Literals header fields use little endian convention

This commit is contained in:
Yann Collet 2016-07-20 20:12:24 +02:00
parent 6fa05a2371
commit 198e6aac44
3 changed files with 95 additions and 89 deletions

View File

@ -572,17 +572,14 @@ static size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void
switch(flSize)
{
case 1: /* 2 - 1 - 5 */
ostart[0] = (BYTE)((lbt_raw<<6) + (0<<5) + srcSize);
ostart[0] = (BYTE)((U32)lbt_raw + (srcSize<<3));
break;
case 2: /* 2 - 2 - 12 */
ostart[0] = (BYTE)((lbt_raw<<6) + (2<<4) + (srcSize >> 8));
ostart[1] = (BYTE)srcSize;
MEM_writeLE16(ostart, (U32)lbt_raw + (1<<2) + (srcSize<<4));
break;
default: /*note : should not be necessary : flSize is within {1,2,3} */
case 3: /* 2 - 2 - 20 */
ostart[0] = (BYTE)((lbt_raw<<6) + (3<<4) + (srcSize >> 16));
ostart[1] = (BYTE)(srcSize>>8);
ostart[2] = (BYTE)srcSize;
MEM_writeLE32(ostart, (U32)lbt_raw + (3<<2) + (srcSize<<4));
break;
}
@ -595,22 +592,19 @@ static size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, cons
BYTE* const ostart = (BYTE* const)dst;
U32 const flSize = 1 + (srcSize>31) + (srcSize>4095);
(void)dstCapacity; /* dstCapacity guaranteed to be >=4, hence large enough */
(void)dstCapacity; /* dstCapacity already guaranteed to be >=4, hence large enough */
switch(flSize)
{
case 1: /* 2 - 1 - 5 */
ostart[0] = (BYTE)((lbt_rle<<6) + (0<<5) + srcSize);
ostart[0] = (BYTE)((U32)lbt_rle + (srcSize<<3));
break;
case 2: /* 2 - 2 - 12 */
ostart[0] = (BYTE)((lbt_rle<<6) + (2<<4) + (srcSize >> 8));
ostart[1] = (BYTE)srcSize;
MEM_writeLE16(ostart, (U32)lbt_rle + (1<<2) + (srcSize<<4));
break;
default: /*note : should not be necessary : flSize is necessarily within {1,2,3} */
case 3: /* 2 - 2 - 20 */
ostart[0] = (BYTE)((lbt_rle<<6) + (3<<4) + (srcSize >> 16));
ostart[1] = (BYTE)(srcSize>>8);
ostart[2] = (BYTE)srcSize;
MEM_writeLE32(ostart, (U32)lbt_rle + (3<<2) + (srcSize<<4));
break;
}
@ -658,24 +652,22 @@ static size_t ZSTD_compressLiterals (ZSTD_CCtx* zc,
switch(lhSize)
{
case 3: /* 2 - 2 - 10 - 10 */
ostart[0] = (BYTE)((srcSize>>6) + (singleStream << 4) + (hType<<6));
ostart[1] = (BYTE)((srcSize<<2) + (cLitSize>>8));
ostart[2] = (BYTE)(cLitSize);
break;
{ U32 const lhc = hType + (singleStream << 2) + (srcSize<<4) + (cLitSize<<14);
MEM_writeLE24(ostart, lhc);
break;
}
case 4: /* 2 - 2 - 14 - 14 */
ostart[0] = (BYTE)((srcSize>>10) + (2<<4) + (hType<<6));
ostart[1] = (BYTE)(srcSize>> 2);
ostart[2] = (BYTE)((srcSize<<6) + (cLitSize>>8));
ostart[3] = (BYTE)(cLitSize);
break;
{ U32 const lhc = hType + (2 << 2) + (srcSize<<4) + (cLitSize<<18);
MEM_writeLE32(ostart, lhc);
break;
}
default: /* should not be necessary, lhSize is only {3,4,5} */
case 5: /* 2 - 2 - 18 - 18 */
ostart[0] = (BYTE)((srcSize>>14) + (3<<4) + (hType<<6));
ostart[1] = (BYTE)(srcSize>>6);
ostart[2] = (BYTE)((srcSize<<2) + (cLitSize>>16));
ostart[3] = (BYTE)(cLitSize>>8);
ostart[4] = (BYTE)(cLitSize);
break;
{ U32 const lhc = hType + (3 << 2) + (srcSize<<4) + (cLitSize<<22);
MEM_writeLE32(ostart, lhc);
ostart[4] = (BYTE)(cLitSize >> 10);
break;
}
}
return lhSize+cLitSize;
}
@ -2735,8 +2727,7 @@ size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
BYTE* op = (BYTE*)dst;
size_t fhSize = 0;
/* not even init ! */
if (cctx->stage==0) return ERROR(stage_wrong);
if (cctx->stage==0) return ERROR(stage_wrong); /*< not even init ! */
/* special case : empty frame */
if (cctx->stage==1) {
@ -2748,7 +2739,7 @@ size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
}
/* frame epilogue */
if (dstCapacity < 3) return ERROR(dstSize_tooSmall);
if (dstCapacity < ZSTD_blockHeaderSize) return ERROR(dstSize_tooSmall);
{ U32 const checksum = cctx->params.fParams.checksumFlag ?
(U32)(XXH64_digest(&cctx->xxhState) >> 11) :
0;
@ -2756,7 +2747,7 @@ size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
}
cctx->stage = 0; /* return to "created but not init" status */
return 3+fhSize;
return ZSTD_blockHeaderSize+fhSize;
}

