Overview
Comment: | Check in the new decompressor implementation in a separate branch |
---|---|
Downloads: | Tarball | ZIP archive | SQL archive |
Timelines: | family | ancestors | descendants | both | decompressor2 |
Files: | files | file ages | folders |
SHA1: |
bd1368b81facc3a0f23d6d04f2ed9aba |
User & Date: | spaskalev on 2014-12-21 22:12:43 |
Other Links: | branch diff | manifest | tags |
Context
2014-12-21
| ||
23:26 | Closing the decompressor2 branch as this implementation is slower than the naive one. check-in: 52e14c83da user: spaskalev tags: decompressor2 | |
22:12 | Check in the new decompressor implementation in a separate branch check-in: bd1368b81f user: spaskalev tags: decompressor2 | |
19:38 | Added debug/pprof to ease basic cpu profiling check-in: 1a4bdf36e2 user: spaskalev tags: trunk | |
Changes
Modified src/0dev.org/predictor/predictor.go from [d2a3bd9d21] to [84146b7c8c].
1 2 3 4 5 6 7 8 9 10 11 12 | // Package predictor implements the predictor compression/decompression algorithm // as specified by RFC1978 - PPP Predictor Compression Protocol package predictor import ( "io" ) type context struct { table [1 << 16]byte input []byte hash uint16 | > | 1 2 3 4 5 6 7 8 9 10 11 12 13 | // Package predictor implements the predictor compression/decompression algorithm // as specified by RFC1978 - PPP Predictor Compression Protocol package predictor import ( bits "0dev.org/bits" "io" ) type context struct { table [1 << 16]byte input []byte hash uint16 |
︙ | ︙ | |||
68 69 70 71 72 73 74 | return err } // ... and stage the rest of the data in the buffer ctx.input = append(ctx.input, data[blockSize-bufferLength:]...) return nil } | < | 69 70 71 72 73 74 75 76 77 78 79 80 81 82 | return err } // ... and stage the rest of the data in the buffer ctx.input = append(ctx.input, data[blockSize-bufferLength:]...) return nil } var buf []byte = make([]byte, 1, blockSize+1) for block := 0; block < len(data)/blockSize; block++ { for i := 0; i < blockSize; i++ { var current byte = data[(block*blockSize)+i] if ctx.table[ctx.hash] == current { // Guess was right - don't output buf[0] |= 1 << uint(i) |
︙ | ︙ | |||
115 116 117 118 119 120 121 | // Required to implement io.Reader func (r decompressor) Read(output []byte) (int, error) { return r(output) } // Returns an io.Reader implementation that wraps the provided io.Reader // and decompresses data according to the predictor algorithm | | | | | < < < < | | > > > | | > > > > > > > > | < < < | < | > > > > | > | | | < < | | > | > > | | | | > > | | | | > | | 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 | // Required to implement io.Reader func (r decompressor) Read(output []byte) (int, error) { return r(output) } // Returns an io.Reader implementation that wraps the provided io.Reader // and decompresses data according to the predictor algorithm func Decompressor(reader io.Reader) io.Reader { var ctx context ctx.input = make([]byte, 0, 8) return decompressor(func(output []byte) (int, error) { var ( err error flags byte readCount, available int ) // Sanity check for space to read into if len(output) == 0 { return 0, nil } // Check whether we have leftover data in the buffer if len(ctx.input) > 0 { readCount = copy(output, ctx.input) // Check whether we still have leftover data in the buffer :) if readCount < len(ctx.input) { ctx.input = ctx.input[:copy(ctx.input, ctx.input[readCount:])] } return readCount, nil } // Read the next prediction header readCount, err = reader.Read(ctx.input[:1]) // Fail on error unless it is EOF if err != nil && err != io.EOF { return 0, err } else if readCount == 0 { return 0, err } // Extend the buffer, copy the prediction header // and calculate the number of subsequent bytes to read ctx.input = ctx.input[:8] flags = ctx.input[0] available = 8 - int(bits.Hamming(flags)) // Read the non-predicted bytes according to header. readCount, err = reader.Read(ctx.input[:available]) retryData: if readCount < int(available) && err == nil { // Retry the read if we have fewer bytes than what the prediction header indicates var rc int rc, err = reader.Read(ctx.input[readCount:available]) readCount += rc goto retryData } // Continue on any error, try to decompress and return it along the result // Spread the read bytes right to left to avoid overlapping for i, a := 7, available-1; i >= 0; i-- { if ((flags >> uint(i)) & 1) == 0 { ctx.input[i] = ctx.input[a] a-- } } // Walk the buffer, fill in the predicted blanks and update the guess table for i := uint(0); i < 8; i++ { if (flags & (1 << i)) > 0 { // Guess succeeded, fill in from the table ctx.input[i] = ctx.table[ctx.hash] readCount++ } else { // Guess failed, update the table ctx.table[ctx.hash] = ctx.input[i] } // Update the hash ctx.hash = (ctx.hash << 4) ^ uint16(ctx.input[i]) } // readCount now contains the precise amount of populated data ctx.input = ctx.input[:readCount] available = copy(output, ctx.input) // Check for remaining bytes that dont fit in the output buffer if available < readCount { ctx.input = ctx.input[:copy(ctx.input, ctx.input[available:])] } else { // Clear the buffer ctx.input = ctx.input[:0] } return available, err }) } |