diff options
Diffstat (limited to 'vendor/github.com/klauspost/compress/s2/encode_go.go')
| -rw-r--r-- | vendor/github.com/klauspost/compress/s2/encode_go.go | 729 |
1 files changed, 729 insertions, 0 deletions
diff --git a/vendor/github.com/klauspost/compress/s2/encode_go.go b/vendor/github.com/klauspost/compress/s2/encode_go.go new file mode 100644 index 0000000..6b393c3 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_go.go | |||
| @@ -0,0 +1,729 @@ | |||
| 1 | //go:build !amd64 || appengine || !gc || noasm | ||
| 2 | // +build !amd64 appengine !gc noasm | ||
| 3 | |||
| 4 | package s2 | ||
| 5 | |||
| 6 | import ( | ||
| 7 | "bytes" | ||
| 8 | "math/bits" | ||
| 9 | ) | ||
| 10 | |||
| 11 | const hasAmd64Asm = false | ||
| 12 | |||
| 13 | // encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It | ||
| 14 | // assumes that the varint-encoded length of the decompressed bytes has already | ||
| 15 | // been written. | ||
| 16 | // | ||
| 17 | // It also assumes that: | ||
| 18 | // | ||
| 19 | // len(dst) >= MaxEncodedLen(len(src)) | ||
| 20 | func encodeBlock(dst, src []byte) (d int) { | ||
| 21 | if len(src) < minNonLiteralBlockSize { | ||
| 22 | return 0 | ||
| 23 | } | ||
| 24 | return encodeBlockGo(dst, src) | ||
| 25 | } | ||
| 26 | |||
| 27 | // encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It | ||
| 28 | // assumes that the varint-encoded length of the decompressed bytes has already | ||
| 29 | // been written. | ||
| 30 | // | ||
| 31 | // It also assumes that: | ||
| 32 | // | ||
| 33 | // len(dst) >= MaxEncodedLen(len(src)) | ||
| 34 | func encodeBlockBetter(dst, src []byte) (d int) { | ||
| 35 | return encodeBlockBetterGo(dst, src) | ||
| 36 | } | ||
| 37 | |||
| 38 | // encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It | ||
| 39 | // assumes that the varint-encoded length of the decompressed bytes has already | ||
| 40 | // been written. | ||
| 41 | // | ||
| 42 | // It also assumes that: | ||
| 43 | // | ||
| 44 | // len(dst) >= MaxEncodedLen(len(src)) | ||
| 45 | func encodeBlockBetterSnappy(dst, src []byte) (d int) { | ||
| 46 | return encodeBlockBetterSnappyGo(dst, src) | ||
| 47 | } | ||
| 48 | |||
| 49 | // encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It | ||
| 50 | // assumes that the varint-encoded length of the decompressed bytes has already | ||
| 51 | // been written. | ||
| 52 | // | ||
| 53 | // It also assumes that: | ||
| 54 | // | ||
| 55 | // len(dst) >= MaxEncodedLen(len(src)) | ||
| 56 | func encodeBlockSnappy(dst, src []byte) (d int) { | ||
| 57 | if len(src) < minNonLiteralBlockSize { | ||
| 58 | return 0 | ||
| 59 | } | ||
| 60 | return encodeBlockSnappyGo(dst, src) | ||
| 61 | } | ||
| 62 | |||
| 63 | // emitLiteral writes a literal chunk and returns the number of bytes written. | ||
| 64 | // | ||
| 65 | // It assumes that: | ||
| 66 | // | ||
| 67 | // dst is long enough to hold the encoded bytes | ||
| 68 | // 0 <= len(lit) && len(lit) <= math.MaxUint32 | ||
| 69 | func emitLiteral(dst, lit []byte) int { | ||
| 70 | if len(lit) == 0 { | ||
| 71 | return 0 | ||
| 72 | } | ||
| 73 | const num = 63<<2 | tagLiteral | ||
| 74 | i, n := 0, uint(len(lit)-1) | ||
| 75 | switch { | ||
| 76 | case n < 60: | ||
| 77 | dst[0] = uint8(n)<<2 | tagLiteral | ||