View File

@ -397,9 +397,9 @@ size_t ZSTD_getFrameParams(ZSTD_frameParams* fparamsPtr, const void* src, size_t
* compatible with legacy mode
* @return : decompressed size if known, 0 otherwise
note : 0 can mean any of the following :
- decompressed size is not provided within frame header
- decompressed size is not present within frame header
- frame header unknown / not supported
- frame header not completely provided (`srcSize` too small) */
- frame header not complete (`srcSize` too small) */
unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
{
#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT==1)
@ -464,33 +464,42 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected);
switch((litBlockType_t)(istart[0]>> 6))
switch((litBlockType_t)(istart[0] & 3))
{
case lbt_huffman:
{ size_t litSize, litCSize, singleStream=0;
U32 lhSize = (istart[0] >> 4) & 3;
{ size_t lhSize, litSize, litCSize, singleStream=0;
U32 const lhlCode = (istart[0] >> 2) & 3;
if (srcSize < 5) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for lhSize, + cSize (+nbSeq) */
switch(lhSize)
switch(lhlCode)
{
case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */
case 1:
singleStream = 1;
/* fall through */
case 0: default: /* note : default is impossible, since lhlCode into [0..3] */
/* 2 - 2 - 10 - 10 */
lhSize=3;
singleStream = istart[0] & 16;
litSize = ((istart[0] & 15) << 6) + (istart[1] >> 2);
litCSize = ((istart[1] & 3) << 8) + istart[2];
break;
{ U32 const lhc = MEM_readLE24(istart) >> 4;
lhSize = 3;
litSize = lhc & 0x3FF;
litCSize = lhc >> 10;
break;
}
case 2:
/* 2 - 2 - 14 - 14 */
lhSize=4;
litSize = ((istart[0] & 15) << 10) + (istart[1] << 2) + (istart[2] >> 6);
litCSize = ((istart[2] & 63) << 8) + istart[3];
break;
{ U32 const lhc = MEM_readLE32(istart) >> 4;
lhSize = 4;
litSize = lhc & 0x3FFF;
litCSize = lhc >> 14;
break;
}
case 3:
/* 2 - 2 - 18 - 18 */
lhSize=5;
litSize = ((istart[0] & 15) << 14) + (istart[1] << 6) + (istart[2] >> 2);
litCSize = ((istart[2] & 3) << 16) + (istart[3] << 8) + istart[4];
break;
{ U64 const lhc = (MEM_readLE32(istart) + (((U64)istart[4]) << 32)) >> 4;
lhSize = 5;
litSize = lhc & 0x3FFFF;
litCSize = lhc >> 18;
break;
}
}
if (litSize > ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(corruption_detected);
if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
@ -501,23 +510,23 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
return ERROR(corruption_detected);
dctx->litPtr = dctx->litBuffer;
dctx->litBufSize = ZSTD_BLOCKSIZE_ABSOLUTEMAX+8;
dctx->litBufSize = ZSTD_BLOCKSIZE_ABSOLUTEMAX+WILDCOPY_OVERLENGTH;
dctx->litSize = litSize;
dctx->litEntropy = 1;
return litCSize + lhSize;
}
case lbt_repeat:
{ size_t litSize, litCSize;
U32 lhSize = ((istart[0]) >> 4) & 3;
if (lhSize != 1) /* only case supported for now : small litSize, single stream */
{ size_t litSize, litCSize, lhSize;
U32 const lhc = MEM_readLE24(istart) >> 4;
if ((((istart[0]) >> 2) & 3) != 1) /* only case supported for now : small litSize, single stream */
return ERROR(corruption_detected);
if (dctx->litEntropy==0)
return ERROR(dictionary_corrupted);
/* 2 - 2 - 10 - 10 */
lhSize=3;
litSize = ((istart[0] & 15) << 6) + (istart[1] >> 2);
litCSize = ((istart[1] & 3) << 8) + istart[2];
lhSize = 3;
litSize = lhc & 0x3FF;
litCSize = lhc >> 10;
if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
{ size_t const errorCode = HUF_decompress1X4_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->hufTable);
@ -529,19 +538,21 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
return litCSize + lhSize;
}
case lbt_raw:
{ size_t litSize;
U32 lhSize = ((istart[0]) >> 4) & 3;
switch(lhSize)
{ size_t litSize, lhSize;
U32 const lhlCode = ((istart[0]) >> 2) & 3;
switch(lhlCode)
{
case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */
lhSize=1;
litSize = istart[0] & 31;
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
lhSize = 1;
litSize = istart[0] >> 3;
break;
case 2:
litSize = ((istart[0] & 15) << 8) + istart[1];
case 1:
lhSize = 2;
litSize = MEM_readLE16(istart) >> 4;
break;
case 3:
litSize = ((istart[0] & 15) << 16) + (istart[1] << 8) + istart[2];
lhSize = 3;
litSize = MEM_readLE24(istart) >> 4;
break;
}
@ -560,19 +571,21 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
return lhSize+litSize;
}
case lbt_rle:
{ size_t litSize;
U32 lhSize = ((istart[0]) >> 4) & 3;
switch(lhSize)
{ U32 const lhlCode = ((istart[0]) >> 2) & 3;
size_t litSize, lhSize;
switch(lhlCode)
{
case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
lhSize = 1;
litSize = istart[0] & 31;
litSize = istart[0] >> 3;
break;
case 2:
litSize = ((istart[0] & 15) << 8) + istart[1];
case 1:
lhSize = 2;
litSize = MEM_readLE16(istart) >> 4;
break;
case 3:
litSize = ((istart[0] & 15) << 16) + (istart[1] << 8) + istart[2];
lhSize = 3;
litSize = MEM_readLE24(istart) >> 4;
if (srcSize<4) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */
break;
}

