[wimlib] / src / lzx-compress.c

/*
 * lzx-compress.c
 *
 * LZX compression routines
 */

/*
 * Copyright (C) 2012, 2013 Eric Biggers
 *
 * This file is part of wimlib, a library for working with WIM files.
 *
 * wimlib 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 3 of the License, or (at your option)
 * any later version.
 *
 * wimlib 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 wimlib; if not, see http://www.gnu.org/licenses/.
 */


/*
 * This file contains a compressor for the LZX compression format, as used in
 * the WIM file format.
 *
 * Format
 * ======
 *
 * First, the primary reference for the LZX compression format is the
 * specification released by Microsoft.
 *
 * Second, the comments in lzx-decompress.c provide some more information about
 * the LZX compression format, including errors in the Microsoft specification.
 *
 * Do note that LZX shares many similarities with DEFLATE, the algorithm used by
 * zlib and gzip.  Both LZX and DEFLATE use LZ77 matching and Huffman coding,
 * and certain other details are quite similar, such as the method for storing
 * Huffman codes.  However, some of the main differences are:
 *
 * - LZX preprocesses the data before attempting to compress it.
 * - LZX uses a "main" alphabet which combines literals and matches, with the
 *   match symbols containing a "length header" (giving all or part of the match
 *   length) and a "position footer" (giving, roughly speaking, the order of
 *   magnitude of the match offset).
 * - LZX does not have static Huffman blocks; however it does have two types of
 *   dynamic Huffman blocks ("aligned offset" and "verbatim").
 * - LZX has a minimum match length of 2 rather than 3.
 * - In LZX, match offsets 0 through 2 actually represent entries in a LRU queue
 *   of match offsets.
 *
 * Algorithms
 * ==========
 *
 * There are actually two distinct overall algorithms implemented here.  We
 * shall refer to them as the "slow" algorithm and the "fast" algorithm.  The
 * "slow" algorithm spends more time compressing to achieve a higher compression
 * ratio compared to the "fast" algorithm.  More details are presented below.
 *
 * Slow algorithm
 * --------------
 *
 * The "slow" algorithm to generate LZX-compressed data is roughly as follows:
 *
 * 1. Preprocess the input data to translate the targets of x86 call instructions
 *    to absolute offsets.
 *
 * 2. Determine the best known sequence of LZ77 matches ((offset, length) pairs)
 *    and literal bytes to divide the input into.  Raw match-finding is done
 *    using a very clever binary tree search based on the "Bt3" algorithm from
 *    7-Zip.  Parsing, or match-choosing, is solved essentially as a
 *    minimum-cost path problem, but using a heuristic forward search based on
 *    the Deflate encoder from 7-Zip rather than a more intuitive backward
 *    search, the latter of which would naively require that all matches be
 *    found.  This heuristic search, as well as other heuristics such as limits
 *    on the matches considered, considerably speed up this part of the
 *    algorithm, which is the main bottleneck.  Finally, after matches and
 *    literals are chosen, the needed Huffman codes needed to output them are
 *    built.
 *
 * 3. Up to a certain number of iterations, use the resulting Huffman codes to
 *    refine a cost model and go back to Step #2 to determine an improved
 *    sequence of matches and literals.
 *
 * 4. Up to a certain depth, try splitting the current block to see if the
 *    compression ratio can be improved.  This may be the case if parts of the
 *    input differ greatly from each other and could benefit from different
 *    Huffman codes.
 *
 * 5. Output the resulting block(s) using the match/literal sequences and the
 *    Huffman codes that were computed for each block.
 *
 * Fast algorithm
 * --------------
 *
 * The fast algorithm (and the only one available in wimlib v1.5.1 and earlier)
 * spends much less time on the main bottlenecks of the compression process ---
 * that is the match finding, match choosing, and block splitting.  Matches are
 * found and chosen with hash chains using a greedy parse with one position of
 * look-ahead.  No block splitting is done; only compressing the full input into
 * an aligned offset block is considered.
 *
 * API
 * ===
 *
 * The old API (retained for backward compatibility) consists of just one function:
 *
 *      wimlib_lzx_compress()
 *
 * The new compressor has more potential parameters and needs more memory, so
 * the new API ties up memory allocations and compression parameters into a
 * context:
 *
 *      wimlib_lzx_alloc_context()
 *      wimlib_lzx_compress2()
 *      wimlib_lzx_free_context()
 *
 * Both wimlib_lzx_compress() and wimlib_lzx_compress2() are designed to
 * compress an in-memory buffer of up to 32768 bytes.  There is no sliding
 * window.  This is suitable for the WIM format, which uses fixed-size chunks
 * that are seemingly always 32768 bytes.  If needed, the compressor potentially
 * could be extended to support a larger and/or sliding window.
 *
 * Both wimlib_lzx_compress() and wimlib_lzx_compress2() return 0 if the data
 * could not be compressed to less than the size of the uncompressed data.
 * Again, this is suitable for the WIM format, which stores such data chunks
 * uncompressed.
 *
 * The functions in this API are exported from the library, although this is
 * only in case other programs happen to have uses for it other than WIM
 * reading/writing as already handled through the rest of the library.
 *
 * Acknowledgments
 * ===============
 *
 * Acknowledgments to several other open-source projects that made it possible
 * to implement this code:
 *
 * - 7-Zip (author: Igor Pavlov), for the binary tree match-finding
 *   algorithm, the heuristic near-optimal forward match-choosing
 *   algorithm, and the block splitting algorithm.
 *
 * - zlib (author: Jean-loup Gailly and Mark Adler), for the hash table
 *   match-finding algorithm.
 *
 * - lzx-compress (author: Matthew T. Russotto), on which some parts of this
 *   code were originally based.
 */

#ifdef HAVE_CONFIG_H
#  include "config.h"
#endif

#include "wimlib.h"
#include "wimlib/compress.h"
#include "wimlib/error.h"
#include "wimlib/lzx.h"
#include "wimlib/util.h"

#ifdef ENABLE_LZX_DEBUG
#  include <wimlib/decompress.h>
#endif

#include <string.h>

/* Experimental parameters not exposed through the API  */
#define LZX_PARAM_OPTIM_ARRAY_SIZE      1024
#define LZX_PARAM_ACCOUNT_FOR_LRU       1
#define LZX_PARAM_DONT_SKIP_MATCHES     0
#define LZX_PARAM_USE_EMPIRICAL_DEFAULT_COSTS 1

/* Currently, this constant can't simply be changed because the code currently
 * uses a static number of position slots.  */
#define LZX_MAX_WINDOW_SIZE     32768

/* This may be WIM-specific  */
#define LZX_DEFAULT_BLOCK_SIZE  32768

#define LZX_LZ_HASH_BITS        15
#define LZX_LZ_HASH_SIZE        (1 << LZX_LZ_HASH_BITS)
#define LZX_LZ_HASH_MASK        (LZX_LZ_HASH_SIZE - 1)
#define LZX_LZ_HASH_SHIFT       5

/* Codewords for the LZX main, length, and aligned offset Huffman codes  */
struct lzx_codewords {
        u16 main[LZX_MAINTREE_NUM_SYMBOLS];
        u16 len[LZX_LENTREE_NUM_SYMBOLS];
        u16 aligned[LZX_ALIGNEDTREE_NUM_SYMBOLS];
};

/* Lengths for the LZX main, length, and aligned offset Huffman codes  */
struct lzx_lens {
        u8 main[LZX_MAINTREE_NUM_SYMBOLS];
        u8 len[LZX_LENTREE_NUM_SYMBOLS];
        u8 aligned[LZX_ALIGNEDTREE_NUM_SYMBOLS];
};

/* The LZX main, length, and aligned offset Huffman codes  */
struct lzx_codes {
        struct lzx_codewords codewords;
        struct lzx_lens lens;
};

/* Tables for tallying symbol frequencies in the three LZX alphabets  */
struct lzx_freqs {
        freq_t main[LZX_MAINTREE_NUM_SYMBOLS];
        freq_t len[LZX_LENTREE_NUM_SYMBOLS];
        freq_t aligned[LZX_ALIGNEDTREE_NUM_SYMBOLS];
};

/* LZX intermediate match/literal format  */
struct lzx_match {
        /* Bit     Description
         *
         * 31      1 if a match, 0 if a literal.
         *
         * 30-25   position slot.  This can be at most 50, so it will fit in 6
         *         bits.
         *
         * 8-24    position footer.  This is the offset of the real formatted
         *         offset from the position base.  This can be at most 17 bits
         *         (since lzx_extra_bits[LZX_NUM_POSITION_SLOTS - 1] is 17).
         *
         * 0-7     length of match, minus 2.  This can be at most
         *         (LZX_MAX_MATCH - 2) == 255, so it will fit in 8 bits.  */
        u32 data;
};

/* Raw LZ match/literal format: just a length and offset.
 *
 * The length is the number of bytes of the match, and the offset is the number
 * of bytes back in the input the match is from the matched text.
 *
 * If @len < LZX_MIN_MATCH, then it's really just a literal byte.  */
struct raw_match {
        u16 len;
        u16 offset;
};

/* Specification for a LZX block  */
struct lzx_block_spec {

        /* Set to 1 if this block has been split (in two --- we only considser
         * binary splits).  In such cases the rest of the fields are
         * unimportant, since the relevant information is rather in the
         * structures for the sub-blocks.  */
        u8 is_split : 1;

        /* One of the LZX_BLOCKTYPE_* constants indicating which type of this
         * block.  */
        u8 block_type : 2;

        /* 0-based position in the window at which this block starts.  */
        u16 window_pos;

        /* The number of bytes of uncompressed data this block represents.  */
        u16 block_size;

        /* The position in the 'chosen_matches' array in the `struct
         * lzx_compressor' at which the match/literal specifications for
         * this block begin.  */
        unsigned chosen_matches_start_pos;

        /* The number of match/literal specifications for this block.  */
        u16 num_chosen_matches;

        /* Huffman codes for this block.  */
        struct lzx_codes codes;
};

/*
 * An array of these structures is used during the match-choosing algorithm.
 * They correspond to consecutive positions in the window and are used to keep
 * track of the cost to reach each position, and the match/literal choices that
 * need to be chosen to reach that position.
 */
struct lzx_optimal {
        /* The approximate minimum cost, in bits, to reach this position in the
         * window which has been found so far.  */
        u32 cost;

        /* The union here is just for clarity, since the fields are used in two
         * slightly different ways.  Initially, the @prev structure is filled in
         * first, and links go from later in the window to earlier in the
         * window.  Later, @next structure is filled in and links go from
         * earlier in the window to later in the window.  */
        union {
                struct {
                        /* Position of the start of the match or literal that
                         * was taken to get to this position in the approximate
                         * minimum-cost parse.  */
                        u16 link;

                        /* Offset, relative to its starting position, of the
                         * match or literal that was taken to get to this
                         * position in the approximate minimum-cost parse.  */
                        u16 match_offset;
                } prev;
                struct {
                        /* Position at which the match or literal starting at
                         * this position ends in the minimum-cost parse.  */
                        u16 link;

                        /* Offset, relative to its starting position, of the
                         * match or literal starting at this position in the
                         * approximate minimum-cost parse.  */
                        u16 match_offset;
                } next;
        };
#if LZX_PARAM_ACCOUNT_FOR_LRU
        struct lzx_lru_queue queue;
#endif
};

/* State of the LZX compressor  */
struct lzx_compressor {

        /* The parameters that were used to create the compressor.  */
        struct wimlib_lzx_params params;

        /* The buffer of data to be compressed.
         *
         * 0xe8 byte preprocessing is done directly on the data here before
         * further compression.
         *
         * Note that this compressor does *not* use a sliding window!!!!
         * It's not needed in the WIM format, since every chunk is compressed
         * independently.  This is by design, to allow random access to the
         * chunks.
         *
         * We reserve a few extra bytes to potentially allow reading off the end
         * of the array in the match-finding code for optimization purposes.
         */
        u8 window[LZX_MAX_WINDOW_SIZE + 12];

        /* Number of bytes of data to be compressed, which is the number of
         * bytes of data in @window that are actually valid.  */
        unsigned window_size;

        /* The current match offset LRU queue.  */
        struct lzx_lru_queue queue;

        /* Space for sequence of matches/literals that were chosen.
         *
         * Each LZX_MAX_WINDOW_SIZE-sized portion of this array is used for a
         * different block splitting level.  */
        struct lzx_match *chosen_matches;

        /* Structures used during block splitting.
         *
         * This can be thought of as a binary tree.  block_specs[(1) - 1]
         * represents to the top-level block (root node), and block_specs[(i*2)
         * - 1] and block_specs[(i*2+1) - 1] represent the sub-blocks (child
         * nodes) resulting from a binary split of the block represented by
         * block_spec[(i) - 1].
         */
        struct lzx_block_spec *block_specs;

        /* This is simply filled in with zeroes and used to avoid special-casing
         * the output of the first compressed Huffman code, which conceptually
         * has a delta taken from a code with all symbols having zero-length
         * codewords.  */
        struct lzx_codes zero_codes;

        /* Slow algorithm only: The current cost model.  */
        struct lzx_lens costs;

