williamp/chatservice_concept
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

implement Scylla database

Browse Source
This commit is contained in:
William P
2025-05-17 21:45:18 -04:00
parent 252b49ae6a
commit f8a550883d
199 changed files with 71243 additions and 424 deletions
Download Patch File Download Diff File Expand all files Collapse all files

304
vendor/github.com/klauspost/compress/LICENSE generated vendored Normal file
View File

@@ -0,0 +1,304 @@
Copyright (c) 2012 The Go Authors. All rights reserved.
Copyright (c) 2019 Klaus Post. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
------------------
Files: gzhttp/*
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright 2016-2017 The New York Times Company
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
------------------
Files: s2/cmd/internal/readahead/*
The MIT License (MIT)
Copyright (c) 2015 Klaus Post
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
---------------------
Files: snappy/*
Files: internal/snapref/*
Copyright (c) 2011 The Snappy-Go Authors. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-----------------
Files: s2/cmd/internal/filepathx/*
Copyright 2016 The filepathx Authors
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

13
vendor/github.com/klauspost/compress/internal/race/norace.go generated vendored Normal file
View File

@@ -0,0 +1,13 @@
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build !race
package race
func ReadSlice[T any](s []T) {
}
func WriteSlice[T any](s []T) {
}

26
vendor/github.com/klauspost/compress/internal/race/race.go generated vendored Normal file
View File

@@ -0,0 +1,26 @@
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build race
package race
import (
"runtime"
"unsafe"
)
func ReadSlice[T any](s []T) {
if len(s) == 0 {
return
}
runtime.RaceReadRange(unsafe.Pointer(&s[0]), len(s)*int(unsafe.Sizeof(s[0])))
}
func WriteSlice[T any](s []T) {
if len(s) == 0 {
return
}
runtime.RaceWriteRange(unsafe.Pointer(&s[0]), len(s)*int(unsafe.Sizeof(s[0])))
}

15
vendor/github.com/klauspost/compress/s2/.gitignore generated vendored Normal file
View File

@@ -0,0 +1,15 @@
testdata/bench
# These explicitly listed benchmark data files are for an obsolete version of
# snappy_test.go.
testdata/alice29.txt
testdata/asyoulik.txt
testdata/fireworks.jpeg
testdata/geo.protodata
testdata/html
testdata/html_x_4
testdata/kppkn.gtb
testdata/lcet10.txt
testdata/paper-100k.pdf
testdata/plrabn12.txt
testdata/urls.10K

28
vendor/github.com/klauspost/compress/s2/LICENSE generated vendored Normal file
View File

@@ -0,0 +1,28 @@
Copyright (c) 2011 The Snappy-Go Authors. All rights reserved.
Copyright (c) 2019 Klaus Post. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

1120
vendor/github.com/klauspost/compress/s2/README.md generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff

443
vendor/github.com/klauspost/compress/s2/decode.go generated vendored Normal file
View File

@@ -0,0 +1,443 @@
// Copyright 2011 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"encoding/binary"
"errors"
"fmt"
"strconv"
"github.com/klauspost/compress/internal/race"
)
var (
// ErrCorrupt reports that the input is invalid.
ErrCorrupt = errors.New("s2: corrupt input")
// ErrCRC reports that the input failed CRC validation (streams only)
ErrCRC = errors.New("s2: corrupt input, crc mismatch")
// ErrTooLarge reports that the uncompressed length is too large.
ErrTooLarge = errors.New("s2: decoded block is too large")
// ErrUnsupported reports that the input isn't supported.
ErrUnsupported = errors.New("s2: unsupported input")
)
// DecodedLen returns the length of the decoded block.
func DecodedLen(src []byte) (int, error) {
v, _, err := decodedLen(src)
return v, err
}
// decodedLen returns the length of the decoded block and the number of bytes
// that the length header occupied.
func decodedLen(src []byte) (blockLen, headerLen int, err error) {
v, n := binary.Uvarint(src)
if n <= 0 || v > 0xffffffff {
return 0, 0, ErrCorrupt
}
const wordSize = 32 << (^uint(0) >> 32 & 1)
if wordSize == 32 && v > 0x7fffffff {
return 0, 0, ErrTooLarge
}
return int(v), n, nil
}
const (
decodeErrCodeCorrupt = 1
)
// Decode returns the decoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire decoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
func Decode(dst, src []byte) ([]byte, error) {
dLen, s, err := decodedLen(src)
if err != nil {
return nil, err
}
if dLen <= cap(dst) {
dst = dst[:dLen]
} else {
dst = make([]byte, dLen)
}
race.WriteSlice(dst)
race.ReadSlice(src[s:])
if s2Decode(dst, src[s:]) != 0 {
return nil, ErrCorrupt
}
return dst, nil
}
// s2DecodeDict writes the decoding of src to dst. It assumes that the varint-encoded
// length of the decompressed bytes has already been read, and that len(dst)
// equals that length.
//
// It returns 0 on success or a decodeErrCodeXxx error code on failure.
func s2DecodeDict(dst, src []byte, dict *Dict) int {
if dict == nil {
return s2Decode(dst, src)
}
const debug = false
const debugErrs = debug
if debug {
fmt.Println("Starting decode, dst len:", len(dst))
}
var d, s, length int
offset := len(dict.dict) - dict.repeat
// As long as we can read at least 5 bytes...
for s < len(src)-5 {
// Removing bounds checks is SLOWER, when if doing
// in := src[s:s+5]
// Checked on Go 1.18
switch src[s] & 0x03 {
case tagLiteral:
x := uint32(src[s] >> 2)
switch {
case x < 60:
s++
case x == 60:
s += 2
x = uint32(src[s-1])
case x == 61:
in := src[s : s+3]
x = uint32(in[1]) | uint32(in[2])<<8
s += 3
case x == 62:
in := src[s : s+4]
// Load as 32 bit and shift down.
x = uint32(in[0]) | uint32(in[1])<<8 | uint32(in[2])<<16 | uint32(in[3])<<24
x >>= 8
s += 4
case x == 63:
in := src[s : s+5]
x = uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24
s += 5
}
length = int(x) + 1
if debug {
fmt.Println("literals, length:", length, "d-after:", d+length)
}
if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) {
if debugErrs {
fmt.Println("corrupt literal: length:", length, "d-left:", len(dst)-d, "src-left:", len(src)-s)
}
return decodeErrCodeCorrupt
}
copy(dst[d:], src[s:s+length])
d += length
s += length
continue
case tagCopy1:
s += 2
toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
length = int(src[s-2]) >> 2 & 0x7
if toffset == 0 {
if debug {
fmt.Print("(repeat) ")
}
// keep last offset
switch length {
case 5:
length = int(src[s]) + 4
s += 1
case 6:
in := src[s : s+2]
length = int(uint32(in[0])|(uint32(in[1])<<8)) + (1 << 8)
s += 2
case 7:
in := src[s : s+3]
length = int((uint32(in[2])<<16)|(uint32(in[1])<<8)|uint32(in[0])) + (1 << 16)
s += 3
default: // 0-> 4
}
} else {
offset = toffset
}
length += 4
case tagCopy2:
in := src[s : s+3]
offset = int(uint32(in[1]) | uint32(in[2])<<8)
length = 1 + int(in[0])>>2
s += 3
case tagCopy4:
in := src[s : s+5]
offset = int(uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24)
length = 1 + int(in[0])>>2
s += 5
}
if offset <= 0 || length > len(dst)-d {
if debugErrs {
fmt.Println("match error; offset:", offset, "length:", length, "dst-left:", len(dst)-d)
}
return decodeErrCodeCorrupt
}
// copy from dict
if d < offset {
if d > MaxDictSrcOffset {
if debugErrs {
fmt.Println("dict after", MaxDictSrcOffset, "d:", d, "offset:", offset, "length:", length)
}
return decodeErrCodeCorrupt
}
startOff := len(dict.dict) - offset + d
if startOff < 0 || startOff+length > len(dict.dict) {
if debugErrs {
fmt.Printf("offset (%d) + length (%d) bigger than dict (%d)\n", offset, length, len(dict.dict))
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("dict copy, length:", length, "offset:", offset, "d-after:", d+length, "dict start offset:", startOff)
}
copy(dst[d:d+length], dict.dict[startOff:])
d += length
continue
}
if debug {
fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length)
}
// Copy from an earlier sub-slice of dst to a later sub-slice.
// If no overlap, use the built-in copy:
if offset > length {
copy(dst[d:d+length], dst[d-offset:])
d += length
continue
}
// Unlike the built-in copy function, this byte-by-byte copy always runs
// forwards, even if the slices overlap. Conceptually, this is:
//
// d += forwardCopy(dst[d:d+length], dst[d-offset:])
//
// We align the slices into a and b and show the compiler they are the same size.
// This allows the loop to run without bounds checks.
a := dst[d : d+length]
b := dst[d-offset:]
b = b[:len(a)]
for i := range a {
a[i] = b[i]
}
d += length
}
// Remaining with extra checks...
for s < len(src) {
switch src[s] & 0x03 {
case tagLiteral:
x := uint32(src[s] >> 2)
switch {
case x < 60:
s++
case x == 60:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
x = uint32(src[s-1])
case x == 61:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
x = uint32(src[s-2]) | uint32(src[s-1])<<8
case x == 62:
s += 4
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16
case x == 63:
s += 5
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24
}
length = int(x) + 1
if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) {
if debugErrs {
fmt.Println("corrupt literal: length:", length, "d-left:", len(dst)-d, "src-left:", len(src)-s)
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("literals, length:", length, "d-after:", d+length)
}
copy(dst[d:], src[s:s+length])
d += length
s += length
continue
case tagCopy1:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = int(src[s-2]) >> 2 & 0x7
toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
if toffset == 0 {
if debug {
fmt.Print("(repeat) ")
}
// keep last offset
switch length {
case 5:
s += 1
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-1])) + 4
case 6:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-2])|(uint32(src[s-1])<<8)) + (1 << 8)
case 7:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-3])|(uint32(src[s-2])<<8)|(uint32(src[s-1])<<16)) + (1 << 16)
default: // 0-> 4
}
} else {
offset = toffset
}
length += 4
case tagCopy2:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = 1 + int(src[s-3])>>2
offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8)
case tagCopy4:
s += 5
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
if debugErrs {
fmt.Println("src went oob")
}
return decodeErrCodeCorrupt
}
length = 1 + int(src[s-5])>>2
offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24)
}
if offset <= 0 || length > len(dst)-d {
if debugErrs {
fmt.Println("match error; offset:", offset, "length:", length, "dst-left:", len(dst)-d)
}
return decodeErrCodeCorrupt
}
// copy from dict
if d < offset {
if d > MaxDictSrcOffset {
if debugErrs {
fmt.Println("dict after", MaxDictSrcOffset, "d:", d, "offset:", offset, "length:", length)
}
return decodeErrCodeCorrupt
}
rOff := len(dict.dict) - (offset - d)
if debug {
fmt.Println("starting dict entry from dict offset", len(dict.dict)-rOff)
}
if rOff+length > len(dict.dict) {
if debugErrs {
fmt.Println("err: END offset", rOff+length, "bigger than dict", len(dict.dict), "dict offset:", rOff, "length:", length)
}
return decodeErrCodeCorrupt
}
if rOff < 0 {
if debugErrs {
fmt.Println("err: START offset", rOff, "less than 0", len(dict.dict), "dict offset:", rOff, "length:", length)
}
return decodeErrCodeCorrupt
}
copy(dst[d:d+length], dict.dict[rOff:])
d += length
continue
}
if debug {
fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length)
}
// Copy from an earlier sub-slice of dst to a later sub-slice.
// If no overlap, use the built-in copy:
if offset > length {
copy(dst[d:d+length], dst[d-offset:])
d += length
continue
}
// Unlike the built-in copy function, this byte-by-byte copy always runs
// forwards, even if the slices overlap. Conceptually, this is:
//
// d += forwardCopy(dst[d:d+length], dst[d-offset:])
//
// We align the slices into a and b and show the compiler they are the same size.
// This allows the loop to run without bounds checks.
a := dst[d : d+length]
b := dst[d-offset:]
b = b[:len(a)]
for i := range a {
a[i] = b[i]
}
d += length
}
if d != len(dst) {
if debugErrs {
fmt.Println("wanted length", len(dst), "got", d)
}
return decodeErrCodeCorrupt
}
return 0
}

568
vendor/github.com/klauspost/compress/s2/decode_amd64.s generated vendored Normal file
View File