| 78 | i = 1 | ||
| 79 | case n < 1<<8: | ||
| 80 | dst[1] = uint8(n) | ||
| 81 | dst[0] = 60<<2 | tagLiteral | ||
| 82 | i = 2 | ||
| 83 | case n < 1<<16: | ||
| 84 | dst[2] = uint8(n >> 8) | ||
| 85 | dst[1] = uint8(n) | ||
| 86 | dst[0] = 61<<2 | tagLiteral | ||
| 87 | i = 3 | ||
| 88 | case n < 1<<24: | ||
| 89 | dst[3] = uint8(n >> 16) | ||
| 90 | dst[2] = uint8(n >> 8) | ||
| 91 | dst[1] = uint8(n) | ||
| 92 | dst[0] = 62<<2 | tagLiteral | ||
| 93 | i = 4 | ||
| 94 | default: | ||
| 95 | dst[4] = uint8(n >> 24) | ||
| 96 | dst[3] = uint8(n >> 16) | ||
| 97 | dst[2] = uint8(n >> 8) | ||
| 98 | dst[1] = uint8(n) | ||
| 99 | dst[0] = 63<<2 | tagLiteral | ||
| 100 | i = 5 | ||
| 101 | } | ||
| 102 | return i + copy(dst[i:], lit) | ||
| 103 | } | ||
| 104 | |||
| 105 | // emitRepeat writes a repeat chunk and returns the number of bytes written. | ||
| 106 | // Length must be at least 4 and < 1<<24 | ||
| 107 | func emitRepeat(dst []byte, offset, length int) int { | ||
| 108 | // Repeat offset, make length cheaper | ||
| 109 | length -= 4 | ||
| 110 | if length <= 4 { | ||
| 111 | dst[0] = uint8(length)<<2 | tagCopy1 | ||
| 112 | dst[1] = 0 | ||
| 113 | return 2 | ||
| 114 | } | ||
| 115 | if length < 8 && offset < 2048 { | ||
| 116 | // Encode WITH offset | ||
| 117 | dst[1] = uint8(offset) | ||
| 118 | dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1 | ||
| 119 | return 2 | ||
| 120 | } | ||
| 121 | if length < (1<<8)+4 { | ||
| 122 | length -= 4 | ||
| 123 | dst[2] = uint8(length) | ||
| 124 | dst[1] = 0 | ||
| 125 | dst[0] = 5<<2 | tagCopy1 | ||
| 126 | return 3 | ||
| 127 | } | ||
| 128 | if length < (1<<16)+(1<<8) { | ||
| 129 | length -= 1 << 8 | ||
| 130 | dst[3] = uint8(length >> 8) | ||
| 131 | dst[2] = uint8(length >> 0) | ||
| 132 | dst[1] = 0 | ||
| 133 | dst[0] = 6<<2 | tagCopy1 | ||
| 134 | return 4 | ||
| 135 | } | ||
| 136 | const maxRepeat = (1 << 24) - 1 | ||
| 137 | length -= 1 << 16 | ||
| 138 | left := 0 | ||
| 139 | if length > maxRepeat { | ||
| 140 | left = length - maxRepeat + 4 | ||
| 141 | length = maxRepeat - 4 | ||
| 142 | } | ||
| 143 | dst[4] = uint8(length >> 16) | ||
| 144 | dst[3] = uint8(length >> 8) | ||
| 145 | dst[2] = uint8(length >> 0) | ||
| 146 | dst[1] = 0 | ||
| 147 | dst[0] = 7<<2 | tagCopy1 | ||
| 148 | if left > 0 { | ||
| 149 | return 5 + emitRepeat(dst[5:], offset, left) | ||
| 150 | } | ||
| 151 | return 5 | ||
| 152 | } | ||
| 153 | |||
| 154 | // emitCopy writes a copy chunk and returns the number of bytes written. | ||
| 155 | // | ||
| 156 | // It assumes that: | ||
| 157 | // | ||
| 158 | // dst is long enough to hold the encoded bytes | ||
| 159 | // 1 <= offset && offset <= math.MaxUint32 | ||
| 160 | // 4 <= length && length <= 1 << 24 | ||
| 161 | func emitCopy(dst []byte, offset, length int) int { | ||
| 162 | if offset >= 65536 { | ||
| 163 | i := 0 | ||
| 164 | if length > 64 { | ||
| 165 | // Emit a length 64 copy, encoded as 5 bytes. | ||
| 166 | dst[4] = uint8(offset >> 24) | ||
| 167 | dst[3] = uint8(offset >> 16) | ||
| 168 | dst[2] = uint8(offset >> 8) | ||
| 169 | dst[1] = uint8(offset) | ||
| 170 | dst[0] = 63<<2 | tagCopy4 | ||
| 171 | length -= 64 | ||
| 172 | if length >= 4 { | ||
| 173 | // Emit remaining as repeats | ||