        /* Slow algorithm only:  Table that maps the hash codes for 3 character
         * sequences to the most recent position that sequence (or a sequence
         * sharing the same hash code) appeared in the window.  */
        u16 *hash_tab;

        /* Slow algorithm only:  Table that maps 2-character sequences to the
         * most recent position that sequence appeared in the window.  */
        u16 *digram_tab;

        /* Slow algorithm only: Table that contains the logical child pointers
         * in the binary trees in the match-finding code.
         *
         * child_tab[i*2] and child_tab[i*2+1] are the left and right pointers,
         * respectively, from the binary tree root corresponding to window
         * position i.  */
        u16 *child_tab;

        /* Slow algorithm only: Matches that were already found and are saved in
         * memory for subsequent queries (e.g. when block splitting).  */
        struct raw_match *cached_matches;

        /* Slow algorithm only: Next position in 'cached_matches' to either
         * return or fill in.  */
        unsigned cached_matches_pos;

        /* Slow algorithm only: %true if reading from 'cached_matches'; %false
         * if writing to 'cached_matches'.  */
        bool matches_already_found;

        /* Slow algorithm only: Position in window of next match to return.  */
        unsigned match_window_pos;

        /* Slow algorithm only: No matches returned shall reach past this
         * position.  */
        unsigned match_window_end;

        /* Slow algorithm only: Temporary space used for match-choosing
         * algorithm.
         *
         * The size of this array must be at least LZX_MAX_MATCH but otherwise
         * is arbitrary.  More space simply allows the match-choosing algorithm
         * to find better matches (depending on the input, as always).  */
        struct lzx_optimal *optimum;

        /* Slow algorithm only: Variables used by the match-choosing algorithm.
         *
         * When matches have been chosen, optimum_cur_idx is set to the position
         * in the window of the next match/literal to return and optimum_end_idx
         * is set to the position in the window at the end of the last
         * match/literal to return.  */
        u32 optimum_cur_idx;
        u32 optimum_end_idx;
};

/* Returns the LZX position slot that corresponds to a given formatted offset.
 *
 * Logically, this returns the smallest i such that
 * formatted_offset >= lzx_position_base[i].
 *
 * The actual implementation below takes advantage of the regularity of the
 * numbers in the lzx_position_base array to calculate the slot directly from
 * the formatted offset without actually looking at the array.
 */
static unsigned
lzx_get_position_slot(unsigned formatted_offset)
{
#if 0
        /*
         * Slots 36-49 (formatted_offset >= 262144) can be found by
         * (formatted_offset/131072) + 34 == (formatted_offset >> 17) + 34;
         * however, this check for formatted_offset >= 262144 is commented out
         * because WIM chunks cannot be that large.
         */
        if (formatted_offset >= 262144) {
                return (formatted_offset >> 17) + 34;
        } else
#endif
        {
                /* Note: this part here only works if:
                 *
                 *    2 <= formatted_offset < 655360
                 *
                 * It is < 655360 because the frequency of the position bases
                 * increases starting at the 655360 entry, and it is >= 2
                 * because the below calculation fails if the most significant
                 * bit is lower than the 2's place. */
                LZX_ASSERT(2 <= formatted_offset  && formatted_offset < 655360);
                unsigned mssb_idx = bsr32(formatted_offset);
                return (mssb_idx << 1) |
                        ((formatted_offset >> (mssb_idx - 1)) & 1);
        }
}

/* Compute the hash code for the next 3-character sequence in the window.  */
static unsigned
lzx_lz_compute_hash(const u8 *window)
{
        unsigned hash;

        hash = window[0];
        hash <<= LZX_LZ_HASH_SHIFT;
        hash ^= window[1];
        hash <<= LZX_LZ_HASH_SHIFT;
        hash ^= window[2];
        return hash & LZX_LZ_HASH_MASK;
}

/* Build the main, length, and aligned offset Huffman codes used in LZX.
 *
 * This takes as input the frequency tables for each code and produces as output
 * a set of tables that map symbols to codewords and lengths.  */
static void
lzx_make_huffman_codes(const struct lzx_freqs *freqs,
                       struct lzx_codes *codes)
{
        make_canonical_huffman_code(LZX_MAINTREE_NUM_SYMBOLS,
                                    LZX_MAX_CODEWORD_LEN,
                                    freqs->main,
                                    codes->lens.main,
                                    codes->codewords.main);

        make_canonical_huffman_code(LZX_LENTREE_NUM_SYMBOLS,
                                    LZX_MAX_CODEWORD_LEN,
                                    freqs->len,
                                    codes->lens.len,
                                    codes->codewords.len);

        make_canonical_huffman_code(LZX_ALIGNEDTREE_NUM_SYMBOLS, 8,
                                    freqs->aligned,
                                    codes->lens.aligned,
                                    codes->codewords.aligned);
}

/*
 * Output a LZX match.
 *
 * @out:         The bitstream to write the match to.
 * @block_type:  The type of the LZX block (LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM)
 * @match:       The match.
 * @codes:       Pointer to a structure that contains the codewords for the
 *               main, length, and aligned offset Huffman codes.
 */
static void
lzx_write_match(struct output_bitstream *out, int block_type,
                struct lzx_match match, const struct lzx_codes *codes)
{
        /* low 8 bits are the match length minus 2 */
        unsigned match_len_minus_2 = match.data & 0xff;
        /* Next 17 bits are the position footer */
        unsigned position_footer = (match.data >> 8) & 0x1ffff; /* 17 bits */
        /* Next 6 bits are the position slot. */
        unsigned position_slot = (match.data >> 25) & 0x3f;     /* 6 bits */
        unsigned len_header;
        unsigned len_footer;
        unsigned len_pos_header;
        unsigned main_symbol;
        unsigned num_extra_bits;
        unsigned verbatim_bits;
        unsigned aligned_bits;

        /* If the match length is less than MIN_MATCH (= 2) +
         * NUM_PRIMARY_LENS (= 7), the length header contains
         * the match length minus MIN_MATCH, and there is no
         * length footer.
         *
         * Otherwise, the length header contains
         * NUM_PRIMARY_LENS, and the length footer contains
         * the match length minus NUM_PRIMARY_LENS minus
         * MIN_MATCH. */
        if (match_len_minus_2 < LZX_NUM_PRIMARY_LENS) {
                len_header = match_len_minus_2;
                /* No length footer-- mark it with a special
                 * value. */
                len_footer = (unsigned)(-1);
        } else {
                len_header = LZX_NUM_PRIMARY_LENS;
                len_footer = match_len_minus_2 - LZX_NUM_PRIMARY_LENS;
        }

        /* Combine the position slot with the length header into
         * a single symbol that will be encoded with the main
         * tree. */
        len_pos_header = (position_slot << 3) | len_header;

        /* The actual main symbol is offset by LZX_NUM_CHARS because
         * values under LZX_NUM_CHARS are used to indicate a literal
         * byte rather than a match. */
        main_symbol = len_pos_header + LZX_NUM_CHARS;

        /* Output main symbol. */
        bitstream_put_bits(out, codes->codewords.main[main_symbol],
                           codes->lens.main[main_symbol]);

        /* If there is a length footer, output it using the
         * length Huffman code. */
        if (len_footer != (unsigned)(-1)) {
                bitstream_put_bits(out, codes->codewords.len[len_footer],
                                   codes->lens.len[len_footer]);
        }

        num_extra_bits = lzx_get_num_extra_bits(position_slot);

        /* For aligned offset blocks with at least 3 extra bits, output the
         * verbatim bits literally, then the aligned bits encoded using the
         * aligned offset tree.  Otherwise, only the verbatim bits need to be
         * output. */
        if ((block_type == LZX_BLOCKTYPE_ALIGNED) && (num_extra_bits >= 3)) {

                verbatim_bits = position_footer >> 3;
                bitstream_put_bits(out, verbatim_bits,
                                   num_extra_bits - 3);

                aligned_bits = (position_footer & 7);
                bitstream_put_bits(out,
                                   codes->codewords.aligned[aligned_bits],
                                   codes->lens.aligned[aligned_bits]);
        } else {
                /* verbatim bits is the same as the position
                 * footer, in this case. */
                bitstream_put_bits(out, position_footer, num_extra_bits);
        }
}

static unsigned
lzx_build_precode(const u8 lens[restrict],
                  const u8 prev_lens[restrict],
                  unsigned num_syms,
                  freq_t precode_freqs[restrict LZX_PRETREE_NUM_SYMBOLS],
                  u8 output_syms[restrict num_syms],
                  u8 precode_lens[restrict LZX_PRETREE_NUM_SYMBOLS],
                  u16 precode_codewords[restrict LZX_PRETREE_NUM_SYMBOLS],
                  unsigned * num_additional_bits_ret)
{
        unsigned output_syms_idx;
        unsigned cur_run_len;
        unsigned i;
        unsigned len_in_run;
        unsigned additional_bits;
        signed char delta;
        unsigned num_additional_bits = 0;

        memset(precode_freqs, 0,
               LZX_PRETREE_NUM_SYMBOLS * sizeof(precode_freqs[0]));

        /* Since the code word lengths use a form of RLE encoding, the goal here
         * is to find each run of identical lengths when going through them in
         * symbol order (including runs of length 1).  For each run, as many
         * lengths are encoded using RLE as possible, and the rest are output
         * literally.
         *
         * output_syms[] will be filled in with the length symbols that will be
         * output, including RLE codes, not yet encoded using the pre-tree.
         *
         * cur_run_len keeps track of how many code word lengths are in the
         * current run of identical lengths.
         */
        output_syms_idx = 0;
        cur_run_len = 1;
        for (i = 1; i <= num_syms; i++) {

                if (i != num_syms && lens[i] == lens[i - 1]) {
                        /* Still in a run--- keep going. */
                        cur_run_len++;
                        continue;
                }

                /* Run ended! Check if it is a run of zeroes or a run of
                 * nonzeroes. */

                /* The symbol that was repeated in the run--- not to be confused
                 * with the length *of* the run (cur_run_len) */
                len_in_run = lens[i - 1];

                if (len_in_run == 0) {
                        /* A run of 0's.  Encode it in as few length
                         * codes as we can. */

                        /* The magic length 18 indicates a run of 20 + n zeroes,
                         * where n is an uncompressed literal 5-bit integer that
                         * follows the magic length. */
                        while (cur_run_len >= 20) {

                                additional_bits = min(cur_run_len - 20, 0x1f);
                                num_additional_bits += 5;
                                precode_freqs[18]++;
                                output_syms[output_syms_idx++] = 18;
                                output_syms[output_syms_idx++] = additional_bits;
                                cur_run_len -= 20 + additional_bits;
                        }

                        /* The magic length 17 indicates a run of 4 + n zeroes,
                         * where n is an uncompressed literal 4-bit integer that
                         * follows the magic length. */
                        while (cur_run_len >= 4) {
                                additional_bits = min(cur_run_len - 4, 0xf);
                                num_additional_bits += 4;
                                precode_freqs[17]++;
                                output_syms[output_syms_idx++] = 17;
                                output_syms[output_syms_idx++] = additional_bits;
                                cur_run_len -= 4 + additional_bits;
                        }

                } else {

                        /* A run of nonzero lengths. */

                        /* The magic length 19 indicates a run of 4 + n
                         * nonzeroes, where n is a literal bit that follows the
                         * magic length, and where the value of the lengths in
                         * the run is given by an extra length symbol, encoded
                         * with the precode, that follows the literal bit.
                         *
                         * The extra length symbol is encoded as a difference
                         * from the length of the codeword for the first symbol
                         * in the run in the previous tree.
                         * */
                        while (cur_run_len >= 4) {
                                additional_bits = (cur_run_len > 4);
                                num_additional_bits += 1;
                                delta = (signed char)prev_lens[i - cur_run_len] -
                                        (signed char)len_in_run;
                                if (delta < 0)
                                        delta += 17;
                                precode_freqs[19]++;
                                precode_freqs[(unsigned char)delta]++;
                                output_syms[output_syms_idx++] = 19;
                                output_syms[output_syms_idx++] = additional_bits;
                                output_syms[output_syms_idx++] = delta;
                                cur_run_len -= 4 + additional_bits;
                        }
                }

                /* Any remaining lengths in the run are outputted without RLE,
                 * as a difference from the length of that codeword in the
                 * previous tree. */
                while (cur_run_len > 0) {
                        delta = (signed char)prev_lens[i - cur_run_len] -
                                (signed char)len_in_run;
                        if (delta < 0)
                                delta += 17;

