path: root/src/lzo1.c

/* lzo1.c -- implementation of the LZO1 algorithm

   This file is part of the LZO real-time data compression library.

   Copyright (C) 2011 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2010 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2009 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2008 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2007 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2006 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2005 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2004 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2003 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2002 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2001 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 2000 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 1999 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 1998 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 1997 Markus Franz Xaver Johannes Oberhumer
   Copyright (C) 1996 Markus Franz Xaver Johannes Oberhumer
   All Rights Reserved.

   The LZO library is free software; you can redistribute it and/or
   modify it under the terms of the GNU General Public License as
   published by the Free Software Foundation; either version 2 of
   the License, or (at your option) any later version.

   The LZO library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with the LZO library; see the file COPYING.
   If not, write to the Free Software Foundation, Inc.,
   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.

   Markus F.X.J. Oberhumer
   <markus@oberhumer.com>
   http://www.oberhumer.com/opensource/lzo/
 */


#include "lzo_conf.h"
#include "lzo/lzo1.h"


/***********************************************************************
// The next two defines can be changed to customize LZO1.
// The default version is LZO1-5/1.
************************************************************************/

/* run bits (3 - 5) - the compressor and the decompressor
 * must use the same value. */
#if !defined(RBITS)
#  define RBITS     5
#endif

/* compression level (1 - 9) - this only affects the compressor.
 * 1 is fastest, 9 is best compression ratio */
#if !defined(CLEVEL)
#  define CLEVEL    1           /* fastest by default */
#endif


/* check configuration */
#if (RBITS < 3 || RBITS > 5)
#  error "invalid RBITS"
#endif
#if (CLEVEL < 1 || CLEVEL > 9)
#  error "invalid CLEVEL"
#endif


/***********************************************************************
// You should not have to change anything below this line.
************************************************************************/

/*
     Format of the marker byte


     76543210
     --------
     00000000   a long run (a 'R0' run) - there are short and long R0 runs
     000rrrrr   a short run with len r
     mmmooooo   a short match (len = 2+m, o = offset low bits)
     111ooooo   a long match (o = offset low bits)
*/


#define RSIZE   (1 << RBITS)
#define RMASK   (RSIZE - 1)

#define OBITS   RBITS               /* offset and run-length use same bits */
#define OSIZE   (1 << OBITS)
#define OMASK   (OSIZE - 1)

#define MBITS   (8 - OBITS)
#define MSIZE   (1 << MBITS)
#define MMASK   (MSIZE - 1)


/* sanity checks */
#if (OBITS < 3 || OBITS > 5)
#  error "invalid OBITS"
#endif
#if (MBITS < 3 || MBITS > 5)
#  error "invalid MBITS"
#endif


/***********************************************************************
// some macros to improve readability
************************************************************************/

/* Minimum len of a match */
#define MIN_MATCH           3
#define THRESHOLD           (MIN_MATCH - 1)

/* Minimum len of match coded in 2 bytes */
#define MIN_MATCH_SHORT     MIN_MATCH

/* Maximum len of match coded in 2 bytes */
#define MAX_MATCH_SHORT     (THRESHOLD + (MSIZE - 2))
/* MSIZE - 2: 0 is used to indicate runs,
 *            MSIZE-1 is used to indicate a long match */

/* Minimum len of match coded in 3 bytes */
#define MIN_MATCH_LONG      (MAX_MATCH_SHORT + 1)

/* Maximum len of match coded in 3 bytes */
#define MAX_MATCH_LONG      (MIN_MATCH_LONG + 255)

/* Maximum offset of a match */
#define MAX_OFFSET          (1 << (8 + OBITS))


/*

RBITS | MBITS  MIN  THR.  MSIZE  MAXS  MINL  MAXL   MAXO  R0MAX R0FAST
======+===============================================================
  3   |   5      3    2     32    32    33    288   2048    263   256
  4   |   4      3    2     16    16    17    272   4096    271   264
  5   |   3      3    2      8     8     9    264   8192    287   280

 */


/***********************************************************************
// internal configuration
// all of these affect compression only
************************************************************************/

