Dinesh Belwalkar | e63f7f9 | 2019-11-22 23:11:16 +0000 | [diff] [blame] | 1 | // Copyright 2019+ Klaus Post. All rights reserved. |
| 2 | // License information can be found in the LICENSE file. |
| 3 | // Based on work by Yann Collet, released under BSD License. |
| 4 | |
| 5 | package zstd |
| 6 | |
| 7 | import ( |
| 8 | "errors" |
| 9 | "fmt" |
| 10 | "io" |
| 11 | ) |
| 12 | |
| 13 | type seq struct { |
| 14 | litLen uint32 |
| 15 | matchLen uint32 |
| 16 | offset uint32 |
| 17 | |
| 18 | // Codes are stored here for the encoder |
| 19 | // so they only have to be looked up once. |
| 20 | llCode, mlCode, ofCode uint8 |
| 21 | } |
| 22 | |
| 23 | func (s seq) String() string { |
| 24 | if s.offset <= 3 { |
| 25 | if s.offset == 0 { |
| 26 | return fmt.Sprint("litLen:", s.litLen, ", matchLen:", s.matchLen+zstdMinMatch, ", offset: INVALID (0)") |
| 27 | } |
| 28 | return fmt.Sprint("litLen:", s.litLen, ", matchLen:", s.matchLen+zstdMinMatch, ", offset:", s.offset, " (repeat)") |
| 29 | } |
| 30 | return fmt.Sprint("litLen:", s.litLen, ", matchLen:", s.matchLen+zstdMinMatch, ", offset:", s.offset-3, " (new)") |
| 31 | } |
| 32 | |
| 33 | type seqCompMode uint8 |
| 34 | |
| 35 | const ( |
| 36 | compModePredefined seqCompMode = iota |
| 37 | compModeRLE |
| 38 | compModeFSE |
| 39 | compModeRepeat |
| 40 | ) |
| 41 | |
| 42 | type sequenceDec struct { |
| 43 | // decoder keeps track of the current state and updates it from the bitstream. |
| 44 | fse *fseDecoder |
| 45 | state fseState |
| 46 | repeat bool |
| 47 | } |
| 48 | |
| 49 | // init the state of the decoder with input from stream. |
| 50 | func (s *sequenceDec) init(br *bitReader) error { |
| 51 | if s.fse == nil { |
| 52 | return errors.New("sequence decoder not defined") |
| 53 | } |
| 54 | s.state.init(br, s.fse.actualTableLog, s.fse.dt[:1<<s.fse.actualTableLog]) |
| 55 | return nil |
| 56 | } |
| 57 | |
| 58 | // sequenceDecs contains all 3 sequence decoders and their state. |
| 59 | type sequenceDecs struct { |
| 60 | litLengths sequenceDec |
| 61 | offsets sequenceDec |
| 62 | matchLengths sequenceDec |
| 63 | prevOffset [3]int |
| 64 | hist []byte |
| 65 | literals []byte |
| 66 | out []byte |
| 67 | maxBits uint8 |
| 68 | } |
| 69 | |
| 70 | // initialize all 3 decoders from the stream input. |
| 71 | func (s *sequenceDecs) initialize(br *bitReader, hist *history, literals, out []byte) error { |
| 72 | if err := s.litLengths.init(br); err != nil { |
| 73 | return errors.New("litLengths:" + err.Error()) |
| 74 | } |
| 75 | if err := s.offsets.init(br); err != nil { |
| 76 | return errors.New("offsets:" + err.Error()) |
| 77 | } |
| 78 | if err := s.matchLengths.init(br); err != nil { |
| 79 | return errors.New("matchLengths:" + err.Error()) |
| 80 | } |
| 81 | s.literals = literals |
| 82 | s.hist = hist.b |
| 83 | s.prevOffset = hist.recentOffsets |
| 84 | s.maxBits = s.litLengths.fse.maxBits + s.offsets.fse.maxBits + s.matchLengths.fse.maxBits |
