Move modules/gzip to gitea.com/macaron/gzip (#9058)

* Move modules/gzip to gitea.com/macaron/gzip

* Fix vendor
pull/9059/head
Lunny Xiao 4 years ago committed by GitHub
parent ba4e8f221b
commit 9ff6312627
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
  1. 5
      go.mod
  2. 10
      go.sum
  3. 2
      integrations/lfs_getobject_test.go
  4. 131
      modules/gzip/gzip_test.go
  5. 2
      routers/routes/routes.go
  6. 9
      vendor/gitea.com/macaron/gzip/go.mod
  7. 42
      vendor/gitea.com/macaron/gzip/go.sum
  8. 0
      vendor/gitea.com/macaron/gzip/gzip.go
  9. 1
      vendor/github.com/klauspost/compress/LICENSE
  10. 32
      vendor/github.com/klauspost/compress/flate/copy.go
  11. 41
      vendor/github.com/klauspost/compress/flate/crc32_amd64.go
  12. 213
      vendor/github.com/klauspost/compress/flate/crc32_amd64.s
  13. 35
      vendor/github.com/klauspost/compress/flate/crc32_noasm.go
  14. 895
      vendor/github.com/klauspost/compress/flate/deflate.go
  15. 257
      vendor/github.com/klauspost/compress/flate/fast_encoder.go
  16. 532
      vendor/github.com/klauspost/compress/flate/huffman_bit_writer.go
  17. 72
      vendor/github.com/klauspost/compress/flate/huffman_code.go
  18. 169
      vendor/github.com/klauspost/compress/flate/inflate.go
  19. 174
      vendor/github.com/klauspost/compress/flate/level1.go
  20. 199
      vendor/github.com/klauspost/compress/flate/level2.go
  21. 225
      vendor/github.com/klauspost/compress/flate/level3.go
  22. 210
      vendor/github.com/klauspost/compress/flate/level4.go
  23. 276
      vendor/github.com/klauspost/compress/flate/level5.go
  24. 279
      vendor/github.com/klauspost/compress/flate/level6.go
  25. 48
      vendor/github.com/klauspost/compress/flate/reverse_bits.go
  26. 856
      vendor/github.com/klauspost/compress/flate/snappy.go
  27. 252
      vendor/github.com/klauspost/compress/flate/stateless.go
  28. 298
      vendor/github.com/klauspost/compress/flate/token.go
  29. 2
      vendor/github.com/klauspost/compress/gzip/gunzip.go
  30. 32
      vendor/github.com/klauspost/compress/gzip/gzip.go
  31. 24
      vendor/github.com/klauspost/cpuid/.gitignore
  32. 8
      vendor/github.com/klauspost/cpuid/.travis.yml
  33. 22
      vendor/github.com/klauspost/cpuid/LICENSE
  34. 145
      vendor/github.com/klauspost/cpuid/README.md
  35. 1022
      vendor/github.com/klauspost/cpuid/cpuid.go
  36. 42
      vendor/github.com/klauspost/cpuid/cpuid_386.s
  37. 42
      vendor/github.com/klauspost/cpuid/cpuid_amd64.s
  38. 17
      vendor/github.com/klauspost/cpuid/detect_intel.go
  39. 23
      vendor/github.com/klauspost/cpuid/detect_ref.go
  40. 3
      vendor/github.com/klauspost/cpuid/generate.go
  41. 24
      vendor/github.com/klauspost/crc32/.gitignore
  42. 13
      vendor/github.com/klauspost/crc32/.travis.yml
  43. 28
      vendor/github.com/klauspost/crc32/LICENSE
  44. 87
      vendor/github.com/klauspost/crc32/README.md
  45. 207
      vendor/github.com/klauspost/crc32/crc32.go
  46. 230
      vendor/github.com/klauspost/crc32/crc32_amd64.go
  47. 319
      vendor/github.com/klauspost/crc32/crc32_amd64.s
  48. 43
      vendor/github.com/klauspost/crc32/crc32_amd64p32.go
  49. 67
      vendor/github.com/klauspost/crc32/crc32_amd64p32.s
  50. 89
      vendor/github.com/klauspost/crc32/crc32_generic.go
  51. 15
      vendor/github.com/klauspost/crc32/crc32_otherarch.go
  52. 91
      vendor/github.com/klauspost/crc32/crc32_s390x.go
  53. 249
      vendor/github.com/klauspost/crc32/crc32_s390x.s
  54. 8
      vendor/modules.txt

@ -9,6 +9,7 @@ require (
gitea.com/macaron/captcha v0.0.0-20190822015246-daa973478bae
gitea.com/macaron/cors v0.0.0-20190821152825-7dcef4a17175
gitea.com/macaron/csrf v0.0.0-20190822024205-3dc5a4474439
gitea.com/macaron/gzip v0.0.0-20191118033930-0c4c5566a0e5
gitea.com/macaron/i18n v0.0.0-20190822004228-474e714e2223
gitea.com/macaron/inject v0.0.0-20190805023432-d4c86e31027a
gitea.com/macaron/macaron v1.3.3-0.20190821202302-9646c0587edb
@ -55,9 +56,7 @@ require (
github.com/joho/godotenv v1.3.0 // indirect
github.com/kballard/go-shellquote v0.0.0-20170619183022-cd60e84ee657
github.com/keybase/go-crypto v0.0.0-20170605145657-00ac4db533f6
github.com/klauspost/compress v0.0.0-20161025140425-8df558b6cb6f
github.com/klauspost/cpuid v0.0.0-20160302075316-09cded8978dc // indirect
github.com/klauspost/crc32 v0.0.0-20161016154125-cb6bfca970f6 // indirect
github.com/klauspost/compress v1.9.2
github.com/lafriks/xormstore v1.3.2
github.com/lib/pq v1.2.0
github.com/lunny/dingtalk_webhook v0.0.0-20171025031554-e3534c89ef96

