Range request 
+is sent at the start of the stream. 
+
+With a little bit of extra code it is still possible to use indexes
+to forward to specific offset with a single forward skip. 
+
+It is possible to load the index manually like this: 
+```
+	var index s2.Index
+	_, err = index.Load(idxBytes)
+```
+
+This can be used to figure out how much to offset the compressed stream:
+
+```
+	compressedOffset, uncompressedOffset, err := index.Find(wantOffset)
+```
+
+The `compressedOffset` is the number of bytes that should be skipped 
+from the beginning of the compressed file.
+
+The `uncompressedOffset` will then be offset of the uncompressed bytes returned
+when decoding from that position. This will always be <= wantOffset.
+
+When creating a decoder it must be specified that it should *not* expect a stream identifier
+at the beginning of the stream. Assuming the io.Reader `r` has been forwarded to `compressedOffset`
+we create the decoder like this:
+
+```
+	dec := s2.NewReader(r, s2.ReaderIgnoreStreamIdentifier())
+```
+
+We are not completely done. We still need to forward the stream the uncompressed bytes we didn't want.
+This is done using the regular "Skip" function:
+
+```
+	err = dec.Skip(wantOffset - uncompressedOffset)
+```
+
+This will ensure that we are at exactly the offset we want, and reading from `dec` will start at the requested offset.
+
+## Index Format:
+
+Each block is structured as a snappy skippable block, with the chunk ID 0x99.
+
+The block can be read from the front, but contains information so it can be read from the back as well.
+
+Numbers are stored as fixed size little endian values or [zigzag encoded](https://developers.google.com/protocol-buffers/docs/encoding#signed_integers) [base 128 varints](https://developers.google.com/protocol-buffers/docs/encoding), 
+with un-encoded value length of 64 bits, unless other limits are specified. 
+
+| Content                                                                   | Format                                                                                                                      |
+|---------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------|
+| ID, `[1]byte`                                                           | Always 0x99.                                                                                                                  |
+| Data Length, `[3]byte`                                                  | 3 byte little-endian length of the chunk in bytes, following this.                                                            |
+| Header `[6]byte`                                                        | Header, must be `[115, 50, 105, 100, 120, 0]` or in text: "s2idx\x00".                                                        |
+| UncompressedSize, Varint                                                | Total Uncompressed size.                                                                                                      |
+| CompressedSize, Varint                                                  | Total Compressed size if known. Should be -1 if unknown.                                                                      |
+| EstBlockSize, Varint                                                    | Block Size, used for guessing uncompressed offsets. Must be >= 0.                                                             |
+| Entries, Varint                                                         | Number of Entries in index, must be < 65536 and >=0.                                                                          |
+| HasUncompressedOffsets `byte`                                           | 0 if no uncompressed offsets are present, 1 if present. Other values are invalid.                                             |
+| UncompressedOffsets, [Entries]VarInt                                    | Uncompressed offsets. See below how to decode.                                                                                |
+| CompressedOffsets, [Entries]VarInt                                      | Compressed offsets. See below how to decode.                                                                                  |
+| Block Size, `[4]byte`                                                   | Little Endian total encoded size (including header and trailer). Can be used for searching backwards to start of block.       |
+| Trailer `[6]byte`                                                       | Trailer, must be `[0, 120, 100, 105, 50, 115]` or in text: "\x00xdi2s". Can be used for identifying block from end of stream. |
+
+For regular streams the uncompressed offsets are fully predictable,
+so `HasUncompressedOffsets` allows to specify that compressed blocks all have 
+exactly `EstBlockSize` bytes of uncompressed content.
+
+Entries *must* be in order, starting with the lowest offset, 
+and there *must* be no uncompressed offset duplicates.  
+Entries *may* point to the start of a skippable block, 
+but it is then not allowed to also have an entry for the next block since 
+that would give an uncompressed offset duplicate.
+
+There is no requirement for all blocks to be represented in the index. 
+In fact there is a maximum of 65536 block entries in an index.
+
+The writer can use any method to reduce the number of entries.
+An implicit block start at 0,0 can be assumed.
+
+### Decoding entries:
+
+```
+// Read Uncompressed entries.
+// Each assumes EstBlockSize delta from previous.
+for each entry {
+    uOff = 0
+    if HasUncompressedOffsets == 1 {
+        uOff = ReadVarInt // Read value from stream
+    }
+   
+    // Except for the first entry, use previous values.
+    if entryNum == 0 {
+        entry[entryNum].UncompressedOffset = uOff
+        continue
+    }
+    
+    // Uncompressed uses previous offset and adds EstBlockSize
+    entry[entryNum].UncompressedOffset = entry[entryNum-1].UncompressedOffset + EstBlockSize
+}
+
+
+// Guess that the first block will be 50% of uncompressed size.
+// Integer truncating division must be used.
+CompressGuess := EstBlockSize / 2
+
+// Read Compressed entries.
+// Each assumes CompressGuess delta from previous.
+// CompressGuess is adjusted for each value.
+for each entry {
+    cOff = ReadVarInt // Read value from stream
+    
+    // Except for the first entry, use previous values.
+    if entryNum == 0 {
+        entry[entryNum].CompressedOffset = cOff
+        continue
+    }
+    
+    // Compressed uses previous and our estimate.
+    entry[entryNum].CompressedOffset = entry[entryNum-1].CompressedOffset + CompressGuess + cOff
+        
+     // Adjust compressed offset for next loop, integer truncating division must be used. 
+     CompressGuess += cOff/2               
+}
+```

 # Format Extensions

--- a/vendor/github.com/klauspost/compress/s2/decode.go
+++ b/vendor/github.com/klauspost/compress/s2/decode.go
@@ -8,7 +8,9 @@ package s2
 import (
 	"encoding/binary"
 	"errors"
+	"fmt"
 	"io"
+	"io/ioutil"
 )

 var (
@@ -22,6 +24,16 @@ var (
 	ErrUnsupported = errors.New("s2: unsupported input")
 )

+// ErrCantSeek is returned if the stream cannot be seeked.
+type ErrCantSeek struct {
+	Reason string
+}
+
+// Error returns the error as string.
+func (e ErrCantSeek) Error() string {
+	return fmt.Sprintf("s2: Can't seek because %s", e.Reason)
+}
+
 // DecodedLen returns the length of the decoded block.
 func DecodedLen(src []byte) (int, error) {
 	v, _, err := decodedLen(src)
@@ -88,6 +100,7 @@ func NewReader(r io.Reader, opts ...ReaderOption) *Reader {
 	} else {
 		nr.buf = make([]byte, MaxEncodedLen(defaultBlockSize)+checksumSize)
 	}
+	nr.readHeader = nr.ignoreStreamID
 	nr.paramsOK = true
 	return &nr
 }
@@ -131,12 +144,41 @@ func ReaderAllocBlock(blockSize int) ReaderOption {
 	}
 }

+// ReaderIgnoreStreamIdentifier will make the reader skip the expected
+// stream identifier at the beginning of the stream.
+// This can be used when serving a stream that has been forwarded to a specific point.
+func ReaderIgnoreStreamIdentifier() ReaderOption {
+	return func(r *Reader) error {
+		r.ignoreStreamID = true
+		return nil
+	}
+}
+
+// ReaderSkippableCB will register a callback for chuncks with the specified ID.
+// ID must be a Reserved skippable chunks ID, 0x80-0xfd (inclusive).
+// For each chunk with the ID, the callback is called with the content.
+// Any returned non-nil error will abort decompression.
+// Only one callback per ID is supported, latest sent will be used.
+func ReaderSkippableCB(id uint8, fn func(r io.Reader) error) ReaderOption {
+	return func(r *Reader) error {
+		if id < 0x80 || id > 0xfd {
+			return fmt.Errorf("ReaderSkippableCB: Invalid id provided, must be 0x80-0xfd (inclusive)")
+		}
+		r.skippableCB[id] = fn
+		return nil
+	}
+}
+
 // Reader is an io.Reader that can read Snappy-compressed bytes.
 type Reader struct {
-	r       io.Reader
-	err     error
-	decoded []byte
-	buf     []byte
+	r           io.Reader
+	err         error
+	decoded     []byte
+	buf         []byte
+	skippableCB [0x80]func(r io.Reader) error
+	blockStart  int64 // Uncompressed offset at start of current.
+	index       *Index
+
 	// decoded[i:j] contains decoded bytes that have not yet been passed on.
 	i, j int
 	// maximum block size allowed.
@@ -144,10 +186,11 @@ type Reader struct {
 	// maximum expected buffer size.
 	maxBufSize int
 	// alloc a buffer this size if > 0.
-	lazyBuf     int
-	readHeader  bool
-	paramsOK    bool
-	snappyFrame bool
+	lazyBuf        int
+	readHeader     bool
+	paramsOK       bool
+	snappyFrame    bool
+	ignoreStreamID bool
 }

 // ensureBufferSize will ensure that the buffer can take at least n bytes.
@@ -172,11 +215,12 @@ func (r *Reader) Reset(reader io.Reader) {
 	if !r.paramsOK {
 		return
 	}
+	r.index = nil
 	r.r = reader
 	r.err = nil
 	r.i = 0
 	r.j = 0
-	r.readHeader = false
+	r.readHeader = r.ignoreStreamID
 }

 func (r *Reader) readFull(p []byte, allowEOF bool) (ok bool) {
@@ -189,11 +233,24 @@ func (r *Reader) readFull(p []byte, allowEOF bool) (ok bool) {
 	return true
 }

-// skipN will skip n bytes.
+// skippable will skip n bytes.
 // If the supplied reader supports seeking that is used.
 // tmp is used as a temporary buffer for reading.
 // The supplied slice does not need to be the size of the read.
-func (r *Reader) skipN(tmp []byte, n int, allowEOF bool) (ok bool) {
+func (r *Reader) skippable(tmp []byte, n int, allowEOF bool, id uint8) (ok bool) {
+	if id < 0x80 {
+		r.err = fmt.Errorf("interbal error: skippable id < 0x80")
+		return false
+	}
+	if fn := r.skippableCB[id-0x80]; fn != nil {
+		rd := io.LimitReader(r.r, int64(n))
+		r.err = fn(rd)
+		if r.err != nil {
+			return false
+		}
+		_, r.err = io.CopyBuffer(ioutil.Discard, rd, tmp)
+		return r.err == nil
+	}
 	if rs, ok := r.r.(io.ReadSeeker); ok {
 		_, err := rs.Seek(int64(n), io.SeekCurrent)
 		if err == nil {
@@ -247,6 +304,7 @@ func (r *Reader) Read(p []byte) (int, error) {
 		// https://github.com/google/snappy/blob/master/framing_format.txt
 		switch chunkType {
 		case chunkTypeCompressedData:
+			r.blockStart += int64(r.j)
 			// Section 4.2. Compressed data (chunk type 0x00).
 			if chunkLen < checksumSize {
 				r.err = ErrCorrupt
@@ -294,6 +352,7 @@ func (r *Reader) Read(p []byte) (int, error) {
 			continue

 		case chunkTypeUncompressedData:
+			r.blockStart += int64(r.j)
 			// Section 4.3. Uncompressed data (chunk type 0x01).
 			if chunkLen < checksumSize {
 				r.err = ErrCorrupt
@@ -357,17 +416,20 @@ func (r *Reader) Read(p []byte) (int, error) {

 		if chunkType <= 0x7f {
 			// Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f).
+			// fmt.Printf("ERR chunktype: 0x%x\n", chunkType)
 			r.err = ErrUnsupported
 			return 0, r.err
 		}
 		// Section 4.4 Padding (chunk type 0xfe).
 		// Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd).
-		if chunkLen > maxBlockSize {
+		if chunkLen > maxChunkSize {
+			// fmt.Printf("ERR chunkLen: 0x%x\n", chunkLen)
 			r.err = ErrUnsupported
 			return 0, r.err
 		}

-		if !r.skipN(r.buf, chunkLen, false) {
+		// fmt.Printf("skippable: ID: 0x%x, len: 0x%x\n", chunkType, chunkLen)
+		if !r.skippable(r.buf, chunkLen, false, chunkType) {
 			return 0, r.err
 		}
 	}
@@ -396,7 +458,7 @@ func (r *Reader) Skip(n int64) error {
 				return nil
 			}
 			n -= int64(r.j - r.i)
-			r.i, r.j = 0, 0
+			r.i = r.j
 		}

 		// Buffer empty; read blocks until we have content.
@@ -420,6 +482,7 @@ func (r *Reader) Skip(n int64) error {
 		// https://github.com/google/snappy/blob/master/framing_format.txt
 		switch chunkType {
 		case chunkTypeCompressedData:
+			r.blockStart += int64(r.j)
 			// Section 4.2. Compressed data (chunk type 0x00).
 			if chunkLen < checksumSize {
 				r.err = ErrCorrupt
@@ -468,6 +531,7 @@ func (r *Reader) Skip(n int64) error {
 			r.i, r.j = 0, dLen
 			continue
 		case chunkTypeUncompressedData:
+			r.blockStart += int64(r.j)
 			// Section 4.3. Uncompressed data (chunk type 0x01).
 			if chunkLen < checksumSize {
 				r.err = ErrCorrupt
@@ -528,19 +592,138 @@ func (r *Reader) Skip(n int64) error {
 			r.err = ErrUnsupported
 			return r.err
 		}
-		if chunkLen > maxBlockSize {
+		if chunkLen > maxChunkSize {
 			r.err = ErrUnsupported
 			return r.err
 		}
 		// Section 4.4 Padding (chunk type 0xfe).
 		// Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd).
-		if !r.skipN(r.buf, chunkLen, false) {
+		if !r.skippable(r.buf, chunkLen, false, chunkType) {
 			return r.err
 		}
 	}
 	return nil
 }

