rich/MeshCore
1
0 Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fit Nordic Full KISS with lossless assets and firmware lifetime trimming

Browse Source
This commit is contained in:
mikecarper committed 2026-10-26 18:21:18 -07:00
1 parent 2369a3d28c
commit b4bec19e88
51 files changed
+7739 -17
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

.github/workflows/run-unit-tests.yml
+3 -1
View File
@@ -584,6 +584,7 @@ jobs:
python3 -B test/test_host_kiss_usb.py -v
python3 -B test/test_host_kiss_ingress.py -v
python3 -B test/test_host_kiss_capabilities.py -v
python3 -B test/test_host_kiss_ram.py -v
python3 -B test/test_ethernet_kiss.py -v
python3 -B test/test_ethernet_radio.py -v
python3 -B test/test_companion_ethernet.py -v
@@ -1008,7 +1009,7 @@ jobs:
working-directory: test
run: >-
python3 -B -m unittest
test_message_navigation test_companion_reader test_reader_touch_coordinates
test_message_navigation test_companion_reader test_reader_stream test_small_font_metrics test_arm_instruction_meter test_reader_touch_coordinates
test_pocket_gui test_pocket_world_map test_pocket_eink_refresh test_pocket_panel_spi
test_touch_debug_overlay test_touch_debug_cli test_color_theme
test_indicator_messages_profile test_t096_message_footer_profile
@@ -1132,6 +1133,7 @@ jobs:
- name: Verify nRF52 infrastructure trims and unchanged Companion scope
run: |
python3 -B test/test_nrf52_flash_trim.py -v
python3 -B test/test_nrf52_program_lifetime.py -v
python3 -B test/test_nrf52_tft_font_trim.py -v
python3 -B test/test_ota_sd_cache.py -v
python3 -B test/test_rak15002_sd_seeder.py -v
scripts/arm_instruction_meter.c
+27
View File
@@ -0,0 +1,27 @@
/* Host-only instruction measurements for the pinned Unicorn 2.1.4 checker.
* Match the Python callback exactly: read SP before every ARM instruction. */
#include <stdint.h>
#include <stddef.h>
typedef int (*mesh_reg_read)(void *, int, void *);
struct mesh_arm_measurement {
mesh_reg_read read_register;
int sp_register, error;
uint64_t minimum_sp, instructions;
};
size_t mesh_arm_measurement_size(void) { return sizeof(struct mesh_arm_measurement); }
void mesh_arm_measure_instruction(void *engine, uint64_t address,
uint32_t size, void *user)
{
struct mesh_arm_measurement *state = (struct mesh_arm_measurement *) user;
uint32_t sp = 0;
int error;
(void) address;
(void) size;
error = state->read_register(engine, state->sp_register, &sp);
if (error) state->error = error;
if (sp < state->minimum_sp) state->minimum_sp = sp;
++state->instructions;
}
scripts/arm_instruction_meter.py
+93
View File
@@ -0,0 +1,93 @@
"""Measure every ARM instruction with a C callback, or the identical Python one.
The optional native callback avoids billions of crossings into Python while
keeping instruction/stack coverage. It uses the reviewed private handle API of
pinned Unicorn 2.1.4. Other versions or hosts without a C compiler use Python.
No Unicorn headers, linker development package, or network access are needed.
"""
import ctypes
from functools import lru_cache
from pathlib import Path
import shutil
import subprocess
import sys
import tempfile
import unicorn
from unicorn import UC_HOOK_CODE
from unicorn import arm_const as arm
class _NativeState(ctypes.Structure):
_fields_ = [('read_register', ctypes.c_void_p), ('sp_register', ctypes.c_int),
('error', ctypes.c_int), ('minimum_sp', ctypes.c_uint64),
('instructions', ctypes.c_uint64)]
@lru_cache(maxsize=1)
def _native_library():
compiler = shutil.which('cc') or shutil.which('gcc') or shutil.which('clang')
if not compiler:
raise RuntimeError('no host C compiler; use the Python instruction callback')
directory = tempfile.TemporaryDirectory(prefix='mesh-arm-instruction-meter-')
library = Path(directory.name) / 'meter.so'
result = subprocess.run([compiler, '-std=c11', '-O2', '-shared', '-fPIC',
str(Path(__file__).with_suffix('.c')), '-o', str(library)],
capture_output=True, text=True)
if result.returncode:
raise RuntimeError('native callback compilation failed: ' + result.stderr)
loaded = ctypes.CDLL(str(library))
loaded.mesh_arm_measurement_size.restype = ctypes.c_size_t
if loaded.mesh_arm_measurement_size() != ctypes.sizeof(_NativeState):
raise RuntimeError('native/Python measurement layout mismatch')
return loaded, directory
class ArmInstructionMeter:
def __init__(self, machine, prefer_native=True):
self.backend = 'python'
self._state = {'minimum_sp': 0, 'instructions': 0}
self._native = None
if prefer_native:
try:
self._install_native(machine)
self.backend = 'native'
return
except (RuntimeError, OSError, AttributeError) as error:
print('ARM meter Python fallback: ' + str(error), file=sys.stderr)
self.handle = machine.hook_add(UC_HOOK_CODE, self._measure_python)
def _install_native(self, machine):
if unicorn.__version__ != '2.1.4':
raise RuntimeError('native callback requires reviewed Unicorn 2.1.4 bindings')
from unicorn.unicorn_py3 import unicorn as bindings
library, directory = _native_library()
state = _NativeState(ctypes.cast(bindings.uclib.uc_reg_read, ctypes.c_void_p),
arm.UC_ARM_REG_SP, 0, 0, 0)
callback = bindings.HOOK_CODE_CFUNC(('mesh_arm_measure_instruction', library))
handle = bindings.uc_hook_h()
result = bindings.uclib.uc_hook_add(machine._uch, ctypes.byref(handle), UC_HOOK_CODE,
callback, ctypes.byref(state), ctypes.c_uint64(1), ctypes.c_uint64(0))
if result:
raise RuntimeError('native Unicorn hook registration failed: ' + str(result))
# Keep the C callback, its context, loaded code and temporary file alive
# for at least the lifetime of the registered machine hook.
machine._callbacks[handle.value] = (callback, state, library, directory)
self._native, self.handle = state, handle.value
def _measure_python(self, machine, address, size, data):
self._state['minimum_sp'] = min(self._state['minimum_sp'], machine.reg_read(arm.UC_ARM_REG_SP))
self._state['instructions'] += 1
def reset(self, top):
if self._native is not None:
self._native.minimum_sp, self._native.instructions, self._native.error = top, 0, 0
else:
self._state.update(minimum_sp=top, instructions=0)
def snapshot(self):
if self._native is not None:
if self._native.error:
raise RuntimeError('Unicorn SP read failed: ' + str(self._native.error))
return {'minimum_sp': self._native.minimum_sp,
'instructions': self._native.instructions}
return dict(self._state)
scripts/check_nrf52_companion_reader.py
+13 -9
View File
@@ -1,7 +1,8 @@
#!/usr/bin/env python3
"""Check canonical Reader replies and stack using the linked RAK Full ARM ELF.
Requires unicorn==2.1.4. Every first page and all pages of the three longest
Requires unicorn==2.1.4. A host C compiler accelerates the identical per-
instruction measurement callback; otherwise Python is used. Every first page and all pages of the three longest
verses are compared with the independently ASCII-normalized pinned source.
--every-page reconstructs every verse. Only replyVerse and its callee frames
are measured; the fixed caller reserve is not a physical high-water reading
@@ -14,7 +15,8 @@ from pathlib import Path
import re
import sys
from firmware_elf import FirmwareElf
from unicorn import Uc, UC_ARCH_ARM, UC_MODE_THUMB, UC_MODE_MCLASS, UC_HOOK_CODE
from unicorn import Uc, UC_ARCH_ARM, UC_MODE_THUMB, UC_MODE_MCLASS
from arm_instruction_meter import ArmInstructionMeter
from unicorn import arm_const as arm
ROOT = Path(__file__).resolve().parents[1]
@@ -41,11 +43,7 @@ def check(path, every_page=False):
uc.mem_write(section[3], elf.section_bytes(section))
stop, top, reply = 0x1000, 0x2003e000, 0x20030000
uc.mem_write(stop, b'\0\xbe')
state = {'minimum_sp': top, 'instructions': 0}
def instruction(machine, address, size, data):
state['minimum_sp'] = min(state['minimum_sp'], machine.reg_read(arm.UC_ARM_REG_SP))
state['instructions'] += 1
uc.hook_add(UC_HOOK_CODE, instruction)
meter = ArmInstructionMeter(uc)
document = json.loads((ROOT / 'tools/bible/reader-web.json').read_text(encoding='utf-8'))
counts = (51, 25, 36, 54, 47, 71, 53, 59, 41, 42, 57,
50, 38, 31, 27, 33, 26, 40, 42, 31, 25)
@@ -59,6 +57,7 @@ def check(path, every_page=False):
reverse=True)[:3])
maximum, calls, total_instructions, checked_bytes = 0, 0, 0, 0
max_ref = None
maximum_instructions, instruction_ref, instruction_page = 0, None, None
guard = b'\x91\xa2\xb3\xc4'
for index, ref in enumerate(references):
raw = document['verses'][ref]
@@ -72,7 +71,7 @@ def check(path, every_page=False):
for register, value in zip((arm.UC_ARM_REG_R0, arm.UC_ARM_REG_R1, arm.UC_ARM_REG_R2, arm.UC_ARM_REG_R3), (selection, page, reply, capacity)):
uc.reg_write(register, value)
uc.mem_write(reply-4, guard + b'?'*capacity + guard)
state['minimum_sp'], state['instructions'] = top, 0
meter.reset(top)
uc.emu_start(elf.address(name) | 1, stop, timeout=10000000, count=10000000)
if uc.reg_read(arm.UC_ARM_REG_PC) != stop or uc.reg_read(arm.UC_ARM_REG_SP) != top:
raise ValueError('Function return/stack restoration failed for ' + ref)
@@ -92,6 +91,7 @@ def check(path, every_page=False):
raise ValueError('Canonical linked text mismatch for ' + ref + ' page ' + str(page))
reconstructed += fragment
checked_bytes += len(fragment)
state = meter.snapshot()
used = top - state['minimum_sp']
if used + 1536 > 8192:
raise ValueError('Reader frames exceed 8 KiB with caller reserve: ' + str(used))
@@ -99,6 +99,8 @@ def check(path, every_page=False):
maximum, max_ref = used, ref
calls += 1
total_instructions += state['instructions']
if state['instructions'] > maximum_instructions:
maximum_instructions, instruction_ref, instruction_page = state['instructions'], ref, page
if next_command is None:
if reconstructed != expected:
raise ValueError('Canonical text incomplete for ' + ref)
@@ -120,7 +122,9 @@ def check(path, every_page=False):
'canonical_source_sha256': source_digest,
'maximum_reply_verse_stack_bytes': maximum, 'maximum_stack_reference': max_ref,
'caller_reserve_bytes': 1536, 'loop_stack_bytes': 8192,
'total_arm_instructions': total_instructions,
'total_arm_instructions': total_instructions, 'measurement_backend': meter.backend,
'maximum_reply_arm_instructions': maximum_instructions,
'maximum_instruction_reference': instruction_ref, 'maximum_instruction_page': instruction_page,
'scope': 'Actual linked replyVerse instructions and callees; excludes firmware caller/ISR/task and physical high-water.'}
scripts/nrf52_flash_trim.py
+15
View File
@@ -48,6 +48,11 @@ TRIM_DEFINES = {
# Keep MeshCore's own startup logo; only omit the invisible library bitmap.
"SSD1306_NO_SPLASH": 1,
}
# The reviewed main runs the Arduino loop forever. Scheduler failure and every
# normal reset/update path reset the MCU; global process-exit cleanup cannot run.
# Unknown/custom core revisions retain the standard ABI registration behavior.
NRF52_MAIN_NO_EXIT_SHA256 = "25c3a05b2631a6daff58048a517b8f86c33e97aa978d8b365f03a5b4fa371231"
CC310_ARCHIVE_SHA256 = "97dc648d44520e47252f62c402a71f976e36d7b3aef99235bbdc4de79a928577"
CC310_ARCHIVE = "src/cortex-m4/fpv4-sp-d16-hard/libnrf_cc310_0.9.13-no-interrupts.a"
@@ -149,6 +154,13 @@ def supported_cc310(framework_dir, mcu, defines):
return archive.is_file() and hashlib.sha256(archive.read_bytes()).hexdigest() == CC310_ARCHIVE_SHA256
def supported_program_lifetime(framework_dir):
if not framework_dir:
return False
main = Path(framework_dir) / "cores/nRF5/main.cpp"
return main.is_file() and hashlib.sha256(main.read_bytes()).hexdigest() == NRF52_MAIN_NO_EXIT_SHA256
def companion_trim_policy(platform, mcu, src_filter, flags=(), cppdefines=()):
"""Qualify the actual Companion sources, independent of board/alias name."""
if platform != "nordicnrf52" or not str(mcu).lower().startswith("nrf52"):
@@ -237,6 +249,9 @@ def install(build_env):
print("nRF52 Companion flash trim: small crypto table, unused USB device classes/GPS cache/vendor splash; fonts and Reader retained")
framework = platform.get_package_dir("framework-arduinoadafruitnrf52")
values = flag_defines(flags, cppdefines)
if "MESH_NRF52_NO_PROGRAM_EXIT" not in values and supported_program_lifetime(framework):
build_env.AppendUnique(CPPDEFINES=[("MESH_NRF52_NO_PROGRAM_EXIT", 1)])
print("nRF52 reviewed firmware lifetime: omit unreachable process-exit registrations")
if infrastructure and values.get("ST7789") is not None:
# Keep the font license with this qualified display build, including
# cached DFU packages. The hook independently checks the effective
src/Nrf52ProgramLifetime.cpp
+15
View File
@@ -0,0 +1,15 @@
#if defined(NRF52_PLATFORM) && defined(MESH_NRF52_NO_PROGRAM_EXIT) && MESH_NRF52_NO_PROGRAM_EXIT
// The qualified Adafruit nRF52 Arduino main never exits: the scheduler runs
// forever and its only return path invokes NVIC_SystemReset(). Device reboot
// and update paths also reset the MCU rather than terminating a C++ process.
// Do not retain registrations for global process-exit cleanup which cannot
// run. Automatic RAII destructors, explicit close/shutdown and dynamic delete
// remain ordinary C++ operations and are unaffected by this ABI entry point.
// nrf52_flash_trim.py enables this only for the exact reviewed core main hash.
extern "C" __attribute__((visibility("hidden"))) int __aeabi_atexit(
void*, void (*)(void*), void*) {
return 0;
}
#endif
src/helpers/CompanionReader.cpp
+4
View File
@@ -4,7 +4,11 @@
#include <Arduino.h>
#include "bible/ReaderLookup.h"
#if defined(NRF52_PLATFORM)
#include "bible/ReaderStreamData.generated.h"
#else
#include "bible/ReaderData.generated.h"
#endif
namespace mesh {
bible::LookupResult readReaderVerse(bible::Reference ref, char* scratch,
src/helpers/bible/ReaderStreamData.generated.h
+1461
View File
File diff suppressed because it is too large. Load diff
src/helpers/bible/ReaderStreamLookup.h
+126
View File
@@ -0,0 +1,126 @@
#pragma once
#include "ReaderLookup.h"
namespace mesh { namespace bible {
// Independent bounded-history streams trade cold Reader decode time for flash.
// The existing ESP32 block corpus and its decoder remain unchanged.
struct StreamTokenCorpus {
const char* translation;
const char* attribution;
const uint8_t* data;
uint32_t data_size;
uint32_t encoded_size;
const uint8_t* words;
uint16_t words_size;
uint16_t word_plain_size;
uint16_t word_count;
uint16_t verse_count;
uint16_t chunk_count; // Zero retains the single-stream fixture format.
};
inline LookupResult lookup(const StreamTokenCorpus& corpus, Reference ref,
char* scratch, size_t capacity, const char*& text) {
text = nullptr;
uint16_t index;
if (!verseNumber(ref, index)) return LookupResult::Invalid;
if (!scratch || capacity < kBlockSize || !corpus.data || !corpus.data_size
|| !corpus.encoded_size || corpus.verse_count != kVerseCount
|| corpus.chunk_count > kVerseCount
|| corpus.word_count > 128 || corpus.word_plain_size > kBlockSize
|| (corpus.word_count && (!corpus.words || !corpus.words_size
|| corpus.word_plain_size < corpus.word_count * 2U))
|| (!corpus.word_count && (corpus.words_size || corpus.word_plain_size)))
return LookupResult::Corrupt;
const uint8_t* source = corpus.data;
uint32_t source_size = corpus.data_size, output_limit = corpus.encoded_size;
if (corpus.chunk_count) {
// Each independently compressed stream has a bounded six-byte header:
// first verse, decoded byte limit and compressed byte length, little-endian.
const uint8_t* cursor = corpus.data;
uint32_t remaining = corpus.data_size, total = 0;
const uint16_t wanted = index;
uint16_t previous = 0;
for (uint16_t n = 0; n < corpus.chunk_count; ++n) {
if (remaining < 6) return LookupResult::Corrupt;
const uint16_t first = cursor[0] | (uint16_t(cursor[1]) << 8);
const uint16_t plain = cursor[2] | (uint16_t(cursor[3]) << 8);
const uint16_t packed = cursor[4] | (uint16_t(cursor[5]) << 8);
cursor += 6;
remaining -= 6;
if (!plain || !packed || packed > remaining || first >= kVerseCount
|| (n ? first <= previous : first != 0)) return LookupResult::Corrupt;
if (first <= wanted) {
source = cursor;
source_size = packed;
output_limit = plain;
index = wanted - first;
}
previous = first;
total += plain;
cursor += packed;
remaining -= packed;
}
if (remaining || total != corpus.encoded_size) return LookupResult::Corrupt;
}
unsigned int encoded_size = 0;
if (tinf_uncompress_record(scratch, kBlockSize, &encoded_size,
source, source_size, output_limit, index) != TINF_OK
|| encoded_size >= kBlockSize) return LookupResult::Corrupt;
if (!encoded_size) return LookupResult::Missing;
bool tokenized = false;
for (unsigned int i = 0; i < encoded_size; ++i) {
const uint8_t byte = static_cast<uint8_t>(scratch[i]);
if (byte >= 128) {
if (byte - 128U >= corpus.word_count) return LookupResult::Corrupt;
tokenized = true;
} else if (byte < 32 || byte >= 127) return LookupResult::Corrupt;
}
if (!tokenized) { text = scratch; return LookupResult::Found; }
// Reuse the history after inflation. Keep the selected encoded record at
// the far end, the decoded vocabulary at the beginning, and expand into the
// gap. The host packer uses literal fallback for a long verse that would
// not fit these three regions, preserving the original 4095-byte limit.
uint8_t* encoded = reinterpret_cast<uint8_t*>(scratch) + kBlockSize - encoded_size - 1;
if (encoded <= reinterpret_cast<uint8_t*>(scratch) + corpus.word_plain_size)
return LookupResult::Corrupt;
memmove(encoded, scratch, encoded_size + 1);
unsigned int dictionary_size = corpus.word_plain_size;
if (tinf_uncompress_exact(scratch, &dictionary_size, corpus.words, corpus.words_size) != TINF_OK
|| dictionary_size != corpus.word_plain_size) return LookupResult::Corrupt;
const char* dictionary_end = scratch + dictionary_size;
const char* check = scratch;
for (unsigned int n = 0; n < corpus.word_count; ++n) {
const char* zero = static_cast<const char*>(memchr(check, 0, dictionary_end - check));
if (!zero || zero == check) return LookupResult::Corrupt;
for (const char* p = check; p != zero; ++p)
if (*p < 32 || static_cast<uint8_t>(*p) >= 127) return LookupResult::Corrupt;
check = zero + 1;
}
if (check != dictionary_end) return LookupResult::Corrupt;
char* dest = scratch + dictionary_size;
const char* const output = dest;
const char* const end = reinterpret_cast<char*>(encoded) - 1;
while (*encoded) {
const uint8_t byte = *encoded++;
const uint8_t* word = &byte;
size_t length = 1;
if (byte >= 128) {
const unsigned int token = byte - 128;
const char* value = scratch;
for (unsigned int n = 0; n < token; ++n) value += strlen(value) + 1;
word = reinterpret_cast<const uint8_t*>(value);
length = strlen(value);
}
if (length > static_cast<size_t>(end - dest)) return LookupResult::Corrupt;
memcpy(dest, word, length);
dest += length;
}
*dest = 0;
text = output;
return LookupResult::Found;
}
}} // namespace mesh::bible
src/helpers/ota/OtaTinf.c
+7
View File
@@ -11,8 +11,15 @@
/* nRF52's Arduino recipe defaults to -Ofast, which more than doubles this
* decoder. Keep the size policy local to tinf instead of changing MeshCore. */
#pragma GCC push_options
#if defined(NRF52_PLATFORM)
#pragma GCC optimize ("Oz")
#else
#pragma GCC optimize ("Os")
#endif
#endif
#if (defined(NRF52_PLATFORM) && COMPANION_FEATURE_READER) || defined(TINF_READER_STREAM_TEST)
#define MESHCORE_TINF_RECORD 1
#endif
#define MESHCORE_TINF_IMPLEMENTATION 1
#include "tinf/tinflate.c"
#if MESHCORE_OTA_DEVICE_DEFLATE
src/helpers/ota/tinf/README.meshcore.txt
+8
View File
@@ -22,3 +22,11 @@ with neither feature links no decoder code or persistent RAM.
Only the raw RFC 1951 decoder is integrated. The zlib/gzip wrappers and checksum
sources are intentionally omitted. The original zlib license is preserved in
LICENSE and in the source headers.
The nRF52 Full Companion Reader also enables reader_record.inc. This MeshCore
extension shares the same bounded bit reader and Huffman trees, keeps only
4096 bytes of circular history in the caller's existing Reader scratch, and
stops at the requested NUL-delimited record. It rejects distances beyond that
history, records exceeding the original 4095-byte limit, truncated inputs and
output beyond the generated corpus limit. No heap or retained decode cache is
used. The ordinary linear decoder and ESP32 Reader remain available unchanged.
src/helpers/ota/tinf/reader_record.inc
+137
View File
@@ -0,0 +1,137 @@
/* MeshCore extension (2026): read a NUL-delimited record through a caller-owned
* 4 KiB DEFLATE history. It shares tinf's bounded bit reader and Huffman trees.
