vendor: github.com/klauspost/compress v1.17.9

full diff: https://github.com/klauspost/compress/compare/v1.17.4...v1.17.9

Signed-off-by: Sebastiaan van Stijn <github@gone.nl>

vendor/github.com/klauspost/compress/zstd/blockdec.go

@@ -554,6 +554,9 @@ func (b *blockDec) prepareSequences(in []byte, hist *history) (err error) {
 		if debugDecoder {
 			printf("Compression modes: 0b%b", compMode)
 		}
+		if compMode&3 != 0 {
+			return errors.New("corrupt block: reserved bits not zero")
+		}
 		for i := uint(0); i < 3; i++ {
 			mode := seqCompMode((compMode >> (6 - i*2)) & 3)
 			if debugDecoder {

vendor/github.com/klauspost/compress/zstd/blockenc.go

@@ -427,6 +427,16 @@ func (b *blockEnc) encodeLits(lits []byte, raw bool) error {
 	return nil
 }
 
+// encodeRLE will encode an RLE block.
+func (b *blockEnc) encodeRLE(val byte, length uint32) {
+	var bh blockHeader
+	bh.setLast(b.last)
+	bh.setSize(length)
+	bh.setType(blockTypeRLE)
+	b.output = bh.appendTo(b.output)
+	b.output = append(b.output, val)
+}
+
 // fuzzFseEncoder can be used to fuzz the FSE encoder.
 func fuzzFseEncoder(data []byte) int {
 	if len(data) > maxSequences || len(data) < 2 {
@@ -479,6 +489,16 @@ func (b *blockEnc) encode(org []byte, raw, rawAllLits bool) error {
 	if len(b.sequences) == 0 {
 		return b.encodeLits(b.literals, rawAllLits)
 	}
+	if len(b.sequences) == 1 && len(org) > 0 && len(b.literals) <= 1 {
+		// Check common RLE cases.
+		seq := b.sequences[0]
+		if seq.litLen == uint32(len(b.literals)) && seq.offset-3 == 1 {
+			// Offset == 1 and 0 or 1 literals.
+			b.encodeRLE(org[0], b.sequences[0].matchLen+zstdMinMatch+seq.litLen)
+			return nil
+		}
+	}
+
 	// We want some difference to at least account for the headers.
 	saved := b.size - len(b.literals) - (b.size >> 6)
 	if saved < 16 {

vendor/github.com/klauspost/compress/zstd/decodeheader.go

@@ -95,42 +95,54 @@ 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 {
+	_, err := h.DecodeAndStrip(in)
+	return err
+}
+
+// DecodeAndStrip will decode the header from the beginning of the stream
+// and on success return the remaining bytes.
+// This will decode the frame header and the first block header if enough bytes are provided.
+// It is recommended to provide at least HeaderMaxSize bytes.
+// If the frame header cannot be read an error will be returned.
+// 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) DecodeAndStrip(in []byte) (remain []byte, err error) {
 	*h = Header{}
 	if len(in) < 4 {
-		return io.ErrUnexpectedEOF
+		return nil, io.ErrUnexpectedEOF
 	}
 	h.HeaderSize += 4
 	b, in := in[:4], in[4:]
 	if string(b) != frameMagic {
 		if string(b[1:4]) != skippableFrameMagic || b[0]&0xf0 != 0x50 {
-			return ErrMagicMismatch
+			return nil, ErrMagicMismatch
 		}
 		if len(in) < 4 {
-			return io.ErrUnexpectedEOF
+			return nil, io.ErrUnexpectedEOF
 		}
 		h.HeaderSize += 4
 		h.Skippable = true
 		h.SkippableID = int(b[0] & 0xf)
 		h.SkippableSize = binary.LittleEndian.Uint32(in)
-		return nil
+		return in[4:], nil
 	}
 
