X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Flzx-compress.c;h=bc50a858bc9b62cf26d9d1b03d7b822591a1eb33;hp=056dc83bb52944c63ad9b98cdad8f966b2340e76;hb=bcf3ccecc8071729f34fe5b393fab64fc02e3d47;hpb=dbfee435692344cccd48bb4c7deb3af23ac80176 diff --git a/src/lzx-compress.c b/src/lzx-compress.c index 056dc83b..bc50a858 100644 --- a/src/lzx-compress.c +++ b/src/lzx-compress.c @@ -1,11 +1,11 @@ /* * lzx-compress.c * - * LZX compression routines + * A compressor that produces output compatible with the LZX compression format. */ /* - * Copyright (C) 2012, 2013 Eric Biggers + * Copyright (C) 2012, 2013, 2014 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -25,165 +25,200 @@ /* - * This file contains a compressor for the LZX compression format, as used in - * the WIM file format. + * This file contains a compressor for the LZX ("Lempel-Ziv eXtended"?) + * compression format, as used in the WIM (Windows IMaging) file format. This + * code may need some slight modifications to be used outside of the WIM format. + * In particular, in other situations the LZX block header might be slightly + * different, and a sliding window rather than a fixed-size window might be + * required. * - * Format - * ====== + * ---------------------------------------------------------------------------- * - * First, the primary reference for the LZX compression format is the - * specification released by Microsoft. + * Format Overview * - * Second, the comments in lzx-decompress.c provide some more information about - * the LZX compression format, including errors in the Microsoft specification. + * The primary reference for LZX is the specification released by Microsoft. + * However, the comments in lzx-decompress.c provide more information about LZX + * and note some 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 shares many similarities with DEFLATE, the format used by zlib and gzip. + * Both LZX and DEFLATE use LZ77 matching and Huffman coding. Certain details + * are quite similar, such as the method for storing Huffman codes. However, + * the main differences are: * * - LZX preprocesses the data to attempt to make x86 machine code slightly more * compressible before attempting to compress it further. + * * - 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 slot" (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 does not have static Huffman blocks (that is, the kind with preset + * Huffman codes); however it does have two types of dynamic Huffman blocks + * ("verbatim" and "aligned"). + * * - LZX has a minimum match length of 2 rather than 3. + * * - In LZX, match offsets 0 through 2 actually represent entries in an LRU * queue of match offsets. This is very useful for certain types of files, * such as binary files that have repeating records. * - * 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. + * Algorithmic Overview * - * Slow algorithm - * -------------- + * At a high level, any implementation of LZX compression must operate as + * follows: * - * The "slow" algorithm to generate LZX-compressed data is roughly as follows: + * 1. Preprocess the input data to translate the targets of 32-bit x86 call + * instructions to absolute offsets. (Actually, this is required for WIM, + * but might not be in other places LZX is used.) * - * 1. Preprocess the input data to translate the targets of x86 call - * instructions to absolute offsets. + * 2. Find a sequence of LZ77-style matches and literal bytes that expands to + * the preprocessed data. * - * 2. Build the suffix array and inverse suffix array for the input data. The - * suffix array contains the indices of all suffixes of the input data, - * sorted lexcographically by the corresponding suffixes. The "position" of - * a suffix is the index of that suffix in the original string, whereas the - * "rank" of a suffix is the index at which that suffix's position is found - * in the suffix array. + * 3. Divide the match/literal sequence into one or more LZX blocks, each of + * which may be "uncompressed", "verbatim", or "aligned". * - * 3. Build the longest common prefix array corresponding to the suffix array. + * 4. Output each LZX block. * - * 4. For each suffix, find the highest lower ranked suffix that has a lower - * position, the lowest higher ranked suffix that has a lower position, and - * the length of the common prefix shared between each. This information is - * later used to link suffix ranks into a doubly-linked list for searching - * the suffix array. + * Step (1) is fairly straightforward. It requires looking for 0xe8 bytes in + * the input data and performing a translation on the 4 bytes following each + * one. * - * 5. Set a default cost model for matches/literals. + * Step (4) is complicated, but it is mostly determined by the LZX format. The + * only real choice we have is what algorithm to use to build the length-limited + * canonical Huffman codes. See lzx_write_all_blocks() for details. * - * 6. Determine the lowest cost sequence of LZ77 matches ((offset, length) - * pairs) and literal bytes to divide the input into. Raw match-finding is - * done by searching the suffix array using a linked list to avoid - * considering any suffixes that start after the current position. Each run - * of the match-finder returns the approximate lowest-cost longest match as - * well as any shorter matches that have even lower approximate costs. Each - * such run also adds the suffix rank of the current position into the linked - * list being used to search the suffix array. Parsing, or match-choosing, - * is solved as a minimum-cost path problem using a forward "optimal parsing" - * algorithm based on the Deflate encoder from 7-Zip. This algorithm moves - * forward calculating the minimum cost to reach each byte until either a - * very long match is found or until a position is found at which no matches - * start or overlap. + * That leaves steps (2) and (3) as where all the hard stuff happens. Focusing + * on step (2), we need to do LZ77-style parsing on the input data, or "window", + * to divide it into a sequence of matches and literals. Each position in the + * window might have multiple matches associated with it, and we need to choose + * which one, if any, to actually use. Therefore, the problem can really be + * divided into two areas of concern: (a) finding matches at a given position, + * which we shall call "match-finding", and (b) choosing whether to use a + * match or a literal at a given position, and if using a match, which one (if + * there is more than one available). We shall call this "match-choosing". We + * first consider match-finding, then match-choosing. * - * 7. Build the Huffman codes needed to output the matches/literals. + * ---------------------------------------------------------------------------- * - * 8. Up to a certain number of iterations, use the resulting Huffman codes to - * refine a cost model and go back to Step #6 to determine an improved - * sequence of matches and literals. + * Match-finding * - * 9. Output the resulting block using the match/literal sequences and the - * Huffman codes that were computed for the block. + * Given a position in the window, we want to find LZ77-style "matches" with + * that position at previous positions in the window. With LZX, the minimum + * match length is 2 and the maximum match length is 257. The only restriction + * on offsets is that LZX does not allow the last 2 bytes of the window to match + * the beginning of the window. * - * Note: the algorithm does not yet attempt to split the input into multiple LZX - * blocks, instead using a series of blocks of LZX_DIV_BLOCK_SIZE bytes. + * There are a number of algorithms that can be used for this, including hash + * chains, binary trees, and suffix arrays. Binary trees generally work well + * for LZX compression since it uses medium-size windows (2^15 to 2^21 bytes). + * However, when compressing in a fast mode where many positions are skipped + * (not searched for matches), hash chains are faster. * - * Fast algorithm - * -------------- + * Since the match-finders are not specific to LZX, I will not explain them in + * detail here. Instead, see lz_hash_chains.c and lz_binary_trees.c. * - * 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 and match choosing. 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. + * ---------------------------------------------------------------------------- * - * Acknowledgments - * =============== + * Match-choosing * - * Acknowledgments to several open-source projects and research papers that made - * it possible to implement this code: + * Usually, choosing the longest match is best because it encodes the most data + * in that one item. However, sometimes the longest match is not optimal + * because (a) choosing a long match now might prevent using an even longer + * match later, or (b) more generally, what we actually care about is the number + * of bits it will ultimately take to output each match or literal, which is + * actually dependent on the entropy encoding using by the underlying + * compression format. Consequently, a longer match usually, but not always, + * takes fewer bits to encode than multiple shorter matches or literals that + * cover the same data. * - * - divsufsort (author: Yuta Mori), for the suffix array construction code, - * located in a separate directory (divsufsort/). + * This problem of choosing the truly best match/literal sequence is probably + * impossible to solve efficiently when combined with entropy encoding. If we + * knew how many bits it takes to output each match/literal, then we could + * choose the optimal sequence using shortest-path search a la Dijkstra's + * algorithm. However, with entropy encoding, the chosen match/literal sequence + * affects its own encoding. Therefore, we can't know how many bits it will + * take to actually output any one match or literal until we have actually + * chosen the full sequence of matches and literals. * - * - "Linear-Time Longest-Common-Prefix Computation in Suffix Arrays and Its - * Applications" (Kasai et al. 2001), for the LCP array computation. + * Notwithstanding the entropy encoding problem, we also aren't guaranteed to + * choose the optimal match/literal sequence unless the match-finder (see + * section "Match-finder") provides the match-chooser with all possible matches + * at each position. However, this is not computationally efficient. For + * example, there might be many matches of the same length, and usually (but not + * always) the best choice is the one with the smallest offset. So in practice, + * it's fine to only consider the smallest offset for a given match length at a + * given position. (Actually, for LZX, it's also worth considering repeat + * offsets.) * - * - "LPF computation revisited" (Crochemore et al. 2009) for the prev and next - * array computations. + * In addition, as mentioned earlier, in LZX we have the choice of using + * multiple blocks, each of which resets the Huffman codes. This expands the + * search space even further. Therefore, to simplify the problem, we currently + * we don't attempt to actually choose the LZX blocks based on the data. + * Instead, we just divide the data into fixed-size blocks of LZX_DIV_BLOCK_SIZE + * bytes each, and always use verbatim or aligned blocks (never uncompressed). + * A previous version of this code recursively split the input data into + * equal-sized blocks, up to a maximum depth, and chose the lowest-cost block + * divisions. However, this made compression much slower and did not actually + * help very much. It remains an open question whether a sufficiently fast and + * useful block-splitting algorithm is possible for LZX. Essentially the same + * problem also applies to DEFLATE. The Microsoft LZX compressor seemingly does + * do block splitting, although I don't know how fast or useful it is, + * specifically. * - * - 7-Zip (author: Igor Pavlov) for the algorithm for forward optimal parsing - * (match-choosing). + * Now, back to the entropy encoding problem. The "solution" is to use an + * iterative approach to compute a good, but not necessarily optimal, + * match/literal sequence. Start with a fixed assignment of symbol costs and + * choose an "optimal" match/literal sequence based on those costs, using + * shortest-path seach a la Dijkstra's algorithm. Then, for each iteration of + * the optimization, update the costs based on the entropy encoding of the + * current match/literal sequence, then choose a new match/literal sequence + * based on the updated costs. Usually, the actual cost to output the current + * match/literal sequence will decrease in each iteration until it converges on + * a fixed point. This result may not be the truly optimal match/literal + * sequence, but it usually is much better than one chosen by doing a "greedy" + * parse where we always chooe the longest match. * - * - zlib (author: Jean-loup Gailly and Mark Adler), for the hash table - * match-finding algorithm (used in lz77.c). - * - * - lzx-compress (author: Matthew T. Russotto), on which some parts of this - * code were originally based. + * An alternative to both greedy parsing and iterative, near-optimal parsing is + * "lazy" parsing. Briefly, "lazy" parsing considers just the longest match at + * each position, but it waits to choose that match until it has also examined + * the next position. This is actually a useful approach; it's used by zlib, + * for example. Therefore, for fast compression we combine lazy parsing with + * the hash chain max-finder. For normal/high compression we combine + * near-optimal