| 174 | return 5 + emitRepeat(dst[5:], offset, length) | ||
| 175 | } | ||
| 176 | i = 5 | ||
| 177 | } | ||
| 178 | if length == 0 { | ||
| 179 | return i | ||
| 180 | } | ||
| 181 | // Emit a copy, offset encoded as 4 bytes. | ||
| 182 | dst[i+0] = uint8(length-1)<<2 | tagCopy4 | ||
| 183 | dst[i+1] = uint8(offset) | ||
| 184 | dst[i+2] = uint8(offset >> 8) | ||
| 185 | dst[i+3] = uint8(offset >> 16) | ||
| 186 | dst[i+4] = uint8(offset >> 24) | ||
| 187 | return i + 5 | ||
| 188 | } | ||
| 189 | |||
| 190 | // Offset no more than 2 bytes. | ||
| 191 | if length > 64 { | ||
| 192 | off := 3 | ||
| 193 | if offset < 2048 { | ||
| 194 | // emit 8 bytes as tagCopy1, rest as repeats. | ||
| 195 | dst[1] = uint8(offset) | ||
| 196 | dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1 | ||
| 197 | length -= 8 | ||
| 198 | off = 2 | ||
| 199 | } else { | ||
| 200 | // Emit a length 60 copy, encoded as 3 bytes. | ||
| 201 | // Emit remaining as repeat value (minimum 4 bytes). | ||
| 202 | dst[2] = uint8(offset >> 8) | ||
| 203 | dst[1] = uint8(offset) | ||
| 204 | dst[0] = 59<<2 | tagCopy2 | ||
| 205 | length -= 60 | ||
| 206 | } | ||
| 207 | // Emit remaining as repeats, at least 4 bytes remain. | ||
| 208 | return off + emitRepeat(dst[off:], offset, length) | ||
| 209 | } | ||
| 210 | if length >= 12 || offset >= 2048 { | ||
| 211 | // Emit the remaining copy, encoded as 3 bytes. | ||
| 212 | dst[2] = uint8(offset >> 8) | ||
| 213 | dst[1] = uint8(offset) | ||
| 214 | dst[0] = uint8(length-1)<<2 | tagCopy2 | ||
| 215 | return 3 | ||
| 216 | } | ||
| 217 | // Emit the remaining copy, encoded as 2 bytes. | ||
| 218 | dst[1] = uint8(offset) | ||
| 219 | dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 | ||
| 220 | return 2 | ||
| 221 | } | ||
| 222 | |||
| 223 | // emitCopyNoRepeat writes a copy chunk and returns the number of bytes written. | ||
| 224 | // | ||
| 225 | // It assumes that: | ||
| 226 | // | ||
| 227 | // dst is long enough to hold the encoded bytes | ||
| 228 | // 1 <= offset && offset <= math.MaxUint32 | ||
| 229 | // 4 <= length && length <= 1 << 24 | ||
| 230 | func emitCopyNoRepeat(dst []byte, offset, length int) int { | ||
| 231 | if offset >= 65536 { | ||
| 232 | i := 0 | ||
| 233 | if length > 64 { | ||
| 234 | // Emit a length 64 copy, encoded as 5 bytes. | ||
| 235 | dst[4] = uint8(offset >> 24) | ||
| 236 | dst[3] = uint8(offset >> 16) | ||
| 237 | dst[2] = uint8(offset >> 8) | ||
| 238 | dst[1] = uint8(offset) | ||
| 239 | dst[0] = 63<<2 | tagCopy4 | ||
| 240 | length -= 64 | ||
| 241 | if length >= 4 { | ||
| 242 | // Emit remaining as repeats | ||
| 243 | return 5 + emitCopyNoRepeat(dst[5:], offset, length) | ||
| 244 | } | ||
| 245 | i = 5 | ||
| 246 | } | ||
| 247 | if length == 0 { | ||
| 248 | return i | ||
| 249 | } | ||
| 250 | // Emit a copy, offset encoded as 4 bytes. | ||
| 251 | dst[i+0] = uint8(length-1)<<2 | tagCopy4 | ||
| 252 | dst[i+1] = uint8(offset) | ||
| 253 | dst[i+2] = uint8(offset >> 8) | ||
| 254 | dst[i+3] = uint8(offset >> 16) | ||
| 255 | dst[i+4] = uint8(offset >> 24) | ||
| 256 | return i + 5 | ||
| 257 | } | ||
| 258 | |||
| 259 | // Offset no more than 2 bytes. | ||
| 260 | if length > 64 { | ||
| 261 | // Emit a length 60 copy, encoded as 3 bytes. | ||