* No heap, permanent cache, or second output/history buffer is used. */
struct tinf_record_output {
unsigned char *window;
unsigned int size, produced, limit, remaining, length;
};
static int tinf_record_emit(struct tinf_record_output *out, unsigned char byte)
{
if (out->produced >= out->limit) return TINF_DATA_ERROR;
out->window[out->produced++ & (out->size - 1)] = byte;
if (byte == 0) {
if (!out->remaining) return 1;
--out->remaining;
out->length = 0;
} else if (++out->length >= out->size) {
return TINF_BUF_ERROR;
}
return TINF_OK;
}
static int tinf_record_block(struct tinf_data *d, struct tinf_record_output *out)
{
static const unsigned char length_bits[29] = {
0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0
};
static const unsigned short length_base[29] = {
3,4,5,6,7,8,9,10,11,13,15,17,19,23,27,31,35,43,51,59,67,83,99,115,131,163,195,227,258
};
static const unsigned char dist_bits[30] = {
0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13
};
static const unsigned short dist_base[30] = {
1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,
1025,1537,2049,3073,4097,6145,8193,12289,16385,24577
};
for (;;) {
int sym = tinf_decode_symbol(d, &d->ltree);
if (d->overflow) return TINF_DATA_ERROR;
if (sym == 256) return TINF_OK;
if (sym < 256) {
const int result = tinf_record_emit(out, (unsigned char) sym);
if (result) return result;
} else {
int length, distance, offset;
if (sym > d->ltree.max_sym || sym - 257 > 28 || d->dtree.max_sym == -1)
return TINF_DATA_ERROR;
sym -= 257;
length = tinf_getbits_base(d, length_bits[sym], length_base[sym]);
distance = tinf_decode_symbol(d, &d->dtree);
if (distance > d->dtree.max_sym || distance > 29) return TINF_DATA_ERROR;
offset = tinf_getbits_base(d, dist_bits[distance], dist_base[distance]);
if (d->overflow || (unsigned int) offset > out->produced
|| (unsigned int) offset > out->size) return TINF_DATA_ERROR;
while (length--) {
const unsigned char byte = out->window[(out->produced - offset) & (out->size - 1)];
const int result = tinf_record_emit(out, byte);
if (result) return result;
}
}
}
}
static void tinf_record_reverse(unsigned char *begin, unsigned char *end)
{
while (begin < end) {
unsigned char byte;
--end;
if (begin == end) break;
byte = *begin;
*begin++ = *end;
*end = byte;
}
}
int tinf_uncompress_record(void *window, unsigned int window_size,
unsigned int *record_length, const void *source, unsigned int source_len,
unsigned int output_limit, unsigned int record_index)
{
struct tinf_data d;
struct tinf_record_output out;
int final = 0;
if (!window || !record_length || !source || !source_len || !output_limit
|| window_size != 4096) return TINF_DATA_ERROR;
*record_length = 0;
d.source = (const unsigned char *) source;
d.source_end = d.source + source_len;
d.tag = 0;
d.bitcount = d.overflow = 0;
out.window = (unsigned char *) window;
out.size = window_size;
out.produced = out.length = 0;
out.limit = output_limit;
out.remaining = record_index;
do {
unsigned int type;
int result;
final = tinf_getbits(&d, 1);
type = tinf_getbits(&d, 2);
if (d.overflow) return TINF_DATA_ERROR;
if (type == 0) {
unsigned int length;
if (d.source_end - d.source < 4) return TINF_DATA_ERROR;
length = read_le16(d.source);
if (length != (~read_le16(d.source + 2) & 0xffff)) return TINF_DATA_ERROR;
d.source += 4;
if ((unsigned int) (d.source_end - d.source) < length) return TINF_DATA_ERROR;
d.tag = 0;
d.bitcount = 0;
result = TINF_OK;
while (length--) {
result = tinf_record_emit(&out, *d.source++);
if (result) break;
}
} else if (type == 1 || type == 2) {
if (type == 1) {
tinf_build_fixed_trees(&d.ltree, &d.dtree);
result = TINF_OK;
} else result = tinf_decode_trees(&d, &d.ltree, &d.dtree);
if (result == TINF_OK) result = tinf_record_block(&d, &out);
} else return TINF_DATA_ERROR;
if (result == 1) {
/* Rotate in place: the selected record may cross the circular boundary.
* Include its NUL and use only initialized history bytes. */
const unsigned int used = out.produced < out.size ? out.produced : out.size;
const unsigned int begin = (out.produced - out.length - 1) & (out.size - 1);
tinf_record_reverse(out.window, out.window + begin);
tinf_record_reverse(out.window + begin, out.window + used);
tinf_record_reverse(out.window, out.window + used);
*record_length = out.length;
return TINF_OK;
}
if (result != TINF_OK) return result;
} while (!final);
return TINF_DATA_ERROR; /* Selected record or its terminator is missing. */
}
src/helpers/ota/tinf/tinf.h
+7
View File
@@ -93,6 +93,13 @@ int TINFCC tinf_uncompress(void *dest, unsigned int *destLen,
int TINFCC tinf_uncompress_exact(void *dest, unsigned int *destLen,
const void *source, unsigned int sourceLen);
/* MeshCore: stop after one NUL-delimited record, using the caller's 4096-byte
* circular history as the final contiguous record buffer. output_limit bounds
* total inflated bytes, including preceding records. No heap is used. */
int TINFCC tinf_uncompress_record(void *window, unsigned int window_size,
unsigned int *record_length, const void *source, unsigned int source_len,
unsigned int output_limit, unsigned int record_index);
/**
* Decompress `sourceLen` bytes of gzip data from `source` to `dest`.
*
src/helpers/ota/tinf/tinflate.c
+4
View File
@@ -643,6 +643,10 @@ int tinf_uncompress_exact(void *dest, unsigned int *destLen,
return tinf_uncompress_internal(dest, destLen, source, sourceLen, 1);
}
#if defined(MESHCORE_TINF_RECORD)
#include "reader_record.inc"
#endif
/* clang -g -O1 -fsanitize=fuzzer,address -DTINF_FUZZING tinflate.c */
#if defined(TINF_FUZZING)
#include <limits.h>
src/helpers/ui/SmallFontGlyph.h
+13
View File
@@ -6,11 +6,24 @@ namespace mesh {
namespace ui {
// Packed monochrome bitmap metrics, with GFX-style baseline offsets.
#if defined(NRF52_PLATFORM)
// Both fixed ASCII fonts use offsets <= 209, dimensions <= 7, advances <= 6
// and baseline offsets -5..1. Keep every metric/pixel while storing the two
// 95-glyph tables in 570 bytes instead of 1520; no decoded cache is needed.
// Native and Cortex-M4 tests check every source metric and rendered glyph.
struct __attribute__((packed)) SmallFontGlyph {
uint8_t bitmap_offset;
uint16_t width : 3, height : 3, advance : 3;
int16_t x_offset : 3, y_offset : 4;
};
static_assert(sizeof(SmallFontGlyph) == 3, "Compact font metric layout changed");
#else
struct SmallFontGlyph {
uint16_t bitmap_offset;
uint8_t width, height, advance;
int8_t x_offset, y_offset;
};
#endif
} // namespace ui
} // namespace mesh
test/test_arm_instruction_meter.py
+75
View File
@@ -0,0 +1,75 @@
#!/usr/bin/env python3
"""Keep native and Python ARM per-instruction measurements identical."""
from pathlib import Path
import shutil
import sys
import unittest
from unittest import mock
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / 'scripts'))
try:
import unicorn
from unicorn import Uc, UC_ARCH_ARM, UC_MODE_THUMB, UC_MODE_MCLASS
from unicorn import arm_const as arm
import arm_instruction_meter as module
except ImportError:
unicorn = None
@unittest.skipUnless(unicorn is not None, 'Unicorn is not installed in this host job')
class ArmInstructionMeterTest(unittest.TestCase):
def machine(self, native):
uc = Uc(UC_ARCH_ARM, UC_MODE_THUMB | UC_MODE_MCLASS)
uc.mem_map(0x1000, 0x1000)
uc.mem_map(0x20000000, 0x1000)
# push {r7,lr}; sub sp,32; movs r0,100; loop: subs r0,1;
# bne loop; add sp,32; pop {r7,pc}. Exactly 205 instructions.
uc.mem_write(0x1100, bytes.fromhex('80b588b064200138fdd108b080bd'))
return uc, module.ArmInstructionMeter(uc, prefer_native=native)
def run_machine(self, machine, meter, top):
machine.reg_write(arm.UC_ARM_REG_SP, top)
machine.reg_write(arm.UC_ARM_REG_LR, 0x1001)
meter.reset(top)
machine.emu_start(0x1101, 0x1000, timeout=1000000, count=1000)
self.assertEqual(machine.reg_read(arm.UC_ARM_REG_PC), 0x1000)
self.assertEqual(machine.reg_read(arm.UC_ARM_REG_SP), top)
self.assertEqual(machine.reg_read(arm.UC_ARM_REG_R0), 0)
self.assertEqual(meter.snapshot(), {'minimum_sp': top - 40, 'instructions': 205})
return meter.snapshot()
def test_native_python_exact_measurements_repeated_and_independent_contexts(self):
if unicorn.__version__ != '2.1.4':
self.skipTest('native private handle API is pinned to Unicorn 2.1.4')
if not any(shutil.which(name) for name in ('cc', 'gcc', 'clang')):
self.skipTest('host C compiler unavailable')
python_uc, python_meter = self.machine(False)
native_uc, native_meter = self.machine(True)
second_uc, second_meter = self.machine(True)
self.assertEqual(native_meter.backend, 'native')
self.assertEqual(second_meter.backend, 'native')
for top in (0x20001000, 0x20000f00, 0x20001000):
expected = self.run_machine(python_uc, python_meter, top)
self.assertEqual(self.run_machine(native_uc, native_meter, top), expected)
self.assertEqual(self.run_machine(second_uc, second_meter, top), expected)
def test_missing_native_compiler_uses_identical_python_measurement(self):
module._native_library.cache_clear()
try:
with mock.patch.object(module.shutil, 'which', return_value=None):
machine, meter = self.machine(True)
finally:
module._native_library.cache_clear()
self.assertEqual(meter.backend, 'python')
self.run_machine(machine, meter, 0x20001000)
def test_unreviewed_unicorn_version_uses_python_measurement(self):
with mock.patch.object(module.unicorn, '__version__', 'unreviewed-core'):
machine, meter = self.machine(True)
self.assertEqual(meter.backend, 'python')
self.run_machine(machine, meter, 0x20001000)
if __name__ == '__main__':
unittest.main()
test/test_companion_reader.py
+21 -7
View File
@@ -93,11 +93,22 @@ class CompanionReaderTest(unittest.TestCase):
return found[0]
reader, decoder = frames("reader"), frames("tinf")
# Summing all helper frames overbounds mutually exclusive branches
# (tree building and block decoding), including their leaf calls.
decoder_bytes = sum(count for name, count in decoder.items()
if not name.endswith(":tinf_init") and not name.endswith(":tinf_uncompress"))
# Tree building and block decoding are mutually exclusive too;
# sum their helpers as a conservative overbound. Linear OTA and
# circular Reader entry points never nest, so retain the larger
# entry family instead of adding both independent histories.
linear_names = (":tinf_uncompress_internal", ":tinf_uncompress_exact",
":tinf_inflate_block_data")
ring_names = (":tinf_uncompress_record", ":tinf_record_block",
":tinf_record_emit", ":tinf_record_reverse")
linear = sum(count for name, count in decoder.items() if name.endswith(linear_names))
ring = sum(count for name, count in decoder.items() if name.endswith(ring_names))
shared = sum(count for name, count in decoder.items()
if not name.endswith(linear_names + ring_names + (":tinf_init", ":tinf_uncompress")))
self.assertGreater(ring, 0, "actual nRF52 Reader ring decoder must be measured")
decoder_bytes = shared + max(linear, ring)
lookup = frame(reader, "mesh::readReaderVerse(")
lookup += sum(count for name, count in reader.items() if "mesh::bible::lookup(" in name)
terminal = (frame(reader, "mesh::handleReaderCommand(")
+ frame(reader, "::printReplyPage(") + frame(reader, "::replyVerse(")
+ lookup + decoder_bytes)
@@ -434,7 +445,10 @@ int main() {
"-I" + str(ROOT / "src"), "-I" + str(FIXTURE),
str(FIXTURE / "test_reader_screen.cpp"), str(ROOT / "src/helpers/CompanionReader.cpp"),
str(obj), "-o", str(binary)])
self.run_checked([str(binary)])
# All 879 verses are laid out forwards/backwards at 18 panel
# geometries. The 4 KiB streaming nRF corpus does more
# bounded decode work per cold lookup; keep every case.
self.run_checked([str(binary)], timeout=600)
def test_actual_reader_stream_pauses_and_decode_entrypoints_reject_during_morse(self):
cc, cxx = shutil.which("gcc"), shutil.which("g++")
@@ -651,9 +665,9 @@ int main() {
str(ROOT / "src/helpers/ui/DisplayDriver.cpp"), "-o", str(binary)])
self.run_checked([str(binary)])
def run_checked(self, command):
def run_checked(self, command, timeout=60):
result = subprocess.run(command, capture_output=True, encoding="utf-8",
errors="backslashreplace", timeout=60)
errors="backslashreplace", timeout=timeout)
self.assertEqual(result.returncode, 0, result.stdout + result.stderr)
return result.stdout
test/test_nrf52_program_lifetime.py
+81
View File
@@ -0,0 +1,81 @@
#!/usr/bin/env python3
"""Keep local cleanup while omitting unreachable nRF52 process-exit callbacks."""
import importlib.util
from pathlib import Path
import shutil
import subprocess
import tempfile
import unittest
ROOT = Path(__file__).resolve().parents[1]
spec = importlib.util.spec_from_file_location("nrf52_flash_trim", ROOT / "scripts/nrf52_flash_trim.py")
policy = importlib.util.module_from_spec(spec)
spec.loader.exec_module(policy)
class Nrf52ProgramLifetimeTest(unittest.TestCase):
def test_automatic_delete_and_explicit_shutdown_remain_and_exit_callback_is_not_registered(self):
compiler = shutil.which("g++")
if not compiler:
self.skipTest("native G++ unavailable")
source = r'''
#include <cassert>
extern "C" int __aeabi_atexit(void*, void (*)(void*), void*);
extern "C" int __cxa_atexit(void (*)(void*), void*, void*);
extern "C" void __cxa_finalize(void*);
unsigned closed = 0, destroyed = 0, callback = 0;
void cleanup(void* count) { ++*static_cast<unsigned*>(count); }
struct Resource {
void close() { ++closed; }
~Resource() { close(); ++destroyed; }
};
int main() {
{ Resource automatic; }
auto* allocated = new Resource;
delete allocated;
assert(closed == 2 && destroyed == 2);
Resource explicit_close;
explicit_close.close();
assert(closed == 3 && destroyed == 2);
unsigned dso = 0;
assert(__aeabi_atexit(&callback, cleanup, &dso) == 0);
__cxa_finalize(&dso);
assert(callback == 0);
// Positive control: the finalizer really can run a registered callback.
assert(__cxa_atexit(cleanup, &callback, &dso) == 0);
__cxa_finalize(&dso);
assert(callback == 1);
}
'''
with tempfile.TemporaryDirectory(prefix="mesh-nrf52-lifetime-") as directory:
directory = Path(directory)
main, binary = directory / "main.cpp", directory / "checks"
main.write_text(source, encoding="ascii")
subprocess.run([compiler, "-std=c++11", "-Os", "-flto", "-DNRF52_PLATFORM=1",
"-DMESH_NRF52_NO_PROGRAM_EXIT=1", str(main),
str(ROOT / "src/Nrf52ProgramLifetime.cpp"), "-o", str(binary)], check=True)
subprocess.run([str(binary)], check=True)
def test_unknown_and_absent_framework_main_keep_standard_registration(self):
self.assertFalse(policy.supported_program_lifetime(None))
with tempfile.TemporaryDirectory(prefix="mesh-nrf52-lifetime-core-") as directory:
self.assertFalse(policy.supported_program_lifetime(directory))
main = Path(directory) / "cores/nRF5/main.cpp"
main.parent.mkdir(parents=True)
main.write_text("int main() { return 0; }\n", encoding="ascii")
self.assertFalse(policy.supported_program_lifetime(directory))
def test_current_pinned_core_is_the_reviewed_nonreturning_main_when_installed(self):
framework = Path.home() / ".platformio/packages/framework-arduinoadafruitnrf52"
if not framework.is_dir():
self.skipTest("pinned nRF52 framework not installed in this host job")
if not policy.supported_program_lifetime(framework):
self.skipTest("custom or newer core retains ordinary exit registration")
main = (framework / "cores/nRF5/main.cpp").read_text()
self.assertIn("while (1)", main)
self.assertIn("vTaskStartScheduler();", main)
self.assertIn("NVIC_SystemReset();", main)
if __name__ == "__main__":
unittest.main()
test/test_reader_stream.py
+415
View File
@@ -0,0 +1,415 @@
#!/usr/bin/env python3
"""Exercise the actual bounded Reader stream decoder and selected-verse lookup."""
import importlib.util
import json
from pathlib import Path
import random
import shutil
import subprocess
import tempfile
import unittest
import zlib
ROOT = Path(__file__).resolve().parents[1]
def raw_deflate(data, level=9, strategy=zlib.Z_DEFAULT_STRATEGY):
compressor = zlib.compressobj(level, zlib.DEFLATED, -12, 8, strategy)
return compressor.compress(data) + compressor.flush()
def stored(data, final=True):
assert len(data) <= 65535
size = len(data)
return bytes([int(final)]) + size.to_bytes(2, "little") + (size ^ 65535).to_bytes(2, "little") + data
class Bits:
def __init__(self):
self.bits = []
def put(self, value, count):
self.bits.extend((value >> bit) & 1 for bit in range(count))
def huffman(self, value, count):
self.put(int(f"{value:0{count}b}"[::-1], 2), count)
def literal(self, symbol):
if symbol <= 143:
self.huffman(48 + symbol, 8)
elif symbol <= 255:
self.huffman(400 + symbol - 144, 9)
elif symbol <= 279:
self.huffman(symbol - 256, 7)
else:
self.huffman(192 + symbol - 280, 8)
def finish(self):
self.bits.extend([0] * ((-len(self.bits)) % 8))
return bytes(sum(self.bits[start + bit] << bit for bit in range(8))
for start in range(0, len(self.bits), 8))
def back_reference(distance):
bits = Bits()
bits.put(1, 1)
bits.put(1, 2) # final fixed-Huffman block
bits.literal(257) # length 3
if distance == 4096:
bits.huffman(23, 5)
bits.put(1023, 10)
elif distance == 4097:
bits.huffman(24, 5)
bits.put(0, 11)
else:
raise AssertionError(distance)
bits.literal(0)
bits.literal(256)
return bits.finish()
def array(data):
return "{" + ",".join(str(byte) for byte in data) + "}"
class ReaderStreamTest(unittest.TestCase):
@classmethod
def setUpClass(cls):
if not shutil.which("g++") or not shutil.which("gcc"):
raise unittest.SkipTest("native C/C++ compilers unavailable")
cls.temporary = tempfile.TemporaryDirectory(prefix="mesh-reader-stream-")
cls.directory = Path(cls.temporary.name)
cls.source = cls.directory / "main.cpp"
cls.decoder = cls.directory / "tinf.o"
flags = ["-O1", "-g", "-fsanitize=address,undefined", "-fno-sanitize-recover=all",
"-fno-omit-frame-pointer", "-fno-pie", "-I" + str(ROOT / "src"),
"-DOTA_TRANSPORT_DEFLATE_TEST=1", "-DTINF_READER_STREAM_TEST=1"]
subprocess.run(["gcc", "-std=c11", *flags, "-c", str(ROOT / "src/helpers/ota/OtaTinf.c"),
"-o", str(cls.decoder)], check=True, capture_output=True)
cls.flags = flags
@classmethod
def tearDownClass(cls):
cls.temporary.cleanup()
def compile(self, body, name="test"):
source = self.directory / (name + ".cpp")
source.write_text(body, encoding="ascii")
binary = self.directory / name
result = subprocess.run(["g++", "-std=c++17", *self.flags, "-no-pie", str(source),
str(self.decoder), "-o", str(binary)], capture_output=True, text=True)
self.assertEqual(result.returncode, 0, result.stdout + result.stderr)
return binary
def run_binary(self, binary, *args):
result = subprocess.run([str(binary), *args], capture_output=True)
self.assertEqual(result.returncode, 0, result.stdout.decode(errors="replace") + result.stderr.decode(errors="replace"))
return result.stdout
def test_canonical_stream_header_and_token_vocabulary_are_reproducible(self):
import sys
path = str(ROOT / "tools/bible")
sys.path.insert(0, path)
try:
spec = importlib.util.spec_from_file_location("reader_stream_repro_packer", ROOT / "tools/bible/pack_reader_stream.py")
packer = importlib.util.module_from_spec(spec)
spec.loader.exec_module(packer)
document = json.loads(packer.blocks.SOURCE.read_text(encoding="utf-8"))
self.assertEqual(packer.HEADER.read_text(encoding="ascii"), packer.render(document))
pinned = json.loads(packer.TOKENS.read_text(encoding="ascii"))
self.assertLessEqual(len(pinned), 128)
self.assertEqual(len(pinned), len(set(pinned)))
for word in pinned:
self.assertTrue(word)
self.assertTrue(all(32 <= byte < 127 for byte in word.encode("ascii")))
finally:
sys.path.remove(path)
def test_all_879_verses_match_independently_normalized_source(self):
binary = self.compile(r"""
#include <helpers/bible/ReaderStreamData.generated.h>
#include <cassert>
#include <cstdio>
#include <cstring>
int main() {
using namespace mesh::bible;
struct { unsigned char before[32]; char scratch[kBlockSize]; unsigned char after[32]; } buffer;
memset(&buffer, 0xa5, sizeof(buffer));
for (unsigned chapter=1; chapter<=21; ++chapter) {
for (unsigned verse=1; verse<=kChapterVerses[chapter-1]; ++verse) {
const char* text=nullptr;
assert(lookup(generated::readerCorpus, Reference{uint8_t(chapter),uint8_t(verse)},
buffer.scratch,sizeof(buffer.scratch),text)==LookupResult::Found);
assert(text>=buffer.scratch && text<buffer.scratch+sizeof(buffer.scratch));
assert(memchr(text,0,buffer.scratch+sizeof(buffer.scratch)-text));
for(auto c:buffer.before) assert(c==0xa5);
for(auto c:buffer.after) assert(c==0xa5);
fwrite(text,1,strlen(text)+1,stdout);
}
}
}
""", "corpus")
document = json.loads((ROOT / "tools/bible/reader-web.json").read_text())
typography = str.maketrans({"\u00a0": " ", "\u2013": "-", "\u2014": "-",
"\u2018": "'", "\u2019": "'", "\u201c": '\"', "\u201d": '\"'})
expected = b"".join(text.translate(typography).encode("ascii") + b"\0"
for text in document["verses"].values())
self.assertEqual(self.run_binary(binary), expected)
def test_ring_bounds_truncation_distances_and_deterministic_mutations(self):
cases = []
def good(plain, record, expected, compressor=raw_deflate):
packed = compressor(plain)
cases.append((packed, len(plain), record, 0, expected))
return packed
for compressor in (raw_deflate, stored,
lambda plain: raw_deflate(plain, strategy=zlib.Z_FIXED)):
good(b"hello\0world\0", 1, b"world", compressor)
good(b"p" * 4090 + b"\0" + b"crossing-boundary-record\0", 1,
b"crossing-boundary-record", compressor)
good(b"q" * 4095 + b"\0", 0, b"q" * 4095, compressor)
good(b"\0missing\0", 0, b"", compressor)
good(b"a\0" * 3000 + b"last record\0", 3000, b"last record")
prefix = b"abc\0" * 1024
packed = stored(prefix, False) + back_reference(4096)
self.assertEqual(zlib.decompress(packed, -15), prefix + b"abc\0")
cases.append((packed, len(prefix) + 4, 1024, 0, b"abc"))
prefix = b"a\0" * 2049
packed = stored(prefix, False) + back_reference(4097)
self.assertEqual(zlib.decompress(packed, -15), prefix + b"\0a\0\0")
cases.append((packed, len(prefix) + 4, 2049, -1, b""))
for bad in (b"\x07", b"\x01\x01\x00\x00\x00x", b"\x05\x00\x00"):
cases.append((bad, 4096, 0, -1, b""))
packed = raw_deflate(b"a\0second\0")
cases.append((packed, 8, 1, -1, b"")) # terminator exceeds total-output limit
cases.append((packed, 20, 2, -1, b"")) # selected record absent
cases.append((raw_deflate(b"q" * 4096 + b"\0"), 5000, 0, -1, b""))
for size in range(len(packed)):
cases.append((packed[:size], 20, 2, -1, b""))
definitions, calls = [], []
for index, (packed, limit, record, status, expected) in enumerate(cases):
definitions.append(f"const unsigned char p{index}[]={array(packed or b'x')};")
definitions.append(f"const unsigned char e{index}[]={array(expected + bytes([0]))};")
calls.append(f"check(p{index},{len(packed)},{limit},{record},{status},e{index},{len(expected)});")
body = r"""
#include <helpers/ota/tinf/tinf.h>
#include <cassert>
#include <cstdio>
#include <cstring>
#include <cstdint>
struct Guard { unsigned char before[32]; unsigned char window[4096]; unsigned char after[32]; };
void check(const unsigned char* source,unsigned size,unsigned limit,unsigned record,int expected,
const unsigned char* text,unsigned length) {
Guard buffer; memset(&buffer,0xa5,sizeof(buffer)); unsigned actual=777;
int result=tinf_uncompress_record(buffer.window,4096,&actual,source,size,limit,record);
if(expected==0) { assert(result==TINF_OK);assert(actual==length);assert(!memcmp(buffer.window,text,length+1)); }
else assert(result!=TINF_OK);
for(auto c:buffer.before) assert(c==0xa5);
for(auto c:buffer.after) assert(c==0xa5);
}
""" + "\n".join(definitions) + r"""
int main() {
""" + "\n".join(calls) + r"""
Guard buffer; memset(&buffer,0xa5,sizeof(buffer)); unsigned length=777;
const unsigned char sample[]={3,0};
assert(tinf_uncompress_record(nullptr,4096,&length,sample,2,4096,0)!=TINF_OK);
assert(tinf_uncompress_record(buffer.window,2048,&length,sample,2,4096,0)!=TINF_OK);
assert(tinf_uncompress_record(buffer.window,4096,nullptr,sample,2,4096,0)!=TINF_OK);
assert(tinf_uncompress_record(buffer.window,4096,&length,nullptr,2,4096,0)!=TINF_OK);
assert(tinf_uncompress_record(buffer.window,4096,&length,sample,2,0,0)!=TINF_OK);
uint32_t seed=0x84a123f1; unsigned char input[512];
for(unsigned pass=0;pass<10000;++pass) {
auto next=[&]() { seed^=seed<<13;seed^=seed>>17;seed^=seed<<5;return seed; };
const unsigned size=next()%sizeof(input);
for(unsigned i=0;i<size;++i) input[i]=next();
length=777;
int result=tinf_uncompress_record(buffer.window,4096,&length,input,size,16384,next()%12);
if(result==TINF_OK) { assert(length<4096);assert(buffer.window[length]==0); }
for(auto c:buffer.before) assert(c==0xa5);
for(auto c:buffer.after) assert(c==0xa5);
}
}
"""
self.run_binary(self.compile(body, "malformed"))
def test_literal_capacity_and_lookup_rejects_invalid_metadata(self):