@ -20,6 +20,8 @@ gitea.com/macaron/cors v0.0.0-20190821152825-7dcef4a17175 h1:ikzdAGB6SsUGByW5wKl
gitea.com/macaron/cors v0.0.0-20190821152825-7dcef4a17175/go.mod h1:rtOK4J20kpMD9XcNsnO5YA843YSTe/MUMbDj/TJ/Q7A=
gitea.com/macaron/csrf v0.0.0-20190822024205-3dc5a4474439 h1:88c34YM29a1GlWLrLBaG/GTT2htDdJz1u3n9+lmPolg=
gitea.com/macaron/csrf v0.0.0-20190822024205-3dc5a4474439/go.mod h1:IsQPHx73HnnqFBYiVHjg87q4XBZyGXXu77xANukvZuk=
gitea.com/macaron/gzip v0.0.0-20191118033930-0c4c5566a0e5 h1:G/a7r0r2jEelSynBlv1+PAEZQKfsdRHQUMb1PlNvemM=
gitea.com/macaron/gzip v0.0.0-20191118033930-0c4c5566a0e5/go.mod h1:jGHtoovArcQj+sw7NJxyPgjuRxOSG9a/oFu3VkLRTKQ=
gitea.com/macaron/i18n v0.0.0-20190822004228-474e714e2223 h1:iZWwQif/LHMjBgfY/ua8CFVa4XMDfbbs7EZ0Q1dYguU=
gitea.com/macaron/i18n v0.0.0-20190822004228-474e714e2223/go.mod h1:+qsc10s4hBsHKU/9luGGumFh4m5FFVc7uih+8/mM1NY=
gitea.com/macaron/inject v0.0.0-20190803172902-8375ba841591/go.mod h1:h6E4kLao1Yko6DOU6QDnQPcuoNzvbZqzj2mtPcEn1aM=
@ -334,12 +336,8 @@ github.com/keybase/go-crypto v0.0.0-20170605145657-00ac4db533f6/go.mod h1:ghbZsc
github.com/kisielk/errcheck v1.1.0/go.mod h1:EZBBE59ingxPouuu3KfxchcWSUPOHkagtvWXihfKN4Q=
github.com/kisielk/errcheck v1.2.0/go.mod h1:/BMXB+zMLi60iA8Vv6Ksmxu/1UDYcXs4uQLJ+jE2L00=
github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
github.com/klauspost/compress v0.0.0-20161025140425-8df558b6cb6f h1:tCnZKEmDovgV4jmsclh6CuKk9AMzTzyVWfejgkgccVg=
github.com/klauspost/compress v0.0.0-20161025140425-8df558b6cb6f/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
github.com/klauspost/cpuid v0.0.0-20160302075316-09cded8978dc h1:WW8B7p7QBnFlqRVv/k6ro/S8Z7tCnYjJHcQNScx9YVs=
github.com/klauspost/cpuid v0.0.0-20160302075316-09cded8978dc/go.mod h1:Pj4uuM528wm8OyEC2QMXAi2YiTZ96dNQPGgoMS4s3ek=
github.com/klauspost/crc32 v0.0.0-20161016154125-cb6bfca970f6 h1:KAZ1BW2TCmT6PRihDPpocIy1QTtsAsrx6TneU/4+CMg=
github.com/klauspost/crc32 v0.0.0-20161016154125-cb6bfca970f6/go.mod h1:+ZoRqAPRLkC4NPOvfYeR5KNOrY6TD+/sAC3HXPZgDYg=
github.com/klauspost/compress v1.9.2 h1:LfVyl+ZlLlLDeQ/d2AqfGIIH4qEDu0Ed2S5GyhCWIWY=
github.com/klauspost/compress v1.9.2/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
github.com/konsorten/go-windows-terminal-sequences v1.0.1/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
github.com/konsorten/go-windows-terminal-sequences v1.0.2/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
github.com/kr/logfmt v0.0.0-20140226030751-b84e30acd515/go.mod h1:+0opPa2QZZtGFBFZlji/RkVcI2GknAs/DXo4wKdlNEc=

@ -15,10 +15,10 @@ import (
"testing"
"code.gitea.io/gitea/models"
"code.gitea.io/gitea/modules/gzip"
"code.gitea.io/gitea/modules/lfs"
"code.gitea.io/gitea/modules/setting"
"gitea.com/macaron/gzip"
gzipp "github.com/klauspost/compress/gzip"
"github.com/stretchr/testify/assert"
)

