mirror of https://github.com/tongzx/nt5src
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
422 lines
11 KiB
422 lines
11 KiB
/*
|
|
* encdata.c
|
|
*
|
|
* Encode a block into the output stream
|
|
*/
|
|
|
|
#include "encoder.h"
|
|
|
|
#define OUT_CHAR \
|
|
c = context->enc_LitData[l]; \
|
|
OUTPUT_BITS(context->enc_main_tree_len[c], context->enc_main_tree_code[c]);
|
|
|
|
|
|
/*
|
|
* Macro to output bits into the encoding stream
|
|
*/
|
|
#define OUTPUT_BITS(N,X) \
|
|
{ \
|
|
context->enc_bitbuf |= (((ulong) (X)) << (context->enc_bitcount-(N))); \
|
|
context->enc_bitcount -= (N); \
|
|
while (context->enc_bitcount <= 16) \
|
|
{ \
|
|
if (context->enc_output_buffer_curpos >= context->enc_output_buffer_end) \
|
|
{ \
|
|
context->enc_output_overflow = true; \
|
|
context->enc_output_buffer_curpos = context->enc_output_buffer_start; \
|
|
} \
|
|
*context->enc_output_buffer_curpos++ = (byte) ((context->enc_bitbuf >> 16) & 255); \
|
|
*context->enc_output_buffer_curpos++ = (byte) (context->enc_bitbuf >> 24); \
|
|
context->enc_bitbuf <<= 16; \
|
|
context->enc_bitcount += 16; \
|
|
} \
|
|
}
|
|
|
|
|
|
/*
|
|
* Given the initial state of the repeated offset buffers at
|
|
* the beginning of this block, calculate the final state of the
|
|
* repeated offset buffers after outputting this block as if it
|
|
* were compressed data.
|
|
*
|
|
* First try to do it the quick way, by starting at the last
|
|
* match and working backwards, to find three consecutive matches
|
|
* which don't use repeated offsets. If this fails, we'll have
|
|
* to take the initial state of the three offsets at the beginning
|
|
* of the block, and evolve them to the end of the block.
|
|
*/
|
|
void get_final_repeated_offset_states(t_encoder_context *context, ulong distances)
|
|
{
|
|
ulong MatchPos;
|
|
signed long d; /* must be signed */
|
|
byte consecutive;
|
|
|
|
consecutive = 0;
|
|
|
|
for (d = distances-1; d >= 0; d--)
|
|
{
|
|
if (context->enc_DistData[d] > 2)
|
|
{
|
|
/* NOT a repeated offset */
|
|
consecutive++;
|
|
|
|
/* do we have three consecutive non-repeated-offsets? */
|
|
if (consecutive >= 3)
|
|
break;
|
|
}
|
|
else
|
|
{
|
|
consecutive = 0;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* If we didn't find three consecutive matches which
|
|
* don't use repeated offsets, then we have to start
|
|
* from the beginning and evolve the repeated offsets.
|
|
*
|
|
* Otherwise, we start at the first of the consecutive
|
|
* matches.
|
|
*/
|
|
if (consecutive < 3)
|
|
{
|
|
d = 0;
|
|
}
|
|
|
|
for (; d < (signed long) distances; d++)
|
|
{
|
|
MatchPos = context->enc_DistData[d];
|
|
|
|
if (MatchPos == 0)
|
|
{
|
|
}
|
|
else if (MatchPos <= 2)
|
|
{
|
|
ulong temp;
|
|
|
|
temp = context->enc_repeated_offset_at_literal_zero[MatchPos];
|
|
context->enc_repeated_offset_at_literal_zero[MatchPos] = context->enc_repeated_offset_at_literal_zero[0];
|
|
context->enc_repeated_offset_at_literal_zero[0] = temp;
|
|
}
|
|
else
|
|
{
|
|
context->enc_repeated_offset_at_literal_zero[2] = context->enc_repeated_offset_at_literal_zero[1];
|
|
context->enc_repeated_offset_at_literal_zero[1] = context->enc_repeated_offset_at_literal_zero[0];
|
|
context->enc_repeated_offset_at_literal_zero[0] = MatchPos-2;
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Encode a block with no compression
|
|
*
|
|
* bufpos is the position in the file from which the first
|
|
* literal in this block starts. To reference memory, we will
|
|
* use enc_MemWindow[bufpos] (remember that enc_MemWindow is
|
|
* moved backwards every time we copymem).