                        precode_freqs[(unsigned char)delta]++;
                        output_syms[output_syms_idx++] = delta;
                        cur_run_len--;
                }

                cur_run_len = 1;
        }

        /* Build the precode from the frequencies of the length symbols. */

        make_canonical_huffman_code(LZX_PRETREE_NUM_SYMBOLS,
                                    LZX_MAX_CODEWORD_LEN,
                                    precode_freqs, precode_lens,
                                    precode_codewords);

        if (num_additional_bits_ret)
                *num_additional_bits_ret = num_additional_bits;

        return output_syms_idx;
}

/*
 * Writes a compressed Huffman code to the output, preceded by the precode for
 * it.
 *
 * The Huffman code is represented in the output as a series of path lengths
 * from which the canonical Huffman code can be reconstructed.  The path lengths
 * themselves are compressed using a separate Huffman code, the precode, which
 * consists of LZX_PRETREE_NUM_SYMBOLS (= 20) symbols that cover all possible
 * code lengths, plus extra codes for repeated lengths.  The path lengths of the
 * precode precede the path lengths of the larger code and are uncompressed,
 * consisting of 20 entries of 4 bits each.
 *
 * @out:                Bitstream to write the code to.
 * @lens:               The code lengths for the Huffman code, indexed by symbol.
 * @prev_lens:          Code lengths for this Huffman code, indexed by symbol,
 *                      in the *previous block*, or all zeroes if this is the
 *                      first block.
 * @num_syms:           The number of symbols in the code.
 */
static void
lzx_write_compressed_code(struct output_bitstream *out,
                          const u8 lens[restrict],
                          const u8 prev_lens[restrict],
                          unsigned num_syms)
{
        freq_t precode_freqs[LZX_PRETREE_NUM_SYMBOLS];
        u8 output_syms[num_syms];
        u8 precode_lens[LZX_PRETREE_NUM_SYMBOLS];
        u16 precode_codewords[LZX_PRETREE_NUM_SYMBOLS];
        unsigned i;
        unsigned num_output_syms;
        u8 precode_sym;

        num_output_syms = lzx_build_precode(lens,
                                            prev_lens,
                                            num_syms,
                                            precode_freqs,
                                            output_syms,
                                            precode_lens,
                                            precode_codewords,
                                            NULL);

        /* Write the lengths of the precode codes to the output. */
        for (i = 0; i < LZX_PRETREE_NUM_SYMBOLS; i++)
                bitstream_put_bits(out, precode_lens[i],
                                   LZX_PRETREE_ELEMENT_SIZE);

        /* Write the length symbols, encoded with the precode, to the output. */

        for (i = 0; i < num_output_syms; ) {
                precode_sym = output_syms[i++];

                bitstream_put_bits(out, precode_codewords[precode_sym],
                                   precode_lens[precode_sym]);
                switch (precode_sym) {
                case 17:
                        bitstream_put_bits(out, output_syms[i++], 4);
                        break;
                case 18:
                        bitstream_put_bits(out, output_syms[i++], 5);
                        break;
                case 19:
                        bitstream_put_bits(out, output_syms[i++], 1);
                        bitstream_put_bits(out,
                                           precode_codewords[output_syms[i]],
                                           precode_lens[output_syms[i]]);
                        i++;
                        break;
                default:
                        break;
                }
        }
}

static unsigned
lzx_huffman_code_output_cost(const u8 lens[restrict],
                             const freq_t freqs[restrict],
                             unsigned num_syms)
{
        unsigned cost = 0;

        for (unsigned i = 0; i < num_syms; i++)
                cost += (unsigned)lens[i] * (unsigned)freqs[i];

        return cost;
}

/* Return the number of bits required to output the lengths for the specified
 * Huffman code in compressed format (encoded with a precode).  */
static unsigned
lzx_code_cost(const u8 lens[], const u8 prev_lens[], unsigned num_syms)
{
        u8 output_syms[num_syms];
        freq_t precode_freqs[LZX_PRETREE_NUM_SYMBOLS];
        u8 precode_lens[LZX_PRETREE_NUM_SYMBOLS];
        u16 precode_codewords[LZX_PRETREE_NUM_SYMBOLS];
        unsigned cost = 0;
        unsigned num_additional_bits;

        /* Acount for the lengths of the precode itself.  */
        cost += LZX_PRETREE_NUM_SYMBOLS * LZX_PRETREE_ELEMENT_SIZE;

        lzx_build_precode(lens, prev_lens, num_syms,
                          precode_freqs, output_syms,
                          precode_lens, precode_codewords,
                          &num_additional_bits);

        /* Account for all precode symbols output.  */
        cost += lzx_huffman_code_output_cost(precode_lens, precode_freqs,
                                             LZX_PRETREE_NUM_SYMBOLS);

        /* Account for additional bits.  */
        cost += num_additional_bits;

        return cost;
}

/*
 * Writes all compressed matches and literal bytes in a LZX block to the the
 * output bitstream.
 *
 * @out:         The output bitstream.
 * @block_type:  The type of the block (LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM)
 * @match_tab[]:   The array of matches that will be output.  It has length
 *                      of @num_compressed_literals.
 * @num_compressed_literals:  Number of compressed literals to be output.
 * @codes:      Pointer to a structure that contains the codewords for the
 *                      main, length, and aligned offset Huffman codes.
 */
static void
lzx_write_matches_and_literals(struct output_bitstream *ostream,
                               int block_type,
                               const struct lzx_match match_tab[],
                               unsigned match_count,
                               const struct lzx_codes *codes)
{
        for (unsigned i = 0; i < match_count; i++) {
                struct lzx_match match = match_tab[i];

                /* High bit of the match indicates whether the match is an
                 * actual match (1) or a literal uncompressed byte (0)  */
                if (match.data & 0x80000000) {
                        /* match */
                        lzx_write_match(ostream, block_type,
                                        match, codes);
                } else {
                        /* literal byte */
                        bitstream_put_bits(ostream,
                                           codes->codewords.main[match.data],
                                           codes->lens.main[match.data]);
                }
        }
}


static void
lzx_assert_codes_valid(const struct lzx_codes * codes)
{
#ifdef ENABLE_LZX_DEBUG
        unsigned i;

        for (i = 0; i < LZX_MAINTREE_NUM_SYMBOLS; i++)
                LZX_ASSERT(codes->lens.main[i] <= LZX_MAX_CODEWORD_LEN);

        for (i = 0; i < LZX_LENTREE_NUM_SYMBOLS; i++)
                LZX_ASSERT(codes->lens.len[i] <= LZX_MAX_CODEWORD_LEN);

        for (i = 0; i < LZX_ALIGNEDTREE_NUM_SYMBOLS; i++)
                LZX_ASSERT(codes->lens.aligned[i] <= 8);

        const unsigned tablebits = 10;
        u16 decode_table[(1 << tablebits) +
                         (2 * max(LZX_MAINTREE_NUM_SYMBOLS, LZX_LENTREE_NUM_SYMBOLS))]
                         _aligned_attribute(DECODE_TABLE_ALIGNMENT);
        LZX_ASSERT(0 == make_huffman_decode_table(decode_table,
                                                  LZX_MAINTREE_NUM_SYMBOLS,
                                                  tablebits,
                                                  codes->lens.main,
                                                  LZX_MAX_CODEWORD_LEN));
        LZX_ASSERT(0 == make_huffman_decode_table(decode_table,
                                                  LZX_LENTREE_NUM_SYMBOLS,
                                                  tablebits,
                                                  codes->lens.len,
                                                  LZX_MAX_CODEWORD_LEN));
        LZX_ASSERT(0 == make_huffman_decode_table(decode_table,
                                                  LZX_ALIGNEDTREE_NUM_SYMBOLS,
                                                  min(tablebits, 6),
                                                  codes->lens.aligned,
                                                  8));
#endif /* ENABLE_LZX_DEBUG */
}

/* Write a LZX aligned offset or verbatim block to the output.  */
static void
lzx_write_compressed_block(int block_type,
                           unsigned block_size,
                           struct lzx_match * chosen_matches,
                           unsigned num_chosen_matches,
                           const struct lzx_codes * codes,
                           const struct lzx_codes * prev_codes,
                           struct output_bitstream * ostream)
{
        unsigned i;

        LZX_ASSERT(block_type == LZX_BLOCKTYPE_ALIGNED ||
                   block_type == LZX_BLOCKTYPE_VERBATIM);
        LZX_ASSERT(block_size <= LZX_MAX_WINDOW_SIZE);
        LZX_ASSERT(num_chosen_matches <= LZX_MAX_WINDOW_SIZE);
        lzx_assert_codes_valid(codes);

        /* The first three bits indicate the type of block and are one of the
         * LZX_BLOCKTYPE_* constants.  */
        bitstream_put_bits(ostream, block_type, LZX_BLOCKTYPE_NBITS);

        /* The next bit indicates whether the block size is the default (32768),
         * indicated by a 1 bit, or whether the block size is given by the next
         * 16 bits, indicated by a 0 bit.  */
        if (block_size == LZX_DEFAULT_BLOCK_SIZE) {
                bitstream_put_bits(ostream, 1, 1);
        } else {
                bitstream_put_bits(ostream, 0, 1);
                bitstream_put_bits(ostream, block_size, LZX_BLOCKSIZE_NBITS);
        }

        /* Write out lengths of the main code. Note that the LZX specification
         * incorrectly states that the aligned offset code comes after the
         * length code, but in fact it is the very first tree to be written
         * (before the main code).  */
        if (block_type == LZX_BLOCKTYPE_ALIGNED)
                for (i = 0; i < LZX_ALIGNEDTREE_NUM_SYMBOLS; i++)
                        bitstream_put_bits(ostream, codes->lens.aligned[i],
                                           LZX_ALIGNEDTREE_ELEMENT_SIZE);

        LZX_DEBUG("Writing main code...");

        /* Write the pre-tree and lengths for the first LZX_NUM_CHARS symbols in
         * the main code, which are the codewords for literal bytes.  */
        lzx_write_compressed_code(ostream,
                                  codes->lens.main,
                                  prev_codes->lens.main,
                                  LZX_NUM_CHARS);

        /* Write the pre-tree and lengths for the rest of the main code, which
         * are the codewords for match headers.  */
        lzx_write_compressed_code(ostream,
                                  codes->lens.main + LZX_NUM_CHARS,
                                  prev_codes->lens.main + LZX_NUM_CHARS,
                                  LZX_MAINTREE_NUM_SYMBOLS - LZX_NUM_CHARS);

        LZX_DEBUG("Writing length code...");

        /* Write the pre-tree and lengths for the length code.  */
        lzx_write_compressed_code(ostream,
                                  codes->lens.len,
                                  prev_codes->lens.len,
                                  LZX_LENTREE_NUM_SYMBOLS);

        LZX_DEBUG("Writing matches and literals...");

        /* Write the actual matches and literals.  */
        lzx_write_matches_and_literals(ostream, block_type,
                                       chosen_matches, num_chosen_matches,
                                       codes);

        LZX_DEBUG("Done writing block.");
}

/* Write the LZX block of index @block_number, or recurse to its children
 * recursively if it is a split block.
 *
 * Return a pointer to the Huffman codes for the last block written.
 */
static struct lzx_codes *
lzx_write_block_recursive(struct lzx_compressor *ctx,
                          unsigned block_number,
                          struct lzx_codes * prev_codes,
                          struct output_bitstream *ostream)
{
        struct lzx_block_spec *spec = &ctx->block_specs[block_number - 1];

        if (spec->is_split) {
                prev_codes = lzx_write_block_recursive(ctx, block_number * 2 + 0,
                                                       prev_codes, ostream);
                prev_codes = lzx_write_block_recursive(ctx, block_number * 2 + 1,
                                                       prev_codes, ostream);
        } else {
                LZX_DEBUG("Writing block #%u (type=%d, size=%u, num_chosen_matches=%u)...",
                          block_number, spec->block_type, spec->block_size,
                          spec->num_chosen_matches);
                lzx_write_compressed_block(spec->block_type,
                                           spec->block_size,
                                           &ctx->chosen_matches[spec->chosen_matches_start_pos],
                                           spec->num_chosen_matches,
                                           &spec->codes,
                                           prev_codes,
                                           ostream);
                prev_codes = &spec->codes;
        }
        return prev_codes;
}

/* Write out the LZX blocks that were computed.  */
static void
lzx_write_all_blocks(struct lzx_compressor *ctx, struct output_bitstream *ostream)
{
        lzx_write_block_recursive(ctx, 1, &ctx->zero_codes, ostream);
}

static u32
lzx_record_literal(u8 literal, void *_freqs)
{
        struct lzx_freqs *freqs = _freqs;

        freqs->main[literal]++;

        return (u32)literal;
}

/* Constructs a match from an offset and a length, and updates the LRU queue and
 * the frequency of symbols in the main, length, and aligned offset alphabets.
 * The return value is a 32-bit number that provides the match in an
 * intermediate representation documented below.  */
static u32
lzx_record_match(unsigned match_offset, unsigned match_len,
                 void *_freqs, void *_queue)
{
        struct lzx_freqs *freqs = _freqs;
        struct lzx_lru_queue *queue = _queue;
        unsigned position_slot;
        unsigned position_footer = 0;
        u32 len_header;
        u32 len_pos_header;
        unsigned len_footer;
        unsigned adjusted_match_len;

        LZX_ASSERT(match_len >= LZX_MIN_MATCH && match_len <= LZX_MAX_MATCH);

        /* If possible, encode this offset as a repeated offset. */
        if (match_offset == queue->R0) {
                position_slot = 0;
        } else if (match_offset == queue->R1) {
                swap(queue->R0, queue->R1);
                position_slot = 1;
        } else if (match_offset == queue->R2) {
                swap(queue->R0, queue->R2);
                position_slot = 2;
        } else {
                /* Not a repeated offset. */

                /* offsets of 0, 1, and 2 are reserved for the repeated offset
                 * codes, so non-repeated offsets must be encoded as 3+.  The
                 * minimum offset is 1, so encode the offsets offset by 2. */
                unsigned formatted_offset = match_offset + 2;

                queue->R2 = queue->R1;
                queue->R1 = queue->R0;
                queue->R0 = match_offset;