+// ReadSeeker provides random or forward seeking in compressed content.
+// See Reader.ReadSeeker
+type ReadSeeker struct {
+	*Reader
+}
+
+// ReadSeeker will return an io.ReadSeeker compatible version of the reader.
+// If 'random' is specified the returned io.Seeker can be used for
+// random seeking, otherwise only forward seeking is supported.
+// Enabling random seeking requires the original input to support
+// the io.Seeker interface.
+// A custom index can be specified which will be used if supplied.
+// When using a custom index, it will not be read from the input stream.
+// The returned ReadSeeker contains a shallow reference to the existing Reader,
+// meaning changes performed to one is reflected in the other.
+func (r *Reader) ReadSeeker(random bool, index []byte) (*ReadSeeker, error) {
+	// Read index if provided.
+	if len(index) != 0 {
+		if r.index == nil {
+			r.index = &Index{}
+		}
+		if _, err := r.index.Load(index); err != nil {
+			return nil, ErrCantSeek{Reason: "loading index returned: " + err.Error()}
+		}
+	}
+
+	// Check if input is seekable
+	rs, ok := r.r.(io.ReadSeeker)
+	if !ok {
+		if !random {
+			return &ReadSeeker{Reader: r}, nil
+		}
+		return nil, ErrCantSeek{Reason: "input stream isn't seekable"}
+	}
+
+	if r.index != nil {
+		// Seekable and index, ok...
+		return &ReadSeeker{Reader: r}, nil
+	}
+
+	// Load from stream.
+	r.index = &Index{}
+
+	// Read current position.
+	pos, err := rs.Seek(0, io.SeekCurrent)
+	if err != nil {
+		return nil, ErrCantSeek{Reason: "seeking input returned: " + err.Error()}
+	}
+	err = r.index.LoadStream(rs)
+	if err != nil {
+		if err == ErrUnsupported {
+			return nil, ErrCantSeek{Reason: "input stream does not contain an index"}
+		}
+		return nil, ErrCantSeek{Reason: "reading index returned: " + err.Error()}
+	}
+
+	// reset position.
+	_, err = rs.Seek(pos, io.SeekStart)
+	if err != nil {
+		return nil, ErrCantSeek{Reason: "seeking input returned: " + err.Error()}
+	}
+	return &ReadSeeker{Reader: r}, nil
+}
+
+// Seek allows seeking in compressed data.
+func (r *ReadSeeker) Seek(offset int64, whence int) (int64, error) {
+	if r.err != nil {
+		return 0, r.err
+	}
+	if offset == 0 && whence == io.SeekCurrent {
+		return r.blockStart + int64(r.i), nil
+	}
+	if !r.readHeader {
+		// Make sure we read the header.
+		_, r.err = r.Read([]byte{})
+	}
+	rs, ok := r.r.(io.ReadSeeker)
+	if r.index == nil || !ok {
+		if whence == io.SeekCurrent && offset >= 0 {
+			err := r.Skip(offset)
+			return r.blockStart + int64(r.i), err
+		}
+		if whence == io.SeekStart && offset >= r.blockStart+int64(r.i) {
+			err := r.Skip(offset - r.blockStart - int64(r.i))
+			return r.blockStart + int64(r.i), err
+		}
+		return 0, ErrUnsupported
+
+	}
+
+	switch whence {
+	case io.SeekCurrent:
+		offset += r.blockStart + int64(r.i)
+	case io.SeekEnd:
+		offset = -offset
+	}
+	c, u, err := r.index.Find(offset)
+	if err != nil {
+		return r.blockStart + int64(r.i), err
+	}
+
+	// Seek to next block
+	_, err = rs.Seek(c, io.SeekStart)
+	if err != nil {
+		return 0, err
+	}
+
+	if offset < 0 {
+		offset = r.index.TotalUncompressed + offset
+	}
+
+	r.i = r.j // Remove rest of current block.
+	if u < offset {
+		// Forward inside block
+		return offset, r.Skip(offset - u)
+	}
+	return offset, nil
+}
+
 // ReadByte satisfies the io.ByteReader interface.
 func (r *Reader) ReadByte() (byte, error) {
 	if r.err != nil {
@@ -563,3 +746,17 @@ func (r *Reader) ReadByte() (byte, error) {
 	}
 	return 0, io.ErrNoProgress
 }
+
+// SkippableCB will register a callback for chunks with the specified ID.
+// ID must be a Reserved skippable chunks ID, 0x80-0xfe (inclusive).
+// For each chunk with the ID, the callback is called with the content.
+// Any returned non-nil error will abort decompression.
+// Only one callback per ID is supported, latest sent will be used.
+// Sending a nil function will disable previous callbacks.
+func (r *Reader) SkippableCB(id uint8, fn func(r io.Reader) error) error {
+	if id < 0x80 || id > chunkTypePadding {
+		return fmt.Errorf("ReaderSkippableCB: Invalid id provided, must be 0x80-0xfe (inclusive)")
+	}
+	r.skippableCB[id] = fn
+	return nil
+}
--- a/vendor/github.com/klauspost/compress/s2/encode.go
+++ b/vendor/github.com/klauspost/compress/s2/encode.go
@@ -395,23 +395,26 @@ type Writer struct {
 	// ibuf is a buffer for the incoming (uncompressed) bytes.
 	ibuf []byte

-	blockSize   int
-	obufLen     int
-	concurrency int
-	written     int64
-	output      chan chan result
-	buffers     sync.Pool
-	pad         int
+	blockSize     int
+	obufLen       int
+	concurrency   int
+	written       int64
+	uncompWritten int64 // Bytes sent to compression
+	output        chan chan result
+	buffers       sync.Pool
+	pad           int

 	writer   io.Writer
 	randSrc  io.Reader
 	writerWg sync.WaitGroup
+	index    Index

 	// wroteStreamHeader is whether we have written the stream header.
 	wroteStreamHeader bool
 	paramsOK          bool
 	snappy            bool
 	flushOnWrite      bool
+	appendIndex       bool
 	level             uint8
 }

@@ -422,7 +425,11 @@ const (
 	levelBest
 )

-type result []byte
+type result struct {
+	b []byte
+	// Uncompressed start offset
+	startOffset int64
+}

 // err returns the previously set error.
 // If no error has been set it is set to err if not nil.
@@ -454,6 +461,9 @@ func (w *Writer) Reset(writer io.Writer) {
 	w.wroteStreamHeader = false
 	w.written = 0
 	w.writer = writer
+	w.uncompWritten = 0
+	w.index.reset(w.blockSize)
+
 	// If we didn't get a writer, stop here.
 	if writer == nil {
 		return
@@ -474,7 +484,8 @@ func (w *Writer) Reset(writer io.Writer) {
 		// Get a queued write.
 		for write := range toWrite {
 			// Wait for the data to be available.
-			in := <-write
+			input := <-write
+			in := input.b
 			if len(in) > 0 {
 				if w.err(nil) == nil {
 					// Don't expose data from previous buffers.
@@ -485,11 +496,12 @@ func (w *Writer) Reset(writer io.Writer) {
 						err = io.ErrShortBuffer
 					}
 					_ = w.err(err)
+					w.err(w.index.add(w.written, input.startOffset))
 					w.written += int64(n)
 				}
 			}
 			if cap(in) >= w.obufLen {
-				w.buffers.Put([]byte(in))
+				w.buffers.Put(in)
 			}
 			// close the incoming write request.
 			// This can be used for synchronizing flushes.
@@ -500,6 +512,9 @@ func (w *Writer) Reset(writer io.Writer) {

 // Write satisfies the io.Writer interface.
 func (w *Writer) Write(p []byte) (nRet int, errRet error) {
+	if err := w.err(nil); err != nil {
+		return 0, err
+	}
 	if w.flushOnWrite {
 		return w.write(p)
 	}
@@ -535,6 +550,9 @@ func (w *Writer) Write(p []byte) (nRet int, errRet error) {
 // The return value n is the number of bytes read.
 // Any error except io.EOF encountered during the read is also returned.
 func (w *Writer) ReadFrom(r io.Reader) (n int64, err error) {
+	if err := w.err(nil); err != nil {
+		return 0, err
+	}
 	if len(w.ibuf) > 0 {
 		err := w.Flush()
 		if err != nil {
@@ -577,6 +595,85 @@ func (w *Writer) ReadFrom(r io.Reader) (n int64, err error) {
 	return n, w.err(nil)
 }

+// AddSkippableBlock will add a skippable block to the stream.
+// The ID must be 0x80-0xfe (inclusive).
+// Length of the skippable block must be <= 16777215 bytes.
+func (w *Writer) AddSkippableBlock(id uint8, data []byte) (err error) {
+	if err := w.err(nil); err != nil {
+		return err
+	}
+	if len(data) == 0 {
+		return nil
+	}
+	if id < 0x80 || id > chunkTypePadding {
+		return fmt.Errorf("invalid skippable block id %x", id)
+	}
+	if len(data) > maxChunkSize {
+		return fmt.Errorf("skippable block excessed maximum size")
+	}
+	var header [4]byte
+	chunkLen := 4 + len(data)
+	header[0] = id
+	header[1] = uint8(chunkLen >> 0)
+	header[2] = uint8(chunkLen >> 8)
+	header[3] = uint8(chunkLen >> 16)
+	if w.concurrency == 1 {
+		write := func(b []byte) error {
+			n, err := w.writer.Write(b)
+			if err = w.err(err); err != nil {
+				return err
+			}
+			if n != len(data) {
+				return w.err(io.ErrShortWrite)
+			}
+			w.written += int64(n)
+			return w.err(nil)
+		}
+		if !w.wroteStreamHeader {
+			w.wroteStreamHeader = true
+			if w.snappy {
+				if err := write([]byte(magicChunkSnappy)); err != nil {
+					return err
+				}
+			} else {
+				if err := write([]byte(magicChunk)); err != nil {
+					return err
+				}
+			}
+		}
+		if err := write(header[:]); err != nil {
+			return err
+		}
+		if err := write(data); err != nil {
+			return err
+		}
+	}
+
+	// Create output...
+	if !w.wroteStreamHeader {
+		w.wroteStreamHeader = true
+		hWriter := make(chan result)
+		w.output <- hWriter
+		if w.snappy {
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunkSnappy)}
+		} else {
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunk)}
+		}
+	}
+
+	// Copy input.
+	inbuf := w.buffers.Get().([]byte)[:4]
+	copy(inbuf, header[:])
+	inbuf = append(inbuf, data...)
+
+	output := make(chan result, 1)
+	// Queue output.
+	w.output <- output
+	output <- result{startOffset: w.uncompWritten, b: inbuf}
+
+	return nil
+}
+
 // EncodeBuffer will add a buffer to the stream.
 // This is the fastest way to encode a stream,
 // but the input buffer cannot be written to by the caller
@@ -614,9 +711,9 @@ func (w *Writer) EncodeBuffer(buf []byte) (err error) {
 		hWriter := make(chan result)
 		w.output <- hWriter
 		if w.snappy {
-			hWriter <- []byte(magicChunkSnappy)
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunkSnappy)}
 		} else {
-			hWriter <- []byte(magicChunk)
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunk)}
 		}
 	}

@@ -632,6 +729,10 @@ func (w *Writer) EncodeBuffer(buf []byte) (err error) {
 		output := make(chan result)
 		// Queue output now, so we keep order.
 		w.output <- output
+		res := result{
+			startOffset: w.uncompWritten,
+		}
+		w.uncompWritten += int64(len(uncompressed))
 		go func() {
 			checksum := crc(uncompressed)

@@ -664,7 +765,8 @@ func (w *Writer) EncodeBuffer(buf []byte) (err error) {
 			obuf[7] = uint8(checksum >> 24)

 			// Queue final output.
-			output <- obuf
+			res.b = obuf
+			output <- res
 		}()
 	}
 	return nil
@@ -708,9 +810,9 @@ func (w *Writer) write(p []byte) (nRet int, errRet error) {
 			hWriter := make(chan result)
 			w.output <- hWriter
 			if w.snappy {
-				hWriter <- []byte(magicChunkSnappy)
+				hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunkSnappy)}
 			} else {
-				hWriter <- []byte(magicChunk)
+				hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunk)}
 			}
 		}

@@ -731,6 +833,11 @@ func (w *Writer) write(p []byte) (nRet int, errRet error) {
 		output := make(chan result)
 		// Queue output now, so we keep order.
 		w.output <- output
+		res := result{
+			startOffset: w.uncompWritten,
+		}
+		w.uncompWritten += int64(len(uncompressed))
+
 		go func() {
 			checksum := crc(uncompressed)

@@ -763,7 +870,8 @@ func (w *Writer) write(p []byte) (nRet int, errRet error) {
 			obuf[7] = uint8(checksum >> 24)

 			// Queue final output.
-			output <- obuf
+			res.b = obuf
+			output <- res

 			// Put unused buffer back in pool.
 			w.buffers.Put(inbuf)
@@ -793,9 +901,9 @@ func (w *Writer) writeFull(inbuf []byte) (errRet error) {
 		hWriter := make(chan result)
 		w.output <- hWriter
 		if w.snappy {
-			hWriter <- []byte(magicChunkSnappy)
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunkSnappy)}
 		} else {
-			hWriter <- []byte(magicChunk)
+			hWriter <- result{startOffset: w.uncompWritten, b: []byte(magicChunk)}
 		}
 	}

@@ -806,6 +914,11 @@ func (w *Writer) writeFull(inbuf []byte) (errRet error) {
 	output := make(chan result)
 	// Queue output now, so we keep order.
 	w.output <- output
+	res := result{
+		startOffset: w.uncompWritten,
+	}
+	w.uncompWritten += int64(len(uncompressed))
+
 	go func() {
 		checksum := crc(uncompressed)

@@ -838,7 +951,8 @@ func (w *Writer) writeFull(inbuf []byte) (errRet error) {
 		obuf[7] = uint8(checksum >> 24)

 		// Queue final output.
-		output <- obuf
+		res.b = obuf
+		output <- res

 		// Put unused buffer back in pool.
 		w.buffers.Put(inbuf)
@@ -912,7 +1026,10 @@ func (w *Writer) writeSync(p []byte) (nRet int, errRet error) {
 		if n != len(obuf) {
 			return 0, w.err(io.ErrShortWrite)
 		}
+		w.err(w.index.add(w.written, w.uncompWritten))
 		w.written += int64(n)
+		w.uncompWritten += int64(len(uncompressed))
+
 		if chunkType == chunkTypeUncompressedData {
 			// Write uncompressed data.
 			n, err := w.writer.Write(uncompressed)
@@ -961,38 +1078,87 @@ func (w *Writer) Flush() error {
 	res := make(chan result)
 	w.output <- res
 	// Block until this has been picked up.
-	res <- nil
+	res <- result{b: nil, startOffset: w.uncompWritten}
 	// When it is closed, we have flushed.
 	<-res
 	return w.err(nil)
 }