| 262 | // Emit remaining as repeat value (minimum 4 bytes). | ||
| 263 | dst[2] = uint8(offset >> 8) | ||
| 264 | dst[1] = uint8(offset) | ||
| 265 | dst[0] = 59<<2 | tagCopy2 | ||
| 266 | length -= 60 | ||
| 267 | // Emit remaining as repeats, at least 4 bytes remain. | ||
| 268 | return 3 + emitCopyNoRepeat(dst[3:], offset, length) | ||
| 269 | } | ||
| 270 | if length >= 12 || offset >= 2048 { | ||
| 271 | // Emit the remaining copy, encoded as 3 bytes. | ||
| 272 | dst[2] = uint8(offset >> 8) | ||
| 273 | dst[1] = uint8(offset) | ||
| 274 | dst[0] = uint8(length-1)<<2 | tagCopy2 | ||
| 275 | return 3 | ||
| 276 | } | ||
| 277 | // Emit the remaining copy, encoded as 2 bytes. | ||
| 278 | dst[1] = uint8(offset) | ||
| 279 | dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 | ||
| 280 | return 2 | ||
| 281 | } | ||
| 282 | |||
| 283 | // matchLen returns how many bytes match in a and b | ||
| 284 | // | ||
| 285 | // It assumes that: | ||
| 286 | // | ||
| 287 | // len(a) <= len(b) | ||
| 288 | func matchLen(a []byte, b []byte) int { | ||
| 289 | b = b[:len(a)] | ||
| 290 | var checked int | ||
| 291 | if len(a) > 4 { | ||
| 292 | // Try 4 bytes first | ||
| 293 | if diff := load32(a, 0) ^ load32(b, 0); diff != 0 { | ||
| 294 | return bits.TrailingZeros32(diff) >> 3 | ||
| 295 | } | ||
| 296 | // Switch to 8 byte matching. | ||
| 297 | checked = 4 | ||
| 298 | a = a[4:] | ||
| 299 | b = b[4:] | ||
| 300 | for len(a) >= 8 { | ||
| 301 | b = b[:len(a)] | ||
| 302 | if diff := load64(a, 0) ^ load64(b, 0); diff != 0 { | ||
| 303 | return checked + (bits.TrailingZeros64(diff) >> 3) | ||
| 304 | } | ||
| 305 | checked += 8 | ||
| 306 | a = a[8:] | ||
| 307 | b = b[8:] | ||
| 308 | } | ||
| 309 | } | ||
| 310 | b = b[:len(a)] | ||
| 311 | for i := range a { | ||
| 312 | if a[i] != b[i] { | ||
| 313 | return int(i) + checked | ||
| 314 | } | ||
| 315 | } | ||
| 316 | return len(a) + checked | ||
| 317 | } | ||
| 318 | |||
| 319 | // input must be > inputMargin | ||
| 320 | func calcBlockSize(src []byte) (d int) { | ||
| 321 | // Initialize the hash table. | ||
| 322 | const ( | ||
| 323 | tableBits = 13 | ||
| 324 | maxTableSize = 1 << tableBits | ||
| 325 | ) | ||
| 326 | |||
| 327 | var table [maxTableSize]uint32 | ||
| 328 | |||
| 329 | // sLimit is when to stop looking for offset/length copies. The inputMargin | ||
| 330 | // lets us use a fast path for emitLiteral in the main loop, while we are | ||
| 331 | // looking for copies. | ||
| 332 | sLimit := len(src) - inputMargin | ||
| 333 | |||
| 334 | // Bail if we can't compress to at least this. | ||
| 335 | dstLimit := len(src) - len(src)>>5 - 5 | ||
| 336 | |||
| 337 | // nextEmit is where in src the next emitLiteral should start from. | ||
| 338 | nextEmit := 0 | ||
| 339 | |||
| 340 | // The encoded form must start with a literal, as there are no previous | ||
| 341 | // bytes to copy, so we start looking for hash matches at s == 1. | ||
| 342 | s := 1 | ||
| 343 | cv := load64(src, s) | ||
| 344 | |||
| 345 | // We search for a repeat at -1, but don't output repeats when nextEmit == 0 | ||
| 346 | repeat := 1 | ||
| 347 | |||
| 348 | for { | ||
| 349 | candidate := 0 | ||
| 350 | for { | ||
| 351 | // Next src position to check | ||