| 85 | s.out = out |
| 86 | return nil |
| 87 | } |
| 88 | |
| 89 | // decode sequences from the stream with the provided history. |
| 90 | func (s *sequenceDecs) decode(seqs int, br *bitReader, hist []byte) error { |
| 91 | startSize := len(s.out) |
| 92 | // Grab full sizes tables, to avoid bounds checks. |
| 93 | llTable, mlTable, ofTable := s.litLengths.fse.dt[:maxTablesize], s.matchLengths.fse.dt[:maxTablesize], s.offsets.fse.dt[:maxTablesize] |
| 94 | llState, mlState, ofState := s.litLengths.state.state, s.matchLengths.state.state, s.offsets.state.state |
| 95 | |
| 96 | for i := seqs - 1; i >= 0; i-- { |
| 97 | if br.overread() { |
| 98 | printf("reading sequence %d, exceeded available data\n", seqs-i) |
| 99 | return io.ErrUnexpectedEOF |
| 100 | } |
| 101 | var litLen, matchOff, matchLen int |
| 102 | if br.off > 4+((maxOffsetBits+16+16)>>3) { |
| 103 | litLen, matchOff, matchLen = s.nextFast(br, llState, mlState, ofState) |
| 104 | br.fillFast() |
| 105 | } else { |
| 106 | litLen, matchOff, matchLen = s.next(br, llState, mlState, ofState) |
| 107 | br.fill() |
| 108 | } |
| 109 | |
| 110 | if debugSequences { |
| 111 | println("Seq", seqs-i-1, "Litlen:", litLen, "matchOff:", matchOff, "(abs) matchLen:", matchLen) |
| 112 | } |
| 113 | |
| 114 | if litLen > len(s.literals) { |
| 115 | return fmt.Errorf("unexpected literal count, want %d bytes, but only %d is available", litLen, len(s.literals)) |
| 116 | } |
| 117 | size := litLen + matchLen + len(s.out) |
| 118 | if size-startSize > maxBlockSize { |
| 119 | return fmt.Errorf("output (%d) bigger than max block size", size) |
| 120 | } |
| 121 | if size > cap(s.out) { |
| 122 | // Not enough size, will be extremely rarely triggered, |
| 123 | // but could be if destination slice is too small for sync operations. |
| 124 | // We add maxBlockSize to the capacity. |
| 125 | s.out = append(s.out, make([]byte, maxBlockSize)...) |
| 126 | s.out = s.out[:len(s.out)-maxBlockSize] |
| 127 | } |
| 128 | if matchLen > maxMatchLen { |
| 129 | return fmt.Errorf("match len (%d) bigger than max allowed length", matchLen) |
| 130 | } |
| 131 | if matchOff > len(s.out)+len(hist)+litLen { |
| 132 | return fmt.Errorf("match offset (%d) bigger than current history (%d)", matchOff, len(s.out)+len(hist)+litLen) |
| 133 | } |
| 134 | if matchOff == 0 && matchLen > 0 { |
| 135 | return fmt.Errorf("zero matchoff and matchlen > 0") |
| 136 | } |
| 137 | |
| 138 | s.out = append(s.out, s.literals[:litLen]...) |
| 139 | s.literals = s.literals[litLen:] |
| 140 | out := s.out |
| 141 | |
| 142 | // Copy from history. |
| 143 | // TODO: Blocks without history could be made to ignore this completely. |
| 144 | if v := matchOff - len(s.out); v > 0 { |
| 145 | // v is the start position in history from end. |
| 146 | start := len(s.hist) - v |
| 147 | if matchLen > v { |
| 148 | // Some goes into current block. |