/* choose the hashing strategy */
#ifndef LZO_HASH
#define LZO_HASH    LZO_HASH_LZO_INCREMENTAL_A
#endif
#define D_INDEX1(d,p)       d = DM(DMUL(0x21,DX2(p,5,5)) >> 5)
#define D_INDEX2(d,p)       d = d ^ D_MASK

#define DBITS       (8 + RBITS)
#include "lzo_dict.h"
#define DVAL_LEN    DVAL_LOOKAHEAD


/***********************************************************************
// get algorithm info, return memory required for compression
************************************************************************/

LZO_EXTERN(lzo_uint) lzo1_info ( int *rbits, int *clevel );

LZO_PUBLIC(lzo_uint)
lzo1_info ( int *rbits, int *clevel )
{
    if (rbits)
        *rbits = RBITS;
    if (clevel)
        *clevel = CLEVEL;
    return D_SIZE * lzo_sizeof(lzo_bytep);
}


/***********************************************************************
// decode a R0 literal run (a long run)
************************************************************************/

#define R0MIN   (RSIZE)             /* Minimum len of R0 run of literals */
#define R0MAX   (R0MIN + 255)       /* Maximum len of R0 run of literals */
#define R0FAST  (R0MAX & ~7u)       /* R0MAX aligned to 8 byte boundary */

#if (R0MAX - R0FAST != 7) || ((R0FAST & 7) != 0)
#  error "something went wrong"
#endif

/* 7 special codes from R0FAST+1 .. R0MAX
 * these codes mean long R0 runs with lengths
 * 512, 1024, 2048, 4096, 8192, 16384, 32768 */


/***********************************************************************
// LZO1 decompress a block of data.
//
// Could be easily translated into assembly code.
************************************************************************/

LZO_PUBLIC(int)
lzo1_decompress  ( const lzo_bytep in , lzo_uint  in_len,
                         lzo_bytep out, lzo_uintp out_len,
                         lzo_voidp wrkmem )
{
    lzo_bytep op;
    const lzo_bytep ip;
    const lzo_bytep const ip_end = in + in_len;
    lzo_uint t;

    LZO_UNUSED(wrkmem);

    op = out;
    ip = in;
    while (ip < ip_end)
    {
        t = *ip++;  /* get marker */

        if (t < R0MIN)          /* a literal run */
        {
            if (t == 0)             /* a R0 literal run */
            {
                t = *ip++;
                if (t >= R0FAST - R0MIN)            /* a long R0 run */
                {
                    t -= R0FAST - R0MIN;
                    if (t == 0)
                        t = R0FAST;
                    else
                    {
#if 0
                        t = 256u << ((unsigned) t);
#else
                        /* help the optimizer */
                        lzo_uint tt = 256;
                        do tt <<= 1; while (--t > 0);
                        t = tt;
#endif
                    }
                    MEMCPY8_DS(op,ip,t);
                    continue;
                }
                t += R0MIN;
            }
            MEMCPY_DS(op,ip,t);
        }
        else                    /* a match */
        {
            lzo_uint tt;
            /* get match offset */
            const lzo_bytep m_pos = op - 1;
            m_pos -= (lzo_uint)(t & OMASK) | (((lzo_uint) *ip++) << OBITS);

            /* get match len */
            if (t >= ((MSIZE - 1) << OBITS))                /* all m-bits set */
                tt = (MIN_MATCH_LONG - THRESHOLD) + *ip++;  /* a long match */
            else
                tt = t >> OBITS;                            /* a short match */

            assert(m_pos >= out);
            assert(m_pos <  op);
            /* a half unrolled loop */
            *op++ = *m_pos++;
            *op++ = *m_pos++;
            MEMCPY_DS(op,m_pos,tt);
        }
    }

    *out_len = pd(op, out);

    /* the next line is the only check in the decompressor ! */
    return (ip == ip_end ? LZO_E_OK :
           (ip < ip_end  ? LZO_E_INPUT_NOT_CONSUMED : LZO_E_INPUT_OVERRUN));
}