 	// Read Window_Descriptor
 	// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#window_descriptor
 	if len(in) < 1 {
-		return io.ErrUnexpectedEOF
+		return nil, 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")
+		return nil, errors.New("reserved bit set on frame header")
 	}
 
 	if !h.SingleSegment {
 		if len(in) < 1 {
-			return io.ErrUnexpectedEOF
+			return nil, io.ErrUnexpectedEOF
 		}
 		var wd byte
 		wd, in = in[0], in[1:]
@@ -148,7 +160,7 @@ func (h *Header) Decode(in []byte) error {
 			size = 4
 		}
 		if len(in) < int(size) {
-			return io.ErrUnexpectedEOF
+			return nil, io.ErrUnexpectedEOF
 		}
 		b, in = in[:size], in[size:]
 		h.HeaderSize += int(size)
@@ -178,7 +190,7 @@ func (h *Header) Decode(in []byte) error {
 	if fcsSize > 0 {
 		h.HasFCS = true
 		if len(in) < fcsSize {
-			return io.ErrUnexpectedEOF
+			return nil, io.ErrUnexpectedEOF
 		}
 		b, in = in[:fcsSize], in[fcsSize:]
 		h.HeaderSize += int(fcsSize)
@@ -199,7 +211,7 @@ func (h *Header) Decode(in []byte) error {
 
 	// Frame Header done, we will not fail from now on.
 	if len(in) < 3 {
-		return nil
+		return in, nil
 	}
 	tmp := in[:3]
 	bh := uint32(tmp[0]) | (uint32(tmp[1]) << 8) | (uint32(tmp[2]) << 16)
@@ -209,7 +221,7 @@ func (h *Header) Decode(in []byte) error {
 	cSize := int(bh >> 3)
 	switch blockType {
 	case blockTypeReserved:
-		return nil
+		return in, nil
 	case blockTypeRLE:
 		h.FirstBlock.Compressed = true
 		h.FirstBlock.DecompressedSize = cSize
@@ -225,5 +237,25 @@ func (h *Header) Decode(in []byte) error {
 	}
 
 	h.FirstBlock.OK = true
-	return nil
+	return in, nil
+}
+
+// AppendTo will append the encoded header to the dst slice.
+// There is no error checking performed on the header values.
+func (h *Header) AppendTo(dst []byte) ([]byte, error) {
+	if h.Skippable {
+		magic := [4]byte{0x50, 0x2a, 0x4d, 0x18}
+		magic[0] |= byte(h.SkippableID & 0xf)
+		dst = append(dst, magic[:]...)
+		f := h.SkippableSize
+		return append(dst, uint8(f), uint8(f>>8), uint8(f>>16), uint8(f>>24)), nil
+	}
+	f := frameHeader{
+		ContentSize:   h.FrameContentSize,
+		WindowSize:    uint32(h.WindowSize),
+		SingleSegment: h.SingleSegment,
+		Checksum:      h.HasCheckSum,
+		DictID:        h.DictionaryID,
+	}
+	return f.appendTo(dst), nil
 }

vendor/github.com/klauspost/compress/zstd/decoder.go

@@ -82,7 +82,7 @@ var (
 // can run multiple concurrent stateless decodes. It is even possible to
 // use stateless decodes while a stream is being decoded.
 //
-// The Reset function can be used to initiate a new stream, which is will considerably
+// The Reset function can be used to initiate a new stream, which will considerably
 // reduce the allocations normally caused by NewReader.
 func NewReader(r io.Reader, opts ...DOption) (*Decoder, error) {
 	initPredefined()