|
|
*
|
|
* Since this data was originally matched into the compressor,
|
|
* our recent match offsets will have been changed; however,
|
|
* since this is an uncompressed block, the decoder won't be
|
|
* updating them. Therefore, we need to tell the decoder
|
|
* the state of the match offsets after it has finished
|
|
* decoding the uncompressed data - we store these in this
|
|
* block.
|
|
*/
|
|
void encode_uncompressed_block(t_encoder_context *context, ulong bufpos, ulong block_size)
|
|
{
|
|
int i;
|
|
int j;
|
|
bool block_size_odd;
|
|
ulong val;
|
|
|
|
/*
|
|
* Align on a byte boundary
|
|
*/
|
|
output_bits(context, context->enc_bitcount-16, 0);
|
|
|
|
/*
|
|
* Now output the contents of the repeated offset
|
|
* buffers, since we need to preserve the state of
|
|
* the encoder
|
|
*/
|
|
for (i = 0; i < NUM_REPEATED_OFFSETS; i++)
|
|
{
|
|
val = context->enc_repeated_offset_at_literal_zero[i];
|
|
|
|
for (j = 0; j < sizeof(long); j++)
|
|
{
|
|
*context->enc_output_buffer_curpos++ = (byte) val;
|
|
val >>= 8;
|
|
}
|
|
}
|
|
|
|
block_size_odd = block_size & 1;
|
|
|
|
/*
|
|
* Write out uncompressed data
|
|
*/
|
|
while (block_size > 0)
|
|
{
|
|
*context->enc_output_buffer_curpos++ = context->enc_MemWindow[bufpos];
|
|
|
|
bufpos++;
|
|
block_size--;
|
|
context->enc_input_running_total++;
|
|
|
|
if (context->enc_input_running_total == CHUNK_SIZE)
|
|
{
|
|
perform_flush_output_callback(context);
|
|
context->enc_num_block_splits = 0;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Add pad byte to keep the output word-aligned
|
|
*/
|
|
if (block_size_odd)
|
|
{
|
|
*context->enc_output_buffer_curpos++ = 0;
|
|
}
|
|
|
|
context->enc_bitcount = 32;
|
|
context->enc_bitbuf = 0;
|
|
}
|
|
|
|
|
|
/*
|
|
* Estimate the size of the data in the buffer, in bytes
|
|
*/
|
|
ulong estimate_compressed_block_size(t_encoder_context *context)
|
|
{
|
|
ulong block_size = 0; /* output size in bits */
|
|
ulong i;
|
|
byte mpslot;
|
|
|
|
/* Estimation of tree size */
|
|
block_size = 150*8;
|
|
|
|
/* Tally bits to output characters */
|
|
for (i = 0; i < NUM_CHARS; i++)
|
|
block_size += (context->enc_main_tree_len[i]*context->enc_main_tree_freq[i]);
|
|
|
|
/* Tally bits to output matches */
|
|
for (mpslot = 0; mpslot < context->enc_num_position_slots; mpslot++)
|
|
{
|
|
long element;
|
|
int primary;
|
|
|
|
element = NUM_CHARS + (mpslot << NL_SHIFT);
|
|
|
|
/* For primary == NUM_PRIMARY_LENGTHS we have secondary lengths */
|
|
for (primary = 0; primary <= NUM_PRIMARY_LENGTHS; primary++)
|
|
{
|
|