/***********************************************************************
// code a literal run
************************************************************************/

static
#if LZO_ARCH_AVR
__lzo_noinline
#endif
lzo_bytep
store_run(lzo_bytep op, const lzo_bytep ii, lzo_uint r_len)
{
    assert(r_len > 0);

    /* code a long R0 run */
    if (r_len >= 512)
    {
        unsigned r_bits = 7;        /* 256 << 7 == 32768 */
        do {
            while (r_len >= (256u << r_bits))
            {
                r_len -= (256u << r_bits);
                *op++ = 0; *op++ = LZO_BYTE((R0FAST - R0MIN) + r_bits);
                MEMCPY8_DS(op, ii, (256u << r_bits));
            }
        } while (--r_bits > 0);
    }
    while (r_len >= R0FAST)
    {
        r_len -= R0FAST;
        *op++ = 0; *op++ = R0FAST - R0MIN;
        MEMCPY8_DS(op, ii, R0FAST);
    }

    if (r_len >= R0MIN)
    {
        /* code a short R0 run */
        *op++ = 0; *op++ = LZO_BYTE(r_len - R0MIN);
        MEMCPY_DS(op, ii, r_len);
    }
    else if (r_len > 0)
    {
        /* code a 'normal' run */
        *op++ = LZO_BYTE(r_len);
        MEMCPY_DS(op, ii, r_len);
    }

    assert(r_len == 0);
    return op;
}



/***********************************************************************
// LZO1 compress a block of data.
//
// Could be translated into assembly code without too much effort.
//
// I apologize for the spaghetti code, but it really helps the optimizer.
************************************************************************/

static int
do_compress    ( const lzo_bytep in , lzo_uint  in_len,
                       lzo_bytep out, lzo_uintp out_len,
                       lzo_voidp wrkmem )
{
    const lzo_bytep ip;
#if defined(__LZO_HASH_INCREMENTAL)
    lzo_xint dv;
#endif
    lzo_bytep op;
    const lzo_bytep m_pos;
    const lzo_bytep const ip_end = in+in_len - DVAL_LEN - MIN_MATCH_LONG;
    const lzo_bytep const in_end = in+in_len - DVAL_LEN;
    const lzo_bytep ii;
    lzo_dict_p const dict = (lzo_dict_p) wrkmem;

#if !defined(NDEBUG)
    const lzo_bytep m_pos_sav;
#endif

    op = out;
    ip = in;
    ii = ip;                /* point to start of literal run */
    if (in_len <= MIN_MATCH_LONG + DVAL_LEN + 1)
        goto the_end;

    /* init dictionary */
#if (LZO_DETERMINISTIC)
    BZERO8_PTR(wrkmem,sizeof(lzo_dict_t),D_SIZE);
#endif

    DVAL_FIRST(dv,ip);
    UPDATE_D(dict,0,dv,ip,in);
    ip++;
    DVAL_NEXT(dv,ip);

    do {
        LZO_DEFINE_UNINITIALIZED_VAR(lzo_uint, m_off, 0);
        lzo_uint dindex;

        DINDEX1(dindex,ip);
        GINDEX(m_pos,m_off,dict,dindex,in);
        if (LZO_CHECK_MPOS(m_pos,m_off,in,ip,MAX_OFFSET))
            goto literal;
        if (m_pos[0] == ip[0] && m_pos[1] == ip[1] && m_pos[2] == ip[2])
            goto match;
        DINDEX2(dindex,ip);
        GINDEX(m_pos,m_off,dict,dindex,in);
        if (LZO_CHECK_MPOS(m_pos,m_off,in,ip,MAX_OFFSET))
            goto literal;
        if (m_pos[0] == ip[0] && m_pos[1] == ip[1] && m_pos[2] == ip[2])
            goto match;
        goto literal;


literal:
        UPDATE_I(dict,0,dindex,ip,in);
        if (++ip >= ip_end)
            break;
        continue;