 // Close calls Flush and then closes the Writer.
-// Calling Close multiple times is ok.
+// Calling Close multiple times is ok,
+// but calling CloseIndex after this will make it not return the index.
 func (w *Writer) Close() error {
+	_, err := w.closeIndex(w.appendIndex)
+	return err
+}
+
+// CloseIndex calls Close and returns an index on first call.
+// This is not required if you are only adding index to a stream.
+func (w *Writer) CloseIndex() ([]byte, error) {
+	return w.closeIndex(true)
+}
+
+func (w *Writer) closeIndex(idx bool) ([]byte, error) {
 	err := w.Flush()
 	if w.output != nil {
 		close(w.output)
 		w.writerWg.Wait()
 		w.output = nil
 	}
-	if w.err(nil) == nil && w.writer != nil && w.pad > 0 {
-		add := calcSkippableFrame(w.written, int64(w.pad))
-		frame, err := skippableFrame(w.ibuf[:0], add, w.randSrc)
-		if err = w.err(err); err != nil {
-			return err
-		}
-		_, err2 := w.writer.Write(frame)
-		_ = w.err(err2)
-	}
-	_ = w.err(errClosed)
-	if err == errClosed {
-		return nil
-	}
-	return err
-}

-const skippableFrameHeader = 4
+	var index []byte
+	if w.err(nil) == nil && w.writer != nil {
+		// Create index.
+		if idx {
+			compSize := int64(-1)
+			if w.pad <= 1 {
+				compSize = w.written
+			}
+			index = w.index.appendTo(w.ibuf[:0], w.uncompWritten, compSize)
+			// Count as written for padding.
+			if w.appendIndex {
+				w.written += int64(len(index))
+			}
+			if true {
+				_, err := w.index.Load(index)
+				if err != nil {
+					panic(err)
+				}
+			}
+		}
+
+		if w.pad > 1 {
+			tmp := w.ibuf[:0]
+			if len(index) > 0 {
+				// Allocate another buffer.
+				tmp = w.buffers.Get().([]byte)[:0]
+				defer w.buffers.Put(tmp)
+			}
+			add := calcSkippableFrame(w.written, int64(w.pad))
+			frame, err := skippableFrame(tmp, add, w.randSrc)
+			if err = w.err(err); err != nil {
+				return nil, err
+			}
+			n, err2 := w.writer.Write(frame)
+			if err2 == nil && n != len(frame) {
+				err2 = io.ErrShortWrite
+			}
+			_ = w.err(err2)
+		}
+		if len(index) > 0 && w.appendIndex {
+			n, err2 := w.writer.Write(index)
+			if err2 == nil && n != len(index) {
+				err2 = io.ErrShortWrite
+			}
+			_ = w.err(err2)
+		}
+	}
+	err = w.err(errClosed)
+	if err == errClosed {
+		return index, nil
+	}
+	return nil, err
+}

 // calcSkippableFrame will return a total size to be added for written
 // to be divisible by multiple.
@@ -1057,6 +1223,15 @@ func WriterConcurrency(n int) WriterOption {
 	}
 }

+// WriterAddIndex will append an index to the end of a stream
+// when it is closed.
+func WriterAddIndex() WriterOption {
+	return func(w *Writer) error {
+		w.appendIndex = true
+		return nil
+	}
+}
+
 // WriterBetterCompression will enable better compression.
 // EncodeBetter compresses better than Encode but typically with a
 // 10-40% speed decrease on both compression and decompression.
--- a/vendor/github.com/klauspost/compress/s2/index.go
+++ b/vendor/github.com/klauspost/compress/s2/index.go
@@ -0,0 +1,525 @@
+// Copyright (c) 2022+ Klaus Post. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package s2
+
+import (
+	"bytes"
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"io"
+)
+
+const (
+	S2IndexHeader   = "s2idx\x00"
+	S2IndexTrailer  = "\x00xdi2s"
+	maxIndexEntries = 1 << 16
+)
+
+// Index represents an S2/Snappy index.
+type Index struct {
+	TotalUncompressed int64 // Total Uncompressed size if known. Will be -1 if unknown.
+	TotalCompressed   int64 // Total Compressed size if known. Will be -1 if unknown.
+	info              []struct {
+		compressedOffset   int64
+		uncompressedOffset int64
+	}
+	estBlockUncomp int64
+}
+
+func (i *Index) reset(maxBlock int) {
+	i.estBlockUncomp = int64(maxBlock)
+	i.TotalCompressed = -1
+	i.TotalUncompressed = -1
+	if len(i.info) > 0 {
+		i.info = i.info[:0]
+	}
+}
+
+// allocInfos will allocate an empty slice of infos.
+func (i *Index) allocInfos(n int) {
+	if n > maxIndexEntries {
+		panic("n > maxIndexEntries")
+	}
+	i.info = make([]struct {
+		compressedOffset   int64
+		uncompressedOffset int64
+	}, 0, n)
+}
+
+// add an uncompressed and compressed pair.
+// Entries must be sent in order.
+func (i *Index) add(compressedOffset, uncompressedOffset int64) error {
+	if i == nil {
+		return nil
+	}
+	lastIdx := len(i.info) - 1
+	if lastIdx >= 0 {
+		latest := i.info[lastIdx]
+		if latest.uncompressedOffset == uncompressedOffset {
+			// Uncompressed didn't change, don't add entry,
+			// but update start index.
+			latest.compressedOffset = compressedOffset
+			i.info[lastIdx] = latest
+			return nil
+		}
+		if latest.uncompressedOffset > uncompressedOffset {
+			return fmt.Errorf("internal error: Earlier uncompressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset)
+		}
+		if latest.compressedOffset > compressedOffset {
+			return fmt.Errorf("internal error: Earlier compressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset)
+		}
+	}
+	i.info = append(i.info, struct {
+		compressedOffset   int64
+		uncompressedOffset int64
+	}{compressedOffset: compressedOffset, uncompressedOffset: uncompressedOffset})
+	return nil
+}
+
+// Find the offset at or before the wanted (uncompressed) offset.
+// If offset is 0 or positive it is the offset from the beginning of the file.
+// If the uncompressed size is known, the offset must be within the file.
+// If an offset outside the file is requested io.ErrUnexpectedEOF is returned.
+// If the offset is negative, it is interpreted as the distance from the end of the file,
+// where -1 represents the last byte.
+// If offset from the end of the file is requested, but size is unknown,
+// ErrUnsupported will be returned.
+func (i *Index) Find(offset int64) (compressedOff, uncompressedOff int64, err error) {
+	if i.TotalUncompressed < 0 {
+		return 0, 0, ErrCorrupt
+	}
+	if offset < 0 {
+		offset = i.TotalUncompressed + offset
+		if offset < 0 {
+			return 0, 0, io.ErrUnexpectedEOF
+		}
+	}
+	if offset > i.TotalUncompressed {
+		return 0, 0, io.ErrUnexpectedEOF
+	}
+	for _, info := range i.info {
+		if info.uncompressedOffset > offset {
+			break
+		}
+		compressedOff = info.compressedOffset
+		uncompressedOff = info.uncompressedOffset
+	}
+	return compressedOff, uncompressedOff, nil
+}
+
+// reduce to stay below maxIndexEntries
+func (i *Index) reduce() {
+	if len(i.info) < maxIndexEntries && i.estBlockUncomp >= 1<<20 {
+		return
+	}
+
+	// Algorithm, keep 1, remove removeN entries...
+	removeN := (len(i.info) + 1) / maxIndexEntries
+	src := i.info
+	j := 0
+
+	// Each block should be at least 1MB, but don't reduce below 1000 entries.
+	for i.estBlockUncomp*(int64(removeN)+1) < 1<<20 && len(i.info)/(removeN+1) > 1000 {
+		removeN++
+	}
+	for idx := 0; idx < len(src); idx++ {
+		i.info[j] = src[idx]
+		j++
+		idx += removeN
+	}
+	i.info = i.info[:j]
+	// Update maxblock estimate.
+	i.estBlockUncomp += i.estBlockUncomp * int64(removeN)
+}
+
+func (i *Index) appendTo(b []byte, uncompTotal, compTotal int64) []byte {
+	i.reduce()
+	var tmp [binary.MaxVarintLen64]byte
+
+	initSize := len(b)
+	// We make the start a skippable header+size.
+	b = append(b, ChunkTypeIndex, 0, 0, 0)
+	b = append(b, []byte(S2IndexHeader)...)
+	// Total Uncompressed size
+	n := binary.PutVarint(tmp[:], uncompTotal)
+	b = append(b, tmp[:n]...)
+	// Total Compressed size
+	n = binary.PutVarint(tmp[:], compTotal)
+	b = append(b, tmp[:n]...)
+	// Put EstBlockUncomp size
+	n = binary.PutVarint(tmp[:], i.estBlockUncomp)
+	b = append(b, tmp[:n]...)
+	// Put length
+	n = binary.PutVarint(tmp[:], int64(len(i.info)))
+	b = append(b, tmp[:n]...)
+
+	// Check if we should add uncompressed offsets
+	var hasUncompressed byte
+	for idx, info := range i.info {
+		if idx == 0 {
+			if info.uncompressedOffset != 0 {
+				hasUncompressed = 1
+				break
+			}
+			continue
+		}
+		if info.uncompressedOffset != i.info[idx-1].uncompressedOffset+i.estBlockUncomp {
+			hasUncompressed = 1
+			break
+		}
+	}
+	b = append(b, hasUncompressed)
+
+	// Add each entry
+	if hasUncompressed == 1 {
+		for idx, info := range i.info {
+			uOff := info.uncompressedOffset
+			if idx > 0 {
+				prev := i.info[idx-1]
+				uOff -= prev.uncompressedOffset + (i.estBlockUncomp)
+			}
+			n = binary.PutVarint(tmp[:], uOff)
+			b = append(b, tmp[:n]...)
+		}
+	}
+
+	// Initial compressed size estimate.
+	cPredict := i.estBlockUncomp / 2
+
+	for idx, info := range i.info {
+		cOff := info.compressedOffset
+		if idx > 0 {
+			prev := i.info[idx-1]
+			cOff -= prev.compressedOffset + cPredict
+			// Update compressed size prediction, with half the error.
+			cPredict += cOff / 2
+		}
+		n = binary.PutVarint(tmp[:], cOff)
+		b = append(b, tmp[:n]...)
+	}
+
+	// Add Total Size.
+	// Stored as fixed size for easier reading.
+	binary.LittleEndian.PutUint32(tmp[:], uint32(len(b)-initSize+4+len(S2IndexTrailer)))
+	b = append(b, tmp[:4]...)
+	// Trailer
+	b = append(b, []byte(S2IndexTrailer)...)
+
+	// Update size
+	chunkLen := len(b) - initSize - skippableFrameHeader
+	b[initSize+1] = uint8(chunkLen >> 0)
+	b[initSize+2] = uint8(chunkLen >> 8)
+	b[initSize+3] = uint8(chunkLen >> 16)
+	//fmt.Printf("chunklen: 0x%x Uncomp:%d, Comp:%d\n", chunkLen, uncompTotal, compTotal)
+	return b
+}
+
+// Load a binary index.
+// A zero value Index can be used or a previous one can be reused.
+func (i *Index) Load(b []byte) ([]byte, error) {
+	if len(b) <= 4+len(S2IndexHeader)+len(S2IndexTrailer) {
+		return b, io.ErrUnexpectedEOF
+	}
+	if b[0] != ChunkTypeIndex {
+		return b, ErrCorrupt
+	}
+	chunkLen := int(b[1]) | int(b[2])<<8 | int(b[3])<<16
+	b = b[4:]
+
+	// Validate we have enough...
+	if len(b) < chunkLen {
+		return b, io.ErrUnexpectedEOF
+	}
+	if !bytes.Equal(b[:len(S2IndexHeader)], []byte(S2IndexHeader)) {
+		return b, ErrUnsupported
+	}
+	b = b[len(S2IndexHeader):]
+
+	// Total Uncompressed
+	if v, n := binary.Varint(b); n <= 0 || v < 0 {
+		return b, ErrCorrupt
+	} else {
+		i.TotalUncompressed = v
+		b = b[n:]
+	}
+
+	// Total Compressed
+	if v, n := binary.Varint(b); n <= 0 {
+		return b, ErrCorrupt
+	} else {
+		i.TotalCompressed = v
+		b = b[n:]
+	}
+
+	// Read EstBlockUncomp
+	if v, n := binary.Varint(b); n <= 0 {
+		return b, ErrCorrupt
+	} else {
+		if v < 0 {
+			return b, ErrCorrupt
+		}
+		i.estBlockUncomp = v
+		b = b[n:]
+	}
+
+	var entries int
+	if v, n := binary.Varint(b); n <= 0 {
+		return b, ErrCorrupt
+	} else {
+		if v < 0 || v > maxIndexEntries {
+			return b, ErrCorrupt
+		}
+		entries = int(v)
+		b = b[n:]
+	}
+	if cap(i.info) < entries {
+		i.allocInfos(entries)
+	}
+	i.info = i.info[:entries]
+
+	if len(b) < 1 {
+		return b, io.ErrUnexpectedEOF
+	}
+	hasUncompressed := b[0]
+	b = b[1:]
+	if hasUncompressed&1 != hasUncompressed {
+		return b, ErrCorrupt
+	}
+
+	// Add each uncompressed entry
+	for idx := range i.info {
+		var uOff int64
+		if hasUncompressed != 0 {
+			// Load delta
+			if v, n := binary.Varint(b); n <= 0 {
+				return b, ErrCorrupt
+			} else {
+				uOff = v
+				b = b[n:]
+			}
+		}
+
+		if idx > 0 {
+			prev := i.info[idx-1].uncompressedOffset
+			uOff += prev + (i.estBlockUncomp)
+			if uOff <= prev {
+				return b, ErrCorrupt
+			}
+		}
+		if uOff < 0 {
+			return b, ErrCorrupt
+		}
+		i.info[idx].uncompressedOffset = uOff
+	}
+
+	// Initial compressed size estimate.
+	cPredict := i.estBlockUncomp / 2
+
+	// Add each compressed entry
+	for idx := range i.info {
+		var cOff int64
+		if v, n := binary.Varint(b); n <= 0 {
+			return b, ErrCorrupt
+		} else {
+			cOff = v
+			b = b[n:]
+		}
+
+		if idx > 0 {
+			// Update compressed size prediction, with half the error.
+			cPredictNew := cPredict + cOff/2
+
+			prev := i.info[idx-1].compressedOffset
+			cOff += prev + cPredict
+			if cOff <= prev {
+				return b, ErrCorrupt
+			}
+			cPredict = cPredictNew
+		}
+		if cOff < 0 {
+			return b, ErrCorrupt
+		}
+		i.info[idx].compressedOffset = cOff
+	}
+	if len(b) < 4+len(S2IndexTrailer) {
+		return b, io.ErrUnexpectedEOF
+	}
+	// Skip size...
+	b = b[4:]
+
+	// Check trailer...
+	if !bytes.Equal(b[:len(S2IndexTrailer)], []byte(S2IndexTrailer)) {
+		return b, ErrCorrupt
+	}
+	return b[len(S2IndexTrailer):], nil
+}
+
+// LoadStream will load an index from the end of the supplied stream.
+// ErrUnsupported will be returned if the signature cannot be found.
+// ErrCorrupt will be returned if unexpected values are found.
+// io.ErrUnexpectedEOF is returned if there are too few bytes.
+// IO errors are returned as-is.
+func (i *Index) LoadStream(rs io.ReadSeeker) error {
+	// Go to end.
+	_, err := rs.Seek(-10, io.SeekEnd)
+	if err != nil {
+		return err
+	}
+	var tmp [10]byte
+	_, err = io.ReadFull(rs, tmp[:])
+	if err != nil {
+		return err
+	}
+	// Check trailer...
+	if !bytes.Equal(tmp[4:4+len(S2IndexTrailer)], []byte(S2IndexTrailer)) {
+		return ErrUnsupported
+	}
+	sz := binary.LittleEndian.Uint32(tmp[:4])
+	if sz > maxChunkSize+skippableFrameHeader {
+		return ErrCorrupt
+	}
+	_, err = rs.Seek(-int64(sz), io.SeekEnd)
+	if err != nil {
+		return err
+	}
+
+	// Read index.
+	buf := make([]byte, sz)
+	_, err = io.ReadFull(rs, buf)
+	if err != nil {
+		return err
+	}
+	_, err = i.Load(buf)
+	return err
+}
+
+// IndexStream will return an index for a stream.
+// The stream structure will be checked, but
+// data within blocks is not verified.
+// The returned index can either be appended to the end of the stream
+// or stored separately.
+func IndexStream(r io.Reader) ([]byte, error) {
+	var i Index
+	var buf [maxChunkSize]byte
+	var readHeader bool
+	for {
+		_, err := io.ReadFull(r, buf[:4])
+		if err != nil {
+			if err == io.EOF {
+				return i.appendTo(nil, i.TotalUncompressed, i.TotalCompressed), nil
+			}
+			return nil, err
+		}
+		// Start of this chunk.
+		startChunk := i.TotalCompressed
+		i.TotalCompressed += 4
+
+		chunkType := buf[0]
+		if !readHeader {
+			if chunkType != chunkTypeStreamIdentifier {
+				return nil, ErrCorrupt
+			}
+			readHeader = true
+		}
+		chunkLen := int(buf[1]) | int(buf[2])<<8 | int(buf[3])<<16
+		if chunkLen < checksumSize {
+			return nil, ErrCorrupt
+		}
+
+		i.TotalCompressed += int64(chunkLen)
+		_, err = io.ReadFull(r, buf[:chunkLen])
+		if err != nil {
+			return nil, io.ErrUnexpectedEOF
+		}
+		// The chunk types are specified at
+		// https://github.com/google/snappy/blob/master/framing_format.txt
+		switch chunkType {
+		case chunkTypeCompressedData:
+			// Section 4.2. Compressed data (chunk type 0x00).
+			// Skip checksum.
+			dLen, err := DecodedLen(buf[checksumSize:])
+			if err != nil {
+				return nil, err
+			}
+			if dLen > maxBlockSize {
+				return nil, ErrCorrupt
+			}
+			if i.estBlockUncomp == 0 {
+				// Use first block for estimate...
+				i.estBlockUncomp = int64(dLen)
+			}
+			err = i.add(startChunk, i.TotalUncompressed)
+			if err != nil {
+				return nil, err
+			}
+			i.TotalUncompressed += int64(dLen)
+			continue
+		case chunkTypeUncompressedData:
+			n2 := chunkLen - checksumSize
+			if n2 > maxBlockSize {
+				return nil, ErrCorrupt
+			}
+			if i.estBlockUncomp == 0 {
+				// Use first block for estimate...
+				i.estBlockUncomp = int64(n2)
+			}
+			err = i.add(startChunk, i.TotalUncompressed)
+			if err != nil {
+				return nil, err
+			}
+			i.TotalUncompressed += int64(n2)
+			continue
+		case chunkTypeStreamIdentifier:
+			// Section 4.1. Stream identifier (chunk type 0xff).
+			if chunkLen != len(magicBody) {
+				return nil, ErrCorrupt
+			}
+
+			if string(buf[:len(magicBody)]) != magicBody {
+				if string(buf[:len(magicBody)]) != magicBodySnappy {
+					return nil, ErrCorrupt
+				}
+			}
+
+			continue
+		}
+
+		if chunkType <= 0x7f {
+			// Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f).
+			return nil, ErrUnsupported
+		}
+		if chunkLen > maxChunkSize {
+			return nil, ErrUnsupported
+		}
+		// Section 4.4 Padding (chunk type 0xfe).
+		// Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd).
+	}
+}
+
+// JSON returns the index as JSON text.
+func (i *Index) JSON() []byte {
+	x := struct {
+		TotalUncompressed int64 `json:"total_uncompressed"` // Total Uncompressed size if known. Will be -1 if unknown.
+		TotalCompressed   int64 `json:"total_compressed"`   // Total Compressed size if known. Will be -1 if unknown.
+		Offsets           []struct {
+			CompressedOffset   int64 `json:"compressed"`
+			UncompressedOffset int64 `json:"uncompressed"`
+		} `json:"offsets"`
+		EstBlockUncomp int64 `json:"est_block_uncompressed"`
+	}{
+		TotalUncompressed: i.TotalUncompressed,
+		TotalCompressed:   i.TotalCompressed,
+		EstBlockUncomp:    i.estBlockUncomp,
+	}
+	for _, v := range i.info {
+		x.Offsets = append(x.Offsets, struct {
+			CompressedOffset   int64 `json:"compressed"`
+			UncompressedOffset int64 `json:"uncompressed"`
+		}{CompressedOffset: v.compressedOffset, UncompressedOffset: v.uncompressedOffset})
+	}
+	b, _ := json.MarshalIndent(x, "", "  ")
+	return b
+}
--- a/vendor/github.com/klauspost/compress/s2/s2.go
+++ b/vendor/github.com/klauspost/compress/s2/s2.go
@@ -87,6 +87,9 @@ const (
 	// minBlockSize is the minimum size of block setting when creating a writer.
 	minBlockSize = 4 << 10