@ -1,131 +0,0 @@
// Copyright 2019 The Gitea Authors. All rights reserved.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package gzip
import (
"archive/zip"
"bytes"
"io/ioutil"
"net/http"
"net/http/httptest"
"testing"
"gitea.com/macaron/macaron"
gzipp "github.com/klauspost/compress/gzip"
"github.com/stretchr/testify/assert"
)
func setup(sampleResponse []byte) (*macaron.Macaron, *[]byte) {
m := macaron.New()
m.Use(Middleware())
m.Get("/", func() *[]byte { return &sampleResponse })
return m, &sampleResponse
}
func reqNoAcceptGzip(t *testing.T, m *macaron.Macaron, sampleResponse *[]byte) {
// Request without accept gzip: Should not gzip
resp := httptest.NewRecorder()
req, err := http.NewRequest("GET", "/", nil)
assert.NoError(t, err)
m.ServeHTTP(resp, req)
_, ok := resp.HeaderMap[contentEncodingHeader]
assert.False(t, ok)
contentEncoding := resp.Header().Get(contentEncodingHeader)
assert.NotContains(t, contentEncoding, "gzip")
result := resp.Body.Bytes()
assert.Equal(t, *sampleResponse, result)
}
func reqAcceptGzip(t *testing.T, m *macaron.Macaron, sampleResponse *[]byte, expectGzip bool) {
// Request without accept gzip: Should not gzip
resp := httptest.NewRecorder()
req, err := http.NewRequest("GET", "/", nil)
assert.NoError(t, err)
req.Header.Set(acceptEncodingHeader, "gzip")
m.ServeHTTP(resp, req)
_, ok := resp.HeaderMap[contentEncodingHeader]
assert.Equal(t, ok, expectGzip)
contentEncoding := resp.Header().Get(contentEncodingHeader)
if expectGzip {
assert.Contains(t, contentEncoding, "gzip")
gzippReader, err := gzipp.NewReader(resp.Body)
assert.NoError(t, err)
result, err := ioutil.ReadAll(gzippReader)
assert.NoError(t, err)
assert.Equal(t, *sampleResponse, result)
} else {
assert.NotContains(t, contentEncoding, "gzip")
result := resp.Body.Bytes()
assert.Equal(t, *sampleResponse, result)
}
}
func TestMiddlewareSmall(t *testing.T) {
m, sampleResponse := setup([]byte("Small response"))
reqNoAcceptGzip(t, m, sampleResponse)
reqAcceptGzip(t, m, sampleResponse, false)
}
func TestMiddlewareLarge(t *testing.T) {
b := make([]byte, MinSize+1)
for i := range b {
b[i] = byte(i % 256)
}
m, sampleResponse := setup(b)
reqNoAcceptGzip(t, m, sampleResponse)
// This should be gzipped as we accept gzip
reqAcceptGzip(t, m, sampleResponse, true)
}
func TestMiddlewareGzip(t *testing.T) {
b := make([]byte, MinSize*10)
for i := range b {
b[i] = byte(i % 256)
}
outputBuffer := bytes.NewBuffer([]byte{})
gzippWriter := gzipp.NewWriter(outputBuffer)
gzippWriter.Write(b)
gzippWriter.Flush()
gzippWriter.Close()
output := outputBuffer.Bytes()
m, sampleResponse := setup(output)
reqNoAcceptGzip(t, m, sampleResponse)
// This should not be gzipped even though we accept gzip
reqAcceptGzip(t, m, sampleResponse, false)
}
func TestMiddlewareZip(t *testing.T) {
b := make([]byte, MinSize*10)
for i := range b {
b[i] = byte(i % 256)
}
outputBuffer := bytes.NewBuffer([]byte{})
zipWriter := zip.NewWriter(outputBuffer)
fileWriter, err := zipWriter.Create("default")
assert.NoError(t, err)
fileWriter.Write(b)
//fileWriter.Close()
zipWriter.Close()
output := outputBuffer.Bytes()
m, sampleResponse := setup(output)
reqNoAcceptGzip(t, m, sampleResponse)
// This should not be gzipped even though we accept gzip
reqAcceptGzip(t, m, sampleResponse, false)
}

@ -16,7 +16,6 @@ import (
"code.gitea.io/gitea/models"
"code.gitea.io/gitea/modules/auth"
"code.gitea.io/gitea/modules/context"
"code.gitea.io/gitea/modules/gzip"
"code.gitea.io/gitea/modules/lfs"
"code.gitea.io/gitea/modules/log"
"code.gitea.io/gitea/modules/metrics"
@ -44,6 +43,7 @@ import (
"gitea.com/macaron/captcha"
"gitea.com/macaron/cors"
"gitea.com/macaron/csrf"
"gitea.com/macaron/gzip"
"gitea.com/macaron/i18n"
"gitea.com/macaron/macaron"
"gitea.com/macaron/session"

@ -0,0 +1,9 @@
module gitea.com/macaron/gzip
go 1.12
require (
gitea.com/macaron/macaron v1.3.3-0.20190821202302-9646c0587edb
github.com/klauspost/compress v1.9.2
github.com/stretchr/testify v1.4.0
)

@ -0,0 +1,42 @@
gitea.com/macaron/inject v0.0.0-20190803172902-8375ba841591 h1:UbCTjPcLrNxR9LzKDjQBMT2zoxZuEnca1pZCpgeMuhQ=
gitea.com/macaron/inject v0.0.0-20190803172902-8375ba841591/go.mod h1:h6E4kLao1Yko6DOU6QDnQPcuoNzvbZqzj2mtPcEn1aM=
gitea.com/macaron/macaron v1.3.3-0.20190821202302-9646c0587edb h1:amL0md6orTj1tXY16ANzVU9FmzQB+W7aJwp8pVDbrmA=
gitea.com/macaron/macaron v1.3.3-0.20190821202302-9646c0587edb/go.mod h1:0coI+mSPSwbsyAbOuFllVS38awuk9mevhLD52l50Gjs=
github.com/davecgh/go-spew v1.1.0 h1:ZDRjVQ15GmhC3fiQ8ni8+OwkZQO4DARzQgrnXU1Liz8=
github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
github.com/gopherjs/gopherjs v0.0.0-20181017120253-0766667cb4d1/go.mod h1:wJfORRmW1u3UXTncJ5qlYoELFm8eSnnEO6hX4iZ3EWY=
github.com/gopherjs/gopherjs v0.0.0-20181103185306-d547d1d9531e h1:JKmoR8x90Iww1ks85zJ1lfDGgIiMDuIptTOhJq+zKyg=
github.com/gopherjs/gopherjs v0.0.0-20181103185306-d547d1d9531e/go.mod h1:wJfORRmW1u3UXTncJ5qlYoELFm8eSnnEO6hX4iZ3EWY=
github.com/jtolds/gls v4.2.1+incompatible/go.mod h1:QJZ7F/aHp+rZTRtaJ1ow/lLfFfVYBRgL+9YlvaHOwJU=
github.com/jtolds/gls v4.20.0+incompatible h1:xdiiI2gbIgH/gLH7ADydsJ1uDOEzR8yvV7C0MuV77Wo=
github.com/jtolds/gls v4.20.0+incompatible/go.mod h1:QJZ7F/aHp+rZTRtaJ1ow/lLfFfVYBRgL+9YlvaHOwJU=
github.com/klauspost/compress v1.9.2 h1:LfVyl+ZlLlLDeQ/d2AqfGIIH4qEDu0Ed2S5GyhCWIWY=
github.com/klauspost/compress v1.9.2/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
github.com/smartystreets/assertions v0.0.0-20180927180507-b2de0cb4f26d/go.mod h1:OnSkiWE9lh6wB0YB77sQom3nweQdgAjqCqsofrRNTgc=
github.com/smartystreets/assertions v0.0.0-20190116191733-b6c0e53d7304 h1:Jpy1PXuP99tXNrhbq2BaPz9B+jNAvH1JPQQpG/9GCXY=
github.com/smartystreets/assertions v0.0.0-20190116191733-b6c0e53d7304/go.mod h1:OnSkiWE9lh6wB0YB77sQom3nweQdgAjqCqsofrRNTgc=
github.com/smartystreets/goconvey v0.0.0-20181108003508-044398e4856c/go.mod h1:XDJAKZRPZ1CvBcN2aX5YOUTYGHki24fSF0Iv48Ibg0s=
github.com/smartystreets/goconvey v0.0.0-20190731233626-505e41936337 h1:WN9BUFbdyOsSH/XohnWpXOlq9NBD5sGAB2FciQMUEe8=
github.com/smartystreets/goconvey v0.0.0-20190731233626-505e41936337/go.mod h1:syvi0/a8iFYH4r/RixwvyeAJjdLS9QV7WQ/tjFTllLA=
github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
github.com/stretchr/testify v1.4.0 h1:2E4SXV/wtOkTonXsotYi4li6zVWxYlZuYNCXe9XRJyk=
github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
github.com/unknwon/com v0.0.0-20190804042917-757f69c95f3e h1:GSGeB9EAKY2spCABz6xOX5DbxZEXolK+nBSvmsQwRjM=
github.com/unknwon/com v0.0.0-20190804042917-757f69c95f3e/go.mod h1:tOOxU81rwgoCLoOVVPHb6T/wt8HZygqH5id+GNnlCXM=
golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
golang.org/x/crypto v0.0.0-20190701094942-4def268fd1a4 h1:HuIa8hRrWRSrqYzx1qI49NNxhdi2PrY7gxVSq1JjLDc=
golang.org/x/crypto v0.0.0-20190701094942-4def268fd1a4/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
golang.org/x/net v0.0.0-20190311183353-d8887717615a/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
golang.org/x/tools v0.0.0-20190328211700-ab21143f2384/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
gopkg.in/ini.v1 v1.44.0 h1:YRJzTUp0kSYWUVFF5XAbDFfyiqwsl0Vb9R8TVP5eRi0=
gopkg.in/ini.v1 v1.44.0/go.mod h1:pNLf8WUiyNEtQjuu5G5vTm06TEv9tsIgeAvK8hOrP4k=
gopkg.in/yaml.v2 v2.2.2 h1:ZCJp+EgiOT7lHqUV2J862kp8Qj64Jo6az82+3Td9dZw=
gopkg.in/yaml.v2 v2.2.2/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=

