/*
 * $RCSfile: MQCoder.java,v $
 * $Revision: 1.1 $
 * $Date: 2005/02/11 05:02:09 $
 * $State: Exp $
 *
 * Class:                   MQCoder
 *
 * Description:             Class that encodes a number of bits using the
 *                          MQ arithmetic coder
 *
 *
 *                          Diego SANTA CRUZ, Jul-26-1999 (improved speed)
 *
 * COPYRIGHT:
 *
 * This software module was originally developed by Raphaël Grosbois and
 * Diego Santa Cruz (Swiss Federal Institute of Technology-EPFL); Joel
 * Askelöf (Ericsson Radio Systems AB); and Bertrand Berthelot, David
 * Bouchard, Félix Henry, Gerard Mozelle and Patrice Onno (Canon Research
 * Centre France S.A) in the course of development of the JPEG2000
 * standard as specified by ISO/IEC 15444 (JPEG 2000 Standard). This
 * software module is an implementation of a part of the JPEG 2000
 * Standard. Swiss Federal Institute of Technology-EPFL, Ericsson Radio
 * Systems AB and Canon Research Centre France S.A (collectively JJ2000
 * Partners) agree not to assert against ISO/IEC and users of the JPEG
 * 2000 Standard (Users) any of their rights under the copyright, not
 * including other intellectual property rights, for this software module
 * with respect to the usage by ISO/IEC and Users of this software module
 * or modifications thereof for use in hardware or software products
 * claiming conformance to the JPEG 2000 Standard. Those intending to use
 * this software module in hardware or software products are advised that
 * their use may infringe existing patents. The original developers of
 * this software module, JJ2000 Partners and ISO/IEC assume no liability
 * for use of this software module or modifications thereof. No license
 * or right to this software module is granted for non JPEG 2000 Standard
 * conforming products. JJ2000 Partners have full right to use this
 * software module for his/her own purpose, assign or donate this
 * software module to any third party and to inhibit third parties from
 * using this software module for non JPEG 2000 Standard conforming
 * products. This copyright notice must be included in all copies or
 * derivative works of this software module.
 *
 * Copyright (c) 1999/2000 JJ2000 Partners.
 * */
package jj2000.j2k.entropy.encoder;

import jj2000.j2k.entropy.StdEntropyCoderOptions;
import jj2000.j2k.util.ArrayUtil;

/**
 * This class implements the MQ arithmetic coder. When initialized a specific
 * state can be specified for each context, which may be adapted to the
 * probability distribution that is expected for that context.
 *
 * <P>The type of length calculation and termination can be chosen at
 * construction time.
 *
 * ---- Tricks that have been tried to improve speed ----
 *
 * 1) Merging Qe and mPS and doubling the lookup tables
 *
 * Merge the mPS into Qe, as the sign bit (if Qe>=0 the sense of MPS is 0, if
 * Qe<0 the sense is 1), and double the lookup tables. The first half of the
 * lookup tables correspond to Qe>=0 (i.e. the sense of MPS is 0) and the
 * second half to Qe<0 (i.e. the sense of MPS is 1). The nLPS lookup table is
 * modified to incorporate the changes in the sense of MPS, by making it jump
 * from the first to the second half and vice-versa, when a change is
 * specified by the swicthLM lookup table. See JPEG book, section 13.2, page
 * 225.
 *
 * There is NO speed improvement in doing this, actually there is a slight
 * decrease, probably due to the fact that often Q has to be negated. Also the
 * fact that a brach of the type "if (bit==mPS[li])" is replaced by two
 * simpler braches of the type "if (bit==0)" and "if (q<0)" may contribute to
 * that.
 *
 * 2) Removing cT
 *
 * It is possible to remove the cT counter by setting a flag bit in the high
 * bits of the C register. This bit will be automatically shifted left
 * whenever a renormalization shift occurs, which is equivalent to decreasing
 * cT. When the flag bit reaches the sign bit (leftmost bit), which is
 * equivalenet to cT==0, the byteOut() procedure is called. This test can be
 * done efficiently with "c<0" since C is a signed quantity. Care must be
 * taken in byteOut() to reset the bit in order to not interfere with other
 * bits in the C register. See JPEG book, page 228.
 *
 * There is NO speed improvement in doing this. I don't really know why since
 * the number of operations whenever a renormalization occurs is
 * decreased. Maybe it is due to the number of extra operations in the
 * byteOut(), terminate() and getNumCodedBytes() procedures.
 *
 *
 * 3) Change the convention of MPS and LPS.
 *
 * Making the LPS interval be above the MPS interval (MQ coder convention is
 * the opposite) can reduce the number of operations along the MPS path. In
 * order to generate the same bit stream as with the MQ convention the output
 * bytes need to be modified accordingly. The basic rule for this is that C =
 * (C'^0xFF...FF)-A, where C is the codestream for the MQ convention and C' is
 * the codestream generated by this other convention. Note that this affects
 * bit-stuffing as well.
 *
 * This has not been tested yet.
 *
 * 4) Removing normalization while loop on MPS path
 *
 * Since in the MPS path Q is guaranteed to be always greater than 0x4000
 * (decimal 0.375) it is never necessary to do more than 1 renormalization
 * shift. Therefore the test of the while loop, and the loop itself, can be
 * removed.
 *
 * 5) Simplifying test on A register
 *
 * Since A is always less than or equal to 0xFFFF, the test "(a & 0x8000)==0"
 * can be replaced by the simplete test "a < 0x8000". This test is simpler in
 * Java since it involves only 1 operation (although the original test can be
 * converted to only one operation by  smart Just-In-Time compilers)
 *
 * This change has been integrated in the decoding procedures.
 *
 * 6) Speedup mode
 *
 * Implemented a method that uses the speedup mode of the MQ-coder if
 * possible. This should greately improve performance when coding long runs of
 * MPS symbols that have high probability. However, to take advantage of this,
 * the entropy coder implementation has to explicetely use it. The generated
 * bit stream is the same as if no speedup mode would have been used.
 *
 * Implemented but performance not tested yet.
 *
 * 7) Multiple-symbol coding
 *
 * Since the time spent in a method call is non-negligable, coding several
 * symbols with one method call reduces the overhead per coded symbol. The
 * decodeSymbols() method implements this. However, to take advantage of it,
 * the implementation of the entropy coder has to explicitely use it.
 *
 * Implemented but performance not tested yet.
 *  */
public class MQCoder {