parsing with the binary tree match-finder. */ #ifdef HAVE_CONFIG_H # include "config.h" #endif -#include "wimlib.h" #include "wimlib/compressor_ops.h" #include "wimlib/compress_common.h" #include "wimlib/endianness.h" #include "wimlib/error.h" -#include "wimlib/lz_hash.h" -#include "wimlib/lz_sarray.h" +#include "wimlib/lz_mf.h" +#include "wimlib/lz_repsearch.h" #include "wimlib/lzx.h" #include "wimlib/util.h" -#include -#include #include -#ifdef ENABLE_LZX_DEBUG -# include "wimlib/decompress_common.h" -#endif +#define LZX_OPTIM_ARRAY_LENGTH 4096 -typedef u32 block_cost_t; -#define INFINITE_BLOCK_COST ((block_cost_t)~0U) +#define LZX_DIV_BLOCK_SIZE 32768 -#define LZX_OPTIM_ARRAY_SIZE 4096 +#define LZX_CACHE_PER_POS 8 -#define LZX_DIV_BLOCK_SIZE 32768 +#define LZX_MAX_MATCHES_PER_POS (LZX_MAX_MATCH_LEN - LZX_MIN_MATCH_LEN + 1) -#define LZX_MAX_CACHE_PER_POS 10 +#define LZX_CACHE_LEN (LZX_DIV_BLOCK_SIZE * (LZX_CACHE_PER_POS + 1)) /* Codewords for the LZX main, length, and aligned offset Huffman codes */ struct lzx_codewords { - u16 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; - u16 len[LZX_LENCODE_NUM_SYMBOLS]; - u16 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; + u32 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; + u32 len[LZX_LENCODE_NUM_SYMBOLS]; + u32 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; /* Codeword lengths (in bits) for the LZX main, length, and aligned offset @@ -201,7 +236,7 @@ struct lzx_lens { * * If a codeword has zero frequency, it must still be assigned some nonzero cost * --- generally a high cost, since even if it gets used in the next iteration, - * it probably will not be used very times. */ + * it probably will not be used very many times. */ struct lzx_costs { u8 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; u8 len[LZX_LENCODE_NUM_SYMBOLS]; @@ -216,13 +251,13 @@ struct lzx_codes { /* Tables for tallying symbol frequencies in the three LZX alphabets */ struct lzx_freqs { - input_idx_t main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; - input_idx_t len[LZX_LENCODE_NUM_SYMBOLS]; - input_idx_t aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; + u32 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; + u32 len[LZX_LENCODE_NUM_SYMBOLS]; + u32 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; /* LZX intermediate match/literal format */ -struct lzx_match { +struct lzx_item { /* Bit Description * * 31 1 if a match, 0 if a literal. @@ -247,74 +282,35 @@ struct lzx_block_spec { int block_type; /* 0-based position in the window at which this block starts. */ - input_idx_t window_pos; + u32 window_pos; /* The number of bytes of uncompressed data this block represents. */ - input_idx_t block_size; + u32 block_size; - /* The position in the 'chosen_matches' array in the `struct - * lzx_compressor' at which the match/literal specifications for - * this block begin. */ - input_idx_t chosen_matches_start_pos; + /* The match/literal sequence for this block. */ + struct lzx_item *chosen_items; - /* The number of match/literal specifications for this block. */ - input_idx_t num_chosen_matches; + /* The length of the @chosen_items sequence. */ + u32 num_chosen_items; /* 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. */ - block_cost_t 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. */ - input_idx_t link; - - /* Offset (as in an LZ (length, offset) pair) of the - * match or literal that was taken to get to this - * position in the approximate minimum-cost parse. */ - input_idx_t match_offset; - } prev; - struct { - /* Position at which the match or literal starting at - * this position ends in the minimum-cost parse. */ - input_idx_t link; - - /* Offset (as in an LZ (length, offset) pair) of the - * match or literal starting at this position in the - * approximate minimum-cost parse. */ - input_idx_t match_offset; - } next; - }; +struct lzx_compressor; - /* The match offset LRU queue that will exist when the approximate - * minimum-cost path to reach this position is taken. */ - struct lzx_lru_queue queue; +struct lzx_compressor_params { + struct lz_match (*choose_item_func)(struct lzx_compressor *); + enum lz_mf_algo mf_algo; + u32 num_optim_passes; + u32 min_match_length; + u32 nice_match_length; + u32 max_search_depth; }; /* State of the LZX compressor. */ struct lzx_compressor { - /* The parameters that were used to create the compressor. */ - struct wimlib_lzx_compressor_params params; - /* The buffer of data to be compressed. * * 0xe8 byte preprocessing is done directly on the data here before @@ -323,22 +319,34 @@ struct lzx_compressor { * Note that this compressor does *not* use a real 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; + * chunks. */ + u8 *cur_window; /* Number of bytes of data to be compressed, which is the number of - * bytes of data in @window that are actually valid. */ - input_idx_t window_size; + * bytes of data in @cur_window that are actually valid. */ + u32 cur_window_size; + + /* Allocated size of @cur_window. */ + u32 max_window_size; + + /* log2 order of the LZX window size for LZ match offset encoding + * purposes. Will be >= LZX_MIN_WINDOW_ORDER and <= + * LZX_MAX_WINDOW_ORDER. + * + * Note: 1 << @window_order is normally equal to @max_window_size, but + * it will be greater than @max_window_size in the event that the + * compressor was created with a non-power-of-2 block size. (See + * lzx_get_window_order().) */ + unsigned window_order; - /* Allocated size of the @window. */ - input_idx_t max_window_size; + /* Compression parameters. */ + struct lzx_compressor_params params; - /* Number of symbols in the main alphabet (depends on the - * @max_window_size since it determines the maximum allowed offset). */ + unsigned (*get_matches_func)(struct lzx_compressor *, const struct lz_match **); + void (*skip_bytes_func)(struct lzx_compressor *, unsigned n); + + /* Number of symbols in the main alphabet (depends on the @window_order + * since it determines the maximum allowed offset). */ unsigned num_main_syms; /* The current match offset LRU queue. */ @@ -346,7 +354,7 @@ struct lzx_compressor { /* Space for the sequences of matches/literals that were chosen for each * block. */ - struct lzx_match *chosen_matches; + struct lzx_item *chosen_items; /* Information about the LZX blocks the preprocessed input was divided * into. */ @@ -365,57 +373,233 @@ struct lzx_compressor { /* The current cost model. */ struct lzx_costs costs; - /* Fast algorithm only: Array of hash table links. */ - input_idx_t *prev_tab; - - /* Slow algorithm only: Suffix array match-finder. */ - struct lz_sarray lz_sarray; + /* Lempel-Ziv match-finder. */ + struct lz_mf *mf; /* Position in window of next match to return. */ - input_idx_t match_window_pos; + u32 match_window_pos; - /* The match-finder shall ensure the length of matches does not exceed - * this position in the input. */ - input_idx_t match_window_end; + /* The end-of-block position. We can't allow any matches to span this + * position. */ + u32 match_window_end; - /* Matches found by the match-finder are cached in the following array - * to achieve a slight speedup when the same matches are needed on + /* When doing more than one match-choosing pass over the data, matches + * found by the match-finder are cached in the following array to + * achieve a slight speedup when the same matches are needed on * subsequent passes. This is suboptimal because different matches may * be preferred with different cost models, but seems to be a worthwhile * speedup. */ - struct raw_match *cached_matches; - unsigned cached_matches_pos; - bool matches_cached; - - /* Slow algorithm only: Temporary space used for match-choosing - * algorithm. - * - * The size of this array must be at least LZX_MAX_MATCH_LEN but - * otherwise is arbitrary. More space simply allows the match-choosing - * algorithm to potentially find better matches (depending on the input, - * as always). */ - struct lzx_optimal *optimum; + struct lz_match *cached_matches; + struct lz_match *cache_ptr; + struct lz_match *cache_limit; - /* Slow algorithm only: Variables used by the match-choosing algorithm. + /* Match-chooser state, used when doing near-optimal parsing. * * 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; + struct lzx_mc_pos_data *optimum; + unsigned optimum_cur_idx; + unsigned optimum_end_idx; + + /* Previous match, used when doing lazy parsing. */ + struct lz_match prev_match; +}; + +/* + * Match chooser position data: + * + * 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_mc_pos_data { + /* The approximate minimum cost, in bits, to reach this position in the + * window which has been found so far. */ + u32 cost; +#define MC_INFINITE_COST ((u32)~0UL) + + /* 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. */ + u32 link; + + /* Offset (as in an LZ (length, offset) pair) of the + * match or literal that was taken to get to this + * position in the approximate minimum-cost parse. */ + u32 match_offset; + } prev; + struct { + /* Position at which the match or literal starting at + * this position ends in the minimum-cost parse. */ + u32 link; + + /* Offset (as in an LZ (length, offset) pair) of the + * match or literal starting at this position in the + * approximate minimum-cost parse. */ + u32 match_offset; + } next; + }; + + /* Adaptive state that exists after an approximate minimum-cost path to + * reach this position is taken. + * + * Note: we update this whenever we update the pending minimum-cost + * path. This is in contrast to LZMA, which also has an optimal parser + * that maintains a repeat offset queue per position, but will only + * compute the queue once that position is actually reached in the + * parse, meaning that matches are being considered *starting* at that + * position. However, the two methods seem to have approximately the + * same performance if appropriate optimizations are used. Intuitively + * the LZMA method seems faster, but it actually suffers from 1-2 extra + * hard-to-predict branches at each position. Probably it works better + * for LZMA than LZX because LZMA has a larger adaptive state than LZX, + * and the LZMA encoder considers more possibilities. */ + struct lzx_lru_queue queue; +}; + + +/* + * Structure to keep track of the current state of sending bits to the + * compressed output buffer. + * + * The LZX bitstream is encoded as a sequence of 16-bit coding units. + */ +struct lzx_output_bitstream { + + /* Bits that haven't yet been written to the output buffer. */ + u32 bitbuf; + + /* Number of bits currently held in @bitbuf. */ + u32 bitcount; + + /* Pointer to the start of the output buffer. */ + le16 *start; + + /* Pointer to the position in the output buffer at which the next coding + * unit should be written. */ + le16 *next; + + /* Pointer past the end of the output buffer. */ + le16 *end; }; +/* + * Initialize the output bitstream. + * + * @os + * The output bitstream structure to initialize. + * @buffer + * The buffer being written to. + * @size + * Size of @buffer, in bytes. + */ +static void +lzx_init_output(struct lzx_output_bitstream *os, void *buffer, u32 size) +{ + os->bitbuf = 0; + os->bitcount = 0; + os->start = buffer; + os->next = os->start; + os->end = os->start + size / sizeof(le16); +} + +/* + * Write some bits to the output bitstream. + * + * The bits are given by the low-order @num_bits bits of @bits. Higher-order + * bits in @bits cannot be set. At most 17 bits can be written at once. + * + * @max_bits is a compile-time constant that specifies the maximum number of + * bits that can ever be written at the call site. Currently, it is used to + * optimize away the conditional code for writing a second 16-bit coding unit + * when writing fewer than 17 bits. + * + * If the output buffer space is exhausted, then the bits will be ignored, and + * lzx_flush_output() will return 0 when it gets called. + */ +static _always_inline_attribute void +lzx_write_varbits(struct lzx_output_bitstream *os, + const u32 bits, const unsigned int num_bits, + const unsigned int max_num_bits) +{ + /* This code is optimized for LZX, which never needs to write more than + * 17 bits at once. */ + LZX_ASSERT(num_bits <= 17); + LZX_ASSERT(num_bits <= max_num_bits); + LZX_ASSERT(os->bitcount <= 15); + + /* Add the bits to the bit buffer variable. @bitcount will be at most + * 15, so there will be just enough space for the maximum possible + * @num_bits of 17. */ + os->bitcount += num_bits; + os->bitbuf = (os->bitbuf << num_bits) | bits; + + /* Check whether any coding units need to be written. */ + if (os->bitcount >= 16) { + + os->bitcount -= 16; + + /* Write a coding unit, unless it would overflow the buffer. */ + if (os->next != os->end) + *os->next++ = cpu_to_le16(os->bitbuf >> os->bitcount); + + /* If writing 17 bits, a second coding unit might need to be + * written. But because 'max_num_bits' is a compile-time + * constant, the compiler will optimize away this code at most + * call sites. */ + if (max_num_bits == 17 && os->bitcount == 16) { + if (os->next != os->end) + *os->next++ = cpu_to_le16(os->bitbuf); + os->bitcount = 0; + } + } +} + +/* Use when @num_bits is a compile-time constant. Otherwise use + * lzx_write_varbits(). */ +static _always_inline_attribute void +lzx_write_bits(struct lzx_output_bitstream *os, + const u32 bits, const unsigned int num_bits) +{ + lzx_write_varbits(os, bits, num_bits, num_bits); +} + +/* + * Flush the last coding unit to the output buffer if needed. Return the total + * number of bytes written to the output buffer, or 0 if an overflow occurred. + */ +static u32 +lzx_flush_output(struct lzx_output_bitstream *os) +{ + if (os->next == os->end) + return 0; + + if (os->bitcount != 0) + *os->next++ = cpu_to_le16(os->bitbuf << (16 - os->bitcount)); + + return (const u8 *)os->next - (const u8 *)os->start; +} + /* Returns the LZX position slot that corresponds to a given match offset, * taking into account the recent offset queue and updating it if the offset is * found in it. */ static unsigned -lzx_get_position_slot(unsigned offset, struct lzx_lru_queue *queue) +lzx_get_position_slot(u32 offset, struct lzx_lru_queue *queue) { unsigned position_slot; /* See if the offset was recently used. */ - for (unsigned i = 0; i < LZX_NUM_RECENT_OFFSETS; i++) { + for (int i = 0; i < LZX_NUM_RECENT_OFFSETS; i++) { if (offset == queue->R[i]) { /* Found it. */ @@ -435,7 +619,7 @@ lzx_get_position_slot(unsigned offset, struct lzx_lru_queue *queue) position_slot = lzx_get_position_slot_raw(offset + LZX_OFFSET_OFFSET); /* Bring the new offset to the front of the queue. */ - for (unsigned i = LZX_NUM_RECENT_OFFSETS - 1; i > 0; i--) + for (int i = LZX_NUM_RECENT_OFFSETS - 1; i > 0; i--) queue->R[i] = queue->R[i - 1]; queue->R[0] = offset; @@ -471,45 +655,40 @@ lzx_make_huffman_codes(const struct lzx_freqs *freqs, } /* - * Output an LZX match. + * Output a precomputed 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. + * @os: + * The bitstream to which to write the match. + * @ones_if_aligned + * A mask of all ones if the block is of type LZX_BLOCKTYPE_ALIGNED, + * otherwise 0. + * @match: + * The match data. + * @codes: + * Pointer to a structure that contains the codewords for the main, length, + * and aligned offset Huffman codes for the current LZX compressed block. */ static void -lzx_write_match(struct output_bitstream *out, int block_type, - struct lzx_match match, const struct lzx_codes *codes) +lzx_write_match(struct lzx_output_bitstream *os, unsigned ones_if_aligned, + struct lzx_item 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 */ + u32 position_footer = (match.data >> 8) & 0x1ffff; + unsigned position_slot = (match.data >> 25) & 0x3f; unsigned len_header; unsigned len_footer; unsigned main_symbol; unsigned num_extra_bits; - unsigned verbatim_bits; - unsigned aligned_bits; /* If the match length is less than MIN_MATCH_LEN (= 2) + - * NUM_PRIMARY_LENS (= 7), the length header contains - * the match length minus MIN_MATCH_LEN, and there is no - * length footer. + * NUM_PRIMARY_LENS (= 7), the length header contains the match length + * minus MIN_MATCH_LEN, 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 + * Otherwise, the length header contains NUM_PRIMARY_LENS, and the + * length footer contains the match length minus NUM_PRIMARY_LENS minus * MIN_MATCH_LEN. */ 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; @@ -524,351 +703,282 @@ lzx_write_match(struct output_bitstream *out, int block_type, main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS; /* Output main symbol. */ - bitstream_put_bits(out, codes->codewords.main[main_symbol], - codes->lens.main[main_symbol]); + lzx_write_varbits(os, codes->codewords.main[main_symbol], + codes->lens.main[main_symbol], + LZX_MAX_MAIN_CODEWORD_LEN); /* 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]); + if (len_header == LZX_NUM_PRIMARY_LENS) { + lzx_write_varbits(os, codes->codewords.len[len_footer], + codes->lens.len[len_footer], + LZX_MAX_LEN_CODEWORD_LEN); } + /* Output the position 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 code. Otherwise, only the verbatim bits need to be - * output. */ - if ((block_type == LZX_BLOCKTYPE_ALIGNED) && (num_extra_bits >= 3)) { + if ((num_extra_bits & ones_if_aligned) >= 3) { - verbatim_bits = position_footer >> 3; - bitstream_put_bits(out, verbatim_bits, - num_extra_bits - 3); + /* Aligned offset blocks: The low 3 bits of the position footer + * are Huffman-encoded using the aligned offset code. The + * remaining bits are output literally. */ - aligned_bits = (position_footer & 7); - bitstream_put_bits(out, - codes->codewords.aligned[aligned_bits], - codes->lens.aligned[aligned_bits]); + lzx_write_varbits(os, + position_footer >> 3, num_extra_bits - 3, 14); + + lzx_write_varbits(os, + codes->codewords.aligned[position_footer & 7], + codes->lens.aligned[position_footer & 7], + LZX_MAX_ALIGNED_CODEWORD_LEN); } else { - /* verbatim bits is the same as the position - * footer, in this case. */ - bitstream_put_bits(out, position_footer, num_extra_bits); + /* Verbatim blocks, or fewer than 3 extra bits: All position + * footer bits are output literally. */ + lzx_write_varbits(os, position_footer, num_extra_bits, 17); } } +/* Output an LZX literal (encoded with the main Huffman code). */ +static void +lzx_write_literal(struct lzx_output_bitstream *os, unsigned literal, + const struct lzx_codes *codes) +{ + lzx_write_varbits(os, codes->codewords.main[literal], + codes->lens.main[literal], LZX_MAX_MAIN_CODEWORD_LEN); +} + static unsigned -lzx_build_precode(const u8 lens[restrict], - const u8 prev_lens[restrict], - const unsigned num_syms, - input_idx_t precode_freqs[restrict LZX_PRECODE_NUM_SYMBOLS], - u8 output_syms[restrict num_syms], - u8 precode_lens[restrict LZX_PRECODE_NUM_SYMBOLS], - u16 precode_codewords[restrict LZX_PRECODE_NUM_SYMBOLS], - unsigned *num_additional_bits_ret) +lzx_compute_precode_items(const u8 lens[restrict], + const u8 prev_lens[restrict], + const unsigned num_lens, + u32 precode_freqs[restrict], + unsigned precode_items[restrict]) { - memset(precode_freqs, 0, - LZX_PRECODE_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 precode. - * - * cur_run_len keeps track of how many code word lengths are in the - * current run of identical lengths. */ - unsigned output_syms_idx = 0; - unsigned cur_run_len = 1; - unsigned num_additional_bits = 0; - for (unsigned 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; - } + unsigned *itemptr; + unsigned run_start; + unsigned run_end; + unsigned extra_bits; + int delta; + u8 len; + + itemptr = precode_items; + run_start = 0; + do { + /* Find the next run of codeword lengths. */ + + /* len = the length being repeated */ + len = lens[run_start]; - /* Run ended! Check if it is a run of zeroes or a run of - * nonzeroes. */ + run_end = run_start + 1; - /* The symbol that was repeated in the run--- not to be confused - * with the length *of* the run (cur_run_len) */ - unsigned len_in_run = lens[i - 1]; + /* Fast case for a single length. */ + if (likely(run_end == num_lens || len != lens[run_end])) { + delta = prev_lens[run_start] - len; + if (delta < 0) + delta += 17; + precode_freqs[delta]++; + *itemptr++ = delta; + run_start++; + continue; + } - if (len_in_run == 0) { - /* A run of 0's. Encode it in as few length - * codes as we can. */ + /* Extend the run. */ + do { + run_end++; + } while (run_end != num_lens && len == lens[run_end]); - /* 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) { - unsigned additional_bits; + if (len == 0) { + /* Run of zeroes. */ - additional_bits = min(cur_run_len - 20, 0x1f); - num_additional_bits += 5; + /* Symbol 18: RLE 20 to 51 zeroes at a time. */ + while ((run_end - run_start) >= 20) { + extra_bits = min((run_end - run_start) - 20, 0x1f); precode_freqs[18]++; - output_syms[output_syms_idx++] = 18; - output_syms[output_syms_idx++] = additional_bits; - cur_run_len -= 20 + additional_bits; + *itemptr++ = 18 | (extra_bits << 5); + run_start += 20 + extra_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) { - unsigned additional_bits; - - additional_bits = min(cur_run_len - 4, 0xf); - num_additional_bits += 4; + /* Symbol 17: RLE 4 to 19 zeroes at a time. */ + if ((run_end - run_start) >= 4) { + extra_bits = min((run_end - run_start) - 4, 0xf); precode_freqs[17]++; - output_syms[output_syms_idx++] = 17; - output_syms[output_syms_idx++] = additional_bits; - cur_run_len -= 4 + additional_bits; + *itemptr++ = 17 | (extra_bits << 5); + run_start += 4 + extra_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 code. - * */ - while (cur_run_len >= 4) { - unsigned additional_bits; - signed char delta; - - additional_bits = (cur_run_len > 4); - num_additional_bits += 1; - delta = (signed char)prev_lens[i - cur_run_len] - - (signed char)len_in_run; + /* Symbol 19: RLE 4 to 5 of any length at a time. */ + while ((run_end - run_start) >= 4) { + extra_bits = (run_end - run_start) > 4; + delta = prev_lens[run_start] - len; 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; + precode_freqs[delta]++; + *itemptr++ = 19 | (extra_bits << 5) | (delta << 6); + run_start += 4 + extra_bits; } } - /* Any remaining lengths in the run are outputted without RLE, - * as a difference from the length of that codeword in the - * previous code. */ - while (cur_run_len > 0) { - signed char delta; - - delta = (signed char)prev_lens[i - cur_run_len] - - (signed char)len_in_run; + /* Output any remaining lengths without RLE. */ + while (run_start != run_end) { + delta = prev_lens[run_start] - len; if (delta < 0) delta += 17; - - precode_freqs[(unsigned char)delta]++; - output_syms[output_syms_idx++] = delta; - cur_run_len--; + precode_freqs[delta]++; + *itemptr++ = delta; + run_start++; } + } while (run_start != num_lens); - cur_run_len = 1; - } - - /* Build the precode from the frequencies of the length symbols. */ - - make_canonical_huffman_code(LZX_PRECODE_NUM_SYMBOLS, - LZX_MAX_PRE_CODEWORD_LEN, - precode_freqs, precode_lens, - precode_codewords); - - *num_additional_bits_ret = num_additional_bits; - - return output_syms_idx; + return itemptr - precode_items; } /* - * 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_PRECODE_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. + * Output a Huffman code in the compressed form used in LZX. + * + * The Huffman code is represented in the output as a logical series of codeword + * lengths from which the Huffman code, which must be in canonical form, can be + * reconstructed. + * + * The codeword lengths are themselves compressed using a separate Huffman code, + * the "precode", which contains a symbol for each possible codeword length in + * the larger code as well as several special symbols to represent repeated + * codeword lengths (a form of run-length encoding). The precode is itself + * constructed in canonical form, and its codeword lengths are represented + * literally in 20 4-bit fields that immediately precede the compressed codeword + * lengths of the larger code. + * + * Furthermore, the codeword lengths of the larger code are actually represented + * as deltas from the codeword lengths of the corresponding code in the previous + * block. + * + * @os: + * Bitstream to which to write the compressed Huffman code. + * @lens: + * The codeword lengths, indexed by symbol, in the Huffman code. + * @prev_lens: + * The codeword lengths, indexed by symbol, in the corresponding Huffman + * code in the previous block, or all zeroes if this is the first block. + * @num_lens: + * The number of symbols in the Huffman code. */ static void -lzx_write_compressed_code(struct output_bitstream *out, +lzx_write_compressed_code(struct lzx_output_bitstream *os, const u8 lens[restrict], const u8 prev_lens[restrict], - unsigned num_syms) + unsigned num_lens) { - input_idx_t precode_freqs[LZX_PRECODE_NUM_SYMBOLS]; - u8 output_syms[num_syms]; + u32 precode_freqs[LZX_PRECODE_NUM_SYMBOLS]; u8 precode_lens[LZX_PRECODE_NUM_SYMBOLS]; - u16 precode_codewords[LZX_PRECODE_NUM_SYMBOLS]; + u32 precode_codewords[LZX_PRECODE_NUM_SYMBOLS]; + unsigned precode_items[num_lens]; + unsigned num_precode_items; + unsigned precode_item; + unsigned precode_sym; unsigned i; - unsigned num_output_syms; - u8 precode_sym; - unsigned dummy; - - num_output_syms = lzx_build_precode(lens, - prev_lens, - num_syms, - precode_freqs, - output_syms, - precode_lens, - precode_codewords, - &dummy); - - /* Write the lengths of the precode codes to the output. */ + + for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) + precode_freqs[i] = 0; + + /* Compute the "items" (RLE / literal tokens and extra bits) with which + * the codeword lengths in the larger code will be output. */ + num_precode_items = lzx_compute_precode_items(lens, + prev_lens, + num_lens, + precode_freqs, + precode_items); + + /* Build the precode. */ + make_canonical_huffman_code(LZX_PRECODE_NUM_SYMBOLS, + LZX_MAX_PRE_CODEWORD_LEN, + precode_freqs, precode_lens, + precode_codewords); + + /* Output the lengths of the codewords in the precode. */ for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) - bitstream_put_bits(out, precode_lens[i], - LZX_PRECODE_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; + lzx_write_bits(os, precode_lens[i], LZX_PRECODE_ELEMENT_SIZE); + + /* Output the encoded lengths of the codewords in the larger code. */ + for (i = 0; i < num_precode_items; i++) { + precode_item = precode_items[i]; + precode_sym = precode_item & 0x1F; + lzx_write_varbits(os, precode_codewords[precode_sym], + precode_lens[precode_sym], + LZX_MAX_PRE_CODEWORD_LEN); + if (precode_sym >= 17) { + if (precode_sym == 17) { + lzx_write_bits(os, precode_item >> 5, 4); + } else if (precode_sym == 18) { + lzx_write_bits(os, precode_item >> 5, 5); + } else { + lzx_write_bits(os, (precode_item >> 5) & 1, 1); + precode_sym = precode_item >> 6; + lzx_write_varbits(os, precode_codewords[precode_sym], + precode_lens[precode_sym], + LZX_MAX_PRE_CODEWORD_LEN); + } } } } /* - * Writes all compressed matches and literal bytes in an LZX block to the the - * output bitstream. + * Write all matches and literal bytes (which were precomputed) in an LZX + * compressed block to the output bitstream in the final