+	skippableFrameHeader = 4
+	maxChunkSize         = 1<<24 - 1 // 16777215
+
 	// Default block size
 	defaultBlockSize = 1 << 20

@@ -99,6 +102,7 @@ const (
 const (
 	chunkTypeCompressedData   = 0x00
 	chunkTypeUncompressedData = 0x01
+	ChunkTypeIndex            = 0x99
 	chunkTypePadding          = 0xfe
 	chunkTypeStreamIdentifier = 0xff
 )
--- a/vendor/github.com/klauspost/compress/zstd/bitreader.go
+++ b/vendor/github.com/klauspost/compress/zstd/bitreader.go
@@ -50,16 +50,23 @@ func (b *bitReader) getBits(n uint8) int {
 	if n == 0 /*|| b.bitsRead >= 64 */ {
 		return 0
 	}
-	return b.getBitsFast(n)
+	return int(b.get32BitsFast(n))
 }

-// getBitsFast requires that at least one bit is requested every time.
+// get32BitsFast requires that at least one bit is requested every time.
 // There are no checks if the buffer is filled.
-func (b *bitReader) getBitsFast(n uint8) int {
+func (b *bitReader) get32BitsFast(n uint8) uint32 {
 	const regMask = 64 - 1
 	v := uint32((b.value << (b.bitsRead & regMask)) >> ((regMask + 1 - n) & regMask))
 	b.bitsRead += n
-	return int(v)
+	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.
--- a/vendor/github.com/klauspost/compress/zstd/bitwriter.go
+++ b/vendor/github.com/klauspost/compress/zstd/bitwriter.go
@@ -38,7 +38,7 @@ func (b *bitWriter) addBits16NC(value uint16, bits uint8) {
 	b.nBits += bits
 }

-// addBits32NC will add up to 32 bits.
+// addBits32NC will add up to 31 bits.
 // It will not check if there is space for them,
 // so the caller must ensure that it has flushed recently.
 func (b *bitWriter) addBits32NC(value uint32, bits uint8) {
@@ -46,6 +46,26 @@ func (b *bitWriter) addBits32NC(value uint32, bits uint8) {
 	b.nBits += bits
 }

+// addBits64NC will add up to 64 bits.
+// There must be space for 32 bits.
+func (b *bitWriter) addBits64NC(value uint64, bits uint8) {
+	if bits <= 31 {
+		b.addBits32Clean(uint32(value), bits)
+		return
+	}
+	b.addBits32Clean(uint32(value), 32)
+	b.flush32()
+	b.addBits32Clean(uint32(value>>32), bits-32)
+}
+
+// addBits32Clean will add up to 32 bits.
+// It will not check if there is space for them.
+// The input must not contain more bits than specified.
+func (b *bitWriter) addBits32Clean(value uint32, bits uint8) {
+	b.bitContainer |= uint64(value) << (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) {
--- a/vendor/github.com/klauspost/compress/zstd/blockdec.go
+++ b/vendor/github.com/klauspost/compress/zstd/blockdec.go
@@ -76,12 +76,11 @@ type blockDec struct {
 	// Window size of the block.
 	WindowSize uint64

-	history     chan *history
-	input       chan struct{}
-	result      chan decodeOutput
-	sequenceBuf []seq
-	err         error
-	decWG       sync.WaitGroup
+	history chan *history
+	input   chan struct{}
+	result  chan decodeOutput
+	err     error
+	decWG   sync.WaitGroup

 	// Frame to use for singlethreaded decoding.
 	// Should not be used by the decoder itself since parent may be another frame.
@@ -512,18 +511,7 @@ func (b *blockDec) decodeCompressed(hist *history) error {
 		nSeqs = 0x7f00 + int(in[1]) + (int(in[2]) << 8)
 		in = in[3:]
 	}
-	// Allocate sequences
-	if cap(b.sequenceBuf) < nSeqs {
-		if b.lowMem {
-			b.sequenceBuf = make([]seq, nSeqs)
-		} else {
-			// Allocate max
-			b.sequenceBuf = make([]seq, nSeqs, maxSequences)
-		}
-	} else {
-		// Reuse buffer
-		b.sequenceBuf = b.sequenceBuf[:nSeqs]
-	}
+
 	var seqs = &sequenceDecs{}
 	if nSeqs > 0 {
 		if len(in) < 1 {
--- a/vendor/github.com/klauspost/compress/zstd/blockenc.go
+++ b/vendor/github.com/klauspost/compress/zstd/blockenc.go
@@ -51,7 +51,7 @@ func (b *blockEnc) init() {
 		if cap(b.literals) < maxCompressedBlockSize {
 			b.literals = make([]byte, 0, maxCompressedBlockSize)
 		}
-		const defSeqs = 200
+		const defSeqs = 2000
 		if cap(b.sequences) < defSeqs {
 			b.sequences = make([]seq, 0, defSeqs)
 		}
@@ -426,7 +426,7 @@ func fuzzFseEncoder(data []byte) int {
 		return 0
 	}
 	enc := fseEncoder{}
-	hist := enc.Histogram()[:256]
+	hist := enc.Histogram()
 	maxSym := uint8(0)
 	for i, v := range data {
 		v = v & 63
@@ -722,52 +722,53 @@ func (b *blockEnc) encode(org []byte, raw, rawAllLits bool) error {
 		println("Encoded seq", seq, s, "codes:", s.llCode, s.mlCode, s.ofCode, "states:", ll.state, ml.state, of.state, "bits:", llB, mlB, ofB)
 	}
 	seq--
-	if llEnc.maxBits+mlEnc.maxBits+ofEnc.maxBits <= 32 {
-		// No need to flush (common)
-		for seq >= 0 {
-			s = b.sequences[seq]
-			wr.flush32()
-			llB, ofB, mlB := llTT[s.llCode], ofTT[s.ofCode], mlTT[s.mlCode]
-			// tabelog max is 8 for all.
-			of.encode(ofB)
-			ml.encode(mlB)
-			ll.encode(llB)
-			wr.flush32()
+	// Store sequences in reverse...
+	for seq >= 0 {
+		s = b.sequences[seq]

-			// We checked that all can stay within 32 bits
-			wr.addBits32NC(s.litLen, llB.outBits)
-			wr.addBits32NC(s.matchLen, mlB.outBits)
-			wr.addBits32NC(s.offset, ofB.outBits)
+		ofB := ofTT[s.ofCode]
+		wr.flush32() // tablelog max is below 8 for each, so it will fill max 24 bits.
+		//of.encode(ofB)
+		nbBitsOut := (uint32(of.state) + ofB.deltaNbBits) >> 16
+		dstState := int32(of.state>>(nbBitsOut&15)) + int32(ofB.deltaFindState)
+		wr.addBits16NC(of.state, uint8(nbBitsOut))
+		of.state = of.stateTable[dstState]