@ -1,4 +1,5 @@
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

@ -1,32 +0,0 @@
// Copyright 2012 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.
package flate
// forwardCopy is like the built-in copy function except that it always goes
// forward from the start, even if the dst and src overlap.
// It is equivalent to:
// for i := 0; i < n; i++ {
// mem[dst+i] = mem[src+i]
// }
func forwardCopy(mem []byte, dst, src, n int) {
if dst <= src {
copy(mem[dst:dst+n], mem[src:src+n])
return
}
for {
if dst >= src+n {
copy(mem[dst:dst+n], mem[src:src+n])
return
}
// There is some forward overlap. The destination
// will be filled with a repeated pattern of mem[src:src+k].
// We copy one instance of the pattern here, then repeat.
// Each time around this loop k will double.
k := dst - src
copy(mem[dst:dst+k], mem[src:src+k])
n -= k
dst += k
}
}

@ -1,41 +0,0 @@
//+build !noasm
//+build !appengine
// Copyright 2015, Klaus Post, see LICENSE for details.
package flate
import (
"github.com/klauspost/cpuid"
)
// crc32sse returns a hash for the first 4 bytes of the slice
// len(a) must be >= 4.
//go:noescape
func crc32sse(a []byte) uint32
// crc32sseAll calculates hashes for each 4-byte set in a.
// dst must be east len(a) - 4 in size.
// The size is not checked by the assembly.
//go:noescape
func crc32sseAll(a []byte, dst []uint32)
// matchLenSSE4 returns the number of matching bytes in a and b
// up to length 'max'. Both slices must be at least 'max'
// bytes in size.
//
// TODO: drop the "SSE4" name, since it doesn't use any SSE instructions.
//
//go:noescape
func matchLenSSE4(a, b []byte, max int) int
// histogram accumulates a histogram of b in h.
// h must be at least 256 entries in length,
// and must be cleared before calling this function.
//go:noescape
func histogram(b []byte, h []int32)
// Detect SSE 4.2 feature.
func init() {
useSSE42 = cpuid.CPU.SSE42()
}