    /** Identifier for the lazy length calculation. The lazy length
     * calculation is not optimal but is extremely simple. */
    public static final int LENGTH_LAZY = 0;

    /** Identifier for a very simple length calculation. This provides better
     * results than the 'LENGTH_LAZY' computation. This is the old length
     * calculation that was implemented in this class. */
    public static final int LENGTH_LAZY_GOOD = 1;

    /** Identifier for the near optimal length calculation. This calculation
     * is more complex than the lazy one but provides an almost optimal length
     * calculation. */
    public static final int LENGTH_NEAR_OPT = 2;

    /** The identifier fort the termination that uses a full flush. This is
     * the less efficient termination. */
    public static final int TERM_FULL = 0;

    /** The identifier for the termination that uses the near optimal length
     * calculation to terminate the arithmetic codewrod */
    public static final int TERM_NEAR_OPT = 1;

    /** The identifier for the easy termination that is simpler than the
     * 'TERM_NEAR_OPT' one but slightly less efficient. */
    public static final int TERM_EASY = 2;

    /** The identifier for the predictable termination policy for error
     * resilience. This is the same as the 'TERM_EASY' one but an special
     * sequence of bits is embodied in the spare bits for error resilience
     * purposes. */
    public static final int TERM_PRED_ER = 3;

    /** The data structures containing the probabilities for the LPS */
    final static
        int qe[]={0x5601, 0x3401, 0x1801, 0x0ac1, 0x0521, 0x0221, 0x5601,
                  0x5401, 0x4801, 0x3801, 0x3001, 0x2401, 0x1c01, 0x1601,
                  0x5601, 0x5401, 0x5101, 0x4801, 0x3801, 0x3401, 0x3001,
                  0x2801, 0x2401, 0x2201, 0x1c01, 0x1801, 0x1601, 0x1401,
                  0x1201, 0x1101, 0x0ac1, 0x09c1, 0x08a1, 0x0521, 0x0441,
                  0x02a1, 0x0221, 0x0141, 0x0111, 0x0085, 0x0049, 0x0025,
                  0x0015, 0x0009, 0x0005, 0x0001, 0x5601 };

    /** The indexes of the next MPS */
    final static
        int nMPS[]={ 1 , 2, 3, 4, 5,38, 7, 8, 9,10,11,12,13,29,15,16,17,
                     18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,
                     35,36,37,38,39,40,41,42,43,44,45,45,46 };

    /** The indexes of the next LPS */
    final static
        int nLPS[]={ 1 , 6, 9,12,29,33, 6,14,14,14,17,18,20,21,14,14,15,
                     16,17,18,19,19,20,21,22,23,24,25,26,27,28,29,30,31,
                     32,33,34,35,36,37,38,39,40,41,42,43,46 };

    /** Whether LPS and MPS should be switched */
    final static        // at indices 0, 6, and 14 we switch
        int switchLM[]={ 1,0,0,0,0,0,1,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,
                         0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 };
    // Having ints proved to be more efficient than booleans

    /** The ByteOutputBuffer used to write the compressed bit stream. */
    ByteOutputBuffer out;

    /** The current most probable signal for each context */
    int[] mPS;

    /** The current index of each context */
    int[] I;

    /** The current bit code */
    int c;

    /** The bit code counter */
    int cT;

    /** The current interval */
    int a;

    /** The last encoded byte of data */
    int b;

    /** If a 0xFF byte has been delayed and not yet been written to the output
     * (in the MQ we can never have more than 1 0xFF byte in a row). */
    boolean delFF;

    /** The number of written bytes so far, excluding any delayed 0xFF
     * bytes. Upon initialization it is -1 to indicated that the byte buffer
     * 'b' is empty as well. */
    int nrOfWrittenBytes = -1;

    /** The initial state of each context */
    int initStates[];

    /** The termination type to use. One of 'TERM_FULL', 'TERM_NEAR_OPT',
     * 'TERM_EASY' or 'TERM_PRED_ER'. */
    int ttype;

    /** The length calculation type to use. One of 'LENGTH_LAZY',
     * 'LENGTH_LAZY_GOOD', 'LENGTH_NEAR_OPT'. */
    int ltype;

    /** Saved values of the C register. Used for the LENGTH_NEAR_OPT length
     * calculation. */
    int savedC[];

    /** Saved values of CT counter. Used for the LENGTH_NEAR_OPT length
     * calculation. */
    int savedCT[];

    /** Saved values of the A register. Used for the LENGTH_NEAR_OPT length
     * calculation. */
    int savedA[];

    /** Saved values of the B byte buffer. Used for the LENGTH_NEAR_OPT length
     * calculation. */
    int savedB[];

    /** Saved values of the delFF (i.e. delayed 0xFF) state. Used for the
     * LENGTH_NEAR_OPT length calculation. */
    boolean savedDelFF[];