-			if debugSequences {
-				println("Encoded seq", seq, s)
-			}
+		// Accumulate extra bits.
+		outBits := ofB.outBits & 31
+		extraBits := uint64(s.offset & bitMask32[outBits])
+		extraBitsN := outBits

-			seq--
+		mlB := mlTT[s.mlCode]
+		//ml.encode(mlB)
+		nbBitsOut = (uint32(ml.state) + mlB.deltaNbBits) >> 16
+		dstState = int32(ml.state>>(nbBitsOut&15)) + int32(mlB.deltaFindState)
+		wr.addBits16NC(ml.state, uint8(nbBitsOut))
+		ml.state = ml.stateTable[dstState]
+
+		outBits = mlB.outBits & 31
+		extraBits = extraBits<<outBits | uint64(s.matchLen&bitMask32[outBits])
+		extraBitsN += outBits
+
+		llB := llTT[s.llCode]
+		//ll.encode(llB)
+		nbBitsOut = (uint32(ll.state) + llB.deltaNbBits) >> 16
+		dstState = int32(ll.state>>(nbBitsOut&15)) + int32(llB.deltaFindState)
+		wr.addBits16NC(ll.state, uint8(nbBitsOut))
+		ll.state = ll.stateTable[dstState]
+
+		outBits = llB.outBits & 31
+		extraBits = extraBits<<outBits | uint64(s.litLen&bitMask32[outBits])
+		extraBitsN += outBits
+
+		wr.flush32()
+		wr.addBits64NC(extraBits, extraBitsN)
+
+		if debugSequences {
+			println("Encoded seq", seq, s)
 		}
-	} else {
-		for seq >= 0 {
-			s = b.sequences[seq]
-			wr.flush32()
-			llB, ofB, mlB := llTT[s.llCode], ofTT[s.ofCode], mlTT[s.mlCode]
-			// tabelog max is below 8 for each.
-			of.encode(ofB)
-			ml.encode(mlB)
-			ll.encode(llB)
-			wr.flush32()

-			// ml+ll = max 32 bits total
-			wr.addBits32NC(s.litLen, llB.outBits)
-			wr.addBits32NC(s.matchLen, mlB.outBits)
-			wr.flush32()
-			wr.addBits32NC(s.offset, ofB.outBits)
-
-			if debugSequences {
-				println("Encoded seq", seq, s)
-			}
-
-			seq--
-		}
+		seq--
 	}
 	ml.flush(mlEnc.actualTableLog)
 	of.flush(ofEnc.actualTableLog)
@@ -801,14 +802,13 @@ func (b *blockEnc) genCodes() {
 		// nothing to do
 		return
 	}
-
 	if len(b.sequences) > math.MaxUint16 {
 		panic("can only encode up to 64K sequences")
 	}
 	// No bounds checks after here:
-	llH := b.coders.llEnc.Histogram()[:256]
-	ofH := b.coders.ofEnc.Histogram()[:256]
-	mlH := b.coders.mlEnc.Histogram()[:256]
+	llH := b.coders.llEnc.Histogram()
+	ofH := b.coders.ofEnc.Histogram()
+	mlH := b.coders.mlEnc.Histogram()
 	for i := range llH {
 		llH[i] = 0
 	}
@@ -820,7 +820,8 @@ func (b *blockEnc) genCodes() {
 	}

 	var llMax, ofMax, mlMax uint8
-	for i, seq := range b.sequences {
+	for i := range b.sequences {
+		seq := &b.sequences[i]
 		v := llCode(seq.litLen)
 		seq.llCode = v
 		llH[v]++
@@ -844,7 +845,6 @@ func (b *blockEnc) genCodes() {
 				panic(fmt.Errorf("mlMax > maxMatchLengthSymbol (%d), matchlen: %d", mlMax, seq.matchLen))
 			}
 		}
-		b.sequences[i] = seq
 	}
 	maxCount := func(a []uint32) int {
 		var max uint32
--- a/vendor/github.com/klauspost/compress/zstd/decodeheader.go
+++ b/vendor/github.com/klauspost/compress/zstd/decodeheader.go
@@ -5,6 +5,7 @@ package zstd

 import (
 	"bytes"
+	"encoding/binary"
 	"errors"
 	"io"
 )
@@ -15,18 +16,50 @@ const HeaderMaxSize = 14 + 3

 // Header contains information about the first frame and block within that.
 type Header struct {
-	// Window Size the window of data to keep while decoding.
-	// Will only be set if HasFCS is false.
-	WindowSize uint64
+	// SingleSegment specifies whether the data is to be decompressed into a
+	// single contiguous memory segment.
+	// It implies that WindowSize is invalid and that FrameContentSize is valid.
+	SingleSegment bool

-	// Frame content size.
-	// Expected size of the entire frame.
-	FrameContentSize uint64
+	// WindowSize is the window of data to keep while decoding.
+	// Will only be set if SingleSegment is false.
+	WindowSize uint64

 	// Dictionary ID.
 	// If 0, no dictionary.
 	DictionaryID uint32

+	// HasFCS specifies whether FrameContentSize has a valid value.
+	HasFCS bool
+
+	// FrameContentSize is the expected uncompressed size of the entire frame.
+	FrameContentSize uint64
+
+	// Skippable will be true if the frame is meant to be skipped.
+	// This implies that FirstBlock.OK is false.
+	Skippable bool
+
+	// SkippableID is the user-specific ID for the skippable frame.
+	// Valid values are between 0 to 15, inclusive.
+	SkippableID int
+
+	// SkippableSize is the length of the user data to skip following
+	// the header.
+	SkippableSize uint32
+
+	// HeaderSize is the raw size of the frame header.
+	//
+	// For normal frames, it includes the size of the magic number and
+	// the size of the header (per section 3.1.1.1).
+	// It does not include the size for any data blocks (section 3.1.1.2) nor
+	// the size for the trailing content checksum.
+	//
+	// For skippable frames, this counts the size of the magic number
+	// along with the size of the size field of the payload.
+	// It does not include the size of the skippable payload itself.
+	// The total frame size is the HeaderSize plus the SkippableSize.
+	HeaderSize int
+
 	// First block information.
 	FirstBlock struct {
 		// OK will be set if first block could be decoded.
@@ -51,17 +84,9 @@ type Header struct {
 		CompressedSize int
 	}

-	// Skippable will be true if the frame is meant to be skipped.
-	// No other information will be populated.
-	Skippable bool
-
 	// If set there is a checksum present for the block content.
+	// The checksum field at the end is always 4 bytes long.
 	HasCheckSum bool
-
-	// If this is true FrameContentSize will have a valid value
-	HasFCS bool
-
-	SingleSegment bool
 }

 // Decode the header from the beginning of the stream.
@@ -71,39 +96,46 @@ type Header struct {
 // If there isn't enough input, io.ErrUnexpectedEOF is returned.
 // The FirstBlock.OK will indicate if enough information was available to decode the first block header.
 func (h *Header) Decode(in []byte) error {
+	*h = Header{}
 	if len(in) < 4 {
 		return io.ErrUnexpectedEOF
 	}
+	h.HeaderSize += 4
 	b, in := in[:4], in[4:]
 	if !bytes.Equal(b, frameMagic) {
 		if !bytes.Equal(b[1:4], skippableFrameMagic) || b[0]&0xf0 != 0x50 {
 			return ErrMagicMismatch
 		}
-		*h = Header{Skippable: true}
+		if len(in) < 4 {
+			return io.ErrUnexpectedEOF
+		}
+		h.HeaderSize += 4
+		h.Skippable = true
+		h.SkippableID = int(b[0] & 0xf)
+		h.SkippableSize = binary.LittleEndian.Uint32(in)
 		return nil
 	}
-	if len(in) < 1 {
-		return io.ErrUnexpectedEOF
-	}
-
-	// Clear output
-	*h = Header{}
-	fhd, in := in[0], in[1:]
-	h.SingleSegment = fhd&(1<<5) != 0
-	h.HasCheckSum = fhd&(1<<2) != 0
-
-	if fhd&(1<<3) != 0 {
-		return errors.New("reserved bit set on frame header")
-	}

 	// Read Window_Descriptor
 	// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#window_descriptor
+	if len(in) < 1 {
+		return io.ErrUnexpectedEOF
+	}
+	fhd, in := in[0], in[1:]
+	h.HeaderSize++
+	h.SingleSegment = fhd&(1<<5) != 0
+	h.HasCheckSum = fhd&(1<<2) != 0
+	if fhd&(1<<3) != 0 {
+		return errors.New("reserved bit set on frame header")
+	}
+
 	if !h.SingleSegment {
 		if len(in) < 1 {
 			return io.ErrUnexpectedEOF
 		}
 		var wd byte
 		wd, in = in[0], in[1:]
+		h.HeaderSize++
 		windowLog := 10 + (wd >> 3)
 		windowBase := uint64(1) << windowLog
 		windowAdd := (windowBase / 8) * uint64(wd&0x7)
@@ -120,9 +152,7 @@ func (h *Header) Decode(in []byte) error {
 			return io.ErrUnexpectedEOF
 		}
 		b, in = in[:size], in[size:]
-		if b == nil {
-			return io.ErrUnexpectedEOF
-		}
+		h.HeaderSize += int(size)
 		switch size {
 		case 1:
 			h.DictionaryID = uint32(b[0])
@@ -152,9 +182,7 @@ func (h *Header) Decode(in []byte) error {
 			return io.ErrUnexpectedEOF
 		}
 		b, in = in[:fcsSize], in[fcsSize:]
-		if b == nil {
-			return io.ErrUnexpectedEOF
-		}
+		h.HeaderSize += int(fcsSize)
 		switch fcsSize {
 		case 1:
 			h.FrameContentSize = uint64(b[0])
--- a/vendor/github.com/klauspost/compress/zstd/enc_base.go
+++ b/vendor/github.com/klauspost/compress/zstd/enc_base.go
@@ -108,11 +108,6 @@ func (e *fastBase) UseBlock(enc *blockEnc) {
 	e.blk = enc
 }

-func (e *fastBase) matchlenNoHist(s, t int32, src []byte) int32 {
-	// Extend the match to be as long as possible.
-	return int32(matchLen(src[s:], src[t:]))
-}
-
 func (e *fastBase) matchlen(s, t int32, src []byte) int32 {
 	if debugAsserts {
 		if s < 0 {
@@ -131,9 +126,24 @@ func (e *fastBase) matchlen(s, t int32, src []byte) int32 {
 			panic(fmt.Sprintf("len(src)-s (%d) > maxCompressedBlockSize (%d)", len(src)-int(s), maxCompressedBlockSize))
 		}
 	}
+	a := src[s:]
+	b := src[t:]
+	b = b[:len(a)]
+	end := int32((len(a) >> 3) << 3)
+	for i := int32(0); i < end; i += 8 {
+		if diff := load6432(a, i) ^ load6432(b, i); diff != 0 {
+			return i + int32(bits.TrailingZeros64(diff)>>3)
+		}
+	}

-	// Extend the match to be as long as possible.
-	return int32(matchLen(src[s:], src[t:]))
+	a = a[end:]
+	b = b[end:]
+	for i := range a {
+		if a[i] != b[i] {
+			return int32(i) + end
+		}
+	}
+	return int32(len(a)) + end
 }

 // Reset the encoding table.
--- a/vendor/github.com/klauspost/compress/zstd/enc_fast.go
+++ b/vendor/github.com/klauspost/compress/zstd/enc_fast.go
@@ -6,8 +6,6 @@ package zstd

 import (
 	"fmt"
-	"math"
-	"math/bits"
 )

 const (
@@ -136,20 +134,7 @@ encodeLoop:
 				// Consider history as well.
 				var seq seq
 				var length int32
-				// length = 4 + e.matchlen(s+6, repIndex+4, src)
-				{
-					a := src[s+6:]
-					b := src[repIndex+4:]
-					endI := len(a) & (math.MaxInt32 - 7)
-					length = int32(endI) + 4
-					for i := 0; i < endI; i += 8 {
-						if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-							length = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-							break
-						}
-					}
-				}
-
+				length = 4 + e.matchlen(s+6, repIndex+4, src)
 				seq.matchLen = uint32(length - zstdMinMatch)

 				// We might be able to match backwards.
@@ -236,20 +221,7 @@ encodeLoop:
 		}

 		// Extend the 4-byte match as long as possible.
-		//l := e.matchlen(s+4, t+4, src) + 4
-		var l int32
-		{
-			a := src[s+4:]
-			b := src[t+4:]
-			endI := len(a) & (math.MaxInt32 - 7)
-			l = int32(endI) + 4
-			for i := 0; i < endI; i += 8 {
-				if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-					l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-					break
-				}
-			}
-		}
+		l := e.matchlen(s+4, t+4, src) + 4

 		// Extend backwards
 		tMin := s - e.maxMatchOff
@@ -286,20 +258,7 @@ encodeLoop:
 		if o2 := s - offset2; canRepeat && load3232(src, o2) == uint32(cv) {
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
-			//l := 4 + e.matchlen(s+4, o2+4, src)
-			var l int32
-			{
-				a := src[s+4:]
-				b := src[o2+4:]
-				endI := len(a) & (math.MaxInt32 - 7)
-				l = int32(endI) + 4
-				for i := 0; i < endI; i += 8 {
-					if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-						l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-						break
-					}
-				}
-			}
+			l := 4 + e.matchlen(s+4, o2+4, src)

 			// Store this, since we have it.
 			nextHash := hashLen(cv, hashLog, tableFastHashLen)
@@ -418,21 +377,7 @@ encodeLoop:
 			if len(blk.sequences) > 2 && load3232(src, repIndex) == uint32(cv>>16) {
 				// Consider history as well.
 				var seq seq
-				// length := 4 + e.matchlen(s+6, repIndex+4, src)
-				// length := 4 + int32(matchLen(src[s+6:], src[repIndex+4:]))
-				var length int32
-				{
-					a := src[s+6:]
-					b := src[repIndex+4:]
-					endI := len(a) & (math.MaxInt32 - 7)
-					length = int32(endI) + 4
-					for i := 0; i < endI; i += 8 {
-						if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-							length = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-							break
-						}
-					}
-				}
+				length := 4 + e.matchlen(s+6, repIndex+4, src)

 				seq.matchLen = uint32(length - zstdMinMatch)