vendor/github.com/klauspost/compress/zstd/dict.go

@@ -273,6 +273,9 @@ func BuildDict(o BuildDictOptions) ([]byte, error) {
 		enc.Encode(&block, b)
 		addValues(&remain, block.literals)
 		litTotal += len(block.literals)
+		if len(block.sequences) == 0 {
+			continue
+		}
 		seqs += len(block.sequences)
 		block.genCodes()
 		addHist(&ll, block.coders.llEnc.Histogram())
@@ -286,6 +289,9 @@ func BuildDict(o BuildDictOptions) ([]byte, error) {
 			if offset == 0 {
 				continue
 			}
+			if int(offset) >= len(o.History) {
+				continue
+			}
 			if offset > 3 {
 				newOffsets[offset-3]++
 			} else {
@@ -336,6 +342,9 @@ func BuildDict(o BuildDictOptions) ([]byte, error) {
 	if seqs/nUsed < 512 {
 		// Use 512 as minimum.
 		nUsed = seqs / 512
+		if nUsed == 0 {
+			nUsed = 1
+		}
 	}
 	copyHist := func(dst *fseEncoder, src *[256]int) ([]byte, error) {
 		hist := dst.Histogram()
@@ -358,6 +367,28 @@ func BuildDict(o BuildDictOptions) ([]byte, error) {
 			fakeLength += v
 			hist[i] = uint32(v)
 		}
+
+		// Ensure we aren't trying to represent RLE.
+		if maxCount == fakeLength {
+			for i := range hist {
+				if uint8(i) == maxSym {
+					fakeLength++
+					maxSym++
+					hist[i+1] = 1
+					if maxSym > 1 {
+						break
+					}
+				}
+				if hist[0] == 0 {
+					fakeLength++
+					hist[i] = 1
+					if maxSym > 1 {
+						break
+					}
+				}
+			}
+		}
+
 		dst.HistogramFinished(maxSym, maxCount)
 		dst.reUsed = false
 		dst.useRLE = false

vendor/github.com/klauspost/compress/zstd/enc_best.go

@@ -135,8 +135,20 @@ func (e *bestFastEncoder) Encode(blk *blockEnc, src []byte) {
 		break
 	}
 
+	// Add block to history
 	s := e.addBlock(src)
 	blk.size = len(src)
+
+	// Check RLE first
+	if len(src) > zstdMinMatch {
+		ml := matchLen(src[1:], src)
+		if ml == len(src)-1 {
+			blk.literals = append(blk.literals, src[0])
+			blk.sequences = append(blk.sequences, seq{litLen: 1, matchLen: uint32(len(src)-1) - zstdMinMatch, offset: 1 + 3})
+			return
+		}
+	}
+
 	if len(src) < minNonLiteralBlockSize {
 		blk.extraLits = len(src)
 		blk.literals = blk.literals[:len(src)]
@@ -201,14 +213,6 @@ encodeLoop:
 			if delta >= e.maxMatchOff || delta <= 0 || load3232(src, offset) != first {
 				return
 			}
-			if debugAsserts {
-				if offset >= s {
-					panic(fmt.Sprintf("offset: %d - s:%d - rep: %d - cur :%d - max: %d", offset, s, rep, e.cur, e.maxMatchOff))
-				}
-				if !bytes.Equal(src[s:s+4], src[offset:offset+4]) {
-					panic(fmt.Sprintf("first match mismatch: %v != %v, first: %08x", src[s:s+4], src[offset:offset+4], first))
-				}
-			}
 			// Try to quick reject if we already have a long match.
 			if m.length > 16 {
 				left := len(src) - int(m.s+m.length)
@@ -227,8 +231,10 @@ encodeLoop:
 				}
 			}
 			l := 4 + e.matchlen(s+4, offset+4, src)
-			if true {
+			if m.rep <= 0 {
 				// Extend candidate match backwards as far as possible.
+				// Do not extend repeats as we can assume they are optimal
+				// and offsets change if s == nextEmit.
 				tMin := s - e.maxMatchOff
 				if tMin < 0 {
 					tMin = 0
@@ -239,7 +245,14 @@ encodeLoop:
 					l++
 				}
 			}
-
+			if debugAsserts {
+				if offset >= s {
+					panic(fmt.Sprintf("offset: %d - s:%d - rep: %d - cur :%d - max: %d", offset, s, rep, e.cur, e.maxMatchOff))
+				}
+				if !bytes.Equal(src[s:s+l], src[offset:offset+l]) {
+					panic(fmt.Sprintf("second match mismatch: %v != %v, first: %08x", src[s:s+4], src[offset:offset+4], first))
+				}
+			}
 			cand := match{offset: offset, s: s, length: l, rep: rep}
 			cand.estBits(bitsPerByte)
 			if m.est >= highScore || cand.est-m.est+(cand.s-m.s)*bitsPerByte>>10 < 0 {
@@ -336,24 +349,31 @@ encodeLoop:
 		}
 
