docker/buildx
1
0
Fork 0
mirror of https://gitea.com/Lydanne/buildx.git synced 2025-07-09 21:17:09 +08:00
Code Issues Packages Projects Releases 1 Wiki Activity

vendor: update buildkit to opentelemetry support

Browse Source 
Signed-off-by: Tonis Tiigi <tonistiigi@gmail.com>
This commit is contained in:
Tonis Tiigi
2021-06-15 21:02:39 -07:00
parent 6ba080d337
commit 334c93fbbe
829 changed files with 89541 additions and 24438 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
vendor/github.com/klauspost/compress/zstd/README.md generated vendored
View File

@ -16,30 +16,28 @@ Currently the package is heavily optimized for 64 bit processors and will be sig
Install using `go get -u github.com/klauspost/compress`. The package is located in `github.com/klauspost/compress/zstd`.
Godoc Documentation: https://godoc.org/github.com/klauspost/compress/zstd
[![Go Reference](https://pkg.go.dev/badge/github.com/klauspost/compress/zstd.svg)](https://pkg.go.dev/github.com/klauspost/compress/zstd)
## Compressor
### Status:
STABLE - there may always be subtle bugs, a wide variety of content has been tested and the library is actively
used by several projects. This library is being continuously [fuzz-tested](https://github.com/klauspost/compress-fuzz),
kindly supplied by [fuzzit.dev](https://fuzzit.dev/).
used by several projects. This library is being [fuzz-tested](https://github.com/klauspost/compress-fuzz) for all updates.
There may still be specific combinations of data types/size/settings that could lead to edge cases,
so as always, testing is recommended.
For now, a high speed (fastest) and medium-fast (default) compressor has been implemented.
The "Fastest" compression ratio is roughly equivalent to zstd level 1.
The "Default" compression ratio is roughly equivalent to zstd level 3 (default).
* The "Fastest" compression ratio is roughly equivalent to zstd level 1.
* The "Default" compression ratio is roughly equivalent to zstd level 3 (default).
* The "Better" compression ratio is roughly equivalent to zstd level 7.
* The "Best" compression ratio is roughly equivalent to zstd level 11.
In terms of speed, it is typically 2x as fast as the stdlib deflate/gzip in its fastest mode.
The compression ratio compared to stdlib is around level 3, but usually 3x as fast.
Compared to cgo zstd, the speed is around level 3 (default), but compression slightly worse, between level 1&2.
### Usage
@ -140,7 +138,7 @@ I have collected some speed examples to compare speed and compression against ot
* `file` is the input file.
* `out` is the compressor used. `zskp` is this package. `zstd` is the Datadog cgo library. `gzstd/gzkp` is gzip standard and this library.
* `level` is the compression level used. For `zskp` level 1 is "fastest", level 2 is "default".
* `level` is the compression level used. For `zskp` level 1 is "fastest", level 2 is "default"; 3 is "better", 4 is "best".
* `insize`/`outsize` is the input/output size.
* `millis` is the number of milliseconds used for compression.
* `mb/s` is megabytes (2^20 bytes) per second.
@ -153,12 +151,14 @@ This package:
file out level insize outsize millis mb/s
silesia.tar zskp 1 211947520 73101992 643 313.87
silesia.tar zskp 2 211947520 67504318 969 208.38
silesia.tar zskp 3 211947520 65177448 1899 106.44
silesia.tar zskp 3 211947520 64595893 2007 100.68
silesia.tar zskp 4 211947520 60995370 7691 26.28
cgo zstd:
silesia.tar zstd 1 211947520 73605392 543 371.56
silesia.tar zstd 3 211947520 66793289 864 233.68
silesia.tar zstd 6 211947520 62916450 1913 105.66
silesia.tar zstd 9 211947520 60212393 5063 39.92
gzip, stdlib/this package:
silesia.tar gzstd 1 211947520 80007735 1654 122.21
@ -170,10 +170,12 @@ https://files.klauspost.com/compress/gob-stream.7z
file out level insize outsize millis mb/s
gob-stream zskp 1 1911399616 235022249 3088 590.30
gob-stream zskp 2 1911399616 205669791 3786 481.34
gob-stream zskp 3 1911399616 185792019 9324 195.48
gob-stream zskp 3 1911399616 175034659 9636 189.17
gob-stream zskp 4 1911399616 167273881 29337 62.13
gob-stream zstd 1 1911399616 249810424 2637 691.26
gob-stream zstd 3 1911399616 208192146 3490 522.31
gob-stream zstd 6 1911399616 193632038 6687 272.56
gob-stream zstd 9 1911399616 177620386 16175 112.70
gob-stream gzstd 1 1911399616 357382641 10251 177.82
gob-stream gzkp 1 1911399616 362156523 5695 320.08
@ -184,10 +186,12 @@ http://mattmahoney.net/dc/textdata.html
file out level insize outsize millis mb/s
enwik9 zskp 1 1000000000 343848582 3609 264.18
enwik9 zskp 2 1000000000 317276632 5746 165.97
enwik9 zskp 3 1000000000 294540704 11725 81.34
enwik9 zskp 3 1000000000 292243069 12162 78.41
enwik9 zskp 4 1000000000 275241169 36430 26.18
enwik9 zstd 1 1000000000 358072021 3110 306.65
enwik9 zstd 3 1000000000 313734672 4784 199.35
enwik9 zstd 6 1000000000 295138875 10290 92.68
enwik9 zstd 9 1000000000 278348700 28549 33.40
enwik9 gzstd 1 1000000000 382578136 9604 99.30
enwik9 gzkp 1 1000000000 383825945 6544 145.73
@ -197,10 +201,12 @@ https://files.klauspost.com/compress/github-june-2days-2019.json.zst
file out level insize outsize millis mb/s
github-june-2days-2019.json zskp 1 6273951764 699045015 10620 563.40
github-june-2days-2019.json zskp 2 6273951764 617881763 11687 511.96
github-june-2days-2019.json zskp 3 6273951764 537511906 29252 204.54
github-june-2days-2019.json zskp 3 6273951764 524340691 34043 175.75
github-june-2days-2019.json zskp 4 6273951764 503314661 93811 63.78
github-june-2days-2019.json zstd 1 6273951764 766284037 8450 708.00
github-june-2days-2019.json zstd 3 6273951764 661889476 10927 547.57
github-june-2days-2019.json zstd 6 6273951764 642756859 22996 260.18
github-june-2days-2019.json zstd 9 6273951764 601974523 52413 114.16
github-june-2days-2019.json gzstd 1 6273951764 1164400847 29948 199.79
github-june-2days-2019.json gzkp 1 6273951764 1128755542 19236 311.03
@ -210,10 +216,12 @@ https://files.klauspost.com/compress/rawstudio-mint14.7z
file out level insize outsize millis mb/s
rawstudio-mint14.tar zskp 1 8558382592 3667489370 20210 403.84
rawstudio-mint14.tar zskp 2 8558382592 3364592300 31873 256.07
rawstudio-mint14.tar zskp 3 8558382592 3224594213 71751 113.75
rawstudio-mint14.tar zskp 3 8558382592 3158085214 77675 105.08
rawstudio-mint14.tar zskp 4 8558382592 3020370044 404956 20.16
rawstudio-mint14.tar zstd 1 8558382592 3609250104 17136 476.27
rawstudio-mint14.tar zstd 3 8558382592 3341679997 29262 278.92
rawstudio-mint14.tar zstd 6 8558382592 3235846406 77904 104.77
rawstudio-mint14.tar zstd 9 8558382592 3160778861 140946 57.91
rawstudio-mint14.tar gzstd 1 8558382592 3926257486 57722 141.40
rawstudio-mint14.tar gzkp 1 8558382592 3970463184 41749 195.49
@ -223,10 +231,12 @@ https://files.klauspost.com/compress/nyc-taxi-data-10M.csv.zst
file out level insize outsize millis mb/s
nyc-taxi-data-10M.csv zskp 1 3325605752 641339945 8925 355.35
nyc-taxi-data-10M.csv zskp 2 3325605752 591748091 11268 281.44
nyc-taxi-data-10M.csv zskp 3 3325605752 538490114 19880 159.53
nyc-taxi-data-10M.csv zskp 3 3325605752 530289687 25239 125.66
nyc-taxi-data-10M.csv zskp 4 3325605752 490907191 65939 48.10
nyc-taxi-data-10M.csv zstd 1 3325605752 687399637 8233 385.18
nyc-taxi-data-10M.csv zstd 3 3325605752 598514411 10065 315.07
nyc-taxi-data-10M.csv zstd 6 3325605752 570522953 20038 158.27
nyc-taxi-data-10M.csv zstd 9 3325605752 517554797 64565 49.12
nyc-taxi-data-10M.csv gzstd 1 3325605752 928656485 23876 132.83
nyc-taxi-data-10M.csv gzkp 1 3325605752 924718719 16388 193.53
```
@ -394,13 +404,28 @@ BenchmarkDecoder_DecodeAllParallelCgo/comp-data.bin.zst-16 749938
This reflects the performance around May 2020, but this may be out of date.
## Zstd inside ZIP files
It is possible to use zstandard to compress individual files inside zip archives.
While this isn't widely supported it can be useful for internal files.
To support the compression and decompression of these files you must register a compressor and decompressor.
It is highly recommended registering the (de)compressors on individual zip Reader/Writer and NOT
use the global registration functions. The main reason for this is that 2 registrations from
different packages will result in a panic.
It is a good idea to only have a single compressor and decompressor, since they can be used for multiple zip
files concurrently, and using a single instance will allow reusing some resources.
See [this example](https://pkg.go.dev/github.com/klauspost/compress/zstd#example-ZipCompressor) for
how to compress and decompress files inside zip archives.
# Contributions
Contributions are always welcome.
For new features/fixes, remember to add tests and for performance enhancements include benchmarks.
For sending files for reproducing errors use a service like [goobox](https://goobox.io/#/upload) or similar to share your files.
For general feedback and experience reports, feel free to open an issue or write me on [Twitter](https://twitter.com/sh0dan).
This package includes the excellent [`github.com/cespare/xxhash`](https://github.com/cespare/xxhash) package Copyright (c) 2016 Caleb Spare.

13
vendor/github.com/klauspost/compress/zstd/blockdec.go generated vendored
View File

@ -123,12 +123,10 @@ func newBlockDec(lowMem bool) *blockDec {
// Input must be a start of a block and will be at the end of the block when returned.
func (b *blockDec) reset(br byteBuffer, windowSize uint64) error {
b.WindowSize = windowSize
tmp := br.readSmall(3)
if tmp == nil {
if debug {
println("Reading block header:", io.ErrUnexpectedEOF)
}
return io.ErrUnexpectedEOF
tmp, err := br.readSmall(3)
if err != nil {
println("Reading block header:", err)
return err
}
bh := uint32(tmp[0]) | (uint32(tmp[1]) << 8) | (uint32(tmp[2]) << 16)
b.Last = bh&1 != 0
@ -179,7 +177,6 @@ func (b *blockDec) reset(br byteBuffer, windowSize uint64) error {
if cap(b.dst) <= maxSize {
b.dst = make([]byte, 0, maxSize+1)
}
var err error
b.data, err = br.readBig(cSize, b.dataStorage)
if err != nil {
if debug {
@ -613,7 +610,7 @@ func (b *blockDec) decodeCompressed(hist *history) error {
// Decode treeless literal block.
if litType == literalsBlockTreeless {
// TODO: We could send the history early WITHOUT the stream history.
// This would allow decoding treeless literials before the byte history is available.
// This would allow decoding treeless literals before the byte history is available.
// Silencia stats: Treeless 4393, with: 32775, total: 37168, 11% treeless.
// So not much obvious gain here.

55
vendor/github.com/klauspost/compress/zstd/blockenc.go generated vendored
View File

@ -22,28 +22,44 @@ type blockEnc struct {
dictLitEnc *huff0.Scratch
wr bitWriter
extraLits int
last bool
extraLits int
output []byte
recentOffsets [3]uint32
prevRecentOffsets [3]uint32
last bool
lowMem bool
}
// init should be used once the block has been created.