@@ -522,21 +467,7 @@ encodeLoop:
 			panic(fmt.Sprintf("t (%d) < 0 ", t))
 		}
 		// Extend the 4-byte match as long as possible.
-		//l := e.matchlenNoHist(s+4, t+4, src) + 4
-		// l := int32(matchLen(src[s+4:], src[t+4:])) + 4
-		var l int32
-		{
-			a := src[s+4:]
-			b := src[t+4:]
-			endI := len(a) & (math.MaxInt32 - 7)
-			l = int32(endI) + 4
-			for i := 0; i < endI; i += 8 {
-				if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-					l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-					break
-				}
-			}
-		}
+		l := e.matchlen(s+4, t+4, src) + 4

 		// Extend backwards
 		tMin := s - e.maxMatchOff
@@ -573,21 +504,7 @@ encodeLoop:
 		if o2 := s - offset2; len(blk.sequences) > 2 && load3232(src, o2) == uint32(cv) {
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
-			//l := 4 + e.matchlenNoHist(s+4, o2+4, src)
-			// l := 4 + int32(matchLen(src[s+4:], src[o2+4:]))
-			var l int32
-			{
-				a := src[s+4:]
-				b := src[o2+4:]
-				endI := len(a) & (math.MaxInt32 - 7)
-				l = int32(endI) + 4
-				for i := 0; i < endI; i += 8 {
-					if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-						l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-						break
-					}
-				}
-			}
+			l := 4 + e.matchlen(s+4, o2+4, src)

 			// Store this, since we have it.
 			nextHash := hashLen(cv, hashLog, tableFastHashLen)
@@ -731,19 +648,7 @@ encodeLoop:
 				// Consider history as well.
 				var seq seq
 				var length int32
-				// length = 4 + e.matchlen(s+6, repIndex+4, src)
-				{
-					a := src[s+6:]
-					b := src[repIndex+4:]
-					endI := len(a) & (math.MaxInt32 - 7)
-					length = int32(endI) + 4
-					for i := 0; i < endI; i += 8 {
-						if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-							length = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-							break
-						}
-					}
-				}
+				length = 4 + e.matchlen(s+6, repIndex+4, src)

 				seq.matchLen = uint32(length - zstdMinMatch)

@@ -831,20 +736,7 @@ encodeLoop:
 		}

 		// Extend the 4-byte match as long as possible.
-		//l := e.matchlen(s+4, t+4, src) + 4
-		var l int32
-		{
-			a := src[s+4:]
-			b := src[t+4:]
-			endI := len(a) & (math.MaxInt32 - 7)
-			l = int32(endI) + 4
-			for i := 0; i < endI; i += 8 {
-				if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-					l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-					break
-				}
-			}
-		}
+		l := e.matchlen(s+4, t+4, src) + 4

 		// Extend backwards
 		tMin := s - e.maxMatchOff
@@ -881,20 +773,7 @@ encodeLoop:
 		if o2 := s - offset2; canRepeat && load3232(src, o2) == uint32(cv) {
 			// We have at least 4 byte match.
 			// No need to check backwards. We come straight from a match
-			//l := 4 + e.matchlen(s+4, o2+4, src)
-			var l int32
-			{
-				a := src[s+4:]
-				b := src[o2+4:]
-				endI := len(a) & (math.MaxInt32 - 7)
-				l = int32(endI) + 4
-				for i := 0; i < endI; i += 8 {
-					if diff := load64(a, i) ^ load64(b, i); diff != 0 {
-						l = int32(i+bits.TrailingZeros64(diff)>>3) + 4
-						break
-					}
-				}
-			}
+			l := 4 + e.matchlen(s+4, o2+4, src)

 			// Store this, since we have it.
 			nextHash := hashLen(cv, hashLog, tableFastHashLen)
--- a/vendor/github.com/klauspost/compress/zstd/encoder_options.go
+++ b/vendor/github.com/klauspost/compress/zstd/encoder_options.go
@@ -24,6 +24,7 @@ type encoderOptions struct {
 	allLitEntropy   bool
 	customWindow    bool
 	customALEntropy bool
+	customBlockSize bool
 	lowMem          bool
 	dict            *dict
 }
@@ -33,7 +34,7 @@ func (o *encoderOptions) setDefault() {
 		concurrent:    runtime.GOMAXPROCS(0),
 		crc:           true,
 		single:        nil,
-		blockSize:     1 << 16,
+		blockSize:     maxCompressedBlockSize,
 		windowSize:    8 << 20,
 		level:         SpeedDefault,
 		allLitEntropy: true,
@@ -106,6 +107,7 @@ func WithWindowSize(n int) EOption {
 		o.customWindow = true
 		if o.blockSize > o.windowSize {
 			o.blockSize = o.windowSize
+			o.customBlockSize = true
 		}
 		return nil
 	}
@@ -188,10 +190,9 @@ func EncoderLevelFromZstd(level int) EncoderLevel {
 		return SpeedDefault
 	case level >= 6 && level < 10:
 		return SpeedBetterCompression
-	case level >= 10:
+	default:
 		return SpeedBestCompression
 	}
-	return SpeedDefault
 }

 // String provides a string representation of the compression level.
@@ -222,6 +223,9 @@ func WithEncoderLevel(l EncoderLevel) EOption {
 			switch o.level {
 			case SpeedFastest:
 				o.windowSize = 4 << 20
+				if !o.customBlockSize {
+					o.blockSize = 1 << 16
+				}
 			case SpeedDefault:
 				o.windowSize = 8 << 20
 			case SpeedBetterCompression:
--- a/vendor/github.com/klauspost/compress/zstd/fse_decoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/fse_decoder.go
@@ -379,7 +379,7 @@ func (s decSymbol) final() (int, uint8) {
 // 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 := uint16(br.getBitsFast(s.state.nbBits()))
+	lowBits := br.get16BitsFast(s.state.nbBits())
 	s.state = s.dt[s.state.newState()+lowBits]
 	return s.state.baseline(), s.state.addBits()
 }
--- a/vendor/github.com/klauspost/compress/zstd/fse_encoder.go
+++ b/vendor/github.com/klauspost/compress/zstd/fse_encoder.go
@@ -62,9 +62,8 @@ func (s symbolTransform) String() string {
 // To indicate that you have populated the histogram call HistogramFinished
 // with the value of the highest populated symbol, as well as the number of entries
 // in the most populated entry. These are accepted at face value.
-// The returned slice will always be length 256.
-func (s *fseEncoder) Histogram() []uint32 {
-	return s.count[:]
+func (s *fseEncoder) Histogram() *[256]uint32 {
+	return &s.count
 }

 // HistogramFinished can be called to indicate that the histogram has been populated.
--- a/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_amd64.s
+++ b/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_amd64.s
@@ -1,6 +1,7 @@
 // +build !appengine
 // +build gc
 // +build !purego
+// +build !noasm

 #include "textflag.h"

--- a/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_arm64.s
+++ b/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_arm64.s
@@ -0,0 +1,186 @@
+// +build gc,!purego,!noasm
+
+#include "textflag.h"
+
+// Register allocation.
+#define digest	R1
+#define h	R2 // Return value.
+#define p	R3 // Input pointer.
+#define len	R4
+#define nblocks	R5 // len / 32.
+#define prime1	R7
+#define prime2	R8
+#define prime3	R9
+#define prime4	R10
+#define prime5	R11
+#define v1	R12
+#define v2	R13
+#define v3	R14
+#define v4	R15
+#define x1	R20
+#define x2	R21
+#define x3	R22
+#define x4	R23
+
+#define round(acc, x) \
+	MADD prime2, acc, x, acc \
+	ROR  $64-31, acc         \
+	MUL  prime1, acc         \
+
+// x = round(0, x).
+#define round0(x) \
+	MUL prime2, x \
+	ROR $64-31, x \
+	MUL prime1, x \
+
+#define mergeRound(x) \
+	round0(x)                 \
+	EOR  x, h                 \
+	MADD h, prime4, prime1, h \
+
+// Update v[1-4] with 32-byte blocks. Assumes len >= 32.
+#define blocksLoop() \
+	LSR     $5, len, nblocks \
+	PCALIGN $16              \
+	loop:                    \
+	LDP.P   32(p), (x1, x2)  \
+	round(v1, x1)            \
+	LDP     -16(p), (x3, x4) \
+	round(v2, x2)            \
+	SUB     $1, nblocks      \
+	round(v3, x3)            \
+	round(v4, x4)            \
+	CBNZ    nblocks, loop    \
+
+// The primes are repeated here to ensure that they're stored
+// in a contiguous array, so we can load them with LDP.
+DATA primes<> +0(SB)/8, $11400714785074694791
+DATA primes<> +8(SB)/8, $14029467366897019727
+DATA primes<>+16(SB)/8, $1609587929392839161
+DATA primes<>+24(SB)/8, $9650029242287828579
+DATA primes<>+32(SB)/8, $2870177450012600261
+GLOBL primes<>(SB), NOPTR+RODATA, $40
+
+// func Sum64(b []byte) uint64
+TEXT ·Sum64(SB), NOFRAME+NOSPLIT, $0-32
+	LDP b_base+0(FP), (p, len)
+
+	LDP  primes<> +0(SB), (prime1, prime2)
+	LDP  primes<>+16(SB), (prime3, prime4)
+	MOVD primes<>+32(SB), prime5
+
+	CMP  $32, len
+	CSEL LO, prime5, ZR, h // if len < 32 { h = prime5 } else { h = 0 }
+	BLO  afterLoop
+
+	ADD  prime1, prime2, v1
+	MOVD prime2, v2
+	MOVD $0, v3
+	NEG  prime1, v4
+
+	blocksLoop()
+
+	ROR $64-1, v1, x1
+	ROR $64-7, v2, x2
+	ADD x1, x2
+	ROR $64-12, v3, x3
+	ROR $64-18, v4, x4
+	ADD x3, x4
+	ADD x2, x4, h
+
+	mergeRound(v1)
+	mergeRound(v2)
+	mergeRound(v3)
+	mergeRound(v4)
+
+afterLoop:
+	ADD len, h
+
+	TBZ   $4, len, try8
+	LDP.P 16(p), (x1, x2)
+
+	round0(x1)
+	ROR  $64-27, h
+	EOR  x1 @> 64-27, h, h
+	MADD h, prime4, prime1, h
+
+	round0(x2)
+	ROR  $64-27, h
+	EOR  x2 @> 64-27, h
+	MADD h, prime4, prime1, h
+
+try8:
+	TBZ    $3, len, try4
+	MOVD.P 8(p), x1
+
+	round0(x1)
+	ROR  $64-27, h
+	EOR  x1 @> 64-27, h
+	MADD h, prime4, prime1, h
+
+try4:
+	TBZ     $2, len, try2
+	MOVWU.P 4(p), x2
+
+	MUL  prime1, x2
+	ROR  $64-23, h
+	EOR  x2 @> 64-23, h
+	MADD h, prime3, prime2, h
+
+try2:
+	TBZ     $1, len, try1
+	MOVHU.P 2(p), x3
+	AND     $255, x3, x1
+	LSR     $8, x3, x2
+
+	MUL prime5, x1
+	ROR $64-11, h
+	EOR x1 @> 64-11, h
+	MUL prime1, h
+
+	MUL prime5, x2
+	ROR $64-11, h
+	EOR x2 @> 64-11, h
+	MUL prime1, h
+
+try1:
+	TBZ   $0, len, end
+	MOVBU (p), x4
+
+	MUL prime5, x4
+	ROR $64-11, h
+	EOR x4 @> 64-11, h
+	MUL prime1, h
+
+end:
+	EOR h >> 33, h
+	MUL prime2, h
+	EOR h >> 29, h
+	MUL prime3, h
+	EOR h >> 32, h
+
+	MOVD h, ret+24(FP)
+	RET
+
+// func writeBlocks(d *Digest, b []byte) int
+//
+// Assumes len(b) >= 32.
+TEXT ·writeBlocks(SB), NOFRAME+NOSPLIT, $0-40
+	LDP primes<>(SB), (prime1, prime2)
+
+	// Load state. Assume v[1-4] are stored contiguously.
+	MOVD d+0(FP), digest
+	LDP  0(digest), (v1, v2)
+	LDP  16(digest), (v3, v4)
+
+	LDP b_base+8(FP), (p, len)
+
+	blocksLoop()
+
+	// Store updated state.
+	STP (v1, v2), 0(digest)
+	STP (v3, v4), 16(digest)
+
+	BIC  $31, len
+	MOVD len, ret+32(FP)
+	RET
--- a/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_amd64.go
+++ b/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_amd64.go
@@ -1,5 +1,9 @@
-//go:build !appengine && gc && !purego
-// +build !appengine,gc,!purego
+//go:build (amd64 || arm64) && !appengine && gc && !purego && !noasm
+// +build amd64 arm64
+// +build !appengine
+// +build gc
+// +build !purego
+// +build !noasm

 package xxhash

--- a/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_other.go
+++ b/vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_other.go
@@ -1,5 +1,5 @@
-//go:build !amd64 || appengine || !gc || purego
-// +build !amd64 appengine !gc purego
+//go:build (!amd64 && !arm64) || appengine || !gc || purego || noasm
+// +build !amd64,!arm64 appengine !gc purego noasm

 package xxhash

--- a/vendor/github.com/klauspost/compress/zstd/seqdec.go
+++ b/vendor/github.com/klauspost/compress/zstd/seqdec.go
@@ -278,7 +278,7 @@ func (s *sequenceDecs) decode(seqs int, br *bitReader, hist []byte) error {
 			mlState = mlTable[mlState.newState()&maxTableMask]
 			ofState = ofTable[ofState.newState()&maxTableMask]
 		} else {
-			bits := br.getBitsFast(nBits)
+			bits := br.get32BitsFast(nBits)
 			lowBits := uint16(bits >> ((ofState.nbBits() + mlState.nbBits()) & 31))
 			llState = llTable[(llState.newState()+lowBits)&maxTableMask]