    /** Number of saved states. Used for the LENGTH_NEAR_OPT length
     * calculation. */
    int nSaved;

    /** The initial length of the arrays to save sates */
    final static int SAVED_LEN = 32*StdEntropyCoderOptions.NUM_PASSES;

    /** The increase in length for the arrays to save states */
    final static int SAVED_INC = 4*StdEntropyCoderOptions.NUM_PASSES;

    /**
     * Set the length calculation type to the specified type
     *
     * @param ltype The type of length calculation to use. One of
     * 'LENGTH_LAZY', 'LENGTH_LAZY_GOOD' or 'LENGTH_NEAR_OPT'.
     * */
    public void setLenCalcType(int ltype){
        // Verify the ttype and ltype
        if (ltype != LENGTH_LAZY && ltype != LENGTH_LAZY_GOOD &&
            ltype != LENGTH_NEAR_OPT) {
            throw new IllegalArgumentException("Unrecognized length "+
                                               "calculation type code: "+ltype);
        }

        if(ltype == LENGTH_NEAR_OPT){
            if(savedC==null)
                savedC = new int[SAVED_LEN];
            if(savedCT==null)
                savedCT = new int[SAVED_LEN];
            if(savedA==null)
                savedA = new int[SAVED_LEN];
            if(savedB==null)
                savedB = new int[SAVED_LEN];
            if(savedDelFF==null)
                savedDelFF = new boolean[SAVED_LEN];
        }

        this.ltype = ltype;
    }

    /**
     * Set termination type to the specified type
     *
     * @param ttype The type of termination to use. One of 'TERM_FULL',
     * 'TERM_NEAR_OPT', 'TERM_EASY' or 'TERM_PRED_ER'.
     * */
    public void setTermType(int ttype){
        if (ttype != TERM_FULL && ttype != TERM_NEAR_OPT &&
            ttype != TERM_EASY && ttype != TERM_PRED_ER ) {
            throw new IllegalArgumentException("Unrecognized termination type "+
                                               "code: "+ttype);
        }

        this.ttype = ttype;
    }

    /**
     * Instantiates a new MQ-coder, with the specified number of contexts and
     * initial states. The compressed bytestream is written to the 'oStream'
     * object.
     *
     * @param oStream where to output the compressed data
     *
     * @param nrOfContexts The number of contexts used
     *
     * @param init The initial state for each context. A reference is kept to
     * this array to reinitialize the contexts whenever 'reset()' or
     * 'resetCtxts()' is called.
     * */
    public MQCoder(ByteOutputBuffer oStream, int nrOfContexts, int init[]){
        out = oStream;

        // --- INITENC

        // Default initialization of the statistics bins is MPS=0 and
        // I=0
        I=new int[nrOfContexts];
        mPS=new int[nrOfContexts];
        initStates = init;

        a=0x8000;
        c=0;
        if(b==0xFF)
            cT=13;
        else
            cT=12;

        resetCtxts();

        // End of INITENC ---

        b=0;
    }

    /**
     * This method performs the coding of the symbol 'bit', using context
     * 'ctxt', 'n' times, using the MQ-coder speedup mode if possible.
     *
     * <P>If the symbol 'bit' is the current more probable symbol (MPS) and
     * qe[ctxt]<=0x4000, and (A-0x8000)>=qe[ctxt], speedup mode will be
     * used. Otherwise the normal mode will be used. The speedup mode can
     * significantly improve the speed of arithmetic coding when several MPS
     * symbols, with a high probability distribution, must be coded with the
     * same context. The generated bit stream is the same as if the normal mode
     * was used.
     *
     * <P>This method is also faster than the 'codeSymbols()' and
     * 'codeSymbol()' ones, for coding the same symbols with the same context
     * several times, when speedup mode can not be used, although not
     * significantly.
     *
     * @param bit The symbol do code, 0 or 1.
     *
     * @param ctxt The context to us in coding the symbol
     *
     * @param n The number of times that the symbol must be coded.
     * */
    public final void fastCodeSymbols(int bit, int ctxt, int n) {
        int q;  // cache for context's Qe
        int la; // cache for A register
        int nc; // counter for renormalization shifts
        int ns; // the maximum length of a speedup mode run
        int li; // cache for I[ctxt]

        li = I[ctxt]; // cache current index
        q=qe[li];     // retrieve current LPS prob.