// If called more than once, the effect is the same as calling reset.
func (b *blockEnc) init() {
if cap(b.literals) < maxCompressedLiteralSize {
b.literals = make([]byte, 0, maxCompressedLiteralSize)
}
const defSeqs = 200
b.literals = b.literals[:0]
if cap(b.sequences) < defSeqs {
b.sequences = make([]seq, 0, defSeqs)
}
if cap(b.output) < maxCompressedBlockSize {
b.output = make([]byte, 0, maxCompressedBlockSize)
if b.lowMem {
// 1K literals
if cap(b.literals) < 1<<10 {
b.literals = make([]byte, 0, 1<<10)
}
const defSeqs = 20
if cap(b.sequences) < defSeqs {
b.sequences = make([]seq, 0, defSeqs)
}
// 1K
if cap(b.output) < 1<<10 {
b.output = make([]byte, 0, 1<<10)
}
} else {
if cap(b.literals) < maxCompressedBlockSize {
b.literals = make([]byte, 0, maxCompressedBlockSize)
}
const defSeqs = 200
if cap(b.sequences) < defSeqs {
b.sequences = make([]seq, 0, defSeqs)
}
if cap(b.output) < maxCompressedBlockSize {
b.output = make([]byte, 0, maxCompressedBlockSize)
}
}
if b.coders.mlEnc == nil {
b.coders.mlEnc = &fseEncoder{}
b.coders.mlPrev = &fseEncoder{}
@ -76,6 +92,7 @@ func (b *blockEnc) reset(prev *blockEnc) {
if prev != nil {
b.recentOffsets = prev.prevRecentOffsets
}
b.dictLitEnc = nil
}
// reset will reset the block for a new encode, but in the same stream,
@ -369,9 +386,9 @@ func (b *blockEnc) encodeLits(lits []byte, raw bool) error {
b.output = bh.appendTo(b.output)
b.output = append(b.output, lits[0])
return nil
case nil:
default:
return err
case nil:
}
// Compressed...
// Now, allow reuse
@ -511,11 +528,6 @@ func (b *blockEnc) encode(org []byte, raw, rawAllLits bool) error {
if debug {
println("Adding literals RLE")
}
default:
if debug {
println("Adding literals ERROR:", err)
}
return err
case nil:
// Compressed litLen...
if reUsed {
@ -546,6 +558,11 @@ func (b *blockEnc) encode(org []byte, raw, rawAllLits bool) error {
if debug {
println("Adding literals compressed")
}
default:
if debug {
println("Adding literals ERROR:", err)
}
return err
}
// Sequence compression

21
vendor/github.com/klauspost/compress/zstd/bytebuf.go generated vendored
View File

@ -12,8 +12,8 @@ import (
type byteBuffer interface {
// Read up to 8 bytes.
// Returns nil if no more input is available.
readSmall(n int) []byte
// Returns io.ErrUnexpectedEOF if this cannot be satisfied.
readSmall(n int) ([]byte, error)
// Read >8 bytes.
// MAY use the destination slice.
@ -29,17 +29,17 @@ type byteBuffer interface {
// in-memory buffer
type byteBuf []byte
func (b *byteBuf) readSmall(n int) []byte {
func (b *byteBuf) readSmall(n int) ([]byte, error) {
if debugAsserts && n > 8 {
panic(fmt.Errorf("small read > 8 (%d). use readBig", n))
}
bb := *b
if len(bb) < n {
return nil
return nil, io.ErrUnexpectedEOF
}
r := bb[:n]
*b = bb[n:]
return r
return r, nil
}
func (b *byteBuf) readBig(n int, dst []byte) ([]byte, error) {
@ -81,19 +81,22 @@ type readerWrapper struct {
tmp [8]byte
}
func (r *readerWrapper) readSmall(n int) []byte {
func (r *readerWrapper) readSmall(n int) ([]byte, error) {
if debugAsserts && n > 8 {
panic(fmt.Errorf("small read > 8 (%d). use readBig", n))
}
n2, err := io.ReadFull(r.r, r.tmp[:n])
// We only really care about the actual bytes read.
if n2 != n {
if err != nil {
if err == io.EOF {
return nil, io.ErrUnexpectedEOF
}
if debug {
println("readSmall: got", n2, "want", n, "err", err)
}
return nil
return nil, err
}
return r.tmp[:n]
return r.tmp[:n], nil
}
func (r *readerWrapper) readBig(n int, dst []byte) ([]byte, error) {

202
vendor/github.com/klauspost/compress/zstd/decodeheader.go generated vendored Normal file
View File

@ -0,0 +1,202 @@
// Copyright 2020+ Klaus Post. All rights reserved.
// License information can be found in the LICENSE file.
package zstd
import (
"bytes"
"errors"
"io"
)
// HeaderMaxSize is the maximum size of a Frame and Block Header.
// If less is sent to Header.Decode it *may* still contain enough information.
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
// Frame content size.
// Expected size of the entire frame.
FrameContentSize uint64
// Dictionary ID.
// If 0, no dictionary.
DictionaryID uint32
// First block information.
FirstBlock struct {
// OK will be set if first block could be decoded.
OK bool
// Is this the last block of a frame?
Last bool
// Is the data compressed?
// If true CompressedSize will be populated.
// Unfortunately DecompressedSize cannot be determined
// without decoding the blocks.
Compressed bool
// DecompressedSize is the expected decompressed size of the block.
// Will be 0 if it cannot be determined.
DecompressedSize int
// CompressedSize of the data in the block.
// Does not include the block header.
// Will be equal to DecompressedSize if not Compressed.
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.
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.
// 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) Decode(in []byte) error {
if len(in) < 4 {
return io.ErrUnexpectedEOF
}
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}
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 !h.SingleSegment {
if len(in) < 1 {
return io.ErrUnexpectedEOF
}
var wd byte
wd, in = in[0], in[1:]
windowLog := 10 + (wd >> 3)
windowBase := uint64(1) << windowLog
windowAdd := (windowBase / 8) * uint64(wd&0x7)
h.WindowSize = windowBase + windowAdd
}
// Read Dictionary_ID
// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#dictionary_id
if size := fhd & 3; size != 0 {
if size == 3 {
size = 4
}
if len(in) < int(size) {
return io.ErrUnexpectedEOF
}
b, in = in[:size], in[size:]
if b == nil {
return io.ErrUnexpectedEOF
}
switch size {
case 1:
h.DictionaryID = uint32(b[0])
case 2:
h.DictionaryID = uint32(b[0]) | (uint32(b[1]) << 8)
case 4:
h.DictionaryID = uint32(b[0]) | (uint32(b[1]) << 8) | (uint32(b[2]) << 16) | (uint32(b[3]) << 24)
}
}
// Read Frame_Content_Size
// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#frame_content_size
var fcsSize int
v := fhd >> 6
switch v {
case 0:
if h.SingleSegment {
fcsSize = 1
}
default:
fcsSize = 1 << v
}
if fcsSize > 0 {
h.HasFCS = true
if len(in) < fcsSize {
return io.ErrUnexpectedEOF
}
b, in = in[:fcsSize], in[fcsSize:]
if b == nil {
return io.ErrUnexpectedEOF
}
switch fcsSize {
case 1:
h.FrameContentSize = uint64(b[0])
case 2:
// When FCS_Field_Size is 2, the offset of 256 is added.
h.FrameContentSize = uint64(b[0]) | (uint64(b[1]) << 8) + 256
case 4:
h.FrameContentSize = uint64(b[0]) | (uint64(b[1]) << 8) | (uint64(b[2]) << 16) | (uint64(b[3]) << 24)
case 8:
d1 := uint32(b[0]) | (uint32(b[1]) << 8) | (uint32(b[2]) << 16) | (uint32(b[3]) << 24)
d2 := uint32(b[4]) | (uint32(b[5]) << 8) | (uint32(b[6]) << 16) | (uint32(b[7]) << 24)
h.FrameContentSize = uint64(d1) | (uint64(d2) << 32)
}
}
// Frame Header done, we will not fail from now on.
if len(in) < 3 {
return nil
}
tmp := in[:3]
bh := uint32(tmp[0]) | (uint32(tmp[1]) << 8) | (uint32(tmp[2]) << 16)
h.FirstBlock.Last = bh&1 != 0
blockType := blockType((bh >> 1) & 3)
// find size.
cSize := int(bh >> 3)
switch blockType {
case blockTypeReserved:
return nil
case blockTypeRLE:
h.FirstBlock.Compressed = true
h.FirstBlock.DecompressedSize = cSize
h.FirstBlock.CompressedSize = 1
case blockTypeCompressed:
h.FirstBlock.Compressed = true
h.FirstBlock.CompressedSize = cSize
case blockTypeRaw:
h.FirstBlock.DecompressedSize = cSize
h.FirstBlock.CompressedSize = cSize
default:
panic("Invalid block type")
}
h.FirstBlock.OK = true
return nil
}

49
vendor/github.com/klauspost/compress/zstd/decoder.go generated vendored
View File

@ -5,7 +5,6 @@
package zstd
import (
"bytes"
"errors"
"io"
"sync"
@ -85,6 +84,10 @@ func NewReader(r io.Reader, opts ...DOption) (*Decoder, error) {
d.current.output = make(chan decodeOutput, d.o.concurrent)
d.current.flushed = true
if r == nil {
d.current.err = ErrDecoderNilInput
}
// Transfer option dicts.
d.dicts = make(map[uint32]dict, len(d.o.dicts))
for _, dc := range d.o.dicts {
@ -111,7 +114,7 @@ func NewReader(r io.Reader, opts ...DOption) (*Decoder, error) {
// When the stream is done, io.EOF will be returned.
func (d *Decoder) Read(p []byte) (int, error) {
if d.stream == nil {
return 0, errors.New("no input has been initialized")
return 0, ErrDecoderNilInput
}
var n int
for {
@ -152,12 +155,20 @@ func (d *Decoder) Read(p []byte) (int, error) {
// Reset will reset the decoder the supplied stream after the current has finished processing.
// Note that this functionality cannot be used after Close has been called.
// Reset can be called with a nil reader to release references to the previous reader.
// After being called with a nil reader, no other operations than Reset or DecodeAll or Close
// should be used.
func (d *Decoder) Reset(r io.Reader) error {
if d.current.err == ErrDecoderClosed {
return d.current.err
}
d.drainOutput()
if r == nil {
return errors.New("nil Reader sent as input")
d.current.err = ErrDecoderNilInput
d.current.flushed = true
return nil
}
if d.stream == nil {
@ -166,14 +177,13 @@ func (d *Decoder) Reset(r io.Reader) error {
go d.startStreamDecoder(d.stream)
}
d.drainOutput()
// If bytes buffer and < 1MB, do sync decoding anyway.
if bb, ok := r.(*bytes.Buffer); ok && bb.Len() < 1<<20 {
if bb, ok := r.(byter); ok && bb.Len() < 1<<20 {
bb2 := bb
if debug {
println("*bytes.Buffer detected, doing sync decode, len:", bb.Len())
}
b := bb.Bytes()
b := bb2.Bytes()
var dst []byte
if cap(d.current.b) > 0 {
dst = d.current.b
@ -226,20 +236,17 @@ func (d *Decoder) drainOutput() {
println("current already flushed")
return
}
for {
select {
case v := <-d.current.output:
if v.d != nil {
if debug {
printf("re-adding decoder %p", v.d)
}
d.decoders <- v.d
}
if v.err == errEndOfStream {
println("current flushed")
d.current.flushed = true
return
for v := range d.current.output {
if v.d != nil {
if debug {
printf("re-adding decoder %p", v.d)
}
d.decoders <- v.d
}
if v.err == errEndOfStream {
println("current flushed")
d.current.flushed = true
return
}
}
}
@ -249,7 +256,7 @@ func (d *Decoder) drainOutput() {
// Any error encountered during the write is also returned.
func (d *Decoder) WriteTo(w io.Writer) (int64, error) {
if d.stream == nil {
return 0, errors.New("no input has been initialized")
return 0, ErrDecoderNilInput
}
var n int64
for {

5
vendor/github.com/klauspost/compress/zstd/decoder_options.go generated vendored
View File

@ -6,7 +6,6 @@ package zstd
import (
"errors"
"fmt"
"runtime"
)
@ -43,7 +42,7 @@ func WithDecoderLowmem(b bool) DOption {
func WithDecoderConcurrency(n int) DOption {
return func(o *decoderOptions) error {
if n <= 0 {
return fmt.Errorf("Concurrency must be at least 1")
return errors.New("concurrency must be at least 1")
}
o.concurrent = n
return nil
@ -61,7 +60,7 @@ func WithDecoderMaxMemory(n uint64) DOption {
return errors.New("WithDecoderMaxMemory must be at least 1")
}
if n > 1<<63 {
return fmt.Errorf("WithDecoderMaxmemory must be less than 1 << 63")
return errors.New("WithDecoderMaxmemory must be less than 1 << 63")
}
o.maxDecodedSize = n
return nil

53
vendor/github.com/klauspost/compress/zstd/enc_base.go generated vendored
View File

@ -7,6 +7,10 @@ import (
"github.com/klauspost/compress/zstd/internal/xxhash"
)
const (
dictShardBits = 6
)
type fastBase struct {
// cur is the offset at the start of hist
cur int32
@ -17,6 +21,7 @@ type fastBase struct {
tmp [8]byte
blk *blockEnc
lastDictID uint32
lowMem bool
}
// CRC returns the underlying CRC writer.