@@ -326,7 +326,7 @@ func (s *sequenceDecs) updateAlt(br *bitReader) {
 		s.offsets.state.state = s.offsets.state.dt[c.newState()]
 		return
 	}
-	bits := br.getBitsFast(nBits)
+	bits := br.get32BitsFast(nBits)
 	lowBits := uint16(bits >> ((c.nbBits() + b.nbBits()) & 31))
 	s.litLengths.state.state = s.litLengths.state.dt[a.newState()+lowBits]

--- a/vendor/github.com/klauspost/cpuid/v2/README.md
+++ b/vendor/github.com/klauspost/cpuid/v2/README.md
@@ -39,10 +39,10 @@ func main() {
 	fmt.Println("ThreadsPerCore:", CPU.ThreadsPerCore)
 	fmt.Println("LogicalCores:", CPU.LogicalCores)
 	fmt.Println("Family", CPU.Family, "Model:", CPU.Model, "Vendor ID:", CPU.VendorID)
-	fmt.Println("Features:", fmt.Sprintf(strings.Join(CPU.FeatureSet(), ",")))
+	fmt.Println("Features:", strings.Join(CPU.FeatureSet(), ","))
 	fmt.Println("Cacheline bytes:", CPU.CacheLine)
 	fmt.Println("L1 Data Cache:", CPU.Cache.L1D, "bytes")
-	fmt.Println("L1 Instruction Cache:", CPU.Cache.L1D, "bytes")
+	fmt.Println("L1 Instruction Cache:", CPU.Cache.L1I, "bytes")
 	fmt.Println("L2 Cache:", CPU.Cache.L2, "bytes")
 	fmt.Println("L3 Cache:", CPU.Cache.L3, "bytes")
 	fmt.Println("Frequency", CPU.Hz, "hz")
--- a/vendor/github.com/klauspost/cpuid/v2/cpuid.go
+++ b/vendor/github.com/klauspost/cpuid/v2/cpuid.go
@@ -95,10 +95,13 @@ const (
 	AVXSLOW                             // Indicates the CPU performs 2 128 bit operations instead of one.
 	BMI1                                // Bit Manipulation Instruction Set 1
 	BMI2                                // Bit Manipulation Instruction Set 2
+	CETIBT                              // Intel CET Indirect Branch Tracking
+	CETSS                               // Intel CET Shadow Stack
 	CLDEMOTE                            // Cache Line Demote
 	CLMUL                               // Carry-less Multiplication
 	CLZERO                              // CLZERO instruction supported
 	CMOV                                // i686 CMOV
+	CMPXCHG8                            // CMPXCHG8 instruction
 	CPBOOST                             // Core Performance Boost
 	CX16                                // CMPXCHG16B Instruction
 	ENQCMD                              // Enqueue Command
@@ -106,6 +109,8 @@ const (
 	F16C                                // Half-precision floating-point conversion
 	FMA3                                // Intel FMA 3. Does not imply AVX.
 	FMA4                                // Bulldozer FMA4 functions
+	FXSR                                // FXSAVE, FXRESTOR instructions, CR4 bit 9
+	FXSROPT                             // FXSAVE/FXRSTOR optimizations
 	GFNI                                // Galois Field New Instructions
 	HLE                                 // Hardware Lock Elision
 	HTT                                 // Hyperthreading (enabled)
@@ -123,16 +128,19 @@ const (
 	IBSRIPINVALIDCHK                    // Instruction Based Sampling Feature (AMD)
 	INT_WBINVD                          // WBINVD/WBNOINVD are interruptible.
 	INVLPGB                             // NVLPGB and TLBSYNC instruction supported
+	LAHF                                // LAHF/SAHF in long mode
 	LZCNT                               // LZCNT instruction
 	MCAOVERFLOW                         // MCA overflow recovery support.
 	MCOMMIT                             // MCOMMIT instruction supported
 	MMX                                 // standard MMX
 	MMXEXT                              // SSE integer functions or AMD MMX ext
+	MOVBE                               // MOVBE instruction (big-endian)
 	MOVDIR64B                           // Move 64 Bytes as Direct Store
 	MOVDIRI                             // Move Doubleword as Direct Store
 	MPX                                 // Intel MPX (Memory Protection Extensions)
 	MSRIRC                              // Instruction Retired Counter MSR available
 	NX                                  // NX (No-Execute) bit
+	OSXSAVE                             // XSAVE enabled by OS
 	POPCNT                              // POPCNT instruction
 	RDPRU                               // RDPRU instruction supported
 	RDRAND                              // RDRAND instruction is available
@@ -140,6 +148,7 @@ const (
 	RDTSCP                              // RDTSCP Instruction
 	RTM                                 // Restricted Transactional Memory
 	RTM_ALWAYS_ABORT                    // Indicates that the loaded microcode is forcing RTM abort.
+	SCE                                 // SYSENTER and SYSEXIT instructions
 	SERIALIZE                           // Serialize Instruction Execution
 	SGX                                 // Software Guard Extensions
 	SGXLC                               // Software Guard Extensions Launch Control
@@ -160,7 +169,9 @@ const (
 	VPCLMULQDQ                          // Carry-Less Multiplication Quadword
 	WAITPKG                             // TPAUSE, UMONITOR, UMWAIT
 	WBNOINVD                            // Write Back and Do Not Invalidate Cache
+	X87                                 // FPU
 	XOP                                 // Bulldozer XOP functions
+	XSAVE                               // XSAVE, XRESTOR, XSETBV, XGETBV

 	// ARM features:
 	AESARM   // AES instructions
@@ -311,6 +322,31 @@ func (c CPUInfo) Has(id FeatureID) bool {
 	return c.featureSet.inSet(id)
 }

+// https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels
+var level1Features = flagSetWith(CMOV, CMPXCHG8, X87, FXSR, MMX, SCE, SSE, SSE2)
+var level2Features = flagSetWith(CMOV, CMPXCHG8, X87, FXSR, MMX, SCE, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3)
+var level3Features = flagSetWith(CMOV, CMPXCHG8, X87, FXSR, MMX, SCE, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3, AVX, AVX2, BMI1, BMI2, F16C, FMA3, LZCNT, MOVBE, OSXSAVE)
+var level4Features = flagSetWith(CMOV, CMPXCHG8, X87, FXSR, MMX, SCE, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3, AVX, AVX2, BMI1, BMI2, F16C, FMA3, LZCNT, MOVBE, OSXSAVE, AVX512F, AVX512BW, AVX512CD, AVX512DQ, AVX512VL)
+
+// X64Level returns the microarchitecture level detected on the CPU.
+// If features are lacking or non x64 mode, 0 is returned.
+// See https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels
+func (c CPUInfo) X64Level() int {
+	if c.featureSet.hasSet(level4Features) {
+		return 4
+	}
+	if c.featureSet.hasSet(level3Features) {
+		return 3
+	}
+	if c.featureSet.hasSet(level2Features) {
+		return 2
+	}
+	if c.featureSet.hasSet(level1Features) {
+		return 1
+	}
+	return 0
+}
+
 // Disable will disable one or several features.
 func (c *CPUInfo) Disable(ids ...FeatureID) bool {
 	for _, id := range ids {
@@ -335,9 +371,7 @@ func (c CPUInfo) IsVendor(v Vendor) bool {

 func (c CPUInfo) FeatureSet() []string {
 	s := make([]string, 0)
-	for _, f := range c.featureSet.Strings() {
-		s = append(s, f)
-	}
+	s = append(s, c.featureSet.Strings()...)
 	return s
 }

@@ -499,6 +533,24 @@ func (s *flagSet) or(other flagSet) {
 	}
 }

+// hasSet returns whether all features are present.
+func (s flagSet) hasSet(other flagSet) bool {
+	for i, v := range other[:] {
+		if s[i]&v != v {
+			return false
+		}
+	}
+	return true
+}
+
+func flagSetWith(feat ...FeatureID) flagSet {
+	var res flagSet
+	for _, f := range feat {
+		res.set(f)
+	}
+	return res
+}
+
 // ParseFeature will parse the string and return the ID of the matching feature.
 // Will return UNKNOWN if not found.
 func ParseFeature(s string) FeatureID {
@@ -708,6 +760,7 @@ func (c *CPUInfo) cacheSize() {
 		if maxFunctionID() < 4 {
 			return
 		}
+		c.Cache.L1I, c.Cache.L1D, c.Cache.L2, c.Cache.L3 = 0, 0, 0, 0
 		for i := uint32(0); ; i++ {
 			eax, ebx, ecx, _ := cpuidex(4, i)
 			cacheType := eax & 15
@@ -800,8 +853,6 @@ func (c *CPUInfo) cacheSize() {
 			}
 		}
 	}
-
-	return
 }

 type SGXEPCSection struct {
@@ -865,9 +916,14 @@ func support() flagSet {
 	family, model := familyModel()

 	_, _, c, d := cpuid(1)
+	fs.setIf((d&(1<<0)) != 0, X87)
+	fs.setIf((d&(1<<8)) != 0, CMPXCHG8)
+	fs.setIf((d&(1<<11)) != 0, SCE)
 	fs.setIf((d&(1<<15)) != 0, CMOV)
+	fs.setIf((d&(1<<22)) != 0, MMXEXT)
 	fs.setIf((d&(1<<23)) != 0, MMX)
-	fs.setIf((d&(1<<25)) != 0, MMXEXT)
+	fs.setIf((d&(1<<24)) != 0, FXSR)
+	fs.setIf((d&(1<<25)) != 0, FXSROPT)
 	fs.setIf((d&(1<<25)) != 0, SSE)
 	fs.setIf((d&(1<<26)) != 0, SSE2)
 	fs.setIf((c&1) != 0, SSE3)
@@ -877,6 +933,7 @@ func support() flagSet {
 	fs.setIf((c&0x00100000) != 0, SSE42)
 	fs.setIf((c&(1<<25)) != 0, AESNI)
 	fs.setIf((c&(1<<1)) != 0, CLMUL)
+	fs.setIf(c&(1<<22) != 0, MOVBE)
 	fs.setIf(c&(1<<23) != 0, POPCNT)
 	fs.setIf(c&(1<<30) != 0, RDRAND)

@@ -892,6 +949,8 @@ func support() flagSet {
 	if vend == AMD && (d&(1<<28)) != 0 && mfi >= 4 {
 		fs.setIf(threadsPerCore() > 1, HTT)
 	}
+	fs.setIf(c&1<<26 != 0, XSAVE)
+	fs.setIf(c&1<<27 != 0, OSXSAVE)
 	// Check XGETBV/XSAVE (26), OXSAVE (27) and AVX (28) bits
 	const avxCheck = 1<<26 | 1<<27 | 1<<28
 	if c&avxCheck == avxCheck {
@@ -936,6 +995,7 @@ func support() flagSet {
 		fs.setIf(ebx&(1<<29) != 0, SHA)
 		// CPUID.(EAX=7, ECX=0).ECX
 		fs.setIf(ecx&(1<<5) != 0, WAITPKG)
+		fs.setIf(ecx&(1<<7) != 0, CETSS)
 		fs.setIf(ecx&(1<<25) != 0, CLDEMOTE)
 		fs.setIf(ecx&(1<<27) != 0, MOVDIRI)
 		fs.setIf(ecx&(1<<28) != 0, MOVDIR64B)
@@ -945,6 +1005,7 @@ func support() flagSet {
 		fs.setIf(edx&(1<<11) != 0, RTM_ALWAYS_ABORT)
 		fs.setIf(edx&(1<<14) != 0, SERIALIZE)
 		fs.setIf(edx&(1<<16) != 0, TSXLDTRK)
+		fs.setIf(edx&(1<<20) != 0, CETIBT)
 		fs.setIf(edx&(1<<26) != 0, IBPB)
 		fs.setIf(edx&(1<<27) != 0, STIBP)

@@ -996,6 +1057,7 @@ func support() flagSet {
 			fs.set(LZCNT)
 			fs.set(POPCNT)
 		}
+		fs.setIf((c&(1<<0)) != 0, LAHF)
 		fs.setIf((c&(1<<10)) != 0, IBS)
 		fs.setIf((d&(1<<31)) != 0, AMD3DNOW)
 		fs.setIf((d&(1<<30)) != 0, AMD3DNOWEXT)
--- a/vendor/github.com/klauspost/cpuid/v2/detect_arm64.go
+++ b/vendor/github.com/klauspost/cpuid/v2/detect_arm64.go
@@ -1,6 +1,7 @@
 // Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file.

-//+build arm64,!gccgo,!noasm,!appengine
+//go:build arm64 && !gccgo && !noasm && !appengine
+// +build arm64,!gccgo,!noasm,!appengine

 package cpuid

--- a/vendor/github.com/klauspost/cpuid/v2/detect_ref.go
+++ b/vendor/github.com/klauspost/cpuid/v2/detect_ref.go
@@ -1,6 +1,7 @@
 // Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file.

-//+build !amd64,!386,!arm64 gccgo noasm appengine
+//go:build (!amd64 && !386 && !arm64) || gccgo || noasm || appengine
+// +build !amd64,!386,!arm64 gccgo noasm appengine

 package cpuid

--- a/vendor/github.com/klauspost/cpuid/v2/detect_x86.go
+++ b/vendor/github.com/klauspost/cpuid/v2/detect_x86.go
@@ -1,6 +1,7 @@
 // Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file.