                /* The (now-formatted) offset will actually be encoded as a
                 * small position slot number that maps to a certain hard-coded
                 * offset (position base), followed by a number of extra bits---
                 * the position footer--- that are added to the position base to
                 * get the original formatted offset. */

                position_slot = lzx_get_position_slot(formatted_offset);
                position_footer = formatted_offset &
                                  ((1 << lzx_get_num_extra_bits(position_slot)) - 1);
        }

        adjusted_match_len = match_len - LZX_MIN_MATCH;


        /* The match length must be at least 2, so let the adjusted match length
         * be the match length minus 2.
         *
         * If it is less than 7, the adjusted match length is encoded as a 3-bit
         * number offset by 2.  Otherwise, the 3-bit length header is all 1's
         * and the actual adjusted length is given as a symbol encoded with the
         * length tree, offset by 7.
         */
        if (adjusted_match_len < LZX_NUM_PRIMARY_LENS) {
                len_header = adjusted_match_len;
        } else {
                len_header = LZX_NUM_PRIMARY_LENS;
                len_footer = adjusted_match_len - LZX_NUM_PRIMARY_LENS;
                freqs->len[len_footer]++;
        }
        len_pos_header = (position_slot << 3) | len_header;

        freqs->main[len_pos_header + LZX_NUM_CHARS]++;

        /* Equivalent to:
         * if (lzx_extra_bits[position_slot] >= 3) */
        if (position_slot >= 8)
                freqs->aligned[position_footer & 7]++;

        /* Pack the position slot, position footer, and match length into an
         * intermediate representation.
         *
         * bits    description
         * ----    -----------------------------------------------------------
         *
         * 31      1 if a match, 0 if a literal.
         *
         * 30-25   position slot.  This can be at most 50, so it will fit in 6
         *         bits.
         *
         * 8-24    position footer.  This is the offset of the real formatted
         *         offset from the position base.  This can be at most 17 bits
         *         (since lzx_extra_bits[LZX_NUM_POSITION_SLOTS - 1] is 17).
         *
         * 0-7     length of match, offset by 2.  This can be at most
         *         (LZX_MAX_MATCH - 2) == 255, so it will fit in 8 bits.  */
        return 0x80000000 |
                (position_slot << 25) |
                (position_footer << 8) |
                (adjusted_match_len);
}

/* Set the cost model @ctx->costs from the Huffman codeword lengths specified in
 * @lens.
 *
 * These are basically the same thing, except that Huffman codewords with length
 * 0 corresponds to symbols with zero frequency.  These need to be assigned
 * actual costs.  The specific values assigned are arbitrary, but they should be
 * fairly high (near the maximum codeword length) to take into account the fact
 * that uses of these symbols are expected to be rare.
 */
static void
lzx_set_costs(struct lzx_compressor * ctx, const struct lzx_lens * lens)
{
        unsigned i;

        memcpy(&ctx->costs, lens, sizeof(struct lzx_lens));

        for (i = 0; i < LZX_MAINTREE_NUM_SYMBOLS; i++)
                if (ctx->costs.main[i] == 0)
                        ctx->costs.main[i] = ctx->params.slow.main_nostat_cost;

        for (i = 0; i < LZX_LENTREE_NUM_SYMBOLS; i++)
                if (ctx->costs.len[i] == 0)
                        ctx->costs.len[i] = ctx->params.slow.len_nostat_cost;

        for (i = 0; i < LZX_ALIGNEDTREE_NUM_SYMBOLS; i++)
                if (ctx->costs.aligned[i] == 0)
                        ctx->costs.aligned[i] = ctx->params.slow.aligned_nostat_cost;
}

static u32
lzx_literal_cost(u8 c, const struct lzx_lens * costs)
{
        return costs->main[c];
}

/* Given a (length, offset) pair that could be turned into a valid LZX match as
 * well as costs for the codewords in the main, length, and aligned Huffman
 * codes, return the approximate number of bits it will take to represent this
 * match in the compressed output.  */
static unsigned
lzx_match_cost(unsigned length, unsigned offset, const struct lzx_lens *costs

#if LZX_PARAM_ACCOUNT_FOR_LRU
               , struct lzx_lru_queue *queue
#endif
        )
{
        unsigned position_slot, len_header, main_symbol;
        unsigned cost = 0;

        /* Calculate position slot and length header, then combine them into the
         * main symbol.  */

#if LZX_PARAM_ACCOUNT_FOR_LRU
        if (offset == queue->R0) {
                position_slot = 0;
        } else if (offset == queue->R1) {
                swap(queue->R0, queue->R1);
                position_slot = 1;
        } else if (offset == queue->R2) {
                swap(queue->R0, queue->R2);
                position_slot = 2;
        } else
#endif
                position_slot = lzx_get_position_slot(offset + 2);

        len_header = min(length - LZX_MIN_MATCH, LZX_NUM_PRIMARY_LENS);
        main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS;

        /* Account for main symbol.  */
        cost += costs->main[main_symbol];

        /* Account for extra position information.  */
        unsigned num_extra_bits = lzx_get_num_extra_bits(position_slot);
        if (num_extra_bits >= 3) {
                cost += num_extra_bits - 3;
                cost += costs->aligned[(offset + LZX_MIN_MATCH) & 7];
        } else {
                cost += num_extra_bits;
        }

        /* Account for extra length information.  */
        if (length - LZX_MIN_MATCH >= LZX_NUM_PRIMARY_LENS)
                cost += costs->len[length - LZX_MIN_MATCH - LZX_NUM_PRIMARY_LENS];

        return cost;
}

/* This procedure effectively creates a new binary tree corresponding to the
 * current string at the same time that it searches the existing tree nodes for
 * matches.  */
static unsigned
lzx_lz_get_matches(const u8 window[restrict],
                   const unsigned bytes_remaining,
                   const unsigned strstart,
                   const unsigned max_length,
                   u16 child_tab[restrict],
                   unsigned cur_match,
                   const unsigned prev_len,
                   struct raw_match * const matches)
{
        u16 *new_tree_lt_ptr = &child_tab[strstart * 2];
        u16 *new_tree_gt_ptr = &child_tab[strstart * 2 + 1];

        u16 longest_lt_match_len = 0;
        u16 longest_gt_match_len = 0;

        /* Maximum number of nodes to walk down before stopping  */
        unsigned depth = max_length;

        /* Length of longest match found so far  */
        unsigned longest_match_len = prev_len;

        /* Maximum length of match to return  */
        unsigned len_limit = min(bytes_remaining, max_length);

        /* Number of matches found so far  */
        unsigned num_matches = 0;

        for (;;) {

                /* Stop if too many nodes were traversed or if there is no next
                 * node  */
                if (depth-- == 0 || cur_match == 0) {
                        *new_tree_gt_ptr = 0;
                        *new_tree_lt_ptr = 0;
                        return num_matches;
                }

                /* Load the pointers to the children of the binary tree node
                 * corresponding to the current match  */
                u16 * const cur_match_ptrs = &child_tab[cur_match * 2];

                /* Set up pointers to the current match and to the current
                 * string  */
                const u8 * const matchptr = &window[cur_match];
                const u8 * const strptr = &window[strstart];

                u16 len = min(longest_lt_match_len,
                              longest_gt_match_len);

                if (matchptr[len] == strptr[len]) {
                        while (++len != len_limit)
                                if (matchptr[len] != strptr[len])
                                        break;

                        if (len > longest_match_len) {
                                longest_match_len = len;
                                matches[num_matches].len = len;
                                matches[num_matches].offset = strstart - cur_match;
                                num_matches++;

                                if (len == len_limit) {
                                        /* Length limit was reached.  Link left pointer
                                         * in the new tree with left subtree of current
                                         * match tree, and link the right pointer in the
                                         * new tree with the right subtree of the
                                         * current match tree.  This in effect deletes
                                         * the node for the currrent match, which is
                                         * desirable because the current match is the
                                         * same as the current string up until the
                                         * length limit, so in subsequent queries it
                                         * will never be preferable to the current
                                         * position.  */
                                        *new_tree_lt_ptr = cur_match_ptrs[0];
                                        *new_tree_gt_ptr = cur_match_ptrs[1];
                                        return num_matches;
                                }
                        }
                }

                if (matchptr[len] < strptr[len]) {
                        /* Case 1:  The current match is lexicographically less
                         * than the current string.
                         *
                         * Since we are searching the binary tree structures, we
                         * need to walk down to the *right* subtree of the
                         * current match's node to get to a match that is
                         * lexicographically *greater* than the current match
                         * but still lexicographically *lesser* than the current
                         * string.
                         *
                         * At the same time, we link the entire binary tree
                         * corresponding to the current match into the
                         * appropriate place in the new binary tree being built
                         * for the current string.  */
                        *new_tree_lt_ptr = cur_match;
                        new_tree_lt_ptr = &cur_match_ptrs[1];
                        cur_match = *new_tree_lt_ptr;
                        longest_lt_match_len = len;
                } else {
                        /* Case 2:  The current match is lexicographically
                         * greater than the current string.
                         *
                         * This is analogous to Case 1 above, but everything
                         * happens in the other direction.
                         */
                        *new_tree_gt_ptr = cur_match;
                        new_tree_gt_ptr = &cur_match_ptrs[0];
                        cur_match = *new_tree_gt_ptr;
                        longest_gt_match_len = len;
                }
        }
}

/* Equivalent to lzx_lz_get_matches(), but only updates the tree and doesn't
 * return matches.  See that function for details (including comments).  */
static void
lzx_lz_skip_matches(const u8 window[restrict],
                    const unsigned bytes_remaining,
                    const unsigned strstart,
                    const unsigned max_length,
                    u16 child_tab[restrict],
                    unsigned cur_match,
                    const unsigned prev_len)
{
        u16 *new_tree_lt_ptr = &child_tab[strstart * 2];
        u16 *new_tree_gt_ptr = &child_tab[strstart * 2 + 1];

        u16 longest_lt_match_len = 0;
        u16 longest_gt_match_len = 0;

        unsigned depth = max_length;

        unsigned longest_match_len = prev_len;

        unsigned len_limit = min(bytes_remaining, max_length);

        for (;;) {
                if (depth-- == 0 || cur_match == 0) {
                        *new_tree_gt_ptr = 0;
                        *new_tree_lt_ptr = 0;
                        return;
                }

                u16 * const cur_match_ptrs = &child_tab[cur_match * 2];

                const u8 * const matchptr = &window[cur_match];
                const u8 * const strptr = &window[strstart];

                u16 len = min(longest_lt_match_len,
                              longest_gt_match_len);

                if (matchptr[len] == strptr[len]) {
                        while (++len != len_limit)
                                if (matchptr[len] != strptr[len])
                                        break;

                        if (len > longest_match_len) {
                                longest_match_len = len;

                                if (len == len_limit) {
                                        *new_tree_lt_ptr = cur_match_ptrs[0];
                                        *new_tree_gt_ptr = cur_match_ptrs[1];
                                        return;
                                }
                        }
                }

                if (matchptr[len] < strptr[len]) {
                        *new_tree_lt_ptr = cur_match;
                        new_tree_lt_ptr = &cur_match_ptrs[1];
                        cur_match = *new_tree_lt_ptr;
                        longest_lt_match_len = len;
                } else {
                        *new_tree_gt_ptr = cur_match;
                        new_tree_gt_ptr = &cur_match_ptrs[0];
                        cur_match = *new_tree_gt_ptr;
                        longest_gt_match_len = len;
                }
        }
}

static unsigned
lzx_lz_get_matches_caching(struct lzx_compressor *ctx,
                           struct raw_match **matches_ret);

/* Tell the match-finder to skip the specified number of bytes (@n) in the
 * input.  */
static void
lzx_lz_skip_bytes(struct lzx_compressor *ctx, unsigned n)
{

#if LZX_PARAM_DONT_SKIP_MATCHES
        /* Option 1: Still cache the matches from the positions skipped.  They
         * will then be available in later passes.  */
        struct raw_match *matches;
        while (n--)
                lzx_lz_get_matches_caching(ctx, &matches);
#else
        /* Option 2: Simply mark the positions skipped as having no matches
         * available.  */
        LZX_ASSERT(n <= ctx->match_window_end - ctx->match_window_pos);
        if (ctx->matches_already_found) {
                while (n--) {
                        LZX_ASSERT(ctx->cached_matches[ctx->cached_matches_pos].offset ==
                                   ctx->match_window_pos);
                        ctx->cached_matches_pos += ctx->cached_matches[ctx->cached_matches_pos].len + 1;
                        ctx->match_window_pos++;
                }
        } else {
                while (n--) {
                        if (ctx->params.slow.use_len2_matches &&
                            ctx->match_window_end - ctx->match_window_pos >= 2) {
                                unsigned c1 = ctx->window[ctx->match_window_pos];
                                unsigned c2 = ctx->window[ctx->match_window_pos + 1];
                                unsigned digram = c1 | (c2 << 8);
                                ctx->digram_tab[digram] = ctx->match_window_pos;
                        }
                        if (ctx->match_window_end - ctx->match_window_pos >= 3) {
                                unsigned hash;
                                unsigned cur_match;