@ -57,15 +62,10 @@ func (e *fastBase) addBlock(src []byte) int32 {
// check if we have space already
if len(e.hist)+len(src) > cap(e.hist) {
if cap(e.hist) == 0 {
l := e.maxMatchOff * 2
// Make it at least 1MB.
if l < 1<<20 {
l = 1 << 20
}
e.hist = make([]byte, 0, l)
e.ensureHist(len(src))
} else {
if cap(e.hist) < int(e.maxMatchOff*2) {
panic("unexpected buffer size")
if cap(e.hist) < int(e.maxMatchOff+maxCompressedBlockSize) {
panic(fmt.Errorf("unexpected buffer cap %d, want at least %d with window %d", cap(e.hist), e.maxMatchOff+maxCompressedBlockSize, e.maxMatchOff))
}
// Move down
offset := int32(len(e.hist)) - e.maxMatchOff
@ -79,6 +79,28 @@ func (e *fastBase) addBlock(src []byte) int32 {
return s
}
// ensureHist will ensure that history can keep at least this many bytes.
func (e *fastBase) ensureHist(n int) {
if cap(e.hist) >= n {
return
}
l := e.maxMatchOff
if (e.lowMem && e.maxMatchOff > maxCompressedBlockSize) || e.maxMatchOff <= maxCompressedBlockSize {
l += maxCompressedBlockSize
} else {
l += e.maxMatchOff
}
// Make it at least 1MB.
if l < 1<<20 && !e.lowMem {
l = 1 << 20
}
// Make it at least the requested size.
if l < int32(n) {
l = int32(n)
}
e.hist = make([]byte, 0, l)
}
// useBlock will replace the block with the provided one,
// but transfer recent offsets from the previous.
func (e *fastBase) UseBlock(enc *blockEnc) {
@ -117,7 +139,7 @@ func (e *fastBase) matchlen(s, t int32, src []byte) int32 {
// Reset the encoding table.
func (e *fastBase) resetBase(d *dict, singleBlock bool) {
if e.blk == nil {
e.blk = &blockEnc{}
e.blk = &blockEnc{lowMem: e.lowMem}
e.blk.init()
} else {
e.blk.reset(nil)
@ -128,14 +150,15 @@ func (e *fastBase) resetBase(d *dict, singleBlock bool) {
} else {
e.crc.Reset()
}
if (!singleBlock || d.DictContentSize() > 0) && cap(e.hist) < int(e.maxMatchOff*2)+d.DictContentSize() {
l := e.maxMatchOff*2 + int32(d.DictContentSize())
// Make it at least 1MB.
if l < 1<<20 {
l = 1 << 20
if d != nil {
low := e.lowMem
if singleBlock {
e.lowMem = true
}
e.hist = make([]byte, 0, l)
e.ensureHist(d.DictContentSize() + maxCompressedBlockSize)
e.lowMem = low
}
// We offset current position so everything will be out of reach.
// If above reset line, history will be purged.
if e.cur < bufferReset {

501
vendor/github.com/klauspost/compress/zstd/enc_best.go generated vendored Normal file
View File

@ -0,0 +1,501 @@
// Copyright 2019+ Klaus Post. All rights reserved.
// License information can be found in the LICENSE file.
// Based on work by Yann Collet, released under BSD License.
package zstd
import (
"fmt"
"math/bits"
)
const (
bestLongTableBits = 20 // Bits used in the long match table
bestLongTableSize = 1 << bestLongTableBits // Size of the table
// Note: Increasing the short table bits or making the hash shorter
// can actually lead to compression degradation since it will 'steal' more from the
// long match table and match offsets are quite big.
// This greatly depends on the type of input.
bestShortTableBits = 16 // Bits used in the short match table
bestShortTableSize = 1 << bestShortTableBits // Size of the table
)
// bestFastEncoder uses 2 tables, one for short matches (5 bytes) and one for long matches.
// The long match table contains the previous entry with the same hash,
// effectively making it a "chain" of length 2.
// When we find a long match we choose between the two values and select the longest.
// When we find a short match, after checking the long, we check if we can find a long at n+1
// and that it is longer (lazy matching).
type bestFastEncoder struct {
fastBase
table [bestShortTableSize]prevEntry
longTable [bestLongTableSize]prevEntry
dictTable []prevEntry
dictLongTable []prevEntry
}
// Encode improves compression...
func (e *bestFastEncoder) Encode(blk *blockEnc, src []byte) {
const (
// Input margin is the number of bytes we read (8)
// and the maximum we will read ahead (2)
inputMargin = 8 + 4
minNonLiteralBlockSize = 16
)
// Protect against e.cur wraparound.
for e.cur >= bufferReset {
if len(e.hist) == 0 {
for i := range e.table[:] {
e.table[i] = prevEntry{}
}
for i := range e.longTable[:] {
e.longTable[i] = prevEntry{}
}
e.cur = e.maxMatchOff
break
}
// Shift down everything in the table that isn't already too far away.
minOff := e.cur + int32(len(e.hist)) - e.maxMatchOff
for i := range e.table[:] {
v := e.table[i].offset
v2 := e.table[i].prev
if v < minOff {
v = 0
v2 = 0
} else {
v = v - e.cur + e.maxMatchOff
if v2 < minOff {
v2 = 0
} else {
v2 = v2 - e.cur + e.maxMatchOff
}
}
e.table[i] = prevEntry{
offset: v,
prev: v2,
}
}
for i := range e.longTable[:] {
v := e.longTable[i].offset
v2 := e.longTable[i].prev
if v < minOff {
v = 0
v2 = 0
} else {
v = v - e.cur + e.maxMatchOff
if v2 < minOff {
v2 = 0
} else {
v2 = v2 - e.cur + e.maxMatchOff
}
}
e.longTable[i] = prevEntry{
offset: v,
prev: v2,
}
}
e.cur = e.maxMatchOff
break
}
s := e.addBlock(src)
blk.size = len(src)
if len(src) < minNonLiteralBlockSize {
blk.extraLits = len(src)
blk.literals = blk.literals[:len(src)]
copy(blk.literals, src)
return
}
// Override src
src = e.hist
sLimit := int32(len(src)) - inputMargin
const kSearchStrength = 10
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := s
cv := load6432(src, s)
// Relative offsets
offset1 := int32(blk.recentOffsets[0])
offset2 := int32(blk.recentOffsets[1])
offset3 := int32(blk.recentOffsets[2])
addLiterals := func(s *seq, until int32) {
if until == nextEmit {
return
}
blk.literals = append(blk.literals, src[nextEmit:until]...)
s.litLen = uint32(until - nextEmit)
}
_ = addLiterals
if debug {
println("recent offsets:", blk.recentOffsets)
}
encodeLoop:
for {
// We allow the encoder to optionally turn off repeat offsets across blocks
canRepeat := len(blk.sequences) > 2
if debugAsserts && canRepeat && offset1 == 0 {
panic("offset0 was 0")
}
type match struct {
offset int32
s int32
length int32
rep int32
}
matchAt := func(offset int32, s int32, first uint32, rep int32) match {
if s-offset >= e.maxMatchOff || load3232(src, offset) != first {
return match{offset: offset, s: s}
}
return match{offset: offset, s: s, length: 4 + e.matchlen(s+4, offset+4, src), rep: rep}
}
bestOf := func(a, b match) match {
aScore := b.s - a.s + a.length
bScore := a.s - b.s + b.length
if a.rep < 0 {
aScore = aScore - int32(bits.Len32(uint32(a.offset)))/8
}
if b.rep < 0 {
bScore = bScore - int32(bits.Len32(uint32(b.offset)))/8
}
if aScore >= bScore {
return a
}
return b
}
const goodEnough = 100
nextHashL := hash8(cv, bestLongTableBits)
nextHashS := hash4x64(cv, bestShortTableBits)
candidateL := e.longTable[nextHashL]
candidateS := e.table[nextHashS]
best := bestOf(matchAt(candidateL.offset-e.cur, s, uint32(cv), -1), matchAt(candidateL.prev-e.cur, s, uint32(cv), -1))
best = bestOf(best, matchAt(candidateS.offset-e.cur, s, uint32(cv), -1))
best = bestOf(best, matchAt(candidateS.prev-e.cur, s, uint32(cv), -1))
if canRepeat && best.length < goodEnough {
best = bestOf(best, matchAt(s-offset1+1, s+1, uint32(cv>>8), 1))
best = bestOf(best, matchAt(s-offset2+1, s+1, uint32(cv>>8), 2))
best = bestOf(best, matchAt(s-offset3+1, s+1, uint32(cv>>8), 3))
if best.length > 0 {
best = bestOf(best, matchAt(s-offset1+3, s+3, uint32(cv>>24), 1))
best = bestOf(best, matchAt(s-offset2+3, s+3, uint32(cv>>24), 2))
best = bestOf(best, matchAt(s-offset3+3, s+3, uint32(cv>>24), 3))
}
}
// Load next and check...
e.longTable[nextHashL] = prevEntry{offset: s + e.cur, prev: candidateL.offset}
e.table[nextHashS] = prevEntry{offset: s + e.cur, prev: candidateS.offset}
// Look far ahead, unless we have a really long match already...
if best.length < goodEnough {
// No match found, move forward on input, no need to check forward...
if best.length < 4 {
s += 1 + (s-nextEmit)>>(kSearchStrength-1)
if s >= sLimit {
break encodeLoop
}
cv = load6432(src, s)
continue
}
s++
candidateS = e.table[hash4x64(cv>>8, bestShortTableBits)]
cv = load6432(src, s)
cv2 := load6432(src, s+1)
candidateL = e.longTable[hash8(cv, bestLongTableBits)]
candidateL2 := e.longTable[hash8(cv2, bestLongTableBits)]
best = bestOf(best, matchAt(candidateS.offset-e.cur, s, uint32(cv), -1))
best = bestOf(best, matchAt(candidateL.offset-e.cur, s, uint32(cv), -1))
best = bestOf(best, matchAt(candidateL.prev-e.cur, s, uint32(cv), -1))
best = bestOf(best, matchAt(candidateL2.offset-e.cur, s+1, uint32(cv2), -1))
best = bestOf(best, matchAt(candidateL2.prev-e.cur, s+1, uint32(cv2), -1))
// See if we can find a better match by checking where the current best ends.
// Use that offset to see if we can find a better full match.
if sAt := best.s + best.length; sAt < sLimit {
nextHashL := hash8(load6432(src, sAt), bestLongTableBits)
candidateEnd := e.longTable[nextHashL]
if pos := candidateEnd.offset - e.cur - best.length; pos >= 0 {
bestEnd := bestOf(best, matchAt(pos, best.s, load3232(src, best.s), -1))
if pos := candidateEnd.prev - e.cur - best.length; pos >= 0 {
bestEnd = bestOf(bestEnd, matchAt(pos, best.s, load3232(src, best.s), -1))
}
best = bestEnd
}
}
}
// We have a match, we can store the forward value
if best.rep > 0 {
s = best.s
var seq seq
seq.matchLen = uint32(best.length - zstdMinMatch)
// We might be able to match backwards.
// Extend as long as we can.
start := best.s
// We end the search early, so we don't risk 0 literals
// and have to do special offset treatment.
startLimit := nextEmit + 1
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
repIndex := best.offset
for repIndex > tMin && start > startLimit && src[repIndex-1] == src[start-1] && seq.matchLen < maxMatchLength-zstdMinMatch-1 {
repIndex--
start--
seq.matchLen++
}
addLiterals(&seq, start)
// rep 0
seq.offset = uint32(best.rep)
if debugSequences {
println("repeat sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Index match start+1 (long) -> s - 1
index0 := s
s = best.s + best.length
nextEmit = s
if s >= sLimit {
if debug {
println("repeat ended", s, best.length)
}
break encodeLoop
}
// Index skipped...
off := index0 + e.cur
for index0 < s-1 {
cv0 := load6432(src, index0)
h0 := hash8(cv0, bestLongTableBits)
h1 := hash4x64(cv0, bestShortTableBits)
e.longTable[h0] = prevEntry{offset: off, prev: e.longTable[h0].offset}
e.table[h1] = prevEntry{offset: off, prev: e.table[h1].offset}
off++
index0++
}
switch best.rep {
case 2:
offset1, offset2 = offset2, offset1
case 3:
offset1, offset2, offset3 = offset3, offset1, offset2
}
cv = load6432(src, s)
continue
}
// 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
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && canRepeat && int(offset1) > len(src) {
panic("invalid offset")
}
// Extend the n-byte match as long as possible.
l := best.length
// Extend backwards
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for t > tMin && s > nextEmit && src[t-1] == src[s-1] && l < maxMatchLength {
s--
t--
l++
}
// Write our sequence
var seq seq
seq.litLen = uint32(s - nextEmit)
seq.matchLen = uint32(l - zstdMinMatch)
if seq.litLen > 0 {
blk.literals = append(blk.literals, src[nextEmit:s]...)
