diff options
Diffstat (limited to 'vendor/github.com/klauspost/compress/snappy')
| -rw-r--r-- | vendor/github.com/klauspost/compress/snappy/decode_amd64.s | 16 | ||||
| -rw-r--r-- | vendor/github.com/klauspost/compress/snappy/decode_other.go | 22 |
2 files changed, 22 insertions, 16 deletions
diff --git a/vendor/github.com/klauspost/compress/snappy/decode_amd64.s b/vendor/github.com/klauspost/compress/snappy/decode_amd64.s index e6179f65e..1c66e3723 100644 --- a/vendor/github.com/klauspost/compress/snappy/decode_amd64.s +++ b/vendor/github.com/klauspost/compress/snappy/decode_amd64.s @@ -184,9 +184,7 @@ tagLit60Plus: // checks. In the asm version, we code it once instead of once per switch case. ADDQ CX, SI SUBQ $58, SI - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ SI, R13 JA errCorrupt // case x == 60: @@ -232,9 +230,7 @@ tagCopy4: ADDQ $5, SI // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ SI, R13 JA errCorrupt // length = 1 + int(src[s-5])>>2 @@ -251,9 +247,7 @@ tagCopy2: ADDQ $3, SI // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ SI, R13 JA errCorrupt // length = 1 + int(src[s-3])>>2 @@ -277,9 +271,7 @@ tagCopy: ADDQ $2, SI // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ SI, R13 JA errCorrupt // offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) diff --git a/vendor/github.com/klauspost/compress/snappy/decode_other.go b/vendor/github.com/klauspost/compress/snappy/decode_other.go index 8c9f2049b..94a96c5d7 100644 --- a/vendor/github.com/klauspost/compress/snappy/decode_other.go +++ b/vendor/github.com/klauspost/compress/snappy/decode_other.go @@ -85,14 +85,28 @@ func decode(dst, src []byte) int { if offset <= 0 || d < offset || length > len(dst)-d { return decodeErrCodeCorrupt } - // Copy from an earlier sub-slice of dst to a later sub-slice. Unlike - // the built-in copy function, this byte-by-byte copy always runs + // Copy from an earlier sub-slice of dst to a later sub-slice. + // If no overlap, use the built-in copy: + if offset > length { + copy(dst[d:d+length], dst[d-offset:]) + d += length + continue + } + + // Unlike the built-in copy function, this byte-by-byte copy always runs // forwards, even if the slices overlap. Conceptually, this is: // // d += forwardCopy(dst[d:d+length], dst[d-offset:]) - for end := d + length; d != end; d++ { - dst[d] = dst[d-offset] + // + // We align the slices into a and b and show the compiler they are the same size. + // This allows the loop to run without bounds checks. + a := dst[d : d+length] + b := dst[d-offset:] + b = b[:len(a)] + for i := range a { + a[i] = b[i] } + d += length } if d != len(dst) { return decodeErrCodeCorrupt |