block_size += ((context->enc_main_tree_len[element] + enc_extra_bits[mpslot]) *
|
|
context->enc_main_tree_freq[element]);
|
|
element++;
|
|
}
|
|
}
|
|
|
|
for (i = 0; i < NUM_SECONDARY_LENGTHS; i++)
|
|
block_size += (context->enc_secondary_tree_freq[i] * context->enc_secondary_tree_len[i]);
|
|
|
|
/* round up */
|
|
return (block_size+7) >> 3;
|
|
}
|
|
|
|
/*
|
|
* Encode block with NO special encoding of the lower 3
|
|
* position bits
|
|
*/
|
|
void encode_verbatim_block(t_encoder_context *context, ulong literal_to_end_at)
|
|
{
|
|
ulong MatchPos;
|
|
ulong d = 0;
|
|
ulong l = 0;
|
|
byte MatchLength;
|
|
byte c;
|
|
byte mpSlot;
|
|
|
|
while (l < literal_to_end_at)
|
|
{
|
|
if (!IsMatch(l))
|
|
{
|
|
OUT_CHAR;
|
|
l++;
|
|
context->enc_input_running_total++;
|
|
}
|
|
else
|
|
{
|
|
/* Note, 0 means MatchLen=3, 1 means MatchLen=4, ... */
|
|
MatchLength = context->enc_LitData[l++];
|
|
|
|
/* Delta match pos */
|
|
MatchPos = context->enc_DistData[d++];
|
|
|
|
mpSlot = (byte) MP_SLOT(MatchPos);
|
|
|
|
if (MatchLength < NUM_PRIMARY_LENGTHS)
|
|
{
|
|
OUTPUT_BITS(
|
|
context->enc_main_tree_len[ NUM_CHARS+(mpSlot<<NL_SHIFT)+MatchLength],
|
|
context->enc_main_tree_code[NUM_CHARS+(mpSlot<<NL_SHIFT)+MatchLength]
|
|
);
|
|
}
|
|
else
|
|
{
|
|
OUTPUT_BITS(
|
|
context->enc_main_tree_len [(NUM_CHARS+NUM_PRIMARY_LENGTHS)+(mpSlot<<NL_SHIFT)],
|
|
context->enc_main_tree_code[(NUM_CHARS+NUM_PRIMARY_LENGTHS)+(mpSlot<<NL_SHIFT)]
|
|
);
|
|
|
|
OUTPUT_BITS(
|
|
context->enc_secondary_tree_len[ MatchLength - NUM_PRIMARY_LENGTHS],
|
|
context->enc_secondary_tree_code[MatchLength - NUM_PRIMARY_LENGTHS]
|
|
);
|
|
}
|
|
|
|
if (enc_extra_bits[ mpSlot ])
|
|
{
|
|
OUTPUT_BITS(
|
|
enc_extra_bits[mpSlot],
|
|
MatchPos & enc_slot_mask[mpSlot]
|
|
);
|
|
}
|
|
|
|
context->enc_input_running_total += (MatchLength+MIN_MATCH);
|
|
}
|
|
|
|
if (context->enc_input_running_total == CHUNK_SIZE)
|
|
{
|
|
perform_flush_output_callback(context);
|
|
context->enc_num_block_splits = 0;
|
|
}
|
|
|
|
_ASSERTE (context->enc_input_running_total < CHUNK_SIZE);
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* aligned block encoding
|
|
*/
|
|
void encode_aligned_block(t_encoder_context *context, ulong literal_to_end_at)
|
|
{
|
|
ulong MatchPos;
|
|
byte MatchLength;
|
|
byte c;
|
|
byte mpSlot;
|
|
byte Lower;
|
|
ulong l = 0;
|
|
ulong d = 0;
|
|
|
|
while (l < literal_to_end_at)
|
|
{
|
|
if (!IsMatch(l))
|
|
{
|
|
OUT_CHAR;
|
|
l++;
|
|
context->enc_input_running_total++;
|
|
}
|
|
else
|
|
{
|
|
/* Note, 0 means MatchLen=3, 1 means MatchLen=4, ... */