| 352 | nextS := s + (s-nextEmit)>>6 + 4 | ||
| 353 | if nextS > sLimit { | ||
| 354 | goto emitRemainder | ||
| 355 | } | ||
| 356 | hash0 := hash6(cv, tableBits) | ||
| 357 | hash1 := hash6(cv>>8, tableBits) | ||
| 358 | candidate = int(table[hash0]) | ||
| 359 | candidate2 := int(table[hash1]) | ||
| 360 | table[hash0] = uint32(s) | ||
| 361 | table[hash1] = uint32(s + 1) | ||
| 362 | hash2 := hash6(cv>>16, tableBits) | ||
| 363 | |||
| 364 | // Check repeat at offset checkRep. | ||
| 365 | const checkRep = 1 | ||
| 366 | if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { | ||
| 367 | base := s + checkRep | ||
| 368 | // Extend back | ||
| 369 | for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { | ||
| 370 | i-- | ||
| 371 | base-- | ||
| 372 | } | ||
| 373 | d += emitLiteralSize(src[nextEmit:base]) | ||
| 374 | |||
| 375 | // Extend forward | ||
| 376 | candidate := s - repeat + 4 + checkRep | ||
| 377 | s += 4 + checkRep | ||
| 378 | for s <= sLimit { | ||
| 379 | if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { | ||
| 380 | s += bits.TrailingZeros64(diff) >> 3 | ||
| 381 | break | ||
| 382 | } | ||
| 383 | s += 8 | ||
| 384 | candidate += 8 | ||
| 385 | } | ||
| 386 | |||
| 387 | d += emitCopyNoRepeatSize(repeat, s-base) | ||
| 388 | nextEmit = s | ||
| 389 | if s >= sLimit { | ||
| 390 | goto emitRemainder | ||
| 391 | } | ||
| 392 | |||
| 393 | cv = load64(src, s) | ||
| 394 | continue | ||
| 395 | } | ||
| 396 | |||
| 397 | if uint32(cv) == load32(src, candidate) { | ||
| 398 | break | ||
| 399 | } | ||
| 400 | candidate = int(table[hash2]) | ||
| 401 | if uint32(cv>>8) == load32(src, candidate2) { | ||
| 402 | table[hash2] = uint32(s + 2) | ||
| 403 | candidate = candidate2 | ||
| 404 | s++ | ||
| 405 | break | ||
| 406 | } | ||
| 407 | table[hash2] = uint32(s + 2) | ||
| 408 | if uint32(cv>>16) == load32(src, candidate) { | ||
| 409 | s += 2 | ||
| 410 | break | ||
| 411 | } | ||
| 412 | |||
| 413 | cv = load64(src, nextS) | ||
| 414 | s = nextS | ||
| 415 | } | ||
| 416 | |||
| 417 | // Extend backwards | ||
| 418 | for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { | ||
| 419 | candidate-- | ||
| 420 | s-- | ||
| 421 | } | ||
| 422 | |||
| 423 | // Bail if we exceed the maximum size. | ||
| 424 | if d+(s-nextEmit) > dstLimit { | ||
| 425 | return 0 | ||
| 426 | } | ||
| 427 | |||
| 428 | // A 4-byte match has been found. We'll later see if more than 4 bytes | ||
| 429 | // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit | ||
| 430 | // them as literal bytes. | ||
| 431 | |||
| 432 | d += emitLiteralSize(src[nextEmit:s]) | ||
| 433 | |||
| 434 | // Call emitCopy, and then see if another emitCopy could be our next | ||
| 435 | // move. Repeat until we find no match for the input immediately after | ||
| 436 | // what was consumed by the last emitCopy call. | ||
| 437 | // | ||
| 438 | // If we exit this loop normally then we need to call emitLiteral next, | ||
| 439 | // though we don't yet know how big the literal will be. We handle that | ||
| 440 | // by proceeding to the next iteration of the main loop. We also can | ||
| 441 | // exit this loop via goto if we get close to exhausting the input. | ||
| 442 | for { | ||
| 443 | // Invariant: we have a 4-byte match at s, and no need to emit any | ||