 		if debugAsserts {
+			if best.offset >= best.s {
+				panic(fmt.Sprintf("best.offset > s: %d >= %d", best.offset, best.s))
+			}
+			if best.s < nextEmit {
+				panic(fmt.Sprintf("s %d < nextEmit %d", best.s, nextEmit))
+			}
+			if best.offset < s-e.maxMatchOff {
+				panic(fmt.Sprintf("best.offset < s-e.maxMatchOff: %d < %d", best.offset, s-e.maxMatchOff))
+			}
 			if !bytes.Equal(src[best.s:best.s+best.length], src[best.offset:best.offset+best.length]) {
 				panic(fmt.Sprintf("match mismatch: %v != %v", src[best.s:best.s+best.length], src[best.offset:best.offset+best.length]))
 			}
 		}
 
 		// We have a match, we can store the forward value
+		s = best.s
 		if best.rep > 0 {
 			var seq seq
 			seq.matchLen = uint32(best.length - zstdMinMatch)
-			if debugAsserts && s < nextEmit {
-				panic("s < nextEmit")
-			}
 			addLiterals(&seq, best.s)
 
 			// Repeat. If bit 4 is set, this is a non-lit repeat.
 			seq.offset = uint32(best.rep & 3)
 			if debugSequences {
-				println("repeat sequence", seq, "next s:", s)
+				println("repeat sequence", seq, "next s:", best.s, "off:", best.s-best.offset)
 			}
 			blk.sequences = append(blk.sequences, seq)
 
@@ -396,7 +416,6 @@ encodeLoop:
 
 		// A 4-byte match has been found. Update recent offsets.
 		// We'll later see if more than 4 bytes.
-		s = best.s
 		t := best.offset
 		offset1, offset2, offset3 = s-t, offset1, offset2
 

vendor/github.com/klauspost/compress/zstd/enc_better.go

@@ -102,9 +102,20 @@ func (e *betterFastEncoder) Encode(blk *blockEnc, src []byte) {
 		e.cur = e.maxMatchOff
 		break
 	}
-
+	// Add block to history
 	s := e.addBlock(src)
 	blk.size = len(src)
+
+	// Check RLE first
+	if len(src) > zstdMinMatch {
+		ml := matchLen(src[1:], src)
+		if ml == len(src)-1 {
+			blk.literals = append(blk.literals, src[0])
+			blk.sequences = append(blk.sequences, seq{litLen: 1, matchLen: uint32(len(src)-1) - zstdMinMatch, offset: 1 + 3})
+			return
+		}
+	}
+
 	if len(src) < minNonLiteralBlockSize {
 		blk.extraLits = len(src)
 		blk.literals = blk.literals[:len(src)]

vendor/github.com/klauspost/compress/zstd/encoder_options.go

@@ -94,7 +94,7 @@ func WithEncoderConcurrency(n int) EOption {
 // The value must be a power of two between MinWindowSize and MaxWindowSize.
 // A larger value will enable better compression but allocate more memory and,
 // for above-default values, take considerably longer.
-// The default value is determined by the compression level.
+// The default value is determined by the compression level and max 8MB.
 func WithWindowSize(n int) EOption {
 	return func(o *encoderOptions) error {
 		switch {
@@ -232,9 +232,9 @@ func WithEncoderLevel(l EncoderLevel) EOption {
 			case SpeedDefault:
 				o.windowSize = 8 << 20
 			case SpeedBetterCompression:
-				o.windowSize = 16 << 20
+				o.windowSize = 8 << 20
 			case SpeedBestCompression:
-				o.windowSize = 32 << 20
+				o.windowSize = 8 << 20
 			}
 		}
 		if !o.customALEntropy {

vendor/github.com/klauspost/compress/zstd/frameenc.go

@@ -76,7 +76,7 @@ func (f frameHeader) appendTo(dst []byte) []byte {
 		if f.SingleSegment {
 			dst = append(dst, uint8(f.ContentSize))
 		}
-		// Unless SingleSegment is set, framessizes < 256 are nto stored.
+		// Unless SingleSegment is set, framessizes < 256 are not stored.
 	case 1:
 		f.ContentSize -= 256
 		dst = append(dst, uint8(f.ContentSize), uint8(f.ContentSize>>8))

vendor/github.com/klauspost/compress/zstd/fse_decoder_generic.go

@@ -20,10 +20,9 @@ func (s *fseDecoder) buildDtable() error {
 			if v == -1 {
 				s.dt[highThreshold].setAddBits(uint8(i))
 				highThreshold--
-				symbolNext[i] = 1
-			} else {
-				symbolNext[i] = uint16(v)
+				v = 1
 			}
+			symbolNext[i] = uint16(v)
 		}
 	}
 