}
seq.offset = uint32(s-t) + 3
s += l
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
nextEmit = s
if s >= sLimit {
break encodeLoop
}
// Index match start+1 (long) -> s - 1
index0 := s - l + 1
// every entry
for index0 < s-1 {
cv0 := load6432(src, index0)
h0 := hash8(cv0, bestLongTableBits)
h1 := hash4x64(cv0, bestShortTableBits)
off := index0 + e.cur
e.longTable[h0] = prevEntry{offset: off, prev: e.longTable[h0].offset}
e.table[h1] = prevEntry{offset: off, prev: e.table[h1].offset}
index0++
}
cv = load6432(src, s)
if !canRepeat {
continue
}
// Check offset 2
for {
o2 := s - offset2
if load3232(src, o2) != uint32(cv) {
// Do regular search
break
}
// Store this, since we have it.
nextHashS := hash4x64(cv, bestShortTableBits)
nextHashL := hash8(cv, bestLongTableBits)
// 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)
e.longTable[nextHashL] = prevEntry{offset: s + e.cur, prev: e.longTable[nextHashL].offset}
e.table[nextHashS] = prevEntry{offset: s + e.cur, prev: e.table[nextHashS].offset}
seq.matchLen = uint32(l) - zstdMinMatch
seq.litLen = 0
// Since litlen is always 0, this is offset 1.
seq.offset = 1
s += l
nextEmit = s
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Swap offset 1 and 2.
offset1, offset2 = offset2, offset1
if s >= sLimit {
// Finished
break encodeLoop
}
cv = load6432(src, s)
}
}
if int(nextEmit) < len(src) {
blk.literals = append(blk.literals, src[nextEmit:]...)
blk.extraLits = len(src) - int(nextEmit)
}
blk.recentOffsets[0] = uint32(offset1)
blk.recentOffsets[1] = uint32(offset2)
blk.recentOffsets[2] = uint32(offset3)
if debug {
println("returning, recent offsets:", blk.recentOffsets, "extra literals:", blk.extraLits)
}
}
// EncodeNoHist will encode a block with no history and no following blocks.
// Most notable difference is that src will not be copied for history and
// we do not need to check for max match length.
func (e *bestFastEncoder) EncodeNoHist(blk *blockEnc, src []byte) {
e.ensureHist(len(src))
e.Encode(blk, src)
}
// ResetDict will reset and set a dictionary if not nil
func (e *bestFastEncoder) Reset(d *dict, singleBlock bool) {
e.resetBase(d, singleBlock)
if d == nil {
return
}
// Init or copy dict table
if len(e.dictTable) != len(e.table) || d.id != e.lastDictID {
if len(e.dictTable) != len(e.table) {
e.dictTable = make([]prevEntry, len(e.table))
}
end := int32(len(d.content)) - 8 + e.maxMatchOff
for i := e.maxMatchOff; i < end; i += 4 {
const hashLog = bestShortTableBits
cv := load6432(d.content, i-e.maxMatchOff)
nextHash := hash4x64(cv, hashLog) // 0 -> 4
nextHash1 := hash4x64(cv>>8, hashLog) // 1 -> 5
nextHash2 := hash4x64(cv>>16, hashLog) // 2 -> 6
nextHash3 := hash4x64(cv>>24, hashLog) // 3 -> 7
e.dictTable[nextHash] = prevEntry{
prev: e.dictTable[nextHash].offset,
offset: i,
}
e.dictTable[nextHash1] = prevEntry{
prev: e.dictTable[nextHash1].offset,
offset: i + 1,
}
e.dictTable[nextHash2] = prevEntry{
prev: e.dictTable[nextHash2].offset,
offset: i + 2,
}
e.dictTable[nextHash3] = prevEntry{
prev: e.dictTable[nextHash3].offset,
offset: i + 3,
}
}
e.lastDictID = d.id
}
// Init or copy dict table
if len(e.dictLongTable) != len(e.longTable) || d.id != e.lastDictID {
if len(e.dictLongTable) != len(e.longTable) {
e.dictLongTable = make([]prevEntry, len(e.longTable))
}
if len(d.content) >= 8 {
cv := load6432(d.content, 0)
h := hash8(cv, bestLongTableBits)
e.dictLongTable[h] = prevEntry{
offset: e.maxMatchOff,
prev: e.dictLongTable[h].offset,
}
end := int32(len(d.content)) - 8 + e.maxMatchOff
off := 8 // First to read
for i := e.maxMatchOff + 1; i < end; i++ {
cv = cv>>8 | (uint64(d.content[off]) << 56)
h := hash8(cv, bestLongTableBits)
e.dictLongTable[h] = prevEntry{
offset: i,
prev: e.dictLongTable[h].offset,
}
off++
}
}
e.lastDictID = d.id
}
// Reset table to initial state
copy(e.longTable[:], e.dictLongTable)
e.cur = e.maxMatchOff
// Reset table to initial state
copy(e.table[:], e.dictTable)
}

660
vendor/github.com/klauspost/compress/zstd/enc_better.go generated vendored
View File

@ -16,6 +16,12 @@ const (
// This greatly depends on the type of input.
betterShortTableBits = 13 // Bits used in the short match table
betterShortTableSize = 1 << betterShortTableBits // Size of the table
betterLongTableShardCnt = 1 << (betterLongTableBits - dictShardBits) // Number of shards in the table
betterLongTableShardSize = betterLongTableSize / betterLongTableShardCnt // Size of an individual shard
betterShortTableShardCnt = 1 << (betterShortTableBits - dictShardBits) // Number of shards in the table
betterShortTableShardSize = betterShortTableSize / betterShortTableShardCnt // Size of an individual shard
)
type prevEntry struct {
@ -31,10 +37,17 @@ type prevEntry struct {
// and that it is longer (lazy matching).
type betterFastEncoder struct {
fastBase
table [betterShortTableSize]tableEntry
longTable [betterLongTableSize]prevEntry
dictTable []tableEntry
dictLongTable []prevEntry
table [betterShortTableSize]tableEntry
longTable [betterLongTableSize]prevEntry
}
type betterFastEncoderDict struct {
betterFastEncoder
dictTable []tableEntry
dictLongTable []prevEntry
shortTableShardDirty [betterShortTableShardCnt]bool
longTableShardDirty [betterLongTableShardCnt]bool
allDirty bool
}
// Encode improves compression...
@ -399,8 +412,41 @@ encodeLoop:
cv = load6432(src, s)
}
// A 4-byte match has been found. Update recent offsets.
// We'll later see if more than 4 bytes.
// Try to find a better match by searching for a long match at the end of the current best match
if true && s+matched < sLimit {
nextHashL := hash8(load6432(src, s+matched), betterLongTableBits)
cv := load3232(src, s)
candidateL := e.longTable[nextHashL]
coffsetL := candidateL.offset - e.cur - matched
if coffsetL >= 0 && coffsetL < s && s-coffsetL < e.maxMatchOff && cv == load3232(src, coffsetL) {
// Found a long match, at least 4 bytes.
matchedNext := e.matchlen(s+4, coffsetL+4, src) + 4
if matchedNext > matched {
t = coffsetL
matched = matchedNext
if debugMatches {
println("long match at end-of-match")
}
}
}
// Check prev long...
if true {
coffsetL = candidateL.prev - e.cur - matched
if coffsetL >= 0 && coffsetL < s && s-coffsetL < e.maxMatchOff && cv == load3232(src, coffsetL) {
// Found a long match, at least 4 bytes.
matchedNext := e.matchlen(s+4, coffsetL+4, src) + 4
if matchedNext > matched {
t = coffsetL
matched = matchedNext
if debugMatches {
println("prev long match at end-of-match")
}
}
}
}
}
// A match has been found. Update recent offsets.
offset2 = offset1
offset1 = s - t
@ -516,11 +562,543 @@ encodeLoop:
// Most notable difference is that src will not be copied for history and
// we do not need to check for max match length.
func (e *betterFastEncoder) EncodeNoHist(blk *blockEnc, src []byte) {
e.ensureHist(len(src))
e.Encode(blk, src)
}
// Encode improves compression...
func (e *betterFastEncoderDict) Encode(blk *blockEnc, src []byte) {
const (
// Input margin is the number of bytes we read (8)
// and the maximum we will read ahead (2)
inputMargin = 8 + 2
minNonLiteralBlockSize = 16
)
// Protect against e.cur wraparound.
for e.cur >= bufferReset {
if len(e.hist) == 0 {
for i := range e.table[:] {
e.table[i] = tableEntry{}
}
for i := range e.longTable[:] {
e.longTable[i] = prevEntry{}
}
e.cur = e.maxMatchOff
e.allDirty = true
break
}
// Shift down everything in the table that isn't already too far away.
minOff := e.cur + int32(len(e.hist)) - e.maxMatchOff
for i := range e.table[:] {
v := e.table[i].offset
if v < minOff {
v = 0
} else {
v = v - e.cur + e.maxMatchOff
}
e.table[i].offset = v
}
for i := range e.longTable[:] {
v := e.longTable[i].offset
v2 := e.longTable[i].prev
if v < minOff {
v = 0
v2 = 0
} else {
v = v - e.cur + e.maxMatchOff
if v2 < minOff {
v2 = 0
} else {
v2 = v2 - e.cur + e.maxMatchOff
}
}
e.longTable[i] = prevEntry{
offset: v,
prev: v2,
}
}
e.allDirty = true
e.cur = e.maxMatchOff
break
}
s := e.addBlock(src)
blk.size = len(src)
if len(src) < minNonLiteralBlockSize {
blk.extraLits = len(src)
blk.literals = blk.literals[:len(src)]
copy(blk.literals, src)
return
}
// Override src
src = e.hist
sLimit := int32(len(src)) - inputMargin
// stepSize is the number of bytes to skip on every main loop iteration.
// It should be >= 1.
const stepSize = 1
const kSearchStrength = 9
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := s
cv := load6432(src, s)
// Relative offsets
offset1 := int32(blk.recentOffsets[0])
offset2 := int32(blk.recentOffsets[1])
addLiterals := func(s *seq, until int32) {
if until == nextEmit {
return
}
blk.literals = append(blk.literals, src[nextEmit:until]...)
s.litLen = uint32(until - nextEmit)
}
if debug {
println("recent offsets:", blk.recentOffsets)
}
encodeLoop:
for {
var t int32
// We allow the encoder to optionally turn off repeat offsets across blocks
canRepeat := len(blk.sequences) > 2
var matched int32
for {
if debugAsserts && canRepeat && offset1 == 0 {
panic("offset0 was 0")
}
nextHashS := hash5(cv, betterShortTableBits)
nextHashL := hash8(cv, betterLongTableBits)
candidateL := e.longTable[nextHashL]
candidateS := e.table[nextHashS]
const repOff = 1
repIndex := s - offset1 + repOff
off := s + e.cur
e.longTable[nextHashL] = prevEntry{offset: off, prev: candidateL.offset}
e.markLongShardDirty(nextHashL)
e.table[nextHashS] = tableEntry{offset: off, val: uint32(cv)}
e.markShortShardDirty(nextHashS)
if canRepeat {
if repIndex >= 0 && load3232(src, repIndex) == uint32(cv>>(repOff*8)) {
// Consider history as well.
var seq seq
lenght := 4 + e.matchlen(s+4+repOff, repIndex+4, src)
seq.matchLen = uint32(lenght - zstdMinMatch)
// We might be able to match backwards.
// Extend as long as we can.
start := s + repOff
// We end the search early, so we don't risk 0 literals
// and have to do special offset treatment.
startLimit := nextEmit + 1
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for repIndex > tMin && start > startLimit && src[repIndex-1] == src[start-1] && seq.matchLen < maxMatchLength-zstdMinMatch-1 {
repIndex--
start--
seq.matchLen++
}
addLiterals(&seq, start)
// rep 0
seq.offset = 1
if debugSequences {
println("repeat sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Index match start+1 (long) -> s - 1
index0 := s + repOff
s += lenght + repOff
nextEmit = s
if s >= sLimit {
if debug {
println("repeat ended", s, lenght)
}
break encodeLoop
}
// Index skipped...
for index0 < s-1 {
cv0 := load6432(src, index0)
cv1 := cv0 >> 8
h0 := hash8(cv0, betterLongTableBits)
off := index0 + e.cur
e.longTable[h0] = prevEntry{offset: off, prev: e.longTable[h0].offset}
e.markLongShardDirty(h0)
h1 := hash5(cv1, betterShortTableBits)
e.table[h1] = tableEntry{offset: off + 1, val: uint32(cv1)}
e.markShortShardDirty(h1)
index0 += 2
}
cv = load6432(src, s)
continue
}
const repOff2 = 1
// We deviate from the reference encoder and also check offset 2.
// Still slower and not much better, so disabled.
// repIndex = s - offset2 + repOff2
if false && repIndex >= 0 && load6432(src, repIndex) == load6432(src, s+repOff) {
// Consider history as well.
var seq seq
lenght := 8 + e.matchlen(s+8+repOff2, repIndex+8, src)
seq.matchLen = uint32(lenght - zstdMinMatch)
// We might be able to match backwards.