# Generated input keeps every reference and the original 4095-byte
# literal capacity, including corpora with no profitable dictionary.
import sys
path = str(ROOT / "tools/bible")
sys.path.insert(0, path)
try:
spec = importlib.util.spec_from_file_location("reader_stream_fixture_packer", ROOT / "tools/bible/pack_reader_stream.py")
packer = importlib.util.module_from_spec(spec)
spec.loader.exec_module(packer)
document = json.loads(packer.blocks.SOURCE.read_text())
document["verses"] = dict.fromkeys(packer.blocks.references(), None)
document["verses"]["1:1"] = "q" * 4095
rendered = packer.render(document)
finally:
sys.path.remove(path)
header = self.directory / "literal.h"
header.write_text(rendered.replace('#include "ReaderStreamLookup.h"', '#include <helpers/bible/ReaderStreamLookup.h>'), encoding="ascii")
binary = self.compile(r"""
#include "literal.h"
#include <cassert>
#include <cstring>
int main() {
using namespace mesh::bible;
char scratch[kBlockSize];const char* text=nullptr;
const auto& c=generated::readerCorpus;
assert(lookup(c,{1,1},scratch,sizeof(scratch),text)==LookupResult::Found);
assert(strlen(text)==4095);for(unsigned n=0;n<4095;++n) assert(text[n]=='q');
assert(lookup(c,{1,2},scratch,sizeof(scratch),text)==LookupResult::Missing);assert(!text);
assert(lookup(c,{0,1},scratch,sizeof(scratch),text)==LookupResult::Invalid);assert(!text);
assert(lookup(c,{1,1},scratch,sizeof(scratch)-1,text)==LookupResult::Corrupt);assert(!text);
assert(lookup(c,{1,1},nullptr,sizeof(scratch),text)==LookupResult::Corrupt);assert(!text);
auto bad=c;bad.data=nullptr;assert(lookup(bad,{1,1},scratch,sizeof(scratch),text)==LookupResult::Corrupt);
bad=c;bad.data_size=1;assert(lookup(bad,{1,1},scratch,sizeof(scratch),text)==LookupResult::Corrupt);
bad=c;bad.encoded_size=1;assert(lookup(bad,{1,1},scratch,sizeof(scratch),text)==LookupResult::Corrupt);
bad=c;--bad.verse_count;assert(lookup(bad,{1,1},scratch,sizeof(scratch),text)==LookupResult::Corrupt);
}
""", "literal")
self.run_binary(binary)
def test_independent_stream_header_bounds_and_boundary_records(self):
binary = self.compile(r"""
#include <helpers/bible/ReaderStreamData.generated.h>
#include <cassert>
#include <cstring>
#include <vector>
int main() {
using namespace mesh::bible;
const auto& original=generated::readerCorpus;
assert(original.chunk_count==3);
struct { unsigned char before[32];char scratch[kBlockSize];unsigned char after[32]; } buffer;
auto check=[&](const StreamTokenCorpus& corpus) {
memset(&buffer,0xa5,sizeof(buffer));const char* text=nullptr;
assert(lookup(corpus,{1,1},buffer.scratch,sizeof(buffer.scratch),text)==LookupResult::Corrupt);
assert(!text);
for(auto byte:buffer.before) assert(byte==0xa5);
for(auto byte:buffer.after) assert(byte==0xa5);
};
auto bad=original;bad.chunk_count=4;check(bad);
bad=original;bad.chunk_count=kVerseCount+1;check(bad);
bad=original;bad.data_size=5;check(bad);
bad=original;--bad.data_size;check(bad);
bad=original;++bad.encoded_size;check(bad);
std::vector<uint8_t> bytes(original.data,original.data+original.data_size);
const size_t second=6+bytes[4]+(unsigned(bytes[5])<<8);
for(auto field: {size_t(0),size_t(2),size_t(4),second,second+2,second+4}) {
bytes.assign(original.data,original.data+original.data_size);
bytes[field]=bytes[field+1]=0;
if(field==0) bytes[0]=1; // First record must start at verse zero.
bad=original;bad.data=bytes.data();check(bad);
}
bytes.assign(original.data,original.data+original.data_size);
bytes[4]=bytes[5]=0xff;bad=original;bad.data=bytes.data();check(bad);
bytes.assign(original.data,original.data+original.data_size);
bytes.push_back(0);bad=original;bad.data=bytes.data();++bad.data_size;check(bad);
// Independently inspect all stream boundaries and byte bounds. The complete
// canonical comparison also checks every record immediately around them.
unsigned offset=0,total=0,first=0;
for(unsigned n=0;n<original.chunk_count;++n) {
const auto* p=original.data+offset;
unsigned current=p[0]+(unsigned(p[1])<<8),plain=p[2]+(unsigned(p[3])<<8);
unsigned packed=p[4]+(unsigned(p[5])<<8);
assert((n ? current>first : current==0) && plain<=17000 && packed>0);
first=current;offset+=6+packed;total+=plain;
}
assert(offset==original.data_size && total==original.encoded_size);
}
""", "chunks")
self.run_binary(binary)
def test_dictionary_corruption_tokens_and_workspace_are_bounded(self):
cases = []
def fixture(encoded=b"\x80 suffix", words=b"Hello\0world\0", count=2,
*, plain_size=None, packed_words=None, expected=3, output=b""):
plain=encoded+b"\0"+bytes(878)
cases.append((raw_deflate(plain),len(plain),
raw_deflate(words) if packed_words is None else packed_words,
len(words) if plain_size is None else plain_size,count,expected,output))
fixture(expected=0,output=b"Hello suffix")
fixture(encoded=b"\x81",expected=0,output=b"world")
fixture(encoded=b"\x82") # token outside the declared vocabulary
fixture(encoded=b"literal\x1f")
fixture(encoded=b"literal\x7f")
fixture(count=0)
fixture(count=129)
fixture(words=b"Hello\0\0")
fixture(words=b"Hello\0wo\x7frld\0")
fixture(words=b"Hello\0wo\x80rld\0")
fixture(words=b"Hello\0world")
fixture(words=b"Hello\0world\0extra\0")
fixture(plain_size=1)
fixture(plain_size=4096)
fixture(plain_size=4097)
fixture(packed_words=raw_deflate(b"Hello\0world\0")[:-1])
fixture(packed_words=raw_deflate(b"Hello\0world\0")+b"x")
fixture(packed_words=b"")
fixture(encoded=b"\x80"*100,words=b"h"*64+b"\0",count=1)
fixture(encoded=b"\x80"*4095,words=b"Hello\0",count=1)
definitions,calls=[],[]
for n,(data,size,words,word_plain,count,expected,output) in enumerate(cases):
definitions.extend((f"const unsigned char p{n}[]={array(data)};",
f"const unsigned char w{n}[]={array(words or b'x')};",
f"const unsigned char e{n}[]={array(output+bytes([0]))};"))
calls.append(f"check(p{n},sizeof(p{n}),{size},w{n},{len(words)},{word_plain},{count},{expected},e{n});")
body=r"""
#include <helpers/bible/ReaderStreamLookup.h>
#include <cassert>
#include <cstring>
void check(const uint8_t* data,unsigned data_size,unsigned plain_size,const uint8_t* words,
unsigned words_size,unsigned word_plain,unsigned count,unsigned expected,const unsigned char* output) {
using namespace mesh::bible;
StreamTokenCorpus c={"fixture","test",data,data_size,plain_size,words,
uint16_t(words_size),uint16_t(word_plain),uint16_t(count),kVerseCount};
struct { unsigned char before[32];char scratch[kBlockSize];unsigned char after[32]; } buffer;
memset(&buffer,0xa5,sizeof(buffer));const char* text=nullptr;
auto result=lookup(c,{1,1},buffer.scratch,sizeof(buffer.scratch),text);
assert(unsigned(result)==expected);
if(result==LookupResult::Found) assert(text && !strcmp(text,reinterpret_cast<const char*>(output)));
else assert(!text);
for(auto byte:buffer.before) assert(byte==0xa5);
for(auto byte:buffer.after) assert(byte==0xa5);
}
"""+"\n".join(definitions)+"\nint main() {\n"+"\n".join(calls)+"\n}\n"
self.run_binary(self.compile(body,"dictionary"))
def test_worst_case_lookup_host_benchmark(self):
# Keep timings informational: hosts vary, but report the CPU tradeoff
# against the unchanged block decoder using identical host options.
binary = self.compile(r"""
#include <helpers/bible/ReaderStreamData.generated.h>
#define generated old_generated
#include <helpers/bible/ReaderData.generated.h>
#undef generated
#include <cassert>
#include <chrono>
#include <cstdio>
#include <cstring>
template<class Corpus> double bench(const Corpus& corpus) {
using namespace mesh::bible;
char scratch[kBlockSize];const char* text=nullptr;
volatile unsigned sum=0;
auto start=std::chrono::steady_clock::now();
for(unsigned n=0;n<80;++n) {
assert(lookup(corpus,{21,25},scratch,sizeof(scratch),text)==LookupResult::Found);
sum+=strlen(text);
}
assert(sum>0);
return std::chrono::duration<double,std::milli>(std::chrono::steady_clock::now()-start).count()/80;
}
int main() { printf("Reader last verse host ms: block=%.4f stream=%.4f\n",
bench(mesh::bible::old_generated::readerCorpus),bench(mesh::bible::generated::readerCorpus)); }
""", "benchmark")
print(self.run_binary(binary).decode().strip())
if __name__ == "__main__":
unittest.main()
test/test_small_font_metrics.py
+108
View File
@@ -0,0 +1,108 @@
#!/usr/bin/env python3
"""Prove compact nRF glyph metrics retain every original field and pixel."""
from pathlib import Path
import re
import shutil
import subprocess
import tempfile
import unittest
ROOT=Path(__file__).resolve().parents[1]
FONT_FILES=("Pixel5FontData.h","Squeezed6FontData.h")
HARNESS=r'''
#include <helpers/ui/Pixel5FontData.h>
#include <helpers/ui/Squeezed6FontData.h>
#include <cassert>
#include <cstdio>
#include <cstring>
#if defined(NRF52_PLATFORM)
static_assert(sizeof(mesh::ui::SmallFontGlyph)==3,"Cortex-M4 glyph packing");
#else
static_assert(sizeof(mesh::ui::SmallFontGlyph)==8,"Existing non-nRF glyph ABI");
#endif
void show(const mesh::ui::SmallFontGlyph* glyphs,const uint8_t* bitmap) {
for(unsigned n=0;n<95;++n) {
const auto& g=glyphs[n];
printf("%u,%u,%u,%u,%d,%d:",unsigned(g.bitmap_offset),unsigned(g.width),
unsigned(g.height),unsigned(g.advance),int(g.x_offset),int(g.y_offset));
for(unsigned row=0;row<g.height;++row) {
for(unsigned col=0;col<g.width;++col) {
const unsigned bit=row*g.width+col;
putchar(bitmap[g.bitmap_offset+bit/8]&(0x80>>(bit%8))?'1':'0');
}
}
putchar('\n');
}
}
int main() {
show(mesh::ui::pixel5Glyphs,mesh::ui::pixel5Bitmaps);
show(mesh::ui::squeezed6Glyphs,mesh::ui::squeezed6Bitmaps);
}
'''
class SmallFontMetricsTest(unittest.TestCase):
def source_metrics(self):
result=[]
for filename in FONT_FILES:
source=(ROOT/"src/helpers/ui"/filename).read_text()
rows=[tuple(map(int,row.split(",")))
for row in re.findall(r"\{([\d, \-]+)\}",source)]
self.assertEqual(len(rows),95)
for offset,width,height,advance,x,y in rows:
self.assertLessEqual(offset,255)
self.assertTrue(0<=width<=7 and 0<=height<=7 and 0<=advance<=7)
self.assertTrue(-4<=x<=3 and -8<=y<=7)
result+=rows
return result
def test_all_190_metrics_and_pixels_match_original_layout(self):
expected=self.source_metrics()
with tempfile.TemporaryDirectory(prefix="mesh-small-font-") as folder:
folder=Path(folder);source=folder/"main.cpp";source.write_text(HARNESS)
output=[]
for nrf in (False,True):
binary=folder/("nrf" if nrf else "original")
result=subprocess.run(["g++","-std=c++11","-O1","-g","-Wall","-Wextra",
"-Werror=overflow","-Werror=conversion","-fsanitize=address,undefined",
"-fno-sanitize-recover=all","-fno-pie","-no-pie","-I"+str(ROOT/"src"),
*(["-DNRF52_PLATFORM=1"] if nrf else []),str(source),"-o",str(binary)],capture_output=True,text=True)
self.assertEqual(result.returncode,0,result.stdout+result.stderr)
run=subprocess.run([str(binary)],capture_output=True,text=True)
self.assertEqual(run.returncode,0,run.stdout+run.stderr)
output.append(run.stdout)
actual=[tuple(map(int,row.split(":")[0].split(","))) for row in run.stdout.splitlines()]
self.assertEqual(actual,expected)
self.assertEqual(output[0],output[1])
def test_actual_cortex_m4_layout_and_flash_saving(self):
compiler=shutil.which("arm-none-eabi-g++")
if not compiler:
path=Path.home()/".platformio/packages/toolchain-gccarmnoneeabi/bin/arm-none-eabi-g++"
if path.is_file():compiler=str(path)
if not compiler:self.skipTest("Cortex-M4 compiler unavailable")
compiler=Path(compiler)
nm=compiler.with_name("arm-none-eabi-nm")
self.source_metrics()
with tempfile.TemporaryDirectory(prefix="mesh-small-font-arm-") as folder:
folder=Path(folder);source=folder/"main.cpp"
source.write_text('#include <helpers/ui/Pixel5FontData.h>\n#include <helpers/ui/Squeezed6FontData.h>\n'
'extern "C" const void* keep(unsigned n) { return n?mesh::ui::pixel5Glyphs:mesh::ui::squeezed6Glyphs; }\n')
sizes=[]
for nrf in (False,True):
obj=folder/("nrf.o" if nrf else "original.o")
result=subprocess.run([str(compiler),"-std=gnu++11","-mcpu=cortex-m4","-mthumb","-Oz",
"-fno-exceptions","-fno-rtti","-Werror=overflow","-Werror=conversion",
"-I"+str(ROOT/"src"),*(["-DNRF52_PLATFORM=1"] if nrf else []),
"-c",str(source),"-o",str(obj)],capture_output=True,text=True)
self.assertEqual(result.returncode,0,result.stdout+result.stderr)
symbols=subprocess.check_output([str(nm),"-S",str(obj)],text=True)
rows=[int(line.split()[1],16) for line in symbols.splitlines()
if "pixel5Glyphs" in line or "squeezed6Glyphs" in line]
self.assertEqual(sorted(rows),[95*(3 if nrf else 8)]*2)
sizes.append(sum(rows))
self.assertEqual(sizes[0]-sizes[1],950)
if __name__=="__main__":unittest.main()
tools/bible/README.md
+30
View File
@@ -302,3 +302,33 @@ match for identity/private key, contacts, channels, self settings, auto-add,
tuning and custom variables. The bootloader and storage were not flashed.
Other panel families were host-tested, not physically flashed or visually
qualified in this step. Physical readability remains a user check.
## nRF52 Full Companion flash encoding
nRF52 Full Companions retain all 879 verses with a separate generated stream
asset. The encoded corpus, stream headers and vocabulary occupy 28,787 flash bytes, compared
with 33,399 bytes for the ESP32 block asset. Both decode to the same ASCII text.
The existing 4 KiB scratch is reused first as circular DEFLATE history, then
for the selected verse and vocabulary. There is no new heap buffer or retained
cache. Long verses that cannot share the scratch with the vocabulary are stored
literally, preserving the original 4095-byte verse capacity.
Three independent streams limit a lookup to at most 16,959 encoded bytes
through the selected verse, using the same scratch for each history. This
makes cold Reader access slower in exchange for nRF52 flash space;
normal packet handling does not enter this path. ESP32 keeps its existing
independent blocks. Native tests compare every verse and output byte, exercise
malformed inputs, and check the actual ARM stack against the existing 8 KiB
loop stack. Linked-ELF tests retain the same text and reply checks.
`reader-stream-tokens.json` is the pinned compression vocabulary. It substitutes
only exact source byte strings, including substrings of words. Host-only Google
Zopfli uses 1,000 iterations with a 4 KiB match window; its Apache-2.0 source,
license and archive hash are in `zopfli4k/README.meshcore.txt`. A host C compiler
is needed only when regenerating this asset. Ordinary firmware builds use the
committed header and require no extra downloads.
```sh
python tools/bible/pack_reader_stream.py --check
python test/test_reader_stream.py
```
tools/bible/pack_reader_stream.py
+142
View File
@@ -0,0 +1,142 @@
#!/usr/bin/env python3
"""Pack the unchanged Reader source for a 4 KiB sliding-window nRF52 decoder."""
import argparse
import hashlib
from functools import lru_cache
import shutil
import subprocess
import tempfile
import json
from pathlib import Path
import re
import zlib
import pack_reader as blocks
HEADER = blocks.ROOT / "src/helpers/bible/ReaderStreamData.generated.h"
TOKENS = Path(__file__).with_name("reader-stream-tokens.json")
COMPRESSOR = Path(__file__).with_name("zopfli4k")
STREAM_LIMIT = 17000
@lru_cache(maxsize=1)
def host_compressor():
compiler = shutil.which("cc") or shutil.which("gcc") or shutil.which("clang")
if not compiler:
raise RuntimeError("Regenerating the nRF52 Reader asset requires a host C compiler")
directory = tempfile.TemporaryDirectory(prefix="mesh-reader-zopfli4k-")
binary = Path(directory.name) / "pack4k"
subprocess.run([compiler, "-std=c11", "-O2", "-DNDEBUG", "-I" + str(COMPRESSOR),
*map(str, sorted(COMPRESSOR.glob("*.c"))), "-lm", "-o", str(binary)], check=True)
return directory, binary
@lru_cache(maxsize=16)
def compress_4k(data):
directory, binary = host_compressor()
source = Path(directory.name) / "input"
target = Path(directory.name) / "output"
source.write_bytes(data)
subprocess.run([str(binary), str(source), str(target), "1000"], check=True, capture_output=True)
packed = target.read_bytes()
assert zlib.decompress(packed, wbits=-12) == data
return packed
def token_dictionary(verses):
# The vocabulary is an independently tunable compression asset. Tokens
# only substitute exact source bytes, including matches inside a word.
selected = json.loads(TOKENS.read_text(encoding="ascii"))
if len(selected) > 128 or len(set(selected)) != len(selected):
raise ValueError("Reader vocabulary must contain at most 128 unique byte strings")
words = [word.encode("ascii") for word in selected]
if any(not word or any(byte < 32 or byte >= 127 for byte in word) for word in words):
raise ValueError("Reader tokens must be nonempty printable ASCII")
return [word for word in words if any(word in verse for verse in verses)]
def encode(document):
verses = blocks.source_verses(document)
words = token_dictionary(verses)
tokens = {word: bytes([128 + index]) for index, word in enumerate(words)}
# Tokens are exact byte strings, including any leading space. Matching
# inside another word is lossless too: expansion restores those same bytes.
pattern = re.compile(b"|".join(re.escape(word) for word in sorted(words, key=lambda x: (-len(x), x)))) if words else None
encoded_verses = []
dictionary_size = sum(map(len, words)) + len(words)
for verse in verses:
encoded = pattern.sub(lambda match: tokens[match.group()], verse) if pattern else verse
if encoded != verse and len(encoded) + len(verse) + dictionary_size + 2 > blocks.BLOCK_SIZE:
encoded = verse # Literal output needs no dictionary workspace.
encoded_verses.append(encoded)
records = [verse + b"\0" for verse in encoded_verses]
chunks, first, current = [], 0, b""
for index, record in enumerate(records):
if current and len(current) + len(record) > STREAM_LIMIT:
chunks.append((first, current))
first, current = index, b""
current += record
chunks.append((first, current))
# Independent 12-bit histories bound cold lookup work without adding a
# buffer. Headers keep both compressed and decoded lengths bounded.
packed = bytearray()
for first, plain in chunks:
compressed = compress_4k(plain)
packed += (first.to_bytes(2, "little") + len(plain).to_bytes(2, "little")
+ len(compressed).to_bytes(2, "little") + compressed)
return bytes(packed), b"".join(records), words, len(chunks)
def render(document):
packed, plain, words, chunk_count = encode(document)
dictionary_plain = b"\0".join(words) + b"\0" if words else b""
dictionary = compress_4k(dictionary_plain) if words else b""
digest = hashlib.sha256(json.dumps(document, ensure_ascii=False, sort_keys=True,
separators=(",", ":")).encode()).hexdigest()
lines = ["// Generated by tools/bible/pack_reader_stream.py; do not edit.",
"// Offline Reader corpus. Public domain; see tools/bible/README.md.",
"// Google Zopfli, 1000 iterations, 4096-byte history; exact ASCII text retained.",
f"// Canonical source SHA-256: {digest}",
"#pragma once", '#include "ReaderStreamLookup.h"',
"namespace mesh { namespace bible { namespace generated {",
"static_assert(kBlockSize == 4096, \"Reader history must remain 4096 bytes\");",
"static const uint8_t readerData[] = {"]
for start in range(0, len(packed), 20):
lines.append(" " + ",".join(f"0x{v:02x}" for v in packed[start:start + 20]) + ",")
lines += ["};", "static const uint8_t readerWords[] = {"]
for start in range(0, len(dictionary), 20):
lines.append(" " + ",".join(f"0x{v:02x}" for v in dictionary[start:start + 20]) + ",")
if not dictionary:
lines.append(" 0, // Empty vocabulary: literal records remain supported.")