match:
        UPDATE_I(dict,0,dindex,ip,in);
#if !defined(NDEBUG) && (LZO_DICT_USE_PTR)
        m_pos_sav = m_pos;
#endif
        m_pos += 3;
        {
    /* we have found a match (of at least length 3) */
#if !defined(NDEBUG) && !(LZO_DICT_USE_PTR)
            assert((m_pos_sav = ip - m_off) == (m_pos - 3));
#endif
            /* 1) store the current literal run */
            if (pd(ip,ii) > 0)
            {
                lzo_uint t = pd(ip,ii);
#if 1
                /* OPTIMIZED: inline the copying of a short run */
                if (t < R0MIN)
                {
                    *op++ = LZO_BYTE(t);
                    MEMCPY_DS(op, ii, t);
                }
                else
#endif
                    op = store_run(op,ii,t);
            }

            /* 2a) compute match len */
            ii = ip;        /* point to start of current match */

            /* we already matched MIN_MATCH bytes,
             * m_pos also already advanced MIN_MATCH bytes */
            ip += MIN_MATCH;
            assert(m_pos < ip);

            /* try to match another MIN_MATCH_LONG - MIN_MATCH bytes
             * to see if we get a long match */

#define PS  *m_pos++ != *ip++

#if (MIN_MATCH_LONG - MIN_MATCH == 2)                   /* MBITS == 2 */
            if (PS || PS)
#elif (MIN_MATCH_LONG - MIN_MATCH == 6)                 /* MBITS == 3 */
            if (PS || PS || PS || PS || PS || PS)
#elif (MIN_MATCH_LONG - MIN_MATCH == 14)                /* MBITS == 4 */
            if (PS || PS || PS || PS || PS || PS || PS ||
                PS || PS || PS || PS || PS || PS || PS)
#elif (MIN_MATCH_LONG - MIN_MATCH == 30)                /* MBITS == 5 */
            if (PS || PS || PS || PS || PS || PS || PS || PS ||
                PS || PS || PS || PS || PS || PS || PS || PS ||
                PS || PS || PS || PS || PS || PS || PS || PS ||
                PS || PS || PS || PS || PS || PS)
#else
#  error "MBITS not yet implemented"
#endif
            {
                lzo_uint m_len;

            /* 2b) code a short match */
                    assert(pd(ip,m_pos) == m_off);
                --ip;   /* ran one too far, point back to non-match */
                m_len = pd(ip, ii);
                    assert(m_len >= MIN_MATCH_SHORT);
                    assert(m_len <= MAX_MATCH_SHORT);
                    assert(m_off > 0);
                    assert(m_off <= MAX_OFFSET);
                    assert(ii-m_off == m_pos_sav);
                    assert(lzo_memcmp(m_pos_sav,ii,m_len) == 0);
                --m_off;
                /* code short match len + low offset bits */
                *op++ = LZO_BYTE(((m_len - THRESHOLD) << OBITS) |
                                 (m_off & OMASK));
                /* code high offset bits */
                *op++ = LZO_BYTE(m_off >> OBITS);


            /* 2c) Insert phrases (beginning with ii+1) into the dictionary. */

#define SI      /* nothing */
#define DI      ++ii; DVAL_NEXT(dv,ii); UPDATE_D(dict,0,dv,ii,in);
#define XI      assert(ii < ip); ii = ip; DVAL_FIRST(dv,(ip));

#if (CLEVEL == 9) || (CLEVEL >= 7 && MBITS <= 4) || (CLEVEL >= 5 && MBITS <= 3)
            /* Insert the whole match (ii+1)..(ip-1) into dictionary.  */
                ++ii;
                do {
                    DVAL_NEXT(dv,ii);
                    UPDATE_D(dict,0,dv,ii,in);
                } while (++ii < ip);
                DVAL_NEXT(dv,ii);
                assert(ii == ip);
                DVAL_ASSERT(dv,ip);
#elif (CLEVEL >= 3)
                SI   DI DI   XI
#elif (CLEVEL >= 2)
                SI   DI      XI
#else
                             XI
#endif