| 149 | // Copy remainder of history |
| 150 | out = append(out, s.hist[start:]...) |
| 151 | matchOff -= v |
| 152 | matchLen -= v |
| 153 | } else { |
| 154 | out = append(out, s.hist[start:start+matchLen]...) |
| 155 | matchLen = 0 |
| 156 | } |
| 157 | } |
| 158 | // We must be in current buffer now |
| 159 | if matchLen > 0 { |
| 160 | start := len(s.out) - matchOff |
| 161 | if matchLen <= len(s.out)-start { |
| 162 | // No overlap |
| 163 | out = append(out, s.out[start:start+matchLen]...) |
| 164 | } else { |
| 165 | // Overlapping copy |
| 166 | // Extend destination slice and copy one byte at the time. |
| 167 | out = out[:len(out)+matchLen] |
| 168 | src := out[start : start+matchLen] |
| 169 | // Destination is the space we just added. |
| 170 | dst := out[len(out)-matchLen:] |
| 171 | dst = dst[:len(src)] |
| 172 | for i := range src { |
| 173 | dst[i] = src[i] |
| 174 | } |
| 175 | } |
| 176 | } |
| 177 | s.out = out |
| 178 | if i == 0 { |
| 179 | // This is the last sequence, so we shouldn't update state. |
| 180 | break |
| 181 | } |
| 182 | |
| 183 | // Manually inlined, ~ 5-20% faster |
| 184 | // Update all 3 states at once. Approx 20% faster. |
| 185 | nBits := llState.nbBits() + mlState.nbBits() + ofState.nbBits() |
| 186 | if nBits == 0 { |
| 187 | llState = llTable[llState.newState()&maxTableMask] |
| 188 | mlState = mlTable[mlState.newState()&maxTableMask] |
| 189 | ofState = ofTable[ofState.newState()&maxTableMask] |
| 190 | } else { |
| 191 | bits := br.getBitsFast(nBits) |
| 192 | lowBits := uint16(bits >> ((ofState.nbBits() + mlState.nbBits()) & 31)) |
| 193 | llState = llTable[(llState.newState()+lowBits)&maxTableMask] |
| 194 | |
| 195 | lowBits = uint16(bits >> (ofState.nbBits() & 31)) |
| 196 | lowBits &= bitMask[mlState.nbBits()&15] |
| 197 | mlState = mlTable[(mlState.newState()+lowBits)&maxTableMask] |
| 198 | |
| 199 | lowBits = uint16(bits) & bitMask[ofState.nbBits()&15] |
| 200 | ofState = ofTable[(ofState.newState()+lowBits)&maxTableMask] |
| 201 | } |
| 202 | } |
| 203 | |
| 204 | // Add final literals |
| 205 | s.out = append(s.out, s.literals...) |
| 206 | return nil |
| 207 | } |
| 208 | |
| 209 | // update states, at least 27 bits must be available. |
| 210 | func (s *sequenceDecs) update(br *bitReader) { |
| 211 | // Max 8 bits |
| 212 | s.litLengths.state.next(br) |
| 213 | // Max 9 bits |
| 214 | s.matchLengths.state.next(br) |
| 215 | // Max 8 bits |
| 216 | s.offsets.state.next(br) |
| 217 | } |
| 218 | |
| 219 | var bitMask [16]uint16 |
| 220 | |
| 221 | func init() { |
| 222 | for i := range bitMask[:] { |
| 223 | bitMask[i] = uint16((1 << uint(i)) - 1) |
| 224 | } |
| 225 | } |
| 226 | |
| 227 | // update states, at least 27 bits must be available. |
| 228 | func (s *sequenceDecs) updateAlt(br *bitReader) { |
| 229 | // Update all 3 states at once. Approx 20% faster. |
| 230 | a, b, c := s.litLengths.state.state, s.matchLengths.state.state, s.offsets.state.state |