// Extend as long as we can.
start := s + repOff2
// We end the search early, so we don't risk 0 literals
// and have to do special offset treatment.
startLimit := nextEmit + 1
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for repIndex > tMin && start > startLimit && src[repIndex-1] == src[start-1] && seq.matchLen < maxMatchLength-zstdMinMatch-1 {
repIndex--
start--
seq.matchLen++
}
addLiterals(&seq, start)
// rep 2
seq.offset = 2
if debugSequences {
println("repeat sequence 2", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
index0 := s + repOff2
s += lenght + repOff2
nextEmit = s
if s >= sLimit {
if debug {
println("repeat ended", s, lenght)
}
break encodeLoop
}
// Index skipped...
for index0 < s-1 {
cv0 := load6432(src, index0)
cv1 := cv0 >> 8
h0 := hash8(cv0, betterLongTableBits)
off := index0 + e.cur
e.longTable[h0] = prevEntry{offset: off, prev: e.longTable[h0].offset}
e.markLongShardDirty(h0)
h1 := hash5(cv1, betterShortTableBits)
e.table[h1] = tableEntry{offset: off + 1, val: uint32(cv1)}
e.markShortShardDirty(h1)
index0 += 2
}
cv = load6432(src, s)
// Swap offsets
offset1, offset2 = offset2, offset1
continue
}
}
// Find the offsets of our two matches.
coffsetL := candidateL.offset - e.cur
coffsetLP := candidateL.prev - e.cur
// Check if we have a long match.
if s-coffsetL < e.maxMatchOff && cv == load6432(src, coffsetL) {
// Found a long match, at least 8 bytes.
matched = e.matchlen(s+8, coffsetL+8, src) + 8
t = coffsetL
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugMatches {
println("long match")
}
if s-coffsetLP < e.maxMatchOff && cv == load6432(src, coffsetLP) {
// Found a long match, at least 8 bytes.
prevMatch := e.matchlen(s+8, coffsetLP+8, src) + 8
if prevMatch > matched {
matched = prevMatch
t = coffsetLP
}
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugMatches {
println("long match")
}
}
break
}
// Check if we have a long match on prev.
if s-coffsetLP < e.maxMatchOff && cv == load6432(src, coffsetLP) {
// Found a long match, at least 8 bytes.
matched = e.matchlen(s+8, coffsetLP+8, src) + 8
t = coffsetLP
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugMatches {
println("long match")
}
break
}
coffsetS := candidateS.offset - e.cur
// Check if we have a short match.
if s-coffsetS < e.maxMatchOff && uint32(cv) == candidateS.val {
// found a regular match
matched = e.matchlen(s+4, coffsetS+4, src) + 4
// See if we can find a long match at s+1
const checkAt = 1
cv := load6432(src, s+checkAt)
nextHashL = hash8(cv, betterLongTableBits)
candidateL = e.longTable[nextHashL]
coffsetL = candidateL.offset - e.cur
// We can store it, since we have at least a 4 byte match.
e.longTable[nextHashL] = prevEntry{offset: s + checkAt + e.cur, prev: candidateL.offset}
e.markLongShardDirty(nextHashL)
if s-coffsetL < e.maxMatchOff && cv == load6432(src, coffsetL) {
// Found a long match, at least 8 bytes.
matchedNext := e.matchlen(s+8+checkAt, coffsetL+8, src) + 8
if matchedNext > matched {
t = coffsetL
s += checkAt
matched = matchedNext
if debugMatches {
println("long match (after short)")
}
break
}
}
// Check prev long...
coffsetL = candidateL.prev - e.cur
if s-coffsetL < e.maxMatchOff && cv == load6432(src, coffsetL) {
// Found a long match, at least 8 bytes.
matchedNext := e.matchlen(s+8+checkAt, coffsetL+8, src) + 8
if matchedNext > matched {
t = coffsetL
s += checkAt
matched = matchedNext
if debugMatches {
println("prev long match (after short)")
}
break
}
}
t = coffsetS
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugAsserts && t < 0 {
panic("t<0")
}
if debugMatches {
println("short match")
}
break
}
// No match found, move forward in input.
s += stepSize + ((s - nextEmit) >> (kSearchStrength - 1))
if s >= sLimit {
break encodeLoop
}
cv = load6432(src, s)
}
// Try to find a better match by searching for a long match at the end of the current best match
if s+matched < sLimit {
nextHashL := hash8(load6432(src, s+matched), betterLongTableBits)
cv := load3232(src, s)
candidateL := e.longTable[nextHashL]
coffsetL := candidateL.offset - e.cur - matched
if coffsetL >= 0 && coffsetL < s && s-coffsetL < e.maxMatchOff && cv == load3232(src, coffsetL) {
// Found a long match, at least 4 bytes.
matchedNext := e.matchlen(s+4, coffsetL+4, src) + 4
if matchedNext > matched {
t = coffsetL
matched = matchedNext
if debugMatches {
println("long match at end-of-match")
}
}
}
// Check prev long...
if true {
coffsetL = candidateL.prev - e.cur - matched
if coffsetL >= 0 && coffsetL < s && s-coffsetL < e.maxMatchOff && cv == load3232(src, coffsetL) {
// Found a long match, at least 4 bytes.
matchedNext := e.matchlen(s+4, coffsetL+4, src) + 4
if matchedNext > matched {
t = coffsetL
matched = matchedNext
if debugMatches {
println("prev long match at end-of-match")
}
}
}
}
}
// A match has been found. Update recent offsets.
offset2 = offset1
offset1 = s - t
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && canRepeat && int(offset1) > len(src) {
panic("invalid offset")
}
// Extend the n-byte match as long as possible.
l := matched
// Extend backwards
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for t > tMin && s > nextEmit && src[t-1] == src[s-1] && l < maxMatchLength {
s--
t--
l++
}
// Write our sequence
var seq seq
seq.litLen = uint32(s - nextEmit)
seq.matchLen = uint32(l - zstdMinMatch)
if seq.litLen > 0 {
blk.literals = append(blk.literals, src[nextEmit:s]...)
}
seq.offset = uint32(s-t) + 3
s += l
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
nextEmit = s
if s >= sLimit {
break encodeLoop
}
// Index match start+1 (long) -> s - 1
index0 := s - l + 1
for index0 < s-1 {
cv0 := load6432(src, index0)
cv1 := cv0 >> 8
h0 := hash8(cv0, betterLongTableBits)
off := index0 + e.cur
e.longTable[h0] = prevEntry{offset: off, prev: e.longTable[h0].offset}
e.markLongShardDirty(h0)
h1 := hash5(cv1, betterShortTableBits)
e.table[h1] = tableEntry{offset: off + 1, val: uint32(cv1)}
e.markShortShardDirty(h1)
index0 += 2
}
cv = load6432(src, s)
if !canRepeat {
continue
}
// Check offset 2
for {
o2 := s - offset2
if load3232(src, o2) != uint32(cv) {
// Do regular search
break
}
// Store this, since we have it.
nextHashS := hash5(cv, betterShortTableBits)
nextHashL := hash8(cv, betterLongTableBits)
// 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)
e.longTable[nextHashL] = prevEntry{offset: s + e.cur, prev: e.longTable[nextHashL].offset}
e.markLongShardDirty(nextHashL)
e.table[nextHashS] = tableEntry{offset: s + e.cur, val: uint32(cv)}
e.markShortShardDirty(nextHashS)
seq.matchLen = uint32(l) - zstdMinMatch
seq.litLen = 0
// Since litlen is always 0, this is offset 1.
seq.offset = 1
s += l
nextEmit = s
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Swap offset 1 and 2.
offset1, offset2 = offset2, offset1
if s >= sLimit {
// Finished
break encodeLoop
}
cv = load6432(src, s)
}
}
if int(nextEmit) < len(src) {
blk.literals = append(blk.literals, src[nextEmit:]...)
blk.extraLits = len(src) - int(nextEmit)
}
blk.recentOffsets[0] = uint32(offset1)
blk.recentOffsets[1] = uint32(offset2)
if debug {
println("returning, recent offsets:", blk.recentOffsets, "extra literals:", blk.extraLits)
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *betterFastEncoder) Reset(d *dict, singleBlock bool) {
e.resetBase(d, singleBlock)
if d != nil {
panic("betterFastEncoder: Reset with dict")
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *betterFastEncoderDict) Reset(d *dict, singleBlock bool) {
e.resetBase(d, singleBlock)
if d == nil {
return
@ -557,6 +1135,7 @@ func (e *betterFastEncoder) Reset(d *dict, singleBlock bool) {
}
}
e.lastDictID = d.id
e.allDirty = true
}
// Init or copy dict table
@ -585,11 +1164,72 @@ func (e *betterFastEncoder) Reset(d *dict, singleBlock bool) {
}
}
e.lastDictID = d.id
e.allDirty = true
}
// Reset table to initial state
copy(e.longTable[:], e.dictLongTable)
e.cur = e.maxMatchOff
// Reset table to initial state
copy(e.table[:], e.dictTable)
{
dirtyShardCnt := 0
if !e.allDirty {
for i := range e.shortTableShardDirty {
if e.shortTableShardDirty[i] {
dirtyShardCnt++
}
}
}
const shardCnt = betterShortTableShardCnt
const shardSize = betterShortTableShardSize
if e.allDirty || dirtyShardCnt > shardCnt*4/6 {
copy(e.table[:], e.dictTable)
for i := range e.shortTableShardDirty {
e.shortTableShardDirty[i] = false
}
} else {
for i := range e.shortTableShardDirty {
if !e.shortTableShardDirty[i] {
continue
}
copy(e.table[i*shardSize:(i+1)*shardSize], e.dictTable[i*shardSize:(i+1)*shardSize])
e.shortTableShardDirty[i] = false
}
}
}
{
dirtyShardCnt := 0
if !e.allDirty {
for i := range e.shortTableShardDirty {
if e.shortTableShardDirty[i] {
dirtyShardCnt++
}
}
}
const shardCnt = betterLongTableShardCnt
const shardSize = betterLongTableShardSize
if e.allDirty || dirtyShardCnt > shardCnt*4/6 {
copy(e.longTable[:], e.dictLongTable)
for i := range e.longTableShardDirty {
e.longTableShardDirty[i] = false
}
} else {
for i := range e.longTableShardDirty {
if !e.longTableShardDirty[i] {
continue
}
copy(e.longTable[i*shardSize:(i+1)*shardSize], e.dictLongTable[i*shardSize:(i+1)*shardSize])
e.longTableShardDirty[i] = false
}
}
}
e.cur = e.maxMatchOff
e.allDirty = false
}
func (e *betterFastEncoderDict) markLongShardDirty(entryNum uint32) {
e.longTableShardDirty[entryNum/betterLongTableShardSize] = true
}
func (e *betterFastEncoderDict) markShortShardDirty(entryNum uint32) {
e.shortTableShardDirty[entryNum/betterShortTableShardSize] = true
}

414
vendor/github.com/klauspost/compress/zstd/enc_dfast.go generated vendored
View File

@ -11,6 +11,9 @@ const (
dFastLongTableSize = 1 << dFastLongTableBits // Size of the table
dFastLongTableMask = dFastLongTableSize - 1 // Mask for table indices. Redundant, but can eliminate bounds checks.
dLongTableShardCnt = 1 << (dFastLongTableBits - dictShardBits) // Number of shards in the table
dLongTableShardSize = dFastLongTableSize / tableShardCnt // Size of an individual shard
dFastShortTableBits = tableBits // Bits used in the short match table
dFastShortTableSize = 1 << dFastShortTableBits // Size of the table
dFastShortTableMask = dFastShortTableSize - 1 // Mask for table indices. Redundant, but can eliminate bounds checks.
@ -18,8 +21,14 @@ const (
type doubleFastEncoder struct {
fastEncoder
longTable [dFastLongTableSize]tableEntry
dictLongTable []tableEntry
longTable [dFastLongTableSize]tableEntry
}
type doubleFastEncoderDict struct {
fastEncoderDict
longTable [dFastLongTableSize]tableEntry
dictLongTable []tableEntry
longTableShardDirty [dLongTableShardCnt]bool
}
// Encode mimmics functionality in zstd_dfast.c
@ -678,9 +687,379 @@ encodeLoop:
}
}
// Encode will encode the content, with a dictionary if initialized for it.
func (e *doubleFastEncoderDict) Encode(blk *blockEnc, src []byte) {
const (
// Input margin is the number of bytes we read (8)
// and the maximum we will read ahead (2)
inputMargin = 8 + 2
minNonLiteralBlockSize = 16
)
// Protect against e.cur wraparound.
for e.cur >= bufferReset {
if len(e.hist) == 0 {
for i := range e.table[:] {
e.table[i] = tableEntry{}
}
for i := range e.longTable[:] {
e.longTable[i] = tableEntry{}
}
e.markAllShardsDirty()
e.cur = e.maxMatchOff
break
}
// Shift down everything in the table that isn't already too far away.
minOff := e.cur + int32(len(e.hist)) - e.maxMatchOff
for i := range e.table[:] {
v := e.table[i].offset
if v < minOff {
v = 0
} else {
v = v - e.cur + e.maxMatchOff
}
e.table[i].offset = v
}
for i := range e.longTable[:] {
v := e.longTable[i].offset
if v < minOff {
v = 0
} else {
v = v - e.cur + e.maxMatchOff
}
e.longTable[i].offset = v
}
e.markAllShardsDirty()
e.cur = e.maxMatchOff
break
}
s := e.addBlock(src)
blk.size = len(src)
if len(src) < minNonLiteralBlockSize {
blk.extraLits = len(src)
blk.literals = blk.literals[:len(src)]
copy(blk.literals, src)
return
}
// Override src
src = e.hist
sLimit := int32(len(src)) - inputMargin
// stepSize is the number of bytes to skip on every main loop iteration.