@ -1,213 +0,0 @@
//+build !noasm
//+build !appengine
// Copyright 2015, Klaus Post, see LICENSE for details.
// func crc32sse(a []byte) uint32
TEXT ยทcrc32sse(SB), 4, $0
MOVQ a+0(FP), R10
XORQ BX, BX
// CRC32 dword (R10), EBX
BYTE $0xF2; BYTE $0x41; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0x1a
MOVL BX, ret+24(FP)
RET
// func crc32sseAll(a []byte, dst []uint32)
TEXT ยทcrc32sseAll(SB), 4, $0
MOVQ a+0(FP), R8 // R8: src
MOVQ a_len+8(FP), R10 // input length
MOVQ dst+24(FP), R9 // R9: dst
SUBQ $4, R10
JS end
JZ one_crc
MOVQ R10, R13
SHRQ $2, R10 // len/4
ANDQ $3, R13 // len&3
XORQ BX, BX
ADDQ $1, R13
TESTQ R10, R10
JZ rem_loop
crc_loop:
MOVQ (R8), R11
XORQ BX, BX
XORQ DX, DX
XORQ DI, DI
MOVQ R11, R12
SHRQ $8, R11
MOVQ R12, AX
MOVQ R11, CX
SHRQ $16, R12
SHRQ $16, R11
MOVQ R12, SI
// CRC32 EAX, EBX
BYTE $0xF2; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0xd8
// CRC32 ECX, EDX
BYTE $0xF2; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0xd1
// CRC32 ESI, EDI
BYTE $0xF2; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0xfe
MOVL BX, (R9)
MOVL DX, 4(R9)
MOVL DI, 8(R9)
XORQ BX, BX
MOVL R11, AX
// CRC32 EAX, EBX
BYTE $0xF2; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0xd8
MOVL BX, 12(R9)
ADDQ $16, R9
ADDQ $4, R8
XORQ BX, BX
SUBQ $1, R10
JNZ crc_loop
rem_loop:
MOVL (R8), AX
// CRC32 EAX, EBX
BYTE $0xF2; BYTE $0x0f
BYTE $0x38; BYTE $0xf1; BYTE $0xd8
MOVL BX, (R9)
ADDQ $4, R9
ADDQ $1, R8
XORQ BX, BX
SUBQ $1, R13
JNZ rem_loop
end:
RET
one_crc:
MOVQ $1, R13
XORQ BX, BX
JMP rem_loop
// func matchLenSSE4(a, b []byte, max int) int
TEXT ยทmatchLenSSE4(SB), 4, $0
MOVQ a_base+0(FP), SI
MOVQ b_base+24(FP), DI
MOVQ DI, DX
MOVQ max+48(FP), CX
cmp8:
// As long as we are 8 or more bytes before the end of max, we can load and
// compare 8 bytes at a time. If those 8 bytes are equal, repeat.
CMPQ CX, $8
JLT cmp1
MOVQ (SI), AX
MOVQ (DI), BX
CMPQ AX, BX
JNE bsf
ADDQ $8, SI
ADDQ $8, DI
SUBQ $8, CX
JMP cmp8
bsf:
// If those 8 bytes were not equal, XOR the two 8 byte values, and return
// the index of the first byte that differs. The BSF instruction finds the
// least significant 1 bit, the amd64 architecture is little-endian, and
// the shift by 3 converts a bit index to a byte index.
XORQ AX, BX
BSFQ BX, BX
SHRQ $3, BX
ADDQ BX, DI
// Subtract off &b[0] to convert from &b[ret] to ret, and return.
SUBQ DX, DI
MOVQ DI, ret+56(FP)
RET
cmp1:
// In the slices' tail, compare 1 byte at a time.
CMPQ CX, $0
JEQ matchLenEnd
MOVB (SI), AX
MOVB (DI), BX
CMPB AX, BX
JNE matchLenEnd
ADDQ $1, SI
ADDQ $1, DI
SUBQ $1, CX
JMP cmp1
matchLenEnd:
// Subtract off &b[0] to convert from &b[ret] to ret, and return.
SUBQ DX, DI
MOVQ DI, ret+56(FP)
RET
// func histogram(b []byte, h []int32)
TEXT ยทhistogram(SB), 4, $0
MOVQ b+0(FP), SI // SI: &b
MOVQ b_len+8(FP), R9 // R9: len(b)
MOVQ h+24(FP), DI // DI: Histogram
MOVQ R9, R8
SHRQ $3, R8
JZ hist1
XORQ R11, R11
loop_hist8:
MOVQ (SI), R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
MOVB R10, R11
INCL (DI)(R11*4)
SHRQ $8, R10
INCL (DI)(R10*4)
ADDQ $8, SI
DECQ R8
JNZ loop_hist8
hist1:
ANDQ $7, R9
JZ end_hist
XORQ R10, R10
loop_hist1:
MOVB (SI), R10
INCL (DI)(R10*4)
INCQ SI
DECQ R9
JNZ loop_hist1
end_hist:
RET

@ -1,35 +0,0 @@
//+build !amd64 noasm appengine
// Copyright 2015, Klaus Post, see LICENSE for details.
package flate
func init() {
useSSE42 = false
}
// crc32sse should never be called.
func crc32sse(a []byte) uint32 {
panic("no assembler")
}
// crc32sseAll should never be called.
func crc32sseAll(a []byte, dst []uint32) {
panic("no assembler")
}
// matchLenSSE4 should never be called.
func matchLenSSE4(a, b []byte, max int) int {
panic("no assembler")
return 0
}
// histogram accumulates a histogram of b in h.
//
// len(h) must be >= 256, and h's elements must be all zeroes.
func histogram(b []byte, h []int32) {
h = h[:256]
for _, t := range b {
h[t]++
}
}