compressed + * representation. * - * @ostream + * @os * The output bitstream. * @block_type - * The type of the block (LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM). - * @match_tab - * The array of matches/literals that will be output (length @match_count). - * @match_count - * Number of matches/literals to be output. + * The chosen type of the LZX compressed block (LZX_BLOCKTYPE_ALIGNED or + * LZX_BLOCKTYPE_VERBATIM). + * @items + * The array of matches/literals to output. + * @num_items + * Number of matches/literals to output (length of @items). * @codes - * Pointer to a structure that contains the codewords for the main, length, - * and aligned offset Huffman codes. + * The main, length, and aligned offset Huffman codes for the current + * LZX compressed block. */ 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) +lzx_write_items(struct lzx_output_bitstream *os, int block_type, + const struct lzx_item items[], u32 num_items, + 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]); - } + unsigned ones_if_aligned = 0U - (block_type == LZX_BLOCKTYPE_ALIGNED); + + for (u32 i = 0; i < num_items; i++) { + /* The high bit of the 32-bit intermediate representation + * indicates whether the item is an actual LZ-style match (1) or + * a literal byte (0). */ + if (items[i].data & 0x80000000) + lzx_write_match(os, ones_if_aligned, items[i], codes); + else + lzx_write_literal(os, items[i].data, codes); } } -static void -lzx_assert_codes_valid(const struct lzx_codes * codes, unsigned num_main_syms) -{ -#ifdef ENABLE_LZX_DEBUG - unsigned i; - - for (i = 0; i < num_main_syms; i++) - LZX_ASSERT(codes->lens.main[i] <= LZX_MAX_MAIN_CODEWORD_LEN); - - for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) - LZX_ASSERT(codes->lens.len[i] <= LZX_MAX_LEN_CODEWORD_LEN); - - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) - LZX_ASSERT(codes->lens.aligned[i] <= LZX_MAX_ALIGNED_CODEWORD_LEN); - - const unsigned tablebits = 10; - u16 decode_table[(1 << tablebits) + - (2 * max(num_main_syms, LZX_LENCODE_NUM_SYMBOLS))] - _aligned_attribute(DECODE_TABLE_ALIGNMENT); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - num_main_syms, - min(tablebits, LZX_MAINCODE_TABLEBITS), - codes->lens.main, - LZX_MAX_MAIN_CODEWORD_LEN)); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - LZX_LENCODE_NUM_SYMBOLS, - min(tablebits, LZX_LENCODE_TABLEBITS), - codes->lens.len, - LZX_MAX_LEN_CODEWORD_LEN)); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - LZX_ALIGNEDCODE_NUM_SYMBOLS, - min(tablebits, LZX_ALIGNEDCODE_TABLEBITS), - codes->lens.aligned, - LZX_MAX_ALIGNED_CODEWORD_LEN)); -#endif /* ENABLE_LZX_DEBUG */ -} - /* Write an LZX aligned offset or verbatim block to the output. */ static void lzx_write_compressed_block(int block_type, - unsigned block_size, - unsigned max_window_size, + u32 block_size, + unsigned window_order, unsigned num_main_syms, - struct lzx_match * chosen_matches, - unsigned num_chosen_matches, + struct lzx_item * chosen_items, + u32 num_chosen_items, const struct lzx_codes * codes, const struct lzx_codes * prev_codes, - struct output_bitstream * ostream) + struct lzx_output_bitstream * os) { - unsigned i; - LZX_ASSERT(block_type == LZX_BLOCKTYPE_ALIGNED || block_type == LZX_BLOCKTYPE_VERBATIM); - lzx_assert_codes_valid(codes, num_main_syms); /* The first three bits indicate the type of block and are one of the * LZX_BLOCKTYPE_* constants. */ - bitstream_put_bits(ostream, block_type, 3); + lzx_write_bits(os, block_type, 3); /* Output the block size. * @@ -886,82 +996,59 @@ lzx_write_compressed_block(int block_type, * because WIMs created with chunk size greater than 32768 can seemingly * only be opened by wimlib anyway. */ if (block_size == LZX_DEFAULT_BLOCK_SIZE) { - bitstream_put_bits(ostream, 1, 1); + lzx_write_bits(os, 1, 1); } else { - bitstream_put_bits(ostream, 0, 1); + lzx_write_bits(os, 0, 1); - if (max_window_size >= 65536) - bitstream_put_bits(ostream, block_size >> 16, 8); + if (window_order >= 16) + lzx_write_bits(os, block_size >> 16, 8); + + lzx_write_bits(os, block_size & 0xFFFF, 16); + } - bitstream_put_bits(ostream, block_size, 16); + /* Output the aligned offset code. */ + if (block_type == LZX_BLOCKTYPE_ALIGNED) { + for (int i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { + lzx_write_bits(os, codes->lens.aligned[i], + LZX_ALIGNEDCODE_ELEMENT_SIZE); + } } - /* 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 code to be written - * (before the main code). */ - if (block_type == LZX_BLOCKTYPE_ALIGNED) - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) - bitstream_put_bits(ostream, codes->lens.aligned[i], - LZX_ALIGNEDCODE_ELEMENT_SIZE); - - LZX_DEBUG("Writing main code..."); - - /* Write the precode 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, + /* Output the main code (two parts). */ + lzx_write_compressed_code(os, codes->lens.main, prev_codes->lens.main, LZX_NUM_CHARS); - - /* Write the precode 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, + lzx_write_compressed_code(os, codes->lens.main + LZX_NUM_CHARS, prev_codes->lens.main + LZX_NUM_CHARS, num_main_syms - LZX_NUM_CHARS); - LZX_DEBUG("Writing length code..."); - - /* Write the precode and lengths for the length code. */ - lzx_write_compressed_code(ostream, - codes->lens.len, + /* Output the length code. */ + lzx_write_compressed_code(os, codes->lens.len, prev_codes->lens.len, LZX_LENCODE_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."); + /* Output the compressed matches and literals. */ + lzx_write_items(os, block_type, chosen_items, num_chosen_items, 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_all_blocks(struct lzx_compressor *c, struct lzx_output_bitstream *os) { - const struct lzx_codes *prev_codes = &ctx->zero_codes; - for (unsigned i = 0; i < ctx->num_blocks; i++) { - const struct lzx_block_spec *spec = &ctx->block_specs[i]; - - LZX_DEBUG("Writing block %u/%u (type=%d, size=%u, num_chosen_matches=%u)...", - i + 1, ctx->num_blocks, - spec->block_type, spec->block_size, - spec->num_chosen_matches); + const struct lzx_codes *prev_codes = &c->zero_codes; + for (unsigned i = 0; i < c->num_blocks; i++) { + const struct lzx_block_spec *spec = &c->block_specs[i]; lzx_write_compressed_block(spec->block_type, spec->block_size, - ctx->max_window_size, - ctx->num_main_syms, - &ctx->chosen_matches[spec->chosen_matches_start_pos], - spec->num_chosen_matches, + c->window_order, + c->num_main_syms, + spec->chosen_items, + spec->num_chosen_items, &spec->codes, prev_codes, - ostream); + os); prev_codes = &spec->codes; } @@ -969,7 +1056,7 @@ lzx_write_all_blocks(struct lzx_compressor *ctx, struct output_bitstream *ostrea /* Constructs an LZX match from a literal byte and updates the main code symbol * frequencies. */ -static u32 +static inline u32 lzx_tally_literal(u8 lit, struct lzx_freqs *freqs) { freqs->main[lit]++; @@ -980,12 +1067,12 @@ lzx_tally_literal(u8 lit, struct lzx_freqs *freqs) * 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_tally_match(unsigned match_len, unsigned match_offset, +static inline u32 +lzx_tally_match(unsigned match_len, u32 match_offset, struct lzx_freqs *freqs, struct lzx_lru_queue *queue) { unsigned position_slot; - unsigned position_footer; + u32 position_footer; u32 len_header; unsigned main_symbol; unsigned len_footer; @@ -997,7 +1084,7 @@ lzx_tally_match(unsigned match_len, unsigned match_offset, * as part of the main symbol) and a position footer. */ position_slot = lzx_get_position_slot(match_offset, queue); position_footer = (match_offset + LZX_OFFSET_OFFSET) & - ((1U << lzx_get_num_extra_bits(position_slot)) - 1); + (((u32)1 << lzx_get_num_extra_bits(position_slot)) - 1); /* The match length shall be encoded as a length header (itself encoded * as part of the main symbol) and an optional length footer. */ @@ -1032,7 +1119,7 @@ lzx_tally_match(unsigned match_len, unsigned match_offset, freqs->aligned[position_footer & 7]++; /* Pack the position slot, position footer, and match length into an - * intermediate representation. See `struct lzx_match' for details. + * intermediate representation. See `struct lzx_item' for details. */ LZX_ASSERT(LZX_MAX_POSITION_SLOTS <= 64); LZX_ASSERT(lzx_get_num_extra_bits(LZX_MAX_POSITION_SLOTS - 1) <= 17); @@ -1047,92 +1134,36 @@ lzx_tally_match(unsigned match_len, unsigned match_offset, (adjusted_match_len); } -struct lzx_record_ctx { - struct lzx_freqs freqs; - struct lzx_lru_queue queue; - struct lzx_match *matches; -}; - -static void -lzx_record_match(unsigned len, unsigned offset, void *_ctx) -{ - struct lzx_record_ctx *ctx = _ctx; - - (ctx->matches++)->data = lzx_tally_match(len, offset, &ctx->freqs, &ctx->queue); -} - -static void -lzx_record_literal(u8 lit, void *_ctx) -{ - struct lzx_record_ctx *ctx = _ctx; - - (ctx->matches++)->data = lzx_tally_literal(lit, &ctx->freqs); -} - /* Returns the cost, in bits, to output a literal byte using the specified cost * model. */ -static unsigned +static u32 lzx_literal_cost(u8 c, const struct lzx_costs * 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. Take into account the match offset LRU - * queue and optionally update it. */ -static unsigned -lzx_match_cost(unsigned length, unsigned offset, const struct lzx_costs *costs, - struct lzx_lru_queue *queue) +/* Returns the cost, in bits, to output a repeat offset match of the specified + * length and position slot (repeat index) using the specified cost model. */ +static u32 +lzx_repmatch_cost(u32 len, unsigned position_slot, const struct lzx_costs *costs) { - unsigned position_slot; unsigned len_header, main_symbol; - unsigned cost = 0; - - position_slot = lzx_get_position_slot(offset, queue); + u32 cost = 0; - len_header = min(length - LZX_MIN_MATCH_LEN, LZX_NUM_PRIMARY_LENS); + len_header = min(len - LZX_MIN_MATCH_LEN, 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_OFFSET_OFFSET) & 7]; - } else { - cost += num_extra_bits; - } - /* Account for extra length information. */ if (len_header == LZX_NUM_PRIMARY_LENS) - cost += costs->len[length - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS]; + cost += costs->len[len - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS]; return cost; - -} - -/* Fast heuristic cost evaluation to use in the inner loop of the match-finder. - * Unlike lzx_match_cost() which does a true cost evaluation, this simply - * prioritize matches based on their offset. */ -static input_idx_t -lzx_match_cost_fast(input_idx_t length, input_idx_t offset, const void *_queue) -{ - const struct lzx_lru_queue *queue = _queue; - - /* It seems well worth it to take the time to give priority to recently - * used offsets. */ - for (input_idx_t i = 0; i < LZX_NUM_RECENT_OFFSETS; i++) - if (offset == queue->R[i]) - return i; - - return offset; } -/* Set the cost model @ctx->costs from the Huffman codeword lengths specified in +/* Set the cost model @c->costs from the Huffman codeword lengths specified in * @lens. * * The cost model and codeword lengths are almost the same thing, but the @@ -1142,385 +1173,840 @@ lzx_match_cost_fast(input_idx_t length, input_idx_t offset, const void *_queue) * 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) +lzx_set_costs(struct lzx_compressor *c, const struct lzx_lens * lens, + unsigned nostat) { unsigned i; - unsigned num_main_syms = ctx->num_main_syms; /* Main code */ - for (i = 0; i < num_main_syms; i++) { - ctx->costs.main[i] = lens->main[i]; - if (ctx->costs.main[i] == 0) - ctx->costs.main[i] = ctx->params.alg_params.slow.main_nostat_cost; - } + for (i = 0; i < c->num_main_syms; i++) + c->costs.main[i] = lens->main[i] ? lens->main[i] : nostat; /* Length code */ - for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) { - ctx->costs.len[i] = lens->len[i]; - if (ctx->costs.len[i] == 0) - ctx->costs.len[i] = ctx->params.alg_params.slow.len_nostat_cost; - } + for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) + c->costs.len[i] = lens->len[i] ? lens->len[i] : nostat; /* Aligned offset code */ - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { - ctx->costs.aligned[i] = lens->aligned[i]; - if (ctx->costs.aligned[i] == 0) - ctx->costs.aligned[i] = ctx->params.alg_params.slow.aligned_nostat_cost; - } + for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) + c->costs.aligned[i] = lens->aligned[i] ? lens->aligned[i] : nostat / 2; } -/* 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) +/* Don't allow matches to span the end of an LZX block. */ +static inline u32 +maybe_truncate_matches(struct lz_match matches[], u32 num_matches, + struct lzx_compressor *c) { - LZX_ASSERT(n <= ctx->match_window_end - ctx->match_window_pos); - if (ctx->matches_cached) { - ctx->match_window_pos += n; - while (n--) { - ctx->cached_matches_pos += - ctx->cached_matches[ctx->cached_matches_pos].len + 1; - } - } else { - while (n--) { - ctx->cached_matches[ctx->cached_matches_pos++].len = 0; - lz_sarray_skip_position(&ctx->lz_sarray); - ctx->match_window_pos++; + if (c->match_window_end < c->cur_window_size && num_matches != 0) { + u32 limit = c->match_window_end - c->match_window_pos; + + if (limit >= LZX_MIN_MATCH_LEN) { + + u32 i = num_matches - 1; + do { + if (matches[i].len >= limit) { + matches[i].len = limit; + + /* Truncation might produce multiple + * matches with length 'limit'. Keep at + * most 1. */ + num_matches = i + 1; + } + } while (i--); + } else { + num_matches = 0; } - LZX_ASSERT(lz_sarray_get_pos(&ctx->lz_sarray) == ctx->match_window_pos); } + return num_matches; } -/* Retrieve a list of matches available at the next position in the input. - * - * The matches are written to ctx->matches in decreasing