| 444 | // literal bytes prior to s. | ||
| 445 | base := s | ||
| 446 | repeat = base - candidate | ||
| 447 | |||
| 448 | // Extend the 4-byte match as long as possible. | ||
| 449 | s += 4 | ||
| 450 | candidate += 4 | ||
| 451 | for s <= len(src)-8 { | ||
| 452 | if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { | ||
| 453 | s += bits.TrailingZeros64(diff) >> 3 | ||
| 454 | break | ||
| 455 | } | ||
| 456 | s += 8 | ||
| 457 | candidate += 8 | ||
| 458 | } | ||
| 459 | |||
| 460 | d += emitCopyNoRepeatSize(repeat, s-base) | ||
| 461 | if false { | ||
| 462 | // Validate match. | ||
| 463 | a := src[base:s] | ||
| 464 | b := src[base-repeat : base-repeat+(s-base)] | ||
| 465 | if !bytes.Equal(a, b) { | ||
| 466 | panic("mismatch") | ||
| 467 | } | ||
| 468 | } | ||
| 469 | |||
| 470 | nextEmit = s | ||
| 471 | if s >= sLimit { | ||
| 472 | goto emitRemainder | ||
| 473 | } | ||
| 474 | |||
| 475 | if d > dstLimit { | ||
| 476 | // Do we have space for more, if not bail. | ||
| 477 | return 0 | ||
| 478 | } | ||
| 479 | // Check for an immediate match, otherwise start search at s+1 | ||
| 480 | x := load64(src, s-2) | ||
| 481 | m2Hash := hash6(x, tableBits) | ||
| 482 | currHash := hash6(x>>16, tableBits) | ||
| 483 | candidate = int(table[currHash]) | ||
| 484 | table[m2Hash] = uint32(s - 2) | ||
| 485 | table[currHash] = uint32(s) | ||
| 486 | if uint32(x>>16) != load32(src, candidate) { | ||
| 487 | cv = load64(src, s+1) | ||
| 488 | s++ | ||
| 489 | break | ||
| 490 | } | ||
| 491 | } | ||
| 492 | } | ||
| 493 | |||
| 494 | emitRemainder: | ||
| 495 | if nextEmit < len(src) { | ||
| 496 | // Bail if we exceed the maximum size. | ||
| 497 | if d+len(src)-nextEmit > dstLimit { | ||
| 498 | return 0 | ||
| 499 | } | ||
| 500 | d += emitLiteralSize(src[nextEmit:]) | ||
| 501 | } | ||
| 502 | return d | ||
| 503 | } | ||
| 504 | |||
| 505 | // length must be > inputMargin. | ||
| 506 | func calcBlockSizeSmall(src []byte) (d int) { | ||
| 507 | // Initialize the hash table. | ||
| 508 | const ( | ||
| 509 | tableBits = 9 | ||
| 510 | maxTableSize = 1 << tableBits | ||
| 511 | ) | ||
| 512 | |||
| 513 | var table [maxTableSize]uint32 | ||
| 514 | |||
| 515 | // sLimit is when to stop looking for offset/length copies. The inputMargin | ||
| 516 | // lets us use a fast path for emitLiteral in the main loop, while we are | ||
| 517 | // looking for copies. | ||
| 518 | sLimit := len(src) - inputMargin | ||
| 519 | |||
| 520 | // Bail if we can't compress to at least this. | ||
| 521 | dstLimit := len(src) - len(src)>>5 - 5 | ||
| 522 | |||
| 523 | // nextEmit is where in src the next emitLiteral should start from. | ||
| 524 | nextEmit := 0 | ||
| 525 | |||
| 526 | // The encoded form must start with a literal, as there are no previous | ||
| 527 | // bytes to copy, so we start looking for hash matches at s == 1. | ||
| 528 | s := 1 | ||
| 529 | cv := load64(src, s) | ||
| 530 | |||
| 531 | // We search for a repeat at -1, but don't output repeats when nextEmit == 0 | ||
| 532 | repeat := 1 | ||
| 533 | |||
| 534 | for { | ||
| 535 | candidate := 0 | ||
| 536 | for { | ||
| 537 | // Next src position to check | ||
| 538 | nextS := s + (s-nextEmit)>>6 + 4 | ||
| 539 | if nextS > sLimit { | ||