| 231 | |
| 232 | nBits := a.nbBits() + b.nbBits() + c.nbBits() |
| 233 | if nBits == 0 { |
| 234 | s.litLengths.state.state = s.litLengths.state.dt[a.newState()] |
| 235 | s.matchLengths.state.state = s.matchLengths.state.dt[b.newState()] |
| 236 | s.offsets.state.state = s.offsets.state.dt[c.newState()] |
| 237 | return |
| 238 | } |
| 239 | bits := br.getBitsFast(nBits) |
| 240 | lowBits := uint16(bits >> ((c.nbBits() + b.nbBits()) & 31)) |
| 241 | s.litLengths.state.state = s.litLengths.state.dt[a.newState()+lowBits] |
| 242 | |
| 243 | lowBits = uint16(bits >> (c.nbBits() & 31)) |
| 244 | lowBits &= bitMask[b.nbBits()&15] |
| 245 | s.matchLengths.state.state = s.matchLengths.state.dt[b.newState()+lowBits] |
| 246 | |
| 247 | lowBits = uint16(bits) & bitMask[c.nbBits()&15] |
| 248 | s.offsets.state.state = s.offsets.state.dt[c.newState()+lowBits] |
| 249 | } |
| 250 | |
| 251 | // nextFast will return new states when there are at least 4 unused bytes left on the stream when done. |
| 252 | func (s *sequenceDecs) nextFast(br *bitReader, llState, mlState, ofState decSymbol) (ll, mo, ml int) { |
| 253 | // Final will not read from stream. |
| 254 | ll, llB := llState.final() |
| 255 | ml, mlB := mlState.final() |
| 256 | mo, moB := ofState.final() |
| 257 | |
| 258 | // extra bits are stored in reverse order. |
| 259 | br.fillFast() |
| 260 | mo += br.getBits(moB) |
| 261 | if s.maxBits > 32 { |
| 262 | br.fillFast() |
| 263 | } |
| 264 | ml += br.getBits(mlB) |
| 265 | ll += br.getBits(llB) |
| 266 | |
| 267 | if moB > 1 { |
| 268 | s.prevOffset[2] = s.prevOffset[1] |
| 269 | s.prevOffset[1] = s.prevOffset[0] |
| 270 | s.prevOffset[0] = mo |
| 271 | return |
| 272 | } |
| 273 | // mo = s.adjustOffset(mo, ll, moB) |
| 274 | // Inlined for rather big speedup |
| 275 | if ll == 0 { |
| 276 | // There is an exception though, when current sequence's literals_length = 0. |
| 277 | // In this case, repeated offsets are shifted by one, so an offset_value of 1 means Repeated_Offset2, |
| 278 | // an offset_value of 2 means Repeated_Offset3, and an offset_value of 3 means Repeated_Offset1 - 1_byte. |
| 279 | mo++ |
| 280 | } |
| 281 | |
| 282 | if mo == 0 { |
| 283 | mo = s.prevOffset[0] |
| 284 | return |
| 285 | } |
| 286 | var temp int |
| 287 | if mo == 3 { |
| 288 | temp = s.prevOffset[0] - 1 |
| 289 | } else { |
| 290 | temp = s.prevOffset[mo] |
| 291 | } |
| 292 | |
| 293 | if temp == 0 { |
| 294 | // 0 is not valid; input is corrupted; force offset to 1 |
| 295 | println("temp was 0") |
| 296 | temp = 1 |
| 297 | } |
| 298 | |
| 299 | if mo != 1 { |
| 300 | s.prevOffset[2] = s.prevOffset[1] |
| 301 | } |
| 302 | s.prevOffset[1] = s.prevOffset[0] |
| 303 | s.prevOffset[0] = temp |
| 304 | mo = temp |
| 305 | return |
| 306 | } |
| 307 | |
| 308 | func (s *sequenceDecs) next(br *bitReader, llState, mlState, ofState decSymbol) (ll, mo, ml int) { |
| 309 | // Final will not read from stream. |
