* is a container format from which the locations and sizes (both compressed and
* uncompressed) of the constituent blocks can be determined.
*
- * A LZMS-compressed block must be read in 16-bit little endian units from both
+ * An LZMS-compressed block must be read in 16-bit little endian units from both
* directions. One logical bitstream starts at the front of the block and
* proceeds forwards. Another logical bitstream starts at the end of the block
* and proceeds backwards. Bits read from the forwards bitstream constitute
* offset is 1, regardless of match length.
*
* For LZ matches, up to 3 repeat offsets are allowed, similar to some other
- * LZ-based formats such as LZX and LZMA. They must updated in a LRU fashion,
+ * LZ-based formats such as LZX and LZMA. They must updated in an LRU fashion,
* except for a quirk: updates to the queue must be delayed by one LZMS item,
* except for the removal of a repeat match. As a result, 4 entries are
* actually needed in the queue, even though it is only possible to decode
* than 15 bits, then all frequencies must be divided by 2, rounding up, and the
* code construction must be attempted again.
*
- * A LZMS-compressed block seemingly cannot have a compressed size greater than
+ * An LZMS-compressed block seemingly cannot have a compressed size greater than
* or equal to the uncompressed size. In such cases the block must be stored
* uncompressed.
*
#include "wimlib/compress_common.h"
#include "wimlib/decompressor_ops.h"
#include "wimlib/decompress_common.h"
-#include "wimlib/error.h"
#include "wimlib/lzms.h"
#include "wimlib/util.h"
* at a time, this needs to be 64 bits rather than 32 bits. */
u64 bitbuf;
- /* Number of bits in @bitbuf that are are used. */
+ /* Number of bits in @bitbuf that are used. */
unsigned num_filled_bits;
/* Pointer to the one past the next little-endian 16-bit integer in the
* read using this decoder. */
const u32 *slot_base_tab;
+ const u8 *extra_bits_tab;
+
/* Number of symbols that have been read using this code far. Reset to
* 0 whenever the code is rebuilt. */
u32 num_syms_read;
u8 lens[LZMS_MAX_NUM_SYMS];
/* The codeword of each symbol in the Huffman code. */
- u16 codewords[LZMS_MAX_NUM_SYMS];
+ u32 codewords[LZMS_MAX_NUM_SYMS];
/* A table for quickly decoding symbols encoded using the Huffman code.
*/
struct lzms_huffman_decoder delta_power_decoder;
struct lzms_huffman_decoder delta_offset_decoder;
- /* LRU (least-recently-used) queues for match information. */
- struct lzms_lru_queues lru;
+ /* LRU (least-recently-used) queues for match information. */
+ struct lzms_lru_queues lru;
/* Used for postprocessing. */
s32 last_target_usages[65536];
static u32
lzms_huffman_decode_symbol(struct lzms_huffman_decoder *dec)
{
- const u8 *lens = dec->lens;
const u16 *decode_table = dec->decode_table;
struct lzms_input_bitstream *is = dec->is;
+ u16 entry;
+ u16 key_bits;
+ u16 sym;
/* The Huffman codes used in LZMS are adaptive and must be rebuilt
* whenever a certain number of symbols have been read. Each such
dec->num_syms_read = 0;
}
- /* In the following Huffman decoding implementation, the first
- * LZMS_DECODE_TABLE_BITS of the input are used as an offset into a
- * decode table. The entry will either provide the decoded symbol
- * directly, or else a "real" Huffman binary tree will be searched to
- * decode the symbol. */
-
+ /* XXX: Copied from read_huffsym() (decompress_common.h), since this
+ * uses a different input bitstream type. Should unify the
+ * implementations. */
lzms_input_bitstream_ensure_bits(is, LZMS_MAX_CODEWORD_LEN);
- u16 key_bits = lzms_input_bitstream_peek_bits(is, LZMS_DECODE_TABLE_BITS);
- u16 sym = decode_table[key_bits];
-
- if (sym < dec->num_syms) {
- /* Fast case: The decode table directly provided the symbol. */
- lzms_input_bitstream_remove_bits(is, lens[sym]);
+ /* Index the decode table by the next table_bits bits of the input. */
+ key_bits = lzms_input_bitstream_peek_bits(is, LZMS_DECODE_TABLE_BITS);
+ entry = decode_table[key_bits];
+ if (likely(entry < 0xC000)) {
+ /* Fast case: The decode table directly provided the symbol and
+ * codeword length. The low 11 bits are the symbol, and the
+ * high 5 bits are the codeword length. */
+ lzms_input_bitstream_remove_bits(is, entry >> 11);
+ sym = entry & 0x7FF;
} else {
- /* Slow case: The symbol took too many bits to include directly
- * in the decode table, so search for it in a binary tree at the
- * end of the decode table. */
+ /* Slow case: The codeword for the symbol is longer than
+ * table_bits, so the symbol does not have an entry directly in
+ * the first (1 << table_bits) entries of the decode table.