File diff suppressed because it is too large Load Diff

@ -0,0 +1,257 @@
// Copyright 2011 The Snappy-Go Authors. All rights reserved.
// Modified for deflate by Klaus Post (c) 2015.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package flate
import (
"fmt"
"math/bits"
)
type fastEnc interface {
Encode(dst *tokens, src []byte)
Reset()
}
func newFastEnc(level int) fastEnc {
switch level {
case 1:
return &fastEncL1{fastGen: fastGen{cur: maxStoreBlockSize}}
case 2:
return &fastEncL2{fastGen: fastGen{cur: maxStoreBlockSize}}
case 3:
return &fastEncL3{fastGen: fastGen{cur: maxStoreBlockSize}}
case 4:
return &fastEncL4{fastGen: fastGen{cur: maxStoreBlockSize}}
case 5:
return &fastEncL5{fastGen: fastGen{cur: maxStoreBlockSize}}
case 6:
return &fastEncL6{fastGen: fastGen{cur: maxStoreBlockSize}}
default:
panic("invalid level specified")
}
}
const (
tableBits = 16 // Bits used in the table
tableSize = 1 << tableBits // Size of the table
tableShift = 32 - tableBits // Right-shift to get the tableBits most significant bits of a uint32.
baseMatchOffset = 1 // The smallest match offset
baseMatchLength = 3 // The smallest match length per the RFC section 3.2.5
maxMatchOffset = 1 << 15 // The largest match offset
bTableBits = 18 // Bits used in the big tables
bTableSize = 1 << bTableBits // Size of the table
allocHistory = maxMatchOffset * 10 // Size to preallocate for history.
bufferReset = (1 << 31) - allocHistory - maxStoreBlockSize // Reset the buffer offset when reaching this.
)
const (
prime3bytes = 506832829
prime4bytes = 2654435761
prime5bytes = 889523592379
prime6bytes = 227718039650203
prime7bytes = 58295818150454627
prime8bytes = 0xcf1bbcdcb7a56463
)
func load32(b []byte, i int) uint32 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:4]
return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
}
func load64(b []byte, i int) uint64 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:8]
return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
}
func load3232(b []byte, i int32) uint32 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:4]
return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
}
func load6432(b []byte, i int32) uint64 {
// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
b = b[i:]
b = b[:8]
return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
}
func hash(u uint32) uint32 {
return (u * 0x1e35a7bd) >> tableShift
}
type tableEntry struct {
val uint32
offset int32
}
// fastGen maintains the table for matches,
// and the previous byte block for level 2.
// This is the generic implementation.
type fastGen struct {
hist []byte
cur int32
}
func (e *fastGen) addBlock(src []byte) int32 {
// check if we have space already
if len(e.hist)+len(src) > cap(e.hist) {
if cap(e.hist) == 0 {
e.hist = make([]byte, 0, allocHistory)
} else {
if cap(e.hist) < maxMatchOffset*2 {
panic("unexpected buffer size")
}
// Move down
offset := int32(len(e.hist)) - maxMatchOffset
copy(e.hist[0:maxMatchOffset], e.hist[offset:])
e.cur += offset
e.hist = e.hist[:maxMatchOffset]
}
}
s := int32(len(e.hist))
e.hist = append(e.hist, src...)
return s
}
// hash4 returns the hash of u to fit in a hash table with h bits.
// Preferably h should be a constant and should always be <32.
func hash4u(u uint32, h uint8) uint32 {
return (u * prime4bytes) >> ((32 - h) & 31)
}
type tableEntryPrev struct {
Cur tableEntry
Prev tableEntry
}
// hash4x64 returns the hash of the lowest 4 bytes of u to fit in a hash table with h bits.
// Preferably h should be a constant and should always be <32.
func hash4x64(u uint64, h uint8) uint32 {
return (uint32(u) * prime4bytes) >> ((32 - h) & 31)
}
// hash7 returns the hash of the lowest 7 bytes of u to fit in a hash table with h bits.
// Preferably h should be a constant and should always be <64.
func hash7(u uint64, h uint8) uint32 {
return uint32(((u << (64 - 56)) * prime7bytes) >> ((64 - h) & 63))
}
// hash8 returns the hash of u to fit in a hash table with h bits.
// Preferably h should be a constant and should always be <64.
func hash8(u uint64, h uint8) uint32 {
return uint32((u * prime8bytes) >> ((64 - h) & 63))
}
// hash6 returns the hash of the lowest 6 bytes of u to fit in a hash table with h bits.
// Preferably h should be a constant and should always be <64.
func hash6(u uint64, h uint8) uint32 {
return uint32(((u << (64 - 48)) * prime6bytes) >> ((64 - h) & 63))
}
// matchlen will return the match length between offsets and t in src.
// The maximum length returned is maxMatchLength - 4.
// It is assumed that s > t, that t >=0 and s < len(src).
func (e *fastGen) matchlen(s, t int32, src []byte) int32 {
if debugDecode {
if t >= s {
panic(fmt.Sprint("t >=s:", t, s))
}
if int(s) >= len(src) {
panic(fmt.Sprint("s >= len(src):", s, len(src)))
}
if t < 0 {
panic(fmt.Sprint("t < 0:", t))
}
if s-t > maxMatchOffset {
panic(fmt.Sprint(s, "-", t, "(", s-t, ") > maxMatchLength (", maxMatchOffset, ")"))
}
}
s1 := int(s) + maxMatchLength - 4
if s1 > len(src) {
s1 = len(src)
}
// Extend the match to be as long as possible.
return int32(matchLen(src[s:s1], src[t:]))
}
// matchlenLong will return the match length between offsets and t in src.
// It is assumed that s > t, that t >=0 and s < len(src).
func (e *fastGen) matchlenLong(s, t int32, src []byte) int32 {
if debugDecode {
if t >= s {
panic(fmt.Sprint("t >=s:", t, s))
}
if int(s) >= len(src) {
panic(fmt.Sprint("s >= len(src):", s, len(src)))
}
if t < 0 {
panic(fmt.Sprint("t < 0:", t))
}
if s-t > maxMatchOffset {
panic(fmt.Sprint(s, "-", t, "(", s-t, ") > maxMatchLength (", maxMatchOffset, ")"))
}
}
// Extend the match to be as long as possible.
return int32(matchLen(src[s:], src[t:]))
}
// Reset the encoding table.
func (e *fastGen) Reset() {
if cap(e.hist) < int(maxMatchOffset*8) {
l := maxMatchOffset * 8
// Make it at least 1MB.
if l < 1<<20 {
l = 1 << 20
}
e.hist = make([]byte, 0, l)
}
// We offset current position so everything will be out of reach
e.cur += maxMatchOffset + int32(len(e.hist))
e.hist = e.hist[:0]
}
// matchLen returns the maximum length.
// 'a' must be the shortest of the two.
func matchLen(a, b []byte) int {
b = b[:len(a)]
var checked int
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
}