| 540 | goto emitRemainder | ||
| 541 | } | ||
| 542 | hash0 := hash6(cv, tableBits) | ||
| 543 | hash1 := hash6(cv>>8, tableBits) | ||
| 544 | candidate = int(table[hash0]) | ||
| 545 | candidate2 := int(table[hash1]) | ||
| 546 | table[hash0] = uint32(s) | ||
| 547 | table[hash1] = uint32(s + 1) | ||
| 548 | hash2 := hash6(cv>>16, tableBits) | ||
| 549 | |||
| 550 | // Check repeat at offset checkRep. | ||
| 551 | const checkRep = 1 | ||
| 552 | if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { | ||
| 553 | base := s + checkRep | ||
| 554 | // Extend back | ||
| 555 | for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { | ||
| 556 | i-- | ||
| 557 | base-- | ||
| 558 | } | ||
| 559 | d += emitLiteralSize(src[nextEmit:base]) | ||
| 560 | |||
| 561 | // Extend forward | ||
| 562 | candidate := s - repeat + 4 + checkRep | ||
| 563 | s += 4 + checkRep | ||
| 564 | for s <= sLimit { | ||
| 565 | if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { | ||
| 566 | s += bits.TrailingZeros64(diff) >> 3 | ||
| 567 | break | ||
| 568 | } | ||
| 569 | s += 8 | ||
| 570 | candidate += 8 | ||
| 571 | } | ||
| 572 | |||
| 573 | d += emitCopyNoRepeatSize(repeat, s-base) | ||
| 574 | nextEmit = s | ||
| 575 | if s >= sLimit { | ||
| 576 | goto emitRemainder | ||
| 577 | } | ||
| 578 | |||
| 579 | cv = load64(src, s) | ||
| 580 | continue | ||
| 581 | } | ||
| 582 | |||
| 583 | if uint32(cv) == load32(src, candidate) { | ||
| 584 | break | ||
| 585 | } | ||
| 586 | candidate = int(table[hash2]) | ||
| 587 | if uint32(cv>>8) == load32(src, candidate2) { | ||
| 588 | table[hash2] = uint32(s + 2) | ||
| 589 | candidate = candidate2 | ||
| 590 | s++ | ||
| 591 | break | ||
| 592 | } | ||
| 593 | table[hash2] = uint32(s + 2) | ||
| 594 | if uint32(cv>>16) == load32(src, candidate) { | ||
| 595 | s += 2 | ||
| 596 | break | ||
| 597 | } | ||
| 598 | |||
| 599 | cv = load64(src, nextS) | ||
| 600 | s = nextS | ||
| 601 | } | ||
| 602 | |||
| 603 | // Extend backwards | ||
| 604 | for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { | ||
| 605 | candidate-- | ||
| 606 | s-- | ||
| 607 | } | ||
| 608 | |||
| 609 | // Bail if we exceed the maximum size. | ||
| 610 | if d+(s-nextEmit) > dstLimit { | ||
| 611 | return 0 | ||
| 612 | } | ||
| 613 | |||
| 614 | // A 4-byte match has been found. We'll later see if more than 4 bytes | ||
| 615 | // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit | ||
| 616 | // them as literal bytes. | ||
| 617 | |||
| 618 | d += emitLiteralSize(src[nextEmit:s]) | ||
| 619 | |||
| 620 | // Call emitCopy, and then see if another emitCopy could be our next | ||
| 621 | // move. Repeat until we find no match for the input immediately after | ||
| 622 | // what was consumed by the last emitCopy call. | ||
| 623 | // | ||
| 624 | // If we exit this loop normally then we need to call emitLiteral next, | ||
| 625 | // though we don't yet know how big the literal will be. We handle that | ||
| 626 | // by proceeding to the next iteration of the main loop. We also can | ||
| 627 | // exit this loop via goto if we get close to exhausting the input. | ||
| 628 | for { | ||
| 629 | // Invariant: we have a 4-byte match at s, and no need to emit any | ||
| 630 | // literal bytes prior to s. | ||