+ * Traverse the appropriate binary tree bit-by-bit in order to
+ * decode the symbol. */
lzms_input_bitstream_remove_bits(is, LZMS_DECODE_TABLE_BITS);
do {
- key_bits = sym + lzms_input_bitstream_pop_bits(is, 1);
- } while ((sym = decode_table[key_bits]) >= dec->num_syms);
+ key_bits = (entry & 0x3FFF) + lzms_input_bitstream_pop_bits(is, 1);
+ } while ((entry = decode_table[key_bits]) >= 0xC000);
+ sym = entry;
}
/* Tally and return the decoded symbol. */
unsigned num_extra_bits;
u32 extra_bits;
+ LZMS_ASSERT(dec->slot_base_tab != NULL);
+ LZMS_ASSERT(dec->extra_bits_tab != NULL);
+
/* Read the slot (position slot, length slot, etc.), which is encoded as
* a Huffman symbol. */
slot = lzms_huffman_decode_symbol(dec);
- LZMS_ASSERT(dec->slot_base_tab != NULL);
-
/* Get the number of extra bits needed to represent the range of values
* that share the slot. */
- num_extra_bits = bsr32(dec->slot_base_tab[slot + 1] -
- dec->slot_base_tab[slot]);
+ num_extra_bits = dec->extra_bits_tab[slot];
- /* Read the number of extra bits and add them to the slot to form the
- * final decoded value. */
+ /* Read the number of extra bits and add them to the slot base to form
+ * the final decoded value. */
extra_bits = lzms_input_bitstream_read_bits(dec->is, num_extra_bits);
return dec->slot_base_tab[slot] + extra_bits;
}
lzms_copy_lz_match(struct lzms_decompressor *ctx, u32 length, u32 offset)
{
u8 *out_next;
- u8 *matchptr;
if (length > ctx->out_end - ctx->out_next) {
LZMS_DEBUG("Match overrun!");
}
out_next = ctx->out_next;
- matchptr = out_next - offset;
- while (length--)
- *out_next++ = *matchptr++;
- ctx->out_next = out_next;
+ lz_copy(out_next, length, offset, ctx->out_end);
+ ctx->out_next = out_next + length;
+
return 0;
}
return lzms_copy_delta_match(ctx, length, power, raw_offset);
}
+/* Decode an LZ or delta match. */
static int
lzms_decode_match(struct lzms_decompressor *ctx)
{
if (ret)
return ret;
- /* Update LRU queues */
- lzms_update_lru_queues(&ctx->lru);
+ lzms_update_lru_queues(&ctx->lru);
return 0;
}
static void
lzms_init_huffman_decoder(struct lzms_huffman_decoder *dec,
struct lzms_input_bitstream *is,
- const u32 *slot_base_tab, unsigned num_syms,
+ const u32 *slot_base_tab,
+ const u8 *extra_bits_tab,
+ unsigned num_syms,
unsigned rebuild_freq)
{
dec->is = is;
dec->slot_base_tab = slot_base_tab;
+ dec->extra_bits_tab = extra_bits_tab;
dec->num_syms = num_syms;
dec->num_syms_read = rebuild_freq;
dec->rebuild_freq = rebuild_freq;
* backwards) */
lzms_input_bitstream_init(&ctx->is, cdata, clen / 2);
- /* Initialize position and length slot bases if not done already. */
- lzms_init_slot_bases();
-
/* Calculate the number of position slots needed for this compressed
* block. */
num_position_slots = lzms_get_position_slot(ulen - 1) + 1;
/* Initialize Huffman decoders for each alphabet used in the compressed
* representation. */
lzms_init_huffman_decoder(&ctx->literal_decoder, &ctx->is,
- NULL, LZMS_NUM_LITERAL_SYMS,
+ NULL, NULL, LZMS_NUM_LITERAL_SYMS,
LZMS_LITERAL_CODE_REBUILD_FREQ);
lzms_init_huffman_decoder(&ctx->lz_offset_decoder, &ctx->is,
- lzms_position_slot_base, num_position_slots,
+ lzms_position_slot_base,
+ lzms_extra_position_bits,
+ num_position_slots,
LZMS_LZ_OFFSET_CODE_REBUILD_FREQ);
lzms_init_huffman_decoder(&ctx->length_decoder, &ctx->is,
- lzms_length_slot_base, LZMS_NUM_LEN_SYMS,
+ lzms_length_slot_base,
+ lzms_extra_length_bits,
+ LZMS_NUM_LEN_SYMS,
LZMS_LENGTH_CODE_REBUILD_FREQ);
lzms_init_huffman_decoder(&ctx->delta_offset_decoder, &ctx->is,
- lzms_position_slot_base, num_position_slots,
+ lzms_position_slot_base,
+ lzms_extra_position_bits,
+ num_position_slots,
LZMS_DELTA_OFFSET_CODE_REBUILD_FREQ);
lzms_init_huffman_decoder(&ctx->delta_power_decoder, &ctx->is,
- NULL, LZMS_NUM_DELTA_POWER_SYMS,
+ NULL, NULL, LZMS_NUM_DELTA_POWER_SYMS,
LZMS_DELTA_POWER_CODE_REBUILD_FREQ);
&ctx->rd, LZMS_NUM_DELTA_REPEAT_MATCH_STATES);
/* Initialize LRU match information. */
- lzms_init_lru_queues(&ctx->lru);
+ lzms_init_lru_queues(&ctx->lru);
LZMS_DEBUG("Decompressor successfully initialized");
}
return -1;
}
- /* A LZMS-compressed data block should be evenly divisible into 16-bit
+ /* An LZMS-compressed data block should be evenly divisible into 16-bit
* integers. */
if (compressed_size % 2 != 0) {
LZMS_DEBUG("Compressed size not divisible by 2 (got %zu)",
{
struct lzms_decompressor *ctx = _ctx;
- FREE(ctx);
+ ALIGNED_FREE(ctx);
}
static int
{
struct lzms_decompressor *ctx;
- ctx = MALLOC(sizeof(struct lzms_decompressor));
+ ctx = ALIGNED_MALLOC(sizeof(struct lzms_decompressor),
+ DECODE_TABLE_ALIGNMENT);
if (ctx == NULL)
return WIMLIB_ERR_NOMEM;
+ /* Initialize position and length slot data if not done already. */
+ lzms_init_slots();
+
*ctx_ret = ctx;
return 0;
}