View File

@ -435,15 +435,17 @@ followed by 1 or 4 streams.
Header is in charge of describing how literals are packed.
It's a byte-aligned variable-size bitfield, ranging from 1 to 5 bytes,
using big-endian convention.
using little-endian convention.
| BlockType | sizes format | [compressed size] | regenerated size |
| --------- | ------------ | ----------------- | ---------------- |
| 2 bits | 1 - 2 bits | 0 - 18 bits | 5 - 20 bits |
| BlockType | sizes format | regenerated size | [compressed size] |
| --------- | ------------ | ---------------- | ----------------- |
| 2 bits | 1 - 2 bits | 5 - 20 bits | 0 - 18 bits |
In this representation, bits on the left are smallest bits.
__Block Type__ :
This is a 2-bits field, describing 4 different block types :
This field uses 2 lowest bits of first byte, describing 4 different block types :
| Value | 0 | 1 | 2 | 3 |
| ---------- | ---------- | ------ | --- | ------- |
@ -466,19 +468,19 @@ Sizes format are divided into 2 families :
and the decompressed size. It will also decode the number of streams.
- For Raw or RLE blocks, it's enough to decode the size to regenerate.
For values spanning several bytes, convention is Big-endian.
For values spanning several bytes, convention is Little-endian.
__Sizes format for Raw or RLE literals block__ :
__Sizes format for Raw and RLE literals block__ :
- Value : 0x : Regenerated size uses 5 bits (0-31).
- Value : x0 : Regenerated size uses 5 bits (0-31).
Total literal header size is 1 byte.
`size = h[0] & 31;`
- Value : 10 : Regenerated size uses 12 bits (0-4095).
`size = h[0]>>3;`
- Value : 01 : Regenerated size uses 12 bits (0-4095).
Total literal header size is 2 bytes.
`size = ((h[0] & 15) << 8) + h[1];`
`size = (h[0]>>4) + (h[1]<<4);`
- Value : 11 : Regenerated size uses 20 bits (0-1048575).
Total literal header size is 3 bytes.
`size = ((h[0] & 15) << 16) + (h[1]<<8) + h[2];`
`size = (h[0]>>4) + (h[1]<<4) + (h[2]<<12);`
Note : it's allowed to represent a short value (ex : `13`)
using a long format, accepting the reduced compacity.
@ -499,7 +501,7 @@ Note : also applicable to "repeat-stats" blocks.
Compressed and regenerated sizes use 18 bits (0-262143).
Total literal header size is 5 bytes.
Compressed and regenerated size fields follow big endian convention.
Compressed and regenerated size fields follow little endian convention.
#### Huffman Tree description