        if ((q <= 0x4000) && (bit == mPS[ctxt])
            && ((ns = (a-0x8000)/q+1) > 1)) { // Do speed up mode
            // coding MPS, no conditional exchange can occur and
            // speedup mode is possible for more than 1 symbol
            do { // do as many speedup runs as necessary
                if (n <= ns) { // All symbols in this run
                    // code 'n' symbols
                    la = n*q; // accumulated Q
                    a -= la;
                    c += la;
                    if (a >= 0x8000) { // no renormalization
                        I[ctxt] = li;  // save the current state
                        return; // done
                    }
                    I[ctxt] = nMPS[li]; // goto next state and save it
                    // -- Renormalization (MPS: no need for while loop)
                    a <<= 1; // a is doubled
                    c <<= 1; // c is doubled
                    cT--;
                    if(cT==0) {
                        byteOut();
                    }
                    // -- End of renormalization
                    return; // done
                }
                else { // Not all symbols in this run
                    // code 'ns' symbols
                    la = ns*q; // accumulated Q
                    c += la;
                    a -= la;
                    // cache li and q for next iteration
                    li = nMPS[li];
                    q = qe[li]; // New q is always less than current one
                    // new I[ctxt] is stored in last run
                    // Renormalization always occurs since we exceed 'ns'
                    // -- Renormalization (MPS: no need for while loop)
                    a <<= 1; // a is doubled
                    c <<= 1; // c is doubled
                    cT--;
                    if(cT==0) {
                        byteOut();
                    }
                    // -- End of renormalization
                    n -= ns; // symbols left to code
                    ns = (a-0x8000)/q+1; // max length of next speedup run
                    continue; // goto next iteration
                }
            } while (n>0);
        } // end speed up mode
        else { // No speedup mode
            // Either speedup mode is not possible or not worth doing it
            // because of probable conditional exchange
            // Code everything as in normal mode
            la = a;       // cache A register in local variable
            do {
                if (bit == mPS[ctxt]) { // -- code MPS
                    la -= q; // Interval division associated with MPS coding
                    if(la>=0x8000){ // Interval big enough
                        c += q;
                    }
                    else { // Interval too short
                        if(la<q) // Probabilities are inverted
                            la = q;
                        else
                            c += q;
                        // cache new li and q for next iteration
                        li = nMPS[li];
                        q = qe[li];
                        // new I[ctxt] is stored after end of loop
                        // -- Renormalization (MPS: no need for while loop)
                        la <<= 1; // a is doubled
                        c <<= 1;  // c is doubled
                        cT--;
                        if(cT==0) {
                            byteOut();
                        }
                        // -- End of renormalization
                    }
                }
                else { // -- code LPS
                    la -= q; // Interval division according to LPS coding
                    if(la<q)
                        c += q;
                    else
                        la = q;
                    if(switchLM[li]!=0){
                        mPS[ctxt]=1-mPS[ctxt];
                    }
                    // cache new li and q for next iteration
                    li = nLPS[li];
                    q = qe[li];
                    // new I[ctxt] is stored after end of loop
                    // -- Renormalization
                    // sligthly better than normal loop
                    nc = 0;
                    do {
                        la <<= 1;
                        nc++; // count number of necessary shifts
                    } while (la<0x8000);
                    if (cT > nc) {
                        c <<= nc;
                        cT -= nc;
                    }
                    else {
                        do {
                            c <<= cT;
                            nc -= cT;
                            // cT = 0; // not necessary
                            byteOut();
                        } while (cT <= nc);
                        c <<= nc;
                        cT -= nc;
                    }
                    // -- End of renormalization
                }
                n--;
            } while (n>0);
            I[ctxt] = li; // store new I[ctxt]
            a = la; // save cached A register
        }
    }

    /**
     * This function performs the arithmetic encoding of several symbols
     * together. The function receives an array of symbols that are to be
     * encoded and an array containing the contexts with which to encode them.
     *
     * <P>The advantage of using this function is that the cost of the method
     * call is amortized by the number of coded symbols per method call.
     *
     * <P>Each context has a current MPS and an index describing what the
     * current probability is for the LPS. Each bit is encoded and if the
     * probability of the LPS exceeds .5, the MPS and LPS are switched.
     *
     * @param bits An array containing the symbols to be encoded. Valid
     * symbols are 0 and 1.
     *
     * @param cX The context for each of the symbols to be encoded
     *
     * @param n The number of symbols to encode.
     * */
    public final void codeSymbols(int[] bits, int[] cX, int n){
        int q;
        int li; // local cache of I[context]
        int la;
        int nc;
        int ctxt; // context of current symbol
        int i; // counter

        // NOTE: here we could use symbol aggregation to speed things up.
        // It remains to be studied.

        la = a; // cache A register in local variable
        for(i=0;i<n;i++){
            // NOTE: (a < 0x8000) is equivalent to ((a & 0x8000)==0)
            // since 'a' is always less than or equal to 0xFFFF

            // NOTE: conditional exchange guarantees that A for MPS is
            // always greater than 0x4000 (i.e. 0.375)
            // => one renormalization shift is enough for MPS
            // => no need to do a renormalization while loop for MPS

            ctxt = cX[i];
            li = I[ctxt];
            q=qe[li]; // Retrieve current LPS prob.

            if(bits[i]==mPS[ctxt]){ // -- Code MPS

                la -= q; // Interval division associated with MPS coding

                if(la>=0x8000){ // Interval big enough
                    c += q;
                }
                else { // Interval too short
                    if(la<q) // Probabilities are inverted
                        la = q;
                    else
                        c += q;

                    I[ctxt]=nMPS[li];

                    // -- Renormalization (MPS: no need for while loop)
                    la <<= 1; // a is doubled
                    c <<= 1; // c is doubled
                    cT--;
                    if(cT==0) {
                        byteOut();
                    }
                    // -- End of renormalization
                }
            }// End Code MPS --
            else{ // -- Code LPS
                la -= q; // Interval division according to LPS coding

                if(la<q)
                    c += q;
                else
                    la = q;
                if(switchLM[li]!=0){
                    mPS[ctxt]=1-mPS[ctxt];
                }
                I[ctxt]=nLPS[li];

                // -- Renormalization

                // sligthly better than normal loop
                nc = 0;
                do {
                    la <<= 1;
                    nc++; // count number of necessary shifts
                } while (la<0x8000);
                if (cT > nc) {
                    c <<= nc;
                    cT -= nc;
                }
                else {
                    do {
                        c <<= cT;
                        nc -= cT;
                        // cT = 0; // not necessary
                        byteOut();
                    } while (cT <= nc);
                    c <<= nc;
                    cT -= nc;
                }

                // -- End of renormalization
            }
        }
        a = la; // save cached A register
    }


