diff options
Diffstat (limited to 'vendor/github.com/klauspost/compress/huff0/decompress_amd64.go')
| -rw-r--r-- | vendor/github.com/klauspost/compress/huff0/decompress_amd64.go | 95 |
1 files changed, 31 insertions, 64 deletions
diff --git a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go index ce8e93bcd..3415e5da2 100644 --- a/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go +++ b/vendor/github.com/klauspost/compress/huff0/decompress_amd64.go @@ -13,19 +13,30 @@ import ( // decompress4x_main_loop_x86 is an x86 assembler implementation // of Decompress4X when tablelog > 8. //go:noescape -func decompress4x_main_loop_x86(pbr0, pbr1, pbr2, pbr3 *bitReaderShifted, - peekBits uint8, buf *byte, tbl *dEntrySingle) uint8 +func decompress4x_main_loop_amd64(ctx *decompress4xContext) // decompress4x_8b_loop_x86 is an x86 assembler implementation // of Decompress4X when tablelog <= 8 which decodes 4 entries // per loop. //go:noescape -func decompress4x_8b_loop_x86(pbr0, pbr1, pbr2, pbr3 *bitReaderShifted, - peekBits uint8, buf *byte, tbl *dEntrySingle) uint8 +func decompress4x_8b_main_loop_amd64(ctx *decompress4xContext) // fallback8BitSize is the size where using Go version is faster. const fallback8BitSize = 800 +type decompress4xContext struct { + pbr0 *bitReaderShifted + pbr1 *bitReaderShifted + pbr2 *bitReaderShifted + pbr3 *bitReaderShifted + peekBits uint8 + out *byte + dstEvery int + tbl *dEntrySingle + decoded int + limit *byte +} + // Decompress4X will decompress a 4X encoded stream. // The length of the supplied input must match the end of a block exactly. // The *capacity* of the dst slice must match the destination size of @@ -42,6 +53,7 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) { if cap(dst) < fallback8BitSize && use8BitTables { return d.decompress4X8bit(dst, src) } + var br [4]bitReaderShifted // Decode "jump table" start := 6 @@ -71,70 +83,28 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) { const tlMask = tlSize - 1 single := d.dt.single[:tlSize] - // Use temp table to avoid bound checks/append penalty. - buf := d.buffer() - var off uint8 var decoded int - const debug = false - - // see: bitReaderShifted.peekBitsFast() - peekBits := uint8((64 - d.actualTableLog) & 63) - - // Decode 2 values from each decoder/loop. - const bufoff = 256 - for { - if br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4 { - break + if len(out) > 4*4 && !(br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4) { + ctx := decompress4xContext{ + pbr0: &br[0], + pbr1: &br[1], + pbr2: &br[2], + pbr3: &br[3], + peekBits: uint8((64 - d.actualTableLog) & 63), // see: bitReaderShifted.peekBitsFast() + out: &out[0], + dstEvery: dstEvery, + tbl: &single[0], + limit: &out[dstEvery-4], // Always stop decoding when first buffer gets here to avoid writing OOB on last. } - if use8BitTables { - off = decompress4x_8b_loop_x86(&br[0], &br[1], &br[2], &br[3], peekBits, &buf[0][0], &single[0]) + decompress4x_8b_main_loop_amd64(&ctx) } else { - off = decompress4x_main_loop_x86(&br[0], &br[1], &br[2], &br[3], peekBits, &buf[0][0], &single[0]) - } - if debug { - fmt.Print("DEBUG: ") - fmt.Printf("off=%d,", off) - for i := 0; i < 4; i++ { - fmt.Printf(" br[%d]={bitsRead=%d, value=%x, off=%d}", - i, br[i].bitsRead, br[i].value, br[i].off) - } - fmt.Println("") - } - - if off != 0 { - break + decompress4x_main_loop_amd64(&ctx) } - if bufoff > dstEvery { - d.bufs.Put(buf) - return nil, errors.New("corruption detected: stream overrun 1") - } - copy(out, buf[0][:]) - copy(out[dstEvery:], buf[1][:]) - copy(out[dstEvery*2:], buf[2][:]) - copy(out[dstEvery*3:], buf[3][:]) - out = out[bufoff:] - decoded += bufoff * 4 - // There must at least be 3 buffers left. - if len(out) < dstEvery*3 { - d.bufs.Put(buf) - return nil, errors.New("corruption detected: stream overrun 2") - } - } - if off > 0 { - ioff := int(off) - if len(out) < dstEvery*3+ioff { - d.bufs.Put(buf) - return nil, errors.New("corruption detected: stream overrun 3") - } - copy(out, buf[0][:off]) - copy(out[dstEvery:], buf[1][:off]) - copy(out[dstEvery*2:], buf[2][:off]) - copy(out[dstEvery*3:], buf[3][:off]) - decoded += int(off) * 4 - out = out[off:] + decoded = ctx.decoded + out = out[decoded/4:] } // Decode remaining. @@ -150,7 +120,6 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) { for bitsLeft > 0 { br.fill() if offset >= endsAt { - d.bufs.Put(buf) return nil, errors.New("corruption detected: stream overrun 4") } @@ -164,7 +133,6 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) { offset++ } if offset != endsAt { - d.bufs.Put(buf) return nil, fmt.Errorf("corruption detected: short output block %d, end %d != %d", i, offset, endsAt) } decoded += offset - dstEvery*i @@ -173,7 +141,6 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) { return nil, err } } - d.bufs.Put(buf) if dstSize != decoded { return nil, errors.New("corruption detected: short output block") } |