lines += ["};", "static const StreamTokenCorpus readerCorpus = {",
f" {json.dumps(document['translation'])},",
f" {json.dumps(document['attribution'])},",
f" readerData, sizeof(readerData), {len(plain)},",
f" readerWords, {len(dictionary)}, {len(dictionary_plain)},",
f" {len(words)}, kVerseCount, {chunk_count}", "};", "}}}", ""]
return "\n".join(lines)
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--source", type=Path, default=blocks.SOURCE)
parser.add_argument("--output", type=Path, default=HEADER)
parser.add_argument("--check", action="store_true")
args = parser.parse_args()
document = json.loads(args.source.read_text(encoding="utf-8"))
header = render(document)
if args.check:
if args.output.read_text(encoding="ascii") != header:
raise SystemExit("Reader stream header differs; regenerate with tools/bible/pack_reader_stream.py")
else:
args.output.write_text(header, encoding="ascii", newline="\n")
packed, plain, words, chunk_count = encode(document)
dictionary_plain = b"\0".join(words) + b"\0" if words else b""
dictionary_size = len(compress_4k(dictionary_plain)) if words else 0
print(f"Reader stream: {len(blocks.references())} unchanged verses, "
f"{len(plain)} encoded bytes -> {len(packed)} DEFLATE + "
f"{dictionary_size} dictionary bytes; {chunk_count} independent streams, 4096-byte history")
if __name__ == "__main__":
main()
tools/bible/reader-stream-tokens.json
+130
View File
@@ -0,0 +1,130 @@
[
" the",
" you",
" and",
" that",
" him",
" said",
" to",
" therefore",
" of",
" have",
" will",
" Jesus",
" Father",
" he",
" follow",
" who",
" answered",
" receive",
" word",
" was",
" in",
" because",
" this",
" me",
" his",
" things",
" is",
" from",
" for",
"Jesus",
" world",
" know",
" believe",
" with",
" but",
" which",
" give",
" I",
" came",
" not",
" saw",
" made",
" are",
" Jews",
" love",
" come",
" has",
" my",
" also",
" about",
" God",
" where",
" judgment",
" it",
" out",
" were",
" had",
" those",
" sent",
" sign",
" certain",
" when",
" than",
" would",
" again",
" man",
" life",
" multitude",
" may",
" judge",
" himself",
" be",
" Pharisees",
" do",
" Peter",
" tell",
" first",
" by",
" speak",
" a",
" gave",
" all",
" up",
" unless",
"You",
"The",
" go",
" now",
" Jerusalem",
"Lord",
" knew",
" take",
" so",
" disciple",
" already",
" revealed",
" did",
" prophet",
" see",
" we",
" Son",
" synagogue",
" say",
" Christ",
" time",
" into",
" she",
" going",
"and",
"Philip",
" Scripture",
"for",
" eternal",
" remain",
" no",
" written",
" concerning",
" place",
" should",
" enter",
" bread",
" what",
" one",
" day",
" spoke",
" on",
" them",
" called"
]
tools/bible/zopfli4k/LICENSE
+201
View File
@@ -0,0 +1,201 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright 2011 Google Inc.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
tools/bible/zopfli4k/README.meshcore.txt
+20
View File
@@ -0,0 +1,20 @@
Host-only Google Zopfli with a 4 KiB match window
================================================
Sources: Google Zopfli, bundled in the pinned python-zopfli 0.4.3 source release.
https://github.com/google/zopfli
https://pypi.org/project/zopfli/0.4.3/
Source archive: zopfli-0.4.3.tar.gz
SHA-256: d3a50f91a13cea9bafe025de8fd87a005eb26de02a4f0c193127ddbf23ac8ebe
License: Apache-2.0, preserved in LICENSE and the upstream source headers.
MeshCore changes: util.h sets ZOPFLI_WINDOW_SIZE to 4096 instead of 32768.
reader_pack.c is a small MeshCore command-line adapter. All other C/header
files are unchanged from that archive. This tool runs only on the build host
when regenerating Reader data. It is never compiled into firmware.
The 4 KiB limit is required by the existing on-device Reader scratch buffer.
The ordinary python-zopfli package has no configurable match-window API.
The generator compiles these pinned sources with a host C compiler and uses
1000 optimization iterations. No network access is required. Ordinary firmware
builds use the committed generated header and do not invoke this tool.
tools/bible/zopfli4k/blocksplitter.c
+332
View File
@@ -0,0 +1,332 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "blocksplitter.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include "deflate.h"
#include "squeeze.h"
#include "tree.h"
#include "util.h"
/*
The "f" for the FindMinimum function below.
i: the current parameter of f(i)
context: for your implementation
*/
typedef double FindMinimumFun(size_t i, void* context);
/*
Finds minimum of function f(i) where is is of type size_t, f(i) is of type
double, i is in range start-end (excluding end).
Outputs the minimum value in *smallest and returns the index of this value.
*/
static size_t FindMinimum(FindMinimumFun f, void* context,
size_t start, size_t end, double* smallest) {
if (end - start < 1024) {
double best = ZOPFLI_LARGE_FLOAT;
size_t result = start;
size_t i;
for (i = start; i < end; i++) {
double v = f(i, context);
if (v < best) {
best = v;
result = i;
}
}
*smallest = best;
return result;
} else {
/* Try to find minimum faster by recursively checking multiple points. */
#define NUM 9 /* Good value: 9. */
size_t i;
size_t p[NUM];
double vp[NUM];
size_t besti;
double best;
double lastbest = ZOPFLI_LARGE_FLOAT;
size_t pos = start;
for (;;) {
if (end - start <= NUM) break;
for (i = 0; i < NUM; i++) {
p[i] = start + (i + 1) * ((end - start) / (NUM + 1));
vp[i] = f(p[i], context);
}
besti = 0;
best = vp[0];
for (i = 1; i < NUM; i++) {
if (vp[i] < best) {
best = vp[i];
besti = i;
}
}
if (best > lastbest) break;
start = besti == 0 ? start : p[besti - 1];
end = besti == NUM - 1 ? end : p[besti + 1];
pos = p[besti];
lastbest = best;
}
*smallest = lastbest;
return pos;
#undef NUM
}
}
/*
Returns estimated cost of a block in bits. It includes the size to encode the
tree and the size to encode all literal, length and distance symbols and their
extra bits.
litlens: lz77 lit/lengths
dists: ll77 distances
lstart: start of block
lend: end of block (not inclusive)
*/
static double EstimateCost(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
return ZopfliCalculateBlockSizeAutoType(lz77, lstart, lend);
}
typedef struct SplitCostContext {
const ZopfliLZ77Store* lz77;
size_t start;
size_t end;
} SplitCostContext;
/*
Gets the cost which is the sum of the cost of the left and the right section
of the data.
type: FindMinimumFun
*/
static double SplitCost(size_t i, void* context) {
SplitCostContext* c = (SplitCostContext*)context;
return EstimateCost(c->lz77, c->start, i) + EstimateCost(c->lz77, i, c->end);
}
static void AddSorted(size_t value, size_t** out, size_t* outsize) {
size_t i;
ZOPFLI_APPEND_DATA(value, out, outsize);
for (i = 0; i + 1 < *outsize; i++) {
if ((*out)[i] > value) {
size_t j;
for (j = *outsize - 1; j > i; j--) {
(*out)[j] = (*out)[j - 1];
}
(*out)[i] = value;
break;
}
}
}
/*
Prints the block split points as decimal and hex values in the terminal.
*/
static void PrintBlockSplitPoints(const ZopfliLZ77Store* lz77,
const size_t* lz77splitpoints,
size_t nlz77points) {
size_t* splitpoints = 0;
size_t npoints = 0;
size_t i;
/* The input is given as lz77 indices, but we want to see the uncompressed
index values. */
size_t pos = 0;
if (nlz77points > 0) {
for (i = 0; i < lz77->size; i++) {
size_t length = lz77->dists[i] == 0 ? 1 : lz77->litlens[i];
if (lz77splitpoints[npoints] == i) {
ZOPFLI_APPEND_DATA(pos, &splitpoints, &npoints);
if (npoints == nlz77points) break;
}
pos += length;
}
}
assert(npoints == nlz77points);
fprintf(stderr, "block split points: ");
for (i = 0; i < npoints; i++) {
fprintf(stderr, "%d ", (int)splitpoints[i]);
}
fprintf(stderr, "(hex:");
for (i = 0; i < npoints; i++) {
fprintf(stderr, " %x", (int)splitpoints[i]);
}
fprintf(stderr, ")\n");
free(splitpoints);
}
/*
Finds next block to try to split, the largest of the available ones.
The largest is chosen to make sure that if only a limited amount of blocks is
requested, their sizes are spread evenly.
lz77size: the size of the LL77 data, which is the size of the done array here.
done: array indicating which blocks starting at that position are no longer
splittable (splitting them increases rather than decreases cost).
splitpoints: the splitpoints found so far.
npoints: the amount of splitpoints found so far.
lstart: output variable, giving start of block.
lend: output variable, giving end of block.
returns 1 if a block was found, 0 if no block found (all are done).
*/
static int FindLargestSplittableBlock(
size_t lz77size, const unsigned char* done,
const size_t* splitpoints, size_t npoints,
size_t* lstart, size_t* lend) {
size_t longest = 0;
int found = 0;
size_t i;
for (i = 0; i <= npoints; i++) {
size_t start = i == 0 ? 0 : splitpoints[i - 1];
size_t end = i == npoints ? lz77size - 1 : splitpoints[i];
if (!done[start] && end - start > longest) {
*lstart = start;
*lend = end;
found = 1;
longest = end - start;
}
}
return found;
}
void ZopfliBlockSplitLZ77(const ZopfliOptions* options,
const ZopfliLZ77Store* lz77, size_t maxblocks,
size_t** splitpoints, size_t* npoints) {
size_t lstart, lend;
size_t i;
size_t llpos = 0;
size_t numblocks = 1;
unsigned char* done;
double splitcost, origcost;
if (lz77->size < 10) return; /* This code fails on tiny files. */
done = (unsigned char*)malloc(lz77->size);
if (!done) exit(-1); /* Allocation failed. */
for (i = 0; i < lz77->size; i++) done[i] = 0;
lstart = 0;
lend = lz77->size;
for (;;) {
SplitCostContext c;
if (maxblocks > 0 && numblocks >= maxblocks) {
break;
}
c.lz77 = lz77;
c.start = lstart;
c.end = lend;
assert(lstart < lend);
llpos = FindMinimum(SplitCost, &c, lstart + 1, lend, &splitcost);
assert(llpos > lstart);
assert(llpos < lend);
origcost = EstimateCost(lz77, lstart, lend);
if (splitcost > origcost || llpos == lstart + 1 || llpos == lend) {
done[lstart] = 1;
} else {
AddSorted(llpos, splitpoints, npoints);
numblocks++;
}
if (!FindLargestSplittableBlock(
lz77->size, done, *splitpoints, *npoints, &lstart, &lend)) {
break; /* No further split will probably reduce compression. */
}
if (lend - lstart < 10) {
break;
}
}
if (options->verbose) {
PrintBlockSplitPoints(lz77, *splitpoints, *npoints);
}
free(done);
}
void ZopfliBlockSplit(const ZopfliOptions* options,
const unsigned char* in, size_t instart, size_t inend,
size_t maxblocks, size_t** splitpoints, size_t* npoints) {
size_t pos = 0;
size_t i;
ZopfliBlockState s;
size_t* lz77splitpoints = 0;
size_t nlz77points = 0;
ZopfliLZ77Store store;
ZopfliHash hash;
ZopfliHash* h = &hash;
ZopfliInitLZ77Store(in, &store);
ZopfliInitBlockState(options, instart, inend, 0, &s);
ZopfliAllocHash(ZOPFLI_WINDOW_SIZE, h);
*npoints = 0;
*splitpoints = 0;
/* Unintuitively, Using a simple LZ77 method here instead of ZopfliLZ77Optimal
results in better blocks. */
ZopfliLZ77Greedy(&s, in, instart, inend, &store, h);
ZopfliBlockSplitLZ77(options,
&store, maxblocks,
&lz77splitpoints, &nlz77points);
/* Convert LZ77 positions to positions in the uncompressed input. */
pos = instart;
if (nlz77points > 0) {
for (i = 0; i < store.size; i++) {
size_t length = store.dists[i] == 0 ? 1 : store.litlens[i];
if (lz77splitpoints[*npoints] == i) {
ZOPFLI_APPEND_DATA(pos, splitpoints, npoints);
if (*npoints == nlz77points) break;
}
pos += length;
}
}
assert(*npoints == nlz77points);
free(lz77splitpoints);
ZopfliCleanBlockState(&s);
ZopfliCleanLZ77Store(&store);
ZopfliCleanHash(h);
}
void ZopfliBlockSplitSimple(const unsigned char* in,
size_t instart, size_t inend,
size_t blocksize,
size_t** splitpoints, size_t* npoints) {
size_t i = instart;
while (i < inend) {
ZOPFLI_APPEND_DATA(i, splitpoints, npoints);
i += blocksize;
}
(void)in;
}
tools/bible/zopfli4k/blocksplitter.h
+73
View File
@@ -0,0 +1,73 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Functions to choose good boundaries for block splitting. Deflate allows encoding
the data in multiple blocks, with a separate Huffman tree for each block. The
Huffman tree itself requires some bytes to encode, so by choosing certain
blocks, you can either hurt, or enhance compression. These functions choose good
ones that enhance it.
*/
#ifndef ZOPFLI_BLOCKSPLITTER_H_
#define ZOPFLI_BLOCKSPLITTER_H_
#include <stdlib.h>
#include "lz77.h"
#include "zopfli.h"
/*
Does blocksplitting on LZ77 data.
The output splitpoints are indices in the LZ77 data.
maxblocks: set a limit to the amount of blocks. Set to 0 to mean no limit.
*/
void ZopfliBlockSplitLZ77(const ZopfliOptions* options,
const ZopfliLZ77Store* lz77, size_t maxblocks,
size_t** splitpoints, size_t* npoints);
/*
Does blocksplitting on uncompressed data.
The output splitpoints are indices in the uncompressed bytes.
options: general program options.
in: uncompressed input data
instart: where to start splitting
inend: where to end splitting (not inclusive)
maxblocks: maximum amount of blocks to split into, or 0 for no limit
splitpoints: dynamic array to put the resulting split point coordinates into.
The coordinates are indices in the input array.
npoints: pointer to amount of splitpoints, for the dynamic array. The amount of
blocks is the amount of splitpoitns + 1.
*/
void ZopfliBlockSplit(const ZopfliOptions* options,
const unsigned char* in, size_t instart, size_t inend,
size_t maxblocks, size_t** splitpoints, size_t* npoints);
/*
Divides the input into equal blocks, does not even take LZ77 lengths into
account.
*/
void ZopfliBlockSplitSimple(const unsigned char* in,
size_t instart, size_t inend,
size_t blocksize,
size_t** splitpoints, size_t* npoints);
#endif /* ZOPFLI_BLOCKSPLITTER_H_ */
tools/bible/zopfli4k/cache.c
+125
View File
@@ -0,0 +1,125 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "cache.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
void ZopfliInitCache(size_t blocksize, ZopfliLongestMatchCache* lmc) {
size_t i;
lmc->length = (unsigned short*)malloc(sizeof(unsigned short) * blocksize);
lmc->dist = (unsigned short*)malloc(sizeof(unsigned short) * blocksize);
/* Rather large amount of memory. */
lmc->sublen = (unsigned char*)malloc(ZOPFLI_CACHE_LENGTH * 3 * blocksize);
if(lmc->sublen == NULL) {
fprintf(stderr,
"Error: Out of memory. Tried allocating %lu bytes of memory.\n",
(unsigned long)ZOPFLI_CACHE_LENGTH * 3 * blocksize);
exit (EXIT_FAILURE);
}
/* length > 0 and dist 0 is invalid combination, which indicates on purpose
that this cache value is not filled in yet. */
for (i = 0; i < blocksize; i++) lmc->length[i] = 1;
for (i = 0; i < blocksize; i++) lmc->dist[i] = 0;
for (i = 0; i < ZOPFLI_CACHE_LENGTH * blocksize * 3; i++) lmc->sublen[i] = 0;
}
void ZopfliCleanCache(ZopfliLongestMatchCache* lmc) {
free(lmc->length);
free(lmc->dist);
free(lmc->sublen);
}
void ZopfliSublenToCache(const unsigned short* sublen,
size_t pos, size_t length,
ZopfliLongestMatchCache* lmc) {
size_t i;
size_t j = 0;
unsigned bestlength = 0;
unsigned char* cache;
#if ZOPFLI_CACHE_LENGTH == 0
return;
#endif
cache = &lmc->sublen[ZOPFLI_CACHE_LENGTH * pos * 3];
if (length < 3) return;
for (i = 3; i <= length; i++) {
if (i == length || sublen[i] != sublen[i + 1]) {
cache[j * 3] = i - 3;
cache[j * 3 + 1] = sublen[i] % 256;
cache[j * 3 + 2] = (sublen[i] >> 8) % 256;
bestlength = i;
j++;
if (j >= ZOPFLI_CACHE_LENGTH) break;
}
}
if (j < ZOPFLI_CACHE_LENGTH) {
assert(bestlength == length);
cache[(ZOPFLI_CACHE_LENGTH - 1) * 3] = bestlength - 3;
} else {
assert(bestlength <= length);
}
assert(bestlength == ZopfliMaxCachedSublen(lmc, pos, length));
}
void ZopfliCacheToSublen(const ZopfliLongestMatchCache* lmc,
size_t pos, size_t length,
unsigned short* sublen) {
size_t i, j;
unsigned maxlength = ZopfliMaxCachedSublen(lmc, pos, length);
unsigned prevlength = 0;
unsigned char* cache;
#if ZOPFLI_CACHE_LENGTH == 0
return;
#endif
if (length < 3) return;
cache = &lmc->sublen[ZOPFLI_CACHE_LENGTH * pos * 3];
for (j = 0; j < ZOPFLI_CACHE_LENGTH; j++) {
unsigned length = cache[j * 3] + 3;
unsigned dist = cache[j * 3 + 1] + 256 * cache[j * 3 + 2];
for (i = prevlength; i <= length; i++) {
sublen[i] = dist;
}
if (length == maxlength) break;
prevlength = length + 1;
}
}
/*
Returns the length up to which could be stored in the cache.
*/
unsigned ZopfliMaxCachedSublen(const ZopfliLongestMatchCache* lmc,
size_t pos, size_t length) {
unsigned char* cache;
#if ZOPFLI_CACHE_LENGTH == 0
return 0;
#endif
cache = &lmc->sublen[ZOPFLI_CACHE_LENGTH * pos * 3];
(void)length;
if (cache[1] == 0 && cache[2] == 0) return 0; /* No sublen cached. */
return cache[(ZOPFLI_CACHE_LENGTH - 1) * 3] + 3;
}
#endif /* ZOPFLI_LONGEST_MATCH_CACHE */
tools/bible/zopfli4k/cache.h
+66
View File
@@ -0,0 +1,66 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
The cache that speeds up ZopfliFindLongestMatch of lz77.c.
*/
#ifndef ZOPFLI_CACHE_H_
#define ZOPFLI_CACHE_H_
#include "util.h"
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
/*
Cache used by ZopfliFindLongestMatch to remember previously found length/dist
values.
This is needed because the squeeze runs will ask these values multiple times for
the same position.
Uses large amounts of memory, since it has to remember the distance belonging
to every possible shorter-than-the-best length (the so called "sublen" array).
*/
typedef struct ZopfliLongestMatchCache {
unsigned short* length;
unsigned short* dist;
unsigned char* sublen;
} ZopfliLongestMatchCache;
/* Initializes the ZopfliLongestMatchCache. */
void ZopfliInitCache(size_t blocksize, ZopfliLongestMatchCache* lmc);
/* Frees up the memory of the ZopfliLongestMatchCache. */
void ZopfliCleanCache(ZopfliLongestMatchCache* lmc);
/* Stores sublen array in the cache. */
void ZopfliSublenToCache(const unsigned short* sublen,
size_t pos, size_t length,
ZopfliLongestMatchCache* lmc);
/* Extracts sublen array from the cache. */
void ZopfliCacheToSublen(const ZopfliLongestMatchCache* lmc,
size_t pos, size_t length,
unsigned short* sublen);
/* Returns the length up to which could be stored in the cache. */
unsigned ZopfliMaxCachedSublen(const ZopfliLongestMatchCache* lmc,
size_t pos, size_t length);
#endif /* ZOPFLI_LONGEST_MATCH_CACHE */
#endif /* ZOPFLI_CACHE_H_ */
tools/bible/zopfli4k/deflate.c
+931
View File
@@ -0,0 +1,931 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "deflate.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include "blocksplitter.h"
#include "squeeze.h"
#include "symbols.h"
#include "tree.h"
/*
bp = bitpointer, always in range [0, 7].
The outsize is number of necessary bytes to encode the bits.
Given the value of bp and the amount of bytes, the amount of bits represented
is not simply bytesize * 8 + bp because even representing one bit requires a
whole byte. It is: (bp == 0) ? (bytesize * 8) : ((bytesize - 1) * 8 + bp)
*/
static void AddBit(int bit,
unsigned char* bp, unsigned char** out, size_t* outsize) {
if (*bp == 0) ZOPFLI_APPEND_DATA(0, out, outsize);
(*out)[*outsize - 1] |= bit << *bp;
*bp = (*bp + 1) & 7;
}
static void AddBits(unsigned symbol, unsigned length,
unsigned char* bp, unsigned char** out, size_t* outsize) {
/* TODO(lode): make more efficient (add more bits at once). */
unsigned i;
for (i = 0; i < length; i++) {
unsigned bit = (symbol >> i) & 1;
if (*bp == 0) ZOPFLI_APPEND_DATA(0, out, outsize);
(*out)[*outsize - 1] |= bit << *bp;
*bp = (*bp + 1) & 7;
}
}
/*
Adds bits, like AddBits, but the order is inverted. The deflate specification
uses both orders in one standard.
*/
static void AddHuffmanBits(unsigned symbol, unsigned length,
unsigned char* bp, unsigned char** out,
size_t* outsize) {
/* TODO(lode): make more efficient (add more bits at once). */
unsigned i;
for (i = 0; i < length; i++) {
unsigned bit = (symbol >> (length - i - 1)) & 1;
if (*bp == 0) ZOPFLI_APPEND_DATA(0, out, outsize);
(*out)[*outsize - 1] |= bit << *bp;
*bp = (*bp + 1) & 7;
}
}
/*
Ensures there are at least 2 distance codes to support buggy decoders.
Zlib 1.2.1 and below have a bug where it fails if there isn't at least 1
distance code (with length > 0), even though it's valid according to the
deflate spec to have 0 distance codes. On top of that, some mobile phones
require at least two distance codes. To support these decoders too (but
potentially at the cost of a few bytes), add dummy code lengths of 1.
References to this bug can be found in the changelog of
Zlib 1.2.2 and here: http://www.jonof.id.au/forum/index.php?topic=515.0.
d_lengths: the 32 lengths of the distance codes.