| 310 | ll, llB := llState.final() |
| 311 | ml, mlB := mlState.final() |
| 312 | mo, moB := ofState.final() |
| 313 | |
| 314 | // extra bits are stored in reverse order. |
| 315 | br.fill() |
| 316 | if s.maxBits <= 32 { |
| 317 | mo += br.getBits(moB) |
| 318 | ml += br.getBits(mlB) |
| 319 | ll += br.getBits(llB) |
| 320 | } else { |
| 321 | mo += br.getBits(moB) |
| 322 | br.fill() |
| 323 | // matchlength+literal length, max 32 bits |
| 324 | ml += br.getBits(mlB) |
| 325 | ll += br.getBits(llB) |
| 326 | |
| 327 | } |
| 328 | mo = s.adjustOffset(mo, ll, moB) |
| 329 | return |
| 330 | } |
| 331 | |
| 332 | func (s *sequenceDecs) adjustOffset(offset, litLen int, offsetB uint8) int { |
| 333 | if offsetB > 1 { |
| 334 | s.prevOffset[2] = s.prevOffset[1] |
| 335 | s.prevOffset[1] = s.prevOffset[0] |
| 336 | s.prevOffset[0] = offset |
| 337 | return offset |
| 338 | } |
| 339 | |
| 340 | if litLen == 0 { |
| 341 | // There is an exception though, when current sequence's literals_length = 0. |
| 342 | // In this case, repeated offsets are shifted by one, so an offset_value of 1 means Repeated_Offset2, |
| 343 | // an offset_value of 2 means Repeated_Offset3, and an offset_value of 3 means Repeated_Offset1 - 1_byte. |
| 344 | offset++ |
| 345 | } |
| 346 | |
| 347 | if offset == 0 { |
| 348 | return s.prevOffset[0] |
| 349 | } |
| 350 | var temp int |
| 351 | if offset == 3 { |
| 352 | temp = s.prevOffset[0] - 1 |
| 353 | } else { |
| 354 | temp = s.prevOffset[offset] |
| 355 | } |
| 356 | |
| 357 | if temp == 0 { |
| 358 | // 0 is not valid; input is corrupted; force offset to 1 |
| 359 | println("temp was 0") |
| 360 | temp = 1 |
| 361 | } |
| 362 | |
| 363 | if offset != 1 { |
| 364 | s.prevOffset[2] = s.prevOffset[1] |
| 365 | } |
| 366 | s.prevOffset[1] = s.prevOffset[0] |
| 367 | s.prevOffset[0] = temp |
| 368 | return temp |
| 369 | } |
| 370 | |
| 371 | // mergeHistory will merge history. |
| 372 | func (s *sequenceDecs) mergeHistory(hist *sequenceDecs) (*sequenceDecs, error) { |
| 373 | for i := uint(0); i < 3; i++ { |
| 374 | var sNew, sHist *sequenceDec |
| 375 | switch i { |
| 376 | default: |
| 377 | // same as "case 0": |
| 378 | sNew = &s.litLengths |
| 379 | sHist = &hist.litLengths |
| 380 | case 1: |
| 381 | sNew = &s.offsets |
| 382 | sHist = &hist.offsets |
| 383 | case 2: |
| 384 | sNew = &s.matchLengths |
| 385 | sHist = &hist.matchLengths |
| 386 | } |
| 387 | if sNew.repeat { |
| 388 | if sHist.fse == nil { |
| 389 | return nil, fmt.Errorf("sequence stream %d, repeat requested, but no history", i) |
| 390 | } |
| 391 | continue |
| 392 | } |
| 393 | if sNew.fse == nil { |
| 394 | return nil, fmt.Errorf("sequence stream %d, no fse found", i) |
| 395 | } |
| 396 | if sHist.fse != nil && !sHist.fse.preDefined { |
| 397 | fseDecoderPool.Put(sHist.fse) |
| 398 | } |
| 399 | sHist.fse = sNew.fse |
| 400 | } |
| 401 | return hist, nil |
| 402 | } |