// It should be >= 1.
const stepSize = 1
const kSearchStrength = 8
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := s
cv := load6432(src, s)
// Relative offsets
offset1 := int32(blk.recentOffsets[0])
offset2 := int32(blk.recentOffsets[1])
addLiterals := func(s *seq, until int32) {
if until == nextEmit {
return
}
blk.literals = append(blk.literals, src[nextEmit:until]...)
s.litLen = uint32(until - nextEmit)
}
if debug {
println("recent offsets:", blk.recentOffsets)
}
encodeLoop:
for {
var t int32
// We allow the encoder to optionally turn off repeat offsets across blocks
canRepeat := len(blk.sequences) > 2
for {
if debugAsserts && canRepeat && offset1 == 0 {
panic("offset0 was 0")
}
nextHashS := hash5(cv, dFastShortTableBits)
nextHashL := hash8(cv, dFastLongTableBits)
candidateL := e.longTable[nextHashL]
candidateS := e.table[nextHashS]
const repOff = 1
repIndex := s - offset1 + repOff
entry := tableEntry{offset: s + e.cur, val: uint32(cv)}
e.longTable[nextHashL] = entry
e.markLongShardDirty(nextHashL)
e.table[nextHashS] = entry
e.markShardDirty(nextHashS)
if canRepeat {
if repIndex >= 0 && load3232(src, repIndex) == uint32(cv>>(repOff*8)) {
// Consider history as well.
var seq seq
lenght := 4 + e.matchlen(s+4+repOff, repIndex+4, src)
seq.matchLen = uint32(lenght - zstdMinMatch)
// We might be able to match backwards.
// Extend as long as we can.
start := s + repOff
// We end the search early, so we don't risk 0 literals
// and have to do special offset treatment.
startLimit := nextEmit + 1
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for repIndex > tMin && start > startLimit && src[repIndex-1] == src[start-1] && seq.matchLen < maxMatchLength-zstdMinMatch-1 {
repIndex--
start--
seq.matchLen++
}
addLiterals(&seq, start)
// rep 0
seq.offset = 1
if debugSequences {
println("repeat sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
s += lenght + repOff
nextEmit = s
if s >= sLimit {
if debug {
println("repeat ended", s, lenght)
}
break encodeLoop
}
cv = load6432(src, s)
continue
}
}
// Find the offsets of our two matches.
coffsetL := s - (candidateL.offset - e.cur)
coffsetS := s - (candidateS.offset - e.cur)
// Check if we have a long match.
if coffsetL < e.maxMatchOff && uint32(cv) == candidateL.val {
// Found a long match, likely at least 8 bytes.
// Reference encoder checks all 8 bytes, we only check 4,
// but the likelihood of both the first 4 bytes and the hash matching should be enough.
t = candidateL.offset - e.cur
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugMatches {
println("long match")
}
break
}
// Check if we have a short match.
if coffsetS < e.maxMatchOff && uint32(cv) == candidateS.val {
// found a regular match
// See if we can find a long match at s+1
const checkAt = 1
cv := load6432(src, s+checkAt)
nextHashL = hash8(cv, dFastLongTableBits)
candidateL = e.longTable[nextHashL]
coffsetL = s - (candidateL.offset - e.cur) + checkAt
// We can store it, since we have at least a 4 byte match.
e.longTable[nextHashL] = tableEntry{offset: s + checkAt + e.cur, val: uint32(cv)}
e.markLongShardDirty(nextHashL)
if coffsetL < e.maxMatchOff && uint32(cv) == candidateL.val {
// Found a long match, likely at least 8 bytes.
// Reference encoder checks all 8 bytes, we only check 4,
// but the likelihood of both the first 4 bytes and the hash matching should be enough.
t = candidateL.offset - e.cur
s += checkAt
if debugMatches {
println("long match (after short)")
}
break
}
t = candidateS.offset - e.cur
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugAsserts && t < 0 {
panic("t<0")
}
if debugMatches {
println("short match")
}
break
}
// No match found, move forward in input.
s += stepSize + ((s - nextEmit) >> (kSearchStrength - 1))
if s >= sLimit {
break encodeLoop
}
cv = load6432(src, s)
}
// A 4-byte match has been found. Update recent offsets.
// We'll later see if more than 4 bytes.
offset2 = offset1
offset1 = s - t
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && canRepeat && int(offset1) > len(src) {
panic("invalid offset")
}
// Extend the 4-byte match as long as possible.
l := e.matchlen(s+4, t+4, src) + 4
// Extend backwards
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for t > tMin && s > nextEmit && src[t-1] == src[s-1] && l < maxMatchLength {
s--
t--
l++
}
// Write our sequence
var seq seq
seq.litLen = uint32(s - nextEmit)
seq.matchLen = uint32(l - zstdMinMatch)
if seq.litLen > 0 {
blk.literals = append(blk.literals, src[nextEmit:s]...)
}
seq.offset = uint32(s-t) + 3
s += l
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
nextEmit = s
if s >= sLimit {
break encodeLoop
}
// Index match start+1 (long) and start+2 (short)
index0 := s - l + 1
// Index match end-2 (long) and end-1 (short)
index1 := s - 2
cv0 := load6432(src, index0)
cv1 := load6432(src, index1)
te0 := tableEntry{offset: index0 + e.cur, val: uint32(cv0)}
te1 := tableEntry{offset: index1 + e.cur, val: uint32(cv1)}
longHash1 := hash8(cv0, dFastLongTableBits)
longHash2 := hash8(cv0, dFastLongTableBits)
e.longTable[longHash1] = te0
e.longTable[longHash2] = te1
e.markLongShardDirty(longHash1)
e.markLongShardDirty(longHash2)
cv0 >>= 8
cv1 >>= 8
te0.offset++
te1.offset++
te0.val = uint32(cv0)
te1.val = uint32(cv1)
hashVal1 := hash5(cv0, dFastShortTableBits)
hashVal2 := hash5(cv1, dFastShortTableBits)
e.table[hashVal1] = te0
e.markShardDirty(hashVal1)
e.table[hashVal2] = te1
e.markShardDirty(hashVal2)
cv = load6432(src, s)
if !canRepeat {
continue
}
// Check offset 2
for {
o2 := s - offset2
if load3232(src, o2) != uint32(cv) {
// Do regular search
break
}
// Store this, since we have it.
nextHashS := hash5(cv, dFastShortTableBits)
nextHashL := hash8(cv, dFastLongTableBits)
// 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)
entry := tableEntry{offset: s + e.cur, val: uint32(cv)}
e.longTable[nextHashL] = entry
e.markLongShardDirty(nextHashL)
e.table[nextHashS] = entry
e.markShardDirty(nextHashS)
seq.matchLen = uint32(l) - zstdMinMatch
seq.litLen = 0
// Since litlen is always 0, this is offset 1.
seq.offset = 1
s += l
nextEmit = s
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Swap offset 1 and 2.
offset1, offset2 = offset2, offset1
if s >= sLimit {
// Finished
break encodeLoop
}
cv = load6432(src, s)
}
}
if int(nextEmit) < len(src) {
blk.literals = append(blk.literals, src[nextEmit:]...)
blk.extraLits = len(src) - int(nextEmit)
}
blk.recentOffsets[0] = uint32(offset1)
blk.recentOffsets[1] = uint32(offset2)
if debug {
println("returning, recent offsets:", blk.recentOffsets, "extra literals:", blk.extraLits)
}
// If we encoded more than 64K mark all dirty.
if len(src) > 64<<10 {
e.markAllShardsDirty()
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *doubleFastEncoder) Reset(d *dict, singleBlock bool) {
e.fastEncoder.Reset(d, singleBlock)
if d != nil {
panic("doubleFastEncoder: Reset with dict not supported")
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *doubleFastEncoderDict) Reset(d *dict, singleBlock bool) {
allDirty := e.allDirty
e.fastEncoderDict.Reset(d, singleBlock)
if d == nil {
return
}
@ -706,8 +1085,37 @@ func (e *doubleFastEncoder) Reset(d *dict, singleBlock bool) {
}
}
e.lastDictID = d.id
e.allDirty = true
}
// Reset table to initial state
e.cur = e.maxMatchOff
copy(e.longTable[:], e.dictLongTable)
dirtyShardCnt := 0
if !allDirty {
for i := range e.longTableShardDirty {
if e.longTableShardDirty[i] {
dirtyShardCnt++
}
}
}
if allDirty || dirtyShardCnt > dLongTableShardCnt/2 {
copy(e.longTable[:], e.dictLongTable)
for i := range e.longTableShardDirty {
e.longTableShardDirty[i] = false
}
return
}
for i := range e.longTableShardDirty {
if !e.longTableShardDirty[i] {
continue
}
copy(e.longTable[i*dLongTableShardSize:(i+1)*dLongTableShardSize], e.dictLongTable[i*dLongTableShardSize:(i+1)*dLongTableShardSize])
e.longTableShardDirty[i] = false
}
}
func (e *doubleFastEncoderDict) markLongShardDirty(entryNum uint32) {
e.longTableShardDirty[entryNum/dLongTableShardSize] = true
}

373
vendor/github.com/klauspost/compress/zstd/enc_fast.go generated vendored
View File

@ -11,9 +11,11 @@ import (
)
const (
tableBits = 15 // Bits used in the table
tableSize = 1 << tableBits // Size of the table
tableMask = tableSize - 1 // Mask for table indices. Redundant, but can eliminate bounds checks.
tableBits = 15 // Bits used in the table
tableSize = 1 << tableBits // Size of the table
tableShardCnt = 1 << (tableBits - dictShardBits) // Number of shards in the table
tableShardSize = tableSize / tableShardCnt // Size of an individual shard
tableMask = tableSize - 1 // Mask for table indices. Redundant, but can eliminate bounds checks.
maxMatchLength = 131074
)
@ -24,8 +26,14 @@ type tableEntry struct {
type fastEncoder struct {
fastBase
table [tableSize]tableEntry
dictTable []tableEntry
table [tableSize]tableEntry
}
type fastEncoderDict struct {
fastEncoder
dictTable []tableEntry
tableShardDirty [tableShardCnt]bool
allDirty bool
}
// Encode mimmics functionality in zstd_fast.c
@ -78,7 +86,7 @@ func (e *fastEncoder) Encode(blk *blockEnc, src []byte) {
// TEMPLATE
const hashLog = tableBits
// seems global, but would be nice to tweak.
const kSearchStrength = 8
const kSearchStrength = 7
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := s
@ -617,8 +625,322 @@ encodeLoop:
}
}
// Encode will encode the content, with a dictionary if initialized for it.
func (e *fastEncoderDict) Encode(blk *blockEnc, src []byte) {
const (
inputMargin = 8
minNonLiteralBlockSize = 1 + 1 + inputMargin
)
if e.allDirty || len(src) > 32<<10 {
e.fastEncoder.Encode(blk, src)
e.allDirty = true
return
}
// Protect against e.cur wraparound.
for e.cur >= bufferReset {
if len(e.hist) == 0 {
for i := range e.table[:] {
e.table[i] = tableEntry{}
}
e.cur = e.maxMatchOff
break
}
// Shift down everything in the table that isn't already too far away.
minOff := e.cur + int32(len(e.hist)) - e.maxMatchOff
for i := range e.table[:] {
v := e.table[i].offset
if v < minOff {
v = 0
} else {
v = v - e.cur + e.maxMatchOff
}
e.table[i].offset = v
}
e.cur = e.maxMatchOff
break
}
s := e.addBlock(src)
blk.size = len(src)
if len(src) < minNonLiteralBlockSize {
blk.extraLits = len(src)
blk.literals = blk.literals[:len(src)]
copy(blk.literals, src)
return
}
// Override src
src = e.hist
sLimit := int32(len(src)) - inputMargin
// stepSize is the number of bytes to skip on every main loop iteration.