            }
            else
            {
            /* we've found a long match - see how far we can still go */
                const lzo_bytep end;
                lzo_uint m_len;

                assert(ip <= in_end);
                assert(ii == ip - MIN_MATCH_LONG);

                if (pd(in_end,ip) <= (MAX_MATCH_LONG - MIN_MATCH_LONG))
                    end = in_end;
                else
                {
                    end = ip + (MAX_MATCH_LONG - MIN_MATCH_LONG);
                    assert(end < in_end);
                }

                while (ip < end  &&  *m_pos == *ip)
                    m_pos++, ip++;
                assert(ip <= in_end);

            /* 2b) code the long match */
                m_len = pd(ip, ii);
                    assert(m_len >= MIN_MATCH_LONG);
                    assert(m_len <= MAX_MATCH_LONG);
                    assert(m_off > 0);
                    assert(m_off <= MAX_OFFSET);
                    assert(ii-m_off == m_pos_sav);
                    assert(lzo_memcmp(m_pos_sav,ii,m_len) == 0);
                    assert(pd(ip,m_pos) == m_off);
                --m_off;
                /* code long match flag + low offset bits */
                *op++ = LZO_BYTE(((MSIZE - 1) << OBITS) | (m_off & OMASK));
                /* code high offset bits */
                *op++ = LZO_BYTE(m_off >> OBITS);
                /* code match len */
                *op++ = LZO_BYTE(m_len - MIN_MATCH_LONG);


            /* 2c) Insert phrases (beginning with ii+1) into the dictionary. */
#if (CLEVEL == 9)
            /* Insert the whole match (ii+1)..(ip-1) into dictionary.  */
            /* This is not recommended because it is slow. */
                ++ii;
                do {
                    DVAL_NEXT(dv,ii);
                    UPDATE_D(dict,0,dv,ii,in);
                } while (++ii < ip);
                DVAL_NEXT(dv,ii);
                assert(ii == ip);
                DVAL_ASSERT(dv,ip);
#elif (CLEVEL >= 8)
                SI   DI DI DI DI DI DI DI DI   XI
#elif (CLEVEL >= 7)
                SI   DI DI DI DI DI DI DI      XI
#elif (CLEVEL >= 6)
                SI   DI DI DI DI DI DI         XI
#elif (CLEVEL >= 5)
                SI   DI DI DI DI               XI
#elif (CLEVEL >= 4)
                SI   DI DI DI                  XI
#elif (CLEVEL >= 3)
                SI   DI DI                     XI
#elif (CLEVEL >= 2)
                SI   DI                        XI
#else
                                               XI
#endif
            }

            /* ii now points to the start of next literal run */
            assert(ii == ip);
        }
    } while (ip < ip_end);



the_end:
    assert(ip <= in_end);


#if defined(LZO_RETURN_IF_NOT_COMPRESSIBLE)
    /* return -1 if op == out to indicate that we
     * couldn't compress and didn't copy anything.
     */
    if (op == out)
    {
        *out_len = 0;
        return LZO_E_NOT_COMPRESSIBLE;
    }
#endif


    /* store the final literal run */
    if (pd(in_end+DVAL_LEN,ii) > 0)
        op = store_run(op,ii,pd(in_end+DVAL_LEN,ii));

    *out_len = pd(op, out);
    return 0;               /* compression went ok */
}


/***********************************************************************
// compress public entry point.
************************************************************************/

LZO_PUBLIC(int)
lzo1_compress ( const lzo_bytep in , lzo_uint  in_len,
                      lzo_bytep out, lzo_uintp out_len,
                      lzo_voidp wrkmem )
{
    int r = LZO_E_OK;

    /* don't try to compress a block that's too short */
    if (in_len == 0)
        *out_len = 0;
    else if (in_len <= MIN_MATCH_LONG + DVAL_LEN + 1)
    {
#if defined(LZO_RETURN_IF_NOT_COMPRESSIBLE)
        r = LZO_E_NOT_COMPRESSIBLE;
#else
        *out_len = pd(store_run(out,in,in_len), out);
#endif
    }
    else
        r = do_compress(in,in_len,out,out_len,wrkmem);

    return r;
}


/*
vi:ts=4:et
*/