@@ -0,0 +1,568 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build !appengine
// +build gc
// +build !noasm
#include "textflag.h"
#define R_TMP0 AX
#define R_TMP1 BX
#define R_LEN CX
#define R_OFF DX
#define R_SRC SI
#define R_DST DI
#define R_DBASE R8
#define R_DLEN R9
#define R_DEND R10
#define R_SBASE R11
#define R_SLEN R12
#define R_SEND R13
#define R_TMP2 R14
#define R_TMP3 R15
// The asm code generally follows the pure Go code in decode_other.go, except
// where marked with a "!!!".
// func decode(dst, src []byte) int
//
// All local variables fit into registers. The non-zero stack size is only to
// spill registers and push args when issuing a CALL. The register allocation:
// - R_TMP0 scratch
// - R_TMP1 scratch
// - R_LEN length or x (shared)
// - R_OFF offset
// - R_SRC &src[s]
// - R_DST &dst[d]
// + R_DBASE dst_base
// + R_DLEN dst_len
// + R_DEND dst_base + dst_len
// + R_SBASE src_base
// + R_SLEN src_len
// + R_SEND src_base + src_len
// - R_TMP2 used by doCopy
// - R_TMP3 used by doCopy
//
// The registers R_DBASE-R_SEND (marked with a "+") are set at the start of the
// function, and after a CALL returns, and are not otherwise modified.
//
// The d variable is implicitly R_DST - R_DBASE, and len(dst)-d is R_DEND - R_DST.
// The s variable is implicitly R_SRC - R_SBASE, and len(src)-s is R_SEND - R_SRC.
TEXT ·s2Decode(SB), NOSPLIT, $48-56
// Initialize R_SRC, R_DST and R_DBASE-R_SEND.
MOVQ dst_base+0(FP), R_DBASE
MOVQ dst_len+8(FP), R_DLEN
MOVQ R_DBASE, R_DST
MOVQ R_DBASE, R_DEND
ADDQ R_DLEN, R_DEND
MOVQ src_base+24(FP), R_SBASE
MOVQ src_len+32(FP), R_SLEN
MOVQ R_SBASE, R_SRC
MOVQ R_SBASE, R_SEND
ADDQ R_SLEN, R_SEND
XORQ R_OFF, R_OFF
loop:
// for s < len(src)
CMPQ R_SRC, R_SEND
JEQ end
// R_LEN = uint32(src[s])
//
// switch src[s] & 0x03
MOVBLZX (R_SRC), R_LEN
MOVL R_LEN, R_TMP1
ANDL $3, R_TMP1
CMPL R_TMP1, $1
JAE tagCopy
// ----------------------------------------
// The code below handles literal tags.
// case tagLiteral:
// x := uint32(src[s] >> 2)
// switch
SHRL $2, R_LEN
CMPL R_LEN, $60
JAE tagLit60Plus
// case x < 60:
// s++
INCQ R_SRC
doLit:
// This is the end of the inner "switch", when we have a literal tag.
//
// We assume that R_LEN == x and x fits in a uint32, where x is the variable
// used in the pure Go decode_other.go code.
// length = int(x) + 1
//
// Unlike the pure Go code, we don't need to check if length <= 0 because
// R_LEN can hold 64 bits, so the increment cannot overflow.
INCQ R_LEN
// Prepare to check if copying length bytes will run past the end of dst or
// src.
//
// R_TMP0 = len(dst) - d
// R_TMP1 = len(src) - s
MOVQ R_DEND, R_TMP0
SUBQ R_DST, R_TMP0
MOVQ R_SEND, R_TMP1
SUBQ R_SRC, R_TMP1
// !!! Try a faster technique for short (16 or fewer bytes) copies.
//
// if length > 16 || len(dst)-d < 16 || len(src)-s < 16 {
// goto callMemmove // Fall back on calling runtime·memmove.
// }
//
// The C++ snappy code calls this TryFastAppend. It also checks len(src)-s
// against 21 instead of 16, because it cannot assume that all of its input
// is contiguous in memory and so it needs to leave enough source bytes to
// read the next tag without refilling buffers, but Go's Decode assumes
// contiguousness (the src argument is a []byte).
CMPQ R_LEN, $16
JGT callMemmove
CMPQ R_TMP0, $16
JLT callMemmove
CMPQ R_TMP1, $16
JLT callMemmove
// !!! Implement the copy from src to dst as a 16-byte load and store.
// (Decode's documentation says that dst and src must not overlap.)
//
// This always copies 16 bytes, instead of only length bytes, but that's
// OK. If the input is a valid Snappy encoding then subsequent iterations
// will fix up the overrun. Otherwise, Decode returns a nil []byte (and a
// non-nil error), so the overrun will be ignored.
//
// Note that on amd64, it is legal and cheap to issue unaligned 8-byte or
// 16-byte loads and stores. This technique probably wouldn't be as
// effective on architectures that are fussier about alignment.
MOVOU 0(R_SRC), X0
MOVOU X0, 0(R_DST)
// d += length
// s += length
ADDQ R_LEN, R_DST
ADDQ R_LEN, R_SRC
JMP loop
callMemmove:
// if length > len(dst)-d || length > len(src)-s { etc }
CMPQ R_LEN, R_TMP0
JGT errCorrupt
CMPQ R_LEN, R_TMP1
JGT errCorrupt
// copy(dst[d:], src[s:s+length])
//
// This means calling runtime·memmove(&dst[d], &src[s], length), so we push
// R_DST, R_SRC and R_LEN as arguments. Coincidentally, we also need to spill those
// three registers to the stack, to save local variables across the CALL.
MOVQ R_DST, 0(SP)
MOVQ R_SRC, 8(SP)
MOVQ R_LEN, 16(SP)
MOVQ R_DST, 24(SP)
MOVQ R_SRC, 32(SP)
MOVQ R_LEN, 40(SP)
MOVQ R_OFF, 48(SP)
CALL runtime·memmove(SB)
// Restore local variables: unspill registers from the stack and
// re-calculate R_DBASE-R_SEND.
MOVQ 24(SP), R_DST
MOVQ 32(SP), R_SRC
MOVQ 40(SP), R_LEN
MOVQ 48(SP), R_OFF
MOVQ dst_base+0(FP), R_DBASE
MOVQ dst_len+8(FP), R_DLEN
MOVQ R_DBASE, R_DEND
ADDQ R_DLEN, R_DEND
MOVQ src_base+24(FP), R_SBASE
MOVQ src_len+32(FP), R_SLEN
MOVQ R_SBASE, R_SEND
ADDQ R_SLEN, R_SEND
// d += length
// s += length
ADDQ R_LEN, R_DST
ADDQ R_LEN, R_SRC
JMP loop
tagLit60Plus:
// !!! This fragment does the
//
// s += x - 58; if uint(s) > uint(len(src)) { etc }
//
// checks. In the asm version, we code it once instead of once per switch case.
ADDQ R_LEN, R_SRC
SUBQ $58, R_SRC
CMPQ R_SRC, R_SEND
JA errCorrupt
// case x == 60:
CMPL R_LEN, $61
JEQ tagLit61
JA tagLit62Plus
// x = uint32(src[s-1])
MOVBLZX -1(R_SRC), R_LEN
JMP doLit
tagLit61:
// case x == 61:
// x = uint32(src[s-2]) | uint32(src[s-1])<<8
MOVWLZX -2(R_SRC), R_LEN
JMP doLit
tagLit62Plus:
CMPL R_LEN, $62
JA tagLit63
// case x == 62:
// x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16
// We read one byte, safe to read one back, since we are just reading tag.
// x = binary.LittleEndian.Uint32(src[s-1:]) >> 8
MOVL -4(R_SRC), R_LEN
SHRL $8, R_LEN
JMP doLit
tagLit63:
// case x == 63:
// x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24
MOVL -4(R_SRC), R_LEN
JMP doLit
// The code above handles literal tags.
// ----------------------------------------
// The code below handles copy tags.
tagCopy4:
// case tagCopy4:
// s += 5
ADDQ $5, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// length = 1 + int(src[s-5])>>2
SHRQ $2, R_LEN
INCQ R_LEN
// offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24)
MOVLQZX -4(R_SRC), R_OFF
JMP doCopy
tagCopy2:
// case tagCopy2:
// s += 3
ADDQ $3, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// length = 1 + int(src[s-3])>>2
SHRQ $2, R_LEN
INCQ R_LEN
// offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8)
MOVWQZX -2(R_SRC), R_OFF
JMP doCopy
tagCopy:
// We have a copy tag. We assume that:
// - R_TMP1 == src[s] & 0x03
// - R_LEN == src[s]
CMPQ R_TMP1, $2
JEQ tagCopy2
JA tagCopy4
// case tagCopy1:
// s += 2
ADDQ $2, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
// length = 4 + int(src[s-2])>>2&0x7
MOVBQZX -1(R_SRC), R_TMP1
MOVQ R_LEN, R_TMP0
SHRQ $2, R_LEN
ANDQ $0xe0, R_TMP0
ANDQ $7, R_LEN
SHLQ $3, R_TMP0
ADDQ $4, R_LEN
ORQ R_TMP1, R_TMP0
// check if repeat code, ZF set by ORQ.
JZ repeatCode
// This is a regular copy, transfer our temporary value to R_OFF (length)
MOVQ R_TMP0, R_OFF
JMP doCopy
// This is a repeat code.
repeatCode:
// If length < 9, reuse last offset, with the length already calculated.
CMPQ R_LEN, $9
JL doCopyRepeat
// Read additional bytes for length.
JE repeatLen1
// Rare, so the extra branch shouldn't hurt too much.
CMPQ R_LEN, $10
JE repeatLen2
JMP repeatLen3
// Read repeat lengths.
repeatLen1:
// s ++
ADDQ $1, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// length = src[s-1] + 8
MOVBQZX -1(R_SRC), R_LEN
ADDL $8, R_LEN
JMP doCopyRepeat
repeatLen2:
// s +=2
ADDQ $2, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// length = uint32(src[s-2]) | (uint32(src[s-1])<<8) + (1 << 8)
MOVWQZX -2(R_SRC), R_LEN
ADDL $260, R_LEN
JMP doCopyRepeat
repeatLen3:
// s +=3
ADDQ $3, R_SRC
// if uint(s) > uint(len(src)) { etc }
CMPQ R_SRC, R_SEND
JA errCorrupt
// length = uint32(src[s-3]) | (uint32(src[s-2])<<8) | (uint32(src[s-1])<<16) + (1 << 16)
// Read one byte further back (just part of the tag, shifted out)
MOVL -4(R_SRC), R_LEN
SHRL $8, R_LEN
ADDL $65540, R_LEN
JMP doCopyRepeat
doCopy:
// This is the end of the outer "switch", when we have a copy tag.
//
// We assume that:
// - R_LEN == length && R_LEN > 0
// - R_OFF == offset
// if d < offset { etc }
MOVQ R_DST, R_TMP1
SUBQ R_DBASE, R_TMP1
CMPQ R_TMP1, R_OFF
JLT errCorrupt
// Repeat values can skip the test above, since any offset > 0 will be in dst.
doCopyRepeat:
// if offset <= 0 { etc }
CMPQ R_OFF, $0
JLE errCorrupt
// if length > len(dst)-d { etc }
MOVQ R_DEND, R_TMP1
SUBQ R_DST, R_TMP1
CMPQ R_LEN, R_TMP1
JGT errCorrupt
// forwardCopy(dst[d:d+length], dst[d-offset:]); d += length
//
// Set:
// - R_TMP2 = len(dst)-d
// - R_TMP3 = &dst[d-offset]
MOVQ R_DEND, R_TMP2
SUBQ R_DST, R_TMP2
MOVQ R_DST, R_TMP3
SUBQ R_OFF, R_TMP3
// !!! Try a faster technique for short (16 or fewer bytes) forward copies.
//
// First, try using two 8-byte load/stores, similar to the doLit technique
// above. Even if dst[d:d+length] and dst[d-offset:] can overlap, this is
// still OK if offset >= 8. Note that this has to be two 8-byte load/stores
// and not one 16-byte load/store, and the first store has to be before the
// second load, due to the overlap if offset is in the range [8, 16).
//
// if length > 16 || offset < 8 || len(dst)-d < 16 {
// goto slowForwardCopy
// }
// copy 16 bytes
// d += length
CMPQ R_LEN, $16
JGT slowForwardCopy
CMPQ R_OFF, $8
JLT slowForwardCopy
CMPQ R_TMP2, $16
JLT slowForwardCopy
MOVQ 0(R_TMP3), R_TMP0
MOVQ R_TMP0, 0(R_DST)
MOVQ 8(R_TMP3), R_TMP1
MOVQ R_TMP1, 8(R_DST)
ADDQ R_LEN, R_DST
JMP loop
slowForwardCopy:
// !!! If the forward copy is longer than 16 bytes, or if offset < 8, we
// can still try 8-byte load stores, provided we can overrun up to 10 extra
// bytes. As above, the overrun will be fixed up by subsequent iterations
// of the outermost loop.
//
// The C++ snappy code calls this technique IncrementalCopyFastPath. Its
// commentary says:
//
// ----
//
// The main part of this loop is a simple copy of eight bytes at a time
// until we've copied (at least) the requested amount of bytes. However,
// if d and d-offset are less than eight bytes apart (indicating a
// repeating pattern of length < 8), we first need to expand the pattern in
// order to get the correct results. For instance, if the buffer looks like
// this, with the eight-byte <d-offset> and <d> patterns marked as
// intervals:
//
// abxxxxxxxxxxxx
// [------] d-offset
// [------] d
//
// a single eight-byte copy from <d-offset> to <d> will repeat the pattern
// once, after which we can move <d> two bytes without moving <d-offset>:
//
// ababxxxxxxxxxx
// [------] d-offset
// [------] d
//
// and repeat the exercise until the two no longer overlap.
//
// This allows us to do very well in the special case of one single byte
// repeated many times, without taking a big hit for more general cases.
//
// The worst case of extra writing past the end of the match occurs when
// offset == 1 and length == 1; the last copy will read from byte positions
// [0..7] and write to [4..11], whereas it was only supposed to write to
// position 1. Thus, ten excess bytes.
//
// ----
//
// That "10 byte overrun" worst case is confirmed by Go's
// TestSlowForwardCopyOverrun, which also tests the fixUpSlowForwardCopy
// and finishSlowForwardCopy algorithm.
//
// if length > len(dst)-d-10 {
// goto verySlowForwardCopy
// }
SUBQ $10, R_TMP2
CMPQ R_LEN, R_TMP2
JGT verySlowForwardCopy
// We want to keep the offset, so we use R_TMP2 from here.
MOVQ R_OFF, R_TMP2
makeOffsetAtLeast8:
// !!! As above, expand the pattern so that offset >= 8 and we can use
// 8-byte load/stores.
//
// for offset < 8 {
// copy 8 bytes from dst[d-offset:] to dst[d:]
// length -= offset
// d += offset
// offset += offset
// // The two previous lines together means that d-offset, and therefore
// // R_TMP3, is unchanged.
// }
CMPQ R_TMP2, $8
JGE fixUpSlowForwardCopy
MOVQ (R_TMP3), R_TMP1
MOVQ R_TMP1, (R_DST)
SUBQ R_TMP2, R_LEN
ADDQ R_TMP2, R_DST
ADDQ R_TMP2, R_TMP2
JMP makeOffsetAtLeast8
fixUpSlowForwardCopy:
// !!! Add length (which might be negative now) to d (implied by R_DST being
// &dst[d]) so that d ends up at the right place when we jump back to the
// top of the loop. Before we do that, though, we save R_DST to R_TMP0 so that, if
// length is positive, copying the remaining length bytes will write to the
// right place.
MOVQ R_DST, R_TMP0
ADDQ R_LEN, R_DST
finishSlowForwardCopy:
// !!! Repeat 8-byte load/stores until length <= 0. Ending with a negative
// length means that we overrun, but as above, that will be fixed up by
// subsequent iterations of the outermost loop.
CMPQ R_LEN, $0
JLE loop
MOVQ (R_TMP3), R_TMP1
MOVQ R_TMP1, (R_TMP0)
ADDQ $8, R_TMP3
ADDQ $8, R_TMP0
SUBQ $8, R_LEN
JMP finishSlowForwardCopy
verySlowForwardCopy:
// verySlowForwardCopy is a simple implementation of forward copy. In C
// parlance, this is a do/while loop instead of a while loop, since we know
// that length > 0. In Go syntax:
//
// for {
// dst[d] = dst[d - offset]
// d++
// length--
// if length == 0 {
// break
// }
// }
MOVB (R_TMP3), R_TMP1
MOVB R_TMP1, (R_DST)
INCQ R_TMP3
INCQ R_DST
DECQ R_LEN
JNZ verySlowForwardCopy
JMP loop
// The code above handles copy tags.
// ----------------------------------------
end:
// This is the end of the "for s < len(src)".
//
// if d != len(dst) { etc }
CMPQ R_DST, R_DEND
JNE errCorrupt
// return 0
MOVQ $0, ret+48(FP)
RET
errCorrupt:
// return decodeErrCodeCorrupt
MOVQ $1, ret+48(FP)
RET

574
vendor/github.com/klauspost/compress/s2/decode_arm64.s generated vendored Normal file
View File

@@ -0,0 +1,574 @@
// Copyright 2020 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build !appengine
// +build gc
// +build !noasm
#include "textflag.h"
#define R_TMP0 R2
#define R_TMP1 R3
#define R_LEN R4
#define R_OFF R5
#define R_SRC R6
#define R_DST R7
#define R_DBASE R8
#define R_DLEN R9
#define R_DEND R10
#define R_SBASE R11
#define R_SLEN R12
#define R_SEND R13
#define R_TMP2 R14
#define R_TMP3 R15
// TEST_SRC will check if R_SRC is <= SRC_END
#define TEST_SRC() \
CMP R_SEND, R_SRC \
BGT errCorrupt
// MOVD R_SRC, R_TMP1
// SUB R_SBASE, R_TMP1, R_TMP1
// CMP R_SLEN, R_TMP1
// BGT errCorrupt
// The asm code generally follows the pure Go code in decode_other.go, except
// where marked with a "!!!".
// func decode(dst, src []byte) int
//
// All local variables fit into registers. The non-zero stack size is only to
// spill registers and push args when issuing a CALL. The register allocation:
// - R_TMP0 scratch
// - R_TMP1 scratch
// - R_LEN length or x
// - R_OFF offset
// - R_SRC &src[s]
// - R_DST &dst[d]
// + R_DBASE dst_base
// + R_DLEN dst_len
// + R_DEND dst_base + dst_len
// + R_SBASE src_base
// + R_SLEN src_len
// + R_SEND src_base + src_len
// - R_TMP2 used by doCopy
// - R_TMP3 used by doCopy
//
// The registers R_DBASE-R_SEND (marked with a "+") are set at the start of the
// function, and after a CALL returns, and are not otherwise modified.
//
// The d variable is implicitly R_DST - R_DBASE, and len(dst)-d is R_DEND - R_DST.
// The s variable is implicitly R_SRC - R_SBASE, and len(src)-s is R_SEND - R_SRC.
TEXT ·s2Decode(SB), NOSPLIT, $56-56
// Initialize R_SRC, R_DST and R_DBASE-R_SEND.
MOVD dst_base+0(FP), R_DBASE
MOVD dst_len+8(FP), R_DLEN
MOVD R_DBASE, R_DST
MOVD R_DBASE, R_DEND
ADD R_DLEN, R_DEND, R_DEND
MOVD src_base+24(FP), R_SBASE
MOVD src_len+32(FP), R_SLEN
MOVD R_SBASE, R_SRC
MOVD R_SBASE, R_SEND
ADD R_SLEN, R_SEND, R_SEND
MOVD $0, R_OFF
loop:
// for s < len(src)
CMP R_SEND, R_SRC
BEQ end
// R_LEN = uint32(src[s])
//
// switch src[s] & 0x03
MOVBU (R_SRC), R_LEN
MOVW R_LEN, R_TMP1
ANDW $3, R_TMP1
MOVW $1, R1
CMPW R1, R_TMP1
BGE tagCopy
// ----------------------------------------
// The code below handles literal tags.
// case tagLiteral:
// x := uint32(src[s] >> 2)
// switch
MOVW $60, R1
LSRW $2, R_LEN, R_LEN
CMPW R_LEN, R1
BLS tagLit60Plus
// case x < 60:
// s++
ADD $1, R_SRC, R_SRC
doLit:
// This is the end of the inner "switch", when we have a literal tag.
//
// We assume that R_LEN == x and x fits in a uint32, where x is the variable
// used in the pure Go decode_other.go code.
// length = int(x) + 1
//
// Unlike the pure Go code, we don't need to check if length <= 0 because
// R_LEN can hold 64 bits, so the increment cannot overflow.
ADD $1, R_LEN, R_LEN
// Prepare to check if copying length bytes will run past the end of dst or
// src.
//
// R_TMP0 = len(dst) - d
// R_TMP1 = len(src) - s
MOVD R_DEND, R_TMP0
SUB R_DST, R_TMP0, R_TMP0
MOVD R_SEND, R_TMP1
SUB R_SRC, R_TMP1, R_TMP1
// !!! Try a faster technique for short (16 or fewer bytes) copies.
//
// if length > 16 || len(dst)-d < 16 || len(src)-s < 16 {
// goto callMemmove // Fall back on calling runtime·memmove.
// }
//
// The C++ snappy code calls this TryFastAppend. It also checks len(src)-s
// against 21 instead of 16, because it cannot assume that all of its input
// is contiguous in memory and so it needs to leave enough source bytes to
// read the next tag without refilling buffers, but Go's Decode assumes
// contiguousness (the src argument is a []byte).
CMP $16, R_LEN
BGT callMemmove
CMP $16, R_TMP0
BLT callMemmove
CMP $16, R_TMP1
BLT callMemmove
// !!! Implement the copy from src to dst as a 16-byte load and store.
// (Decode's documentation says that dst and src must not overlap.)
//
// This always copies 16 bytes, instead of only length bytes, but that's
// OK. If the input is a valid Snappy encoding then subsequent iterations
// will fix up the overrun. Otherwise, Decode returns a nil []byte (and a
// non-nil error), so the overrun will be ignored.
//
// Note that on arm64, it is legal and cheap to issue unaligned 8-byte or
// 16-byte loads and stores. This technique probably wouldn't be as
// effective on architectures that are fussier about alignment.
LDP 0(R_SRC), (R_TMP2, R_TMP3)
STP (R_TMP2, R_TMP3), 0(R_DST)
// d += length
// s += length
ADD R_LEN, R_DST, R_DST
ADD R_LEN, R_SRC, R_SRC
B loop
callMemmove:
// if length > len(dst)-d || length > len(src)-s { etc }
CMP R_TMP0, R_LEN
BGT errCorrupt
CMP R_TMP1, R_LEN
BGT errCorrupt
// copy(dst[d:], src[s:s+length])
//
// This means calling runtime·memmove(&dst[d], &src[s], length), so we push
// R_DST, R_SRC and R_LEN as arguments. Coincidentally, we also need to spill those
// three registers to the stack, to save local variables across the CALL.
MOVD R_DST, 8(RSP)
MOVD R_SRC, 16(RSP)
MOVD R_LEN, 24(RSP)
MOVD R_DST, 32(RSP)
MOVD R_SRC, 40(RSP)
MOVD R_LEN, 48(RSP)
MOVD R_OFF, 56(RSP)
CALL runtime·memmove(SB)
// Restore local variables: unspill registers from the stack and
// re-calculate R_DBASE-R_SEND.
MOVD 32(RSP), R_DST
MOVD 40(RSP), R_SRC
MOVD 48(RSP), R_LEN
MOVD 56(RSP), R_OFF
MOVD dst_base+0(FP), R_DBASE
MOVD dst_len+8(FP), R_DLEN
MOVD R_DBASE, R_DEND
ADD R_DLEN, R_DEND, R_DEND
MOVD src_base+24(FP), R_SBASE
MOVD src_len+32(FP), R_SLEN
MOVD R_SBASE, R_SEND
ADD R_SLEN, R_SEND, R_SEND
// d += length
// s += length
ADD R_LEN, R_DST, R_DST
ADD R_LEN, R_SRC, R_SRC
B loop
tagLit60Plus:
// !!! This fragment does the
//
// s += x - 58; if uint(s) > uint(len(src)) { etc }
//
// checks. In the asm version, we code it once instead of once per switch case.
ADD R_LEN, R_SRC, R_SRC
SUB $58, R_SRC, R_SRC
TEST_SRC()
// case x == 60:
MOVW $61, R1
CMPW R1, R_LEN
BEQ tagLit61
BGT tagLit62Plus
// x = uint32(src[s-1])
MOVBU -1(R_SRC), R_LEN
B doLit
tagLit61:
// case x == 61:
// x = uint32(src[s-2]) | uint32(src[s-1])<<8
MOVHU -2(R_SRC), R_LEN
B doLit
tagLit62Plus:
CMPW $62, R_LEN
BHI tagLit63
// case x == 62:
// x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16
MOVHU -3(R_SRC), R_LEN
MOVBU -1(R_SRC), R_TMP1
ORR R_TMP1<<16, R_LEN
B doLit
tagLit63:
// case x == 63:
// x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24
MOVWU -4(R_SRC), R_LEN
B doLit
// The code above handles literal tags.
// ----------------------------------------
// The code below handles copy tags.
tagCopy4:
// case tagCopy4:
// s += 5
ADD $5, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
MOVD R_SRC, R_TMP1
SUB R_SBASE, R_TMP1, R_TMP1
CMP R_SLEN, R_TMP1
BGT errCorrupt
// length = 1 + int(src[s-5])>>2
MOVD $1, R1
ADD R_LEN>>2, R1, R_LEN
// offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24)
MOVWU -4(R_SRC), R_OFF
B doCopy
tagCopy2:
// case tagCopy2:
// s += 3
ADD $3, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
TEST_SRC()
// length = 1 + int(src[s-3])>>2
MOVD $1, R1
ADD R_LEN>>2, R1, R_LEN
// offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8)
MOVHU -2(R_SRC), R_OFF
B doCopy
tagCopy:
// We have a copy tag. We assume that:
// - R_TMP1 == src[s] & 0x03
// - R_LEN == src[s]
CMP $2, R_TMP1
BEQ tagCopy2
BGT tagCopy4
// case tagCopy1:
// s += 2
ADD $2, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
TEST_SRC()
// offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
// Calculate offset in R_TMP0 in case it is a repeat.
MOVD R_LEN, R_TMP0
AND $0xe0, R_TMP0
MOVBU -1(R_SRC), R_TMP1
ORR R_TMP0<<3, R_TMP1, R_TMP0
// length = 4 + int(src[s-2])>>2&0x7
MOVD $7, R1
AND R_LEN>>2, R1, R_LEN
ADD $4, R_LEN, R_LEN
// check if repeat code with offset 0.
CMP $0, R_TMP0
BEQ repeatCode
// This is a regular copy, transfer our temporary value to R_OFF (offset)
MOVD R_TMP0, R_OFF
B doCopy
// This is a repeat code.
repeatCode:
// If length < 9, reuse last offset, with the length already calculated.
CMP $9, R_LEN
BLT doCopyRepeat
BEQ repeatLen1
CMP $10, R_LEN
BEQ repeatLen2
repeatLen3:
// s +=3
ADD $3, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
TEST_SRC()
// length = uint32(src[s-3]) | (uint32(src[s-2])<<8) | (uint32(src[s-1])<<16) + 65540
MOVBU -1(R_SRC), R_TMP0
MOVHU -3(R_SRC), R_LEN
ORR R_TMP0<<16, R_LEN, R_LEN
ADD $65540, R_LEN, R_LEN
B doCopyRepeat
repeatLen2:
// s +=2
ADD $2, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
TEST_SRC()
// length = uint32(src[s-2]) | (uint32(src[s-1])<<8) + 260
MOVHU -2(R_SRC), R_LEN
ADD $260, R_LEN, R_LEN
B doCopyRepeat
repeatLen1:
// s +=1
ADD $1, R_SRC, R_SRC
// if uint(s) > uint(len(src)) { etc }
TEST_SRC()
// length = src[s-1] + 8
MOVBU -1(R_SRC), R_LEN
ADD $8, R_LEN, R_LEN
B doCopyRepeat
doCopy:
// This is the end of the outer "switch", when we have a copy tag.
//
// We assume that:
// - R_LEN == length && R_LEN > 0
// - R_OFF == offset
// if d < offset { etc }
MOVD R_DST, R_TMP1
SUB R_DBASE, R_TMP1, R_TMP1
CMP R_OFF, R_TMP1
BLT errCorrupt
// Repeat values can skip the test above, since any offset > 0 will be in dst.
doCopyRepeat:
// if offset <= 0 { etc }
CMP $0, R_OFF
BLE errCorrupt
// if length > len(dst)-d { etc }
MOVD R_DEND, R_TMP1
SUB R_DST, R_TMP1, R_TMP1
CMP R_TMP1, R_LEN
BGT errCorrupt
// forwardCopy(dst[d:d+length], dst[d-offset:]); d += length
//
// Set:
// - R_TMP2 = len(dst)-d
// - R_TMP3 = &dst[d-offset]
MOVD R_DEND, R_TMP2
SUB R_DST, R_TMP2, R_TMP2
MOVD R_DST, R_TMP3
SUB R_OFF, R_TMP3, R_TMP3
// !!! Try a faster technique for short (16 or fewer bytes) forward copies.
//
// First, try using two 8-byte load/stores, similar to the doLit technique
// above. Even if dst[d:d+length] and dst[d-offset:] can overlap, this is
// still OK if offset >= 8. Note that this has to be two 8-byte load/stores
// and not one 16-byte load/store, and the first store has to be before the
// second load, due to the overlap if offset is in the range [8, 16).
//
// if length > 16 || offset < 8 || len(dst)-d < 16 {
// goto slowForwardCopy
// }
// copy 16 bytes
// d += length
CMP $16, R_LEN
BGT slowForwardCopy
CMP $8, R_OFF
BLT slowForwardCopy
CMP $16, R_TMP2
BLT slowForwardCopy
MOVD 0(R_TMP3), R_TMP0
MOVD R_TMP0, 0(R_DST)
MOVD 8(R_TMP3), R_TMP1
MOVD R_TMP1, 8(R_DST)
ADD R_LEN, R_DST, R_DST
B loop
slowForwardCopy:
// !!! If the forward copy is longer than 16 bytes, or if offset < 8, we
// can still try 8-byte load stores, provided we can overrun up to 10 extra
// bytes. As above, the overrun will be fixed up by subsequent iterations
// of the outermost loop.
//
// The C++ snappy code calls this technique IncrementalCopyFastPath. Its
// commentary says:
//
// ----
//
// The main part of this loop is a simple copy of eight bytes at a time
// until we've copied (at least) the requested amount of bytes. However,
// if d and d-offset are less than eight bytes apart (indicating a
// repeating pattern of length < 8), we first need to expand the pattern in
// order to get the correct results. For instance, if the buffer looks like
// this, with the eight-byte <d-offset> and <d> patterns marked as
// intervals:
//
// abxxxxxxxxxxxx
// [------] d-offset
// [------] d
//
// a single eight-byte copy from <d-offset> to <d> will repeat the pattern
// once, after which we can move <d> two bytes without moving <d-offset>:
//
// ababxxxxxxxxxx
// [------] d-offset
// [------] d
//
// and repeat the exercise until the two no longer overlap.
//
// This allows us to do very well in the special case of one single byte
// repeated many times, without taking a big hit for more general cases.
//
// The worst case of extra writing past the end of the match occurs when
// offset == 1 and length == 1; the last copy will read from byte positions
// [0..7] and write to [4..11], whereas it was only supposed to write to
// position 1. Thus, ten excess bytes.
//
// ----
//
// That "10 byte overrun" worst case is confirmed by Go's
// TestSlowForwardCopyOverrun, which also tests the fixUpSlowForwardCopy
// and finishSlowForwardCopy algorithm.
//
// if length > len(dst)-d-10 {
// goto verySlowForwardCopy
// }
SUB $10, R_TMP2, R_TMP2
CMP R_TMP2, R_LEN
BGT verySlowForwardCopy
// We want to keep the offset, so we use R_TMP2 from here.
MOVD R_OFF, R_TMP2
makeOffsetAtLeast8:
// !!! As above, expand the pattern so that offset >= 8 and we can use
// 8-byte load/stores.
//
// for offset < 8 {
// copy 8 bytes from dst[d-offset:] to dst[d:]
// length -= offset
// d += offset
// offset += offset
// // The two previous lines together means that d-offset, and therefore
// // R_TMP3, is unchanged.
// }
CMP $8, R_TMP2
BGE fixUpSlowForwardCopy
MOVD (R_TMP3), R_TMP1
MOVD R_TMP1, (R_DST)
SUB R_TMP2, R_LEN, R_LEN
ADD R_TMP2, R_DST, R_DST
ADD R_TMP2, R_TMP2, R_TMP2
B makeOffsetAtLeast8
fixUpSlowForwardCopy:
// !!! Add length (which might be negative now) to d (implied by R_DST being
// &dst[d]) so that d ends up at the right place when we jump back to the
// top of the loop. Before we do that, though, we save R_DST to R_TMP0 so that, if
// length is positive, copying the remaining length bytes will write to the
// right place.
MOVD R_DST, R_TMP0
ADD R_LEN, R_DST, R_DST
finishSlowForwardCopy:
// !!! Repeat 8-byte load/stores until length <= 0. Ending with a negative
// length means that we overrun, but as above, that will be fixed up by
// subsequent iterations of the outermost loop.
MOVD $0, R1
CMP R1, R_LEN
BLE loop
MOVD (R_TMP3), R_TMP1
MOVD R_TMP1, (R_TMP0)
ADD $8, R_TMP3, R_TMP3
ADD $8, R_TMP0, R_TMP0
SUB $8, R_LEN, R_LEN
B finishSlowForwardCopy
verySlowForwardCopy:
// verySlowForwardCopy is a simple implementation of forward copy. In C
// parlance, this is a do/while loop instead of a while loop, since we know
// that length > 0. In Go syntax:
//
// for {
// dst[d] = dst[d - offset]
// d++
// length--
// if length == 0 {
// break
// }
// }
MOVB (R_TMP3), R_TMP1
MOVB R_TMP1, (R_DST)
ADD $1, R_TMP3, R_TMP3
ADD $1, R_DST, R_DST
SUB $1, R_LEN, R_LEN
CBNZ R_LEN, verySlowForwardCopy
B loop
// The code above handles copy tags.
// ----------------------------------------
end:
// This is the end of the "for s < len(src)".
//
// if d != len(dst) { etc }
CMP R_DEND, R_DST
BNE errCorrupt
// return 0
MOVD $0, ret+48(FP)
RET
errCorrupt:
// return decodeErrCodeCorrupt
MOVD $1, R_TMP0
MOVD R_TMP0, ret+48(FP)
RET