                                hash = lzx_lz_compute_hash(&ctx->window[ctx->match_window_pos]);

                                cur_match = ctx->hash_tab[hash];
                                ctx->hash_tab[hash] = ctx->match_window_pos;

                                lzx_lz_skip_matches(ctx->window,
                                                    ctx->match_window_end - ctx->match_window_pos,
                                                    ctx->match_window_pos,
                                                    ctx->params.slow.num_fast_bytes,
                                                    ctx->child_tab,
                                                    cur_match, 1);
                        }
                        ctx->cached_matches[ctx->cached_matches_pos].len = 0;
                        ctx->cached_matches[ctx->cached_matches_pos].offset = ctx->match_window_pos;
                        ctx->cached_matches_pos++;
                        ctx->match_window_pos++;
                }
        }
#endif /* !LZX_PARAM_DONT_SKIP_MATCHES */
}

/* Retrieve a list of matches available at the next position in the input.
 *
 * The return value is the number of matches found, and a pointer to them is
 * written to @matches_ret.  The matches will be sorted in order by length.
 *
 * This is essentially a wrapper around lzx_lz_get_matches() that caches its
 * output the first time and also performs the needed hashing.
 */
static unsigned
lzx_lz_get_matches_caching(struct lzx_compressor *ctx,
                           struct raw_match **matches_ret)
{
        unsigned num_matches;
        struct raw_match *matches;

        LZX_ASSERT(ctx->match_window_end >= ctx->match_window_pos);

        matches = &ctx->cached_matches[ctx->cached_matches_pos + 1];

        if (ctx->matches_already_found) {
                num_matches = ctx->cached_matches[ctx->cached_matches_pos].len;
                LZX_ASSERT(ctx->cached_matches[ctx->cached_matches_pos].offset == ctx->match_window_pos);

                for (int i = (int)num_matches - 1; i >= 0; i--) {
                        if (ctx->match_window_pos + matches[i].len > ctx->match_window_end)
                                matches[i].len = ctx->match_window_end - ctx->match_window_pos;
                        else
                                break;
                }
        } else {
                unsigned prev_len = 1;
                struct raw_match * matches_ret = &ctx->cached_matches[ctx->cached_matches_pos + 1];
                num_matches = 0;

                if (ctx->params.slow.use_len2_matches &&
                    ctx->match_window_end - ctx->match_window_pos >= 3) {
                        unsigned c1 = ctx->window[ctx->match_window_pos];
                        unsigned c2 = ctx->window[ctx->match_window_pos + 1];
                        unsigned digram = c1 | (c2 << 8);
                        unsigned cur_match;

                        cur_match = ctx->digram_tab[digram];
                        ctx->digram_tab[digram] = ctx->match_window_pos;
                        if (cur_match != 0 &&
                            ctx->window[cur_match + 2] != ctx->window[ctx->match_window_pos + 2])
                        {
                                matches_ret->len = 2;
                                matches_ret->offset = ctx->match_window_pos - cur_match;
                                matches_ret++;
                                num_matches++;
                                prev_len = 2;
                        }
                }
                if (ctx->match_window_end - ctx->match_window_pos >= 3) {
                        unsigned hash;
                        unsigned cur_match;

                        hash = lzx_lz_compute_hash(&ctx->window[ctx->match_window_pos]);

                        cur_match = ctx->hash_tab[hash];
                        ctx->hash_tab[hash] = ctx->match_window_pos;
                        num_matches += lzx_lz_get_matches(ctx->window,
                                                          ctx->match_window_end - ctx->match_window_pos,
                                                          ctx->match_window_pos,
                                                          ctx->params.slow.num_fast_bytes,
                                                          ctx->child_tab,
                                                          cur_match,
                                                          prev_len,
                                                          matches_ret);
                }

                ctx->cached_matches[ctx->cached_matches_pos].len = num_matches;
                ctx->cached_matches[ctx->cached_matches_pos].offset = ctx->match_window_pos;

                if (num_matches) {
                        struct raw_match *longest_match_ptr =
                                &ctx->cached_matches[ctx->cached_matches_pos + 1 +
                                                     num_matches - 1];
                        u16 len = longest_match_ptr->len;

                        /* If the longest match returned by the match-finder
                         * reached the number of fast bytes, extend it as much
                         * as possible.  */
                        if (len == ctx->params.slow.num_fast_bytes) {
                                const unsigned maxlen =
                                        min(ctx->match_window_end - ctx->match_window_pos,
                                            LZX_MAX_MATCH);

                                const u8 * const matchptr =
                                        &ctx->window[ctx->match_window_pos - longest_match_ptr->offset];

                                const u8 * const strptr =
                                        &ctx->window[ctx->match_window_pos];

                                while (len < maxlen && matchptr[len] == strptr[len])
                                        len++;
                        }
                        longest_match_ptr->len = len;
                }
        }
        ctx->cached_matches_pos += num_matches + 1;
        *matches_ret = matches;

#if 0
        printf("\n");
        for (unsigned i = 0; i < num_matches; i++)
        {
                printf("Len %u Offset %u\n", matches[i].len, matches[i].offset);
        }
#endif

        for (unsigned i = 0; i < num_matches; i++) {
                LZX_ASSERT(matches[i].len <= LZX_MAX_MATCH);
                if (matches[i].len >= LZX_MIN_MATCH) {
                        LZX_ASSERT(matches[i].offset <= ctx->match_window_pos);
                        LZX_ASSERT(matches[i].len <= ctx->match_window_end - ctx->match_window_pos);
                        LZX_ASSERT(!memcmp(&ctx->window[ctx->match_window_pos],
                                           &ctx->window[ctx->match_window_pos - matches[i].offset],
                                           matches[i].len));
                }
        }

        ctx->match_window_pos++;
        return num_matches;
}

/*
 * Reverse the linked list of near-optimal matches so that they can be returned
 * in forwards order.
 *
 * Returns the first match in the list.
 */
static struct raw_match
lzx_lz_reverse_near_optimal_match_list(struct lzx_compressor *ctx,
                                       unsigned cur_pos)
{
        unsigned prev_link, saved_prev_link;
        unsigned prev_match_offset, saved_prev_match_offset;

        ctx->optimum_end_idx = cur_pos;

        saved_prev_link = ctx->optimum[cur_pos].prev.link;
        saved_prev_match_offset = ctx->optimum[cur_pos].prev.match_offset;

        do {
                prev_link = saved_prev_link;
                prev_match_offset = saved_prev_match_offset;

                saved_prev_link = ctx->optimum[prev_link].prev.link;
                saved_prev_match_offset = ctx->optimum[prev_link].prev.match_offset;

                ctx->optimum[prev_link].next.link = cur_pos;
                ctx->optimum[prev_link].next.match_offset = prev_match_offset;

                cur_pos = prev_link;
        } while (cur_pos != 0);

        ctx->optimum_cur_idx = ctx->optimum[0].next.link;

        return (struct raw_match)
                { .len = ctx->optimum_cur_idx,
                  .offset = ctx->optimum[0].next.match_offset,
                };
}

/*
 * lzx_lz_get_near_optimal_match() -
 *
 * Choose the "best" match or literal to use at the next position in the input.
 *
 * Unlike a "greedy" parser that always takes the longest match, or even a
 * parser with one match/literal look-ahead like zlib, the algorithm used here
 * may look ahead many matches/literals to determine the best match/literal to
 * output next.  The motivation is that the compression ratio is improved if the
 * compressor can do things like use a shorter-than-possible match in order to
 * allow a longer match later, and also take into account the Huffman code cost
 * model rather than simply assuming that longer is better.  It is not a true
 * "optimal" parser, however, since some shortcuts can be taken; for example, if
 * a match is very long, the parser just chooses it immediately before too much
 * time is wasting considering many different alternatives that are unlikely to
 * be better.
 *
 * This algorithm is based on that used in 7-Zip's deflate encoder.
 *
 * Each call to this function does one of two things:
 *
 * 1. Build a near-optimal sequence of matches/literals, up to some point, that
 *    will be returned by subsequent calls to this function, then return the
 *    first one.
 *
 * OR
 *
 * 2. Return the next match/literal previously computed by a call to this
 *    function;
 *
 * This function relies on the following state in the compressor context:
 *
 *      ctx->window          (read-only: preprocessed data being compressed)
 *      ctx->cost            (read-only: cost model to use)
 *      ctx->optimum         (internal state; leave uninitialized)
 *      ctx->optimum_cur_idx (must set to 0 before first call)
 *      ctx->optimum_end_idx (must set to 0 before first call)
 *      ctx->hash_tab        (must set to 0 before first call)
 *      ctx->cached_matches  (internal state; leave uninitialized)
 *      ctx->cached_matches_pos (initialize to 0 before first call; save and
 *                               restore value if restarting parse from a
 *                               certain position)
 *      ctx->match_window_pos (must initialize to position of next match to
 *                             return; subsequent calls return subsequent
 *                             matches)
 *      ctx->match_window_end (must initialize to limit of match-finding region;
 *                             subsequent calls use the same limit)
 *
 * The return value is a (length, offset) pair specifying the match or literal
 * chosen.
 */
static struct raw_match
lzx_lz_get_near_optimal_match(struct lzx_compressor * ctx)
{
#if 0
        /* Testing: literals only  */
        ctx->match_window_pos++;
        return (struct raw_match) { .len = 0 };
#elif 0
        /* Testing: greedy parsing  */
        struct raw_match *matches;
        unsigned num_matches;
        struct raw_match match = {.len = 0};

        num_matches = lzx_lz_get_matches_caching(ctx, &matches);
        if (num_matches) {
                match = matches[num_matches - 1];
                lzx_lz_skip_bytes(ctx, match.len - 1);
        }
        return match;
#else
        unsigned num_possible_matches;
        struct raw_match *possible_matches;
        struct raw_match match;
        unsigned longest_match_len;
        unsigned len, match_idx;

        if (ctx->optimum_cur_idx != ctx->optimum_end_idx) {
                /* Case 2: Return the next match/literal already found.  */
                match.len = ctx->optimum[ctx->optimum_cur_idx].next.link -
                                    ctx->optimum_cur_idx;
                match.offset = ctx->optimum[ctx->optimum_cur_idx].next.match_offset;

                ctx->optimum_cur_idx = ctx->optimum[ctx->optimum_cur_idx].next.link;
                return match;
        }

        /* Case 1:  Compute a new list of matches/literals to return.  */

        ctx->optimum_cur_idx = 0;
        ctx->optimum_end_idx = 0;

        /* Get matches at this position.  */
        num_possible_matches = lzx_lz_get_matches_caching(ctx, &possible_matches);

        /* If no matches found, return literal.  */
        if (num_possible_matches == 0)
                return (struct raw_match){ .len = 0 };

        /* The matches that were found are sorted by length.  Get the length of
         * the longest one.  */
        longest_match_len = possible_matches[num_possible_matches - 1].len;

        /* Greedy heuristic:  if the longest match that was found is greater
         * than LZX_PARAM_NUM_FAST_BYTES, return it immediately; don't both
         * doing more work.  */
        if (longest_match_len > ctx->params.slow.num_fast_bytes) {
                lzx_lz_skip_bytes(ctx, longest_match_len - 1);
                return possible_matches[num_possible_matches - 1];
        }

        /* Calculate the cost to reach the next position by outputting a
         * literal.  */
#if LZX_PARAM_ACCOUNT_FOR_LRU
        ctx->optimum[0].queue = ctx->queue;
        ctx->optimum[1].queue = ctx->optimum[0].queue;
#endif
        ctx->optimum[1].cost = lzx_literal_cost(ctx->window[ctx->match_window_pos],
                                                &ctx->costs);
        ctx->optimum[1].prev.link = 0;

        /* Calculate the cost to reach any position up to and including that
         * reached by the longest match, using the shortest (i.e. closest) match
         * that reaches each position.  */
        match_idx = 0;
        BUILD_BUG_ON(LZX_MIN_MATCH != 2);
        for (len = LZX_MIN_MATCH; len <= longest_match_len; len++) {

                LZX_ASSERT(match_idx < num_possible_matches);

        #if LZX_PARAM_ACCOUNT_FOR_LRU
                ctx->optimum[len].queue = ctx->optimum[0].queue;
        #endif
                ctx->optimum[len].prev.link = 0;
                ctx->optimum[len].prev.match_offset = possible_matches[match_idx].offset;
                ctx->optimum[len].cost = lzx_match_cost(len,
                                                        possible_matches[match_idx].offset,
                                                        &ctx->costs
                                                #if LZX_PARAM_ACCOUNT_FOR_LRU
                                                        , &ctx->optimum[len].queue
                                                #endif
                                                        );
                if (len == possible_matches[match_idx].len)
                        match_idx++;
        }

        unsigned cur_pos = 0;

        /* len_end: greatest index forward at which costs have been calculated
         * so far  */
        unsigned len_end = longest_match_len;


        for (;;) {
                /* Advance to next position.  */
                cur_pos++;

                if (cur_pos == len_end || cur_pos == LZX_PARAM_OPTIM_ARRAY_SIZE)
                        return lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos);