*/
static void PatchDistanceCodesForBuggyDecoders(unsigned* d_lengths) {
int num_dist_codes = 0; /* Amount of non-zero distance codes */
int i;
for (i = 0; i < 30 /* Ignore the two unused codes from the spec */; i++) {
if (d_lengths[i]) num_dist_codes++;
if (num_dist_codes >= 2) return; /* Two or more codes is fine. */
}
if (num_dist_codes == 0) {
d_lengths[0] = d_lengths[1] = 1;
} else if (num_dist_codes == 1) {
d_lengths[d_lengths[0] ? 1 : 0] = 1;
}
}
/*
Encodes the Huffman tree and returns how many bits its encoding takes. If out
is a null pointer, only returns the size and runs faster.
*/
static size_t EncodeTree(const unsigned* ll_lengths,
const unsigned* d_lengths,
int use_16, int use_17, int use_18,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
unsigned lld_total; /* Total amount of literal, length, distance codes. */
/* Runlength encoded version of lengths of litlen and dist trees. */
unsigned* rle = 0;
unsigned* rle_bits = 0; /* Extra bits for rle values 16, 17 and 18. */
size_t rle_size = 0; /* Size of rle array. */
size_t rle_bits_size = 0; /* Should have same value as rle_size. */
unsigned hlit = 29; /* 286 - 257 */
unsigned hdist = 29; /* 32 - 1, but gzip does not like hdist > 29.*/
unsigned hclen;
unsigned hlit2;
size_t i, j;
size_t clcounts[19];
unsigned clcl[19]; /* Code length code lengths. */
unsigned clsymbols[19];
/* The order in which code length code lengths are encoded as per deflate. */
static const unsigned order[19] = {
16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15
};
int size_only = !out;
size_t result_size = 0;
for(i = 0; i < 19; i++) clcounts[i] = 0;
/* Trim zeros. */
while (hlit > 0 && ll_lengths[257 + hlit - 1] == 0) hlit--;
while (hdist > 0 && d_lengths[1 + hdist - 1] == 0) hdist--;
hlit2 = hlit + 257;
lld_total = hlit2 + hdist + 1;
for (i = 0; i < lld_total; i++) {
/* This is an encoding of a huffman tree, so now the length is a symbol */
unsigned char symbol = i < hlit2 ? ll_lengths[i] : d_lengths[i - hlit2];
unsigned count = 1;
if(use_16 || (symbol == 0 && (use_17 || use_18))) {
for (j = i + 1; j < lld_total && symbol ==
(j < hlit2 ? ll_lengths[j] : d_lengths[j - hlit2]); j++) {
count++;
}
}
i += count - 1;
/* Repetitions of zeroes */
if (symbol == 0 && count >= 3) {
if (use_18) {
while (count >= 11) {
unsigned count2 = count > 138 ? 138 : count;
if (!size_only) {
ZOPFLI_APPEND_DATA(18, &rle, &rle_size);
ZOPFLI_APPEND_DATA(count2 - 11, &rle_bits, &rle_bits_size);
}
clcounts[18]++;
count -= count2;
}
}
if (use_17) {
while (count >= 3) {
unsigned count2 = count > 10 ? 10 : count;
if (!size_only) {
ZOPFLI_APPEND_DATA(17, &rle, &rle_size);
ZOPFLI_APPEND_DATA(count2 - 3, &rle_bits, &rle_bits_size);
}
clcounts[17]++;
count -= count2;
}
}
}
/* Repetitions of any symbol */
if (use_16 && count >= 4) {
count--; /* Since the first one is hardcoded. */
clcounts[symbol]++;
if (!size_only) {
ZOPFLI_APPEND_DATA(symbol, &rle, &rle_size);
ZOPFLI_APPEND_DATA(0, &rle_bits, &rle_bits_size);
}
while (count >= 3) {
unsigned count2 = count > 6 ? 6 : count;
if (!size_only) {
ZOPFLI_APPEND_DATA(16, &rle, &rle_size);
ZOPFLI_APPEND_DATA(count2 - 3, &rle_bits, &rle_bits_size);
}
clcounts[16]++;
count -= count2;
}
}
/* No or insufficient repetition */
clcounts[symbol] += count;
while (count > 0) {
if (!size_only) {
ZOPFLI_APPEND_DATA(symbol, &rle, &rle_size);
ZOPFLI_APPEND_DATA(0, &rle_bits, &rle_bits_size);
}
count--;
}
}
ZopfliCalculateBitLengths(clcounts, 19, 7, clcl);
if (!size_only) ZopfliLengthsToSymbols(clcl, 19, 7, clsymbols);
hclen = 15;
/* Trim zeros. */
while (hclen > 0 && clcounts[order[hclen + 4 - 1]] == 0) hclen--;
if (!size_only) {
AddBits(hlit, 5, bp, out, outsize);
AddBits(hdist, 5, bp, out, outsize);
AddBits(hclen, 4, bp, out, outsize);
for (i = 0; i < hclen + 4; i++) {
AddBits(clcl[order[i]], 3, bp, out, outsize);
}
for (i = 0; i < rle_size; i++) {
unsigned symbol = clsymbols[rle[i]];
AddHuffmanBits(symbol, clcl[rle[i]], bp, out, outsize);
/* Extra bits. */
if (rle[i] == 16) AddBits(rle_bits[i], 2, bp, out, outsize);
else if (rle[i] == 17) AddBits(rle_bits[i], 3, bp, out, outsize);
else if (rle[i] == 18) AddBits(rle_bits[i], 7, bp, out, outsize);
}
}
result_size += 14; /* hlit, hdist, hclen bits */
result_size += (hclen + 4) * 3; /* clcl bits */
for(i = 0; i < 19; i++) {
result_size += clcl[i] * clcounts[i];
}
/* Extra bits. */
result_size += clcounts[16] * 2;
result_size += clcounts[17] * 3;
result_size += clcounts[18] * 7;
/* Note: in case of "size_only" these are null pointers so no effect. */
free(rle);
free(rle_bits);
return result_size;
}
static void AddDynamicTree(const unsigned* ll_lengths,
const unsigned* d_lengths,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
int i;
int best = 0;
size_t bestsize = 0;
for(i = 0; i < 8; i++) {
size_t size = EncodeTree(ll_lengths, d_lengths,
i & 1, i & 2, i & 4,
0, 0, 0);
if (bestsize == 0 || size < bestsize) {
bestsize = size;
best = i;
}
}
EncodeTree(ll_lengths, d_lengths,
best & 1, best & 2, best & 4,
bp, out, outsize);
}
/*
Gives the exact size of the tree, in bits, as it will be encoded in DEFLATE.
*/
static size_t CalculateTreeSize(const unsigned* ll_lengths,
const unsigned* d_lengths) {
size_t result = 0;
int i;
for(i = 0; i < 8; i++) {
size_t size = EncodeTree(ll_lengths, d_lengths,
i & 1, i & 2, i & 4,
0, 0, 0);
if (result == 0 || size < result) result = size;
}
return result;
}
/*
Adds all lit/len and dist codes from the lists as huffman symbols. Does not add
end code 256. expected_data_size is the uncompressed block size, used for
assert, but you can set it to 0 to not do the assertion.
*/
static void AddLZ77Data(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
size_t expected_data_size,
const unsigned* ll_symbols, const unsigned* ll_lengths,
const unsigned* d_symbols, const unsigned* d_lengths,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
size_t testlength = 0;
size_t i;
for (i = lstart; i < lend; i++) {
unsigned dist = lz77->dists[i];
unsigned litlen = lz77->litlens[i];
if (dist == 0) {
assert(litlen < 256);
assert(ll_lengths[litlen] > 0);
AddHuffmanBits(ll_symbols[litlen], ll_lengths[litlen], bp, out, outsize);
testlength++;
} else {
unsigned lls = ZopfliGetLengthSymbol(litlen);
unsigned ds = ZopfliGetDistSymbol(dist);
assert(litlen >= 3 && litlen <= 288);
assert(ll_lengths[lls] > 0);
assert(d_lengths[ds] > 0);
AddHuffmanBits(ll_symbols[lls], ll_lengths[lls], bp, out, outsize);
AddBits(ZopfliGetLengthExtraBitsValue(litlen),
ZopfliGetLengthExtraBits(litlen),
bp, out, outsize);
AddHuffmanBits(d_symbols[ds], d_lengths[ds], bp, out, outsize);
AddBits(ZopfliGetDistExtraBitsValue(dist),
ZopfliGetDistExtraBits(dist),
bp, out, outsize);
testlength += litlen;
}
}
assert(expected_data_size == 0 || testlength == expected_data_size);
}
static void GetFixedTree(unsigned* ll_lengths, unsigned* d_lengths) {
size_t i;
for (i = 0; i < 144; i++) ll_lengths[i] = 8;
for (i = 144; i < 256; i++) ll_lengths[i] = 9;
for (i = 256; i < 280; i++) ll_lengths[i] = 7;
for (i = 280; i < 288; i++) ll_lengths[i] = 8;
for (i = 0; i < 32; i++) d_lengths[i] = 5;
}
/*
Same as CalculateBlockSymbolSize, but for block size smaller than histogram
size.
*/
static size_t CalculateBlockSymbolSizeSmall(const unsigned* ll_lengths,
const unsigned* d_lengths,
const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
size_t result = 0;
size_t i;
for (i = lstart; i < lend; i++) {
assert(i < lz77->size);
assert(lz77->litlens[i] < 259);
if (lz77->dists[i] == 0) {
result += ll_lengths[lz77->litlens[i]];
} else {
int ll_symbol = ZopfliGetLengthSymbol(lz77->litlens[i]);
int d_symbol = ZopfliGetDistSymbol(lz77->dists[i]);
result += ll_lengths[ll_symbol];
result += d_lengths[d_symbol];
result += ZopfliGetLengthSymbolExtraBits(ll_symbol);
result += ZopfliGetDistSymbolExtraBits(d_symbol);
}
}
result += ll_lengths[256]; /*end symbol*/
return result;
}
/*
Same as CalculateBlockSymbolSize, but with the histogram provided by the caller.
*/
static size_t CalculateBlockSymbolSizeGivenCounts(const size_t* ll_counts,
const size_t* d_counts,
const unsigned* ll_lengths,
const unsigned* d_lengths,
const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
size_t result = 0;
size_t i;
if (lstart + ZOPFLI_NUM_LL * 3 > lend) {
return CalculateBlockSymbolSizeSmall(
ll_lengths, d_lengths, lz77, lstart, lend);
} else {
for (i = 0; i < 256; i++) {
result += ll_lengths[i] * ll_counts[i];
}
for (i = 257; i < 286; i++) {
result += ll_lengths[i] * ll_counts[i];
result += ZopfliGetLengthSymbolExtraBits(i) * ll_counts[i];
}
for (i = 0; i < 30; i++) {
result += d_lengths[i] * d_counts[i];
result += ZopfliGetDistSymbolExtraBits(i) * d_counts[i];
}
result += ll_lengths[256]; /*end symbol*/
return result;
}
}
/*
Calculates size of the part after the header and tree of an LZ77 block, in bits.
*/
static size_t CalculateBlockSymbolSize(const unsigned* ll_lengths,
const unsigned* d_lengths,
const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
if (lstart + ZOPFLI_NUM_LL * 3 > lend) {
return CalculateBlockSymbolSizeSmall(
ll_lengths, d_lengths, lz77, lstart, lend);
} else {
size_t ll_counts[ZOPFLI_NUM_LL];
size_t d_counts[ZOPFLI_NUM_D];
ZopfliLZ77GetHistogram(lz77, lstart, lend, ll_counts, d_counts);
return CalculateBlockSymbolSizeGivenCounts(
ll_counts, d_counts, ll_lengths, d_lengths, lz77, lstart, lend);
}
}
static size_t AbsDiff(size_t x, size_t y) {
if (x > y)
return x - y;
else
return y - x;
}
/*
Changes the population counts in a way that the consequent Huffman tree
compression, especially its rle-part, will be more likely to compress this data
more efficiently. length contains the size of the histogram.
*/
void OptimizeHuffmanForRle(int length, size_t* counts) {
int i, k, stride;
size_t symbol, sum, limit;
int* good_for_rle;
/* 1) We don't want to touch the trailing zeros. We may break the
rules of the format by adding more data in the distance codes. */
for (; length >= 0; --length) {
if (length == 0) {
return;
}
if (counts[length - 1] != 0) {
/* Now counts[0..length - 1] does not have trailing zeros. */
break;
}
}
/* 2) Let's mark all population counts that already can be encoded
with an rle code.*/
good_for_rle = (int*)malloc((unsigned)length * sizeof(int));
for (i = 0; i < length; ++i) good_for_rle[i] = 0;
/* Let's not spoil any of the existing good rle codes.
Mark any seq of 0's that is longer than 5 as a good_for_rle.
Mark any seq of non-0's that is longer than 7 as a good_for_rle.*/
symbol = counts[0];
stride = 0;
for (i = 0; i < length + 1; ++i) {
if (i == length || counts[i] != symbol) {
if ((symbol == 0 && stride >= 5) || (symbol != 0 && stride >= 7)) {
for (k = 0; k < stride; ++k) {
good_for_rle[i - k - 1] = 1;
}
}
stride = 1;
if (i != length) {
symbol = counts[i];
}
} else {
++stride;
}
}
/* 3) Let's replace those population counts that lead to more rle codes. */
stride = 0;
limit = counts[0];
sum = 0;
for (i = 0; i < length + 1; ++i) {
if (i == length || good_for_rle[i]
/* Heuristic for selecting the stride ranges to collapse. */
|| AbsDiff(counts[i], limit) >= 4) {
if (stride >= 4 || (stride >= 3 && sum == 0)) {
/* The stride must end, collapse what we have, if we have enough (4). */
int count = (sum + stride / 2) / stride;
if (count < 1) count = 1;
if (sum == 0) {
/* Don't make an all zeros stride to be upgraded to ones. */
count = 0;
}
for (k = 0; k < stride; ++k) {
/* We don't want to change value at counts[i],
that is already belonging to the next stride. Thus - 1. */
counts[i - k - 1] = count;
}
}
stride = 0;
sum = 0;
if (i < length - 3) {
/* All interesting strides have a count of at least 4,
at least when non-zeros. */
limit = (counts[i] + counts[i + 1] +
counts[i + 2] + counts[i + 3] + 2) / 4;
} else if (i < length) {
limit = counts[i];
} else {
limit = 0;
}
}
++stride;
if (i != length) {
sum += counts[i];
}
}
free(good_for_rle);
}
/*
Tries out OptimizeHuffmanForRle for this block, if the result is smaller,
uses it, otherwise keeps the original. Returns size of encoded tree and data in
bits, not including the 3-bit block header.
*/
static double TryOptimizeHuffmanForRle(
const ZopfliLZ77Store* lz77, size_t lstart, size_t lend,
const size_t* ll_counts, const size_t* d_counts,
unsigned* ll_lengths, unsigned* d_lengths) {
size_t ll_counts2[ZOPFLI_NUM_LL];
size_t d_counts2[ZOPFLI_NUM_D];
unsigned ll_lengths2[ZOPFLI_NUM_LL];
unsigned d_lengths2[ZOPFLI_NUM_D];
double treesize;
double datasize;
double treesize2;
double datasize2;
treesize = CalculateTreeSize(ll_lengths, d_lengths);
datasize = CalculateBlockSymbolSizeGivenCounts(ll_counts, d_counts,
ll_lengths, d_lengths, lz77, lstart, lend);
memcpy(ll_counts2, ll_counts, sizeof(ll_counts2));
memcpy(d_counts2, d_counts, sizeof(d_counts2));
OptimizeHuffmanForRle(ZOPFLI_NUM_LL, ll_counts2);
OptimizeHuffmanForRle(ZOPFLI_NUM_D, d_counts2);
ZopfliCalculateBitLengths(ll_counts2, ZOPFLI_NUM_LL, 15, ll_lengths2);
ZopfliCalculateBitLengths(d_counts2, ZOPFLI_NUM_D, 15, d_lengths2);
PatchDistanceCodesForBuggyDecoders(d_lengths2);
treesize2 = CalculateTreeSize(ll_lengths2, d_lengths2);
datasize2 = CalculateBlockSymbolSizeGivenCounts(ll_counts, d_counts,
ll_lengths2, d_lengths2, lz77, lstart, lend);
if (treesize2 + datasize2 < treesize + datasize) {
memcpy(ll_lengths, ll_lengths2, sizeof(ll_lengths2));
memcpy(d_lengths, d_lengths2, sizeof(d_lengths2));
return treesize2 + datasize2;
}
return treesize + datasize;
}
/*
Calculates the bit lengths for the symbols for dynamic blocks. Chooses bit
lengths that give the smallest size of tree encoding + encoding of all the
symbols to have smallest output size. This are not necessarily the ideal Huffman
bit lengths. Returns size of encoded tree and data in bits, not including the
3-bit block header.
*/
static double GetDynamicLengths(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
unsigned* ll_lengths, unsigned* d_lengths) {
size_t ll_counts[ZOPFLI_NUM_LL];
size_t d_counts[ZOPFLI_NUM_D];
ZopfliLZ77GetHistogram(lz77, lstart, lend, ll_counts, d_counts);
ll_counts[256] = 1; /* End symbol. */
ZopfliCalculateBitLengths(ll_counts, ZOPFLI_NUM_LL, 15, ll_lengths);
ZopfliCalculateBitLengths(d_counts, ZOPFLI_NUM_D, 15, d_lengths);
PatchDistanceCodesForBuggyDecoders(d_lengths);
return TryOptimizeHuffmanForRle(
lz77, lstart, lend, ll_counts, d_counts, ll_lengths, d_lengths);
}
double ZopfliCalculateBlockSize(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend, int btype) {
unsigned ll_lengths[ZOPFLI_NUM_LL];
unsigned d_lengths[ZOPFLI_NUM_D];
double result = 3; /* bfinal and btype bits */
if (btype == 0) {
size_t length = ZopfliLZ77GetByteRange(lz77, lstart, lend);
size_t rem = length % 65535;
size_t blocks = length / 65535 + (rem ? 1 : 0);
/* An uncompressed block must actually be split into multiple blocks if it's
larger than 65535 bytes long. Eeach block header is 5 bytes: 3 bits,
padding, LEN and NLEN (potential less padding for first one ignored). */
return blocks * 5 * 8 + length * 8;
} if (btype == 1) {
GetFixedTree(ll_lengths, d_lengths);
result += CalculateBlockSymbolSize(
ll_lengths, d_lengths, lz77, lstart, lend);
} else {
result += GetDynamicLengths(lz77, lstart, lend, ll_lengths, d_lengths);
}
return result;
}
double ZopfliCalculateBlockSizeAutoType(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
double uncompressedcost = ZopfliCalculateBlockSize(lz77, lstart, lend, 0);
/* Don't do the expensive fixed cost calculation for larger blocks that are
unlikely to use it. */
double fixedcost = (lz77->size > 1000) ?
uncompressedcost : ZopfliCalculateBlockSize(lz77, lstart, lend, 1);
double dyncost = ZopfliCalculateBlockSize(lz77, lstart, lend, 2);
return (uncompressedcost < fixedcost && uncompressedcost < dyncost)
? uncompressedcost
: (fixedcost < dyncost ? fixedcost : dyncost);
}
/* Since an uncompressed block can be max 65535 in size, it actually adds
multible blocks if needed. */
static void AddNonCompressedBlock(const ZopfliOptions* options, int final,
const unsigned char* in, size_t instart,
size_t inend,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
size_t pos = instart;
(void)options;
for (;;) {
size_t i;
unsigned short blocksize = 65535;
unsigned short nlen;
int currentfinal;
if (pos + blocksize > inend) blocksize = inend - pos;
currentfinal = pos + blocksize >= inend;
nlen = ~blocksize;
AddBit(final && currentfinal, bp, out, outsize);
/* BTYPE 00 */
AddBit(0, bp, out, outsize);
AddBit(0, bp, out, outsize);
/* Any bits of input up to the next byte boundary are ignored. */
*bp = 0;
ZOPFLI_APPEND_DATA(blocksize % 256, out, outsize);
ZOPFLI_APPEND_DATA((blocksize / 256) % 256, out, outsize);
ZOPFLI_APPEND_DATA(nlen % 256, out, outsize);
ZOPFLI_APPEND_DATA((nlen / 256) % 256, out, outsize);
for (i = 0; i < blocksize; i++) {
ZOPFLI_APPEND_DATA(in[pos + i], out, outsize);
}
if (currentfinal) break;
pos += blocksize;
}
}
/*
Adds a deflate block with the given LZ77 data to the output.
options: global program options
btype: the block type, must be 1 or 2
final: whether to set the "final" bit on this block, must be the last block
litlens: literal/length array of the LZ77 data, in the same format as in
ZopfliLZ77Store.
dists: distance array of the LZ77 data, in the same format as in
ZopfliLZ77Store.
lstart: where to start in the LZ77 data
lend: where to end in the LZ77 data (not inclusive)
expected_data_size: the uncompressed block size, used for assert, but you can
set it to 0 to not do the assertion.
bp: output bit pointer
out: dynamic output array to append to
outsize: dynamic output array size
*/
static void AddLZ77Block(const ZopfliOptions* options, int btype, int final,
const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
size_t expected_data_size,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
unsigned ll_lengths[ZOPFLI_NUM_LL];
unsigned d_lengths[ZOPFLI_NUM_D];
unsigned ll_symbols[ZOPFLI_NUM_LL];
unsigned d_symbols[ZOPFLI_NUM_D];
size_t detect_block_size = *outsize;
size_t compressed_size;
size_t uncompressed_size = 0;
size_t i;
if (btype == 0) {
size_t length = ZopfliLZ77GetByteRange(lz77, lstart, lend);
size_t pos = lstart == lend ? 0 : lz77->pos[lstart];
size_t end = pos + length;
AddNonCompressedBlock(options, final,
lz77->data, pos, end, bp, out, outsize);
return;
}
AddBit(final, bp, out, outsize);
AddBit(btype & 1, bp, out, outsize);
AddBit((btype & 2) >> 1, bp, out, outsize);
if (btype == 1) {
/* Fixed block. */
GetFixedTree(ll_lengths, d_lengths);
} else {
/* Dynamic block. */
unsigned detect_tree_size;
assert(btype == 2);
GetDynamicLengths(lz77, lstart, lend, ll_lengths, d_lengths);
detect_tree_size = *outsize;
AddDynamicTree(ll_lengths, d_lengths, bp, out, outsize);
if (options->verbose) {
fprintf(stderr, "treesize: %d\n", (int)(*outsize - detect_tree_size));
}
}
ZopfliLengthsToSymbols(ll_lengths, ZOPFLI_NUM_LL, 15, ll_symbols);
ZopfliLengthsToSymbols(d_lengths, ZOPFLI_NUM_D, 15, d_symbols);
detect_block_size = *outsize;
AddLZ77Data(lz77, lstart, lend, expected_data_size,
ll_symbols, ll_lengths, d_symbols, d_lengths,
bp, out, outsize);
/* End symbol. */
AddHuffmanBits(ll_symbols[256], ll_lengths[256], bp, out, outsize);
for (i = lstart; i < lend; i++) {
uncompressed_size += lz77->dists[i] == 0 ? 1 : lz77->litlens[i];
}
compressed_size = *outsize - detect_block_size;
if (options->verbose) {
fprintf(stderr, "compressed block size: %d (%dk) (unc: %d)\n",
(int)compressed_size, (int)(compressed_size / 1024),
(int)(uncompressed_size));
}
}
static void AddLZ77BlockAutoType(const ZopfliOptions* options, int final,
const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
size_t expected_data_size,
unsigned char* bp,
unsigned char** out, size_t* outsize) {
double uncompressedcost = ZopfliCalculateBlockSize(lz77, lstart, lend, 0);
double fixedcost = ZopfliCalculateBlockSize(lz77, lstart, lend, 1);
double dyncost = ZopfliCalculateBlockSize(lz77, lstart, lend, 2);
/* Whether to perform the expensive calculation of creating an optimal block
with fixed huffman tree to check if smaller. Only do this for small blocks or
blocks which already are pretty good with fixed huffman tree. */
int expensivefixed = (lz77->size < 1000) || fixedcost <= dyncost * 1.1;
ZopfliLZ77Store fixedstore;
if (lstart == lend) {
/* Smallest empty block is represented by fixed block */
AddBits(final, 1, bp, out, outsize);
AddBits(1, 2, bp, out, outsize); /* btype 01 */
AddBits(0, 7, bp, out, outsize); /* end symbol has code 0000000 */
return;
}
ZopfliInitLZ77Store(lz77->data, &fixedstore);
if (expensivefixed) {
/* Recalculate the LZ77 with ZopfliLZ77OptimalFixed */
size_t instart = lz77->pos[lstart];
size_t inend = instart + ZopfliLZ77GetByteRange(lz77, lstart, lend);
ZopfliBlockState s;
ZopfliInitBlockState(options, instart, inend, 1, &s);
ZopfliLZ77OptimalFixed(&s, lz77->data, instart, inend, &fixedstore);
fixedcost = ZopfliCalculateBlockSize(&fixedstore, 0, fixedstore.size, 1);
ZopfliCleanBlockState(&s);
}
if (uncompressedcost < fixedcost && uncompressedcost < dyncost) {
AddLZ77Block(options, 0, final, lz77, lstart, lend,
expected_data_size, bp, out, outsize);
} else if (fixedcost < dyncost) {
if (expensivefixed) {
AddLZ77Block(options, 1, final, &fixedstore, 0, fixedstore.size,
expected_data_size, bp, out, outsize);
} else {
AddLZ77Block(options, 1, final, lz77, lstart, lend,
expected_data_size, bp, out, outsize);
}
} else {
AddLZ77Block(options, 2, final, lz77, lstart, lend,
expected_data_size, bp, out, outsize);
}
ZopfliCleanLZ77Store(&fixedstore);
}
/*
Deflate a part, to allow ZopfliDeflate() to use multiple master blocks if
needed.