17
vendor/github.com/klauspost/compress/s2/decode_asm.go generated vendored Normal file
View File

@@ -0,0 +1,17 @@
// Copyright 2016 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build (amd64 || arm64) && !appengine && gc && !noasm
// +build amd64 arm64
// +build !appengine
// +build gc
// +build !noasm
package s2
// decode has the same semantics as in decode_other.go.
//
//go:noescape
func s2Decode(dst, src []byte) int

292
vendor/github.com/klauspost/compress/s2/decode_other.go generated vendored Normal file
View File

@@ -0,0 +1,292 @@
// Copyright 2016 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build (!amd64 && !arm64) || appengine || !gc || noasm
// +build !amd64,!arm64 appengine !gc noasm
package s2
import (
"fmt"
"strconv"
)
// decode writes the decoding of src to dst. It assumes that the varint-encoded
// length of the decompressed bytes has already been read, and that len(dst)
// equals that length.
//
// It returns 0 on success or a decodeErrCodeXxx error code on failure.
func s2Decode(dst, src []byte) int {
const debug = false
if debug {
fmt.Println("Starting decode, dst len:", len(dst))
}
var d, s, length int
offset := 0
// As long as we can read at least 5 bytes...
for s < len(src)-5 {
// Removing bounds checks is SLOWER, when if doing
// in := src[s:s+5]
// Checked on Go 1.18
switch src[s] & 0x03 {
case tagLiteral:
x := uint32(src[s] >> 2)
switch {
case x < 60:
s++
case x == 60:
s += 2
x = uint32(src[s-1])
case x == 61:
in := src[s : s+3]
x = uint32(in[1]) | uint32(in[2])<<8
s += 3
case x == 62:
in := src[s : s+4]
// Load as 32 bit and shift down.
x = uint32(in[0]) | uint32(in[1])<<8 | uint32(in[2])<<16 | uint32(in[3])<<24
x >>= 8
s += 4
case x == 63:
in := src[s : s+5]
x = uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24
s += 5
}
length = int(x) + 1
if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) {
if debug {
fmt.Println("corrupt: lit size", length)
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("literals, length:", length, "d-after:", d+length)
}
copy(dst[d:], src[s:s+length])
d += length
s += length
continue
case tagCopy1:
s += 2
toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
length = int(src[s-2]) >> 2 & 0x7
if toffset == 0 {
if debug {
fmt.Print("(repeat) ")
}
// keep last offset
switch length {
case 5:
length = int(src[s]) + 4
s += 1
case 6:
in := src[s : s+2]
length = int(uint32(in[0])|(uint32(in[1])<<8)) + (1 << 8)
s += 2
case 7:
in := src[s : s+3]
length = int((uint32(in[2])<<16)|(uint32(in[1])<<8)|uint32(in[0])) + (1 << 16)
s += 3
default: // 0-> 4
}
} else {
offset = toffset
}
length += 4
case tagCopy2:
in := src[s : s+3]
offset = int(uint32(in[1]) | uint32(in[2])<<8)
length = 1 + int(in[0])>>2
s += 3
case tagCopy4:
in := src[s : s+5]
offset = int(uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24)
length = 1 + int(in[0])>>2
s += 5
}
if offset <= 0 || d < offset || length > len(dst)-d {
if debug {
fmt.Println("corrupt: match, length", length, "offset:", offset, "dst avail:", len(dst)-d, "dst pos:", d)
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length)
}
// Copy from an earlier sub-slice of dst to a later sub-slice.
// If no overlap, use the built-in copy:
if offset > length {
copy(dst[d:d+length], dst[d-offset:])
d += length
continue
}
// Unlike the built-in copy function, this byte-by-byte copy always runs
// forwards, even if the slices overlap. Conceptually, this is:
//
// d += forwardCopy(dst[d:d+length], dst[d-offset:])
//
// We align the slices into a and b and show the compiler they are the same size.
// This allows the loop to run without bounds checks.
a := dst[d : d+length]
b := dst[d-offset:]
b = b[:len(a)]
for i := range a {
a[i] = b[i]
}
d += length
}
// Remaining with extra checks...
for s < len(src) {
switch src[s] & 0x03 {
case tagLiteral:
x := uint32(src[s] >> 2)
switch {
case x < 60:
s++
case x == 60:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
x = uint32(src[s-1])
case x == 61:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
x = uint32(src[s-2]) | uint32(src[s-1])<<8
case x == 62:
s += 4
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16
case x == 63:
s += 5
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24
}
length = int(x) + 1
if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) {
if debug {
fmt.Println("corrupt: lit size", length)
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("literals, length:", length, "d-after:", d+length)
}
copy(dst[d:], src[s:s+length])
d += length
s += length
continue
case tagCopy1:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = int(src[s-2]) >> 2 & 0x7
toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1]))
if toffset == 0 {
if debug {
fmt.Print("(repeat) ")
}
// keep last offset
switch length {
case 5:
s += 1
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-1])) + 4
case 6:
s += 2
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-2])|(uint32(src[s-1])<<8)) + (1 << 8)
case 7:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = int(uint32(src[s-3])|(uint32(src[s-2])<<8)|(uint32(src[s-1])<<16)) + (1 << 16)
default: // 0-> 4
}
} else {
offset = toffset
}
length += 4
case tagCopy2:
s += 3
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = 1 + int(src[s-3])>>2
offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8)
case tagCopy4:
s += 5
if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line.
return decodeErrCodeCorrupt
}
length = 1 + int(src[s-5])>>2
offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24)
}
if offset <= 0 || d < offset || length > len(dst)-d {
if debug {
fmt.Println("corrupt: match, length", length, "offset:", offset, "dst avail:", len(dst)-d, "dst pos:", d)
}
return decodeErrCodeCorrupt
}
if debug {
fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length)
}
// Copy from an earlier sub-slice of dst to a later sub-slice.
// If no overlap, use the built-in copy:
if offset > length {
copy(dst[d:d+length], dst[d-offset:])
d += length
continue
}
// Unlike the built-in copy function, this byte-by-byte copy always runs
// forwards, even if the slices overlap. Conceptually, this is:
//
// d += forwardCopy(dst[d:d+length], dst[d-offset:])
//
// We align the slices into a and b and show the compiler they are the same size.
// This allows the loop to run without bounds checks.
a := dst[d : d+length]
b := dst[d-offset:]
b = b[:len(a)]
for i := range a {
a[i] = b[i]
}
d += length
}
if d != len(dst) {
return decodeErrCodeCorrupt
}
return 0
}

350
vendor/github.com/klauspost/compress/s2/dict.go generated vendored Normal file
View File

@@ -0,0 +1,350 @@
// Copyright (c) 2022+ Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"bytes"
"encoding/binary"
"sync"
)
const (
// MinDictSize is the minimum dictionary size when repeat has been read.
MinDictSize = 16
// MaxDictSize is the maximum dictionary size when repeat has been read.
MaxDictSize = 65536
// MaxDictSrcOffset is the maximum offset where a dictionary entry can start.
MaxDictSrcOffset = 65535
)
// Dict contains a dictionary that can be used for encoding and decoding s2
type Dict struct {
dict []byte
repeat int // Repeat as index of dict
fast, better, best sync.Once
fastTable *[1 << 14]uint16
betterTableShort *[1 << 14]uint16
betterTableLong *[1 << 17]uint16
bestTableShort *[1 << 16]uint32
bestTableLong *[1 << 19]uint32
}
// NewDict will read a dictionary.
// It will return nil if the dictionary is invalid.
func NewDict(dict []byte) *Dict {
if len(dict) == 0 {
return nil
}
var d Dict
// Repeat is the first value of the dict
r, n := binary.Uvarint(dict)
if n <= 0 {
return nil
}
dict = dict[n:]
d.dict = dict
if cap(d.dict) < len(d.dict)+16 {
d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...)
}
if len(dict) < MinDictSize || len(dict) > MaxDictSize {
return nil
}
d.repeat = int(r)
if d.repeat > len(dict) {
return nil
}
return &d
}
// Bytes will return a serialized version of the dictionary.
// The output can be sent to NewDict.
func (d *Dict) Bytes() []byte {
dst := make([]byte, binary.MaxVarintLen16+len(d.dict))
return append(dst[:binary.PutUvarint(dst, uint64(d.repeat))], d.dict...)
}
// MakeDict will create a dictionary.
// 'data' must be at least MinDictSize.
// If data is longer than MaxDictSize only the last MaxDictSize bytes will be used.
// If searchStart is set the start repeat value will be set to the last
// match of this content.
// If no matches are found, it will attempt to find shorter matches.
// This content should match the typical start of a block.
// If at least 4 bytes cannot be matched, repeat is set to start of block.
func MakeDict(data []byte, searchStart []byte) *Dict {
if len(data) == 0 {
return nil
}
if len(data) > MaxDictSize {
data = data[len(data)-MaxDictSize:]
}
var d Dict
dict := data
d.dict = dict
if cap(d.dict) < len(d.dict)+16 {
d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...)
}
if len(dict) < MinDictSize {
return nil
}
// Find the longest match possible, last entry if multiple.
for s := len(searchStart); s > 4; s-- {
if idx := bytes.LastIndex(data, searchStart[:s]); idx >= 0 && idx <= len(data)-8 {
d.repeat = idx
break
}
}
return &d
}
// MakeDictManual will create a dictionary.
// 'data' must be at least MinDictSize and less than or equal to MaxDictSize.
// A manual first repeat index into data must be provided.
// It must be less than len(data)-8.
func MakeDictManual(data []byte, firstIdx uint16) *Dict {
if len(data) < MinDictSize || int(firstIdx) >= len(data)-8 || len(data) > MaxDictSize {
return nil
}
var d Dict
dict := data
d.dict = dict
if cap(d.dict) < len(d.dict)+16 {
d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...)
}
d.repeat = int(firstIdx)
return &d
}
// Encode returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func (d *Dict) Encode(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if cap(dst) < n {
dst = make([]byte, n)
} else {
dst = dst[:n]
}
// The block starts with the varint-encoded length of the decompressed bytes.
dstP := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:dstP]
}
if len(src) < minNonLiteralBlockSize {
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
n := encodeBlockDictGo(dst[dstP:], src, d)
if n > 0 {
dstP += n
return dst[:dstP]
}
// Not compressible
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
// EncodeBetter returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// EncodeBetter compresses better than Encode but typically with a
// 10-40% speed decrease on both compression and decompression.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func (d *Dict) EncodeBetter(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if len(dst) < n {
dst = make([]byte, n)
}
// The block starts with the varint-encoded length of the decompressed bytes.
dstP := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:dstP]
}
if len(src) < minNonLiteralBlockSize {
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
n := encodeBlockBetterDict(dst[dstP:], src, d)
if n > 0 {
dstP += n
return dst[:dstP]
}
// Not compressible
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
// EncodeBest returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// EncodeBest compresses as good as reasonably possible but with a
// big speed decrease.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func (d *Dict) EncodeBest(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if len(dst) < n {
dst = make([]byte, n)
}
// The block starts with the varint-encoded length of the decompressed bytes.
dstP := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:dstP]
}
if len(src) < minNonLiteralBlockSize {
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
n := encodeBlockBest(dst[dstP:], src, d)
if n > 0 {
dstP += n
return dst[:dstP]
}
// Not compressible
dstP += emitLiteral(dst[dstP:], src)
return dst[:dstP]
}
// Decode returns the decoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire decoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
func (d *Dict) Decode(dst, src []byte) ([]byte, error) {
dLen, s, err := decodedLen(src)
if err != nil {
return nil, err
}
if dLen <= cap(dst) {
dst = dst[:dLen]
} else {
dst = make([]byte, dLen)
}
if s2DecodeDict(dst, src[s:], d) != 0 {
return nil, ErrCorrupt
}
return dst, nil
}
func (d *Dict) initFast() {
d.fast.Do(func() {
const (
tableBits = 14
maxTableSize = 1 << tableBits
)
var table [maxTableSize]uint16
// We stop so any entry of length 8 can always be read.
for i := 0; i < len(d.dict)-8-2; i += 3 {
x0 := load64(d.dict, i)
h0 := hash6(x0, tableBits)
h1 := hash6(x0>>8, tableBits)
h2 := hash6(x0>>16, tableBits)
table[h0] = uint16(i)
table[h1] = uint16(i + 1)
table[h2] = uint16(i + 2)
}
d.fastTable = &table
})
}
func (d *Dict) initBetter() {
d.better.Do(func() {
const (
// Long hash matches.
lTableBits = 17
maxLTableSize = 1 << lTableBits
// Short hash matches.
sTableBits = 14
maxSTableSize = 1 << sTableBits
)
var lTable [maxLTableSize]uint16
var sTable [maxSTableSize]uint16
// We stop so any entry of length 8 can always be read.
for i := 0; i < len(d.dict)-8; i++ {
cv := load64(d.dict, i)
lTable[hash7(cv, lTableBits)] = uint16(i)
sTable[hash4(cv, sTableBits)] = uint16(i)
}
d.betterTableShort = &sTable
d.betterTableLong = &lTable
})
}
func (d *Dict) initBest() {
d.best.Do(func() {
const (
// Long hash matches.
lTableBits = 19
maxLTableSize = 1 << lTableBits
// Short hash matches.
sTableBits = 16
maxSTableSize = 1 << sTableBits
)
var lTable [maxLTableSize]uint32
var sTable [maxSTableSize]uint32
// We stop so any entry of length 8 can always be read.
for i := 0; i < len(d.dict)-8; i++ {
cv := load64(d.dict, i)
hashL := hash8(cv, lTableBits)
hashS := hash4(cv, sTableBits)
candidateL := lTable[hashL]
candidateS := sTable[hashS]
lTable[hashL] = uint32(i) | candidateL<<16
sTable[hashS] = uint32(i) | candidateS<<16
}
d.bestTableShort = &sTable
d.bestTableLong = &lTable
})
}