    /**
     * This function performs the arithmetic encoding of one symbol. The
     * function receives a bit that is to be encoded and a context with which
     * to encode it.
     *
     * <P>Each context has a current MPS and an index describing what the
     * current probability is for the LPS. Each bit is encoded and if the
     * probability of the LPS exceeds .5, the MPS and LPS are switched.
     *
     * @param bit The symbol to be encoded, must be 0 or 1.
     *
     * @param context the context with which to encode the symbol.
     * */
    public final void codeSymbol(int bit, int context){
        int q;
        int li; // local cache of I[context]
        int la;
        int n;

        // NOTE: (a < 0x8000) is equivalent to ((a & 0x8000)==0)
        // since 'a' is always less than or equal to 0xFFFF

        // NOTE: conditional exchange guarantees that A for MPS is
        // always greater than 0x4000 (i.e. 0.375)
        // => one renormalization shift is enough for MPS
        // => no need to do a renormalization while loop for MPS

        li = I[context];
        q=qe[li]; // Retrieve current LPS prob.

        if(bit==mPS[context]){// -- Code MPS

            a -= q; // Interval division associated with MPS coding

            if(a>=0x8000){ // Interval big enough
                c += q;
            }
            else { // Interval too short
                if(a<q) // Probabilities are inverted
                    a = q;
                else
                    c += q;

                I[context]=nMPS[li];

                // -- Renormalization (MPS: no need for while loop)
                a <<= 1; // a is doubled
                c <<= 1; // c is doubled
                cT--;
                if(cT==0) {
                    byteOut();
                }
                // -- End of renormalization
            }
        }// End Code MPS --
        else{ // -- Code LPS

            la = a; // cache A register in local variable
            la -= q; // Interval division according to LPS coding

            if(la<q)
                c += q;
            else
                la = q;
            if(switchLM[li]!=0){
                mPS[context]=1-mPS[context];
            }
            I[context]=nLPS[li];

            // -- Renormalization

            // sligthly better than normal loop
            n = 0;
            do {
                la <<= 1;
                n++; // count number of necessary shifts
            } while (la<0x8000);
            if (cT > n) {
                c <<= n;
                cT -= n;
            }
            else {
                do {
                    c <<= cT;
                    n -= cT;
                    // cT = 0; // not necessary
                    byteOut();
                } while (cT <= n);
                c <<= n;
                cT -= n;
            }

            // -- End of renormalization
            a = la; // save cached A register
        }
    }

    /**
     * This function puts one byte of compressed bits in the out out stream.
     * the highest 8 bits of c are then put in b to be the next byte to
     * write. This method delays the output of any 0xFF bytes until a non 0xFF
     * byte has to be written to the output bit stream (the 'delFF' variable
     * signals if there is a delayed 0xff byte).
     * */
    private void byteOut(){
        if(nrOfWrittenBytes >= 0){
            if(b==0xFF){
                // Delay 0xFF byte
                delFF = true;
                b=c>>>20;
                c &= 0xFFFFF;
                cT=7;
            }
            else if(c < 0x8000000){
                // Write delayed 0xFF bytes
                if (delFF) {
                    out.write(0xFF);
                    delFF = false;
                    nrOfWrittenBytes++;
                }
                out.write(b);
                nrOfWrittenBytes++;
                b=c>>>19;
                c &= 0x7FFFF;
                cT=8;
            }
            else{
                b++;
                if(b==0xFF){
                    // Delay 0xFF byte
                    delFF = true;
                    c &= 0x7FFFFFF;
                    b=c>>>20;
                    c &= 0xFFFFF;
                    cT=7;
                }
                else{
                    // Write delayed 0xFF bytes
                    if (delFF) {
                        out.write(0xFF);
                        delFF = false;
                        nrOfWrittenBytes++;
                    }
                    out.write(b);
                    nrOfWrittenBytes++;
                    b=((c>>>19)&0xFF);
                    c &= 0x7FFFF;
                    cT=8;
                }
            }
        }
        else {
            // NOTE: carry bit can never be set if the byte buffer was empty
            b= (c>>>19);
            c &= 0x7FFFF;
            cT=8;
            nrOfWrittenBytes++;
        }
    }

    /**
     * This function flushes the remaining encoded bits and makes sure that
     * enough information is written to the bit stream to be able to finish
     * decoding, and then it reinitializes the internal state of the MQ coder
     * but without modifying the context states.
     *
     * <P>After calling this method the 'finishLengthCalculation()' method
     * should be called, after cmopensating the returned length for the length
     * of previous coded segments, so that the length calculation is finalized.
     *
     * <P>The type of termination used depends on the one specified at the
     * constructor.
     *
     * @return The length of the arithmetic codeword after termination, in
     * bytes.
     * */
    public int terminate(){
        switch (ttype) {
        case TERM_FULL:
            //sets the remaining bits of the last byte of the coded bits.
            int tempc=c+a;
            c=c|0xFFFF;
            if(c>=tempc)
                c=c-0x8000;

            int remainingBits = 27-cT;

            // Flushes remainingBits
            do{
                c <<= cT;
                if(b != 0xFF)
                    remainingBits -= 8;
                else
                    remainingBits -= 7;
                byteOut();
            }
            while(remainingBits > 0);

            b |= (1<<(-remainingBits))-1;
            if (b==0xFF) { // Delay 0xFF bytes
                delFF = true;
            }
            else {
                // Write delayed 0xFF bytes
                if (delFF) {
                    out.write(0xFF);
                    delFF = false;
                    nrOfWrittenBytes++;
                }
                out.write(b);
                nrOfWrittenBytes++;
            }
            break;
        case TERM_PRED_ER:
        case TERM_EASY:
            // The predictable error resilient and easy termination are the
            // same, except for the fact that the easy one can modify the
            // spare bits in the last byte to maximize the likelihood of
            // having a 0xFF, while the error resilient one can not touch
            // these bits.