@@ -35,10 +34,12 @@ func (s *fseDecoder) buildDtable() error {
 		for ss, v := range s.norm[:s.symbolLen] {
 			for i := 0; i < int(v); i++ {
 				s.dt[position].setAddBits(uint8(ss))
-				position = (position + step) & tableMask
-				for position > highThreshold {
+				for {
 					// lowprob area
 					position = (position + step) & tableMask
+					if position <= highThreshold {
+						break
+					}
 				}
 			}
 		}

vendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_arm64.s

@@ -162,12 +162,12 @@ finalize:
 	MOVD h, ret+24(FP)
 	RET
 
-// func writeBlocks(d *Digest, b []byte) int
+// func writeBlocks(s *Digest, b []byte) int
 TEXT ·writeBlocks(SB), NOSPLIT|NOFRAME, $0-40
 	LDP ·primes+0(SB), (prime1, prime2)
 
 	// Load state. Assume v[1-4] are stored contiguously.
-	MOVD d+0(FP), digest
+	MOVD s+0(FP), digest
 	LDP  0(digest), (v1, v2)
 	LDP  16(digest), (v3, v4)
 

vendor/github.com/klauspost/compress/zstd/matchlen_amd64.s

@@ -5,7 +5,6 @@
 #include "textflag.h"
 
 // func matchLen(a []byte, b []byte) int
-// Requires: BMI
 TEXT ·matchLen(SB), NOSPLIT, $0-56
 	MOVQ a_base+0(FP), AX
 	MOVQ b_base+24(FP), CX
@@ -17,17 +16,16 @@ TEXT ·matchLen(SB), NOSPLIT, $0-56
 	JB   matchlen_match4_standalone
 
 matchlen_loopback_standalone:
-	MOVQ  (AX)(SI*1), BX
-	XORQ  (CX)(SI*1), BX
-	TESTQ BX, BX
-	JZ    matchlen_loop_standalone
+	MOVQ (AX)(SI*1), BX
+	XORQ (CX)(SI*1), BX
+	JZ   matchlen_loop_standalone
 
 #ifdef GOAMD64_v3
 	TZCNTQ BX, BX
 #else
 	BSFQ BX, BX
 #endif
-	SARQ $0x03, BX
+	SHRL $0x03, BX
 	LEAL (SI)(BX*1), SI
 	JMP  gen_match_len_end
 

vendor/github.com/klauspost/compress/zstd/seqdec_amd64.s

@@ -157,8 +157,7 @@ sequenceDecs_decode_amd64_ll_update_zero:
 
 	// Update Literal Length State
 	MOVBQZX DI, R14
-	SHRQ    $0x10, DI
-	MOVWQZX DI, DI
+	SHRL    $0x10, DI
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -177,8 +176,7 @@ sequenceDecs_decode_amd64_ll_update_zero:
 
 	// Update Match Length State
 	MOVBQZX R8, R14
-	SHRQ    $0x10, R8
-	MOVWQZX R8, R8
+	SHRL    $0x10, R8
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -197,8 +195,7 @@ sequenceDecs_decode_amd64_ll_update_zero:
 
 	// Update Offset State
 	MOVBQZX R9, R14
-	SHRQ    $0x10, R9
-	MOVWQZX R9, R9
+	SHRL    $0x10, R9
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -459,8 +456,7 @@ sequenceDecs_decode_56_amd64_ll_update_zero:
 
 	// Update Literal Length State
 	MOVBQZX DI, R14
-	SHRQ    $0x10, DI
-	MOVWQZX DI, DI
+	SHRL    $0x10, DI
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -479,8 +475,7 @@ sequenceDecs_decode_56_amd64_ll_update_zero:
 