393
vendor/github.com/klauspost/compress/s2/encode.go generated vendored Normal file
View File

@@ -0,0 +1,393 @@
// Copyright 2011 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"encoding/binary"
"math"
"math/bits"
)
// Encode returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func Encode(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if cap(dst) < n {
dst = make([]byte, n)
} else {
dst = dst[:n]
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlock(dst[d:], src)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// EstimateBlockSize will perform a very fast compression
// without outputting the result and return the compressed output size.
// The function returns -1 if no improvement could be achieved.
// Using actual compression will most often produce better compression than the estimate.
func EstimateBlockSize(src []byte) (d int) {
if len(src) <= inputMargin || int64(len(src)) > 0xffffffff {
return -1
}
if len(src) <= 1024 {
d = calcBlockSizeSmall(src)
} else {
d = calcBlockSize(src)
}
if d == 0 {
return -1
}
// Size of the varint encoded block size.
d += (bits.Len64(uint64(len(src))) + 7) / 7
if d >= len(src) {
return -1
}
return d
}
// EncodeBetter returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// EncodeBetter compresses better than Encode but typically with a
// 10-40% speed decrease on both compression and decompression.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func EncodeBetter(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if len(dst) < n {
dst = make([]byte, n)
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlockBetter(dst[d:], src)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// EncodeBest returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// EncodeBest compresses as good as reasonably possible but with a
// big speed decrease.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func EncodeBest(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if len(dst) < n {
dst = make([]byte, n)
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlockBest(dst[d:], src, nil)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// EncodeSnappy returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The output is Snappy compatible and will likely decompress faster.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func EncodeSnappy(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if cap(dst) < n {
dst = make([]byte, n)
} else {
dst = dst[:n]
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlockSnappy(dst[d:], src)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// EncodeSnappyBetter returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The output is Snappy compatible and will likely decompress faster.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func EncodeSnappyBetter(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if cap(dst) < n {
dst = make([]byte, n)
} else {
dst = dst[:n]
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlockBetterSnappy(dst[d:], src)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// EncodeSnappyBest returns the encoded form of src. The returned slice may be a sub-
// slice of dst if dst was large enough to hold the entire encoded block.
// Otherwise, a newly allocated slice will be returned.
//
// The output is Snappy compatible and will likely decompress faster.
//
// The dst and src must not overlap. It is valid to pass a nil dst.
//
// The blocks will require the same amount of memory to decode as encoding,
// and does not make for concurrent decoding.
// Also note that blocks do not contain CRC information, so corruption may be undetected.
//
// If you need to encode larger amounts of data, consider using
// the streaming interface which gives all of these features.
func EncodeSnappyBest(dst, src []byte) []byte {
if n := MaxEncodedLen(len(src)); n < 0 {
panic(ErrTooLarge)
} else if cap(dst) < n {
dst = make([]byte, n)
} else {
dst = dst[:n]
}
// The block starts with the varint-encoded length of the decompressed bytes.
d := binary.PutUvarint(dst, uint64(len(src)))
if len(src) == 0 {
return dst[:d]
}
if len(src) < minNonLiteralBlockSize {
d += emitLiteral(dst[d:], src)
return dst[:d]
}
n := encodeBlockBestSnappy(dst[d:], src)
if n > 0 {
d += n
return dst[:d]
}
// Not compressible
d += emitLiteral(dst[d:], src)
return dst[:d]
}
// ConcatBlocks will concatenate the supplied blocks and append them to the supplied destination.
// If the destination is nil or too small, a new will be allocated.
// The blocks are not validated, so garbage in = garbage out.
// dst may not overlap block data.
// Any data in dst is preserved as is, so it will not be considered a block.
func ConcatBlocks(dst []byte, blocks ...[]byte) ([]byte, error) {
totalSize := uint64(0)
compSize := 0
for _, b := range blocks {
l, hdr, err := decodedLen(b)
if err != nil {
return nil, err
}
totalSize += uint64(l)
compSize += len(b) - hdr
}
if totalSize == 0 {
dst = append(dst, 0)
return dst, nil
}
if totalSize > math.MaxUint32 {
return nil, ErrTooLarge
}
var tmp [binary.MaxVarintLen32]byte
hdrSize := binary.PutUvarint(tmp[:], totalSize)
wantSize := hdrSize + compSize
if cap(dst)-len(dst) < wantSize {
dst = append(make([]byte, 0, wantSize+len(dst)), dst...)
}
dst = append(dst, tmp[:hdrSize]...)
for _, b := range blocks {
_, hdr, err := decodedLen(b)
if err != nil {
return nil, err
}
dst = append(dst, b[hdr:]...)
}
return dst, nil
}
// inputMargin is the minimum number of extra input bytes to keep, inside
// encodeBlock's inner loop. On some architectures, this margin lets us
// implement a fast path for emitLiteral, where the copy of short (<= 16 byte)
// literals can be implemented as a single load to and store from a 16-byte
// register. That literal's actual length can be as short as 1 byte, so this
// can copy up to 15 bytes too much, but that's OK as subsequent iterations of
// the encoding loop will fix up the copy overrun, and this inputMargin ensures
// that we don't overrun the dst and src buffers.
const inputMargin = 8
// minNonLiteralBlockSize is the minimum size of the input to encodeBlock that
// will be accepted by the encoder.
const minNonLiteralBlockSize = 32
const intReduction = 2 - (1 << (^uint(0) >> 63)) // 1 (32 bits) or 0 (64 bits)
// MaxBlockSize is the maximum value where MaxEncodedLen will return a valid block size.
// Blocks this big are highly discouraged, though.
// Half the size on 32 bit systems.
const MaxBlockSize = (1<<(32-intReduction) - 1) - binary.MaxVarintLen32 - 5
// MaxEncodedLen returns the maximum length of a snappy block, given its
// uncompressed length.
//
// It will return a negative value if srcLen is too large to encode.
// 32 bit platforms will have lower thresholds for rejecting big content.
func MaxEncodedLen(srcLen int) int {
n := uint64(srcLen)
if intReduction == 1 {
// 32 bits
if n > math.MaxInt32 {
// Also includes negative.
return -1
}
} else if n > 0xffffffff {
// 64 bits
// Also includes negative.
return -1
}
// Size of the varint encoded block size.
n = n + uint64((bits.Len64(n)+7)/7)
// Add maximum size of encoding block as literals.
n += uint64(literalExtraSize(int64(srcLen)))
if intReduction == 1 {
// 32 bits
if n > math.MaxInt32 {
return -1
}
} else if n > 0xffffffff {
// 64 bits
// Also includes negative.
return -1
}
return int(n)
}

1068
vendor/github.com/klauspost/compress/s2/encode_all.go generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff

162
vendor/github.com/klauspost/compress/s2/encode_amd64.go generated vendored Normal file
View File

@@ -0,0 +1,162 @@
//go:build !appengine && !noasm && gc
// +build !appengine,!noasm,gc
package s2
import "github.com/klauspost/compress/internal/race"
const hasAmd64Asm = true
// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlock(dst, src []byte) (d int) {
race.ReadSlice(src)
race.WriteSlice(dst)
const (
// Use 12 bit table when less than...
limit12B = 16 << 10
// Use 10 bit table when less than...
limit10B = 4 << 10
// Use 8 bit table when less than...
limit8B = 512
)
if len(src) >= 4<<20 {
return encodeBlockAsm(dst, src)
}
if len(src) >= limit12B {
return encodeBlockAsm4MB(dst, src)
}
if len(src) >= limit10B {
return encodeBlockAsm12B(dst, src)
}
if len(src) >= limit8B {
return encodeBlockAsm10B(dst, src)
}
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeBlockAsm8B(dst, src)
}
// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlockBetter(dst, src []byte) (d int) {
race.ReadSlice(src)
race.WriteSlice(dst)
const (
// Use 12 bit table when less than...
limit12B = 16 << 10
// Use 10 bit table when less than...
limit10B = 4 << 10
// Use 8 bit table when less than...
limit8B = 512
)
if len(src) > 4<<20 {
return encodeBetterBlockAsm(dst, src)
}
if len(src) >= limit12B {
return encodeBetterBlockAsm4MB(dst, src)
}
if len(src) >= limit10B {
return encodeBetterBlockAsm12B(dst, src)
}
if len(src) >= limit8B {
return encodeBetterBlockAsm10B(dst, src)
}
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeBetterBlockAsm8B(dst, src)
}
// encodeBlockSnappy encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlockSnappy(dst, src []byte) (d int) {
race.ReadSlice(src)
race.WriteSlice(dst)
const (
// Use 12 bit table when less than...
limit12B = 16 << 10
// Use 10 bit table when less than...
limit10B = 4 << 10
// Use 8 bit table when less than...
limit8B = 512
)
if len(src) >= 64<<10 {
return encodeSnappyBlockAsm(dst, src)
}
if len(src) >= limit12B {
return encodeSnappyBlockAsm64K(dst, src)
}
if len(src) >= limit10B {
return encodeSnappyBlockAsm12B(dst, src)
}
if len(src) >= limit8B {
return encodeSnappyBlockAsm10B(dst, src)
}
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeSnappyBlockAsm8B(dst, src)
}
// encodeBlockSnappy encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlockBetterSnappy(dst, src []byte) (d int) {
race.ReadSlice(src)
race.WriteSlice(dst)
const (
// Use 12 bit table when less than...
limit12B = 16 << 10
// Use 10 bit table when less than...
limit10B = 4 << 10
// Use 8 bit table when less than...
limit8B = 512
)
if len(src) >= 64<<10 {
return encodeSnappyBetterBlockAsm(dst, src)
}
if len(src) >= limit12B {
return encodeSnappyBetterBlockAsm64K(dst, src)
}
if len(src) >= limit10B {
return encodeSnappyBetterBlockAsm12B(dst, src)
}
if len(src) >= limit8B {
return encodeSnappyBetterBlockAsm10B(dst, src)
}
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeSnappyBetterBlockAsm8B(dst, src)
}

796
vendor/github.com/klauspost/compress/s2/encode_best.go generated vendored Normal file
View File