order of length, and - * the return value is the number of matches found. */ static unsigned -lzx_lz_get_matches_caching(struct lzx_compressor *ctx, - const struct lzx_lru_queue *queue, - struct raw_match **matches_ret) +lzx_get_matches_fillcache_singleblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) { + struct lz_match *cache_ptr; + struct lz_match *matches; unsigned num_matches; - struct raw_match *matches; - - LZX_ASSERT(ctx->match_window_pos <= ctx->match_window_end); - matches = &ctx->cached_matches[ctx->cached_matches_pos + 1]; - - if (ctx->matches_cached) { - num_matches = matches[-1].len; + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (likely(cache_ptr <= c->cache_limit)) { + num_matches = lz_mf_get_matches(c->mf, matches); + cache_ptr->len = num_matches; + c->cache_ptr = matches + num_matches; } else { - LZX_ASSERT(lz_sarray_get_pos(&ctx->lz_sarray) == ctx->match_window_pos); - num_matches = lz_sarray_get_matches(&ctx->lz_sarray, - matches, - lzx_match_cost_fast, - queue); - matches[-1].len = num_matches; + num_matches = 0; } - ctx->cached_matches_pos += num_matches + 1; + c->match_window_pos++; *matches_ret = matches; - - /* Cap the length of returned matches to the number of bytes remaining, - * if it is not the whole window. */ - if (ctx->match_window_end < ctx->window_size) { - unsigned maxlen = ctx->match_window_end - ctx->match_window_pos; - for (unsigned i = 0; i < num_matches; i++) - if (matches[i].len > maxlen) - matches[i].len = maxlen; - } -#if 0 - fprintf(stderr, "Pos %u/%u: %u matches\n", - ctx->match_window_pos, ctx->match_window_end, num_matches); - for (unsigned i = 0; i < num_matches; i++) - fprintf(stderr, "\tLen %u Offset %u\n", matches[i].len, matches[i].offset); -#endif - -#ifdef ENABLE_LZX_DEBUG - for (unsigned i = 0; i < num_matches; i++) { - LZX_ASSERT(matches[i].len >= LZX_MIN_MATCH_LEN); - LZX_ASSERT(matches[i].len <= LZX_MAX_MATCH_LEN); - LZX_ASSERT(matches[i].len <= ctx->match_window_end - ctx->match_window_pos); - LZX_ASSERT(matches[i].offset > 0); - LZX_ASSERT(matches[i].offset <= 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)); - } -#endif - - ctx->match_window_pos++; return num_matches; } -/* - * Reverse the linked list of near-optimal matches so that they can be returned +static unsigned +lzx_get_matches_fillcache_multiblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; + + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (likely(cache_ptr <= c->cache_limit)) { + num_matches = lz_mf_get_matches(c->mf, matches); + num_matches = maybe_truncate_matches(matches, num_matches, c); + cache_ptr->len = num_matches; + c->cache_ptr = matches + num_matches; + } else { + num_matches = 0; + } + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} + +static unsigned +lzx_get_matches_usecache(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; + + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (cache_ptr <= c->cache_limit) { + num_matches = cache_ptr->len; + c->cache_ptr = matches + num_matches; + } else { + num_matches = 0; + } + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} + +static unsigned +lzx_get_matches_usecache_nocheck(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; + + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + num_matches = cache_ptr->len; + c->cache_ptr = matches + num_matches; + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} + +static unsigned +lzx_get_matches_nocache_singleblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *matches; + unsigned num_matches; + + matches = c->cache_ptr; + num_matches = lz_mf_get_matches(c->mf, matches); + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} + +static unsigned +lzx_get_matches_nocache_multiblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *matches; + unsigned num_matches; + + matches = c->cache_ptr; + num_matches = lz_mf_get_matches(c->mf, matches); + num_matches = maybe_truncate_matches(matches, num_matches, c); + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} + +/* + * Find matches at the next position in the window. + * + * Returns the number of matches found and sets *matches_ret to point to the + * matches array. The matches will be sorted by strictly increasing length and + * offset. + */ +static inline unsigned +lzx_get_matches(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + return (*c->get_matches_func)(c, matches_ret); +} + +static void +lzx_skip_bytes_fillcache(struct lzx_compressor *c, unsigned n) +{ + struct lz_match *cache_ptr; + + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + lz_mf_skip_positions(c->mf, n); + if (cache_ptr <= c->cache_limit) { + do { + cache_ptr->len = 0; + cache_ptr += 1; + } while (--n && cache_ptr <= c->cache_limit); + } + c->cache_ptr = cache_ptr; +} + +static void +lzx_skip_bytes_usecache(struct lzx_compressor *c, unsigned n) +{ + struct lz_match *cache_ptr; + + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + if (cache_ptr <= c->cache_limit) { + do { + cache_ptr += 1 + cache_ptr->len; + } while (--n && cache_ptr <= c->cache_limit); + } + c->cache_ptr = cache_ptr; +} + +static void +lzx_skip_bytes_usecache_nocheck(struct lzx_compressor *c, unsigned n) +{ + struct lz_match *cache_ptr; + + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + do { + cache_ptr += 1 + cache_ptr->len; + } while (--n); + c->cache_ptr = cache_ptr; +} + +static void +lzx_skip_bytes_nocache(struct lzx_compressor *c, unsigned n) +{ + c->match_window_pos += n; + lz_mf_skip_positions(c->mf, n); +} + +/* + * Skip the specified number of positions in the window (don't search for + * matches at them). + */ +static inline void +lzx_skip_bytes(struct lzx_compressor *c, unsigned n) +{ + return (*c->skip_bytes_func)(c, n); +} + +/* + * 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) +static struct lz_match +lzx_match_chooser_reverse_list(struct lzx_compressor *c, unsigned cur_pos) { unsigned prev_link, saved_prev_link; unsigned prev_match_offset, saved_prev_match_offset; - ctx->optimum_end_idx = cur_pos; + c->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; + saved_prev_link = c->optimum[cur_pos].prev.link; + saved_prev_match_offset = c->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; + saved_prev_link = c->optimum[prev_link].prev.link; + saved_prev_match_offset = c->optimum[prev_link].prev.match_offset; - ctx->optimum[prev_link].next.link = cur_pos; - ctx->optimum[prev_link].next.match_offset = prev_match_offset; + c->optimum[prev_link].next.link = cur_pos; + c->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; + c->optimum_cur_idx = c->optimum[0].next.link; - return (struct raw_match) - { .len = ctx->optimum_cur_idx, - .offset = ctx->optimum[0].next.match_offset, + return (struct lz_match) + { .len = c->optimum_cur_idx, + .offset = c->optimum[0].next.match_offset, }; } /* - * lzx_lz_get_near_optimal_match() - + * Find the longest repeat offset match. * - * Choose the optimal match or literal to use at the next position in the input. + * If no match of at least LZX_MIN_MATCH_LEN bytes is found, then return 0. * - * Unlike a greedy parser that always takes the longest match, or even a + * If a match of at least LZX_MIN_MATCH_LEN bytes is found, then return its + * length and set *slot_ret to the index of its offset in @queue. + */ +static inline u32 +lzx_repsearch(const u8 * const strptr, const u32 bytes_remaining, + const struct lzx_lru_queue *queue, unsigned *slot_ret) +{ + BUILD_BUG_ON(LZX_MIN_MATCH_LEN != 2); + return lz_repsearch(strptr, bytes_remaining, LZX_MAX_MATCH_LEN, + queue->R, LZX_NUM_RECENT_OFFSETS, slot_ret); +} + +/* + * lzx_choose_near_optimal_item() - + * + * Choose an approximately optimal match or literal to use at the next position + * in the string, or "window", being LZ-encoded. + * + * This is based on algorithms used in 7-Zip, including the DEFLATE encoder + * and the LZMA encoder, written by Igor Pavlov. + * + * Unlike a greedy parser that always takes the longest match, or even a "lazy" * parser with one match/literal look-ahead like zlib, the algorithm used here - * may look ahead many matches/literals to determine the optimal 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. + * may look ahead many matches/literals to determine the approximately optimal + * match/literal to code 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 + * estimated real cost of coding each match/literal based on the underlying + * entropy encoding. + * + * Still, this is not a true optimal parser for several reasons: + * + * - Real compression formats use entropy encoding of the literal/match + * sequence, so the real cost of coding each match or literal is unknown until + * the parse is fully determined. It can be approximated based on iterative + * parses, but the end result is not guaranteed to be globally optimal. * - * Still, this is not truly an optimal parser because very long matches are - * taken immediately, and the raw match-finder takes some shortcuts. This is - * done to avoid considering many different alternatives that are unlikely to - * be significantly better. + * - Very long matches are chosen immediately. This is because locations with + * long matches are likely to have many possible alternatives that would cause + * slow optimal parsing, but also such locations are already highly + * compressible so it is not too harmful to just grab the longest match. * - * This algorithm is based on that used in 7-Zip's DEFLATE encoder. + * - Not all possible matches at each location are considered because the + * underlying match-finder limits the number and type of matches produced at + * each position. For example, for a given match length it's usually not + * worth it to only consider matches other than the lowest-offset match, + * except in the case of a repeat offset. + * + * - Although we take into account the adaptive state (in LZX, the recent offset + * queue), coding decisions made with respect to the adaptive state will be + * locally optimal but will not necessarily be globally optimal. This is + * because the algorithm only keeps the least-costly path to get to a given + * location and does not take into account that a slightly more costly path + * could result in a different adaptive state that ultimately results in a + * lower global cost. + * + * - The array space used by this function is bounded, so in degenerate cases it + * is forced to start returning matches/literals before the algorithm has + * really finished. * * Each call to this function does one of two things: * - * 1. Build a near-optimal sequence of matches/literals, up to some point, that + * 1. Build a sequence of near-optimal 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) - * - * Plus any state used by the raw match-finder. + * function. * * The return value is a (length, offset) pair specifying the match or literal - * chosen. For literals, the length is less than LZX_MIN_MATCH_LEN and the - * offset is meaningless. + * chosen. For literals, the length is 0 or 1 and the offset is meaningless. */ -static struct raw_match -lzx_lz_get_near_optimal_match(struct lzx_compressor * ctx) +static struct lz_match +lzx_choose_near_optimal_item(struct lzx_compressor *c) { - unsigned num_possible_matches; - struct raw_match *possible_matches; - struct raw_match match; - unsigned longest_match_len; - - if (ctx->optimum_cur_idx != ctx->optimum_end_idx) { + unsigned num_matches; + const struct lz_match *matches; + struct lz_match match; + u32 longest_len; + u32 longest_rep_len; + unsigned longest_rep_slot; + unsigned cur_pos; + unsigned end_pos; + struct lzx_mc_pos_data *optimum = c->optimum; + + if (c->optimum_cur_idx != c->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; + match.len = optimum[c->optimum_cur_idx].next.link - + c->optimum_cur_idx; + match.offset = optimum[c->optimum_cur_idx].next.match_offset; - ctx->optimum_cur_idx = ctx->optimum[ctx->optimum_cur_idx].next.link; + c->optimum_cur_idx = optimum[c->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; + c->optimum_cur_idx = 0; + c->optimum_end_idx = 0; - /* Get matches at this position. */ - num_possible_matches = lzx_lz_get_matches_caching(ctx, &ctx->queue, &possible_matches); + /* Search for matches at repeat offsets. As a heuristic, we only keep + * the one with the longest match length. */ + if (likely(c->match_window_pos >= 1)) { + longest_rep_len = lzx_repsearch(&c->cur_window[c->match_window_pos], + c->match_window_end - c->match_window_pos, + &c->queue, + &longest_rep_slot); + } else { + longest_rep_len = 0; + } - /* If no matches found, return literal. */ - if (num_possible_matches == 0) - return (struct raw_match){ .len = 0 }; + /* If there's a long match with a repeat offset, choose it immediately. */ + if (longest_rep_len >= c->params.nice_match_length) { + lzx_skip_bytes(c, longest_rep_len); + return (struct lz_match) { + .len = longest_rep_len, + .offset = c->queue.R[longest_rep_slot], + }; + } - /* The matches that were found are sorted in decreasing order by length. - * Get the length of the longest one. */ - longest_match_len = possible_matches[0].len; + /* Find other matches. */ + num_matches = lzx_get_matches(c, &matches); - /* Greedy heuristic: if the longest match that was found is greater - * than the number of fast bytes, return it immediately; don't both - * doing more work. */ - if (longest_match_len > ctx->params.alg_params.slow.num_fast_bytes) { - lzx_lz_skip_bytes(ctx, longest_match_len - 1); - return possible_matches[0]; + /* If there's a long match, choose it immediately. */ + if (num_matches) { + longest_len = matches[num_matches - 1].len; + if (longest_len >= c->params.nice_match_length) { + lzx_skip_bytes(c, longest_len - 1); + return matches[num_matches - 1]; + } + } else { + longest_len = 1; } - /* Calculate the cost to reach the next position by outputting a - * literal. */ - ctx->optimum[0].queue = ctx->queue; - ctx->optimum[1].queue = ctx->optimum[0].queue; - 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 the next position by coding a literal. */ + optimum[1].queue = c->queue; + optimum[1].cost = lzx_literal_cost(c->cur_window[c->match_window_pos - 1], + &c->costs); + 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. */ - BUILD_BUG_ON(LZX_MIN_MATCH_LEN != 2); - for (unsigned len = LZX_MIN_MATCH_LEN, match_idx = num_possible_matches - 1; - len <= longest_match_len; len++) { - - LZX_ASSERT(match_idx < num_possible_matches); - - ctx->optimum[len].queue = ctx->optimum[0].queue; - 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, - &ctx->optimum[len].queue); - if (len == possible_matches[match_idx].len) - match_idx--; + * reached by the longest match. + * + * Note: We consider only the lowest-offset match that reaches each + * position. + * + * Note: Some of the cost calculation stays the same for each offset, + * regardless of how many lengths it gets used for. Therefore, to + * improve performance, we hand-code the cost calculation instead of + * calling lzx_match_cost() to do a from-scratch cost evaluation at each + * length. */ + for (unsigned i = 0, len = 2; i < num_matches; i++) { + u32 offset; + struct lzx_lru_queue queue; + u32 position_cost; + unsigned position_slot; + unsigned num_extra_bits; + + offset = matches[i].offset; + queue = c->queue; + position_cost = 0; + + position_slot = lzx_get_position_slot(offset, &queue); + num_extra_bits = lzx_get_num_extra_bits(position_slot); + if (num_extra_bits >= 3) { + position_cost += num_extra_bits - 3; + position_cost += c->costs.aligned[(offset + LZX_OFFSET_OFFSET) & 7]; + } else { + position_cost += num_extra_bits; + } + + do { + u32 cost; + unsigned len_header; + unsigned main_symbol; + + cost = position_cost; + + if (len - LZX_MIN_MATCH_LEN < LZX_NUM_PRIMARY_LENS) { + len_header = len - LZX_MIN_MATCH_LEN; + } else { + len_header = LZX_NUM_PRIMARY_LENS; + cost += c->costs.len[len - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS]; + } + + main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS; + cost += c->costs.main[main_symbol]; + + optimum[len].queue = queue; + optimum[len].prev.link = 0; + optimum[len].prev.match_offset = offset; + optimum[len].cost = cost; + } while (++len <= matches[i].len); } + end_pos = longest_len; - unsigned cur_pos = 0; + if (longest_rep_len) { - /* len_end: greatest index forward at which costs have been calculated - * so far */ - unsigned len_end = longest_match_len; + LZX_ASSERT(longest_rep_len >= LZX_MIN_MATCH_LEN); - for (;;) { - /* Advance to next position. */ - cur_pos++; + u32 cost; + + while (end_pos < longest_rep_len) + optimum[++end_pos].cost = MC_INFINITE_COST; - if (cur_pos == len_end || cur_pos == LZX_OPTIM_ARRAY_SIZE) - return lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos); + cost = lzx_repmatch_cost(longest_rep_len, longest_rep_slot, + &c->costs); + if (cost <= optimum[longest_rep_len].cost) { + optimum[longest_rep_len].queue = c->queue; + swap(optimum[longest_rep_len].queue.R[0], + optimum[longest_rep_len].queue.R[longest_rep_slot]); + optimum[longest_rep_len].prev.link = 0; + optimum[longest_rep_len].prev.match_offset = + optimum[longest_rep_len].queue.R[0]; + optimum[longest_rep_len].cost = cost; + } + } - /* retrieve the number of matches available at this position */ - num_possible_matches = lzx_lz_get_matches_caching(ctx, &ctx->optimum[cur_pos].queue, - &possible_matches); + /* Step forward, calculating the estimated minimum cost to reach each + * position. The algorithm may find multiple paths to reach each + * position; only the lowest-cost path is saved. + * + * The progress of the parse is tracked in the @optimum array, which for + * each position contains the minimum cost to reach that position, the + * index of the start of the match/literal taken to reach that position + * through the minimum-cost path, the offset of the match taken (not + * relevant for literals), and the adaptive state that will exist at + * that position after the minimum-cost path is taken. The @cur_pos + * variable stores the position at which the algorithm is currently + * considering coding choices, and the @end_pos variable stores the + * greatest position at which the costs of coding choices have been + * saved. + * + * The loop terminates when any one of the following conditions occurs: + * + * 1. A match with length greater than or equal to @nice_match_length is + * found. When this occurs, the algorithm chooses this match + * unconditionally, and consequently the near-optimal match/literal + * sequence up to and including that match is fully determined and it + * can begin returning the match/literal list. + * + * 2. @cur_pos reaches a position not overlapped by a preceding match. + * In such cases, the near-optimal match/literal sequence up to + * @cur_pos is fully determined and it can begin returning the + * match/literal list. + * + * 3. Failing either of the above in a degenerate case, the loop + * terminates when space in the @optimum array is exhausted. + * This terminates the algorithm and forces it to start returning + * matches/literals even though they may not be globally optimal. + * + * Upon loop termination, a nonempty list of matches/literals will have + * been produced and stored in the @optimum array. These + * matches/literals are linked in reverse order, so the last thing this + * function does is reverse this list and return the first + * match/literal, leaving the rest to be returned immediately by + * subsequent calls to this function. + */ + cur_pos = 0; + for (;;) { + u32 cost; - unsigned new_len = 0; + /* Advance to next position. */ + cur_pos++; - if (num_possible_matches != 0) { - new_len = possible_matches[0].len; + /* Check termination conditions (2) and (3) noted above. */ + if (cur_pos == end_pos || cur_pos == LZX_OPTIM_ARRAY_LENGTH) + return lzx_match_chooser_reverse_list(c, cur_pos); + + /* Search for matches at repeat offsets. Again, as a heuristic + * we only keep the longest one. */ + longest_rep_len = lzx_repsearch(&c->cur_window[c->match_window_pos], + c->match_window_end - c->match_window_pos, + &optimum[cur_pos].queue, + &longest_rep_slot); + + /* If we found a long match at a repeat offset, choose it + * immediately. */ + if (longest_rep_len >= c->params.nice_match_length) { + /* Build the list of matches to return and get + * the first one. */ + match = lzx_match_chooser_reverse_list(c, cur_pos); + + /* Append the long match to the end of the list. */ + optimum[cur_pos].next.match_offset = + optimum[cur_pos].queue.R[longest_rep_slot]; + optimum[cur_pos].next.link = cur_pos + longest_rep_len; + c->optimum_end_idx = cur_pos + longest_rep_len; + + /* Skip over the remaining bytes of the long match. */ + lzx_skip_bytes(c, longest_rep_len); + + /* Return first match in the list. */ + return match; + } - /* Greedy heuristic: if we found a match greater than - * the number of fast bytes, stop immediately. */ - if (new_len > ctx->params.alg_params.slow.num_fast_bytes) { + /* Find other matches. */ + num_matches = lzx_get_matches(c, &matches); + /* If there's a long match, choose it immediately. */ + if (num_matches) { + longest_len = matches[num_matches - 1].len; + if (longest_len >= c->params.nice_match_length) { /* Build the list of matches to return and get * the first one. */ - match = lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos); + match = lzx_match_chooser_reverse_list(c, cur_pos); /* Append the long match to the end of the list. */ - ctx->optimum[cur_pos].next.match_offset = - possible_matches[0].offset; - ctx->optimum[cur_pos].next.link = cur_pos + new_len; - ctx->optimum_end_idx = cur_pos + new_len; + optimum[cur_pos].next.match_offset = + matches[num_matches - 1].offset; + optimum[cur_pos].next.link = cur_pos + longest_len; + c->optimum_end_idx = cur_pos + longest_len; /* Skip over the remaining bytes of the long match. */ - lzx_lz_skip_bytes(ctx, new_len - 1); + lzx_skip_bytes(c, longest_len - 1); - /* Return first match in the list */ + /* Return first match in the list. */ return match; } + } else { + longest_len = 1; } - /* Consider proceeding with a literal byte. */ - block_cost_t cur_cost = ctx->optimum[cur_pos].cost; - block_cost_t 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; - ctx->optimum[cur_pos + 1].queue = ctx->optimum[cur_pos].queue; + /* If we are reaching any positions for the first time, we need + * to initialize their costs to infinity. */ + while (end_pos < cur_pos + longest_len) + optimum[++end_pos].cost = MC_INFINITE_COST; + + /* Consider coding a literal. */ + cost = optimum[cur_pos].cost + + lzx_literal_cost(c->cur_window[c->match_window_pos - 1], + &c->costs); + if (cost < optimum[cur_pos + 1].cost) { + optimum[cur_pos + 1].queue = optimum[cur_pos].queue; + optimum[cur_pos + 1].cost = cost; + optimum[cur_pos + 1].prev.link = cur_pos; } - if (num_possible_matches == 0) - continue; + /* Consider coding a match. + * + * The hard-coded cost calculation is done for the same reason + * stated in the comment for the similar loop earlier. + * Actually, it is *this* one that has the biggest effect on + * performance; overall LZX compression is > 10% faster with + * this code compared to calling lzx_match_cost() with each + * length. */ + for (unsigned i = 0, len = 2; i < num_matches; i++) { + u32 offset; + u32 position_cost; + unsigned position_slot; + unsigned num_extra_bits; + + offset = matches[i].offset; + position_cost = optimum[cur_pos].cost; + + /* Yet another optimization: instead of calling + * lzx_get_position_slot(), hand-inline the search of + * the repeat offset queue. Then we can omit the + * extra_bits calculation for repeat offset matches, and + * also only compute the updated queue if we actually do + * find a new lowest cost path. */ + for (position_slot = 0; position_slot < LZX_NUM_RECENT_OFFSETS; position_slot++) + if (offset == optimum[cur_pos].queue.R[position_slot]) + goto have_position_cost; + + position_slot = lzx_get_position_slot_raw(offset + LZX_OFFSET_OFFSET); + + num_extra_bits = lzx_get_num_extra_bits(position_slot); + if (num_extra_bits >= 3) { + position_cost += num_extra_bits - 3; + position_cost += c->costs.aligned[ + (offset + LZX_OFFSET_OFFSET) & 7]; + } else { + position_cost += num_extra_bits; + } + + have_position_cost: + + do { + u32 cost; + unsigned len_header; + unsigned main_symbol; + + cost = position_cost; + + if (len - LZX_MIN_MATCH_LEN < LZX_NUM_PRIMARY_LENS) { + len_header = len - LZX_MIN_MATCH_LEN; + } else { + len_header = LZX_NUM_PRIMARY_LENS; + cost += c->costs.len[len - + LZX_MIN_MATCH_LEN - + LZX_NUM_PRIMARY_LENS]; + } + + main_symbol = ((position_slot << 3) | len_header) + + LZX_NUM_CHARS; + cost += c->costs.main[main_symbol]; + + if (cost < optimum[cur_pos + len].cost) { + if (position_slot < LZX_NUM_RECENT_OFFSETS) { + optimum[cur_pos + len].queue = optimum[cur_pos].queue; + swap(optimum[cur_pos + len].queue.R[0], + optimum[cur_pos + len].queue.R[position_slot]); + } else { + optimum[cur_pos + len].queue.R[0] = offset; + optimum[cur_pos + len].queue.R[1] = optimum[cur_pos].queue.R[0]; + optimum[cur_pos + len].queue.R[2] = optimum[cur_pos].queue.R[1]; + } + optimum[cur_pos + len].prev.link = cur_pos; + optimum[cur_pos + len].prev.match_offset = offset; + optimum[cur_pos + len].cost = cost; + } + } while (++len <= matches[i].len); + } - /* Consider proceeding with a match. */ - - while (len_end < cur_pos + new_len) - ctx->optimum[++len_end].cost = INFINITE_BLOCK_COST; - - for (unsigned len = LZX_MIN_MATCH_LEN, match_idx = num_possible_matches - 1; - len <= new_len; len++) { - LZX_ASSERT(match_idx < num_possible_matches); - struct lzx_lru_queue q = ctx->optimum[cur_pos].queue; - block_cost_t cost = cur_cost + lzx_match_cost(len, - possible_matches[match_idx].offset, - &ctx->costs, - &q); - - 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; - ctx->optimum[cur_pos + len].queue = q; + /* Consider coding a repeat offset match. + * + * As a heuristic, we only consider the longest length of the + * longest repeat offset match. This does not, however, + * necessarily mean that we will never consider any other repeat + * offsets, because above we detect repeat offset matches that + * were found by the regular match-finder. Therefore, this + * special handling of the longest repeat-offset match is only + * helpful for coding a repeat offset match that was *not* found + * by the match-finder, e.g. due to being obscured by a less + * distant match that is at least as long. + * + * Note: an alternative, used in LZMA, is to consider every + * length of every repeat offset match. This is a more thorough + * search, and it makes it unnecessary to detect repeat offset + * matches that were found by the regular match-finder. But by + * my tests, for LZX the LZMA method slows down the compressor + * by ~10% and doesn't actually help the compression ratio too + * much. + * + * Also tested a compromise approach: consider every 3rd length + * of the longest repeat offset match. Still didn't seem quite + * worth it, though. + */ + if (longest_rep_len) { + + LZX_ASSERT(longest_rep_len >= LZX_MIN_MATCH_LEN); + + while (end_pos < cur_pos + longest_rep_len) + optimum[++end_pos].cost = MC_INFINITE_COST; + + cost = optimum[cur_pos].cost + + lzx_repmatch_cost(longest_rep_len, longest_rep_slot, + &c->costs); + if (cost <= optimum[cur_pos + longest_rep_len].cost) { + optimum[cur_pos + longest_rep_len].queue = + optimum[cur_pos].queue; + swap(optimum[cur_pos + longest_rep_len].queue.R[0], + optimum[cur_pos + longest_rep_len].queue.R[longest_rep_slot]); + optimum[cur_pos + longest_rep_len].prev.link = + cur_pos; + optimum[cur_pos + longest_rep_len].prev.match_offset = + optimum[cur_pos + longest_rep_len].queue.R[0]; + optimum[cur_pos + longest_rep_len].cost = + cost; } + } + } +} - if (len == possible_matches[match_idx].len) - match_idx--; +static struct lz_match +lzx_choose_lazy_item(struct lzx_compressor *c) +{ + const struct lz_match *matches; + struct lz_match cur_match; + struct lz_match next_match; + u32 num_matches; + + if (c->prev_match.len) { + cur_match = c->prev_match; + c->prev_match.len = 0; + } else { + num_matches = lzx_get_matches(c, &matches); + if (num_matches == 0 || + (matches[num_matches - 1].len <= 3 && + (matches[num_matches - 1].len <= 2 || + matches[num_matches - 1].offset > 4096))) + { + return (struct lz_match) { }; } + + cur_match = matches[num_matches - 1]; + } + + if (cur_match.len >= c->params.nice_match_length) { + lzx_skip_bytes(c, cur_match.len - 1); + return cur_match; + } + + num_matches = lzx_get_matches(c, &matches); + if (num_matches == 0 || + (matches[num_matches - 1].len <= 3 && + (matches[num_matches - 1].len <= 2 || + matches[num_matches - 1].offset > 4096))) + { + lzx_skip_bytes(c, cur_match.len - 2); + return cur_match; + } + + next_match = matches[num_matches - 1]; + + if (next_match.len <= cur_match.len) { + lzx_skip_bytes(c, cur_match.len - 2); + return cur_match; + } else { + c->prev_match = next_match; + return (struct lz_match) { }; } } /* - * Set default symbol costs. + * Return the next match or literal to use, delegating to the currently selected + * match-choosing algorithm. + * + * If the length of the returned 'struct lz_match' is less than + * LZX_MIN_MATCH_LEN, then it is really a literal. */ +static inline struct lz_match +lzx_choose_item(struct lzx_compressor *c) +{ + return (*c->params.choose_item_func)(c); +} + +/* Set default symbol costs for the LZX Huffman codes. */ static void lzx_set_default_costs(struct lzx_costs * costs, unsigned num_main_syms) { unsigned i; - /* Literal symbols */ + /* Main code (part 1): Literal symbols */ for (i = 0; i < LZX_NUM_CHARS; i++) costs->main[i] = 8; - /* Match header symbols */ + /* Main code (part 2): Match header symbols */ for (; i < num_main_syms; i++) costs->main[i] = 10; - /* Length symbols */ + /* Length code */ for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) costs->len[i] = 8; - /* Aligned offset symbols */ + /* Aligned offset code */ for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) costs->aligned[i] = 3; } -/* Given the frequencies of symbols in a compressed block and the corresponding - * Huffman codes, return LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM if an - * aligned offset or verbatim block, respectively, will take fewer bits to - * output. */ +/* Given the frequencies of symbols in an LZX-compressed block and the + * corresponding Huffman codes, return LZX_BLOCKTYPE_ALIGNED or + * LZX_BLOCKTYPE_VERBATIM if an aligned offset or verbatim block, respectively, + * will take fewer bits to output. */ static int lzx_choose_verbatim_or_aligned(const struct lzx_freqs * freqs, const struct lzx_codes * codes) @@ -1530,8 +2016,8 @@ lzx_choose_verbatim_or_aligned(const struct lzx_freqs * freqs, /* Verbatim blocks have a constant 3 bits per position footer. Aligned * offset blocks have an aligned offset symbol per position footer, plus - * an extra 24 bits to output the lengths necessary to reconstruct the - * aligned offset code itself. */ + * an extra 24 bits per block to output the lengths necessary to + * reconstruct the aligned offset code itself. */ for (unsigned i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { verbatim_cost += 3 * freqs->aligned[i]; aligned_cost += codes->lens.aligned[i] * freqs->aligned[i]; @@ -1543,506 +2029,366 @@ lzx_choose_verbatim_or_aligned(const struct lzx_freqs * freqs, return LZX_BLOCKTYPE_VERBATIM; } -/* Find a near-optimal sequence of matches/literals with which to output the - * specified LZX block, then set its type to that which has the minimum cost to - * output. */ +/* Find a sequence of matches/literals with which to output the specified LZX + * block, then set the block's type to that which has the minimum cost to output + * (either verbatim or aligned). */ static void -lzx_optimize_block(struct lzx_compressor *ctx, struct lzx_block_spec *spec, - unsigned num_passes) +lzx_choose_items_for_block(struct lzx_compressor *c, struct lzx_block_spec *spec) { - const struct lzx_lru_queue orig_queue = ctx->queue; + const struct lzx_lru_queue orig_queue = c->queue; + u32 num_passes_remaining = c->params.num_optim_passes; struct lzx_freqs freqs; + const u8 *window_ptr; + const u8 *window_end; + struct lzx_item *next_chosen_item; + struct lz_match lz_match; + struct lzx_item lzx_item; - unsigned orig_window_pos = spec->window_pos; - unsigned orig_cached_pos = ctx->cached_matches_pos; - - LZX_ASSERT(ctx->match_window_pos == spec->window_pos); + LZX_ASSERT(num_passes_remaining >= 1); + LZX_ASSERT(lz_mf_get_position(c->mf) == spec->window_pos); - ctx->match_window_end = spec->window_pos + spec->block_size; - spec->chosen_matches_start_pos = spec->window_pos; + c->match_window_end = spec->window_pos + spec->block_size; - LZX_ASSERT(num_passes >= 1); + if (c->params.num_optim_passes > 1) { + if (spec->block_size == c->cur_window_size) + c->get_matches_func = lzx_get_matches_fillcache_singleblock; + else + c->get_matches_func = lzx_get_matches_fillcache_multiblock; + c->skip_bytes_func = lzx_skip_bytes_fillcache; + } else { + if (spec->block_size == c->cur_window_size) + c->get_matches_func = lzx_get_matches_nocache_singleblock; + else + c->get_matches_func = lzx_get_matches_nocache_multiblock; + c->skip_bytes_func = lzx_skip_bytes_nocache; + } /* The first optimal parsing pass is done using the cost model already - * set in ctx->costs. Each later pass is done using a cost model - * computed from the previous pass. */ - for (unsigned pass = 0; pass < num_passes; pass++) { - - ctx->match_window_pos = orig_window_pos; - ctx->cached_matches_pos = orig_cached_pos; - ctx->queue = orig_queue; - spec->num_chosen_matches = 0; + * set in c->costs. Each later pass is done using a cost model + * computed from the previous pass. + * + * To improve performance we only generate the array containing the + * matches and literals in intermediate form on the final pass. */ + + while (--num_passes_remaining) { + c->match_window_pos = spec->window_pos; + c->cache_ptr = c->cached_matches; memset(&freqs, 0, sizeof(freqs)); + window_ptr = &c->cur_window[spec->window_pos]; + window_end = window_ptr + spec->block_size; - for (unsigned i = spec->window_pos; i < spec->window_pos + spec->block_size; ) { - struct raw_match raw_match; - struct lzx_match lzx_match; + while (window_ptr != window_end) { - raw_match = lzx_lz_get_near_optimal_match(ctx); - if (raw_match.len >= LZX_MIN_MATCH_LEN) { - lzx_match.data = lzx_tally_match(raw_match.len, raw_match.offset, - &freqs, &ctx->queue); - i += raw_match.len; + lz_match = lzx_choose_item(c); + + LZX_ASSERT(!(lz_match.len == LZX_MIN_MATCH_LEN && + lz_match.offset == c->max_window_size - + LZX_MIN_MATCH_LEN)); + if (lz_match.len >= LZX_MIN_MATCH_LEN) { + lzx_tally_match(lz_match.len, lz_match.offset, + &freqs, &c->queue); + window_ptr += lz_match.len; } else { - lzx_match.data = lzx_tally_literal(ctx->window[i], &freqs); - i += 1; + lzx_tally_literal(*window_ptr, &freqs); + window_ptr += 1; } - ctx->chosen_matches[spec->chosen_matches_start_pos + - spec->num_chosen_matches++] = lzx_match; } - - lzx_make_huffman_codes(&freqs, &spec->codes, - ctx->num_main_syms); - if (pass < num_passes - 1) - lzx_set_costs(ctx, &spec->codes.lens); - ctx->matches_cached = true; + lzx_make_huffman_codes(&freqs, &spec->codes, c->num_main_syms); + lzx_set_costs(c, &spec->codes.lens, 15); + c->queue = orig_queue; + if (c->cache_ptr <= c->cache_limit) { + c->get_matches_func = lzx_get_matches_usecache_nocheck; + c->skip_bytes_func = lzx_skip_bytes_usecache_nocheck; + } else { + c->get_matches_func = lzx_get_matches_usecache; + c->skip_bytes_func = lzx_skip_bytes_usecache; + } } - spec->block_type = lzx_choose_verbatim_or_aligned(&freqs, &spec->codes); - ctx->matches_cached = false; -} -static void -lzx_optimize_blocks(struct lzx_compressor *ctx) -{ - lzx_lru_queue_init(&ctx->queue); - ctx->optimum_cur_idx = 0; - ctx->optimum_end_idx = 0; + c->match_window_pos = spec->window_pos; + c->cache_ptr = c->cached_matches; + memset(&freqs, 0, sizeof(freqs)); + window_ptr = &c->cur_window[spec->window_pos]; + window_end = window_ptr + spec->block_size; + + spec->chosen_items = &c->chosen_items[spec->window_pos]; + next_chosen_item = spec->chosen_items; - const unsigned num_passes = ctx->params.alg_params.slow.num_optim_passes; + unsigned unseen_cost = 9; + while (window_ptr != window_end) { - for (unsigned i = 0; i < ctx->num_blocks; i++) - lzx_optimize_block(ctx, &ctx->block_specs[i], num_passes); + lz_match = lzx_choose_item(c); + + LZX_ASSERT(!(lz_match.len == LZX_MIN_MATCH_LEN && + lz_match.offset == c->max_window_size - + LZX_MIN_MATCH_LEN)); + if (lz_match.len >= LZX_MIN_MATCH_LEN) { + lzx_item.data = lzx_tally_match(lz_match.len, + lz_match.offset, + &freqs, &c->queue); + window_ptr += lz_match.len; + } else { + lzx_item.data = lzx_tally_literal(*window_ptr, &freqs); + window_ptr += 1; + } + *next_chosen_item++ = lzx_item; + + /* When doing one-pass "near-optimal" parsing, update the cost + * model occassionally. */ + if (unlikely((next_chosen_item - spec->chosen_items) % 2048 == 0) && + c->params.choose_item_func == lzx_choose_near_optimal_item && + c->params.num_optim_passes == 1) + { + lzx_make_huffman_codes(&freqs, &spec->codes, c->num_main_syms); + lzx_set_costs(c, &spec->codes.lens, unseen_cost); + if (unseen_cost < 15) + unseen_cost++; + } + } + spec->num_chosen_items = next_chosen_item - spec->chosen_items; + lzx_make_huffman_codes(&freqs, &spec->codes, c->num_main_syms); + spec->block_type = lzx_choose_verbatim_or_aligned(&freqs, &spec->codes); } /* Prepare the input window into one or more LZX blocks ready to be output. */ static void -lzx_prepare_blocks(struct lzx_compressor * ctx) +lzx_prepare_blocks(struct lzx_compressor *c) { - /* Initialize the match-finder. */ - lz_sarray_load_window(&ctx->lz_sarray, ctx->window, ctx->window_size); - ctx->cached_matches_pos = 0; - ctx->matches_cached = false; - ctx->match_window_pos = 0; - /* Set up a default cost model. */ - lzx_set_default_costs(&ctx->costs, ctx->num_main_syms); - - ctx->num_blocks = DIV_ROUND_UP(ctx->window_size, LZX_DIV_BLOCK_SIZE); - for (unsigned i = 0; i < ctx->num_blocks; i++) { - unsigned pos = LZX_DIV_BLOCK_SIZE * i; - ctx->block_specs[i].window_pos = pos; - ctx->block_specs[i].block_size = min(ctx->window_size - pos, LZX_DIV_BLOCK_SIZE); + if (c->params.choose_item_func == lzx_choose_near_optimal_item) + lzx_set_default_costs(&c->costs, c->num_main_syms); + + /* Set up the block specifications. + * TODO: The compression ratio could be slightly improved by performing + * data-dependent block splitting instead of using fixed-size blocks. + * Doing so well is a computationally hard problem, however. */ + c->num_blocks = DIV_ROUND_UP(c->cur_window_size, LZX_DIV_BLOCK_SIZE); + for (unsigned i = 0; i < c->num_blocks; i++) { + u32 pos = LZX_DIV_BLOCK_SIZE * i; + c->block_specs[i].window_pos = pos; + c->block_specs[i].block_size = min(c->cur_window_size - pos, + LZX_DIV_BLOCK_SIZE); } + /* Load the window into the match-finder. */ + lz_mf_load_window(c->mf, c->cur_window, c->cur_window_size); + /* Determine sequence of matches/literals to output for each block. */ - lzx_optimize_blocks(ctx); + lzx_lru_queue_init(&c->queue); + c->optimum_cur_idx = 0; + c->optimum_end_idx = 0; + c->prev_match.len = 0; + for (unsigned i = 0; i < c->num_blocks; i++) + lzx_choose_items_for_block(c, &c->block_specs[i]); } -/* - * This is the fast version of lzx_prepare_blocks(). This version "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 - * - * Output --- the block specification and the corresponding match/literal data: - * - * ctx->block_specs[] - * ctx->num_blocks - * ctx->chosen_matches[] - */ static void -lzx_prepare_block_fast(struct lzx_compressor * ctx) +lzx_build_params(unsigned int compression_level, + u32 max_window_size, + struct lzx_compressor_params *lzx_params) { - struct lzx_record_ctx record_ctx; - struct lzx_block_spec *spec; - - /* Parameters to hash chain LZ match finder - * (lazy with 1 match lookahead) */ - static const struct lz_params lzx_lz_params = { - /* Although LZX_MIN_MATCH_LEN == 2, length 2 matches typically - * aren't worth choosing when using greedy or lazy parsing. */ - .min_match = 3, - .max_match = LZX_MAX_MATCH_LEN, - .max_offset = LZX_MAX_WINDOW_SIZE, - .good_match = LZX_MAX_MATCH_LEN, - .nice_match = LZX_MAX_MATCH_LEN, - .max_chain_len = LZX_MAX_MATCH_LEN, - .max_lazy_match = LZX_MAX_MATCH_LEN, - .too_far = 4096, - }; - - /* Initialize symbol frequencies and match offset LRU queue. */ - memset(&record_ctx.freqs, 0, sizeof(struct lzx_freqs)); - lzx_lru_queue_init(&record_ctx.queue); - record_ctx.matches = ctx->chosen_matches; - - /* Determine series of matches/literals to output. */ - lz_analyze_block(ctx->window, - ctx->window_size, - lzx_record_match, - lzx_record_literal, - &record_ctx, - &lzx_lz_params, - ctx->prev_tab); - - /* Set up block specification. */ - spec = &ctx->block_specs[0]; - spec->block_type = LZX_BLOCKTYPE_ALIGNED; - spec->window_pos = 0; - spec->block_size = ctx->window_size; - spec->num_chosen_matches = (record_ctx.matches - ctx->chosen_matches); - spec->chosen_matches_start_pos = 0; - lzx_make_huffman_codes(&record_ctx.freqs, &spec->codes, - ctx->num_main_syms); - ctx->num_blocks = 1; + if (compression_level < 25) { + lzx_params->choose_item_func = lzx_choose_lazy_item; + lzx_params->num_optim_passes = 1; + if (max_window_size <= 262144) + lzx_params->mf_algo = LZ_MF_HASH_CHAINS; + else + lzx_params->mf_algo = LZ_MF_BINARY_TREES; + lzx_params->min_match_length = 3; + lzx_params->nice_match_length = 25 + compression_level * 2; + lzx_params->max_search_depth = 25 + compression_level; + } else { + lzx_params->choose_item_func = lzx_choose_near_optimal_item; + lzx_params->num_optim_passes = compression_level / 20; + if (max_window_size <= 32768 && lzx_params->num_optim_passes == 1) + lzx_params->mf_algo = LZ_MF_HASH_CHAINS; + else + lzx_params->mf_algo = LZ_MF_BINARY_TREES; + lzx_params->min_match_length = (compression_level >= 45) ? 