548 lines
15 KiB
C++
548 lines
15 KiB
C++
//========= Copyright Valve Corporation, All rights reserved. ============//
|
|
//
|
|
// Purpose:
|
|
//
|
|
// $NoKeywords: $
|
|
//=============================================================================//
|
|
|
|
#include "tier0/platform.h"
|
|
#include "tier0/dbg.h"
|
|
#include "tier1/diff.h"
|
|
#include "mathlib/mathlib.h"
|
|
|
|
// memdbgon must be the last include file in a .cpp file!!!
|
|
#include "tier0/memdbgon.h"
|
|
|
|
|
|
// format of diff output:
|
|
// 0NN (N=1..127) copy next N literaly
|
|
//
|
|
// 1NN (N=1..127) ofs (-128..127) copy next N bytes from original, changin offset by N bytes from
|
|
// last copy end
|
|
// 100 N ofs(-32768..32767) copy next N, with larger delta offset
|
|
// 00 NNNN(1..65535) ofs(-32768..32767) big copy from old
|
|
// 80 00 NN NN NN big raw copy
|
|
//
|
|
// available codes (could be used for additonal compression ops)
|
|
// long offset form whose offset could have fit in short offset
|
|
|
|
// note - this algorithm uses storage equal to 8* the old buffer size. 64k=.5mb
|
|
|
|
|
|
#define MIN_MATCH_LEN 8
|
|
#define ACCEPTABLE_MATCH_LEN 4096
|
|
|
|
struct BlockPtr
|
|
{
|
|
BlockPtr *Next;
|
|
uint8 const *dataptr;
|
|
};
|
|
|
|
template<class T,class V> static inline void AddToHead(T * & head, V * node)
|
|
{
|
|
node->Next=head;
|
|
head=node;
|
|
}
|
|
|
|
void Fail(char const *msg)
|
|
{
|
|
Assert(0);
|
|
}
|
|
|
|
void ApplyDiffs(uint8 const *OldBlock, uint8 const *DiffList,
|
|
int OldSize, int DiffListSize, int &ResultListSize,uint8 *Output,uint32 OutSize)
|
|
{
|
|
uint8 const *copy_src=OldBlock;
|
|
uint8 const *end_of_diff_list=DiffList+DiffListSize;
|
|
uint8 const *obuf=Output;
|
|
while(DiffList<end_of_diff_list)
|
|
{
|
|
// printf("dptr=%x ",DiffList-d);
|
|
uint8 op=*(DiffList++);
|
|
if (op==0)
|
|
{
|
|
uint16 copy_sz=DiffList[0]+256*DiffList[1];
|
|
int copy_ofs=DiffList[2]+DiffList[3]*256;
|
|
if (copy_ofs>32767)
|
|
copy_ofs|=0xffff0000;
|
|
// printf("long cp from %x to %x len=%d\n", copy_src+copy_ofs-OldBlock,Output-obuf,copy_sz);
|
|
|
|
memcpy(Output,copy_src+copy_ofs,copy_sz);
|
|
Output+=copy_sz;
|
|
copy_src=copy_src+copy_ofs+copy_sz;
|
|
DiffList+=4;
|
|
}
|
|
else
|
|
{
|
|
if (op & 0x80)
|
|
{
|
|
int copy_sz=op & 0x7f;
|
|
int copy_ofs;
|
|
if (copy_sz==0)
|
|
{
|
|
copy_sz=DiffList[0];
|
|
if (copy_sz==0)
|
|
{
|
|
// big raw copy
|
|
copy_sz=DiffList[1]+256*DiffList[2]+65536*DiffList[3];
|
|
memcpy(Output,DiffList+4,copy_sz);
|
|
// printf("big rawcopy to %x len=%d\n", Output-obuf,copy_sz);
|
|
|
|
DiffList+=copy_sz+4;
|
|
Output+=copy_sz;
|
|
}
|
|
else
|
|
{
|
|
copy_ofs=DiffList[1]+(DiffList[2]*256);
|
|
if (copy_ofs>32767)
|
|
copy_ofs|=0xffff0000;
|
|
// printf("long ofs cp from %x to %x len=%d\n", copy_src+copy_ofs-OldBlock,Output-obuf,copy_sz);
|
|
|
|
memcpy(Output,copy_src+copy_ofs,copy_sz);
|
|
Output+=copy_sz;
|
|
copy_src=copy_src+copy_ofs+copy_sz;
|
|
DiffList+=3;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
copy_ofs=DiffList[0];
|
|
if (copy_ofs>127)
|
|
copy_ofs|=0xffffff80;
|
|
// printf("cp from %x to %x len=%d\n", copy_src+copy_ofs-OldBlock,Output-obuf,copy_sz);
|
|
|
|
memcpy(Output,copy_src+copy_ofs,copy_sz);
|
|
Output+=copy_sz;
|
|
copy_src=copy_src+copy_ofs+copy_sz;
|
|
DiffList++;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// printf("raw copy %d to %x\n",op & 127,Output-obuf);