@@ -0,0 +1,796 @@
// Copyright 2016 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"fmt"
"math"
"math/bits"
)
// encodeBlockBest encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlockBest(dst, src []byte, dict *Dict) (d int) {
// Initialize the hash tables.
const (
// Long hash matches.
lTableBits = 19
maxLTableSize = 1 << lTableBits
// Short hash matches.
sTableBits = 16
maxSTableSize = 1 << sTableBits
inputMargin = 8 + 2
debug = false
)
// sLimit is when to stop looking for offset/length copies. The inputMargin
// lets us use a fast path for emitLiteral in the main loop, while we are
// looking for copies.
sLimit := len(src) - inputMargin
if len(src) < minNonLiteralBlockSize {
return 0
}
sLimitDict := len(src) - inputMargin
if sLimitDict > MaxDictSrcOffset-inputMargin {
sLimitDict = MaxDictSrcOffset - inputMargin
}
var lTable [maxLTableSize]uint64
var sTable [maxSTableSize]uint64
// Bail if we can't compress to at least this.
dstLimit := len(src) - 5
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := 0
// The encoded form must start with a literal, as there are no previous
// bytes to copy, so we start looking for hash matches at s == 1.
s := 1
repeat := 1
if dict != nil {
dict.initBest()
s = 0
repeat = len(dict.dict) - dict.repeat
}
cv := load64(src, s)
// We search for a repeat at -1, but don't output repeats when nextEmit == 0
const lowbitMask = 0xffffffff
getCur := func(x uint64) int {
return int(x & lowbitMask)
}
getPrev := func(x uint64) int {
return int(x >> 32)
}
const maxSkip = 64
for {
type match struct {
offset int
s int
length int
score int
rep, dict bool
}
var best match
for {
// Next src position to check
nextS := (s-nextEmit)>>8 + 1
if nextS > maxSkip {
nextS = s + maxSkip
} else {
nextS += s
}
if nextS > sLimit {
goto emitRemainder
}
if dict != nil && s >= MaxDictSrcOffset {
dict = nil
if repeat > s {
repeat = math.MinInt32
}
}
hashL := hash8(cv, lTableBits)
hashS := hash4(cv, sTableBits)
candidateL := lTable[hashL]
candidateS := sTable[hashS]
score := func(m match) int {
// Matches that are longer forward are penalized since we must emit it as a literal.
score := m.length - m.s
if nextEmit == m.s {
// If we do not have to emit literals, we save 1 byte
score++
}
offset := m.s - m.offset
if m.rep {
return score - emitRepeatSize(offset, m.length)
}
return score - emitCopySize(offset, m.length)
}
matchAt := func(offset, s int, first uint32, rep bool) match {
if best.length != 0 && best.s-best.offset == s-offset {
// Don't retest if we have the same offset.
return match{offset: offset, s: s}
}
if load32(src, offset) != first {
return match{offset: offset, s: s}
}
m := match{offset: offset, s: s, length: 4 + offset, rep: rep}
s += 4
for s < len(src) {
if len(src)-s < 8 {
if src[s] == src[m.length] {
m.length++
s++
continue
}
break
}
if diff := load64(src, s) ^ load64(src, m.length); diff != 0 {
m.length += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
m.length += 8
}
m.length -= offset
m.score = score(m)
if m.score <= -m.s {
// Eliminate if no savings, we might find a better one.
m.length = 0
}
return m
}
matchDict := func(candidate, s int, first uint32, rep bool) match {
if s >= MaxDictSrcOffset {
return match{offset: candidate, s: s}
}
// Calculate offset as if in continuous array with s
offset := -len(dict.dict) + candidate
if best.length != 0 && best.s-best.offset == s-offset && !rep {
// Don't retest if we have the same offset.
return match{offset: offset, s: s}
}
if load32(dict.dict, candidate) != first {
return match{offset: offset, s: s}
}
m := match{offset: offset, s: s, length: 4 + candidate, rep: rep, dict: true}
s += 4
if !rep {
for s < sLimitDict && m.length < len(dict.dict) {
if len(src)-s < 8 || len(dict.dict)-m.length < 8 {
if src[s] == dict.dict[m.length] {
m.length++
s++
continue
}
break
}
if diff := load64(src, s) ^ load64(dict.dict, m.length); diff != 0 {
m.length += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
m.length += 8
}
} else {
for s < len(src) && m.length < len(dict.dict) {
if len(src)-s < 8 || len(dict.dict)-m.length < 8 {
if src[s] == dict.dict[m.length] {
m.length++
s++
continue
}
break
}
if diff := load64(src, s) ^ load64(dict.dict, m.length); diff != 0 {
m.length += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
m.length += 8
}
}
m.length -= candidate
m.score = score(m)
if m.score <= -m.s {
// Eliminate if no savings, we might find a better one.
m.length = 0
}
return m
}
bestOf := func(a, b match) match {
if b.length == 0 {
return a
}
if a.length == 0 {
return b
}
as := a.score + b.s
bs := b.score + a.s
if as >= bs {
return a
}
return b
}
if s > 0 {
best = bestOf(matchAt(getCur(candidateL), s, uint32(cv), false), matchAt(getPrev(candidateL), s, uint32(cv), false))
best = bestOf(best, matchAt(getCur(candidateS), s, uint32(cv), false))
best = bestOf(best, matchAt(getPrev(candidateS), s, uint32(cv), false))
}
if dict != nil {
candidateL := dict.bestTableLong[hashL]
candidateS := dict.bestTableShort[hashS]
best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false))
best = bestOf(best, matchDict(int(candidateL>>16), s, uint32(cv), false))
best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false))
best = bestOf(best, matchDict(int(candidateS>>16), s, uint32(cv), false))
}
{
if (dict == nil || repeat <= s) && repeat > 0 {
best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8), true))
} else if s-repeat < -4 && dict != nil {
candidate := len(dict.dict) - (repeat - s)
best = bestOf(best, matchDict(candidate, s, uint32(cv), true))
candidate++
best = bestOf(best, matchDict(candidate, s+1, uint32(cv>>8), true))
}
if best.length > 0 {
hashS := hash4(cv>>8, sTableBits)
// s+1
nextShort := sTable[hashS]
s := s + 1
cv := load64(src, s)
hashL := hash8(cv, lTableBits)
nextLong := lTable[hashL]
best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv), false))
best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv), false))
best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv), false))
best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv), false))
// Dict at + 1
if dict != nil {
candidateL := dict.bestTableLong[hashL]
candidateS := dict.bestTableShort[hashS]
best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false))
best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false))
}
// s+2
if true {
hashS := hash4(cv>>8, sTableBits)
nextShort = sTable[hashS]
s++
cv = load64(src, s)
hashL := hash8(cv, lTableBits)
nextLong = lTable[hashL]
if (dict == nil || repeat <= s) && repeat > 0 {
// Repeat at + 2
best = bestOf(best, matchAt(s-repeat, s, uint32(cv), true))
} else if repeat-s > 4 && dict != nil {
candidate := len(dict.dict) - (repeat - s)
best = bestOf(best, matchDict(candidate, s, uint32(cv), true))
}
best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv), false))
best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv), false))
best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv), false))
best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv), false))
// Dict at +2
// Very small gain
if dict != nil {
candidateL := dict.bestTableLong[hashL]
candidateS := dict.bestTableShort[hashS]
best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false))
best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false))
}
}
// Search for a match at best match end, see if that is better.
// Allow some bytes at the beginning to mismatch.
// Sweet spot is around 1-2 bytes, but depends on input.
// The skipped bytes are tested in Extend backwards,
// and still picked up as part of the match if they do.
const skipBeginning = 2
const skipEnd = 1
if sAt := best.s + best.length - skipEnd; sAt < sLimit {
sBack := best.s + skipBeginning - skipEnd
backL := best.length - skipBeginning
// Load initial values
cv = load64(src, sBack)
// Grab candidates...
next := lTable[hash8(load64(src, sAt), lTableBits)]
if checkAt := getCur(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false))
}
if checkAt := getPrev(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false))
}
// Disabled: Extremely small gain
if false {
next = sTable[hash4(load64(src, sAt), sTableBits)]
if checkAt := getCur(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false))
}
if checkAt := getPrev(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false))
}
}
}
}
}
// Update table
lTable[hashL] = uint64(s) | candidateL<<32
sTable[hashS] = uint64(s) | candidateS<<32
if best.length > 0 {
break
}
cv = load64(src, nextS)
s = nextS
}
// Extend backwards, not needed for repeats...
s = best.s
if !best.rep && !best.dict {
for best.offset > 0 && s > nextEmit && src[best.offset-1] == src[s-1] {
best.offset--
best.length++
s--
}
}
if false && best.offset >= s {
panic(fmt.Errorf("t %d >= s %d", best.offset, s))
}
// Bail if we exceed the maximum size.
if d+(s-nextEmit) > dstLimit {
return 0
}
base := s
offset := s - best.offset
s += best.length
if offset > 65535 && s-base <= 5 && !best.rep {
// Bail if the match is equal or worse to the encoding.
s = best.s + 1
if s >= sLimit {
goto emitRemainder
}
cv = load64(src, s)
continue
}
if debug && nextEmit != base {
fmt.Println("EMIT", base-nextEmit, "literals. base-after:", base)
}
d += emitLiteral(dst[d:], src[nextEmit:base])
if best.rep {
if nextEmit > 0 || best.dict {
if debug {
fmt.Println("REPEAT, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best)
}
// same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset.
d += emitRepeat(dst[d:], offset, best.length)
} else {
// First match without dict cannot be a repeat.
if debug {
fmt.Println("COPY, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best)
}
d += emitCopy(dst[d:], offset, best.length)
}
} else {
if debug {
fmt.Println("COPY, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best)
}
d += emitCopy(dst[d:], offset, best.length)
}
repeat = offset
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
if d > dstLimit {
// Do we have space for more, if not bail.
return 0
}
// Fill tables...
for i := best.s + 1; i < s; i++ {
cv0 := load64(src, i)
long0 := hash8(cv0, lTableBits)
short0 := hash4(cv0, sTableBits)
lTable[long0] = uint64(i) | lTable[long0]<<32
sTable[short0] = uint64(i) | sTable[short0]<<32
}
cv = load64(src, s)
}
emitRemainder:
if nextEmit < len(src) {
// Bail if we exceed the maximum size.
if d+len(src)-nextEmit > dstLimit {
return 0
}
if debug && nextEmit != s {
fmt.Println("emitted ", len(src)-nextEmit, "literals")
}
d += emitLiteral(dst[d:], src[nextEmit:])
}
return d
}
// encodeBlockBestSnappy encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src)) &&
// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize
func encodeBlockBestSnappy(dst, src []byte) (d int) {
// Initialize the hash tables.
const (
// Long hash matches.
lTableBits = 19
maxLTableSize = 1 << lTableBits
// Short hash matches.
sTableBits = 16
maxSTableSize = 1 << sTableBits
inputMargin = 8 + 2
)
// sLimit is when to stop looking for offset/length copies. The inputMargin
// lets us use a fast path for emitLiteral in the main loop, while we are
// looking for copies.
sLimit := len(src) - inputMargin
if len(src) < minNonLiteralBlockSize {
return 0
}
var lTable [maxLTableSize]uint64
var sTable [maxSTableSize]uint64
// Bail if we can't compress to at least this.
dstLimit := len(src) - 5
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := 0
// The encoded form must start with a literal, as there are no previous
// bytes to copy, so we start looking for hash matches at s == 1.
s := 1
cv := load64(src, s)
// We search for a repeat at -1, but don't output repeats when nextEmit == 0
repeat := 1
const lowbitMask = 0xffffffff
getCur := func(x uint64) int {
return int(x & lowbitMask)
}
getPrev := func(x uint64) int {
return int(x >> 32)
}
const maxSkip = 64
for {
type match struct {
offset int
s int
length int
score int
}
var best match
for {
// Next src position to check
nextS := (s-nextEmit)>>8 + 1
if nextS > maxSkip {
nextS = s + maxSkip
} else {
nextS += s
}
if nextS > sLimit {
goto emitRemainder
}
hashL := hash8(cv, lTableBits)
hashS := hash4(cv, sTableBits)
candidateL := lTable[hashL]
candidateS := sTable[hashS]
score := func(m match) int {
// Matches that are longer forward are penalized since we must emit it as a literal.
score := m.length - m.s
if nextEmit == m.s {
// If we do not have to emit literals, we save 1 byte
score++
}
offset := m.s - m.offset
return score - emitCopyNoRepeatSize(offset, m.length)
}
matchAt := func(offset, s int, first uint32) match {
if best.length != 0 && best.s-best.offset == s-offset {
// Don't retest if we have the same offset.
return match{offset: offset, s: s}
}
if load32(src, offset) != first {
return match{offset: offset, s: s}
}
m := match{offset: offset, s: s, length: 4 + offset}
s += 4
for s <= sLimit {
if diff := load64(src, s) ^ load64(src, m.length); diff != 0 {
m.length += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
m.length += 8
}
m.length -= offset
m.score = score(m)
if m.score <= -m.s {
// Eliminate if no savings, we might find a better one.
m.length = 0
}
return m
}
bestOf := func(a, b match) match {
if b.length == 0 {
return a
}
if a.length == 0 {
return b
}
as := a.score + b.s
bs := b.score + a.s
if as >= bs {
return a
}
return b
}
best = bestOf(matchAt(getCur(candidateL), s, uint32(cv)), matchAt(getPrev(candidateL), s, uint32(cv)))
best = bestOf(best, matchAt(getCur(candidateS), s, uint32(cv)))
best = bestOf(best, matchAt(getPrev(candidateS), s, uint32(cv)))
{
best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8)))
if best.length > 0 {
// s+1
nextShort := sTable[hash4(cv>>8, sTableBits)]
s := s + 1
cv := load64(src, s)
nextLong := lTable[hash8(cv, lTableBits)]
best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv)))
best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv)))
best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv)))
best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv)))
// Repeat at + 2
best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8)))
// s+2
if true {
nextShort = sTable[hash4(cv>>8, sTableBits)]
s++
cv = load64(src, s)
nextLong = lTable[hash8(cv, lTableBits)]
best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv)))
best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv)))
best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv)))
best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv)))
}
// Search for a match at best match end, see if that is better.
if sAt := best.s + best.length; sAt < sLimit {
sBack := best.s
backL := best.length
// Load initial values
cv = load64(src, sBack)
// Search for mismatch
next := lTable[hash8(load64(src, sAt), lTableBits)]
//next := sTable[hash4(load64(src, sAt), sTableBits)]
if checkAt := getCur(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv)))
}
if checkAt := getPrev(next) - backL; checkAt > 0 {
best = bestOf(best, matchAt(checkAt, sBack, uint32(cv)))
}
}
}
}
// Update table
lTable[hashL] = uint64(s) | candidateL<<32
sTable[hashS] = uint64(s) | candidateS<<32
if best.length > 0 {
break
}
cv = load64(src, nextS)
s = nextS
}
// Extend backwards, not needed for repeats...
s = best.s
if true {
for best.offset > 0 && s > nextEmit && src[best.offset-1] == src[s-1] {
best.offset--
best.length++
s--
}
}
if false && best.offset >= s {
panic(fmt.Errorf("t %d >= s %d", best.offset, s))
}
// Bail if we exceed the maximum size.
if d+(s-nextEmit) > dstLimit {
return 0
}
base := s
offset := s - best.offset
s += best.length
if offset > 65535 && s-base <= 5 {
// Bail if the match is equal or worse to the encoding.
s = best.s + 1
if s >= sLimit {
goto emitRemainder
}
cv = load64(src, s)
continue
}
d += emitLiteral(dst[d:], src[nextEmit:base])
d += emitCopyNoRepeat(dst[d:], offset, best.length)
repeat = offset
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
if d > dstLimit {
// Do we have space for more, if not bail.
return 0
}
// Fill tables...
for i := best.s + 1; i < s; i++ {
cv0 := load64(src, i)
long0 := hash8(cv0, lTableBits)
short0 := hash4(cv0, sTableBits)
lTable[long0] = uint64(i) | lTable[long0]<<32
sTable[short0] = uint64(i) | sTable[short0]<<32
}
cv = load64(src, s)
}
emitRemainder:
if nextEmit < len(src) {
// Bail if we exceed the maximum size.
if d+len(src)-nextEmit > dstLimit {
return 0
}
d += emitLiteral(dst[d:], src[nextEmit:])
}
return d
}
// emitCopySize returns the size to encode the offset+length
//
// It assumes that:
//
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
func emitCopySize(offset, length int) int {
if offset >= 65536 {
i := 0
if length > 64 {
length -= 64
if length >= 4 {
// Emit remaining as repeats
return 5 + emitRepeatSize(offset, length)
}
i = 5
}
if length == 0 {
return i
}
return i + 5
}
// Offset no more than 2 bytes.
if length > 64 {
if offset < 2048 {
// Emit 8 bytes, then rest as repeats...
return 2 + emitRepeatSize(offset, length-8)
}
// Emit remaining as repeats, at least 4 bytes remain.
return 3 + emitRepeatSize(offset, length-60)
}
if length >= 12 || offset >= 2048 {
return 3
}
// Emit the remaining copy, encoded as 2 bytes.
return 2
}
// emitCopyNoRepeatSize returns the size to encode the offset+length
//
// It assumes that:
//
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
func emitCopyNoRepeatSize(offset, length int) int {
if offset >= 65536 {
return 5 + 5*(length/64)
}
// Offset no more than 2 bytes.
if length > 64 {
// Emit remaining as repeats, at least 4 bytes remain.
return 3 + 3*(length/60)
}
if length >= 12 || offset >= 2048 {
return 3
}
// Emit the remaining copy, encoded as 2 bytes.
return 2
}
// emitRepeatSize returns the number of bytes required to encode a repeat.
// Length must be at least 4 and < 1<<24
func emitRepeatSize(offset, length int) int {
// Repeat offset, make length cheaper
if length <= 4+4 || (length < 8+4 && offset < 2048) {
return 2
}
if length < (1<<8)+4+4 {
return 3
}
if length < (1<<16)+(1<<8)+4 {
return 4
}
const maxRepeat = (1 << 24) - 1
length -= (1 << 16) - 4
left := 0
if length > maxRepeat {
left = length - maxRepeat + 4
}
if left > 0 {
return 5 + emitRepeatSize(offset, left)
}
return 5
}

1106
vendor/github.com/klauspost/compress/s2/encode_better.go generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff

729
vendor/github.com/klauspost/compress/s2/encode_go.go generated vendored Normal file
View File