-//+build 386,!gccgo,!noasm,!appengine amd64,!gccgo,!noasm,!appengine
+//go:build (386 && !gccgo && !noasm && !appengine) || (amd64 && !gccgo && !noasm && !appengine)
+// +build 386,!gccgo,!noasm,!appengine amd64,!gccgo,!noasm,!appengine

 package cpuid

--- a/vendor/github.com/klauspost/cpuid/v2/featureid_string.go
+++ b/vendor/github.com/klauspost/cpuid/v2/featureid_string.go
@@ -36,103 +36,114 @@ func _() {
 	_ = x[AVXSLOW-26]
 	_ = x[BMI1-27]
 	_ = x[BMI2-28]
-	_ = x[CLDEMOTE-29]
-	_ = x[CLMUL-30]
-	_ = x[CLZERO-31]
-	_ = x[CMOV-32]
-	_ = x[CPBOOST-33]
-	_ = x[CX16-34]
-	_ = x[ENQCMD-35]
-	_ = x[ERMS-36]
-	_ = x[F16C-37]
-	_ = x[FMA3-38]
-	_ = x[FMA4-39]
-	_ = x[GFNI-40]
-	_ = x[HLE-41]
-	_ = x[HTT-42]
-	_ = x[HWA-43]
-	_ = x[HYPERVISOR-44]
-	_ = x[IBPB-45]
-	_ = x[IBS-46]
-	_ = x[IBSBRNTRGT-47]
-	_ = x[IBSFETCHSAM-48]
-	_ = x[IBSFFV-49]
-	_ = x[IBSOPCNT-50]
-	_ = x[IBSOPCNTEXT-51]
-	_ = x[IBSOPSAM-52]
-	_ = x[IBSRDWROPCNT-53]
-	_ = x[IBSRIPINVALIDCHK-54]
-	_ = x[INT_WBINVD-55]
-	_ = x[INVLPGB-56]
-	_ = x[LZCNT-57]
-	_ = x[MCAOVERFLOW-58]
-	_ = x[MCOMMIT-59]
-	_ = x[MMX-60]
-	_ = x[MMXEXT-61]
-	_ = x[MOVDIR64B-62]
-	_ = x[MOVDIRI-63]
-	_ = x[MPX-64]
-	_ = x[MSRIRC-65]
-	_ = x[NX-66]
-	_ = x[POPCNT-67]
-	_ = x[RDPRU-68]
-	_ = x[RDRAND-69]
-	_ = x[RDSEED-70]
-	_ = x[RDTSCP-71]
-	_ = x[RTM-72]
-	_ = x[RTM_ALWAYS_ABORT-73]
-	_ = x[SERIALIZE-74]
-	_ = x[SGX-75]
-	_ = x[SGXLC-76]
-	_ = x[SHA-77]
-	_ = x[SSE-78]
-	_ = x[SSE2-79]
-	_ = x[SSE3-80]
-	_ = x[SSE4-81]
-	_ = x[SSE42-82]
-	_ = x[SSE4A-83]
-	_ = x[SSSE3-84]
-	_ = x[STIBP-85]
-	_ = x[SUCCOR-86]
-	_ = x[TBM-87]
-	_ = x[TSXLDTRK-88]
-	_ = x[VAES-89]
-	_ = x[VMX-90]
-	_ = x[VPCLMULQDQ-91]
-	_ = x[WAITPKG-92]
-	_ = x[WBNOINVD-93]
-	_ = x[XOP-94]
-	_ = x[AESARM-95]
-	_ = x[ARMCPUID-96]
-	_ = x[ASIMD-97]
-	_ = x[ASIMDDP-98]
-	_ = x[ASIMDHP-99]
-	_ = x[ASIMDRDM-100]
-	_ = x[ATOMICS-101]
-	_ = x[CRC32-102]
-	_ = x[DCPOP-103]
-	_ = x[EVTSTRM-104]
-	_ = x[FCMA-105]
-	_ = x[FP-106]
-	_ = x[FPHP-107]
-	_ = x[GPA-108]
-	_ = x[JSCVT-109]
-	_ = x[LRCPC-110]
-	_ = x[PMULL-111]
-	_ = x[SHA1-112]
-	_ = x[SHA2-113]
-	_ = x[SHA3-114]
-	_ = x[SHA512-115]
-	_ = x[SM3-116]
-	_ = x[SM4-117]
-	_ = x[SVE-118]
-	_ = x[lastID-119]
+	_ = x[CETIBT-29]
+	_ = x[CETSS-30]
+	_ = x[CLDEMOTE-31]
+	_ = x[CLMUL-32]
+	_ = x[CLZERO-33]
+	_ = x[CMOV-34]
+	_ = x[CMPXCHG8-35]
+	_ = x[CPBOOST-36]
+	_ = x[CX16-37]
+	_ = x[ENQCMD-38]
+	_ = x[ERMS-39]
+	_ = x[F16C-40]
+	_ = x[FMA3-41]
+	_ = x[FMA4-42]
+	_ = x[FXSR-43]
+	_ = x[FXSROPT-44]
+	_ = x[GFNI-45]
+	_ = x[HLE-46]
+	_ = x[HTT-47]
+	_ = x[HWA-48]
+	_ = x[HYPERVISOR-49]
+	_ = x[IBPB-50]
+	_ = x[IBS-51]
+	_ = x[IBSBRNTRGT-52]
+	_ = x[IBSFETCHSAM-53]
+	_ = x[IBSFFV-54]
+	_ = x[IBSOPCNT-55]
+	_ = x[IBSOPCNTEXT-56]
+	_ = x[IBSOPSAM-57]
+	_ = x[IBSRDWROPCNT-58]
+	_ = x[IBSRIPINVALIDCHK-59]
+	_ = x[INT_WBINVD-60]
+	_ = x[INVLPGB-61]
+	_ = x[LAHF-62]
+	_ = x[LZCNT-63]
+	_ = x[MCAOVERFLOW-64]
+	_ = x[MCOMMIT-65]
+	_ = x[MMX-66]
+	_ = x[MMXEXT-67]
+	_ = x[MOVBE-68]
+	_ = x[MOVDIR64B-69]
+	_ = x[MOVDIRI-70]
+	_ = x[MPX-71]
+	_ = x[MSRIRC-72]
+	_ = x[NX-73]
+	_ = x[OSXSAVE-74]
+	_ = x[POPCNT-75]
+	_ = x[RDPRU-76]
+	_ = x[RDRAND-77]
+	_ = x[RDSEED-78]
+	_ = x[RDTSCP-79]
+	_ = x[RTM-80]
+	_ = x[RTM_ALWAYS_ABORT-81]
+	_ = x[SCE-82]
+	_ = x[SERIALIZE-83]
+	_ = x[SGX-84]
+	_ = x[SGXLC-85]
+	_ = x[SHA-86]
+	_ = x[SSE-87]
+	_ = x[SSE2-88]
+	_ = x[SSE3-89]
+	_ = x[SSE4-90]
+	_ = x[SSE42-91]
+	_ = x[SSE4A-92]
+	_ = x[SSSE3-93]
+	_ = x[STIBP-94]
+	_ = x[SUCCOR-95]
+	_ = x[TBM-96]
+	_ = x[TSXLDTRK-97]
+	_ = x[VAES-98]
+	_ = x[VMX-99]
+	_ = x[VPCLMULQDQ-100]
+	_ = x[WAITPKG-101]
+	_ = x[WBNOINVD-102]
+	_ = x[X87-103]
+	_ = x[XOP-104]
+	_ = x[XSAVE-105]
+	_ = x[AESARM-106]
+	_ = x[ARMCPUID-107]
+	_ = x[ASIMD-108]
+	_ = x[ASIMDDP-109]
+	_ = x[ASIMDHP-110]
+	_ = x[ASIMDRDM-111]
+	_ = x[ATOMICS-112]
+	_ = x[CRC32-113]
+	_ = x[DCPOP-114]
+	_ = x[EVTSTRM-115]
+	_ = x[FCMA-116]
+	_ = x[FP-117]
+	_ = x[FPHP-118]
+	_ = x[GPA-119]
+	_ = x[JSCVT-120]
+	_ = x[LRCPC-121]
+	_ = x[PMULL-122]
+	_ = x[SHA1-123]
+	_ = x[SHA2-124]
+	_ = x[SHA3-125]
+	_ = x[SHA512-126]
+	_ = x[SM3-127]
+	_ = x[SM4-128]
+	_ = x[SVE-129]
+	_ = x[lastID-130]
 	_ = x[firstID-0]
 }

-const _FeatureID_name = "firstIDADXAESNIAMD3DNOWAMD3DNOWEXTAMXBF16AMXINT8AMXTILEAVXAVX2AVX512BF16AVX512BITALGAVX512BWAVX512CDAVX512DQAVX512ERAVX512FAVX512FP16AVX512IFMAAVX512PFAVX512VBMIAVX512VBMI2AVX512VLAVX512VNNIAVX512VP2INTERSECTAVX512VPOPCNTDQAVXSLOWBMI1BMI2CLDEMOTECLMULCLZEROCMOVCPBOOSTCX16ENQCMDERMSF16CFMA3FMA4GFNIHLEHTTHWAHYPERVISORIBPBIBSIBSBRNTRGTIBSFETCHSAMIBSFFVIBSOPCNTIBSOPCNTEXTIBSOPSAMIBSRDWROPCNTIBSRIPINVALIDCHKINT_WBINVDINVLPGBLZCNTMCAOVERFLOWMCOMMITMMXMMXEXTMOVDIR64BMOVDIRIMPXMSRIRCNXPOPCNTRDPRURDRANDRDSEEDRDTSCPRTMRTM_ALWAYS_ABORTSERIALIZESGXSGXLCSHASSESSE2SSE3SSE4SSE42SSE4ASSSE3STIBPSUCCORTBMTSXLDTRKVAESVMXVPCLMULQDQWAITPKGWBNOINVDXOPAESARMARMCPUIDASIMDASIMDDPASIMDHPASIMDRDMATOMICSCRC32DCPOPEVTSTRMFCMAFPFPHPGPAJSCVTLRCPCPMULLSHA1SHA2SHA3SHA512SM3SM4SVElastID"
+const _FeatureID_name = "firstIDADXAESNIAMD3DNOWAMD3DNOWEXTAMXBF16AMXINT8AMXTILEAVXAVX2AVX512BF16AVX512BITALGAVX512BWAVX512CDAVX512DQAVX512ERAVX512FAVX512FP16AVX512IFMAAVX512PFAVX512VBMIAVX512VBMI2AVX512VLAVX512VNNIAVX512VP2INTERSECTAVX512VPOPCNTDQAVXSLOWBMI1BMI2CETIBTCETSSCLDEMOTECLMULCLZEROCMOVCMPXCHG8CPBOOSTCX16ENQCMDERMSF16CFMA3FMA4FXSRFXSROPTGFNIHLEHTTHWAHYPERVISORIBPBIBSIBSBRNTRGTIBSFETCHSAMIBSFFVIBSOPCNTIBSOPCNTEXTIBSOPSAMIBSRDWROPCNTIBSRIPINVALIDCHKINT_WBINVDINVLPGBLAHFLZCNTMCAOVERFLOWMCOMMITMMXMMXEXTMOVBEMOVDIR64BMOVDIRIMPXMSRIRCNXOSXSAVEPOPCNTRDPRURDRANDRDSEEDRDTSCPRTMRTM_ALWAYS_ABORTSCESERIALIZESGXSGXLCSHASSESSE2SSE3SSE4SSE42SSE4ASSSE3STIBPSUCCORTBMTSXLDTRKVAESVMXVPCLMULQDQWAITPKGWBNOINVDX87XOPXSAVEAESARMARMCPUIDASIMDASIMDDPASIMDHPASIMDRDMATOMICSCRC32DCPOPEVTSTRMFCMAFPFPHPGPAJSCVTLRCPCPMULLSHA1SHA2SHA3SHA512SM3SM4SVElastID"

-var _FeatureID_index = [...]uint16{0, 7, 10, 15, 23, 34, 41, 48, 55, 58, 62, 72, 84, 92, 100, 108, 116, 123, 133, 143, 151, 161, 172, 180, 190, 208, 223, 230, 234, 238, 246, 251, 257, 261, 268, 272, 278, 282, 286, 290, 294, 298, 301, 304, 307, 317, 321, 324, 334, 345, 351, 359, 370, 378, 390, 406, 416, 423, 428, 439, 446, 449, 455, 464, 471, 474, 480, 482, 488, 493, 499, 505, 511, 514, 530, 539, 542, 547, 550, 553, 557, 561, 565, 570, 575, 580, 585, 591, 594, 602, 606, 609, 619, 626, 634, 637, 643, 651, 656, 663, 670, 678, 685, 690, 695, 702, 706, 708, 712, 715, 720, 725, 730, 734, 738, 742, 748, 751, 754, 757, 763}
+var _FeatureID_index = [...]uint16{0, 7, 10, 15, 23, 34, 41, 48, 55, 58, 62, 72, 84, 92, 100, 108, 116, 123, 133, 143, 151, 161, 172, 180, 190, 208, 223, 230, 234, 238, 244, 249, 257, 262, 268, 272, 280, 287, 291, 297, 301, 305, 309, 313, 317, 324, 328, 331, 334, 337, 347, 351, 354, 364, 375, 381, 389, 400, 408, 420, 436, 446, 453, 457, 462, 473, 480, 483, 489, 494, 503, 510, 513, 519, 521, 528, 534, 539, 545, 551, 557, 560, 576, 579, 588, 591, 596, 599, 602, 606, 610, 614, 619, 624, 629, 634, 640, 643, 651, 655, 658, 668, 675, 683, 686, 689, 694, 700, 708, 713, 720, 727, 735, 742, 747, 752, 759, 763, 765, 769, 772, 777, 782, 787, 791, 795, 799, 805, 808, 811, 814, 820}

 func (i FeatureID) String() string {
 	if i < 0 || i >= FeatureID(len(_FeatureID_index)-1) {
--- a/vendor/github.com/klauspost/cpuid/v2/go.mod
+++ b/vendor/github.com/klauspost/cpuid/v2/go.mod
@@ -1,3 +1,3 @@
 module github.com/klauspost/cpuid/v2

-go 1.13
+go 1.15
--- a/vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go
+++ b/vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go
@@ -1,8 +1,7 @@
 // Copyright (c) 2020 Klaus Post, released under MIT License. See LICENSE file.

-// +build arm64
-// +build !linux
-// +build !darwin
+//go:build arm64 && !linux && !darwin
+// +build arm64,!linux,!darwin

 package cpuid

--- a/vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go
+++ b/vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go
@@ -1,6 +1,7 @@
 // Copyright (c) 2021 Klaus Post, released under MIT License. See LICENSE file.

-//+build nounsafe
+//go:build nounsafe
+// +build nounsafe

 package cpuid

--- a/vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go
+++ b/vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go
@@ -1,6 +1,7 @@
 // Copyright (c) 2021 Klaus Post, released under MIT License. See LICENSE file.

-//+build !nounsafe
+//go:build !nounsafe
+// +build !nounsafe

 package cpuid