|
|
memcpy(Output,DiffList,op & 127);
|
|
Output+=op & 127;
|
|
DiffList+=(op & 127);
|
|
}
|
|
}
|
|
}
|
|
ResultListSize=Output-obuf;
|
|
|
|
}
|
|
|
|
static void CopyPending(int len, uint8 const *rawbytes,uint8 * &outbuf, uint8 const *limit)
|
|
{
|
|
// printf("copy raw len=%d\n",len);
|
|
if (len<128)
|
|
{
|
|
if (limit-outbuf < len+1)
|
|
Fail("diff buffer overrun");
|
|
*(outbuf++)=len;
|
|
memcpy(outbuf,rawbytes,len);
|
|
outbuf+=len;
|
|
}
|
|
else
|
|
{
|
|
if (limit-outbuf < len+5)
|
|
Fail("diff buffer overrun");
|
|
*(outbuf++)=0x80;
|
|
*(outbuf++)=0x00;
|
|
*(outbuf++)=(len & 255);
|
|
*(outbuf++)=((len>>8) & 255);
|
|
*(outbuf++)=((len>>16) & 255);
|
|
memcpy(outbuf,rawbytes,len);
|
|
outbuf+=len;
|
|
}
|
|
}
|
|
|
|
static uint32 hasher(uint8 const *mdata)
|
|
{
|
|
// attempt to scramble the bits of h1 and h2 together
|
|
uint32 ret=0;
|
|
for(int i=0;i<MIN_MATCH_LEN;i++)
|
|
{
|
|
ret=ret<<4;
|
|
ret+=(*mdata++);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
int FindDiffsForLargeFiles(uint8 const *NewBlock, uint8 const *OldBlock,
|
|
int NewSize, int OldSize, int &DiffListSize,uint8 *Output,
|
|
uint32 OutSize,
|
|
int hashsize)
|
|
{
|
|
|
|
int ret=0;
|
|
if (OldSize!=NewSize)
|
|
ret=1;
|
|
// first, build the hash table
|
|
BlockPtr **HashedMatches=new BlockPtr* [hashsize];
|
|
memset(HashedMatches,0,sizeof(HashedMatches[0])*hashsize);
|
|
BlockPtr *Blocks=0;
|
|
if (OldSize)
|
|
Blocks=new BlockPtr[OldSize];
|
|
BlockPtr *FreeList=Blocks;
|
|
// now, build the hash table
|
|
uint8 const *walk=OldBlock;
|
|
if (OldBlock && OldSize)
|
|
while(walk<OldBlock+OldSize-MIN_MATCH_LEN)
|
|
{
|
|
uint32 hash1=hasher(walk);
|
|
hash1 &=(hashsize-1);
|
|
BlockPtr *newnode=FreeList;
|
|
FreeList++;
|
|
newnode->dataptr=walk;
|
|
AddToHead(HashedMatches[hash1],newnode);
|
|
walk++;
|
|
}
|
|
else
|
|
ret=1;
|
|
// now, we have the hash table which may be used to search. begin the output step
|
|
int pending_raw_len=0;
|
|
walk=NewBlock;
|
|
uint8 *outbuf=Output;
|
|
uint8 const *lastmatchend=OldBlock;
|
|
while(walk<NewBlock+NewSize)
|
|
{
|
|
int longest=0;
|
|
BlockPtr *longest_block=0;
|
|
if (walk<NewBlock+NewSize-MIN_MATCH_LEN)
|
|
{
|
|
// check for a match
|
|
uint32 hash1=hasher(walk);
|
|
hash1 &= (hashsize-1);
|
|
// now, find the longest match in the hash table. If we find one >MIN_MATCH_LEN, take it
|
|
for(BlockPtr *b=HashedMatches[hash1];b;b=b->Next)
|
|
{
|
|
// find the match length
|
|
int match_of=b->dataptr-lastmatchend;
|
|
if ((match_of>-32768) && (match_of<32767))
|
|
{
|
|
int max_mlength=min(65535,(int)((ptrdiff_t)OldBlock+OldSize-(ptrdiff_t)b->dataptr));
|
|
max_mlength=min(max_mlength,(int)((ptrdiff_t)NewBlock+NewSize-(ptrdiff_t)walk));
|
|
int i;
|
|
for(i=0;i<max_mlength;i++)
|
|
if (walk[i]!=b->dataptr[i])
|
|
break;
|
|
if ((i>MIN_MATCH_LEN) && (i>longest))
|
|
{
|
|
longest=i;
|
|
longest_block=b;
|
|
if (longest>ACCEPTABLE_MATCH_LEN)
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
// now, we have a match maybe
|
|
if (longest_block)
|
|
{
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
// now, output copy block
|
|
int match_of=longest_block->dataptr-lastmatchend;
|
|
int nremaining=OutSize-(outbuf-Output);
|
|
|
|
if (match_of)
|
|
ret=1;
|
|
// printf("copy from %x to %x len=%d\n", match_of,outbuf-Output,longest);
|
|
if (longest>127)
|
|
{
|
|
// use really long encoding