@@ -0,0 +1,729 @@
//go:build !amd64 || appengine || !gc || noasm
// +build !amd64 appengine !gc noasm
package s2
import (
"bytes"
"math/bits"
)
const hasAmd64Asm = false
// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src))
func encodeBlock(dst, src []byte) (d int) {
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeBlockGo(dst, src)
}
// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src))
func encodeBlockBetter(dst, src []byte) (d int) {
return encodeBlockBetterGo(dst, src)
}
// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src))
func encodeBlockBetterSnappy(dst, src []byte) (d int) {
return encodeBlockBetterSnappyGo(dst, src)
}
// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It
// assumes that the varint-encoded length of the decompressed bytes has already
// been written.
//
// It also assumes that:
//
// len(dst) >= MaxEncodedLen(len(src))
func encodeBlockSnappy(dst, src []byte) (d int) {
if len(src) < minNonLiteralBlockSize {
return 0
}
return encodeBlockSnappyGo(dst, src)
}
// emitLiteral writes a literal chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 0 <= len(lit) && len(lit) <= math.MaxUint32
func emitLiteral(dst, lit []byte) int {
if len(lit) == 0 {
return 0
}
const num = 63<<2 | tagLiteral
i, n := 0, uint(len(lit)-1)
switch {
case n < 60:
dst[0] = uint8(n)<<2 | tagLiteral
i = 1
case n < 1<<8:
dst[1] = uint8(n)
dst[0] = 60<<2 | tagLiteral
i = 2
case n < 1<<16:
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 61<<2 | tagLiteral
i = 3
case n < 1<<24:
dst[3] = uint8(n >> 16)
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 62<<2 | tagLiteral
i = 4
default:
dst[4] = uint8(n >> 24)
dst[3] = uint8(n >> 16)
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 63<<2 | tagLiteral
i = 5
}
return i + copy(dst[i:], lit)
}
// emitRepeat writes a repeat chunk and returns the number of bytes written.
// Length must be at least 4 and < 1<<24
func emitRepeat(dst []byte, offset, length int) int {
// Repeat offset, make length cheaper
length -= 4
if length <= 4 {
dst[0] = uint8(length)<<2 | tagCopy1
dst[1] = 0
return 2
}
if length < 8 && offset < 2048 {
// Encode WITH offset
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1
return 2
}
if length < (1<<8)+4 {
length -= 4
dst[2] = uint8(length)
dst[1] = 0
dst[0] = 5<<2 | tagCopy1
return 3
}
if length < (1<<16)+(1<<8) {
length -= 1 << 8
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 6<<2 | tagCopy1
return 4
}
const maxRepeat = (1 << 24) - 1
length -= 1 << 16
left := 0
if length > maxRepeat {
left = length - maxRepeat + 4
length = maxRepeat - 4
}
dst[4] = uint8(length >> 16)
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 7<<2 | tagCopy1
if left > 0 {
return 5 + emitRepeat(dst[5:], offset, left)
}
return 5
}
// emitCopy writes a copy chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
func emitCopy(dst []byte, offset, length int) int {
if offset >= 65536 {
i := 0
if length > 64 {
// Emit a length 64 copy, encoded as 5 bytes.
dst[4] = uint8(offset >> 24)
dst[3] = uint8(offset >> 16)
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 63<<2 | tagCopy4
length -= 64
if length >= 4 {
// Emit remaining as repeats
return 5 + emitRepeat(dst[5:], offset, length)
}
i = 5
}
if length == 0 {
return i
}
// Emit a copy, offset encoded as 4 bytes.
dst[i+0] = uint8(length-1)<<2 | tagCopy4
dst[i+1] = uint8(offset)
dst[i+2] = uint8(offset >> 8)
dst[i+3] = uint8(offset >> 16)
dst[i+4] = uint8(offset >> 24)
return i + 5
}
// Offset no more than 2 bytes.
if length > 64 {
off := 3
if offset < 2048 {
// emit 8 bytes as tagCopy1, rest as repeats.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1
length -= 8
off = 2
} else {
// Emit a length 60 copy, encoded as 3 bytes.
// Emit remaining as repeat value (minimum 4 bytes).
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 59<<2 | tagCopy2
length -= 60
}
// Emit remaining as repeats, at least 4 bytes remain.
return off + emitRepeat(dst[off:], offset, length)
}
if length >= 12 || offset >= 2048 {
// Emit the remaining copy, encoded as 3 bytes.
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = uint8(length-1)<<2 | tagCopy2
return 3
}
// Emit the remaining copy, encoded as 2 bytes.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
return 2
}
// emitCopyNoRepeat writes a copy chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
func emitCopyNoRepeat(dst []byte, offset, length int) int {
if offset >= 65536 {
i := 0
if length > 64 {
// Emit a length 64 copy, encoded as 5 bytes.
dst[4] = uint8(offset >> 24)
dst[3] = uint8(offset >> 16)
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 63<<2 | tagCopy4
length -= 64
if length >= 4 {
// Emit remaining as repeats
return 5 + emitCopyNoRepeat(dst[5:], offset, length)
}
i = 5
}
if length == 0 {
return i
}
// Emit a copy, offset encoded as 4 bytes.
dst[i+0] = uint8(length-1)<<2 | tagCopy4
dst[i+1] = uint8(offset)
dst[i+2] = uint8(offset >> 8)
dst[i+3] = uint8(offset >> 16)
dst[i+4] = uint8(offset >> 24)
return i + 5
}
// Offset no more than 2 bytes.
if length > 64 {
// Emit a length 60 copy, encoded as 3 bytes.
// Emit remaining as repeat value (minimum 4 bytes).
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 59<<2 | tagCopy2
length -= 60
// Emit remaining as repeats, at least 4 bytes remain.
return 3 + emitCopyNoRepeat(dst[3:], offset, length)
}
if length >= 12 || offset >= 2048 {
// Emit the remaining copy, encoded as 3 bytes.
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = uint8(length-1)<<2 | tagCopy2
return 3
}
// Emit the remaining copy, encoded as 2 bytes.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
return 2
}
// matchLen returns how many bytes match in a and b
//
// It assumes that:
//
// len(a) <= len(b)
func matchLen(a []byte, b []byte) int {
b = b[:len(a)]
var checked int
if len(a) > 4 {
// Try 4 bytes first
if diff := load32(a, 0) ^ load32(b, 0); diff != 0 {
return bits.TrailingZeros32(diff) >> 3
}
// Switch to 8 byte matching.
checked = 4
a = a[4:]
b = b[4:]
for len(a) >= 8 {
b = b[:len(a)]
if diff := load64(a, 0) ^ load64(b, 0); diff != 0 {
return checked + (bits.TrailingZeros64(diff) >> 3)
}
checked += 8
a = a[8:]
b = b[8:]
}
}
b = b[:len(a)]
for i := range a {
if a[i] != b[i] {
return int(i) + checked
}
}
return len(a) + checked
}
// input must be > inputMargin
func calcBlockSize(src []byte) (d int) {
// Initialize the hash table.
const (
tableBits = 13
maxTableSize = 1 << tableBits
)
var table [maxTableSize]uint32
// sLimit is when to stop looking for offset/length copies. The inputMargin
// lets us use a fast path for emitLiteral in the main loop, while we are
// looking for copies.
sLimit := len(src) - inputMargin
// Bail if we can't compress to at least this.
dstLimit := len(src) - len(src)>>5 - 5
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := 0
// The encoded form must start with a literal, as there are no previous
// bytes to copy, so we start looking for hash matches at s == 1.
s := 1
cv := load64(src, s)
// We search for a repeat at -1, but don't output repeats when nextEmit == 0
repeat := 1
for {
candidate := 0
for {
// Next src position to check
nextS := s + (s-nextEmit)>>6 + 4
if nextS > sLimit {
goto emitRemainder
}
hash0 := hash6(cv, tableBits)
hash1 := hash6(cv>>8, tableBits)
candidate = int(table[hash0])
candidate2 := int(table[hash1])
table[hash0] = uint32(s)
table[hash1] = uint32(s + 1)
hash2 := hash6(cv>>16, tableBits)
// Check repeat at offset checkRep.
const checkRep = 1
if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) {
base := s + checkRep
// Extend back
for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; {
i--
base--
}
d += emitLiteralSize(src[nextEmit:base])
// Extend forward
candidate := s - repeat + 4 + checkRep
s += 4 + checkRep
for s <= sLimit {
if diff := load64(src, s) ^ load64(src, candidate); diff != 0 {
s += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
candidate += 8
}
d += emitCopyNoRepeatSize(repeat, s-base)
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
cv = load64(src, s)
continue
}
if uint32(cv) == load32(src, candidate) {
break
}
candidate = int(table[hash2])
if uint32(cv>>8) == load32(src, candidate2) {
table[hash2] = uint32(s + 2)
candidate = candidate2
s++
break
}
table[hash2] = uint32(s + 2)
if uint32(cv>>16) == load32(src, candidate) {
s += 2
break
}
cv = load64(src, nextS)
s = nextS
}
// Extend backwards
for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] {
candidate--
s--
}
// Bail if we exceed the maximum size.
if d+(s-nextEmit) > dstLimit {
return 0
}
// A 4-byte match has been found. We'll later see if more than 4 bytes
// match. But, prior to the match, src[nextEmit:s] are unmatched. Emit
// them as literal bytes.
d += emitLiteralSize(src[nextEmit:s])
// Call emitCopy, and then see if another emitCopy could be our next
// move. Repeat until we find no match for the input immediately after
// what was consumed by the last emitCopy call.
//
// If we exit this loop normally then we need to call emitLiteral next,
// though we don't yet know how big the literal will be. We handle that
// by proceeding to the next iteration of the main loop. We also can
// exit this loop via goto if we get close to exhausting the input.
for {
// Invariant: we have a 4-byte match at s, and no need to emit any
// literal bytes prior to s.
base := s
repeat = base - candidate
// Extend the 4-byte match as long as possible.
s += 4
candidate += 4
for s <= len(src)-8 {
if diff := load64(src, s) ^ load64(src, candidate); diff != 0 {
s += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
candidate += 8
}
d += emitCopyNoRepeatSize(repeat, s-base)
if false {
// Validate match.
a := src[base:s]
b := src[base-repeat : base-repeat+(s-base)]
if !bytes.Equal(a, b) {
panic("mismatch")
}
}
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
if d > dstLimit {
// Do we have space for more, if not bail.
return 0
}
// Check for an immediate match, otherwise start search at s+1
x := load64(src, s-2)
m2Hash := hash6(x, tableBits)
currHash := hash6(x>>16, tableBits)
candidate = int(table[currHash])
table[m2Hash] = uint32(s - 2)
table[currHash] = uint32(s)
if uint32(x>>16) != load32(src, candidate) {
cv = load64(src, s+1)
s++
break
}
}
}
emitRemainder:
if nextEmit < len(src) {
// Bail if we exceed the maximum size.
if d+len(src)-nextEmit > dstLimit {
return 0
}
d += emitLiteralSize(src[nextEmit:])
}
return d
}
// length must be > inputMargin.
func calcBlockSizeSmall(src []byte) (d int) {
// Initialize the hash table.
const (
tableBits = 9
maxTableSize = 1 << tableBits
)
var table [maxTableSize]uint32
// sLimit is when to stop looking for offset/length copies. The inputMargin
// lets us use a fast path for emitLiteral in the main loop, while we are
// looking for copies.
sLimit := len(src) - inputMargin
// Bail if we can't compress to at least this.
dstLimit := len(src) - len(src)>>5 - 5
// nextEmit is where in src the next emitLiteral should start from.
nextEmit := 0
// The encoded form must start with a literal, as there are no previous
// bytes to copy, so we start looking for hash matches at s == 1.
s := 1
cv := load64(src, s)
// We search for a repeat at -1, but don't output repeats when nextEmit == 0
repeat := 1
for {
candidate := 0
for {
// Next src position to check
nextS := s + (s-nextEmit)>>6 + 4
if nextS > sLimit {
goto emitRemainder
}
hash0 := hash6(cv, tableBits)
hash1 := hash6(cv>>8, tableBits)
candidate = int(table[hash0])
candidate2 := int(table[hash1])
table[hash0] = uint32(s)
table[hash1] = uint32(s + 1)
hash2 := hash6(cv>>16, tableBits)
// Check repeat at offset checkRep.
const checkRep = 1
if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) {
base := s + checkRep
// Extend back
for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; {
i--
base--
}
d += emitLiteralSize(src[nextEmit:base])
// Extend forward
candidate := s - repeat + 4 + checkRep
s += 4 + checkRep
for s <= sLimit {
if diff := load64(src, s) ^ load64(src, candidate); diff != 0 {
s += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
candidate += 8
}
d += emitCopyNoRepeatSize(repeat, s-base)
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
cv = load64(src, s)
continue
}
if uint32(cv) == load32(src, candidate) {
break
}
candidate = int(table[hash2])
if uint32(cv>>8) == load32(src, candidate2) {
table[hash2] = uint32(s + 2)
candidate = candidate2
s++
break
}
table[hash2] = uint32(s + 2)
if uint32(cv>>16) == load32(src, candidate) {
s += 2
break
}
cv = load64(src, nextS)
s = nextS
}
// Extend backwards
for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] {
candidate--
s--
}
// Bail if we exceed the maximum size.
if d+(s-nextEmit) > dstLimit {
return 0
}
// A 4-byte match has been found. We'll later see if more than 4 bytes
// match. But, prior to the match, src[nextEmit:s] are unmatched. Emit
// them as literal bytes.
d += emitLiteralSize(src[nextEmit:s])
// Call emitCopy, and then see if another emitCopy could be our next
// move. Repeat until we find no match for the input immediately after
// what was consumed by the last emitCopy call.
//
// If we exit this loop normally then we need to call emitLiteral next,
// though we don't yet know how big the literal will be. We handle that
// by proceeding to the next iteration of the main loop. We also can
// exit this loop via goto if we get close to exhausting the input.
for {
// Invariant: we have a 4-byte match at s, and no need to emit any
// literal bytes prior to s.
base := s
repeat = base - candidate
// Extend the 4-byte match as long as possible.
s += 4
candidate += 4
for s <= len(src)-8 {
if diff := load64(src, s) ^ load64(src, candidate); diff != 0 {
s += bits.TrailingZeros64(diff) >> 3
break
}
s += 8
candidate += 8
}
d += emitCopyNoRepeatSize(repeat, s-base)
if false {
// Validate match.
a := src[base:s]
b := src[base-repeat : base-repeat+(s-base)]
if !bytes.Equal(a, b) {
panic("mismatch")
}
}
nextEmit = s
if s >= sLimit {
goto emitRemainder
}
if d > dstLimit {
// Do we have space for more, if not bail.
return 0
}
// Check for an immediate match, otherwise start search at s+1
x := load64(src, s-2)
m2Hash := hash6(x, tableBits)
currHash := hash6(x>>16, tableBits)
candidate = int(table[currHash])
table[m2Hash] = uint32(s - 2)
table[currHash] = uint32(s)
if uint32(x>>16) != load32(src, candidate) {
cv = load64(src, s+1)
s++
break
}
}
}
emitRemainder:
if nextEmit < len(src) {
// Bail if we exceed the maximum size.
if d+len(src)-nextEmit > dstLimit {
return 0
}
d += emitLiteralSize(src[nextEmit:])
}
return d
}
// emitLiteral writes a literal chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 0 <= len(lit) && len(lit) <= math.MaxUint32
func emitLiteralSize(lit []byte) int {
if len(lit) == 0 {
return 0
}
switch {
case len(lit) <= 60:
return len(lit) + 1
case len(lit) <= 1<<8:
return len(lit) + 2
case len(lit) <= 1<<16:
return len(lit) + 3
case len(lit) <= 1<<24:
return len(lit) + 4
default:
return len(lit) + 5
}
}
func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) {
panic("cvtLZ4BlockAsm should be unreachable")
}
func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) {
panic("cvtLZ4BlockSnappyAsm should be unreachable")
}
func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) {
panic("cvtLZ4sBlockAsm should be unreachable")
}
func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) {
panic("cvtLZ4sBlockSnappyAsm should be unreachable")
}

228
vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go generated vendored Normal file
View File

@@ -0,0 +1,228 @@
// Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT.
//go:build !appengine && !noasm && gc && !noasm
package s2
func _dummy_()
// encodeBlockAsm encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4294967295 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBlockAsm(dst []byte, src []byte) int
// encodeBlockAsm4MB encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4194304 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBlockAsm4MB(dst []byte, src []byte) int
// encodeBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 16383 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBlockAsm12B(dst []byte, src []byte) int
// encodeBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4095 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBlockAsm10B(dst []byte, src []byte) int
// encodeBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 511 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBlockAsm8B(dst []byte, src []byte) int
// encodeBetterBlockAsm encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4294967295 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBetterBlockAsm(dst []byte, src []byte) int
// encodeBetterBlockAsm4MB encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4194304 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBetterBlockAsm4MB(dst []byte, src []byte) int
// encodeBetterBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 16383 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBetterBlockAsm12B(dst []byte, src []byte) int
// encodeBetterBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4095 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBetterBlockAsm10B(dst []byte, src []byte) int
// encodeBetterBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 511 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeBetterBlockAsm8B(dst []byte, src []byte) int
// encodeSnappyBlockAsm encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4294967295 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBlockAsm(dst []byte, src []byte) int
// encodeSnappyBlockAsm64K encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 65535 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBlockAsm64K(dst []byte, src []byte) int
// encodeSnappyBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 16383 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBlockAsm12B(dst []byte, src []byte) int
// encodeSnappyBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4095 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBlockAsm10B(dst []byte, src []byte) int
// encodeSnappyBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 511 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBlockAsm8B(dst []byte, src []byte) int
// encodeSnappyBetterBlockAsm encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4294967295 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBetterBlockAsm(dst []byte, src []byte) int
// encodeSnappyBetterBlockAsm64K encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 65535 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBetterBlockAsm64K(dst []byte, src []byte) int
// encodeSnappyBetterBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 16383 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBetterBlockAsm12B(dst []byte, src []byte) int
// encodeSnappyBetterBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4095 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBetterBlockAsm10B(dst []byte, src []byte) int
// encodeSnappyBetterBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 511 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func encodeSnappyBetterBlockAsm8B(dst []byte, src []byte) int
// calcBlockSize encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 4294967295 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func calcBlockSize(src []byte) int
// calcBlockSizeSmall encodes a non-empty src to a guaranteed-large-enough dst.
// Maximum input 1024 bytes.
// It assumes that the varint-encoded length of the decompressed bytes has already been written.
//
//go:noescape
func calcBlockSizeSmall(src []byte) int
// emitLiteral writes a literal chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes with margin of 0 bytes
// 0 <= len(lit) && len(lit) <= math.MaxUint32
//
//go:noescape
func emitLiteral(dst []byte, lit []byte) int
// emitRepeat writes a repeat chunk and returns the number of bytes written.
// Length must be at least 4 and < 1<<32
//
//go:noescape
func emitRepeat(dst []byte, offset int, length int) int
// emitCopy writes a copy chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
//
//go:noescape
func emitCopy(dst []byte, offset int, length int) int
// emitCopyNoRepeat writes a copy chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 1 <= offset && offset <= math.MaxUint32
// 4 <= length && length <= 1 << 24
//
//go:noescape
func emitCopyNoRepeat(dst []byte, offset int, length int) int
// matchLen returns how many bytes match in a and b
//
// It assumes that:
//
// len(a) <= len(b)
//
//go:noescape
func matchLen(a []byte, b []byte) int
// cvtLZ4Block converts an LZ4 block to S2
//
//go:noescape
func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int)
// cvtLZ4sBlock converts an LZ4s block to S2
//
//go:noescape
func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int)
// cvtLZ4Block converts an LZ4 block to Snappy
//
//go:noescape
func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int)
// cvtLZ4sBlock converts an LZ4s block to Snappy
//
//go:noescape
func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int)

21277
vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff

602
vendor/github.com/klauspost/compress/s2/index.go generated vendored Normal file
View File