2 : 3; + lzx_params->nice_match_length = min(((u64)compression_level * 32) / 50, + LZX_MAX_MATCH_LEN); + lzx_params->max_search_depth = min(((u64)compression_level * 50) / 50, + LZX_MAX_MATCH_LEN); + } } static void -do_call_insn_translation(u32 *call_insn_target, int input_pos, - s32 file_size) +lzx_build_mf_params(const struct lzx_compressor_params *lzx_params, + u32 max_window_size, struct lz_mf_params *mf_params) { - 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); - } + memset(mf_params, 0, sizeof(*mf_params)); + + mf_params->algorithm = lzx_params->mf_algo; + mf_params->max_window_size = max_window_size; + mf_params->min_match_len = lzx_params->min_match_length; + mf_params->max_match_len = LZX_MAX_MATCH_LEN; + mf_params->max_search_depth = lzx_params->max_search_depth; + mf_params->nice_match_len = lzx_params->nice_match_length; } -/* 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; - } - } -} +lzx_free_compressor(void *_c); -static size_t -lzx_compress(const void *uncompressed_data, size_t uncompressed_size, - void *compressed_data, size_t compressed_size_avail, void *_ctx) +static u64 +lzx_get_needed_memory(size_t max_block_size, unsigned int compression_level) { - struct lzx_compressor *ctx = _ctx; - struct output_bitstream ostream; - size_t compressed_size; - - if (uncompressed_size < 100) { - LZX_DEBUG("Too small to bother compressing."); - return 0; - } - - if (uncompressed_size > ctx->max_window_size) { - LZX_DEBUG("Can't compress %zu bytes using window of %u bytes!", - uncompressed_size, ctx->max_window_size); - return 0; - } - - LZX_DEBUG("Attempting to compress %zu bytes...", - uncompressed_size); - - /* The input data must be preprocessed. To avoid changing the original - * input, copy it to a temporary buffer. */ - memcpy(ctx->window, uncompressed_data, uncompressed_size); - ctx->window_size = uncompressed_size; - - /* 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); + struct lzx_compressor_params params; + u64 size = 0; + unsigned window_order; + u32 max_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, compressed_size_avail); - lzx_write_all_blocks(ctx, &ostream); - - LZX_DEBUG("Flushing bitstream..."); - compressed_size = flush_output_bitstream(&ostream); - if (compressed_size == ~(input_idx_t)0) { - LZX_DEBUG("Data did not compress to %zu bytes or less!", - compressed_size_avail); + window_order = lzx_get_window_order(max_block_size); + if (window_order == 0) return 0; - } + max_window_size = max_block_size; - LZX_DEBUG("Done: compressed %zu => %zu bytes.", - uncompressed_size, compressed_size); - - /* Verify that we really get the same thing back when decompressing. - * Although this could be disabled by default in all cases, it only - * takes around 2-3% of the running time of the slow algorithm to do the - * verification. */ - if (ctx->params.algorithm == WIMLIB_LZX_ALGORITHM_SLOW - #if defined(ENABLE_LZX_DEBUG) || defined(ENABLE_VERIFY_COMPRESSION) - || 1 - #endif - ) - { - struct wimlib_decompressor *decompressor; + lzx_build_params(compression_level, max_window_size, ¶ms); - if (0 == wimlib_create_decompressor(WIMLIB_COMPRESSION_TYPE_LZX, - ctx->max_window_size, - NULL, - &decompressor)) - { - int ret; - ret = wimlib_decompress(compressed_data, - compressed_size, - ctx->window, - uncompressed_size, - decompressor); - wimlib_free_decompressor(decompressor); - - if (ret) { - ERROR("Failed to decompress data we " - "compressed using LZX algorithm"); - wimlib_assert(0); - return 0; - } - if (memcmp(uncompressed_data, ctx->window, uncompressed_size)) { - ERROR("Data we compressed using LZX algorithm " - "didn't decompress to original"); - wimlib_assert(0); - return 0; - } - } else { - WARNING("Failed to create decompressor for " - "data verification!"); - } - } - return compressed_size; -} + size += sizeof(struct lzx_compressor); -static bool -lzx_params_valid(const struct wimlib_lzx_compressor_params *params) -{ - /* Validate parameters. */ - if (params->hdr.size != sizeof(struct wimlib_lzx_compressor_params)) { - LZX_DEBUG("Invalid parameter structure size!"); - return false; - } + size += max_window_size; - if (params->algorithm != WIMLIB_LZX_ALGORITHM_SLOW && - params->algorithm != WIMLIB_LZX_ALGORITHM_FAST) - { - LZX_DEBUG("Invalid algorithm."); - return false; - } + size += DIV_ROUND_UP(max_window_size, LZX_DIV_BLOCK_SIZE) * + sizeof(struct lzx_block_spec); - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - if (params->alg_params.slow.num_optim_passes < 1) - { - LZX_DEBUG("Invalid number of optimization passes!"); - return false; - } - - if (params->alg_params.slow.main_nostat_cost < 1 || - params->alg_params.slow.main_nostat_cost > 16) - { - LZX_DEBUG("Invalid main_nostat_cost!"); - return false; - } - - if (params->alg_params.slow.len_nostat_cost < 1 || - params->alg_params.slow.len_nostat_cost > 16) - { - LZX_DEBUG("Invalid len_nostat_cost!"); - return false; - } - - if (params->alg_params.slow.aligned_nostat_cost < 1 || - params->alg_params.slow.aligned_nostat_cost > 8) - { - LZX_DEBUG("Invalid aligned_nostat_cost!"); - return false; - } - } - return true; -} + size += max_window_size * sizeof(struct lzx_item); -static void -lzx_free_compressor(void *_ctx) -{ - struct lzx_compressor *ctx = _ctx; - - if (ctx) { - FREE(ctx->chosen_matches); - FREE(ctx->cached_matches); - FREE(ctx->optimum); - lz_sarray_destroy(&ctx->lz_sarray); - FREE(ctx->block_specs); - FREE(ctx->prev_tab); - FREE(ctx->window); - FREE(ctx); + size += lz_mf_get_needed_memory(params.mf_algo, max_window_size); + if (params.choose_item_func == lzx_choose_near_optimal_item) { + size += (LZX_OPTIM_ARRAY_LENGTH + params.nice_match_length) * + sizeof(struct lzx_mc_pos_data); } + if (params.num_optim_passes > 1) + size += LZX_CACHE_LEN * sizeof(struct lz_match); + else + size += LZX_MAX_MATCHES_PER_POS * sizeof(struct lz_match); + return size; } static int -lzx_create_compressor(size_t window_size, - const struct wimlib_compressor_params_header *_params, - void **ctx_ret) +lzx_create_compressor(size_t max_block_size, unsigned int compression_level, + void **c_ret) { - const struct wimlib_lzx_compressor_params *params = - (const struct wimlib_lzx_compressor_params*)_params; - struct lzx_compressor *ctx; - - LZX_DEBUG("Allocating LZX context..."); - - if (!lzx_window_size_valid(window_size)) + struct lzx_compressor *c; + struct lzx_compressor_params params; + struct lz_mf_params mf_params; + unsigned window_order; + u32 max_window_size; + + window_order = lzx_get_window_order(max_block_size); + if (window_order == 0) return WIMLIB_ERR_INVALID_PARAM; + max_window_size = max_block_size; - static const struct wimlib_lzx_compressor_params fast_default = { - .hdr = { - .size = sizeof(struct wimlib_lzx_compressor_params), - }, - .algorithm = WIMLIB_LZX_ALGORITHM_FAST, - .use_defaults = 0, - .alg_params = { - .fast = { - }, - }, - }; - static const struct wimlib_lzx_compressor_params slow_default = { - .hdr = { - .size = sizeof(struct wimlib_lzx_compressor_params), - }, - .algorithm = WIMLIB_LZX_ALGORITHM_SLOW, - .use_defaults = 0, - .alg_params = { - .slow = { - .use_len2_matches = 1, - .num_fast_bytes = 32, - .num_optim_passes = 2, - .max_search_depth = 50, - .max_matches_per_pos = 3, - .main_nostat_cost = 15, - .len_nostat_cost = 15, - .aligned_nostat_cost = 7, - }, - }, - }; - - if (params) { - if (!lzx_params_valid(params)) - return WIMLIB_ERR_INVALID_PARAM; - } else { - LZX_DEBUG("Using default algorithm and parameters."); - params = &slow_default; - } + lzx_build_params(compression_level, max_window_size, ¶ms); + lzx_build_mf_params(¶ms, max_window_size, &mf_params); + if (!lz_mf_params_valid(&mf_params)) + return WIMLIB_ERR_INVALID_PARAM; - if (params->use_defaults) { - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) - params = &slow_default; - else - params = &fast_default; - } + c = CALLOC(1, sizeof(struct lzx_compressor)); + if (!c) + goto oom; - LZX_DEBUG("Allocating memory."); + c->params = params; + c->num_main_syms = lzx_get_num_main_syms(window_order); + c->max_window_size = max_window_size; + c->window_order = window_order; - ctx = CALLOC(1, sizeof(struct lzx_compressor)); - if (ctx == NULL) + c->cur_window = ALIGNED_MALLOC(max_window_size, 16); + if (!c->cur_window) goto oom; - ctx->num_main_syms = lzx_get_num_main_syms(window_size); - ctx->max_window_size = window_size; - ctx->window = MALLOC(window_size + 12); - if (ctx->window == NULL) + c->block_specs = MALLOC(DIV_ROUND_UP(max_window_size, + LZX_DIV_BLOCK_SIZE) * + sizeof(struct lzx_block_spec)); + if (!c->block_specs) goto oom; - if (params->algorithm == WIMLIB_LZX_ALGORITHM_FAST) { - ctx->prev_tab = MALLOC(window_size * sizeof(ctx->prev_tab[0])); - if (ctx->prev_tab == NULL) - goto oom; - } + c->chosen_items = MALLOC(max_window_size * sizeof(struct lzx_item)); + if (!c->chosen_items) + goto oom; - size_t block_specs_length = DIV_ROUND_UP(window_size, LZX_DIV_BLOCK_SIZE); - ctx->block_specs = MALLOC(block_specs_length * sizeof(ctx->block_specs[0])); - if (ctx->block_specs == NULL) + c->mf = lz_mf_alloc(&mf_params); + if (!c->mf) goto oom; - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - unsigned min_match_len = LZX_MIN_MATCH_LEN; - if (!params->alg_params.slow.use_len2_matches) - min_match_len = max(min_match_len, 3); - - if (!lz_sarray_init(&ctx->lz_sarray, - window_size, - min_match_len, - LZX_MAX_MATCH_LEN, - params->alg_params.slow.max_search_depth, - params->alg_params.slow.max_matches_per_pos)) + if (params.choose_item_func == lzx_choose_near_optimal_item) { + c->optimum = MALLOC((LZX_OPTIM_ARRAY_LENGTH + + params.nice_match_length) * + sizeof(struct lzx_mc_pos_data)); + if (!c->optimum) goto oom; } - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - ctx->optimum = MALLOC((LZX_OPTIM_ARRAY_SIZE + LZX_MAX_MATCH_LEN) * - sizeof(ctx->optimum[0])); - if (ctx->optimum == NULL) + if (params.num_optim_passes > 1) { + c->cached_matches = MALLOC(LZX_CACHE_LEN * + sizeof(struct lz_match)); + if (!c->cached_matches) + goto oom; + c->cache_limit = c->cached_matches + LZX_CACHE_LEN - + (LZX_MAX_MATCHES_PER_POS + 1); + } else { + c->cached_matches = MALLOC(LZX_MAX_MATCHES_PER_POS * + sizeof(struct lz_match)); + if (!c->cached_matches) goto oom; } - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - u32 cache_per_pos; + *c_ret = c; + return 0; + +oom: + lzx_free_compressor(c); + return WIMLIB_ERR_NOMEM; +} - cache_per_pos = params->alg_params.slow.max_matches_per_pos; - if (cache_per_pos > LZX_MAX_CACHE_PER_POS) - cache_per_pos = LZX_MAX_CACHE_PER_POS; +static size_t +lzx_compress(const void *uncompressed_data, size_t uncompressed_size, + void *compressed_data, size_t compressed_size_avail, void *_c) +{ + struct lzx_compressor *c = _c; + struct lzx_output_bitstream os; - ctx->cached_matches = MALLOC(window_size * (cache_per_pos + 1) * - sizeof(ctx->cached_matches[0])); - if (ctx->cached_matches == NULL) - goto oom; - } + /* Don't bother compressing very small inputs. */ + if (uncompressed_size < 100) + return 0; - ctx->chosen_matches = MALLOC(window_size * sizeof(ctx->chosen_matches[0])); - if (ctx->chosen_matches == NULL) - goto oom; + /* The input data must be preprocessed. To avoid changing the original + * input, copy it to a temporary buffer. */ + memcpy(c->cur_window, uncompressed_data, uncompressed_size); + c->cur_window_size = uncompressed_size; - memcpy(&ctx->params, params, sizeof(struct wimlib_lzx_compressor_params)); - memset(&ctx->zero_codes, 0, sizeof(ctx->zero_codes)); + /* Preprocess the data. */ + lzx_do_e8_preprocessing(c->cur_window, c->cur_window_size); - LZX_DEBUG("Successfully allocated new LZX context."); + /* Prepare the compressed data. */ + lzx_prepare_blocks(c); - *ctx_ret = ctx; - return 0; + /* Generate the compressed data and return its size, or 0 if an overflow + * occurred. */ + lzx_init_output(&os, compressed_data, compressed_size_avail); + lzx_write_all_blocks(c, &os); + return lzx_flush_output(&os); +} -oom: - lzx_free_compressor(ctx); - return WIMLIB_ERR_NOMEM; +static void +lzx_free_compressor(void *_c) +{ + struct lzx_compressor *c = _c; + + if (c) { + ALIGNED_FREE(c->cur_window); + FREE(c->block_specs); + FREE(c->chosen_items); + lz_mf_free(c->mf); + FREE(c->optimum); + FREE(c->cached_matches); + FREE(c); + } } const struct compressor_ops lzx_compressor_ops = { + .get_needed_memory = lzx_get_needed_memory, .create_compressor = lzx_create_compressor, .compress = lzx_compress, .free_compressor = lzx_free_compressor,