// It should be >= 2.
const stepSize = 2
// TEMPLATE
const hashLog = tableBits
// seems global, but would be nice to tweak.
const kSearchStrength = 7
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := s
cv := load6432(src, s)
// Relative offsets
offset1 := int32(blk.recentOffsets[0])
offset2 := int32(blk.recentOffsets[1])
addLiterals := func(s *seq, until int32) {
if until == nextEmit {
return
}
blk.literals = append(blk.literals, src[nextEmit:until]...)
s.litLen = uint32(until - nextEmit)
}
if debug {
println("recent offsets:", blk.recentOffsets)
}
encodeLoop:
for {
// t will contain the match offset when we find one.
// When existing the search loop, we have already checked 4 bytes.
var t int32
// We will not use repeat offsets across blocks.
// By not using them for the first 3 matches
canRepeat := len(blk.sequences) > 2
for {
if debugAsserts && canRepeat && offset1 == 0 {
panic("offset0 was 0")
}
nextHash := hash6(cv, hashLog)
nextHash2 := hash6(cv>>8, hashLog)
candidate := e.table[nextHash]
candidate2 := e.table[nextHash2]
repIndex := s - offset1 + 2
e.table[nextHash] = tableEntry{offset: s + e.cur, val: uint32(cv)}
e.markShardDirty(nextHash)
e.table[nextHash2] = tableEntry{offset: s + e.cur + 1, val: uint32(cv >> 8)}
e.markShardDirty(nextHash2)
if canRepeat && repIndex >= 0 && load3232(src, repIndex) == uint32(cv>>16) {
// 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
}
}
}
seq.matchLen = uint32(length - zstdMinMatch)
// We might be able to match backwards.
// Extend as long as we can.
start := s + 2
// We end the search early, so we don't risk 0 literals
// and have to do special offset treatment.
startLimit := nextEmit + 1
sMin := s - e.maxMatchOff
if sMin < 0 {
sMin = 0
}
for repIndex > sMin && start > startLimit && src[repIndex-1] == src[start-1] && seq.matchLen < maxMatchLength-zstdMinMatch {
repIndex--
start--
seq.matchLen++
}
addLiterals(&seq, start)
// rep 0
seq.offset = 1
if debugSequences {
println("repeat sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
s += length + 2
nextEmit = s
if s >= sLimit {
if debug {
println("repeat ended", s, length)
}
break encodeLoop
}
cv = load6432(src, s)
continue
}
coffset0 := s - (candidate.offset - e.cur)
coffset1 := s - (candidate2.offset - e.cur) + 1
if coffset0 < e.maxMatchOff && uint32(cv) == candidate.val {
// found a regular match
t = candidate.offset - e.cur
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
break
}
if coffset1 < e.maxMatchOff && uint32(cv>>8) == candidate2.val {
// found a regular match
t = candidate2.offset - e.cur
s++
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && s-t > e.maxMatchOff {
panic("s - t >e.maxMatchOff")
}
if debugAsserts && t < 0 {
panic("t<0")
}
break
}
s += stepSize + ((s - nextEmit) >> (kSearchStrength - 1))
if s >= sLimit {
break encodeLoop
}
cv = load6432(src, s)
}
// A 4-byte match has been found. We'll later see if more than 4 bytes.
offset2 = offset1
offset1 = s - t
if debugAsserts && s <= t {
panic(fmt.Sprintf("s (%d) <= t (%d)", s, t))
}
if debugAsserts && canRepeat && int(offset1) > len(src) {
panic("invalid offset")
}
// 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
}
}
}
// Extend backwards
tMin := s - e.maxMatchOff
if tMin < 0 {
tMin = 0
}
for t > tMin && s > nextEmit && src[t-1] == src[s-1] && l < maxMatchLength {
s--
t--
l++
}
// Write our sequence.
var seq seq
seq.litLen = uint32(s - nextEmit)
seq.matchLen = uint32(l - zstdMinMatch)
if seq.litLen > 0 {
blk.literals = append(blk.literals, src[nextEmit:s]...)
}
// Don't use repeat offsets
seq.offset = uint32(s-t) + 3
s += l
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
nextEmit = s
if s >= sLimit {
break encodeLoop
}
cv = load6432(src, s)
// Check offset 2
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
}
}
}
// Store this, since we have it.
nextHash := hash6(cv, hashLog)
e.table[nextHash] = tableEntry{offset: s + e.cur, val: uint32(cv)}
e.markShardDirty(nextHash)
seq.matchLen = uint32(l) - zstdMinMatch
seq.litLen = 0
// Since litlen is always 0, this is offset 1.
seq.offset = 1
s += l
nextEmit = s
if debugSequences {
println("sequence", seq, "next s:", s)
}
blk.sequences = append(blk.sequences, seq)
// Swap offset 1 and 2.
offset1, offset2 = offset2, offset1
if s >= sLimit {
break encodeLoop
}
// Prepare next loop.
cv = load6432(src, s)
}
}
if int(nextEmit) < len(src) {
blk.literals = append(blk.literals, src[nextEmit:]...)
blk.extraLits = len(src) - int(nextEmit)
}
blk.recentOffsets[0] = uint32(offset1)
blk.recentOffsets[1] = uint32(offset2)
if debug {
println("returning, recent offsets:", blk.recentOffsets, "extra literals:", blk.extraLits)
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *fastEncoder) Reset(d *dict, singleBlock bool) {
e.resetBase(d, singleBlock)
if d != nil {
panic("fastEncoder: Reset with dict")
}
}
// ResetDict will reset and set a dictionary if not nil
func (e *fastEncoderDict) Reset(d *dict, singleBlock bool) {
e.resetBase(d, singleBlock)
if d == nil {
return
@ -653,9 +975,44 @@ func (e *fastEncoder) Reset(d *dict, singleBlock bool) {
}
}
e.lastDictID = d.id
e.allDirty = true
}
e.cur = e.maxMatchOff
// Reset table to initial state
copy(e.table[:], e.dictTable)
dirtyShardCnt := 0
if !e.allDirty {
for i := range e.tableShardDirty {
if e.tableShardDirty[i] {
dirtyShardCnt++
}
}
}
const shardCnt = tableShardCnt
const shardSize = tableShardSize
if e.allDirty || dirtyShardCnt > shardCnt*4/6 {
copy(e.table[:], e.dictTable)
for i := range e.tableShardDirty {
e.tableShardDirty[i] = false
}
e.allDirty = false
return
}
for i := range e.tableShardDirty {
if !e.tableShardDirty[i] {
continue
}
copy(e.table[i*shardSize:(i+1)*shardSize], e.dictTable[i*shardSize:(i+1)*shardSize])
e.tableShardDirty[i] = false
}
e.allDirty = false
}
func (e *fastEncoderDict) markAllShardsDirty() {
e.allDirty = true
}
func (e *fastEncoderDict) markShardDirty(entryNum uint32) {
e.tableShardDirty[entryNum/tableShardSize] = true
}

12
vendor/github.com/klauspost/compress/zstd/encoder.go generated vendored
View File

@ -106,7 +106,7 @@ func (e *Encoder) Reset(w io.Writer) {
s.encoder = e.o.encoder()
}
if s.writing == nil {
s.writing = &blockEnc{}
s.writing = &blockEnc{lowMem: e.o.lowMem}
s.writing.init()
}
s.writing.initNewEncode()
@ -176,6 +176,12 @@ func (e *Encoder) nextBlock(final bool) error {
}
if !s.headerWritten {
// If we have a single block encode, do a sync compression.
if final && len(s.filling) == 0 && !e.o.fullZero {
s.headerWritten = true
s.fullFrameWritten = true
s.eofWritten = true
return nil
}
if final && len(s.filling) > 0 {
s.current = e.EncodeAll(s.filling, s.current[:0])
var n2 int
@ -334,13 +340,13 @@ func (e *Encoder) ReadFrom(r io.Reader) (n int64, err error) {
println("ReadFrom: got EOF final block:", len(e.state.filling))
}
return n, nil
case nil:
default:
if debug {
println("ReadFrom: got error:", err)
}
e.state.err = err
return n, err
case nil:
}
if len(src) > 0 {
if debug {
@ -471,7 +477,7 @@ func (e *Encoder) EncodeAll(src, dst []byte) []byte {
}
// If less than 1MB, allocate a buffer up front.
if len(dst) == 0 && cap(dst) == 0 && len(src) < 1<<20 {
if len(dst) == 0 && cap(dst) == 0 && len(src) < 1<<20 && !e.o.lowMem {
dst = make([]byte, 0, len(src))
}
dst, err := fh.appendTo(dst)

56
vendor/github.com/klauspost/compress/zstd/encoder_options.go generated vendored
View File

@ -24,12 +24,12 @@ type encoderOptions struct {
allLitEntropy bool
customWindow bool
customALEntropy bool
lowMem bool
dict *dict
}
func (o *encoderOptions) setDefault() {
*o = encoderOptions{
// use less ram: true for now, but may change.
concurrent: runtime.GOMAXPROCS(0),
crc: true,
single: nil,
@ -37,18 +37,31 @@ func (o *encoderOptions) setDefault() {
windowSize: 8 << 20,
level: SpeedDefault,
allLitEntropy: true,
lowMem: false,
}
}
// encoder returns an encoder with the selected options.
func (o encoderOptions) encoder() encoder {
switch o.level {
case SpeedDefault:
return &doubleFastEncoder{fastEncoder: fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize)}}}
case SpeedBetterCompression:
return &betterFastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize)}}
case SpeedFastest:
return &fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize)}}
if o.dict != nil {
return &fastEncoderDict{fastEncoder: fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}}
}
return &fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}
case SpeedDefault:
if o.dict != nil {
return &doubleFastEncoderDict{fastEncoderDict: fastEncoderDict{fastEncoder: fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}}}
}
return &doubleFastEncoder{fastEncoder: fastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}}
case SpeedBetterCompression:
if o.dict != nil {
return &betterFastEncoderDict{betterFastEncoder: betterFastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}}
}
return &betterFastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}
case SpeedBestCompression:
return &bestFastEncoder{fastBase: fastBase{maxMatchOff: int32(o.windowSize), lowMem: o.lowMem}}
}
panic("unknown compression level")
}
@ -60,7 +73,7 @@ func WithEncoderCRC(b bool) EOption {
}
// WithEncoderConcurrency will set the concurrency,
// meaning the maximum number of decoders to run concurrently.
// meaning the maximum number of encoders to run concurrently.
// The value supplied must be at least 1.
// By default this will be set to GOMAXPROCS.
func WithEncoderConcurrency(n int) EOption {
@ -143,20 +156,20 @@ const (
// By using this, notice that CPU usage may go up in the future.
SpeedBetterCompression
// SpeedBestCompression will choose the best available compression option.
// This will offer the best compression no matter the CPU cost.
SpeedBestCompression
// speedLast should be kept as the last actual compression option.
// The is not for external usage, but is used to keep track of the valid options.
speedLast
// SpeedBestCompression will choose the best available compression option.
// For now this is not implemented.
SpeedBestCompression = SpeedBetterCompression
)
// EncoderLevelFromString will convert a string representation of an encoding level back
// to a compression level. The compare is not case sensitive.
// If the string wasn't recognized, (false, SpeedDefault) will be returned.
func EncoderLevelFromString(s string) (bool, EncoderLevel) {
for l := EncoderLevel(speedNotSet + 1); l < speedLast; l++ {
for l := speedNotSet + 1; l < speedLast; l++ {
if strings.EqualFold(s, l.String()) {
return true, l
}
@ -173,7 +186,9 @@ func EncoderLevelFromZstd(level int) EncoderLevel {
return SpeedFastest
case level >= 3 && level < 6:
return SpeedDefault
case level > 5:
case level >= 6 && level < 10:
return SpeedBetterCompression
case level >= 10:
return SpeedBetterCompression
}
return SpeedDefault
@ -188,6 +203,8 @@ func (e EncoderLevel) String() string {
return "default"
case SpeedBetterCompression:
return "better"
case SpeedBestCompression:
return "best"
default:
return "invalid"
}
@ -209,6 +226,8 @@ func WithEncoderLevel(l EncoderLevel) EOption {
o.windowSize = 8 << 20
case SpeedBetterCompression:
o.windowSize = 16 << 20
case SpeedBestCompression:
o.windowSize = 32 << 20
}
}
if !o.customALEntropy {
@ -268,6 +287,17 @@ func WithSingleSegment(b bool) EOption {
}
}
// WithLowerEncoderMem will trade in some memory cases trade less memory usage for
// slower encoding speed.
// This will not change the window size which is the primary function for reducing
// memory usage. See WithWindowSize.
func WithLowerEncoderMem(b bool) EOption {
return func(o *encoderOptions) error {
o.lowMem = b
return nil
}
}
// WithEncoderDict allows to register a dictionary that will be used for the encode.