                /* retrieve the number of matches available at this position  */
                num_possible_matches = lzx_lz_get_matches_caching(ctx,
                                                                  &possible_matches);

                unsigned new_len = 0;

                if (num_possible_matches != 0) {
                        new_len = possible_matches[num_possible_matches - 1].len;

                        /* Greedy heuristic:  if we found a match greater than
                         * LZX_PARAM_NUM_FAST_BYTES, stop immediately.  */
                        if (new_len > ctx->params.slow.num_fast_bytes) {

                                /* Build the list of matches to return and get
                                 * the first one.  */
                                match = lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos);

                                /* Append the long match to the end of the list.  */
                                ctx->optimum[cur_pos].next.match_offset =
                                        possible_matches[num_possible_matches - 1].offset;
                                ctx->optimum[cur_pos].next.link = cur_pos + new_len;
                                ctx->optimum_end_idx = cur_pos + new_len;

                                /* Skip over the remaining bytes of the long match.  */
                                lzx_lz_skip_bytes(ctx, new_len - 1);

                                /* Return first match in the list  */
                                return match;
                        }
                }

                /* Consider proceeding with a literal byte.  */
                u32 cur_cost = ctx->optimum[cur_pos].cost;
                u32 cur_plus_literal_cost = cur_cost +
                        lzx_literal_cost(ctx->window[ctx->match_window_pos - 1],
                                         &ctx->costs);
                if (cur_plus_literal_cost < ctx->optimum[cur_pos + 1].cost) {
                        ctx->optimum[cur_pos + 1].cost = cur_plus_literal_cost;
                        ctx->optimum[cur_pos + 1].prev.link = cur_pos;
                #if LZX_PARAM_ACCOUNT_FOR_LRU
                        ctx->optimum[cur_pos + 1].queue = ctx->optimum[cur_pos].queue;
                #endif
                }

                if (num_possible_matches == 0)
                        continue;

                /* Consider proceeding with a match.  */

                while (len_end < cur_pos + new_len)
                        ctx->optimum[++len_end].cost = ~(u32)0;

                match_idx = 0;
                for (len = LZX_MIN_MATCH; len <= new_len; len++) {
                        LZX_ASSERT(match_idx < num_possible_matches);
                #if LZX_PARAM_ACCOUNT_FOR_LRU
                        struct lzx_lru_queue q = ctx->optimum[cur_pos].queue;
                #endif
                        u32 cost = cur_cost + lzx_match_cost(len,
                                                             possible_matches[match_idx].offset,
                                                             &ctx->costs
                                                        #if LZX_PARAM_ACCOUNT_FOR_LRU
                                                             , &q
                                                        #endif
                                                             );

                        if (cost < ctx->optimum[cur_pos + len].cost) {
                                ctx->optimum[cur_pos + len].cost = cost;
                                ctx->optimum[cur_pos + len].prev.link = cur_pos;
                                ctx->optimum[cur_pos + len].prev.match_offset =
                                                possible_matches[match_idx].offset;
                        #if LZX_PARAM_ACCOUNT_FOR_LRU
                                ctx->optimum[cur_pos + len].queue = q;
                        #endif
                        }

                        if (len == possible_matches[match_idx].len)
                                match_idx++;
                }
        }
#endif
}

/* Account for extra bits in the main symbols.  */
static void
lzx_update_mainsym_match_costs(int block_type,
                               u8 main_lens[LZX_MAINTREE_NUM_SYMBOLS])
{
        unsigned i;

        LZX_ASSERT(block_type == LZX_BLOCKTYPE_ALIGNED ||
                   block_type == LZX_BLOCKTYPE_VERBATIM);

        for (i = LZX_NUM_CHARS; i < LZX_MAINTREE_NUM_SYMBOLS; i++) {
                unsigned position_slot = (i >> 3) & 0x1f;

                /* If it's a verbatim block, add the number of extra bits
                 * corresponding to the position slot.
                 *
                 * If it's an aligned block and there would normally be at least
                 * 3 extra bits, count 3 less because they will be output as an
                 * aligned offset symbol instead.  */
                unsigned num_extra_bits = lzx_get_num_extra_bits(position_slot);

                if (block_type == LZX_BLOCKTYPE_ALIGNED && num_extra_bits >= 3)
                        num_extra_bits -= 3;
                main_lens[i] += num_extra_bits;
        }
}

/*
 * Compute the costs, in bits, to output a compressed block as aligned offset
 * and verbatim.
 *
 * @block_size
 *      Number of bytes of uncompressed data this block represents.
 * @codes
 *      Huffman codes that will be used to output the block.
 * @prev_codes
 *      Huffman codes for the previous block, or all zeroes if this is the first
 *      block.
 * @freqs
 *      Frequencies of Huffman symbol that will be output in the block.
 * @aligned_cost_ret
 *      Cost of aligned block will be returned here.
 * @verbatim_cost_ret
 *      Cost of verbatim block will be returned here.
 */
static void
lzx_compute_compressed_block_costs(unsigned block_size,
                                   const struct lzx_codes *codes,
                                   const struct lzx_codes *prev_codes,
                                   const struct lzx_freqs *freqs,
                                   unsigned * aligned_cost_ret,
                                   unsigned * verbatim_cost_ret)
{
        unsigned common_cost = 0;
        unsigned aligned_cost = 0;
        unsigned verbatim_cost = 0;

        u8 updated_main_lens[LZX_MAINTREE_NUM_SYMBOLS];

        /* Account for cost of block header.  */
        common_cost += LZX_BLOCKTYPE_NBITS;
        if (block_size == LZX_DEFAULT_BLOCK_SIZE)
                common_cost += 1;
        else
                common_cost += LZX_BLOCKSIZE_NBITS;

        /* Account for cost of outputting aligned offset code.  */
        aligned_cost += LZX_ALIGNEDTREE_NUM_SYMBOLS * LZX_ALIGNEDTREE_ELEMENT_SIZE;

        /* Account for cost of outputting main and length codes.  */
        common_cost += lzx_code_cost(codes->lens.main,
                                     prev_codes->lens.main,
                                     LZX_NUM_CHARS);
        common_cost += lzx_code_cost(codes->lens.main + LZX_NUM_CHARS,
                                     prev_codes->lens.main + LZX_NUM_CHARS,
                                     LZX_MAINTREE_NUM_SYMBOLS - LZX_NUM_CHARS);
        common_cost += lzx_code_cost(codes->lens.len,
                                     prev_codes->lens.len,
                                     LZX_LENTREE_NUM_SYMBOLS);

        /* Account for cost to output main, length, and aligned symbols, taking
         * into account extra position bits.  */

        memcpy(updated_main_lens, codes->lens.main, LZX_MAINTREE_NUM_SYMBOLS);
        lzx_update_mainsym_match_costs(LZX_BLOCKTYPE_VERBATIM, updated_main_lens);
        verbatim_cost += lzx_huffman_code_output_cost(updated_main_lens,
                                                      freqs->main,
                                                      LZX_MAINTREE_NUM_SYMBOLS);
        memcpy(updated_main_lens, codes->lens.main, LZX_MAINTREE_NUM_SYMBOLS);
        lzx_update_mainsym_match_costs(LZX_BLOCKTYPE_ALIGNED, updated_main_lens);
        aligned_cost += lzx_huffman_code_output_cost(updated_main_lens,
                                                     freqs->main,
                                                     LZX_MAINTREE_NUM_SYMBOLS);

        common_cost += lzx_huffman_code_output_cost(codes->lens.len,
                                                    freqs->len,
                                                    LZX_LENTREE_NUM_SYMBOLS);

        aligned_cost += lzx_huffman_code_output_cost(codes->lens.aligned,
                                                     freqs->aligned,
                                                     LZX_ALIGNEDTREE_NUM_SYMBOLS);

        *aligned_cost_ret = aligned_cost + common_cost;
        *verbatim_cost_ret = verbatim_cost + common_cost;
}

/* Prepare a (nonsplit) compressed block.  */
static unsigned
lzx_prepare_compressed_block(struct lzx_compressor *ctx, unsigned block_number,
                             struct lzx_codes *prev_codes)
{
        struct lzx_block_spec *spec = &ctx->block_specs[block_number - 1];
        unsigned orig_cached_matches_pos = ctx->cached_matches_pos;
        struct lzx_lru_queue orig_queue = ctx->queue;
        struct lzx_freqs freqs;
        unsigned cost;

        /* Here's where the real work happens.  The following loop runs one or
         * more times, each time using a cost model based on the Huffman codes
         * computed from the previous iteration (the first iteration uses a
         * default model).  Each iteration of the loop uses a heuristic
         * algorithm to divide the block into near-optimal matches/literals from
         * beginning to end.  */
        LZX_ASSERT(ctx->params.slow.num_optim_passes >= 1);
        spec->num_chosen_matches = 0;
        for (unsigned pass = 0; pass < ctx->params.slow.num_optim_passes; pass++)
        {
                LZX_DEBUG("Block %u: Match-choosing pass %u of %u",
                          block_number, pass + 1,
                          ctx->params.slow.num_optim_passes);

                /* Reset frequency tables.  */
                memset(&freqs, 0, sizeof(freqs));

                /* Reset match offset LRU queue.  */
                ctx->queue = orig_queue;

                /* Reset match-finding position.  */
                ctx->cached_matches_pos = orig_cached_matches_pos;
                ctx->match_window_pos = spec->window_pos;
                ctx->match_window_end = spec->window_pos + spec->block_size;

                /* Set cost model.  */
                lzx_set_costs(ctx, &spec->codes.lens);

                unsigned window_pos = spec->window_pos;
                unsigned end = window_pos + spec->block_size;

                while (window_pos < end) {
                        struct raw_match match;
                        struct lzx_match lzx_match;

                        match = lzx_lz_get_near_optimal_match(ctx);

                        if (match.len >= LZX_MIN_MATCH) {

                                /* Best to output a match here.  */

                                LZX_ASSERT(match.len <= LZX_MAX_MATCH);
                                LZX_ASSERT(!memcmp(&ctx->window[window_pos],
                                                   &ctx->window[window_pos - match.offset],
                                                   match.len));

                                /* Tally symbol frequencies.  */
                                lzx_match.data = lzx_record_match(match.offset,
                                                                  match.len,
                                                                  &freqs,
                                                                  &ctx->queue);

                                window_pos += match.len;
                        } else {
                                /* Best to output a literal here.  */

                                /* Tally symbol frequencies.  */
                                lzx_match.data = lzx_record_literal(ctx->window[window_pos],
                                                                    &freqs);

                                window_pos += 1;
                        }

                        /* If it's the last pass, save the match/literal in
                         * intermediate form.  */
                        if (pass == ctx->params.slow.num_optim_passes - 1) {
                                ctx->chosen_matches[spec->chosen_matches_start_pos +
                                                    spec->num_chosen_matches] = lzx_match;

                                spec->num_chosen_matches++;
                        }
                }
                LZX_ASSERT(window_pos == end);

                /* Build Huffman codes using the new frequencies.  */
                lzx_make_huffman_codes(&freqs, &spec->codes);

                /* The first time we get here is when the full input has been
                 * processed, so the match-finding is done.  */
                ctx->matches_already_found = true;
        }

        LZX_DEBUG("Block %u: saved %u matches/literals @ %u",
                  block_number, spec->num_chosen_matches,
                  spec->chosen_matches_start_pos);

        unsigned aligned_cost;
        unsigned verbatim_cost;

        lzx_compute_compressed_block_costs(spec->block_size,
                                           &spec->codes,
                                           prev_codes,
                                           &freqs,
                                           &aligned_cost,
                                           &verbatim_cost);

        /* Choose whether to make the block aligned offset or verbatim.  */
        if (aligned_cost < verbatim_cost) {
                spec->block_type = LZX_BLOCKTYPE_ALIGNED;
                cost = aligned_cost;
                LZX_DEBUG("Using aligned block (cost %u vs %u for verbatim)",
                          aligned_cost, verbatim_cost);
        } else {
                spec->block_type = LZX_BLOCKTYPE_VERBATIM;
                cost = verbatim_cost;
                LZX_DEBUG("Using verbatim block (cost %u vs %u for aligned)",
                          verbatim_cost, aligned_cost);
        }

        LZX_DEBUG("Block %u is %u => %u bytes unsplit.",
                  block_number, spec->block_size, cost / 8);

        return cost;
}

/*
 * lzx_prepare_block_recursive() -
 *
 * Given a (possibly nonproper) sub-sequence of the preprocessed input, compute
 * the LZX block(s) that it should be output as.
 *
 * This function initially considers the case where the given sub-sequence of
 * the preprocessed input be output as a single block.  This block is calculated
 * and its cost (number of bits required to output it) is computed.
 *
 * Then, if @max_split_level is greater than zero, a split into two evenly sized
 * subblocks is considered.  The block is recursively split in this way,
 * potentially up to the depth specified by @max_split_level.  The cost of the
 * split block is compared to the cost of the single block, and the lower cost
 * solution is used.
 *
 * For each compressed output block computed, the sequence of matches/literals
 * and the corresponding Huffman codes for the block are produced and saved.
 *
 * The return value is the approximate number of bits the block (or all
 * subblocks, in the case that the split block had lower cast), will take up
 * when written to the compressed output.
 */
static unsigned
lzx_prepare_block_recursive(struct lzx_compressor * ctx,
                            unsigned block_number,
                            unsigned max_split_level,
                            struct lzx_codes **prev_codes_p)
{
        struct lzx_block_spec *spec = &ctx->block_specs[block_number - 1];
        unsigned cost;
        unsigned orig_cached_matches_pos;
        struct lzx_lru_queue orig_queue, nonsplit_queue;
        struct lzx_codes *prev_codes = *prev_codes_p;