It is possible to call this function multiple times in a row, shifting
instart and inend to next bytes of the data. If instart is larger than 0, then
previous bytes are used as the initial dictionary for LZ77.
This function will usually output multiple deflate blocks. If final is 1, then
the final bit will be set on the last block.
*/
void ZopfliDeflatePart(const ZopfliOptions* options, int btype, int final,
const unsigned char* in, size_t instart, size_t inend,
unsigned char* bp, unsigned char** out,
size_t* outsize) {
size_t i;
/* byte coordinates rather than lz77 index */
size_t* splitpoints_uncompressed = 0;
size_t npoints = 0;
size_t* splitpoints = 0;
double totalcost = 0;
ZopfliLZ77Store lz77;
/* If btype=2 is specified, it tries all block types. If a lesser btype is
given, then however it forces that one. Neither of the lesser types needs
block splitting as they have no dynamic huffman trees. */
if (btype == 0) {
AddNonCompressedBlock(options, final, in, instart, inend, bp, out, outsize);
return;
} else if (btype == 1) {
ZopfliLZ77Store store;
ZopfliBlockState s;
ZopfliInitLZ77Store(in, &store);
ZopfliInitBlockState(options, instart, inend, 1, &s);
ZopfliLZ77OptimalFixed(&s, in, instart, inend, &store);
AddLZ77Block(options, btype, final, &store, 0, store.size, 0,
bp, out, outsize);
ZopfliCleanBlockState(&s);
ZopfliCleanLZ77Store(&store);
return;
}
if (options->blocksplitting) {
ZopfliBlockSplit(options, in, instart, inend,
options->blocksplittingmax,
&splitpoints_uncompressed, &npoints);
splitpoints = (size_t*)malloc(sizeof(*splitpoints) * npoints);
}
ZopfliInitLZ77Store(in, &lz77);
for (i = 0; i <= npoints; i++) {
size_t start = i == 0 ? instart : splitpoints_uncompressed[i - 1];
size_t end = i == npoints ? inend : splitpoints_uncompressed[i];
ZopfliBlockState s;
ZopfliLZ77Store store;
ZopfliInitLZ77Store(in, &store);
ZopfliInitBlockState(options, start, end, 1, &s);
ZopfliLZ77Optimal(&s, in, start, end, options->numiterations, &store);
totalcost += ZopfliCalculateBlockSizeAutoType(&store, 0, store.size);
ZopfliAppendLZ77Store(&store, &lz77);
if (i < npoints) splitpoints[i] = lz77.size;
ZopfliCleanBlockState(&s);
ZopfliCleanLZ77Store(&store);
}
/* Second block splitting attempt */
if (options->blocksplitting && npoints > 1) {
size_t* splitpoints2 = 0;
size_t npoints2 = 0;
double totalcost2 = 0;
ZopfliBlockSplitLZ77(options, &lz77,
options->blocksplittingmax, &splitpoints2, &npoints2);
for (i = 0; i <= npoints2; i++) {
size_t start = i == 0 ? 0 : splitpoints2[i - 1];
size_t end = i == npoints2 ? lz77.size : splitpoints2[i];
totalcost2 += ZopfliCalculateBlockSizeAutoType(&lz77, start, end);
}
if (totalcost2 < totalcost) {
free(splitpoints);
splitpoints = splitpoints2;
npoints = npoints2;
} else {
free(splitpoints2);
}
}
for (i = 0; i <= npoints; i++) {
size_t start = i == 0 ? 0 : splitpoints[i - 1];
size_t end = i == npoints ? lz77.size : splitpoints[i];
AddLZ77BlockAutoType(options, i == npoints && final,
&lz77, start, end, 0,
bp, out, outsize);
}
ZopfliCleanLZ77Store(&lz77);
free(splitpoints);
free(splitpoints_uncompressed);
}
void ZopfliDeflate(const ZopfliOptions* options, int btype, int final,
const unsigned char* in, size_t insize,
unsigned char* bp, unsigned char** out, size_t* outsize) {
size_t offset = *outsize;
#if ZOPFLI_MASTER_BLOCK_SIZE == 0
ZopfliDeflatePart(options, btype, final, in, 0, insize, bp, out, outsize);
#else
size_t i = 0;
do {
int masterfinal = (i + ZOPFLI_MASTER_BLOCK_SIZE >= insize);
int final2 = final && masterfinal;
size_t size = masterfinal ? insize - i : ZOPFLI_MASTER_BLOCK_SIZE;
ZopfliDeflatePart(options, btype, final2,
in, i, i + size, bp, out, outsize);
i += size;
} while (i < insize);
#endif
if (options->verbose) {
fprintf(stderr,
"Original Size: %lu, Deflate: %lu, Compression: %f%% Removed\n",
(unsigned long)insize, (unsigned long)(*outsize - offset),
100.0 * (double)(insize - (*outsize - offset)) / (double)insize);
}
}
tools/bible/zopfli4k/deflate.h
+92
View File
@@ -0,0 +1,92 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#ifndef ZOPFLI_DEFLATE_H_
#define ZOPFLI_DEFLATE_H_
/*
Functions to compress according to the DEFLATE specification, using the
"squeeze" LZ77 compression backend.
*/
#include "lz77.h"
#include "zopfli.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
Compresses according to the deflate specification and append the compressed
result to the output.
This function will usually output multiple deflate blocks. If final is 1, then
the final bit will be set on the last block.
options: global program options
btype: the deflate block type. Use 2 for best compression.
-0: non compressed blocks (00)
-1: blocks with fixed tree (01)
-2: blocks with dynamic tree (10)
final: whether this is the last section of the input, sets the final bit to the
last deflate block.
in: the input bytes
insize: number of input bytes
bp: bit pointer for the output array. This must initially be 0, and for
consecutive calls must be reused (it can have values from 0-7). This is
because deflate appends blocks as bit-based data, rather than on byte
boundaries.
out: pointer to the dynamic output array to which the result is appended. Must
be freed after use.
outsize: pointer to the dynamic output array size.
*/
void ZopfliDeflate(const ZopfliOptions* options, int btype, int final,
const unsigned char* in, size_t insize,
unsigned char* bp, unsigned char** out, size_t* outsize);
/*
Like ZopfliDeflate, but allows to specify start and end byte with instart and
inend. Only that part is compressed, but earlier bytes are still used for the
back window.
*/
void ZopfliDeflatePart(const ZopfliOptions* options, int btype, int final,
const unsigned char* in, size_t instart, size_t inend,
unsigned char* bp, unsigned char** out,
size_t* outsize);
/*
Calculates block size in bits.
litlens: lz77 lit/lengths
dists: ll77 distances
lstart: start of block
lend: end of block (not inclusive)
*/
double ZopfliCalculateBlockSize(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend, int btype);
/*
Calculates block size in bits, automatically using the best btype.
*/
double ZopfliCalculateBlockSizeAutoType(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* ZOPFLI_DEFLATE_H_ */
tools/bible/zopfli4k/gzip_container.c
+124
View File
@@ -0,0 +1,124 @@
/*
Copyright 2013 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "gzip_container.h"
#include "util.h"
#include <stdio.h>
#include "deflate.h"
/* CRC polynomial: 0xedb88320 */
static const unsigned long crc32_table[256] = {
0u, 1996959894u, 3993919788u, 2567524794u, 124634137u, 1886057615u,
3915621685u, 2657392035u, 249268274u, 2044508324u, 3772115230u, 2547177864u,
162941995u, 2125561021u, 3887607047u, 2428444049u, 498536548u, 1789927666u,
4089016648u, 2227061214u, 450548861u, 1843258603u, 4107580753u, 2211677639u,
325883990u, 1684777152u, 4251122042u, 2321926636u, 335633487u, 1661365465u,
4195302755u, 2366115317u, 997073096u, 1281953886u, 3579855332u, 2724688242u,
1006888145u, 1258607687u, 3524101629u, 2768942443u, 901097722u, 1119000684u,
3686517206u, 2898065728u, 853044451u, 1172266101u, 3705015759u, 2882616665u,
651767980u, 1373503546u, 3369554304u, 3218104598u, 565507253u, 1454621731u,
3485111705u, 3099436303u, 671266974u, 1594198024u, 3322730930u, 2970347812u,
795835527u, 1483230225u, 3244367275u, 3060149565u, 1994146192u, 31158534u,
2563907772u, 4023717930u, 1907459465u, 112637215u, 2680153253u, 3904427059u,
2013776290u, 251722036u, 2517215374u, 3775830040u, 2137656763u, 141376813u,
2439277719u, 3865271297u, 1802195444u, 476864866u, 2238001368u, 4066508878u,
1812370925u, 453092731u, 2181625025u, 4111451223u, 1706088902u, 314042704u,
2344532202u, 4240017532u, 1658658271u, 366619977u, 2362670323u, 4224994405u,
1303535960u, 984961486u, 2747007092u, 3569037538u, 1256170817u, 1037604311u,
2765210733u, 3554079995u, 1131014506u, 879679996u, 2909243462u, 3663771856u,
1141124467u, 855842277u, 2852801631u, 3708648649u, 1342533948u, 654459306u,
3188396048u, 3373015174u, 1466479909u, 544179635u, 3110523913u, 3462522015u,
1591671054u, 702138776u, 2966460450u, 3352799412u, 1504918807u, 783551873u,
3082640443u, 3233442989u, 3988292384u, 2596254646u, 62317068u, 1957810842u,
3939845945u, 2647816111u, 81470997u, 1943803523u, 3814918930u, 2489596804u,
225274430u, 2053790376u, 3826175755u, 2466906013u, 167816743u, 2097651377u,
4027552580u, 2265490386u, 503444072u, 1762050814u, 4150417245u, 2154129355u,
426522225u, 1852507879u, 4275313526u, 2312317920u, 282753626u, 1742555852u,
4189708143u, 2394877945u, 397917763u, 1622183637u, 3604390888u, 2714866558u,
953729732u, 1340076626u, 3518719985u, 2797360999u, 1068828381u, 1219638859u,
3624741850u, 2936675148u, 906185462u, 1090812512u, 3747672003u, 2825379669u,
829329135u, 1181335161u, 3412177804u, 3160834842u, 628085408u, 1382605366u,
3423369109u, 3138078467u, 570562233u, 1426400815u, 3317316542u, 2998733608u,
733239954u, 1555261956u, 3268935591u, 3050360625u, 752459403u, 1541320221u,
2607071920u, 3965973030u, 1969922972u, 40735498u, 2617837225u, 3943577151u,
1913087877u, 83908371u, 2512341634u, 3803740692u, 2075208622u, 213261112u,
2463272603u, 3855990285u, 2094854071u, 198958881u, 2262029012u, 4057260610u,
1759359992u, 534414190u, 2176718541u, 4139329115u, 1873836001u, 414664567u,
2282248934u, 4279200368u, 1711684554u, 285281116u, 2405801727u, 4167216745u,
1634467795u, 376229701u, 2685067896u, 3608007406u, 1308918612u, 956543938u,
2808555105u, 3495958263u, 1231636301u, 1047427035u, 2932959818u, 3654703836u,
1088359270u, 936918000u, 2847714899u, 3736837829u, 1202900863u, 817233897u,
3183342108u, 3401237130u, 1404277552u, 615818150u, 3134207493u, 3453421203u,
1423857449u, 601450431u, 3009837614u, 3294710456u, 1567103746u, 711928724u,
3020668471u, 3272380065u, 1510334235u, 755167117u
};
/* Returns the CRC32 */
static unsigned long CRC(const unsigned char* data, size_t size) {
unsigned long result = 0xffffffffu;
for (; size > 0; size--) {
result = crc32_table[(result ^ *(data++)) & 0xff] ^ (result >> 8);
}
return result ^ 0xffffffffu;
}
/* Compresses the data according to the gzip specification, RFC 1952. */
void ZopfliGzipCompress(const ZopfliOptions* options,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize) {
unsigned long crcvalue = CRC(in, insize);
unsigned char bp = 0;
ZOPFLI_APPEND_DATA(31, out, outsize); /* ID1 */
ZOPFLI_APPEND_DATA(139, out, outsize); /* ID2 */
ZOPFLI_APPEND_DATA(8, out, outsize); /* CM */
ZOPFLI_APPEND_DATA(0, out, outsize); /* FLG */
/* MTIME */
ZOPFLI_APPEND_DATA(0, out, outsize);
ZOPFLI_APPEND_DATA(0, out, outsize);
ZOPFLI_APPEND_DATA(0, out, outsize);
ZOPFLI_APPEND_DATA(0, out, outsize);
ZOPFLI_APPEND_DATA(2, out, outsize); /* XFL, 2 indicates best compression. */
ZOPFLI_APPEND_DATA(3, out, outsize); /* OS follows Unix conventions. */
ZopfliDeflate(options, 2 /* Dynamic block */, 1,
in, insize, &bp, out, outsize);
/* CRC */
ZOPFLI_APPEND_DATA(crcvalue % 256, out, outsize);
ZOPFLI_APPEND_DATA((crcvalue >> 8) % 256, out, outsize);
ZOPFLI_APPEND_DATA((crcvalue >> 16) % 256, out, outsize);
ZOPFLI_APPEND_DATA((crcvalue >> 24) % 256, out, outsize);
/* ISIZE */
ZOPFLI_APPEND_DATA(insize % 256, out, outsize);
ZOPFLI_APPEND_DATA((insize >> 8) % 256, out, outsize);
ZOPFLI_APPEND_DATA((insize >> 16) % 256, out, outsize);
ZOPFLI_APPEND_DATA((insize >> 24) % 256, out, outsize);
if (options->verbose) {
fprintf(stderr,
"Original Size: %d, Gzip: %d, Compression: %f%% Removed\n",
(int)insize, (int)*outsize,
100.0 * (double)(insize - *outsize) / (double)insize);
}
}
tools/bible/zopfli4k/gzip_container.h
+50
View File
@@ -0,0 +1,50 @@
/*
Copyright 2013 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#ifndef ZOPFLI_GZIP_H_
#define ZOPFLI_GZIP_H_
/*
Functions to compress according to the Gzip specification.
*/
#include "zopfli.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
Compresses according to the gzip specification and append the compressed
result to the output.
options: global program options
out: pointer to the dynamic output array to which the result is appended. Must
be freed after use.
outsize: pointer to the dynamic output array size.
*/
void ZopfliGzipCompress(const ZopfliOptions* options,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* ZOPFLI_GZIP_H_ */
tools/bible/zopfli4k/hash.c
+143
View File
@@ -0,0 +1,143 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "hash.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#define HASH_SHIFT 5
#define HASH_MASK 32767
void ZopfliAllocHash(size_t window_size, ZopfliHash* h) {
h->head = (int*)malloc(sizeof(*h->head) * 65536);
h->prev = (unsigned short*)malloc(sizeof(*h->prev) * window_size);
h->hashval = (int*)malloc(sizeof(*h->hashval) * window_size);
#ifdef ZOPFLI_HASH_SAME
h->same = (unsigned short*)malloc(sizeof(*h->same) * window_size);
#endif
#ifdef ZOPFLI_HASH_SAME_HASH
h->head2 = (int*)malloc(sizeof(*h->head2) * 65536);
h->prev2 = (unsigned short*)malloc(sizeof(*h->prev2) * window_size);
h->hashval2 = (int*)malloc(sizeof(*h->hashval2) * window_size);
#endif
}
void ZopfliResetHash(size_t window_size, ZopfliHash* h) {
size_t i;
h->val = 0;
for (i = 0; i < 65536; i++) {
h->head[i] = -1; /* -1 indicates no head so far. */
}
for (i = 0; i < window_size; i++) {
h->prev[i] = i; /* If prev[j] == j, then prev[j] is uninitialized. */
h->hashval[i] = -1;
}
#ifdef ZOPFLI_HASH_SAME
for (i = 0; i < window_size; i++) {
h->same[i] = 0;
}
#endif
#ifdef ZOPFLI_HASH_SAME_HASH
h->val2 = 0;
for (i = 0; i < 65536; i++) {
h->head2[i] = -1;
}
for (i = 0; i < window_size; i++) {
h->prev2[i] = i;
h->hashval2[i] = -1;
}
#endif
}
void ZopfliCleanHash(ZopfliHash* h) {
free(h->head);
free(h->prev);
free(h->hashval);
#ifdef ZOPFLI_HASH_SAME_HASH
free(h->head2);
free(h->prev2);
free(h->hashval2);
#endif
#ifdef ZOPFLI_HASH_SAME
free(h->same);
#endif
}
/*
Update the sliding hash value with the given byte. All calls to this function
must be made on consecutive input characters. Since the hash value exists out
of multiple input bytes, a few warmups with this function are needed initially.
*/
static void UpdateHashValue(ZopfliHash* h, unsigned char c) {
h->val = (((h->val) << HASH_SHIFT) ^ (c)) & HASH_MASK;
}
void ZopfliUpdateHash(const unsigned char* array, size_t pos, size_t end,
ZopfliHash* h) {
unsigned short hpos = pos & ZOPFLI_WINDOW_MASK;
#ifdef ZOPFLI_HASH_SAME
size_t amount = 0;
#endif
UpdateHashValue(h, pos + ZOPFLI_MIN_MATCH <= end ?
array[pos + ZOPFLI_MIN_MATCH - 1] : 0);
h->hashval[hpos] = h->val;
if (h->head[h->val] != -1 && h->hashval[h->head[h->val]] == h->val) {
h->prev[hpos] = h->head[h->val];
}
else h->prev[hpos] = hpos;
h->head[h->val] = hpos;
#ifdef ZOPFLI_HASH_SAME
/* Update "same". */
if (h->same[(pos - 1) & ZOPFLI_WINDOW_MASK] > 1) {
amount = h->same[(pos - 1) & ZOPFLI_WINDOW_MASK] - 1;
}
while (pos + amount + 1 < end &&
array[pos] == array[pos + amount + 1] && amount < (unsigned short)(-1)) {
amount++;
}
h->same[hpos] = amount;
#endif
#ifdef ZOPFLI_HASH_SAME_HASH
h->val2 = ((h->same[hpos] - ZOPFLI_MIN_MATCH) & 255) ^ h->val;
h->hashval2[hpos] = h->val2;
if (h->head2[h->val2] != -1 && h->hashval2[h->head2[h->val2]] == h->val2) {
h->prev2[hpos] = h->head2[h->val2];
}
else h->prev2[hpos] = hpos;
h->head2[h->val2] = hpos;
#endif
}
void ZopfliWarmupHash(const unsigned char* array, size_t pos, size_t end,
ZopfliHash* h) {
UpdateHashValue(h, array[pos + 0]);
if (pos + 1 < end) UpdateHashValue(h, array[pos + 1]);
}
tools/bible/zopfli4k/hash.h
+73
View File
@@ -0,0 +1,73 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
The hash for ZopfliFindLongestMatch of lz77.c.
*/
#ifndef ZOPFLI_HASH_H_
#define ZOPFLI_HASH_H_
#include "util.h"
typedef struct ZopfliHash {
int* head; /* Hash value to index of its most recent occurrence. */
unsigned short* prev; /* Index to index of prev. occurrence of same hash. */
int* hashval; /* Index to hash value at this index. */
int val; /* Current hash value. */
#ifdef ZOPFLI_HASH_SAME_HASH
/* Fields with similar purpose as the above hash, but for the second hash with
a value that is calculated differently. */
int* head2; /* Hash value to index of its most recent occurrence. */
unsigned short* prev2; /* Index to index of prev. occurrence of same hash. */
int* hashval2; /* Index to hash value at this index. */
int val2; /* Current hash value. */
#endif
#ifdef ZOPFLI_HASH_SAME
unsigned short* same; /* Amount of repetitions of same byte after this .*/
#endif
} ZopfliHash;
/* Allocates ZopfliHash memory. */
void ZopfliAllocHash(size_t window_size, ZopfliHash* h);
/* Resets all fields of ZopfliHash. */
void ZopfliResetHash(size_t window_size, ZopfliHash* h);
/* Frees ZopfliHash memory. */
void ZopfliCleanHash(ZopfliHash* h);
/*
Updates the hash values based on the current position in the array. All calls
to this must be made for consecutive bytes.
*/
void ZopfliUpdateHash(const unsigned char* array, size_t pos, size_t end,
ZopfliHash* h);
/*
Prepopulates hash:
Fills in the initial values in the hash, before ZopfliUpdateHash can be used
correctly.
*/
void ZopfliWarmupHash(const unsigned char* array, size_t pos, size_t end,
ZopfliHash* h);
#endif /* ZOPFLI_HASH_H_ */
tools/bible/zopfli4k/katajainen.c
+262
View File
@@ -0,0 +1,262 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Bounded package merge algorithm, based on the paper
"A Fast and Space-Economical Algorithm for Length-Limited Coding
Jyrki Katajainen, Alistair Moffat, Andrew Turpin".
*/
#include "katajainen.h"
#include <assert.h>
#include <stdlib.h>
#include <limits.h>
typedef struct Node Node;
/*
Nodes forming chains. Also used to represent leaves.
*/
struct Node {
size_t weight; /* Total weight (symbol count) of this chain. */
Node* tail; /* Previous node(s) of this chain, or 0 if none. */
int count; /* Leaf symbol index, or number of leaves before this chain. */
};
/*
Memory pool for nodes.
*/
typedef struct NodePool {
Node* next; /* Pointer to a free node in the pool. */
} NodePool;
/*
Initializes a chain node with the given values and marks it as in use.