@@ -0,0 +1,602 @@
// Copyright (c) 2022+ Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"bytes"
"encoding/binary"
"encoding/json"
"fmt"
"io"
"sort"
)
const (
S2IndexHeader = "s2idx\x00"
S2IndexTrailer = "\x00xdi2s"
maxIndexEntries = 1 << 16
// If distance is less than this, we do not add the entry.
minIndexDist = 1 << 20
)
// Index represents an S2/Snappy index.
type Index struct {
TotalUncompressed int64 // Total Uncompressed size if known. Will be -1 if unknown.
TotalCompressed int64 // Total Compressed size if known. Will be -1 if unknown.
info []struct {
compressedOffset int64
uncompressedOffset int64
}
estBlockUncomp int64
}
func (i *Index) reset(maxBlock int) {
i.estBlockUncomp = int64(maxBlock)
i.TotalCompressed = -1
i.TotalUncompressed = -1
if len(i.info) > 0 {
i.info = i.info[:0]
}
}
// allocInfos will allocate an empty slice of infos.
func (i *Index) allocInfos(n int) {
if n > maxIndexEntries {
panic("n > maxIndexEntries")
}
i.info = make([]struct {
compressedOffset int64
uncompressedOffset int64
}, 0, n)
}
// add an uncompressed and compressed pair.
// Entries must be sent in order.
func (i *Index) add(compressedOffset, uncompressedOffset int64) error {
if i == nil {
return nil
}
lastIdx := len(i.info) - 1
if lastIdx >= 0 {
latest := i.info[lastIdx]
if latest.uncompressedOffset == uncompressedOffset {
// Uncompressed didn't change, don't add entry,
// but update start index.
latest.compressedOffset = compressedOffset
i.info[lastIdx] = latest
return nil
}
if latest.uncompressedOffset > uncompressedOffset {
return fmt.Errorf("internal error: Earlier uncompressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset)
}
if latest.compressedOffset > compressedOffset {
return fmt.Errorf("internal error: Earlier compressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset)
}
if latest.uncompressedOffset+minIndexDist > uncompressedOffset {
// Only add entry if distance is large enough.
return nil
}
}
i.info = append(i.info, struct {
compressedOffset int64
uncompressedOffset int64
}{compressedOffset: compressedOffset, uncompressedOffset: uncompressedOffset})
return nil
}
// Find the offset at or before the wanted (uncompressed) offset.
// If offset is 0 or positive it is the offset from the beginning of the file.
// If the uncompressed size is known, the offset must be within the file.
// If an offset outside the file is requested io.ErrUnexpectedEOF is returned.
// If the offset is negative, it is interpreted as the distance from the end of the file,
// where -1 represents the last byte.
// If offset from the end of the file is requested, but size is unknown,
// ErrUnsupported will be returned.
func (i *Index) Find(offset int64) (compressedOff, uncompressedOff int64, err error) {
if i.TotalUncompressed < 0 {
return 0, 0, ErrCorrupt
}
if offset < 0 {
offset = i.TotalUncompressed + offset
if offset < 0 {
return 0, 0, io.ErrUnexpectedEOF
}
}
if offset > i.TotalUncompressed {
return 0, 0, io.ErrUnexpectedEOF
}
if len(i.info) > 200 {
n := sort.Search(len(i.info), func(n int) bool {
return i.info[n].uncompressedOffset > offset
})
if n == 0 {
n = 1
}
return i.info[n-1].compressedOffset, i.info[n-1].uncompressedOffset, nil
}
for _, info := range i.info {
if info.uncompressedOffset > offset {
break
}
compressedOff = info.compressedOffset
uncompressedOff = info.uncompressedOffset
}
return compressedOff, uncompressedOff, nil
}
// reduce to stay below maxIndexEntries
func (i *Index) reduce() {
if len(i.info) < maxIndexEntries && i.estBlockUncomp >= minIndexDist {
return
}
// Algorithm, keep 1, remove removeN entries...
removeN := (len(i.info) + 1) / maxIndexEntries
src := i.info
j := 0
// Each block should be at least 1MB, but don't reduce below 1000 entries.
for i.estBlockUncomp*(int64(removeN)+1) < minIndexDist && len(i.info)/(removeN+1) > 1000 {
removeN++
}
for idx := 0; idx < len(src); idx++ {
i.info[j] = src[idx]
j++
idx += removeN
}
i.info = i.info[:j]
// Update maxblock estimate.
i.estBlockUncomp += i.estBlockUncomp * int64(removeN)
}
func (i *Index) appendTo(b []byte, uncompTotal, compTotal int64) []byte {
i.reduce()
var tmp [binary.MaxVarintLen64]byte
initSize := len(b)
// We make the start a skippable header+size.
b = append(b, ChunkTypeIndex, 0, 0, 0)
b = append(b, []byte(S2IndexHeader)...)
// Total Uncompressed size
n := binary.PutVarint(tmp[:], uncompTotal)
b = append(b, tmp[:n]...)
// Total Compressed size
n = binary.PutVarint(tmp[:], compTotal)
b = append(b, tmp[:n]...)
// Put EstBlockUncomp size
n = binary.PutVarint(tmp[:], i.estBlockUncomp)
b = append(b, tmp[:n]...)
// Put length
n = binary.PutVarint(tmp[:], int64(len(i.info)))
b = append(b, tmp[:n]...)
// Check if we should add uncompressed offsets
var hasUncompressed byte
for idx, info := range i.info {
if idx == 0 {
if info.uncompressedOffset != 0 {
hasUncompressed = 1
break
}
continue
}
if info.uncompressedOffset != i.info[idx-1].uncompressedOffset+i.estBlockUncomp {
hasUncompressed = 1
break
}
}
b = append(b, hasUncompressed)
// Add each entry
if hasUncompressed == 1 {
for idx, info := range i.info {
uOff := info.uncompressedOffset
if idx > 0 {
prev := i.info[idx-1]
uOff -= prev.uncompressedOffset + (i.estBlockUncomp)
}
n = binary.PutVarint(tmp[:], uOff)
b = append(b, tmp[:n]...)
}
}
// Initial compressed size estimate.
cPredict := i.estBlockUncomp / 2
for idx, info := range i.info {
cOff := info.compressedOffset
if idx > 0 {
prev := i.info[idx-1]
cOff -= prev.compressedOffset + cPredict
// Update compressed size prediction, with half the error.
cPredict += cOff / 2
}
n = binary.PutVarint(tmp[:], cOff)
b = append(b, tmp[:n]...)
}
// Add Total Size.
// Stored as fixed size for easier reading.
binary.LittleEndian.PutUint32(tmp[:], uint32(len(b)-initSize+4+len(S2IndexTrailer)))
b = append(b, tmp[:4]...)
// Trailer
b = append(b, []byte(S2IndexTrailer)...)
// Update size
chunkLen := len(b) - initSize - skippableFrameHeader
b[initSize+1] = uint8(chunkLen >> 0)
b[initSize+2] = uint8(chunkLen >> 8)
b[initSize+3] = uint8(chunkLen >> 16)
//fmt.Printf("chunklen: 0x%x Uncomp:%d, Comp:%d\n", chunkLen, uncompTotal, compTotal)
return b
}
// Load a binary index.
// A zero value Index can be used or a previous one can be reused.
func (i *Index) Load(b []byte) ([]byte, error) {
if len(b) <= 4+len(S2IndexHeader)+len(S2IndexTrailer) {
return b, io.ErrUnexpectedEOF
}
if b[0] != ChunkTypeIndex {
return b, ErrCorrupt
}
chunkLen := int(b[1]) | int(b[2])<<8 | int(b[3])<<16
b = b[4:]
// Validate we have enough...
if len(b) < chunkLen {
return b, io.ErrUnexpectedEOF
}
if !bytes.Equal(b[:len(S2IndexHeader)], []byte(S2IndexHeader)) {
return b, ErrUnsupported
}
b = b[len(S2IndexHeader):]
// Total Uncompressed
if v, n := binary.Varint(b); n <= 0 || v < 0 {
return b, ErrCorrupt
} else {
i.TotalUncompressed = v
b = b[n:]
}
// Total Compressed
if v, n := binary.Varint(b); n <= 0 {
return b, ErrCorrupt
} else {
i.TotalCompressed = v
b = b[n:]
}
// Read EstBlockUncomp
if v, n := binary.Varint(b); n <= 0 {
return b, ErrCorrupt
} else {
if v < 0 {
return b, ErrCorrupt
}
i.estBlockUncomp = v
b = b[n:]
}
var entries int
if v, n := binary.Varint(b); n <= 0 {
return b, ErrCorrupt
} else {
if v < 0 || v > maxIndexEntries {
return b, ErrCorrupt
}
entries = int(v)
b = b[n:]
}
if cap(i.info) < entries {
i.allocInfos(entries)
}
i.info = i.info[:entries]
if len(b) < 1 {
return b, io.ErrUnexpectedEOF
}
hasUncompressed := b[0]
b = b[1:]
if hasUncompressed&1 != hasUncompressed {
return b, ErrCorrupt
}
// Add each uncompressed entry
for idx := range i.info {
var uOff int64
if hasUncompressed != 0 {
// Load delta
if v, n := binary.Varint(b); n <= 0 {
return b, ErrCorrupt
} else {
uOff = v
b = b[n:]
}
}
if idx > 0 {
prev := i.info[idx-1].uncompressedOffset
uOff += prev + (i.estBlockUncomp)
if uOff <= prev {
return b, ErrCorrupt
}
}
if uOff < 0 {
return b, ErrCorrupt
}
i.info[idx].uncompressedOffset = uOff
}
// Initial compressed size estimate.
cPredict := i.estBlockUncomp / 2
// Add each compressed entry
for idx := range i.info {
var cOff int64
if v, n := binary.Varint(b); n <= 0 {
return b, ErrCorrupt
} else {
cOff = v
b = b[n:]
}
if idx > 0 {
// Update compressed size prediction, with half the error.
cPredictNew := cPredict + cOff/2
prev := i.info[idx-1].compressedOffset
cOff += prev + cPredict
if cOff <= prev {
return b, ErrCorrupt
}
cPredict = cPredictNew
}
if cOff < 0 {
return b, ErrCorrupt
}
i.info[idx].compressedOffset = cOff
}
if len(b) < 4+len(S2IndexTrailer) {
return b, io.ErrUnexpectedEOF
}
// Skip size...
b = b[4:]
// Check trailer...
if !bytes.Equal(b[:len(S2IndexTrailer)], []byte(S2IndexTrailer)) {
return b, ErrCorrupt
}
return b[len(S2IndexTrailer):], nil
}
// LoadStream will load an index from the end of the supplied stream.
// ErrUnsupported will be returned if the signature cannot be found.
// ErrCorrupt will be returned if unexpected values are found.
// io.ErrUnexpectedEOF is returned if there are too few bytes.
// IO errors are returned as-is.
func (i *Index) LoadStream(rs io.ReadSeeker) error {
// Go to end.
_, err := rs.Seek(-10, io.SeekEnd)
if err != nil {
return err
}
var tmp [10]byte
_, err = io.ReadFull(rs, tmp[:])
if err != nil {
return err
}
// Check trailer...
if !bytes.Equal(tmp[4:4+len(S2IndexTrailer)], []byte(S2IndexTrailer)) {
return ErrUnsupported
}
sz := binary.LittleEndian.Uint32(tmp[:4])
if sz > maxChunkSize+skippableFrameHeader {
return ErrCorrupt
}
_, err = rs.Seek(-int64(sz), io.SeekEnd)
if err != nil {
return err
}
// Read index.
buf := make([]byte, sz)
_, err = io.ReadFull(rs, buf)
if err != nil {
return err
}
_, err = i.Load(buf)
return err
}
// IndexStream will return an index for a stream.
// The stream structure will be checked, but
// data within blocks is not verified.
// The returned index can either be appended to the end of the stream
// or stored separately.
func IndexStream(r io.Reader) ([]byte, error) {
var i Index
var buf [maxChunkSize]byte
var readHeader bool
for {
_, err := io.ReadFull(r, buf[:4])
if err != nil {
if err == io.EOF {
return i.appendTo(nil, i.TotalUncompressed, i.TotalCompressed), nil
}
return nil, err
}
// Start of this chunk.
startChunk := i.TotalCompressed
i.TotalCompressed += 4
chunkType := buf[0]
if !readHeader {
if chunkType != chunkTypeStreamIdentifier {
return nil, ErrCorrupt
}
readHeader = true
}
chunkLen := int(buf[1]) | int(buf[2])<<8 | int(buf[3])<<16
if chunkLen < checksumSize {
return nil, ErrCorrupt
}
i.TotalCompressed += int64(chunkLen)
_, err = io.ReadFull(r, buf[:chunkLen])
if err != nil {
return nil, io.ErrUnexpectedEOF
}
// The chunk types are specified at
// https://github.com/google/snappy/blob/master/framing_format.txt
switch chunkType {
case chunkTypeCompressedData:
// Section 4.2. Compressed data (chunk type 0x00).
// Skip checksum.
dLen, err := DecodedLen(buf[checksumSize:])
if err != nil {
return nil, err
}
if dLen > maxBlockSize {
return nil, ErrCorrupt
}
if i.estBlockUncomp == 0 {
// Use first block for estimate...
i.estBlockUncomp = int64(dLen)
}
err = i.add(startChunk, i.TotalUncompressed)
if err != nil {
return nil, err
}
i.TotalUncompressed += int64(dLen)
continue
case chunkTypeUncompressedData:
n2 := chunkLen - checksumSize
if n2 > maxBlockSize {
return nil, ErrCorrupt
}
if i.estBlockUncomp == 0 {
// Use first block for estimate...
i.estBlockUncomp = int64(n2)
}
err = i.add(startChunk, i.TotalUncompressed)
if err != nil {
return nil, err
}
i.TotalUncompressed += int64(n2)
continue
case chunkTypeStreamIdentifier:
// Section 4.1. Stream identifier (chunk type 0xff).
if chunkLen != len(magicBody) {
return nil, ErrCorrupt
}
if string(buf[:len(magicBody)]) != magicBody {
if string(buf[:len(magicBody)]) != magicBodySnappy {
return nil, ErrCorrupt
}
}
continue
}
if chunkType <= 0x7f {
// Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f).
return nil, ErrUnsupported
}
if chunkLen > maxChunkSize {
return nil, ErrUnsupported
}
// Section 4.4 Padding (chunk type 0xfe).
// Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd).
}
}
// JSON returns the index as JSON text.
func (i *Index) JSON() []byte {
type offset struct {
CompressedOffset int64 `json:"compressed"`
UncompressedOffset int64 `json:"uncompressed"`
}
x := struct {
TotalUncompressed int64 `json:"total_uncompressed"` // Total Uncompressed size if known. Will be -1 if unknown.
TotalCompressed int64 `json:"total_compressed"` // Total Compressed size if known. Will be -1 if unknown.
Offsets []offset `json:"offsets"`
EstBlockUncomp int64 `json:"est_block_uncompressed"`
}{
TotalUncompressed: i.TotalUncompressed,
TotalCompressed: i.TotalCompressed,
EstBlockUncomp: i.estBlockUncomp,
}
for _, v := range i.info {
x.Offsets = append(x.Offsets, offset{CompressedOffset: v.compressedOffset, UncompressedOffset: v.uncompressedOffset})
}
b, _ := json.MarshalIndent(x, "", " ")
return b
}
// RemoveIndexHeaders will trim all headers and trailers from a given index.
// This is expected to save 20 bytes.
// These can be restored using RestoreIndexHeaders.
// This removes a layer of security, but is the most compact representation.
// Returns nil if headers contains errors.
// The returned slice references the provided slice.
func RemoveIndexHeaders(b []byte) []byte {
const save = 4 + len(S2IndexHeader) + len(S2IndexTrailer) + 4
if len(b) <= save {
return nil
}
if b[0] != ChunkTypeIndex {
return nil
}
chunkLen := int(b[1]) | int(b[2])<<8 | int(b[3])<<16
b = b[4:]
// Validate we have enough...
if len(b) < chunkLen {
return nil
}
b = b[:chunkLen]
if !bytes.Equal(b[:len(S2IndexHeader)], []byte(S2IndexHeader)) {
return nil
}
b = b[len(S2IndexHeader):]
if !bytes.HasSuffix(b, []byte(S2IndexTrailer)) {
return nil
}
b = bytes.TrimSuffix(b, []byte(S2IndexTrailer))
if len(b) < 4 {
return nil
}
return b[:len(b)-4]
}
// RestoreIndexHeaders will index restore headers removed by RemoveIndexHeaders.
// No error checking is performed on the input.
// If a 0 length slice is sent, it is returned without modification.
func RestoreIndexHeaders(in []byte) []byte {
if len(in) == 0 {
return in
}
b := make([]byte, 0, 4+len(S2IndexHeader)+len(in)+len(S2IndexTrailer)+4)
b = append(b, ChunkTypeIndex, 0, 0, 0)
b = append(b, []byte(S2IndexHeader)...)
b = append(b, in...)
var tmp [4]byte
binary.LittleEndian.PutUint32(tmp[:], uint32(len(b)+4+len(S2IndexTrailer)))
b = append(b, tmp[:4]...)
// Trailer
b = append(b, []byte(S2IndexTrailer)...)
chunkLen := len(b) - skippableFrameHeader
b[1] = uint8(chunkLen >> 0)
b[2] = uint8(chunkLen >> 8)
b[3] = uint8(chunkLen >> 16)
return b
}

585
vendor/github.com/klauspost/compress/s2/lz4convert.go generated vendored Normal file
View File

@@ -0,0 +1,585 @@
// Copyright (c) 2022 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"encoding/binary"
"errors"
"fmt"
)
// LZ4Converter provides conversion from LZ4 blocks as defined here:
// https://github.com/lz4/lz4/blob/dev/doc/lz4_Block_format.md
type LZ4Converter struct {
}
// ErrDstTooSmall is returned when provided destination is too small.
var ErrDstTooSmall = errors.New("s2: destination too small")
// ConvertBlock will convert an LZ4 block and append it as an S2
// block without block length to dst.
// The uncompressed size is returned as well.
// dst must have capacity to contain the entire compressed block.
func (l *LZ4Converter) ConvertBlock(dst, src []byte) ([]byte, int, error) {
if len(src) == 0 {
return dst, 0, nil
}
const debug = false
const inline = true
const lz4MinMatch = 4
s, d := 0, len(dst)
dst = dst[:cap(dst)]
if !debug && hasAmd64Asm {
res, sz := cvtLZ4BlockAsm(dst[d:], src)
if res < 0 {
const (
errCorrupt = -1
errDstTooSmall = -2
)
switch res {
case errCorrupt:
return nil, 0, ErrCorrupt
case errDstTooSmall:
return nil, 0, ErrDstTooSmall
default:
return nil, 0, fmt.Errorf("unexpected result: %d", res)
}
}
if d+sz > len(dst) {
return nil, 0, ErrDstTooSmall
}
return dst[:d+sz], res, nil
}
dLimit := len(dst) - 10
var lastOffset uint16
var uncompressed int
if debug {
fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst))
}
for {
if s >= len(src) {
return dst[:d], 0, ErrCorrupt
}
// Read literal info
token := src[s]
ll := int(token >> 4)
ml := int(lz4MinMatch + (token & 0xf))
// If upper nibble is 15, literal length is extended
if token >= 0xf0 {
for {
s++
if s >= len(src) {
if debug {
fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src))
}
return dst[:d], 0, ErrCorrupt
}
val := src[s]
ll += int(val)
if val != 255 {
break
}
}
}
// Skip past token
if s+ll >= len(src) {
if debug {
fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src))
}
return nil, 0, ErrCorrupt
}
s++
if ll > 0 {
if d+ll > dLimit {
return nil, 0, ErrDstTooSmall
}
if debug {
fmt.Printf("emit %d literals\n", ll)
}
d += emitLiteralGo(dst[d:], src[s:s+ll])
s += ll
uncompressed += ll
}
// Check if we are done...
if s == len(src) && ml == lz4MinMatch {
break
}
// 2 byte offset
if s >= len(src)-2 {
if debug {
fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2)
}
return nil, 0, ErrCorrupt
}
offset := binary.LittleEndian.Uint16(src[s:])
s += 2
if offset == 0 {
if debug {
fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s)
}
return nil, 0, ErrCorrupt
}
if int(offset) > uncompressed {
if debug {
fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed)
}
return nil, 0, ErrCorrupt
}
if ml == lz4MinMatch+15 {
for {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
s++
ml += int(val)
if val != 255 {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
break
}
}
}
if offset == lastOffset {
if debug {
fmt.Printf("emit repeat, length: %d, offset: %d\n", ml, offset)
}
if !inline {
d += emitRepeat16(dst[d:], offset, ml)
} else {
length := ml
dst := dst[d:]
for len(dst) > 5 {
// Repeat offset, make length cheaper
length -= 4
if length <= 4 {
dst[0] = uint8(length)<<2 | tagCopy1
dst[1] = 0
d += 2
break
}
if length < 8 && offset < 2048 {
// Encode WITH offset
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1
d += 2
break
}
if length < (1<<8)+4 {
length -= 4
dst[2] = uint8(length)
dst[1] = 0
dst[0] = 5<<2 | tagCopy1
d += 3
break
}
if length < (1<<16)+(1<<8) {
length -= 1 << 8
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 6<<2 | tagCopy1
d += 4
break
}
const maxRepeat = (1 << 24) - 1
length -= 1 << 16
left := 0
if length > maxRepeat {
left = length - maxRepeat + 4
length = maxRepeat - 4
}
dst[4] = uint8(length >> 16)
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 7<<2 | tagCopy1
if left > 0 {
d += 5 + emitRepeat16(dst[5:], offset, left)
break
}
d += 5
break
}
}
} else {
if debug {
fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset)
}
if !inline {
d += emitCopy16(dst[d:], offset, ml)
} else {
length := ml
dst := dst[d:]
for len(dst) > 5 {
// Offset no more than 2 bytes.
if length > 64 {
off := 3
if offset < 2048 {
// emit 8 bytes as tagCopy1, rest as repeats.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1
length -= 8
off = 2
} else {
// Emit a length 60 copy, encoded as 3 bytes.
// Emit remaining as repeat value (minimum 4 bytes).
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 59<<2 | tagCopy2
length -= 60
}
// Emit remaining as repeats, at least 4 bytes remain.
d += off + emitRepeat16(dst[off:], offset, length)
break
}
if length >= 12 || offset >= 2048 {
// Emit the remaining copy, encoded as 3 bytes.
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = uint8(length-1)<<2 | tagCopy2
d += 3
break
}
// Emit the remaining copy, encoded as 2 bytes.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
d += 2
break
}
}
lastOffset = offset
}
uncompressed += ml
if d > dLimit {
return nil, 0, ErrDstTooSmall
}
}
return dst[:d], uncompressed, nil
}
// ConvertBlockSnappy will convert an LZ4 block and append it
// as a Snappy block without block length to dst.
// The uncompressed size is returned as well.
// dst must have capacity to contain the entire compressed block.
func (l *LZ4Converter) ConvertBlockSnappy(dst, src []byte) ([]byte, int, error) {
if len(src) == 0 {
return dst, 0, nil
}
const debug = false
const lz4MinMatch = 4
s, d := 0, len(dst)
dst = dst[:cap(dst)]
// Use assembly when possible
if !debug && hasAmd64Asm {
res, sz := cvtLZ4BlockSnappyAsm(dst[d:], src)
if res < 0 {
const (
errCorrupt = -1
errDstTooSmall = -2
)
switch res {
case errCorrupt:
return nil, 0, ErrCorrupt
case errDstTooSmall:
return nil, 0, ErrDstTooSmall
default:
return nil, 0, fmt.Errorf("unexpected result: %d", res)
}
}
if d+sz > len(dst) {
return nil, 0, ErrDstTooSmall
}
return dst[:d+sz], res, nil
}
dLimit := len(dst) - 10
var uncompressed int
if debug {
fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst))
}
for {
if s >= len(src) {
return nil, 0, ErrCorrupt
}
// Read literal info
token := src[s]
ll := int(token >> 4)
ml := int(lz4MinMatch + (token & 0xf))
// If upper nibble is 15, literal length is extended
if token >= 0xf0 {
for {
s++
if s >= len(src) {
if debug {
fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
ll += int(val)
if val != 255 {
break
}
}
}
// Skip past token
if s+ll >= len(src) {
if debug {
fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src))
}
return nil, 0, ErrCorrupt
}
s++
if ll > 0 {
if d+ll > dLimit {
return nil, 0, ErrDstTooSmall
}
if debug {
fmt.Printf("emit %d literals\n", ll)
}
d += emitLiteralGo(dst[d:], src[s:s+ll])
s += ll
uncompressed += ll
}
// Check if we are done...
if s == len(src) && ml == lz4MinMatch {
break
}
// 2 byte offset
if s >= len(src)-2 {
if debug {
fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2)
}
return nil, 0, ErrCorrupt
}
offset := binary.LittleEndian.Uint16(src[s:])
s += 2
if offset == 0 {
if debug {
fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s)
}
return nil, 0, ErrCorrupt
}
if int(offset) > uncompressed {
if debug {
fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed)
}
return nil, 0, ErrCorrupt
}
if ml == lz4MinMatch+15 {
for {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
s++
ml += int(val)
if val != 255 {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
break
}
}
}
if debug {
fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset)
}
length := ml
// d += emitCopyNoRepeat(dst[d:], int(offset), ml)
for length > 0 {
if d >= dLimit {
return nil, 0, ErrDstTooSmall
}
// Offset no more than 2 bytes.
if length > 64 {
// Emit a length 64 copy, encoded as 3 bytes.
dst[d+2] = uint8(offset >> 8)
dst[d+1] = uint8(offset)
dst[d+0] = 63<<2 | tagCopy2
length -= 64
d += 3
continue
}
if length >= 12 || offset >= 2048 || length < 4 {
// Emit the remaining copy, encoded as 3 bytes.
dst[d+2] = uint8(offset >> 8)
dst[d+1] = uint8(offset)
dst[d+0] = uint8(length-1)<<2 | tagCopy2
d += 3
break
}
// Emit the remaining copy, encoded as 2 bytes.
dst[d+1] = uint8(offset)
dst[d+0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
d += 2
break
}
uncompressed += ml
if d > dLimit {
return nil, 0, ErrDstTooSmall
}
}
return dst[:d], uncompressed, nil
}
// emitRepeat writes a repeat chunk and returns the number of bytes written.
// Length must be at least 4 and < 1<<24
func emitRepeat16(dst []byte, offset uint16, length int) int {
// Repeat offset, make length cheaper
length -= 4
if length <= 4 {
dst[0] = uint8(length)<<2 | tagCopy1
dst[1] = 0
return 2
}
if length < 8 && offset < 2048 {
// Encode WITH offset
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1
return 2
}
if length < (1<<8)+4 {
length -= 4
dst[2] = uint8(length)
dst[1] = 0
dst[0] = 5<<2 | tagCopy1
return 3
}
if length < (1<<16)+(1<<8) {
length -= 1 << 8
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 6<<2 | tagCopy1
return 4
}
const maxRepeat = (1 << 24) - 1
length -= 1 << 16
left := 0
if length > maxRepeat {
left = length - maxRepeat + 4
length = maxRepeat - 4
}
dst[4] = uint8(length >> 16)
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 7<<2 | tagCopy1
if left > 0 {
return 5 + emitRepeat16(dst[5:], offset, left)
}
return 5
}
// emitCopy writes a copy chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 1 <= offset && offset <= math.MaxUint16
// 4 <= length && length <= math.MaxUint32
func emitCopy16(dst []byte, offset uint16, length int) int {
// Offset no more than 2 bytes.
if length > 64 {
off := 3
if offset < 2048 {
// emit 8 bytes as tagCopy1, rest as repeats.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1
length -= 8
off = 2
} else {
// Emit a length 60 copy, encoded as 3 bytes.
// Emit remaining as repeat value (minimum 4 bytes).
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 59<<2 | tagCopy2
length -= 60
}
// Emit remaining as repeats, at least 4 bytes remain.
return off + emitRepeat16(dst[off:], offset, length)
}
if length >= 12 || offset >= 2048 {
// Emit the remaining copy, encoded as 3 bytes.
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = uint8(length-1)<<2 | tagCopy2
return 3
}
// Emit the remaining copy, encoded as 2 bytes.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
return 2
}
// emitLiteral writes a literal chunk and returns the number of bytes written.
//
// It assumes that:
//
// dst is long enough to hold the encoded bytes
// 0 <= len(lit) && len(lit) <= math.MaxUint32
func emitLiteralGo(dst, lit []byte) int {
if len(lit) == 0 {
return 0
}
i, n := 0, uint(len(lit)-1)
switch {
case n < 60:
dst[0] = uint8(n)<<2 | tagLiteral
i = 1
case n < 1<<8:
dst[1] = uint8(n)
dst[0] = 60<<2 | tagLiteral
i = 2
case n < 1<<16:
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 61<<2 | tagLiteral
i = 3
case n < 1<<24:
dst[3] = uint8(n >> 16)
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 62<<2 | tagLiteral
i = 4
default:
dst[4] = uint8(n >> 24)
dst[3] = uint8(n >> 16)
dst[2] = uint8(n >> 8)
dst[1] = uint8(n)
dst[0] = 63<<2 | tagLiteral
i = 5
}
return i + copy(dst[i:], lit)
}