        LZX_DEBUG("Preparing block %u...", block_number);

        /* Save positions of chosen and cached matches, and the match offset LRU
         * queue, so that they can be restored if splitting is attempted.  */
        orig_cached_matches_pos = ctx->cached_matches_pos;
        orig_queue = ctx->queue;

        /* Consider outputting the input subsequence as a single block.  */
        spec->is_split = 0;
        cost = lzx_prepare_compressed_block(ctx, block_number, prev_codes);
        nonsplit_queue = ctx->queue;

        *prev_codes_p = &spec->codes;

        /* If the maximum split level is at least one, consider splitting the
         * block in two.  */
        if (max_split_level--) {

                LZX_DEBUG("Calculating split of block %u...", block_number);

                struct lzx_block_spec *spec1, *spec2;
                unsigned split_cost;

                ctx->cached_matches_pos = orig_cached_matches_pos;
                ctx->queue = orig_queue;

                /* Prepare and get the cost of the first sub-block.  */
                spec1 = &ctx->block_specs[block_number * 2 - 1];
                spec1->codes.lens = spec->codes.lens;
                spec1->window_pos = spec->window_pos;
                spec1->block_size = spec->block_size / 2;
                spec1->chosen_matches_start_pos = spec->chosen_matches_start_pos +
                                                  LZX_MAX_WINDOW_SIZE;
                split_cost = lzx_prepare_block_recursive(ctx,
                                                         block_number * 2,
                                                         max_split_level,
                                                         &prev_codes);

                /* Prepare and get the cost of the second sub-block.  */
                spec2 = spec1 + 1;
                spec2->codes.lens = spec->codes.lens;
                spec2->window_pos = spec->window_pos + spec1->block_size;
                spec2->block_size = spec->block_size - spec1->block_size;
                spec2->chosen_matches_start_pos = spec1->chosen_matches_start_pos +
                                                  spec1->block_size;
                split_cost += lzx_prepare_block_recursive(ctx,
                                                          block_number * 2 + 1,
                                                          max_split_level,
                                                          &prev_codes);

                /* Compare the cost of the whole block with that of the split
                 * block.  Choose the lower cost solution.  */
                if (split_cost < cost) {
                        LZX_DEBUG("Splitting block %u is worth it "
                                  "(%u => %u bytes).",
                                  block_number, cost / 8, split_cost / 8);
                        spec->is_split = 1;
                        cost = split_cost;
                        *prev_codes_p = prev_codes;
                } else {
                        LZX_DEBUG("Splitting block %u is NOT worth it "
                                  "(%u => %u bytes).",
                                  block_number, cost / 8, split_cost / 8);
                        ctx->queue = nonsplit_queue;
                }
        }

        return cost;
}

/* Empirical averages  */
static const u8 lzx_default_mainsym_costs[LZX_MAINTREE_NUM_SYMBOLS] = {
        7, 9, 9, 10, 9, 10, 10, 10, 9, 10, 9, 10, 10, 9, 10, 10, 9, 10, 10, 11,
        10, 10, 10, 11, 10, 11, 11, 11, 10, 11, 11, 11, 8, 11, 9, 10, 9, 10, 11,
        11, 9, 9, 11, 10, 10, 9, 9, 9, 8, 8, 8, 8, 8, 9, 9, 9, 8, 8, 9, 9, 9, 9,
        10, 10, 10, 8, 9, 8, 8, 8, 8, 9, 9, 9, 10, 10, 8, 8, 9, 9, 8, 10, 9, 8,
        8, 9, 8, 9, 9, 10, 10, 10, 9, 10, 11, 9, 10, 8, 9, 8, 8, 8, 8, 9, 8, 8,
        9, 9, 8, 8, 8, 8, 8, 10, 8, 8, 7, 8, 9, 9, 9, 9, 10, 11, 10, 10, 11, 11,
        10, 11, 11, 10, 10, 11, 11, 11, 10, 10, 11, 10, 11, 10, 11, 11, 10, 11,
        11, 12, 11, 11, 11, 12, 11, 11, 11, 11, 11, 11, 11, 12, 10, 11, 11, 11,
        11, 11, 11, 12, 11, 11, 11, 11, 11, 12, 11, 11, 10, 11, 11, 11, 11, 11,
        11, 11, 10, 11, 11, 11, 11, 11, 11, 11, 10, 11, 11, 11, 11, 11, 11, 11,
        10, 11, 11, 11, 11, 11, 11, 11, 10, 11, 11, 11, 11, 12, 11, 11, 10, 11,
        11, 11, 11, 12, 11, 11, 10, 11, 11, 11, 10, 12, 11, 11, 10, 10, 11, 10,
        10, 11, 11, 11, 10, 11, 11, 11, 10, 11, 11, 11, 10, 11, 11, 11, 10, 11,
        10, 9, 8, 7, 10, 10, 11, 10, 11, 7, 9, 9, 11, 11, 11, 12, 11, 9, 10, 10,
        12, 12, 13, 13, 12, 11, 10, 12, 12, 14, 14, 14, 13, 12, 9, 12, 13, 14,
        14, 14, 14, 14, 9, 10, 13, 14, 14, 14, 14, 14, 9, 9, 11, 11, 13, 13, 13,
        14, 9, 9, 11, 12, 12, 13, 13, 13, 8, 8, 11, 11, 12, 12, 12, 11, 9, 9,
        10, 11, 12, 12, 12, 11, 8, 9, 10, 10, 11, 12, 11, 10, 9, 9, 10, 11, 11,
        12, 11, 10, 8, 9, 10, 10, 11, 11, 11, 9, 9, 9, 10, 11, 11, 11, 11, 9, 8,
        8, 10, 10, 11, 11, 11, 9, 9, 9, 10, 10, 11, 11, 11, 9, 9, 8, 9, 10, 11,
        11, 11, 9, 10, 9, 10, 11, 11, 11, 11, 9, 14, 9, 9, 10, 10, 11, 10, 9,
        14, 9, 10, 11, 11, 11, 11, 9, 14, 9, 10, 10, 11, 11, 11, 9, 14, 10, 10,
        11, 11, 12, 11, 10, 14, 10, 10, 10, 11, 11, 11, 10, 14, 11, 11, 11, 11,
        12, 12, 10, 14, 10, 11, 11, 11, 12, 11, 10, 14, 11, 11, 11, 12, 12, 12,
        11, 15, 11, 11, 11, 12, 12, 12, 11, 14, 12, 12, 12, 12, 13, 12, 11, 15,
        12, 12, 12, 13, 13, 13, 12, 15, 14, 13, 14, 14, 14, 14, 13,
};

/* Empirical averages  */
static const u8 lzx_default_lensym_costs[LZX_LENTREE_NUM_SYMBOLS] = {
        5, 5, 5, 5, 5, 6, 5, 5, 6, 7, 7, 7, 8, 8, 7, 8, 9, 9, 9, 9, 10, 9, 9,
        10, 9, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 12, 12, 12, 11, 12, 12,
        12, 12, 12, 12, 13, 12, 12, 12, 13, 12, 13, 13, 12, 12, 13, 12, 13, 13,
        13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 14, 13, 14, 13, 14, 13,
        14, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
        14, 14, 14, 14, 14, 14, 14, 14, 14, 10,
};

/*
 * Set default symbol costs.
 */
static void
lzx_set_default_costs(struct lzx_lens * lens)
{
        unsigned i;

#if LZX_PARAM_USE_EMPIRICAL_DEFAULT_COSTS
        memcpy(&lens->main, lzx_default_mainsym_costs, LZX_MAINTREE_NUM_SYMBOLS);
        memcpy(&lens->len, lzx_default_lensym_costs, LZX_LENTREE_NUM_SYMBOLS);

#else
        /* Literal symbols  */
        for (i = 0; i < LZX_NUM_CHARS; i++)
                lens->main[i] = 8;

        /* Match header symbols  */
        for (; i < LZX_MAINTREE_NUM_SYMBOLS; i++)
                lens->main[i] = 10;

        /* Length symbols  */
        for (i = 0; i < LZX_LENTREE_NUM_SYMBOLS; i++)
                lens->len[i] = 8;
#endif

        /* Aligned offset symbols  */
        for (i = 0; i < LZX_ALIGNEDTREE_NUM_SYMBOLS; i++)
                lens->aligned[i] = 3;
}

/*
 * lzx_prepare_blocks() -
 *
 * Calculate the blocks to split the preprocessed data into.
 *
 * Input ---  the preprocessed data:
 *
 *      ctx->window[]
 *      ctx->window_size
 *
 * Working space:
 *      Match finding:
 *              ctx->hash_tab
 *              ctx->child_tab
 *              ctx->cached_matches
 *              ctx->cached_matches_pos
 *              ctx->matches_already_found
 *
 *      Block cost modeling:
 *              ctx->costs
 *              ctx->block_specs (also an output)
 *
 *      Match choosing:
 *              ctx->optimum
 *              ctx->optimum_cur_idx
 *              ctx->optimum_end_idx
 *              ctx->chosen_matches (also an output)
 *
 * Output --- the block specifications and the corresponding match/literal data:
 *
 *      ctx->block_specs[]
 *      ctx->chosen_matches[]
 *
 * The return value is the approximate number of bits the compressed data will
 * take up.
 */
static unsigned
lzx_prepare_blocks(struct lzx_compressor * ctx)
{
        /* This function merely does some initializations, then passes control
         * to lzx_prepare_block_recursive().  */

        /* 1. Initialize match-finding variables.  */

        /* Zero all entries in the hash table, indicating that no length-3
         * character sequences have been discovered in the input yet.  */
        memset(ctx->hash_tab, 0, LZX_LZ_HASH_SIZE * 2 * sizeof(ctx->hash_tab[0]));
        if (ctx->params.slow.use_len2_matches)
                memset(ctx->digram_tab, 0, 256 * 256 * sizeof(ctx->digram_tab[0]));
        /* Note: ctx->child_tab need not be initialized.  */

        /* No matches have been found and cached yet.  */
        ctx->cached_matches_pos = 0;
        ctx->matches_already_found = false;

        /* 2. Initialize match-choosing variables.  */
        ctx->optimum_cur_idx = 0;
        ctx->optimum_end_idx = 0;
        /* Note: ctx->optimum need not be initialized.  */
        ctx->block_specs[0].chosen_matches_start_pos = 0;

        /* 3. Set block 1 (index 0) to represent the entire input data.  */
        ctx->block_specs[0].block_size = ctx->window_size;
        ctx->block_specs[0].window_pos = 0;

        /* 4. Set up a default Huffman symbol cost model for block 1 (index 0).
         * The model will be refined later.  */
        lzx_set_default_costs(&ctx->block_specs[0].codes.lens);

        /* 5. Initialize the match offset LRU queue.  */
        ctx->queue = (struct lzx_lru_queue){1, 1, 1};

        /* 6. Pass control to recursive procedure.  */
        struct lzx_codes * prev_codes = &ctx->zero_codes;
        return lzx_prepare_block_recursive(ctx, 1,
                                           ctx->params.slow.num_split_passes,
                                           &prev_codes);
}

/*
 * This is the fast version of lzx_prepare_blocks(), which "quickly" prepares a
 * single compressed block containing the entire input.  See the description of
 * the "Fast algorithm" at the beginning of this file for more information.
 *
 * Input ---  the preprocessed data:
 *
 *      ctx->window[]
 *      ctx->window_size
 *
 * Working space:
 *      ctx->queue
 *
 * Output --- the block specifications and the corresponding match/literal data:
 *
 *      ctx->block_specs[]
 *      ctx->chosen_matches[]
 */
static void
lzx_prepare_block_fast(struct lzx_compressor * ctx)
{
        unsigned num_matches;
        struct lzx_freqs freqs;
        struct lzx_block_spec *spec;

        /* Parameters to hash chain LZ match finder  */
        static const struct lz_params lzx_lz_params = {
                /* LZX_MIN_MATCH == 2, but 2-character matches are rarely
                 * useful; the minimum match for compression is set to 3
                 * instead. */
                .min_match      = 3,
                .max_match      = LZX_MAX_MATCH,
                .good_match     = LZX_MAX_MATCH,
                .nice_match     = LZX_MAX_MATCH,
                .max_chain_len  = LZX_MAX_MATCH,
                .max_lazy_match = LZX_MAX_MATCH,
                .too_far        = 4096,
        };

        /* Initialize symbol frequencies and match offset LRU queue.  */
        memset(&freqs, 0, sizeof(struct lzx_freqs));
        ctx->queue = (struct lzx_lru_queue){ 1, 1, 1 };

        /* Determine series of matches/literals to output.  */
        num_matches = lz_analyze_block(ctx->window,
                                       ctx->window_size,
                                       (u32*)ctx->chosen_matches,
                                       lzx_record_match,
                                       lzx_record_literal,
                                       &freqs,
                                       &ctx->queue,
                                       &freqs,
                                       &lzx_lz_params);