 	// Update Match Length State
 	MOVBQZX R8, R14
-	SHRQ    $0x10, R8
-	MOVWQZX R8, R8
+	SHRL    $0x10, R8
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -499,8 +494,7 @@ sequenceDecs_decode_56_amd64_ll_update_zero:
 
 	// Update Offset State
 	MOVBQZX R9, R14
-	SHRQ    $0x10, R9
-	MOVWQZX R9, R9
+	SHRL    $0x10, R9
 	LEAQ    (BX)(R14*1), CX
 	MOVQ    DX, R15
 	MOVQ    CX, BX
@@ -772,11 +766,10 @@ sequenceDecs_decode_bmi2_fill_2_end:
 	BZHIQ   R14, R15, R15
 
 	// Update Offset State
-	BZHIQ  R8, R15, CX
-	SHRXQ  R8, R15, R15
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, R8, R8
-	ADDQ   CX, R8
+	BZHIQ R8, R15, CX
+	SHRXQ R8, R15, R15
+	SHRL  $0x10, R8
+	ADDQ  CX, R8
 
 	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
@@ -784,11 +777,10 @@ sequenceDecs_decode_bmi2_fill_2_end:
 	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	BZHIQ  DI, R15, CX
-	SHRXQ  DI, R15, R15
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, DI, DI
-	ADDQ   CX, DI
+	BZHIQ DI, R15, CX
+	SHRXQ DI, R15, R15
+	SHRL  $0x10, DI
+	ADDQ  CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
@@ -796,10 +788,9 @@ sequenceDecs_decode_bmi2_fill_2_end:
 	MOVQ (CX)(DI*8), DI
 
 	// Update Literal Length State
-	BZHIQ  SI, R15, CX
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, SI, SI
-	ADDQ   CX, SI
+	BZHIQ SI, R15, CX
+	SHRL  $0x10, SI
+	ADDQ  CX, SI
 
 	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
@@ -1032,11 +1023,10 @@ sequenceDecs_decode_56_bmi2_fill_end:
 	BZHIQ   R14, R15, R15
 
 	// Update Offset State
-	BZHIQ  R8, R15, CX
-	SHRXQ  R8, R15, R15
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, R8, R8
-	ADDQ   CX, R8
+	BZHIQ R8, R15, CX
+	SHRXQ R8, R15, R15
+	SHRL  $0x10, R8
+	ADDQ  CX, R8
 
 	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
@@ -1044,11 +1034,10 @@ sequenceDecs_decode_56_bmi2_fill_end:
 	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	BZHIQ  DI, R15, CX
-	SHRXQ  DI, R15, R15
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, DI, DI
-	ADDQ   CX, DI
+	BZHIQ DI, R15, CX
+	SHRXQ DI, R15, R15
+	SHRL  $0x10, DI
+	ADDQ  CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
@@ -1056,10 +1045,9 @@ sequenceDecs_decode_56_bmi2_fill_end:
 	MOVQ (CX)(DI*8), DI
 
 	// Update Literal Length State
-	BZHIQ  SI, R15, CX
-	MOVQ   $0x00001010, R14
-	BEXTRQ R14, SI, SI
-	ADDQ   CX, SI
+	BZHIQ SI, R15, CX
+	SHRL  $0x10, SI
+	ADDQ  CX, SI
 
 	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
@@ -1967,8 +1955,7 @@ sequenceDecs_decodeSync_amd64_ll_update_zero:
 
 	// Update Literal Length State
 	MOVBQZX DI, R13
-	SHRQ    $0x10, DI
-	MOVWQZX DI, DI
+	SHRL    $0x10, DI
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -1987,8 +1974,7 @@ sequenceDecs_decodeSync_amd64_ll_update_zero:
 
 	// Update Match Length State
 	MOVBQZX R8, R13
-	SHRQ    $0x10, R8
-	MOVWQZX R8, R8
+	SHRL    $0x10, R8
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -2007,8 +1993,7 @@ sequenceDecs_decodeSync_amd64_ll_update_zero:
 
 	// Update Offset State
 	MOVBQZX R9, R13
-	SHRQ    $0x10, R9
-	MOVWQZX R9, R9
+	SHRL    $0x10, R9
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -2514,11 +2499,10 @@ sequenceDecs_decodeSync_bmi2_fill_2_end:
 	BZHIQ   R13, R14, R14
 