*/
static void InitNode(size_t weight, int count, Node* tail, Node* node) {
node->weight = weight;
node->count = count;
node->tail = tail;
}
/*
Performs a Boundary Package-Merge step. Puts a new chain in the given list. The
new chain is, depending on the weights, a leaf or a combination of two chains
from the previous list.
lists: The lists of chains.
maxbits: Number of lists.
leaves: The leaves, one per symbol.
numsymbols: Number of leaves.
pool: the node memory pool.
index: The index of the list in which a new chain or leaf is required.
*/
static void BoundaryPM(Node* (*lists)[2], Node* leaves, int numsymbols,
NodePool* pool, int index) {
Node* newchain;
Node* oldchain;
int lastcount = lists[index][1]->count; /* Count of last chain of list. */
if (index == 0 && lastcount >= numsymbols) return;
newchain = pool->next++;
oldchain = lists[index][1];
/* These are set up before the recursive calls below, so that there is a list
pointing to the new node, to let the garbage collection know it's in use. */
lists[index][0] = oldchain;
lists[index][1] = newchain;
if (index == 0) {
/* New leaf node in list 0. */
InitNode(leaves[lastcount].weight, lastcount + 1, 0, newchain);
} else {
size_t sum = lists[index - 1][0]->weight + lists[index - 1][1]->weight;
if (lastcount < numsymbols && sum > leaves[lastcount].weight) {
/* New leaf inserted in list, so count is incremented. */
InitNode(leaves[lastcount].weight, lastcount + 1, oldchain->tail,
newchain);
} else {
InitNode(sum, lastcount, lists[index - 1][1], newchain);
/* Two lookahead chains of previous list used up, create new ones. */
BoundaryPM(lists, leaves, numsymbols, pool, index - 1);
BoundaryPM(lists, leaves, numsymbols, pool, index - 1);
}
}
}
static void BoundaryPMFinal(Node* (*lists)[2],
Node* leaves, int numsymbols, NodePool* pool, int index) {
int lastcount = lists[index][1]->count; /* Count of last chain of list. */
size_t sum = lists[index - 1][0]->weight + lists[index - 1][1]->weight;
if (lastcount < numsymbols && sum > leaves[lastcount].weight) {
Node* newchain = pool->next;
Node* oldchain = lists[index][1]->tail;
lists[index][1] = newchain;
newchain->count = lastcount + 1;
newchain->tail = oldchain;
} else {
lists[index][1]->tail = lists[index - 1][1];
}
}
/*
Initializes each list with as lookahead chains the two leaves with lowest
weights.
*/
static void InitLists(
NodePool* pool, const Node* leaves, int maxbits, Node* (*lists)[2]) {
int i;
Node* node0 = pool->next++;
Node* node1 = pool->next++;
InitNode(leaves[0].weight, 1, 0, node0);
InitNode(leaves[1].weight, 2, 0, node1);
for (i = 0; i < maxbits; i++) {
lists[i][0] = node0;
lists[i][1] = node1;
}
}
/*
Converts result of boundary package-merge to the bitlengths. The result in the
last chain of the last list contains the amount of active leaves in each list.
chain: Chain to extract the bit length from (last chain from last list).
*/
static void ExtractBitLengths(Node* chain, Node* leaves, unsigned* bitlengths) {
int counts[16] = {0};
unsigned end = 16;
unsigned ptr = 15;
unsigned value = 1;
Node* node;
int val;
for (node = chain; node; node = node->tail) {
counts[--end] = node->count;
}
val = counts[15];
while (ptr >= end) {
for (; val > counts[ptr - 1]; val--) {
bitlengths[leaves[val - 1].count] = value;
}
ptr--;
value++;
}
}
/*
Comparator for sorting the leaves. Has the function signature for qsort.
*/
static int LeafComparator(const void* a, const void* b) {
return ((const Node*)a)->weight - ((const Node*)b)->weight;
}
int ZopfliLengthLimitedCodeLengths(
const size_t* frequencies, int n, int maxbits, unsigned* bitlengths) {
NodePool pool;
int i;
int numsymbols = 0; /* Amount of symbols with frequency > 0. */
int numBoundaryPMRuns;
Node* nodes;
/* Array of lists of chains. Each list requires only two lookahead chains at
a time, so each list is a array of two Node*'s. */
Node* (*lists)[2];
/* One leaf per symbol. Only numsymbols leaves will be used. */
Node* leaves = (Node*)malloc(n * sizeof(*leaves));
/* Initialize all bitlengths at 0. */
for (i = 0; i < n; i++) {
bitlengths[i] = 0;
}
/* Count used symbols and place them in the leaves. */
for (i = 0; i < n; i++) {
if (frequencies[i]) {
leaves[numsymbols].weight = frequencies[i];
leaves[numsymbols].count = i; /* Index of symbol this leaf represents. */
numsymbols++;
}
}
/* Check special cases and error conditions. */
if ((1 << maxbits) < numsymbols) {
free(leaves);
return 1; /* Error, too few maxbits to represent symbols. */
}
if (numsymbols == 0) {
free(leaves);
return 0; /* No symbols at all. OK. */
}
if (numsymbols == 1) {
bitlengths[leaves[0].count] = 1;
free(leaves);
return 0; /* Only one symbol, give it bitlength 1, not 0. OK. */
}
if (numsymbols == 2) {
bitlengths[leaves[0].count]++;
bitlengths[leaves[1].count]++;
free(leaves);
return 0;
}
/* Sort the leaves from lightest to heaviest. Add count into the same
variable for stable sorting. */
for (i = 0; i < numsymbols; i++) {
if (leaves[i].weight >=
((size_t)1 << (sizeof(leaves[0].weight) * CHAR_BIT - 9))) {
free(leaves);
return 1; /* Error, we need 9 bits for the count. */
}
leaves[i].weight = (leaves[i].weight << 9) | leaves[i].count;
}
qsort(leaves, numsymbols, sizeof(Node), LeafComparator);
for (i = 0; i < numsymbols; i++) {
leaves[i].weight >>= 9;
}
if (numsymbols - 1 < maxbits) {
maxbits = numsymbols - 1;
}
/* Initialize node memory pool. */
nodes = (Node*)malloc(maxbits * 2 * numsymbols * sizeof(Node));
pool.next = nodes;
lists = (Node* (*)[2])malloc(maxbits * sizeof(*lists));
InitLists(&pool, leaves, maxbits, lists);
/* In the last list, 2 * numsymbols - 2 active chains need to be created. Two
are already created in the initialization. Each BoundaryPM run creates one. */
numBoundaryPMRuns = 2 * numsymbols - 4;
for (i = 0; i < numBoundaryPMRuns - 1; i++) {
BoundaryPM(lists, leaves, numsymbols, &pool, maxbits - 1);
}
BoundaryPMFinal(lists, leaves, numsymbols, &pool, maxbits - 1);
ExtractBitLengths(lists[maxbits - 1][1], leaves, bitlengths);
free(lists);
free(leaves);
free(nodes);
return 0; /* OK. */
}
tools/bible/zopfli4k/katajainen.h
+42
View File
@@ -0,0 +1,42 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#ifndef ZOPFLI_KATAJAINEN_H_
#define ZOPFLI_KATAJAINEN_H_
#include <string.h>
/*
Outputs minimum-redundancy length-limited code bitlengths for symbols with the
given counts. The bitlengths are limited by maxbits.
The output is tailored for DEFLATE: symbols that never occur, get a bit length
of 0, and if only a single symbol occurs at least once, its bitlength will be 1,
and not 0 as would theoretically be needed for a single symbol.
frequencies: The amount of occurrences of each symbol.
n: The amount of symbols.
maxbits: Maximum bit length, inclusive.
bitlengths: Output, the bitlengths for the symbol prefix codes.
return: 0 for OK, non-0 for error.
*/
int ZopfliLengthLimitedCodeLengths(
const size_t* frequencies, int n, int maxbits, unsigned* bitlengths);
#endif /* ZOPFLI_KATAJAINEN_H_ */
tools/bible/zopfli4k/lz77.c
+641
View File
@@ -0,0 +1,641 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "lz77.h"
#include "symbols.h"
#include "util.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
void ZopfliInitLZ77Store(const unsigned char* data, ZopfliLZ77Store* store) {
store->size = 0;
store->litlens = 0;
store->dists = 0;
store->pos = 0;
store->data = data;
store->ll_symbol = 0;
store->d_symbol = 0;
store->ll_counts = 0;
store->d_counts = 0;
}
void ZopfliCleanLZ77Store(ZopfliLZ77Store* store) {
free(store->litlens);
free(store->dists);
free(store->pos);
free(store->ll_symbol);
free(store->d_symbol);
free(store->ll_counts);
free(store->d_counts);
}
static size_t CeilDiv(size_t a, size_t b) {
return (a + b - 1) / b;
}
void ZopfliCopyLZ77Store(
const ZopfliLZ77Store* source, ZopfliLZ77Store* dest) {
size_t i;
size_t llsize = ZOPFLI_NUM_LL * CeilDiv(source->size, ZOPFLI_NUM_LL);
size_t dsize = ZOPFLI_NUM_D * CeilDiv(source->size, ZOPFLI_NUM_D);
ZopfliCleanLZ77Store(dest);
ZopfliInitLZ77Store(source->data, dest);
dest->litlens =
(unsigned short*)malloc(sizeof(*dest->litlens) * source->size);
dest->dists = (unsigned short*)malloc(sizeof(*dest->dists) * source->size);
dest->pos = (size_t*)malloc(sizeof(*dest->pos) * source->size);
dest->ll_symbol =
(unsigned short*)malloc(sizeof(*dest->ll_symbol) * source->size);
dest->d_symbol =
(unsigned short*)malloc(sizeof(*dest->d_symbol) * source->size);
dest->ll_counts = (size_t*)malloc(sizeof(*dest->ll_counts) * llsize);
dest->d_counts = (size_t*)malloc(sizeof(*dest->d_counts) * dsize);
/* Allocation failed. */
if (!dest->litlens || !dest->dists) exit(-1);
if (!dest->pos) exit(-1);
if (!dest->ll_symbol || !dest->d_symbol) exit(-1);
if (!dest->ll_counts || !dest->d_counts) exit(-1);
dest->size = source->size;
for (i = 0; i < source->size; i++) {
dest->litlens[i] = source->litlens[i];
dest->dists[i] = source->dists[i];
dest->pos[i] = source->pos[i];
dest->ll_symbol[i] = source->ll_symbol[i];
dest->d_symbol[i] = source->d_symbol[i];
}
for (i = 0; i < llsize; i++) {
dest->ll_counts[i] = source->ll_counts[i];
}
for (i = 0; i < dsize; i++) {
dest->d_counts[i] = source->d_counts[i];
}
}
/*
Appends the length and distance to the LZ77 arrays of the ZopfliLZ77Store.
context must be a ZopfliLZ77Store*.
*/
void ZopfliStoreLitLenDist(unsigned short length, unsigned short dist,
size_t pos, ZopfliLZ77Store* store) {
size_t i;
/* Needed for using ZOPFLI_APPEND_DATA multiple times. */
size_t origsize = store->size;
size_t llstart = ZOPFLI_NUM_LL * (origsize / ZOPFLI_NUM_LL);
size_t dstart = ZOPFLI_NUM_D * (origsize / ZOPFLI_NUM_D);
/* Everytime the index wraps around, a new cumulative histogram is made: we're
keeping one histogram value per LZ77 symbol rather than a full histogram for
each to save memory. */
if (origsize % ZOPFLI_NUM_LL == 0) {
size_t llsize = origsize;
for (i = 0; i < ZOPFLI_NUM_LL; i++) {
ZOPFLI_APPEND_DATA(
origsize == 0 ? 0 : store->ll_counts[origsize - ZOPFLI_NUM_LL + i],
&store->ll_counts, &llsize);
}
}
if (origsize % ZOPFLI_NUM_D == 0) {
size_t dsize = origsize;
for (i = 0; i < ZOPFLI_NUM_D; i++) {
ZOPFLI_APPEND_DATA(
origsize == 0 ? 0 : store->d_counts[origsize - ZOPFLI_NUM_D + i],
&store->d_counts, &dsize);
}
}
ZOPFLI_APPEND_DATA(length, &store->litlens, &store->size);
store->size = origsize;
ZOPFLI_APPEND_DATA(dist, &store->dists, &store->size);
store->size = origsize;
ZOPFLI_APPEND_DATA(pos, &store->pos, &store->size);
assert(length < 259);
if (dist == 0) {
store->size = origsize;
ZOPFLI_APPEND_DATA(length, &store->ll_symbol, &store->size);
store->size = origsize;
ZOPFLI_APPEND_DATA(0, &store->d_symbol, &store->size);
store->ll_counts[llstart + length]++;
} else {
store->size = origsize;
ZOPFLI_APPEND_DATA(ZopfliGetLengthSymbol(length),
&store->ll_symbol, &store->size);
store->size = origsize;
ZOPFLI_APPEND_DATA(ZopfliGetDistSymbol(dist),
&store->d_symbol, &store->size);
store->ll_counts[llstart + ZopfliGetLengthSymbol(length)]++;
store->d_counts[dstart + ZopfliGetDistSymbol(dist)]++;
}
}
void ZopfliAppendLZ77Store(const ZopfliLZ77Store* store,
ZopfliLZ77Store* target) {
size_t i;
for (i = 0; i < store->size; i++) {
ZopfliStoreLitLenDist(store->litlens[i], store->dists[i],
store->pos[i], target);
}
}
size_t ZopfliLZ77GetByteRange(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend) {
size_t l = lend - 1;
if (lstart == lend) return 0;
return lz77->pos[l] + ((lz77->dists[l] == 0) ?
1 : lz77->litlens[l]) - lz77->pos[lstart];
}
static void ZopfliLZ77GetHistogramAt(const ZopfliLZ77Store* lz77, size_t lpos,
size_t* ll_counts, size_t* d_counts) {
/* The real histogram is created by using the histogram for this chunk, but
all superfluous values of this chunk subtracted. */
size_t llpos = ZOPFLI_NUM_LL * (lpos / ZOPFLI_NUM_LL);
size_t dpos = ZOPFLI_NUM_D * (lpos / ZOPFLI_NUM_D);
size_t i;
for (i = 0; i < ZOPFLI_NUM_LL; i++) {
ll_counts[i] = lz77->ll_counts[llpos + i];
}
for (i = lpos + 1; i < llpos + ZOPFLI_NUM_LL && i < lz77->size; i++) {
ll_counts[lz77->ll_symbol[i]]--;
}
for (i = 0; i < ZOPFLI_NUM_D; i++) {
d_counts[i] = lz77->d_counts[dpos + i];
}
for (i = lpos + 1; i < dpos + ZOPFLI_NUM_D && i < lz77->size; i++) {
if (lz77->dists[i] != 0) d_counts[lz77->d_symbol[i]]--;
}
}
void ZopfliLZ77GetHistogram(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
size_t* ll_counts, size_t* d_counts) {
size_t i;
if (lstart + ZOPFLI_NUM_LL * 3 > lend) {
memset(ll_counts, 0, sizeof(*ll_counts) * ZOPFLI_NUM_LL);
memset(d_counts, 0, sizeof(*d_counts) * ZOPFLI_NUM_D);
for (i = lstart; i < lend; i++) {
ll_counts[lz77->ll_symbol[i]]++;
if (lz77->dists[i] != 0) d_counts[lz77->d_symbol[i]]++;
}
} else {
/* Subtract the cumulative histograms at the end and the start to get the
histogram for this range. */
ZopfliLZ77GetHistogramAt(lz77, lend - 1, ll_counts, d_counts);
if (lstart > 0) {
size_t ll_counts2[ZOPFLI_NUM_LL];
size_t d_counts2[ZOPFLI_NUM_D];
ZopfliLZ77GetHistogramAt(lz77, lstart - 1, ll_counts2, d_counts2);
for (i = 0; i < ZOPFLI_NUM_LL; i++) {
ll_counts[i] -= ll_counts2[i];
}
for (i = 0; i < ZOPFLI_NUM_D; i++) {
d_counts[i] -= d_counts2[i];
}
}
}
}
void ZopfliInitBlockState(const ZopfliOptions* options,
size_t blockstart, size_t blockend, int add_lmc,
ZopfliBlockState* s) {
s->options = options;
s->blockstart = blockstart;
s->blockend = blockend;
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
if (add_lmc) {
s->lmc = (ZopfliLongestMatchCache*)malloc(sizeof(ZopfliLongestMatchCache));
ZopfliInitCache(blockend - blockstart, s->lmc);
} else {
s->lmc = 0;
}
#endif
}
void ZopfliCleanBlockState(ZopfliBlockState* s) {
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
if (s->lmc) {
ZopfliCleanCache(s->lmc);
free(s->lmc);
}
#endif
}
/*
Gets a score of the length given the distance. Typically, the score of the
length is the length itself, but if the distance is very long, decrease the
score of the length a bit to make up for the fact that long distances use large
amounts of extra bits.
This is not an accurate score, it is a heuristic only for the greedy LZ77
implementation. More accurate cost models are employed later. Making this
heuristic more accurate may hurt rather than improve compression.
The two direct uses of this heuristic are:
-avoid using a length of 3 in combination with a long distance. This only has
an effect if length == 3.
-make a slightly better choice between the two options of the lazy matching.
Indirectly, this affects:
-the block split points if the default of block splitting first is used, in a
rather unpredictable way
-the first zopfli run, so it affects the chance of the first run being closer
to the optimal output
*/
static int GetLengthScore(int length, int distance) {
/*
At 1024, the distance uses 9+ extra bits and this seems to be the sweet spot
on tested files.
*/
return distance > 1024 ? length - 1 : length;
}
void ZopfliVerifyLenDist(const unsigned char* data, size_t datasize, size_t pos,
unsigned short dist, unsigned short length) {
/* TODO(lode): make this only run in a debug compile, it's for assert only. */
size_t i;
assert(pos + length <= datasize);
for (i = 0; i < length; i++) {
if (data[pos - dist + i] != data[pos + i]) {
assert(data[pos - dist + i] == data[pos + i]);
break;
}
}
}
/*
Finds how long the match of scan and match is. Can be used to find how many
bytes starting from scan, and from match, are equal. Returns the last byte
after scan, which is still equal to the correspondinb byte after match.
scan is the position to compare
match is the earlier position to compare.
end is the last possible byte, beyond which to stop looking.
safe_end is a few (8) bytes before end, for comparing multiple bytes at once.
*/
static const unsigned char* GetMatch(const unsigned char* scan,
const unsigned char* match,
const unsigned char* end,
const unsigned char* safe_end) {
if (sizeof(size_t) == 8) {
/* 8 checks at once per array bounds check (size_t is 64-bit).
Use memcpy to avoid unaligned access (causes SIGBUS on strict-alignment
architectures like SPARC and MIPS). The compiler will optimize memcpy of
a known size into a single load where the architecture supports it. */
size_t sv, mv;
while (scan < safe_end) {
memcpy(&sv, scan, sizeof(sv));
memcpy(&mv, match, sizeof(mv));
if (sv != mv) break;
scan += 8;
match += 8;
}
} else if (sizeof(unsigned int) == 4) {
/* 4 checks at once per array bounds check (unsigned int is 32-bit). */
unsigned int sv, mv;
while (scan < safe_end) {
memcpy(&sv, scan, sizeof(sv));
memcpy(&mv, match, sizeof(mv));
if (sv != mv) break;
scan += 4;
match += 4;
}
} else {
/* do 8 checks at once per array bounds check. */
while (scan < safe_end && *scan == *match && *++scan == *++match
&& *++scan == *++match && *++scan == *++match
&& *++scan == *++match && *++scan == *++match
&& *++scan == *++match && *++scan == *++match) {
scan++; match++;
}
}
/* The remaining few bytes. */
while (scan != end && *scan == *match) {
scan++; match++;
}
return scan;
}
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
/*
Gets distance, length and sublen values from the cache if possible.
Returns 1 if it got the values from the cache, 0 if not.
Updates the limit value to a smaller one if possible with more limited
information from the cache.
*/
static int TryGetFromLongestMatchCache(ZopfliBlockState* s,
size_t pos, size_t* limit,
unsigned short* sublen, unsigned short* distance, unsigned short* length) {
/* The LMC cache starts at the beginning of the block rather than the
beginning of the whole array. */
size_t lmcpos = pos - s->blockstart;
/* Length > 0 and dist 0 is invalid combination, which indicates on purpose
that this cache value is not filled in yet. */
unsigned char cache_available = s->lmc && (s->lmc->length[lmcpos] == 0 ||
s->lmc->dist[lmcpos] != 0);
unsigned char limit_ok_for_cache = cache_available &&
(*limit == ZOPFLI_MAX_MATCH || s->lmc->length[lmcpos] <= *limit ||
(sublen && ZopfliMaxCachedSublen(s->lmc,
lmcpos, s->lmc->length[lmcpos]) >= *limit));
if (s->lmc && limit_ok_for_cache && cache_available) {
if (!sublen || s->lmc->length[lmcpos]
<= ZopfliMaxCachedSublen(s->lmc, lmcpos, s->lmc->length[lmcpos])) {
*length = s->lmc->length[lmcpos];
if (*length > *limit) *length = *limit;
if (sublen) {
ZopfliCacheToSublen(s->lmc, lmcpos, *length, sublen);
*distance = sublen[*length];
if (*limit == ZOPFLI_MAX_MATCH && *length >= ZOPFLI_MIN_MATCH) {
assert(sublen[*length] == s->lmc->dist[lmcpos]);
}
} else {
*distance = s->lmc->dist[lmcpos];
}
return 1;
}
/* Can't use much of the cache, since the "sublens" need to be calculated,
but at least we already know when to stop. */
*limit = s->lmc->length[lmcpos];
}
return 0;
}
/*
Stores the found sublen, distance and length in the longest match cache, if
possible.