467
vendor/github.com/klauspost/compress/s2/lz4sconvert.go generated vendored Normal file
View File

@@ -0,0 +1,467 @@
// Copyright (c) 2022 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package s2
import (
"encoding/binary"
"fmt"
)
// LZ4sConverter provides conversion from LZ4s.
// (Intel modified LZ4 Blocks)
// https://cdrdv2-public.intel.com/743912/743912-qat-programmers-guide-v2.0.pdf
// LZ4s is a variant of LZ4 block format. LZ4s should be considered as an intermediate compressed block format.
// The LZ4s format is selected when the application sets the compType to CPA_DC_LZ4S in CpaDcSessionSetupData.
// The LZ4s block returned by the Intel® QAT hardware can be used by an external
// software post-processing to generate other compressed data formats.
// The following table lists the differences between LZ4 and LZ4s block format. LZ4s block format uses
// the same high-level formatting as LZ4 block format with the following encoding changes:
// For Min Match of 4 bytes, Copy length value 1-15 means length 4-18 with 18 bytes adding an extra byte.
// ONLY "Min match of 4 bytes" is supported.
type LZ4sConverter struct {
}
// ConvertBlock will convert an LZ4s block and append it as an S2
// block without block length to dst.
// The uncompressed size is returned as well.
// dst must have capacity to contain the entire compressed block.
func (l *LZ4sConverter) ConvertBlock(dst, src []byte) ([]byte, int, error) {
if len(src) == 0 {
return dst, 0, nil
}
const debug = false
const inline = true
const lz4MinMatch = 3
s, d := 0, len(dst)
dst = dst[:cap(dst)]
if !debug && hasAmd64Asm {
res, sz := cvtLZ4sBlockAsm(dst[d:], src)
if res < 0 {
const (
errCorrupt = -1
errDstTooSmall = -2
)
switch res {
case errCorrupt:
return nil, 0, ErrCorrupt
case errDstTooSmall:
return nil, 0, ErrDstTooSmall
default:
return nil, 0, fmt.Errorf("unexpected result: %d", res)
}
}
if d+sz > len(dst) {
return nil, 0, ErrDstTooSmall
}
return dst[:d+sz], res, nil
}
dLimit := len(dst) - 10
var lastOffset uint16
var uncompressed int
if debug {
fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst))
}
for {
if s >= len(src) {
return dst[:d], 0, ErrCorrupt
}
// Read literal info
token := src[s]
ll := int(token >> 4)
ml := int(lz4MinMatch + (token & 0xf))
// If upper nibble is 15, literal length is extended
if token >= 0xf0 {
for {
s++
if s >= len(src) {
if debug {
fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src))
}
return dst[:d], 0, ErrCorrupt
}
val := src[s]
ll += int(val)
if val != 255 {
break
}
}
}
// Skip past token
if s+ll >= len(src) {
if debug {
fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src))
}
return nil, 0, ErrCorrupt
}
s++
if ll > 0 {
if d+ll > dLimit {
return nil, 0, ErrDstTooSmall
}
if debug {
fmt.Printf("emit %d literals\n", ll)
}
d += emitLiteralGo(dst[d:], src[s:s+ll])
s += ll
uncompressed += ll
}
// Check if we are done...
if ml == lz4MinMatch {
if s == len(src) {
break
}
// 0 bytes.
continue
}
// 2 byte offset
if s >= len(src)-2 {
if debug {
fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2)
}
return nil, 0, ErrCorrupt
}
offset := binary.LittleEndian.Uint16(src[s:])
s += 2
if offset == 0 {
if debug {
fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s)
}
return nil, 0, ErrCorrupt
}
if int(offset) > uncompressed {
if debug {
fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed)
}
return nil, 0, ErrCorrupt
}
if ml == lz4MinMatch+15 {
for {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
s++
ml += int(val)
if val != 255 {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
break
}
}
}
if offset == lastOffset {
if debug {
fmt.Printf("emit repeat, length: %d, offset: %d\n", ml, offset)
}
if !inline {
d += emitRepeat16(dst[d:], offset, ml)
} else {
length := ml
dst := dst[d:]
for len(dst) > 5 {
// Repeat offset, make length cheaper
length -= 4
if length <= 4 {
dst[0] = uint8(length)<<2 | tagCopy1
dst[1] = 0
d += 2
break
}
if length < 8 && offset < 2048 {
// Encode WITH offset
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1
d += 2
break
}
if length < (1<<8)+4 {
length -= 4
dst[2] = uint8(length)
dst[1] = 0
dst[0] = 5<<2 | tagCopy1
d += 3
break
}
if length < (1<<16)+(1<<8) {
length -= 1 << 8
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 6<<2 | tagCopy1
d += 4
break
}
const maxRepeat = (1 << 24) - 1
length -= 1 << 16
left := 0
if length > maxRepeat {
left = length - maxRepeat + 4
length = maxRepeat - 4
}
dst[4] = uint8(length >> 16)
dst[3] = uint8(length >> 8)
dst[2] = uint8(length >> 0)
dst[1] = 0
dst[0] = 7<<2 | tagCopy1
if left > 0 {
d += 5 + emitRepeat16(dst[5:], offset, left)
break
}
d += 5
break
}
}
} else {
if debug {
fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset)
}
if !inline {
d += emitCopy16(dst[d:], offset, ml)
} else {
length := ml
dst := dst[d:]
for len(dst) > 5 {
// Offset no more than 2 bytes.
if length > 64 {
off := 3
if offset < 2048 {
// emit 8 bytes as tagCopy1, rest as repeats.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1
length -= 8
off = 2
} else {
// Emit a length 60 copy, encoded as 3 bytes.
// Emit remaining as repeat value (minimum 4 bytes).
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = 59<<2 | tagCopy2
length -= 60
}
// Emit remaining as repeats, at least 4 bytes remain.
d += off + emitRepeat16(dst[off:], offset, length)
break
}
if length >= 12 || offset >= 2048 {
// Emit the remaining copy, encoded as 3 bytes.
dst[2] = uint8(offset >> 8)
dst[1] = uint8(offset)
dst[0] = uint8(length-1)<<2 | tagCopy2
d += 3
break
}
// Emit the remaining copy, encoded as 2 bytes.
dst[1] = uint8(offset)
dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
d += 2
break
}
}
lastOffset = offset
}
uncompressed += ml
if d > dLimit {
return nil, 0, ErrDstTooSmall
}
}
return dst[:d], uncompressed, nil
}
// ConvertBlockSnappy will convert an LZ4s block and append it
// as a Snappy block without block length to dst.
// The uncompressed size is returned as well.
// dst must have capacity to contain the entire compressed block.
func (l *LZ4sConverter) ConvertBlockSnappy(dst, src []byte) ([]byte, int, error) {
if len(src) == 0 {
return dst, 0, nil
}
const debug = false
const lz4MinMatch = 3
s, d := 0, len(dst)
dst = dst[:cap(dst)]
// Use assembly when possible
if !debug && hasAmd64Asm {
res, sz := cvtLZ4sBlockSnappyAsm(dst[d:], src)
if res < 0 {
const (
errCorrupt = -1
errDstTooSmall = -2
)
switch res {
case errCorrupt:
return nil, 0, ErrCorrupt
case errDstTooSmall:
return nil, 0, ErrDstTooSmall
default:
return nil, 0, fmt.Errorf("unexpected result: %d", res)
}
}
if d+sz > len(dst) {
return nil, 0, ErrDstTooSmall
}
return dst[:d+sz], res, nil
}
dLimit := len(dst) - 10
var uncompressed int
if debug {
fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst))
}
for {
if s >= len(src) {
return nil, 0, ErrCorrupt
}
// Read literal info
token := src[s]
ll := int(token >> 4)
ml := int(lz4MinMatch + (token & 0xf))
// If upper nibble is 15, literal length is extended
if token >= 0xf0 {
for {
s++
if s >= len(src) {
if debug {
fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
ll += int(val)
if val != 255 {
break
}
}
}
// Skip past token
if s+ll >= len(src) {
if debug {
fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src))
}
return nil, 0, ErrCorrupt
}
s++
if ll > 0 {
if d+ll > dLimit {
return nil, 0, ErrDstTooSmall
}
if debug {
fmt.Printf("emit %d literals\n", ll)
}
d += emitLiteralGo(dst[d:], src[s:s+ll])
s += ll
uncompressed += ll
}
// Check if we are done...
if ml == lz4MinMatch {
if s == len(src) {
break
}
// 0 bytes.
continue
}
// 2 byte offset
if s >= len(src)-2 {
if debug {
fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2)
}
return nil, 0, ErrCorrupt
}
offset := binary.LittleEndian.Uint16(src[s:])
s += 2
if offset == 0 {
if debug {
fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s)
}
return nil, 0, ErrCorrupt
}
if int(offset) > uncompressed {
if debug {
fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed)
}
return nil, 0, ErrCorrupt
}
if ml == lz4MinMatch+15 {
for {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
val := src[s]
s++
ml += int(val)
if val != 255 {
if s >= len(src) {
if debug {
fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src))
}
return nil, 0, ErrCorrupt
}
break
}
}
}
if debug {
fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset)
}
length := ml
// d += emitCopyNoRepeat(dst[d:], int(offset), ml)
for length > 0 {
if d >= dLimit {
return nil, 0, ErrDstTooSmall
}
// Offset no more than 2 bytes.
if length > 64 {
// Emit a length 64 copy, encoded as 3 bytes.
dst[d+2] = uint8(offset >> 8)
dst[d+1] = uint8(offset)
dst[d+0] = 63<<2 | tagCopy2
length -= 64
d += 3
continue
}
if length >= 12 || offset >= 2048 || length < 4 {
// Emit the remaining copy, encoded as 3 bytes.
dst[d+2] = uint8(offset >> 8)
dst[d+1] = uint8(offset)
dst[d+0] = uint8(length-1)<<2 | tagCopy2
d += 3
break
}
// Emit the remaining copy, encoded as 2 bytes.
dst[d+1] = uint8(offset)
dst[d+0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1
d += 2
break
}
uncompressed += ml
if d > dLimit {
return nil, 0, ErrDstTooSmall
}
}
return dst[:d], uncompressed, nil
}

1075
vendor/github.com/klauspost/compress/s2/reader.go generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff

151
vendor/github.com/klauspost/compress/s2/s2.go generated vendored Normal file
View File

@@ -0,0 +1,151 @@
// Copyright 2011 The Snappy-Go Authors. All rights reserved.
// Copyright (c) 2019 Klaus Post. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package s2 implements the S2 compression format.
//
// S2 is an extension of Snappy. Similar to Snappy S2 is aimed for high throughput,
// which is why it features concurrent compression for bigger payloads.
//
// Decoding is compatible with Snappy compressed content,
// but content compressed with S2 cannot be decompressed by Snappy.
//
// For more information on Snappy/S2 differences see README in: https://github.com/klauspost/compress/tree/master/s2
//
// There are actually two S2 formats: block and stream. They are related,
// but different: trying to decompress block-compressed data as a S2 stream
// will fail, and vice versa. The block format is the Decode and Encode
// functions and the stream format is the Reader and Writer types.
//
// A "better" compression option is available. This will trade some compression
// speed
//
// The block format, the more common case, is used when the complete size (the
// number of bytes) of the original data is known upfront, at the time
// compression starts. The stream format, also known as the framing format, is
// for when that isn't always true.
//
// Blocks to not offer much data protection, so it is up to you to
// add data validation of decompressed blocks.
//
// Streams perform CRC validation of the decompressed data.
// Stream compression will also be performed on multiple CPU cores concurrently
// significantly improving throughput.
package s2
import (
"bytes"
"hash/crc32"
"github.com/klauspost/compress/internal/race"
)
/*
Each encoded block begins with the varint-encoded length of the decoded data,
followed by a sequence of chunks. Chunks begin and end on byte boundaries. The
first byte of each chunk is broken into its 2 least and 6 most significant bits
called l and m: l ranges in [0, 4) and m ranges in [0, 64). l is the chunk tag.
Zero means a literal tag. All other values mean a copy tag.
For literal tags:
- If m < 60, the next 1 + m bytes are literal bytes.
- Otherwise, let n be the little-endian unsigned integer denoted by the next
m - 59 bytes. The next 1 + n bytes after that are literal bytes.
For copy tags, length bytes are copied from offset bytes ago, in the style of
Lempel-Ziv compression algorithms. In particular:
- For l == 1, the offset ranges in [0, 1<<11) and the length in [4, 12).
The length is 4 + the low 3 bits of m. The high 3 bits of m form bits 8-10
of the offset. The next byte is bits 0-7 of the offset.
- For l == 2, the offset ranges in [0, 1<<16) and the length in [1, 65).
The length is 1 + m. The offset is the little-endian unsigned integer
denoted by the next 2 bytes.
- For l == 3, the offset ranges in [0, 1<<32) and the length in
[1, 65). The length is 1 + m. The offset is the little-endian unsigned
integer denoted by the next 4 bytes.
*/
const (
tagLiteral = 0x00
tagCopy1 = 0x01
tagCopy2 = 0x02
tagCopy4 = 0x03
)
const (
checksumSize = 4
chunkHeaderSize = 4
magicChunk = "\xff\x06\x00\x00" + magicBody
magicChunkSnappy = "\xff\x06\x00\x00" + magicBodySnappy
magicBodySnappy = "sNaPpY"
magicBody = "S2sTwO"
// maxBlockSize is the maximum size of the input to encodeBlock.
//
// For the framing format (Writer type instead of Encode function),
// this is the maximum uncompressed size of a block.
maxBlockSize = 4 << 20
// minBlockSize is the minimum size of block setting when creating a writer.
minBlockSize = 4 << 10
skippableFrameHeader = 4
maxChunkSize = 1<<24 - 1 // 16777215
// Default block size
defaultBlockSize = 1 << 20
// maxSnappyBlockSize is the maximum snappy block size.
maxSnappyBlockSize = 1 << 16
obufHeaderLen = checksumSize + chunkHeaderSize
)
const (
chunkTypeCompressedData = 0x00
chunkTypeUncompressedData = 0x01
ChunkTypeIndex = 0x99
chunkTypePadding = 0xfe
chunkTypeStreamIdentifier = 0xff
)
var (
crcTable = crc32.MakeTable(crc32.Castagnoli)
magicChunkSnappyBytes = []byte(magicChunkSnappy) // Can be passed to functions where it escapes.
magicChunkBytes = []byte(magicChunk) // Can be passed to functions where it escapes.
)
// crc implements the checksum specified in section 3 of
// https://github.com/google/snappy/blob/master/framing_format.txt
func crc(b []byte) uint32 {
race.ReadSlice(b)
c := crc32.Update(0, crcTable, b)
return c>>15 | c<<17 + 0xa282ead8
}
// literalExtraSize returns the extra size of encoding n literals.
// n should be >= 0 and <= math.MaxUint32.
func literalExtraSize(n int64) int64 {
if n == 0 {
return 0
}
switch {
case n < 60:
return 1
case n < 1<<8:
return 2
case n < 1<<16:
return 3
case n < 1<<24:
return 4
default:
return 5
}
}
type byter interface {
Bytes() []byte
}
var _ byter = &bytes.Buffer{}

1039
vendor/github.com/klauspost/compress/s2/writer.go generated vendored Normal file
View File

File diff suppressed because it is too large Load Diff