            // In the predictable error resilient case the spare bits will be
            // recalculated by the decoder and it will check if they are the
            // same as as in the codestream and then deduce an error
            // probability from there.

            int k; // number of bits to push out

            k = (11-cT)+1;

            c <<= cT;
            for (; k > 0; k-=cT, c<<=cT){
              byteOut();
            }

            // Make any spare bits 1s if in easy termination
            if (k < 0 && ttype == TERM_EASY) {
                // At this stage there is never a carry bit in C, so we can
                // freely modify the (-k) least significant bits.
                b |= (1<<(-k))-1;
            }

            byteOut(); // Push contents of byte buffer
            break;
        case TERM_NEAR_OPT:

            // This algorithm terminates in the shortest possible way, besides
            // the fact any previous 0xFF 0x7F sequences are not
            // eliminated. The probabalility of having those sequences is
            // extremely low.

            // The calculation of the length is based on the fact that the
            // decoder will pad the codestream with an endless string of
            // (binary) 1s. If the codestream, padded with 1s, is within the
            // bounds of the current interval then correct decoding is
            // guaranteed. The lower inclusive bound of the current interval
            // is the value of C (i.e. if only lower intervals would be coded
            // in the future). The upper exclusive bound of the current
            // interval is C+A (i.e. if only upper intervals would be coded in
            // the future). We therefore calculate the minimum length that
            // would be needed so that padding with 1s gives a codestream
            // within the interval.

            // In general, such a calculation needs the value of the next byte
            // that appears in the codestream. Here, since we are terminating,
            // the next value can be anything we want that lies within the
            // interval, we use the lower bound since this minimizes the
            // length. To calculate the necessary length at any other place
            // than the termination it is necessary to know the next bytes
            // that will appear in the codestream, which involves storing the
            // codestream and the sate of the MQCoder at various points (a
            // worst case approach can be used, but it is much more
            // complicated and the calculated length would be only marginally
            // better than much simple calculations, if not the same).

            int cLow;
            int cUp;
            int bLow;
            int bUp;

            // Initialize the upper (exclusive) and lower bound (inclusive) of
            // the valid interval (the actual interval is the concatenation of
            // bUp and cUp, and bLow and cLow).
            cLow = c;
            cUp = c+a;
            bLow = bUp = b;

            // We start by normalizing the C register to the sate cT = 0
            // (i.e., just before byteOut() is called)
            cLow <<= cT;
            cUp  <<= cT;
            // Progate eventual carry bits and reset them in Clow, Cup NOTE:
            // carry bit can never be set if the byte buffer was empty so no
            // problem with propagating a carry into an empty byte buffer.
            if ((cLow & (1<<27)) != 0) { // Carry bit in cLow
                if (bLow == 0xFF) {
                    // We can not propagate carry bit, do bit stuffing
                    delFF = true; // delay 0xFF
                    // Get next byte buffer
                    bLow = cLow>>>20;
                    bUp = cUp>>>20;
                    cLow &= 0xFFFFF;
                    cUp &= 0xFFFFF;
                    // Normalize to cT = 0
                    cLow <<= 7;
                    cUp <<= 7;
                }
                else { // we can propagate carry bit
                    bLow++; // propagate
                    cLow &= ~(1<<27); // reset carry in cLow
                }
            }
            if ((cUp & (1<<27)) != 0) {
                bUp++; // propagate
                cUp &= ~(1<<27); // reset carry
            }

            // From now on there can never be a carry bit on cLow, since we
            // always output bLow.

            // Loop testing for the condition and doing byte output if they
            // are not met.
            while(true){
                // If decoder's codestream is within interval stop
                // If preceding byte is 0xFF only values [0,127] are valid
                if(delFF){ // If delayed 0xFF
                    if (bLow <= 127 && bUp > 127) break;
                    // We will write more bytes so output delayed 0xFF now
                    out.write(0xFF);
                    nrOfWrittenBytes++;
                    delFF = false;
                }
                else{ // No delayed 0xFF
                    if (bLow <= 255 && bUp > 255) break;
                }

                // Output next byte
                // We could output anything within the interval, but using
                // bLow simplifies things a lot.

                // We should not have any carry bit here

                // Output bLow
                if (bLow < 255) {
                    // Transfer byte bits from C to B
                    // (if the byte buffer was empty output nothing)
                    if (nrOfWrittenBytes >= 0) out.write(bLow);
                    nrOfWrittenBytes++;
                    bUp -= bLow;
                    bUp <<= 8;
                    // Here bLow would be 0
                    bUp |= (cUp >>> 19) & 0xFF;
                    bLow = (cLow>>> 19) & 0xFF;
                    // Clear upper bits (just pushed out) from cUp Clow.
                    cLow &= 0x7FFFF;
                    cUp  &= 0x7FFFF;
                    // Goto next state where CT is 0
                    cLow <<= 8;
                    cUp  <<= 8;
                    // Here there can be no carry on Cup, Clow
                }
                else { // bLow = 0xFF
                    // Transfer byte bits from C to B
                    // Since the byte to output is 0xFF we can delay it
                    delFF = true;
                    bUp -= bLow;
                    bUp <<= 7;
                    // Here bLow would be 0
                    bUp |= (cUp>>20) & 0x7F;
                    bLow = (cLow>>20) & 0x7F;
                    // Clear upper bits (just pushed out) from cUp Clow.
                    cLow &= 0xFFFFF;
                    cUp  &= 0xFFFFF;
                    // Goto next state where CT is 0
                    cLow <<= 7;
                    cUp  <<= 7;
                    // Here there can be no carry on Cup, Clow
                }
            }
            break;
        default:
            throw new Error("Illegal termination type code");
        }