// The encoder *may* choose to use no dictionary instead for certain payloads.
func WithEncoderDict(dict []byte) EOption {

48
vendor/github.com/klauspost/compress/zstd/framedec.go generated vendored
View File

@ -80,9 +80,14 @@ func (d *frameDec) reset(br byteBuffer) error {
d.WindowSize = 0
var b []byte
for {
b = br.readSmall(4)
if b == nil {
var err error
b, err = br.readSmall(4)
switch err {
case io.EOF, io.ErrUnexpectedEOF:
return io.EOF
default:
return err
case nil:
}
if !bytes.Equal(b[1:4], skippableFrameMagic) || b[0]&0xf0 != 0x50 {
if debug {
@ -92,14 +97,14 @@ func (d *frameDec) reset(br byteBuffer) error {
break
}
// Read size to skip
b = br.readSmall(4)
if b == nil {
println("Reading Frame Size EOF")
return io.ErrUnexpectedEOF
b, err = br.readSmall(4)
if err != nil {
println("Reading Frame Size", err)
return err
}
n := uint32(b[0]) | (uint32(b[1]) << 8) | (uint32(b[2]) << 16) | (uint32(b[3]) << 24)
println("Skipping frame with", n, "bytes.")
err := br.skipN(int(n))
err = br.skipN(int(n))
if err != nil {
if debug {
println("Reading discarded frame", err)
@ -121,7 +126,7 @@ func (d *frameDec) reset(br byteBuffer) error {
d.SingleSegment = fhd&(1<<5) != 0
if fhd&(1<<3) != 0 {
return errors.New("Reserved bit set on frame header")
return errors.New("reserved bit set on frame header")
}
// Read Window_Descriptor
@ -147,12 +152,11 @@ func (d *frameDec) reset(br byteBuffer) error {
if size == 3 {
size = 4
}
b = br.readSmall(int(size))
if b == nil {
if debug {
println("Reading Dictionary_ID", io.ErrUnexpectedEOF)
}
return io.ErrUnexpectedEOF
b, err = br.readSmall(int(size))
if err != nil {
println("Reading Dictionary_ID", err)
return err
}
var id uint32
switch size {
@ -187,10 +191,10 @@ func (d *frameDec) reset(br byteBuffer) error {
}
d.FrameContentSize = 0
if fcsSize > 0 {
b := br.readSmall(fcsSize)
if b == nil {
println("Reading Frame content", io.ErrUnexpectedEOF)
return io.ErrUnexpectedEOF
b, err = br.readSmall(fcsSize)
if err != nil {
println("Reading Frame content", err)
return err
}
switch fcsSize {
case 1:
@ -307,10 +311,10 @@ func (d *frameDec) checkCRC() error {
tmp[3] = byte(got >> 24)
// We can overwrite upper tmp now
want := d.rawInput.readSmall(4)
if want == nil {
println("CRC missing?")
return io.ErrUnexpectedEOF
want, err := d.rawInput.readSmall(4)
if err != nil {
println("CRC missing?", err)
return err
}
if !bytes.Equal(tmp[:], want) {

13
vendor/github.com/klauspost/compress/zstd/fse_encoder.go generated vendored
View File

@ -97,7 +97,7 @@ func (s *fseEncoder) prepare() (*fseEncoder, error) {
func (s *fseEncoder) allocCtable() {
tableSize := 1 << s.actualTableLog
// get tableSymbol that is big enough.
if cap(s.ct.tableSymbol) < int(tableSize) {
if cap(s.ct.tableSymbol) < tableSize {
s.ct.tableSymbol = make([]byte, tableSize)
}
s.ct.tableSymbol = s.ct.tableSymbol[:tableSize]
@ -202,13 +202,13 @@ func (s *fseEncoder) buildCTable() error {
case 0:
case -1, 1:
symbolTT[i].deltaNbBits = tl
symbolTT[i].deltaFindState = int16(total - 1)
symbolTT[i].deltaFindState = total - 1
total++
default:
maxBitsOut := uint32(tableLog) - highBit(uint32(v-1))
minStatePlus := uint32(v) << maxBitsOut
symbolTT[i].deltaNbBits = (maxBitsOut << 16) - minStatePlus
symbolTT[i].deltaFindState = int16(total - v)
symbolTT[i].deltaFindState = total - v
total += v
}
}
@ -353,8 +353,8 @@ func (s *fseEncoder) normalizeCount2(length int) error {
distributed uint32
total = uint32(length)
tableLog = s.actualTableLog
lowThreshold = uint32(total >> tableLog)
lowOne = uint32((total * 3) >> (tableLog + 1))
lowThreshold = total >> tableLog
lowOne = (total * 3) >> (tableLog + 1)
)
for i, cnt := range s.count[:s.symbolLen] {
if cnt == 0 {
@ -379,7 +379,7 @@ func (s *fseEncoder) normalizeCount2(length int) error {
if (total / toDistribute) > lowOne {
// risk of rounding to zero
lowOne = uint32((total * 3) / (toDistribute * 2))
lowOne = (total * 3) / (toDistribute * 2)
for i, cnt := range s.count[:s.symbolLen] {
if (s.norm[i] == notYetAssigned) && (cnt <= lowOne) {
s.norm[i] = 1
@ -708,7 +708,6 @@ func (c *cState) init(bw *bitWriter, ct *cTable, first symbolTransform) {
im := int32((nbBitsOut << 16) - first.deltaNbBits)
lu := (im >> nbBitsOut) + int32(first.deltaFindState)
c.state = c.stateTable[lu]
return
}
// encode the output symbol provided and write it to the bitstream.

2
vendor/github.com/klauspost/compress/zstd/fse_predefined.go generated vendored
View File

@ -59,7 +59,7 @@ func fillBase(dst []baseOffset, base uint32, bits ...uint8) {
}
for i, bit := range bits {
if base > math.MaxInt32 {
panic(fmt.Sprintf("invalid decoding table, base overflows int32"))
panic("invalid decoding table, base overflows int32")
}
dst[i] = baseOffset{

15
vendor/github.com/klauspost/compress/zstd/seqdec.go generated vendored
View File

@ -181,11 +181,18 @@ func (s *sequenceDecs) decode(seqs int, br *bitReader, hist []byte) error {
return fmt.Errorf("output (%d) bigger than max block size", size)
}
if size > cap(s.out) {
// Not enough size, will be extremely rarely triggered,
// Not enough size, which can happen under high volume block streaming conditions
// but could be if destination slice is too small for sync operations.
// We add maxBlockSize to the capacity.
s.out = append(s.out, make([]byte, maxBlockSize)...)
s.out = s.out[:len(s.out)-maxBlockSize]
// over-allocating here can create a large amount of GC pressure so we try to keep
// it as contained as possible
used := len(s.out) - startSize
addBytes := 256 + ll + ml + used>>2
// Clamp to max block size.
if used+addBytes > maxBlockSize {
addBytes = maxBlockSize - used
}
s.out = append(s.out, make([]byte, addBytes)...)
s.out = s.out[:len(s.out)-addBytes]
}
if ml > maxMatchLen {
return fmt.Errorf("match len (%d) bigger than max allowed length", ml)

1
vendor/github.com/klauspost/compress/zstd/seqenc.go generated vendored
View File

@ -35,7 +35,6 @@ func (s *seqCoders) setPrev(ll, ml, of *fseEncoder) {
// Ensure we cannot reuse by accident
prevEnc := *prev
prevEnc.symbolLen = 0
return
}
compareSwap(ll, &s.llEnc, &s.llPrev)
compareSwap(ml, &s.mlEnc, &s.mlPrev)

5
vendor/github.com/klauspost/compress/zstd/snappy.go generated vendored
View File

@ -10,8 +10,8 @@ import (
"hash/crc32"
"io"
"github.com/golang/snappy"
"github.com/klauspost/compress/huff0"
"github.com/klauspost/compress/snappy"
)
const (
@ -185,7 +185,6 @@ func (r *SnappyConverter) Convert(in io.Reader, w io.Writer) (int64, error) {
r.block.reset(nil)
r.block.literals, err = snappy.Decode(r.block.literals[:n], r.buf[snappyChecksumSize:chunkLen])
if err != nil {
println("snappy.Decode:", err)
return written, err
}
err = r.block.encodeLits(r.block.literals, false)
@ -417,7 +416,7 @@ var crcTable = crc32.MakeTable(crc32.Castagnoli)
// https://github.com/google/snappy/blob/master/framing_format.txt
func snappyCRC(b []byte) uint32 {
c := crc32.Update(0, crcTable, b)
return uint32(c>>15|c<<17) + 0xa282ead8
return c>>15 | c<<17 + 0xa282ead8
}
// snappyDecodedLen returns the length of the decoded block and the number of bytes

121
vendor/github.com/klauspost/compress/zstd/zip.go generated vendored Normal file
View File

@ -0,0 +1,121 @@
// Copyright 2019+ Klaus Post. All rights reserved.
// License information can be found in the LICENSE file.
package zstd
import (
"errors"
"io"
"sync"
)
// ZipMethodWinZip is the method for Zstandard compressed data inside Zip files for WinZip.
// See https://www.winzip.com/win/en/comp_info.html
const ZipMethodWinZip = 93
// ZipMethodPKWare is the original method number used by PKWARE to indicate Zstandard compression.
// Deprecated: This has been deprecated by PKWARE, use ZipMethodWinZip instead for compression.
// See https://pkware.cachefly.net/webdocs/APPNOTE/APPNOTE-6.3.9.TXT
const ZipMethodPKWare = 20
var zipReaderPool sync.Pool
// newZipReader cannot be used since we would leak goroutines...
func newZipReader(r io.Reader) io.ReadCloser {
dec, ok := zipReaderPool.Get().(*Decoder)
if ok {
dec.Reset(r)
} else {
d, err := NewReader(r, WithDecoderConcurrency(1), WithDecoderLowmem(true))
if err != nil {
panic(err)
}
dec = d
}
return &pooledZipReader{dec: dec}
}
type pooledZipReader struct {
mu sync.Mutex // guards Close and Read
dec *Decoder
}
func (r *pooledZipReader) Read(p []byte) (n int, err error) {
r.mu.Lock()
defer r.mu.Unlock()
if r.dec == nil {
return 0, errors.New("Read after Close")
}
dec, err := r.dec.Read(p)
return dec, err
}
func (r *pooledZipReader) Close() error {
r.mu.Lock()
defer r.mu.Unlock()
var err error
if r.dec != nil {
err = r.dec.Reset(nil)
zipReaderPool.Put(r.dec)
r.dec = nil
}
return err
}
type pooledZipWriter struct {
mu sync.Mutex // guards Close and Read
enc *Encoder
}
func (w *pooledZipWriter) Write(p []byte) (n int, err error) {
w.mu.Lock()
defer w.mu.Unlock()
if w.enc == nil {
return 0, errors.New("Write after Close")
}
return w.enc.Write(p)
}
func (w *pooledZipWriter) Close() error {
w.mu.Lock()
defer w.mu.Unlock()
var err error
if w.enc != nil {
err = w.enc.Close()
zipReaderPool.Put(w.enc)
w.enc = nil
}
return err
}
// ZipCompressor returns a compressor that can be registered with zip libraries.
// The provided encoder options will be used on all encodes.
func ZipCompressor(opts ...EOption) func(w io.Writer) (io.WriteCloser, error) {
var pool sync.Pool
return func(w io.Writer) (io.WriteCloser, error) {
enc, ok := pool.Get().(*Encoder)
if ok {
enc.Reset(w)
} else {
var err error
enc, err = NewWriter(w, opts...)
if err != nil {
return nil, err
}
}
return &pooledZipWriter{enc: enc}, nil
}
}
// ZipDecompressor returns a decompressor that can be registered with zip libraries.
// See ZipCompressor for example.
func ZipDecompressor() func(r io.Reader) io.ReadCloser {
return func(r io.Reader) io.ReadCloser {
d, err := NewReader(r, WithDecoderConcurrency(1), WithDecoderLowmem(true))
if err != nil {
panic(err)
}
return d.IOReadCloser()
}
}

30
vendor/github.com/klauspost/compress/zstd/zstd.go generated vendored
View File

@ -4,6 +4,8 @@
package zstd
import (
"bytes"
"encoding/binary"
"errors"
"log"
"math"
@ -73,6 +75,10 @@ var (
// ErrDecoderClosed will be returned if the Decoder was used after
// Close has been called.
ErrDecoderClosed = errors.New("decoder used after Close")
// ErrDecoderNilInput is returned when a nil Reader was provided
// and an operation other than Reset/DecodeAll/Close was attempted.
ErrDecoderNilInput = errors.New("nil input provided as reader")
)
func println(a ...interface{}) {
@ -121,24 +127,20 @@ func matchLen(a, b []byte) int {
}
func load3232(b []byte, i int32) uint32 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:4]
return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
return binary.LittleEndian.Uint32(b[i:])
}
func load6432(b []byte, i int32) uint64 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:8]
return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
return binary.LittleEndian.Uint64(b[i:])
}
func load64(b []byte, i int) uint64 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:8]
return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
return binary.LittleEndian.Uint64(b[i:])
}
type byter interface {
Bytes() []byte
Len() int
}
var _ byter = &bytes.Buffer{}