 	// Update Offset State
-	BZHIQ  R8, R14, CX
-	SHRXQ  R8, R14, R14
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, R8, R8
-	ADDQ   CX, R8
+	BZHIQ R8, R14, CX
+	SHRXQ R8, R14, R14
+	SHRL  $0x10, R8
+	ADDQ  CX, R8
 
 	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
@@ -2526,11 +2510,10 @@ sequenceDecs_decodeSync_bmi2_fill_2_end:
 	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	BZHIQ  DI, R14, CX
-	SHRXQ  DI, R14, R14
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, DI, DI
-	ADDQ   CX, DI
+	BZHIQ DI, R14, CX
+	SHRXQ DI, R14, R14
+	SHRL  $0x10, DI
+	ADDQ  CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
@@ -2538,10 +2521,9 @@ sequenceDecs_decodeSync_bmi2_fill_2_end:
 	MOVQ (CX)(DI*8), DI
 
 	// Update Literal Length State
-	BZHIQ  SI, R14, CX
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, SI, SI
-	ADDQ   CX, SI
+	BZHIQ SI, R14, CX
+	SHRL  $0x10, SI
+	ADDQ  CX, SI
 
 	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX
@@ -3055,8 +3037,7 @@ sequenceDecs_decodeSync_safe_amd64_ll_update_zero:
 
 	// Update Literal Length State
 	MOVBQZX DI, R13
-	SHRQ    $0x10, DI
-	MOVWQZX DI, DI
+	SHRL    $0x10, DI
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -3075,8 +3056,7 @@ sequenceDecs_decodeSync_safe_amd64_ll_update_zero:
 
 	// Update Match Length State
 	MOVBQZX R8, R13
-	SHRQ    $0x10, R8
-	MOVWQZX R8, R8
+	SHRL    $0x10, R8
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -3095,8 +3075,7 @@ sequenceDecs_decodeSync_safe_amd64_ll_update_zero:
 
 	// Update Offset State
 	MOVBQZX R9, R13
-	SHRQ    $0x10, R9
-	MOVWQZX R9, R9
+	SHRL    $0x10, R9
 	LEAQ    (BX)(R13*1), CX
 	MOVQ    DX, R14
 	MOVQ    CX, BX
@@ -3704,11 +3683,10 @@ sequenceDecs_decodeSync_safe_bmi2_fill_2_end:
 	BZHIQ   R13, R14, R14
 
 	// Update Offset State
-	BZHIQ  R8, R14, CX
-	SHRXQ  R8, R14, R14
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, R8, R8
-	ADDQ   CX, R8
+	BZHIQ R8, R14, CX
+	SHRXQ R8, R14, R14
+	SHRL  $0x10, R8
+	ADDQ  CX, R8
 
 	// Load ctx.ofTable
 	MOVQ ctx+16(FP), CX
@@ -3716,11 +3694,10 @@ sequenceDecs_decodeSync_safe_bmi2_fill_2_end:
 	MOVQ (CX)(R8*8), R8
 
 	// Update Match Length State
-	BZHIQ  DI, R14, CX
-	SHRXQ  DI, R14, R14
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, DI, DI
-	ADDQ   CX, DI
+	BZHIQ DI, R14, CX
+	SHRXQ DI, R14, R14
+	SHRL  $0x10, DI
+	ADDQ  CX, DI
 
 	// Load ctx.mlTable
 	MOVQ ctx+16(FP), CX
@@ -3728,10 +3705,9 @@ sequenceDecs_decodeSync_safe_bmi2_fill_2_end:
 	MOVQ (CX)(DI*8), DI
 
 	// Update Literal Length State
-	BZHIQ  SI, R14, CX
-	MOVQ   $0x00001010, R13
-	BEXTRQ R13, SI, SI
-	ADDQ   CX, SI
+	BZHIQ SI, R14, CX
+	SHRL  $0x10, SI
+	ADDQ  CX, SI
 
 	// Load ctx.llTable
 	MOVQ ctx+16(FP), CX