@ -35,7 +35,7 @@ const (
)
// The number of extra bits needed by length code X - LENGTH_CODES_START.
var lengthExtraBits = []int8{
var lengthExtraBits = [32]int8{
/* 257 */ 0, 0, 0,
/* 260 */ 0, 0, 0, 0, 0, 1, 1, 1, 1, 2,
/* 270 */ 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,
@ -43,14 +43,14 @@ var lengthExtraBits = []int8{
}
// The length indicated by length code X - LENGTH_CODES_START.
var lengthBase = []uint32{
var lengthBase = [32]uint8{
0, 1, 2, 3, 4, 5, 6, 7, 8, 10,
12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
64, 80, 96, 112, 128, 160, 192, 224, 255,
}
// offset code word extra bits.
var offsetExtraBits = []int8{
var offsetExtraBits = [64]int8{
0, 0, 0, 0, 1, 1, 2, 2, 3, 3,
4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
9, 9, 10, 10, 11, 11, 12, 12, 13, 13,
@ -58,7 +58,7 @@ var offsetExtraBits = []int8{
14, 14, 15, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20,
}
var offsetBase = []uint32{
var offsetBase = [64]uint32{
/* normal deflate */
0x000000, 0x000001, 0x000002, 0x000003, 0x000004,
0x000006, 0x000008, 0x00000c, 0x000010, 0x000018,
@ -85,26 +85,48 @@ type huffmanBitWriter struct {
// Data waiting to be written is bytes[0:nbytes]
// and then the low nbits of bits.
bits uint64
nbits uint
bytes [bufferSize]byte
codegenFreq [codegenCodeCount]int32
nbytes int
literalFreq []int32
offsetFreq []int32
codegen []uint8
nbits uint16
nbytes uint8
literalEncoding *huffmanEncoder
offsetEncoding *huffmanEncoder
codegenEncoding *huffmanEncoder
err error
lastHeader int
// Set between 0 (reused block can be up to 2x the size)
logReusePenalty uint
lastHuffMan bool
bytes [256]byte
literalFreq [lengthCodesStart + 32]uint16
offsetFreq [32]uint16
codegenFreq [codegenCodeCount]uint16
// codegen must have an extra space for the final symbol.
codegen [literalCount + offsetCodeCount + 1]uint8
}
// Huffman reuse.
//
// The huffmanBitWriter supports reusing huffman tables and thereby combining block sections.
//
// This is controlled by several variables:
//
// If lastHeader is non-zero the Huffman table can be reused.
// This also indicates that a Huffman table has been generated that can output all
// possible symbols.
// It also indicates that an EOB has not yet been emitted, so if a new tabel is generated
// an EOB with the previous table must be written.
//
// If lastHuffMan is set, a table for outputting literals has been generated and offsets are invalid.
//
// An incoming block estimates the output size of a new table using a 'fresh' by calculating the
// optimal size and adding a penalty in 'logReusePenalty'.
// A Huffman table is not optimal, which is why we add a penalty, and generating a new table
// is slower both for compression and decompression.
func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
return &huffmanBitWriter{
writer: w,
literalFreq: make([]int32, maxNumLit),
offsetFreq: make([]int32, offsetCodeCount),
codegen: make([]uint8, maxNumLit+offsetCodeCount+1),
literalEncoding: newHuffmanEncoder(maxNumLit),
literalEncoding: newHuffmanEncoder(literalCount),
codegenEncoding: newHuffmanEncoder(codegenCodeCount),
offsetEncoding: newHuffmanEncoder(offsetCodeCount),
}
@ -113,7 +135,42 @@ func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
func (w *huffmanBitWriter) reset(writer io.Writer) {
w.writer = writer
w.bits, w.nbits, w.nbytes, w.err = 0, 0, 0, nil
w.bytes = [bufferSize]byte{}
w.bytes = [256]byte{}
w.lastHeader = 0
w.lastHuffMan = false
}
func (w *huffmanBitWriter) canReuse(t *tokens) (offsets, lits bool) {
offsets, lits = true, true
a := t.offHist[:offsetCodeCount]
b := w.offsetFreq[:len(a)]
for i := range a {
if b[i] == 0 && a[i] != 0 {
offsets = false
break
}
}
a = t.extraHist[:literalCount-256]
b = w.literalFreq[256:literalCount]
b = b[:len(a)]
for i := range a {
if b[i] == 0 && a[i] != 0 {
lits = false
break
}
}
if lits {
a = t.litHist[:]
b = w.literalFreq[:len(a)]
for i := range a {
if b[i] == 0 && a[i] != 0 {
lits = false
break
}
}
}
return
}
func (w *huffmanBitWriter) flush() {
@ -144,30 +201,11 @@ func (w *huffmanBitWriter) write(b []byte) {
_, w.err = w.writer.Write(b)
}
func (w *huffmanBitWriter) writeBits(b int32, nb uint) {
if w.err != nil {
return
}
w.bits |= uint64(b) << w.nbits
func (w *huffmanBitWriter) writeBits(b int32, nb uint16) {
w.bits |= uint64(b) << (w.nbits & 63)
w.nbits += nb
if w.nbits >= 48 {
bits := w.bits
w.bits >>= 48
w.nbits -= 48
n := w.nbytes
bytes := w.bytes[n : n+6]
bytes[0] = byte(bits)
bytes[1] = byte(bits >> 8)
bytes[2] = byte(bits >> 16)
bytes[3] = byte(bits >> 24)
bytes[4] = byte(bits >> 32)
bytes[5] = byte(bits >> 40)
n += 6
if n >= bufferFlushSize {
w.write(w.bytes[:n])
n = 0
}
w.nbytes = n
w.writeOutBits()
}
}
@ -213,7 +251,7 @@ func (w *huffmanBitWriter) generateCodegen(numLiterals int, numOffsets int, litE
// a copy of the frequencies, and as the place where we put the result.
// This is fine because the output is always shorter than the input used
// so far.
codegen := w.codegen // cache
codegen := w.codegen[:] // cache
// Copy the concatenated code sizes to codegen. Put a marker at the end.
cgnl := codegen[:numLiterals]
for i := range cgnl {
@ -292,30 +330,54 @@ func (w *huffmanBitWriter) generateCodegen(numLiterals int, numOffsets int, litE
codegen[outIndex] = badCode
}
// dynamicSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
func (w *huffmanBitWriter) codegens() int {
numCodegens := len(w.codegenFreq)
for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
numCodegens--
}
return numCodegens
}
func (w *huffmanBitWriter) headerSize() (size, numCodegens int) {
numCodegens = len(w.codegenFreq)
for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
numCodegens--
}
header := 3 + 5 + 5 + 4 + (3 * numCodegens) +
return 3 + 5 + 5 + 4 + (3 * numCodegens) +
w.codegenEncoding.bitLength(w.codegenFreq[:]) +
int(w.codegenFreq[16])*2 +
int(w.codegenFreq[17])*3 +
int(w.codegenFreq[18])*7
int(w.codegenFreq[18])*7, numCodegens