|
|
if (nremaining<5)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=00;
|
|
*(outbuf++)=(longest & 255);
|
|
*(outbuf++)=((longest>>8) & 255);
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
|
|
}
|
|
else
|
|
{
|
|
if ((match_of>=-128) && (match_of<128))
|
|
{
|
|
if (nremaining<2)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=128+longest;
|
|
*(outbuf++)=(match_of&255);
|
|
}
|
|
else
|
|
{
|
|
// use long encoding
|
|
if (nremaining<4)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=0x80;
|
|
*(outbuf++)=longest;
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
}
|
|
}
|
|
lastmatchend=longest_block->dataptr+longest;
|
|
walk+=longest;
|
|
}
|
|
else
|
|
{
|
|
walk++;
|
|
pending_raw_len++;
|
|
}
|
|
}
|
|
// now, flush pending raw copy
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
delete[] HashedMatches;
|
|
if (Blocks)
|
|
delete[] Blocks;
|
|
DiffListSize=outbuf-Output;
|
|
return ret;
|
|
}
|
|
|
|
|
|
int FindDiffs(uint8 const *NewBlock, uint8 const *OldBlock,
|
|
int NewSize, int OldSize, int &DiffListSize,uint8 *Output,uint32 OutSize)
|
|
{
|
|
|
|
int ret=0;
|
|
if (OldSize!=NewSize)
|
|
ret=1;
|
|
// first, build the hash table
|
|
BlockPtr *HashedMatches[65536];
|
|
memset(HashedMatches,0,sizeof(HashedMatches));
|
|
BlockPtr *Blocks=0;
|
|
if (OldSize)
|
|
Blocks=new BlockPtr[OldSize];
|
|
BlockPtr *FreeList=Blocks;
|
|
// now, build the hash table
|
|
uint8 const *walk=OldBlock;
|
|
if (OldBlock && OldSize)
|
|
while(walk<OldBlock+OldSize-MIN_MATCH_LEN)
|
|
{
|
|
uint16 hash1=*((uint16 const *) walk)+*((uint16 const *) walk+2);
|
|
BlockPtr *newnode=FreeList;
|
|
FreeList++;
|
|
newnode->dataptr=walk;
|
|
AddToHead(HashedMatches[hash1],newnode);
|
|
walk++;
|
|
}
|
|
else
|
|
ret=1;
|
|
// now, we have the hash table which may be used to search. begin the output step
|
|
int pending_raw_len=0;
|
|
walk=NewBlock;
|
|
uint8 *outbuf=Output;
|
|
uint8 const *lastmatchend=OldBlock;
|
|
while(walk<NewBlock+NewSize)
|
|
{
|
|
int longest=0;
|
|
BlockPtr *longest_block=0;
|
|
if (walk<NewBlock+NewSize-MIN_MATCH_LEN)
|
|
{
|
|
// check for a match
|
|
uint16 hash1=*((uint16 const *) walk)+*((uint16 const *) walk+2);
|
|
// now, find the longest match in the hash table. If we find one >MIN_MATCH_LEN, take it
|
|
for(BlockPtr *b=HashedMatches[hash1];b;b=b->Next)
|
|
{
|
|
// find the match length
|
|
int match_of=b->dataptr-lastmatchend;
|
|
if ((match_of>-32768) && (match_of<32767))
|
|
{
|
|
int max_mlength=min(65535,(int)((ptrdiff_t)OldBlock+OldSize-(ptrdiff_t)b->dataptr));
|
|
max_mlength=min(max_mlength,(int)((ptrdiff_t)NewBlock+NewSize-(ptrdiff_t)walk));
|
|
int i;
|
|
for(i=0;i<max_mlength;i++)
|
|
if (walk[i]!=b->dataptr[i])
|
|
break;
|
|
if ((i>MIN_MATCH_LEN) && (i>longest))
|
|
{
|
|
longest=i;
|
|
longest_block=b;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
// now, we have a match maybe
|
|
if (longest_block)
|
|
{
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
// now, output copy block
|
|
int match_of=longest_block->dataptr-lastmatchend;
|
|
int nremaining=OutSize-(outbuf-Output);
|
|
if (match_of)
|
|
ret=1;
|
|
if (longest>127)
|
|
{
|
|
// use really long encoding
|
|
if (nremaining<5)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=00;
|
|
*(outbuf++)=(longest & 255);
|
|
*(outbuf++)=((longest>>8) & 255);
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