| 631 | base := s | ||
| 632 | repeat = base - candidate | ||
| 633 | |||
| 634 | // Extend the 4-byte match as long as possible. | ||
| 635 | s += 4 | ||
| 636 | candidate += 4 | ||
| 637 | for s <= len(src)-8 { | ||
| 638 | if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { | ||
| 639 | s += bits.TrailingZeros64(diff) >> 3 | ||
| 640 | break | ||
| 641 | } | ||
| 642 | s += 8 | ||
| 643 | candidate += 8 | ||
| 644 | } | ||
| 645 | |||
| 646 | d += emitCopyNoRepeatSize(repeat, s-base) | ||
| 647 | if false { | ||
| 648 | // Validate match. | ||
| 649 | a := src[base:s] | ||
| 650 | b := src[base-repeat : base-repeat+(s-base)] | ||
| 651 | if !bytes.Equal(a, b) { | ||
| 652 | panic("mismatch") | ||
| 653 | } | ||
| 654 | } | ||
| 655 | |||
| 656 | nextEmit = s | ||
| 657 | if s >= sLimit { | ||
| 658 | goto emitRemainder | ||
| 659 | } | ||
| 660 | |||
| 661 | if d > dstLimit { | ||
| 662 | // Do we have space for more, if not bail. | ||
| 663 | return 0 | ||
| 664 | } | ||
| 665 | // Check for an immediate match, otherwise start search at s+1 | ||
| 666 | x := load64(src, s-2) | ||
| 667 | m2Hash := hash6(x, tableBits) | ||
| 668 | currHash := hash6(x>>16, tableBits) | ||
| 669 | candidate = int(table[currHash]) | ||
| 670 | table[m2Hash] = uint32(s - 2) | ||
| 671 | table[currHash] = uint32(s) | ||
| 672 | if uint32(x>>16) != load32(src, candidate) { | ||
| 673 | cv = load64(src, s+1) | ||
| 674 | s++ | ||
| 675 | break | ||
| 676 | } | ||
| 677 | } | ||
| 678 | } | ||
| 679 | |||
| 680 | emitRemainder: | ||
| 681 | if nextEmit < len(src) { | ||
| 682 | // Bail if we exceed the maximum size. | ||
| 683 | if d+len(src)-nextEmit > dstLimit { | ||
| 684 | return 0 | ||
| 685 | } | ||
| 686 | d += emitLiteralSize(src[nextEmit:]) | ||
| 687 | } | ||
| 688 | return d | ||
| 689 | } | ||
| 690 | |||
| 691 | // emitLiteral writes a literal chunk and returns the number of bytes written. | ||
| 692 | // | ||
| 693 | // It assumes that: | ||
| 694 | // | ||
| 695 | // dst is long enough to hold the encoded bytes | ||
| 696 | // 0 <= len(lit) && len(lit) <= math.MaxUint32 | ||
| 697 | func emitLiteralSize(lit []byte) int { | ||
| 698 | if len(lit) == 0 { | ||
| 699 | return 0 | ||
| 700 | } | ||
| 701 | switch { | ||
| 702 | case len(lit) <= 60: | ||
| 703 | return len(lit) + 1 | ||
| 704 | case len(lit) <= 1<<8: | ||
| 705 | return len(lit) + 2 | ||
| 706 | case len(lit) <= 1<<16: | ||
| 707 | return len(lit) + 3 | ||
| 708 | case len(lit) <= 1<<24: | ||
| 709 | return len(lit) + 4 | ||
| 710 | default: | ||
| 711 | return len(lit) + 5 | ||
| 712 | } | ||
| 713 | } | ||
| 714 | |||
| 715 | func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { | ||
| 716 | panic("cvtLZ4BlockAsm should be unreachable") | ||
| 717 | } | ||
| 718 | |||
| 719 | func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { | ||
| 720 | panic("cvtLZ4BlockSnappyAsm should be unreachable") | ||
| 721 | } | ||
| 722 | |||
| 723 | func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { | ||
| 724 | panic("cvtLZ4sBlockAsm should be unreachable") | ||
| 725 | } | ||
| 726 | |||
| 727 | func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { | ||
| 728 | panic("cvtLZ4sBlockSnappyAsm should be unreachable") | ||
| 729 | } | ||