        // Reinitialize the state (without modifying the contexts)
        int len;

        len = nrOfWrittenBytes;
        a = 0x8000;
        c = 0;
        b = 0;
        cT = 12;
        delFF = false;
        nrOfWrittenBytes = -1;

        // Return the terminated length
        return len;
    }

    /**
     * Returns the number of contexts in the arithmetic coder.
     *
     * @return The number of contexts
     * */
    public final int getNumCtxts(){
        return I.length;
    }

    /**
     * Resets a context to the original probability distribution, and sets its
     * more probable symbol to 0.
     *
     * @param c The number of the context (it starts at 0).
     * */
    public final void resetCtxt(int c){
        I[c]=initStates[c];
        mPS[c] = 0;
    }

    /**
     * Resets all contexts to their original probability distribution and sets
     * all more probable symbols to 0.
     * */
    public final void resetCtxts(){
        System.arraycopy(initStates,0,I,0,I.length);
        ArrayUtil.intArraySet(mPS,0);
    }

    /**
     * Returns the number of bytes that are necessary from the compressed
     * output stream to decode all the symbols that have been coded this
     * far. The number of returned bytes does not include anything coded
     * previous to the last time the 'terminate()' or 'reset()' methods where
     * called.
     *
     * <P>The values returned by this method are then to be used in finishing
     * the length calculation with the 'finishLengthCalculation()' method,
     * after compensation of the offset in the number of bytes due to previous
     * terminated segments.
     *
     * <P>This method should not be called if the current coding pass is to be
     * terminated. The 'terminate()' method should be called instead.
     *
     * <P>The calculation is done based on the type of length calculation
     * specified at the constructor.
     *
     * @return The number of bytes in the compressed output stream necessary
     * to decode all the information coded this far.
     * */
    public final int getNumCodedBytes(){
        // NOTE: testing these algorithms for correctness is quite
        // difficult. One way is to modify the rate allocator so that not all
        // bit-planes are output if the distortion estimate for last passes is
        // the same as for the previous ones.

        switch (ltype) {
        case LENGTH_LAZY_GOOD:
            // This one is a bit better than LENGTH_LAZY.
            int bitsInN3Bytes; // The minimum amount of bits that can be stored
                               // in the 3 bytes following the current byte
                               // buffer 'b'.
            if (b >= 0xFE) {
                // The byte after b can have a bit stuffed so ther could be
                // one less bit available
                bitsInN3Bytes = 22; // 7 + 8 + 7
            }
            else {
                // We are sure that next byte after current byte buffer has no
                // bit stuffing
                bitsInN3Bytes = 23; // 8 + 7 + 8
            }
            if ((11-cT+16) <= bitsInN3Bytes) {
                return nrOfWrittenBytes+(delFF ? 1 : 0)+1+3;
            }
            else {
                return nrOfWrittenBytes+(delFF ? 1 : 0)+1+4;
            }
        case LENGTH_LAZY:
            // This is the very basic one that appears in the VM text
            if ((27-cT) <= 22) {
                return nrOfWrittenBytes+(delFF ? 1 : 0)+1+3;
            }
            else {
                return nrOfWrittenBytes+(delFF ? 1 : 0)+1+4;
            }
        case LENGTH_NEAR_OPT:
            // This is the best length calculation implemented in this class.
            // It is almost always optimal. In order to calculate the length
            // it is necessary to know which bytes will follow in the MQ
            // bit stream, so we need to wait until termination to perform it.
            // Save the state to perform the calculation later, in
            // finishLengthCalculation()
            saveState();
            // Return current number of output bytes to use it later in
            // finishLengthCalculation()
            return nrOfWrittenBytes;
        default:
            throw new Error("Illegal length calculation type code");
        }
    }

    /**
     * Reinitializes the MQ coder and the underlying 'ByteOutputBuffer' buffer
     * as if a new object was instantaited. All the data in the
     * 'ByteOutputBuffer' buffer is erased and the state and contexts of the
     * MQ coder are reinitialized). Additionally any saved MQ states are
     * discarded.
     * */
    public final void reset() {

        // Reset the output buffer
        out.reset();

        a=0x8000;
        c=0;
        b=0;
        if(b==0xFF)
            cT=13;
        else
            cT=12;
        resetCtxts();
        nrOfWrittenBytes = -1;
        delFF = false;

        nSaved = 0;
    }

    /**
     * Saves the current state of the MQ coder (just the registers, not the
     * contexts) so that a near optimal length calculation can be performed
     * later.
     * */
    private void saveState() {
        // Increase capacity if necessary
        if (nSaved == savedC.length) {
            Object tmp;
            tmp = savedC;
            savedC = new int[nSaved+SAVED_INC];
            System.arraycopy(tmp,0,savedC,0,nSaved);
            tmp = savedCT;
            savedCT = new int[nSaved+SAVED_INC];
            System.arraycopy(tmp,0,savedCT,0,nSaved);
            tmp = savedA;
            savedA = new int[nSaved+SAVED_INC];
            System.arraycopy(tmp,0,savedA,0,nSaved);
            tmp = savedB;
            savedB = new int[nSaved+SAVED_INC];
            System.arraycopy(tmp,0,savedB,0,nSaved);
            tmp = savedDelFF;
            savedDelFF = new boolean[nSaved+SAVED_INC];
            System.arraycopy(tmp,0,savedDelFF,0,nSaved);
        }
        // Save the current sate
        savedC[nSaved] = c;
        savedCT[nSaved] = cT;
        savedA[nSaved] = a;
        savedB[nSaved] = b;
        savedDelFF[nSaved] = delFF;
        nSaved++;
    }