}
// dynamicSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
header, numCodegens := w.headerSize()
size = header +
litEnc.bitLength(w.literalFreq) +
offEnc.bitLength(w.offsetFreq) +
litEnc.bitLength(w.literalFreq[:]) +
offEnc.bitLength(w.offsetFreq[:]) +
extraBits
return size, numCodegens
}
// extraBitSize will return the number of bits that will be written
// as "extra" bits on matches.
func (w *huffmanBitWriter) extraBitSize() int {
total := 0
for i, n := range w.literalFreq[257:literalCount] {
total += int(n) * int(lengthExtraBits[i&31])
}
for i, n := range w.offsetFreq[:offsetCodeCount] {
total += int(n) * int(offsetExtraBits[i&31])
}
return total
}
// fixedSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) fixedSize(extraBits int) int {
return 3 +
fixedLiteralEncoding.bitLength(w.literalFreq) +
fixedOffsetEncoding.bitLength(w.offsetFreq) +
fixedLiteralEncoding.bitLength(w.literalFreq[:]) +
fixedOffsetEncoding.bitLength(w.offsetFreq[:]) +
extraBits
}
@ -333,30 +395,36 @@ func (w *huffmanBitWriter) storedSize(in []byte) (int, bool) {
}
func (w *huffmanBitWriter) writeCode(c hcode) {
if w.err != nil {
return
}
// The function does not get inlined if we "& 63" the shift.
w.bits |= uint64(c.code) << w.nbits
w.nbits += uint(c.len)
w.nbits += c.len
if w.nbits >= 48 {
bits := w.bits
w.bits >>= 48
w.nbits -= 48
n := w.nbytes
bytes := w.bytes[n : n+6]
bytes[0] = byte(bits)
bytes[1] = byte(bits >> 8)
bytes[2] = byte(bits >> 16)
bytes[3] = byte(bits >> 24)
bytes[4] = byte(bits >> 32)
bytes[5] = byte(bits >> 40)
n += 6
if n >= bufferFlushSize {
w.write(w.bytes[:n])
w.writeOutBits()
}
}
// writeOutBits will write bits to the buffer.
func (w *huffmanBitWriter) writeOutBits() {
bits := w.bits
w.bits >>= 48
w.nbits -= 48
n := w.nbytes
w.bytes[n] = byte(bits)
w.bytes[n+1] = byte(bits >> 8)
w.bytes[n+2] = byte(bits >> 16)
w.bytes[n+3] = byte(bits >> 24)
w.bytes[n+4] = byte(bits >> 32)
w.bytes[n+5] = byte(bits >> 40)
n += 6
if n >= bufferFlushSize {
if w.err != nil {
n = 0
return
}
w.nbytes = n
w.write(w.bytes[:n])
n = 0
}
w.nbytes = n
}
// Write the header of a dynamic Huffman block to the output stream.
@ -412,6 +480,11 @@ func (w *huffmanBitWriter) writeStoredHeader(length int, isEof bool) {
if w.err != nil {
return
}
if w.lastHeader > 0 {
// We owe an EOB
w.writeCode(w.literalEncoding.codes[endBlockMarker])
w.lastHeader = 0
}
var flag int32
if isEof {
flag = 1
@ -426,6 +499,12 @@ func (w *huffmanBitWriter) writeFixedHeader(isEof bool) {
if w.err != nil {
return
}
if w.lastHeader > 0 {
// We owe an EOB
w.writeCode(w.literalEncoding.codes[endBlockMarker])
w.lastHeader = 0
}
// Indicate that we are a fixed Huffman block
var value int32 = 2
if isEof {
@ -439,29 +518,23 @@ func (w *huffmanBitWriter) writeFixedHeader(isEof bool) {
// is larger than the original bytes, the data will be written as a
// stored block.
// If the input is nil, the tokens will always be Huffman encoded.
func (w *huffmanBitWriter) writeBlock(tokens []token, eof bool, input []byte) {
func (w *huffmanBitWriter) writeBlock(tokens *tokens, eof bool, input []byte) {
if w.err != nil {
return
}
tokens = append(tokens, endBlockMarker)
numLiterals, numOffsets := w.indexTokens(tokens)
tokens.AddEOB()
if w.lastHeader > 0 {
// We owe an EOB
w.writeCode(w.literalEncoding.codes[endBlockMarker])
w.lastHeader = 0
}
numLiterals, numOffsets := w.indexTokens(tokens, false)
w.generate(tokens)
var extraBits int
storedSize, storable := w.storedSize(input)
if storable {
// We only bother calculating the costs of the extra bits required by
// the length of offset fields (which will be the same for both fixed
// and dynamic encoding), if we need to compare those two encodings
// against stored encoding.
for lengthCode := lengthCodesStart + 8; lengthCode < numLiterals; lengthCode++ {
// First eight length codes have extra size = 0.
extraBits += int(w.literalFreq[lengthCode]) * int(lengthExtraBits[lengthCode-lengthCodesStart])
}
for offsetCode := 4; offsetCode < numOffsets; offsetCode++ {
// First four offset codes have extra size = 0.
extraBits += int(w.offsetFreq[offsetCode]) * int(offsetExtraBits[offsetCode])
}
extraBits = w.extraBitSize()
}
// Figure out smallest code.
@ -500,7 +573,7 @@ func (w *huffmanBitWriter) writeBlock(tokens []token, eof bool, input []byte) {
}
// Write the tokens.
w.writeTokens(tokens, literalEncoding.codes, offsetEncoding.codes)
w.writeTokens(tokens.Slice(), literalEncoding.codes, offsetEncoding.codes)
}
// writeBlockDynamic encodes a block using a dynamic Huffman table.
@ -508,57 +581,103 @@ func (w *huffmanBitWriter) writeBlock(tokens []token, eof bool, input []byte) {
// histogram distribution.
// If input is supplied and the compression savings are below 1/16th of the
// input size the block is stored.
func (w *huffmanBitWriter) writeBlockDynamic(tokens []token, eof bool, input []byte) {
func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []byte, sync bool) {
if w.err != nil {
return
}
tokens = append(tokens, endBlockMarker)
numLiterals, numOffsets := w.indexTokens(tokens)
sync = sync || eof
if sync {
tokens.AddEOB()
}
// Generate codegen and codegenFrequencies, which indicates how to encode
// the literalEncoding and the offsetEncoding.
w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
w.codegenEncoding.generate(w.codegenFreq[:], 7)
size, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, 0)
// We cannot reuse pure huffman table.
if w.lastHuffMan && w.lastHeader > 0 {
// We will not try to reuse.
w.writeCode(w.literalEncoding.codes[endBlockMarker])
w.lastHeader = 0
w.lastHuffMan = false
}
if !sync {
tokens.Fill()
}
numLiterals, numOffsets := w.indexTokens(tokens, !sync)
// Store bytes, if we don't get a reasonable improvement.
if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
w.writeStoredHeader(len(input), eof)
w.writeBytes(input)
return
var size int
// Check if we should reuse.
if w.lastHeader > 0 {