        /* Set up block specification.  */
        spec = &ctx->block_specs[0];
        spec->is_split = 0;
        spec->block_type = LZX_BLOCKTYPE_ALIGNED;
        spec->window_pos = 0;
        spec->block_size = ctx->window_size;
        spec->num_chosen_matches = num_matches;
        spec->chosen_matches_start_pos = 0;
        lzx_make_huffman_codes(&freqs, &spec->codes);
}

static void
do_call_insn_translation(u32 *call_insn_target, int input_pos,
                         s32 file_size)
{
        s32 abs_offset;
        s32 rel_offset;

        rel_offset = le32_to_cpu(*call_insn_target);
        if (rel_offset >= -input_pos && rel_offset < file_size) {
                if (rel_offset < file_size - input_pos) {
                        /* "good translation" */
                        abs_offset = rel_offset + input_pos;
                } else {
                        /* "compensating translation" */
                        abs_offset = rel_offset - file_size;
                }
                *call_insn_target = cpu_to_le32(abs_offset);
        }
}

/* This is the reverse of undo_call_insn_preprocessing() in lzx-decompress.c.
 * See the comment above that function for more information.  */
static void
do_call_insn_preprocessing(u8 data[], int size)
{
        for (int i = 0; i < size - 10; i++) {
                if (data[i] == 0xe8) {
                        do_call_insn_translation((u32*)&data[i + 1], i,
                                                 LZX_WIM_MAGIC_FILESIZE);
                        i += 4;
                }
        }
}

/* API function documented in wimlib.h  */
WIMLIBAPI unsigned
wimlib_lzx_compress2(const void                 * const restrict uncompressed_data,
                     unsigned                     const          uncompressed_len,
                     void                       * const restrict compressed_data,
                     struct wimlib_lzx_context  * const restrict lzx_ctx)
{
        struct lzx_compressor *ctx = (struct lzx_compressor*)lzx_ctx;
        struct output_bitstream ostream;
        unsigned compressed_len;

        if (uncompressed_len < 100) {
                LZX_DEBUG("Too small to bother compressing.");
                return 0;
        }

        if (uncompressed_len > 32768) {
                LZX_DEBUG("Only up to 32768 bytes of uncompressed data are supported.");
                return 0;
        }

        wimlib_assert(lzx_ctx != NULL);

        LZX_DEBUG("Attempting to compress %u bytes...", uncompressed_len);

        /* The input data must be preprocessed.  To avoid changing the original
         * input, copy it to a temporary buffer.  */
        memcpy(ctx->window, uncompressed_data, uncompressed_len);
        ctx->window_size = uncompressed_len;

        /* This line is unnecessary; it just avoids inconsequential accesses of
         * uninitialized memory that would show up in memory-checking tools such
         * as valgrind.  */
        memset(&ctx->window[ctx->window_size], 0, 12);

        LZX_DEBUG("Preprocessing data...");

        /* Before doing any actual compression, do the call instruction (0xe8
         * byte) translation on the uncompressed data.  */
        do_call_insn_preprocessing(ctx->window, ctx->window_size);

        LZX_DEBUG("Preparing blocks...");

        /* Prepare the compressed data.  */
        if (ctx->params.algorithm == WIMLIB_LZX_ALGORITHM_FAST)
                lzx_prepare_block_fast(ctx);
        else
                lzx_prepare_blocks(ctx);

        LZX_DEBUG("Writing compressed blocks...");

        /* Generate the compressed data.  */
        init_output_bitstream(&ostream, compressed_data, ctx->window_size - 1);
        lzx_write_all_blocks(ctx, &ostream);

        LZX_DEBUG("Flushing bitstream...");
        if (flush_output_bitstream(&ostream)) {
                /* If the bitstream cannot be flushed, then the output space was
                 * exhausted.  */
                LZX_DEBUG("Data did not compress to less than original length!");
                return 0;
        }

        /* Compute the length of the compressed data.  */
        compressed_len = ostream.bit_output - (u8*)compressed_data;

        LZX_DEBUG("Done: compressed %u => %u bytes.",
                  uncompressed_len, compressed_len);

#if defined(ENABLE_LZX_DEBUG) || defined(ENABLE_VERIFY_COMPRESSION)
        /* Verify that we really get the same thing back when decompressing.  */
        {
                u8 buf[uncompressed_len];
                int ret;
                unsigned i;

                ret = wimlib_lzx_decompress(compressed_data, compressed_len,
                                            buf, uncompressed_len);
                if (ret) {
                        ERROR("Failed to decompress data we "
                              "compressed using LZX algorithm");
                        wimlib_assert(0);
                        return 0;
                }

                bool bad = false;
                const u8 * udata = uncompressed_data;
                for (i = 0; i < uncompressed_len; i++) {
                        if (buf[i] != udata[i]) {
                                bad = true;
                                ERROR("Data we compressed using LZX algorithm "
                                      "didn't decompress to original "
                                      "(difference at idx %u: c %#02x, u %#02x)",
                                      i, buf[i], udata[i]);
                        }
                }
                if (bad) {
                        wimlib_assert(0);
                        return 0;
                }
        }
#endif
        return compressed_len;
}

static bool
lzx_params_compatible(const struct wimlib_lzx_params *oldparams,
                      const struct wimlib_lzx_params *newparams)
{
        return 0 == memcmp(oldparams, newparams, sizeof(struct wimlib_lzx_params));
}

/* API function documented in wimlib.h  */
WIMLIBAPI int
wimlib_lzx_alloc_context(const struct wimlib_lzx_params *params,
                         struct wimlib_lzx_context **ctx_pp)
{

        LZX_DEBUG("Allocating LZX context...");

        struct lzx_compressor *ctx;

        static const struct wimlib_lzx_params fast_default = {
                .size_of_this = sizeof(struct wimlib_lzx_params),
                .algorithm = WIMLIB_LZX_ALGORITHM_FAST,
                .use_defaults = 0,
                .fast = {
                },
        };
        static const struct wimlib_lzx_params slow_default = {
                .size_of_this = sizeof(struct wimlib_lzx_params),
                .algorithm = WIMLIB_LZX_ALGORITHM_SLOW,
                .use_defaults = 0,
                .slow = {
                        .use_len2_matches = 1,
                        .num_fast_bytes = 32,
                        .num_optim_passes = 3,
                        .num_split_passes = 3,
                        .main_nostat_cost = 15,
                        .len_nostat_cost = 15,
                        .aligned_nostat_cost = 7,
                },
        };

        if (params == NULL) {
                LZX_DEBUG("Using default algorithm and parameters.");
                params = &slow_default;
        }

        if (params->algorithm != WIMLIB_LZX_ALGORITHM_SLOW &&
            params->algorithm != WIMLIB_LZX_ALGORITHM_FAST)
        {
                LZX_DEBUG("Invalid algorithm.");
                return WIMLIB_ERR_INVALID_PARAM;
        }

        if (params->use_defaults) {
                if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW)
                        params = &slow_default;
                else
                        params = &fast_default;
        }

        if (params->size_of_this != sizeof(struct wimlib_lzx_params)) {
                LZX_DEBUG("Invalid parameter structure size!");
                return WIMLIB_ERR_INVALID_PARAM;
        }

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) {
                if (params->slow.num_fast_bytes < 3 ||
                    params->slow.num_fast_bytes > 257)
                {
                        LZX_DEBUG("Invalid number of fast bytes!");
                        return WIMLIB_ERR_INVALID_PARAM;
                }

                if (params->slow.num_optim_passes < 1)
                {
                        LZX_DEBUG("Invalid number of optimization passes!");
                        return WIMLIB_ERR_INVALID_PARAM;
                }

                if (params->slow.main_nostat_cost < 1 ||
                    params->slow.main_nostat_cost > 16)
                {
                        LZX_DEBUG("Invalid main_nostat_cost!");
                        return WIMLIB_ERR_INVALID_PARAM;
                }

                if (params->slow.len_nostat_cost < 1 ||
                    params->slow.len_nostat_cost > 16)
                {
                        LZX_DEBUG("Invalid len_nostat_cost!");
                        return WIMLIB_ERR_INVALID_PARAM;
                }

                if (params->slow.aligned_nostat_cost < 1 ||
                    params->slow.aligned_nostat_cost > 8)
                {
                        LZX_DEBUG("Invalid aligned_nostat_cost!");
                        return WIMLIB_ERR_INVALID_PARAM;
                }
        }

        if (ctx_pp == NULL) {
                LZX_DEBUG("Check parameters only.");
                return 0;
        }

        ctx = *(struct lzx_compressor**)ctx_pp;

        if (ctx && lzx_params_compatible(&ctx->params, params))
                return 0;

        LZX_DEBUG("Allocating memory.");

        ctx = MALLOC(sizeof(struct lzx_compressor));
        if (ctx == NULL)
                goto err;

        size_t block_specs_length;

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW)
                block_specs_length = ((1 << (params->slow.num_split_passes + 1)) - 1);
        else
                block_specs_length = 1;
        ctx->block_specs = MALLOC(block_specs_length * sizeof(ctx->block_specs[0]));
        if (ctx->block_specs == NULL)
                goto err_free_ctx;

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) {
                ctx->hash_tab = MALLOC((LZX_LZ_HASH_SIZE + 2 * LZX_MAX_WINDOW_SIZE) *
                                        sizeof(ctx->hash_tab[0]));
                if (ctx->hash_tab == NULL)
                        goto err_free_block_specs;
                ctx->child_tab = ctx->hash_tab + LZX_LZ_HASH_SIZE;
        } else {
                ctx->hash_tab = NULL;
                ctx->child_tab = NULL;
        }

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW &&
            params->slow.use_len2_matches)
        {
                ctx->digram_tab = MALLOC(256 * 256 * sizeof(ctx->digram_tab[0]));
                if (ctx->digram_tab == NULL)
                        goto err_free_hash_tab;
        } else {
                ctx->digram_tab = NULL;
        }

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) {
                ctx->cached_matches = MALLOC(10 * LZX_MAX_WINDOW_SIZE *
                                             sizeof(ctx->cached_matches[0]));
                if (ctx->cached_matches == NULL)
                        goto err_free_digram_tab;
        } else {
                ctx->cached_matches = NULL;
        }

        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) {
                ctx->optimum = MALLOC((LZX_PARAM_OPTIM_ARRAY_SIZE + LZX_MAX_MATCH) *
                                       sizeof(ctx->optimum[0]));
                if (ctx->optimum == NULL)
                        goto err_free_cached_matches;
        } else {
                ctx->optimum = NULL;
        }

        size_t chosen_matches_length;
        if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW)
                chosen_matches_length = LZX_MAX_WINDOW_SIZE *
                                        (params->slow.num_split_passes + 1);
        else
                chosen_matches_length = LZX_MAX_WINDOW_SIZE;

        ctx->chosen_matches = MALLOC(chosen_matches_length *
                                     sizeof(ctx->chosen_matches[0]));
        if (ctx->chosen_matches == NULL)
                goto err_free_optimum;

        memcpy(&ctx->params, params, sizeof(struct wimlib_lzx_params));
        memset(&ctx->zero_codes, 0, sizeof(ctx->zero_codes));

        LZX_DEBUG("Successfully allocated new LZX context.");

        wimlib_lzx_free_context(*ctx_pp);
        *ctx_pp = (struct wimlib_lzx_context*)ctx;
        return 0;

err_free_optimum:
        FREE(ctx->optimum);
err_free_cached_matches:
        FREE(ctx->cached_matches);
err_free_digram_tab:
        FREE(ctx->digram_tab);
err_free_hash_tab:
        FREE(ctx->hash_tab);
err_free_block_specs:
        FREE(ctx->block_specs);
err_free_ctx:
        FREE(ctx);
err:
        LZX_DEBUG("Ran out of memory.");
        return WIMLIB_ERR_NOMEM;
}

/* API function documented in wimlib.h  */
WIMLIBAPI void
wimlib_lzx_free_context(struct wimlib_lzx_context *_ctx)
{
        struct lzx_compressor *ctx = (struct lzx_compressor*)_ctx;

        if (ctx) {
                FREE(ctx->chosen_matches);
                FREE(ctx->optimum);
                FREE(ctx->cached_matches);
                FREE(ctx->digram_tab);
                FREE(ctx->hash_tab);
                FREE(ctx->block_specs);
                FREE(ctx);
        }
}

/* API function documented in wimlib.h  */
WIMLIBAPI unsigned
wimlib_lzx_compress(const void * const restrict uncompressed_data,
                    unsigned     const          uncompressed_len,
                    void       * const restrict compressed_data)
{
        int ret;
        struct wimlib_lzx_context *ctx;
        unsigned compressed_len;

        ret = wimlib_lzx_alloc_context(NULL, &ctx);
        if (ret) {
                wimlib_assert(ret != WIMLIB_ERR_INVALID_PARAM);
                WARNING("Couldn't allocate LZX compression context: %"TS"",
                        wimlib_get_error_string(ret));
                return 0;
        }

        compressed_len = wimlib_lzx_compress2(uncompressed_data,
                                              uncompressed_len,
                                              compressed_data,
                                              ctx);

        wimlib_lzx_free_context(ctx);

        return compressed_len;
}