    /**
     * Terminates the calculation of the required length for each coding
     * pass. This method must be called just after the 'terminate()' one has
     * been called for each terminated MQ segment.
     *
     * <P>The values in 'rates' must have been compensated for any offset due
     * to previous terminated segments, so that the correct index to the
     * stored coded data is used.
     *
     * @param rates The array containing the values returned by
     * 'getNumCodedBytes()' for each coding pass.
     *
     * @param n The index in the 'rates' array of the last terminated length.
     * */
    public void finishLengthCalculation(int rates[], int n) {
        if (ltype != LENGTH_NEAR_OPT) {
            // For the simple calculations the only thing we need to do is to
            // ensure that the calculated lengths are no greater than the
            // terminated one
            if (n > 0 && rates[n-1] > rates[n]) {
                // We need correction
                int tl = rates[n]; // The terminated length
                n--;
                do {
                    rates[n--] = tl;
               }  while (n >= 0 && rates[n] > tl);
            }
        }
        else {
            // We need to perform the more sophisticated near optimal
            // calculation.

            // The calculation of the length is based on the fact that the
            // decoder will pad the codestream with an endless string of
            // (binary) 1s after termination. If the codestream, padded with
            // 1s, is within the bounds of the current interval then correct
            // decoding is guaranteed. The lower inclusive bound of the
            // current interval is the value of C (i.e. if only lower
            // intervals would be coded in the future). The upper exclusive
            // bound of the current interval is C+A (i.e. if only upper
            // intervals would be coded in the future). We therefore calculate
            // the minimum length that would be needed so that padding with 1s
            // gives a codestream within the interval.

            // In order to know what will be appended to the current base of
            // the interval we need to know what is in the MQ bit stream after
            // the current last output byte until the termination. This is why
            // this calculation has to be performed after the MQ segment has
            // been entirely coded and terminated.

            int cLow; // lower bound on the C register for correct decoding
            int cUp;  // upper bound on the C register for correct decoding
            int bLow; // lower bound on the byte buffer for correct decoding
            int bUp;  // upper bound on the byte buffer for correct decoding
            int ridx; // index in the rates array of the pass we are
            // calculating
            int sidx; // index in the saved state array
            int clen; // current calculated length
            boolean cdFF; // the current delayed FF state
            int nb;   // the next byte of output
            int minlen; // minimum possible length
            int maxlen; // maximum possible length

            // Start on the first pass of this segment
            ridx = n-nSaved;
            // Minimum allowable length is length of previous termination
            minlen = (ridx-1>=0) ? rates[ridx-1] : 0;
            // Maximum possible length is the terminated length
            maxlen = rates[n];
            for (sidx = 0; ridx < n; ridx++, sidx++) {
                // Load the initial values of the bounds
                cLow = savedC[sidx];
                cUp = savedC[sidx]+savedA[sidx];
                bLow = savedB[sidx];
                bUp = savedB[sidx];
                // Normalize to CT = 0 and propagate and reset any carry bits
                cLow <<= savedCT[sidx];
                if ((cLow & 0x8000000) != 0) {
                    bLow++;
                    cLow &= 0x7FFFFFF;
                }
                cUp <<= savedCT[sidx];
                if ((cUp & 0x8000000) != 0) {
                    bUp++;
                    cUp &= 0x7FFFFFF;
                }
                // Initialize current calculated length
                cdFF = savedDelFF[sidx];
                // rates[ridx] contains the number of bytes already output
                // when the state was saved, compensated for the offset in the
                // output stream.
                clen = rates[ridx]+(cdFF? 1 : 0);
                while (true) {
                    // If we are at end of coded data then this is the length
                    if (clen >= maxlen) {
                        clen = maxlen;
                        break;
                    }
                    // Check for sufficiency of coded data
                    if (cdFF) {
                        if (bLow < 128 && bUp >= 128) {
                            // We are done for this pass
                            clen--; // Don't need delayed FF
                            break;
                        }
                    }
                    else {
                        if (bLow < 256 && bUp >= 256) {
                            // We are done for this pass
                            break;
                        }
                    }
                    // Update bounds with next byte of coded data and
                    // normalize to CT = 0 again.
                    nb =  (clen >= minlen) ? out.getByte(clen) : 0;
                    bLow -= nb;
                    bUp -= nb;
                    clen++;
                    if (nb == 0xFF) {
                        bLow <<= 7;
                        bLow |= (cLow >> 20) & 0x7F;
                        cLow &= 0xFFFFF;
                        cLow <<= 7;
                        bUp <<= 7;
                        bUp |= (cUp >> 20) & 0x7F;
                        cUp &= 0xFFFFF;
                        cUp <<= 7;
                        cdFF = true;
                    }
                    else {
                        bLow <<= 8;
                        bLow |= (cLow >> 19) & 0xFF;
                        cLow &= 0x7FFFF;
                        cLow <<= 8;
                        bUp <<= 8;
                        bUp |= (cUp >> 19) & 0xFF;
                        cUp &= 0x7FFFF;
                        cUp <<= 8;
                        cdFF = false;
                    }
                    // Test again
                }
                // Store the rate found
                rates[ridx] = (clen>=minlen) ? clen : minlen;
            }
            // Reset the saved states
            nSaved = 0;
        }
    }
}