}
|
|
else
|
|
{
|
|
if ((match_of>=-128) && (match_of<128))
|
|
{
|
|
if (nremaining<2)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=128+longest;
|
|
*(outbuf++)=(match_of&255);
|
|
}
|
|
else
|
|
{
|
|
// use long encoding
|
|
if (nremaining<4)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=0x80;
|
|
*(outbuf++)=longest;
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
}
|
|
}
|
|
lastmatchend=longest_block->dataptr+longest;
|
|
walk+=longest;
|
|
}
|
|
else
|
|
{
|
|
walk++;
|
|
pending_raw_len++;
|
|
}
|
|
}
|
|
// now, flush pending raw copy
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
if (Blocks)
|
|
delete[] Blocks;
|
|
DiffListSize=outbuf-Output;
|
|
return ret;
|
|
}
|
|
|
|
|
|
int FindDiffsLowMemory(uint8 const *NewBlock, uint8 const *OldBlock,
|
|
int NewSize, int OldSize, int &DiffListSize,uint8 *Output,uint32 OutSize)
|
|
{
|
|
|
|
int ret=0;
|
|
if (OldSize!=NewSize)
|
|
ret=1;
|
|
uint8 const *old_data_hash[256];
|
|
memset(old_data_hash,0,sizeof(old_data_hash));
|
|
int pending_raw_len=0;
|
|
uint8 const *walk=NewBlock;
|
|
uint8 const *oldptr=OldBlock;
|
|
uint8 *outbuf=Output;
|
|
uint8 const *lastmatchend=OldBlock;
|
|
while(walk<NewBlock+NewSize)
|
|
{
|
|
while( (oldptr-OldBlock<walk-NewBlock+40) && (oldptr-OldBlock<OldSize-MIN_MATCH_LEN))
|
|
{
|
|
uint16 hash1=(oldptr[0]+oldptr[1]+oldptr[2]+oldptr[3]) & (NELEMS(old_data_hash)-1);
|
|
old_data_hash[hash1]=oldptr;
|
|
oldptr++;
|
|
}
|
|
int longest=0;
|
|
uint8 const *longest_block=0;
|
|
if (walk<NewBlock+NewSize-MIN_MATCH_LEN)
|
|
{
|
|
// check for a match
|
|
uint16 hash1=(walk[0]+walk[1]+walk[2]+walk[3]) & (NELEMS(old_data_hash)-1);
|
|
if (old_data_hash[hash1])
|
|
{
|
|
int max_bytes_to_compare=min(NewBlock+NewSize-walk,OldBlock+OldSize-old_data_hash[hash1]);
|
|
int nmatches;
|
|
for(nmatches=0;nmatches<max_bytes_to_compare;nmatches++)
|
|
if (walk[nmatches]!=old_data_hash[hash1][nmatches])
|
|
break;
|
|
if (nmatches>MIN_MATCH_LEN)
|
|
{
|
|
longest_block=old_data_hash[hash1];
|
|
longest=nmatches;
|
|
}
|
|
}
|
|
}
|
|
// now, we have a match maybe
|
|
if (longest_block)
|
|
{
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
// now, output copy block
|
|
int match_of=longest_block-lastmatchend;
|
|
int nremaining=OutSize-(outbuf-Output);
|
|
if (match_of)
|
|
ret=1;
|
|
if (longest>127)
|
|
{
|
|
// use really long encoding
|
|
if (nremaining<5)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=00;
|
|
*(outbuf++)=(longest & 255);
|
|
*(outbuf++)=((longest>>8) & 255);
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
}
|
|
else
|
|
{
|
|
if ((match_of>=-128) && (match_of<128))
|
|
{
|
|
if (nremaining<2)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=128+longest;
|
|
*(outbuf++)=(match_of&255);
|
|
}
|
|
else
|
|
{
|
|
// use long encoding
|
|
if (nremaining<4)
|
|
Fail("diff buff needs increase");
|
|
*(outbuf++)=0x80;
|
|
*(outbuf++)=longest;
|
|
*(outbuf++)=(match_of & 255);
|
|
*(outbuf++)=((match_of>>8) & 255);
|
|
}
|
|
}
|
|
lastmatchend=longest_block+longest;
|
|
walk+=longest;
|
|
}
|
|
else
|
|
{
|
|
walk++;
|
|
pending_raw_len++;
|
|
}
|
|
}
|
|
// now, flush pending raw copy
|
|
if (pending_raw_len) // must output
|
|
{
|
|
ret=1;
|
|
CopyPending(pending_raw_len,walk-pending_raw_len,outbuf,Output+OutSize);
|
|
pending_raw_len=0;
|
|
}
|
|
DiffListSize=outbuf-Output;
|
|
return ret;
|
|
}
|
|
|
|
|