|
|
MatchLength = context->enc_LitData[l++];
|
|
|
|
/* Delta match pos */
|
|
MatchPos = context->enc_DistData[d++];
|
|
|
|
mpSlot = (byte) MP_SLOT(MatchPos);
|
|
|
|
if (MatchLength < NUM_PRIMARY_LENGTHS)
|
|
{
|
|
OUTPUT_BITS(
|
|
context->enc_main_tree_len[ NUM_CHARS+(mpSlot<<NL_SHIFT)+MatchLength],
|
|
context->enc_main_tree_code[NUM_CHARS+(mpSlot<<NL_SHIFT)+MatchLength]
|
|
);
|
|
}
|
|
else
|
|
{
|
|
OUTPUT_BITS(
|
|
context->enc_main_tree_len[ (NUM_CHARS+NUM_PRIMARY_LENGTHS)+(mpSlot<<NL_SHIFT)],
|
|
context->enc_main_tree_code[(NUM_CHARS+NUM_PRIMARY_LENGTHS)+(mpSlot<<NL_SHIFT)]
|
|
);
|
|
|
|
OUTPUT_BITS(
|
|
context->enc_secondary_tree_len[ MatchLength - NUM_PRIMARY_LENGTHS],
|
|
context->enc_secondary_tree_code[MatchLength - NUM_PRIMARY_LENGTHS]
|
|
);
|
|
}
|
|
|
|
if (enc_extra_bits[ mpSlot ] >= 3)
|
|
{
|
|
if (enc_extra_bits[ mpSlot ] > 3)
|
|
{
|
|
OUTPUT_BITS(
|
|
enc_extra_bits[mpSlot] - 3,
|
|
(MatchPos >> 3) & ( (1 << (enc_extra_bits[mpSlot]-3)) -1)
|
|
);
|
|
}
|
|
|
|
Lower = (byte) (MatchPos & 7);
|
|
|
|
OUTPUT_BITS(
|
|
context->enc_aligned_tree_len[Lower],
|
|
context->enc_aligned_tree_code[Lower]
|
|
);
|
|
}
|
|
else if (enc_extra_bits[ mpSlot ])
|
|
{
|
|
OUTPUT_BITS(
|
|
enc_extra_bits[mpSlot],
|
|
MatchPos & enc_slot_mask[ mpSlot ]
|
|
);
|
|
}
|
|
|
|
context->enc_input_running_total += (MatchLength+MIN_MATCH);
|
|
}
|
|
|
|
if (context->enc_input_running_total == CHUNK_SIZE)
|
|
{
|
|
perform_flush_output_callback(context);
|
|
context->enc_num_block_splits = 0;
|
|
}
|
|
|
|
_ASSERTE (context->enc_input_running_total < CHUNK_SIZE);
|
|
}
|
|
}
|
|
|
|
|
|
void perform_flush_output_callback(t_encoder_context *context)
|
|
{
|
|
long output_size;
|
|
|
|
/*
|
|
* Do this only if there is any input to account for, so we don't
|
|
* end up outputting blocks where comp_size > 0 and uncmp_size = 0.
|
|
*/
|
|
if (context->enc_input_running_total > 0)
|
|
{
|
|
flush_output_bit_buffer(context);
|
|
|
|
output_size = (long)(context->enc_output_buffer_curpos - context->enc_output_buffer_start);
|
|
|
|
if (output_size > 0)
|
|
{
|
|
(*context->enc_output_callback_function)(
|
|
context->enc_fci_data,
|
|
context->enc_output_buffer_start,
|
|
(long) (context->enc_output_buffer_curpos - context->enc_output_buffer_start),
|
|
context->enc_input_running_total
|
|
);
|
|
}
|
|
}
|
|
|
|
context->enc_input_running_total = 0;
|
|
context->enc_output_buffer_curpos = context->enc_output_buffer_start;
|
|
|
|
/* initialise bit buffer */
|
|
context->enc_bitcount = 32;
|
|
context->enc_bitbuf = 0;
|
|
}
|