*/
static void StoreInLongestMatchCache(ZopfliBlockState* s,
size_t pos, size_t limit,
const unsigned short* sublen,
unsigned short distance, unsigned short length) {
/* The LMC cache starts at the beginning of the block rather than the
beginning of the whole array. */
size_t lmcpos = pos - s->blockstart;
/* Length > 0 and dist 0 is invalid combination, which indicates on purpose
that this cache value is not filled in yet. */
unsigned char cache_available = s->lmc && (s->lmc->length[lmcpos] == 0 ||
s->lmc->dist[lmcpos] != 0);
if (s->lmc && limit == ZOPFLI_MAX_MATCH && sublen && !cache_available) {
assert(s->lmc->length[lmcpos] == 1 && s->lmc->dist[lmcpos] == 0);
s->lmc->dist[lmcpos] = length < ZOPFLI_MIN_MATCH ? 0 : distance;
s->lmc->length[lmcpos] = length < ZOPFLI_MIN_MATCH ? 0 : length;
assert(!(s->lmc->length[lmcpos] == 1 && s->lmc->dist[lmcpos] == 0));
ZopfliSublenToCache(sublen, lmcpos, length, s->lmc);
}
}
#endif
void ZopfliFindLongestMatch(ZopfliBlockState* s, const ZopfliHash* h,
const unsigned char* array,
size_t pos, size_t size, size_t limit,
unsigned short* sublen, unsigned short* distance, unsigned short* length) {
unsigned short hpos = pos & ZOPFLI_WINDOW_MASK, p, pp;
unsigned short bestdist = 0;
unsigned short bestlength = 1;
const unsigned char* scan;
const unsigned char* match;
const unsigned char* arrayend;
const unsigned char* arrayend_safe;
#if ZOPFLI_MAX_CHAIN_HITS < ZOPFLI_WINDOW_SIZE
int chain_counter = ZOPFLI_MAX_CHAIN_HITS; /* For quitting early. */
#endif
unsigned dist = 0; /* Not unsigned short on purpose. */
int* hhead = h->head;
unsigned short* hprev = h->prev;
int* hhashval = h->hashval;
int hval = h->val;
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
if (TryGetFromLongestMatchCache(s, pos, &limit, sublen, distance, length)) {
assert(pos + *length <= size);
return;
}
#endif
assert(limit <= ZOPFLI_MAX_MATCH);
assert(limit >= ZOPFLI_MIN_MATCH);
assert(pos < size);
if (size - pos < ZOPFLI_MIN_MATCH) {
/* The rest of the code assumes there are at least ZOPFLI_MIN_MATCH bytes to
try. */
*length = 0;
*distance = 0;
return;
}
if (pos + limit > size) {
limit = size - pos;
}
arrayend = &array[pos] + limit;
arrayend_safe = arrayend - 8;
assert(hval < 65536);
pp = hhead[hval]; /* During the whole loop, p == hprev[pp]. */
p = hprev[pp];
assert(pp == hpos);
dist = p < pp ? pp - p : ((ZOPFLI_WINDOW_SIZE - p) + pp);
/* Go through all distances. */
while (dist < ZOPFLI_WINDOW_SIZE) {
unsigned short currentlength = 0;
assert(p < ZOPFLI_WINDOW_SIZE);
assert(p == hprev[pp]);
assert(hhashval[p] == hval);
if (dist > 0) {
assert(pos < size);
assert(dist <= pos);
scan = &array[pos];
match = &array[pos - dist];
/* Testing the byte at position bestlength first, goes slightly faster. */
if (pos + bestlength >= size
|| *(scan + bestlength) == *(match + bestlength)) {
#ifdef ZOPFLI_HASH_SAME
unsigned short same0 = h->same[pos & ZOPFLI_WINDOW_MASK];
if (same0 > 2 && *scan == *match) {
unsigned short same1 = h->same[(pos - dist) & ZOPFLI_WINDOW_MASK];
unsigned short same = same0 < same1 ? same0 : same1;
if (same > limit) same = limit;
scan += same;
match += same;
}
#endif
scan = GetMatch(scan, match, arrayend, arrayend_safe);
currentlength = scan - &array[pos]; /* The found length. */
}
if (currentlength > bestlength) {
if (sublen) {
unsigned short j;
for (j = bestlength + 1; j <= currentlength; j++) {
sublen[j] = dist;
}
}
bestdist = dist;
bestlength = currentlength;
if (currentlength >= limit) break;
}
}
#ifdef ZOPFLI_HASH_SAME_HASH
/* Switch to the other hash once this will be more efficient. */
if (hhead != h->head2 && bestlength >= h->same[hpos] &&
h->val2 == h->hashval2[p]) {
/* Now use the hash that encodes the length and first byte. */
hhead = h->head2;
hprev = h->prev2;
hhashval = h->hashval2;
hval = h->val2;
}
#endif
pp = p;
p = hprev[p];
if (p == pp) break; /* Uninited prev value. */
dist += p < pp ? pp - p : ((ZOPFLI_WINDOW_SIZE - p) + pp);
#if ZOPFLI_MAX_CHAIN_HITS < ZOPFLI_WINDOW_SIZE
chain_counter--;
if (chain_counter <= 0) break;
#endif
}
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
StoreInLongestMatchCache(s, pos, limit, sublen, bestdist, bestlength);
#endif
assert(bestlength <= limit);
*distance = bestdist;
*length = bestlength;
assert(pos + *length <= size);
}
void ZopfliLZ77Greedy(ZopfliBlockState* s, const unsigned char* in,
size_t instart, size_t inend,
ZopfliLZ77Store* store, ZopfliHash* h) {
size_t i = 0, j;
unsigned short leng;
unsigned short dist;
int lengthscore;
size_t windowstart = instart > ZOPFLI_WINDOW_SIZE
? instart - ZOPFLI_WINDOW_SIZE : 0;
unsigned short dummysublen[259];
#ifdef ZOPFLI_LAZY_MATCHING
/* Lazy matching. */
unsigned prev_length = 0;
unsigned prev_match = 0;
int prevlengthscore;
int match_available = 0;
#endif
if (instart == inend) return;
ZopfliResetHash(ZOPFLI_WINDOW_SIZE, h);
ZopfliWarmupHash(in, windowstart, inend, h);
for (i = windowstart; i < instart; i++) {
ZopfliUpdateHash(in, i, inend, h);
}
for (i = instart; i < inend; i++) {
ZopfliUpdateHash(in, i, inend, h);
ZopfliFindLongestMatch(s, h, in, i, inend, ZOPFLI_MAX_MATCH, dummysublen,
&dist, &leng);
lengthscore = GetLengthScore(leng, dist);
#ifdef ZOPFLI_LAZY_MATCHING
/* Lazy matching. */
prevlengthscore = GetLengthScore(prev_length, prev_match);
if (match_available) {
match_available = 0;
if (lengthscore > prevlengthscore + 1) {
ZopfliStoreLitLenDist(in[i - 1], 0, i - 1, store);
if (lengthscore >= ZOPFLI_MIN_MATCH && leng < ZOPFLI_MAX_MATCH) {
match_available = 1;
prev_length = leng;
prev_match = dist;
continue;
}
} else {
/* Add previous to output. */
leng = prev_length;
dist = prev_match;
lengthscore = prevlengthscore;
/* Add to output. */
ZopfliVerifyLenDist(in, inend, i - 1, dist, leng);
ZopfliStoreLitLenDist(leng, dist, i - 1, store);
for (j = 2; j < leng; j++) {
assert(i < inend);
i++;
ZopfliUpdateHash(in, i, inend, h);
}
continue;
}
}
else if (lengthscore >= ZOPFLI_MIN_MATCH && leng < ZOPFLI_MAX_MATCH) {
match_available = 1;
prev_length = leng;
prev_match = dist;
continue;
}
/* End of lazy matching. */
#endif
/* Add to output. */
if (lengthscore >= ZOPFLI_MIN_MATCH) {
ZopfliVerifyLenDist(in, inend, i, dist, leng);
ZopfliStoreLitLenDist(leng, dist, i, store);
} else {
leng = 1;
ZopfliStoreLitLenDist(in[i], 0, i, store);
}
for (j = 1; j < leng; j++) {
assert(i < inend);
i++;
ZopfliUpdateHash(in, i, inend, h);
}
}
}
tools/bible/zopfli4k/lz77.h
+142
View File
@@ -0,0 +1,142 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Functions for basic LZ77 compression and utilities for the "squeeze" LZ77
compression.
*/
#ifndef ZOPFLI_LZ77_H_
#define ZOPFLI_LZ77_H_
#include <stdlib.h>
#include "cache.h"
#include "hash.h"
#include "zopfli.h"
/*
Stores lit/length and dist pairs for LZ77.
Parameter litlens: Contains the literal symbols or length values.
Parameter dists: Contains the distances. A value is 0 to indicate that there is
no dist and the corresponding litlens value is a literal instead of a length.
Parameter size: The size of both the litlens and dists arrays.
The memory can best be managed by using ZopfliInitLZ77Store to initialize it,
ZopfliCleanLZ77Store to destroy it, and ZopfliStoreLitLenDist to append values.
*/
typedef struct ZopfliLZ77Store {
unsigned short* litlens; /* Lit or len. */
unsigned short* dists; /* If 0: indicates literal in corresponding litlens,
if > 0: length in corresponding litlens, this is the distance. */
size_t size;
const unsigned char* data; /* original data */
size_t* pos; /* position in data where this LZ77 command begins */
unsigned short* ll_symbol;
unsigned short* d_symbol;
/* Cumulative histograms wrapping around per chunk. Each chunk has the amount
of distinct symbols as length, so using 1 value per LZ77 symbol, we have a
precise histogram at every N symbols, and the rest can be calculated by
looping through the actual symbols of this chunk. */
size_t* ll_counts;
size_t* d_counts;
} ZopfliLZ77Store;
void ZopfliInitLZ77Store(const unsigned char* data, ZopfliLZ77Store* store);
void ZopfliCleanLZ77Store(ZopfliLZ77Store* store);
void ZopfliCopyLZ77Store(const ZopfliLZ77Store* source, ZopfliLZ77Store* dest);
void ZopfliStoreLitLenDist(unsigned short length, unsigned short dist,
size_t pos, ZopfliLZ77Store* store);
void ZopfliAppendLZ77Store(const ZopfliLZ77Store* store,
ZopfliLZ77Store* target);
/* Gets the amount of raw bytes that this range of LZ77 symbols spans. */
size_t ZopfliLZ77GetByteRange(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend);
/* Gets the histogram of lit/len and dist symbols in the given range, using the
cumulative histograms, so faster than adding one by one for large range. Does
not add the one end symbol of value 256. */
void ZopfliLZ77GetHistogram(const ZopfliLZ77Store* lz77,
size_t lstart, size_t lend,
size_t* ll_counts, size_t* d_counts);
/*
Some state information for compressing a block.
This is currently a bit under-used (with mainly only the longest match cache),
but is kept for easy future expansion.
*/
typedef struct ZopfliBlockState {
const ZopfliOptions* options;
#ifdef ZOPFLI_LONGEST_MATCH_CACHE
/* Cache for length/distance pairs found so far. */
ZopfliLongestMatchCache* lmc;
#endif
/* The start (inclusive) and end (not inclusive) of the current block. */
size_t blockstart;
size_t blockend;
} ZopfliBlockState;
void ZopfliInitBlockState(const ZopfliOptions* options,
size_t blockstart, size_t blockend, int add_lmc,
ZopfliBlockState* s);
void ZopfliCleanBlockState(ZopfliBlockState* s);
/*
Finds the longest match (length and corresponding distance) for LZ77
compression.
Even when not using "sublen", it can be more efficient to provide an array,
because only then the caching is used.
array: the data
pos: position in the data to find the match for
size: size of the data
limit: limit length to maximum this value (default should be 258). This allows
finding a shorter dist for that length (= less extra bits). Must be
in the range [ZOPFLI_MIN_MATCH, ZOPFLI_MAX_MATCH].
sublen: output array of 259 elements, or null. Has, for each length, the
smallest distance required to reach this length. Only 256 of its 259 values
are used, the first 3 are ignored (the shortest length is 3. It is purely
for convenience that the array is made 3 longer).
*/
void ZopfliFindLongestMatch(
ZopfliBlockState *s, const ZopfliHash* h, const unsigned char* array,
size_t pos, size_t size, size_t limit,
unsigned short* sublen, unsigned short* distance, unsigned short* length);
/*
Verifies if length and dist are indeed valid, only used for assertion.
*/
void ZopfliVerifyLenDist(const unsigned char* data, size_t datasize, size_t pos,
unsigned short dist, unsigned short length);
/*
Does LZ77 using an algorithm similar to gzip, with lazy matching, rather than
with the slow but better "squeeze" implementation.
The result is placed in the ZopfliLZ77Store.
If instart is larger than 0, it uses values before instart as starting
dictionary.
*/
void ZopfliLZ77Greedy(ZopfliBlockState* s, const unsigned char* in,
size_t instart, size_t inend,
ZopfliLZ77Store* store, ZopfliHash* h);
#endif /* ZOPFLI_LZ77_H_ */
tools/bible/zopfli4k/reader_pack.c
+16
View File
@@ -0,0 +1,16 @@
#include "zopfli.h"
#include <stdio.h>
#include <stdlib.h>
int main(int argc,char **argv) {
if(argc!=4)return 2;
FILE *in=fopen(argv[1],"rb");if(!in)return 3;
fseek(in,0,SEEK_END);long len=ftell(in);rewind(in);
unsigned char*src=malloc((size_t)len);if(!src)return 4;
if(fread(src,1,(size_t)len,in)!=(size_t)len)return 5;fclose(in);
ZopfliOptions options;ZopfliInitOptions(&options);options.numiterations=atoi(argv[3]);
unsigned char*out=NULL;size_t out_size=0;
ZopfliCompress(&options,ZOPFLI_FORMAT_DEFLATE,src,(size_t)len,&out,&out_size);
FILE *dest=fopen(argv[2],"wb");if(!dest)return 6;
if(fwrite(out,1,out_size,dest)!=out_size)return 7;fclose(dest);
printf("%ld -> %zu\n",len,out_size);free(src);free(out);return 0;
}
tools/bible/zopfli4k/squeeze.c
+560
View File
@@ -0,0 +1,560 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "squeeze.h"
#include <assert.h>
#include <math.h>
#include <stdio.h>
#include "blocksplitter.h"
#include "deflate.h"
#include "symbols.h"
#include "tree.h"
#include "util.h"
typedef struct SymbolStats {
/* The literal and length symbols. */
size_t litlens[ZOPFLI_NUM_LL];
/* The 32 unique dist symbols, not the 32768 possible dists. */
size_t dists[ZOPFLI_NUM_D];
/* Length of each lit/len symbol in bits. */
double ll_symbols[ZOPFLI_NUM_LL];
/* Length of each dist symbol in bits. */
double d_symbols[ZOPFLI_NUM_D];
} SymbolStats;
/* Sets everything to 0. */
static void InitStats(SymbolStats* stats) {
memset(stats->litlens, 0, ZOPFLI_NUM_LL * sizeof(stats->litlens[0]));
memset(stats->dists, 0, ZOPFLI_NUM_D * sizeof(stats->dists[0]));
memset(stats->ll_symbols, 0, ZOPFLI_NUM_LL * sizeof(stats->ll_symbols[0]));
memset(stats->d_symbols, 0, ZOPFLI_NUM_D * sizeof(stats->d_symbols[0]));
}
static void CopyStats(SymbolStats* source, SymbolStats* dest) {
memcpy(dest->litlens, source->litlens,
ZOPFLI_NUM_LL * sizeof(dest->litlens[0]));
memcpy(dest->dists, source->dists, ZOPFLI_NUM_D * sizeof(dest->dists[0]));
memcpy(dest->ll_symbols, source->ll_symbols,
ZOPFLI_NUM_LL * sizeof(dest->ll_symbols[0]));
memcpy(dest->d_symbols, source->d_symbols,
ZOPFLI_NUM_D * sizeof(dest->d_symbols[0]));
}
/* Adds the bit lengths. */
static void AddWeighedStatFreqs(const SymbolStats* stats1, double w1,
const SymbolStats* stats2, double w2,
SymbolStats* result) {
size_t i;
for (i = 0; i < ZOPFLI_NUM_LL; i++) {
result->litlens[i] =
(size_t) (stats1->litlens[i] * w1 + stats2->litlens[i] * w2);
}
for (i = 0; i < ZOPFLI_NUM_D; i++) {
result->dists[i] =
(size_t) (stats1->dists[i] * w1 + stats2->dists[i] * w2);
}
result->litlens[256] = 1; /* End symbol. */
}
typedef struct RanState {
unsigned int m_w, m_z;
} RanState;
static void InitRanState(RanState* state) {
state->m_w = 1;
state->m_z = 2;
}
/* Get random number: "Multiply-With-Carry" generator of G. Marsaglia */
static unsigned int Ran(RanState* state) {
state->m_z = 36969 * (state->m_z & 65535) + (state->m_z >> 16);
state->m_w = 18000 * (state->m_w & 65535) + (state->m_w >> 16);
return (state->m_z << 16) + state->m_w; /* 32-bit result. */
}
static void RandomizeFreqs(RanState* state, size_t* freqs, int n) {
int i;
for (i = 0; i < n; i++) {
if ((Ran(state) >> 4) % 3 == 0) freqs[i] = freqs[Ran(state) % n];
}
}
static void RandomizeStatFreqs(RanState* state, SymbolStats* stats) {
RandomizeFreqs(state, stats->litlens, ZOPFLI_NUM_LL);
RandomizeFreqs(state, stats->dists, ZOPFLI_NUM_D);
stats->litlens[256] = 1; /* End symbol. */
}
static void ClearStatFreqs(SymbolStats* stats) {
size_t i;
for (i = 0; i < ZOPFLI_NUM_LL; i++) stats->litlens[i] = 0;
for (i = 0; i < ZOPFLI_NUM_D; i++) stats->dists[i] = 0;
}
/*
Function that calculates a cost based on a model for the given LZ77 symbol.
litlen: means literal symbol if dist is 0, length otherwise.
*/
typedef double CostModelFun(unsigned litlen, unsigned dist, void* context);
/*
Cost model which should exactly match fixed tree.
type: CostModelFun
*/
static double GetCostFixed(unsigned litlen, unsigned dist, void* unused) {
(void)unused;
if (dist == 0) {
if (litlen <= 143) return 8;
else return 9;
} else {
int dbits = ZopfliGetDistExtraBits(dist);
int lbits = ZopfliGetLengthExtraBits(litlen);
int lsym = ZopfliGetLengthSymbol(litlen);
int cost = 0;
if (lsym <= 279) cost += 7;
else cost += 8;
cost += 5; /* Every dist symbol has length 5. */
return cost + dbits + lbits;
}
}
/*
Cost model based on symbol statistics.
type: CostModelFun
*/
static double GetCostStat(unsigned litlen, unsigned dist, void* context) {
SymbolStats* stats = (SymbolStats*)context;
if (dist == 0) {
return stats->ll_symbols[litlen];
} else {
int lsym = ZopfliGetLengthSymbol(litlen);
int lbits = ZopfliGetLengthExtraBits(litlen);
int dsym = ZopfliGetDistSymbol(dist);
int dbits = ZopfliGetDistExtraBits(dist);
return lbits + dbits + stats->ll_symbols[lsym] + stats->d_symbols[dsym];
}
}
/*
Finds the minimum possible cost this cost model can return for valid length and
distance symbols.
*/
static double GetCostModelMinCost(CostModelFun* costmodel, void* costcontext) {
double mincost;
int bestlength = 0; /* length that has lowest cost in the cost model */
int bestdist = 0; /* distance that has lowest cost in the cost model */
int i;
/*
Table of distances that have a different distance symbol in the deflate
specification. Each value is the first distance that has a new symbol. Only
different symbols affect the cost model so only these need to be checked.
See RFC 1951 section 3.2.5. Compressed blocks (length and distance codes).
*/
static const int dsymbols[30] = {
1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513,
769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577
};
mincost = ZOPFLI_LARGE_FLOAT;
for (i = 3; i < 259; i++) {
double c = costmodel(i, 1, costcontext);
if (c < mincost) {
bestlength = i;
mincost = c;
}
}
mincost = ZOPFLI_LARGE_FLOAT;
for (i = 0; i < 30; i++) {
double c = costmodel(3, dsymbols[i], costcontext);
if (c < mincost) {
bestdist = dsymbols[i];
mincost = c;
}
}
return costmodel(bestlength, bestdist, costcontext);
}
static size_t zopfli_min(size_t a, size_t b) {
return a < b ? a : b;
}
/*
Performs the forward pass for "squeeze". Gets the most optimal length to reach
every byte from a previous byte, using cost calculations.
s: the ZopfliBlockState
in: the input data array
instart: where to start
inend: where to stop (not inclusive)
costmodel: function to calculate the cost of some lit/len/dist pair.
costcontext: abstract context for the costmodel function
length_array: output array of size (inend - instart) which will receive the best
length to reach this byte from a previous byte.
returns the cost that was, according to the costmodel, needed to get to the end.
*/
static double GetBestLengths(ZopfliBlockState *s,
const unsigned char* in,
size_t instart, size_t inend,
CostModelFun* costmodel, void* costcontext,
unsigned short* length_array,
ZopfliHash* h, float* costs) {
/* Best cost to get here so far. */
size_t blocksize = inend - instart;
size_t i = 0, k, kend;
unsigned short leng;
unsigned short dist;
unsigned short sublen[259];
size_t windowstart = instart > ZOPFLI_WINDOW_SIZE
? instart - ZOPFLI_WINDOW_SIZE : 0;
double result;
double mincost = GetCostModelMinCost(costmodel, costcontext);
double mincostaddcostj;
if (instart == inend) return 0;
ZopfliResetHash(ZOPFLI_WINDOW_SIZE, h);
ZopfliWarmupHash(in, windowstart, inend, h);
for (i = windowstart; i < instart; i++) {
ZopfliUpdateHash(in, i, inend, h);
}
for (i = 1; i < blocksize + 1; i++) costs[i] = ZOPFLI_LARGE_FLOAT;
costs[0] = 0; /* Because it's the start. */
length_array[0] = 0;
for (i = instart; i < inend; i++) {
size_t j = i - instart; /* Index in the costs array and length_array. */
ZopfliUpdateHash(in, i, inend, h);
#ifdef ZOPFLI_SHORTCUT_LONG_REPETITIONS
/* If we're in a long repetition of the same character and have more than
ZOPFLI_MAX_MATCH characters before and after our position. */
if (h->same[i & ZOPFLI_WINDOW_MASK] > ZOPFLI_MAX_MATCH * 2
&& i > instart + ZOPFLI_MAX_MATCH + 1
&& i + ZOPFLI_MAX_MATCH * 2 + 1 < inend
&& h->same[(i - ZOPFLI_MAX_MATCH) & ZOPFLI_WINDOW_MASK]
> ZOPFLI_MAX_MATCH) {
double symbolcost = costmodel(ZOPFLI_MAX_MATCH, 1, costcontext);
/* Set the length to reach each one to ZOPFLI_MAX_MATCH, and the cost to
the cost corresponding to that length. Doing this, we skip
ZOPFLI_MAX_MATCH values to avoid calling ZopfliFindLongestMatch. */
for (k = 0; k < ZOPFLI_MAX_MATCH; k++) {
costs[j + ZOPFLI_MAX_MATCH] = costs[j] + symbolcost;
length_array[j + ZOPFLI_MAX_MATCH] = ZOPFLI_MAX_MATCH;
i++;
j++;
ZopfliUpdateHash(in, i, inend, h);
}
}
#endif
ZopfliFindLongestMatch(s, h, in, i, inend, ZOPFLI_MAX_MATCH, sublen,
&dist, &leng);
/* Literal. */
if (i + 1 <= inend) {
double newCost = costmodel(in[i], 0, costcontext) + costs[j];
assert(newCost >= 0);
if (newCost < costs[j + 1]) {
costs[j + 1] = newCost;
length_array[j + 1] = 1;
}
}
/* Lengths. */
kend = zopfli_min(leng, inend-i);
mincostaddcostj = mincost + costs[j];
for (k = 3; k <= kend; k++) {
double newCost;
/* Calling the cost model is expensive, avoid this if we are already at
the minimum possible cost that it can return. */
if (costs[j + k] <= mincostaddcostj) continue;
newCost = costmodel(k, sublen[k], costcontext) + costs[j];
assert(newCost >= 0);
if (newCost < costs[j + k]) {
assert(k <= ZOPFLI_MAX_MATCH);
costs[j + k] = newCost;
length_array[j + k] = k;
}
}
}
assert(costs[blocksize] >= 0);
result = costs[blocksize];
return result;
}
/*
Calculates the optimal path of lz77 lengths to use, from the calculated
length_array. The length_array must contain the optimal length to reach that
byte. The path will be filled with the lengths to use, so its data size will be
the amount of lz77 symbols.
*/
static void TraceBackwards(size_t size, const unsigned short* length_array,
unsigned short** path, size_t* pathsize) {
size_t index = size;
if (size == 0) return;
for (;;) {
ZOPFLI_APPEND_DATA(length_array[index], path, pathsize);
assert(length_array[index] <= index);
assert(length_array[index] <= ZOPFLI_MAX_MATCH);
assert(length_array[index] != 0);
index -= length_array[index];
if (index == 0) break;
}
/* Mirror result. */
for (index = 0; index < *pathsize / 2; index++) {
unsigned short temp = (*path)[index];
(*path)[index] = (*path)[*pathsize - index - 1];
(*path)[*pathsize - index - 1] = temp;
}
}
static void FollowPath(ZopfliBlockState* s,
const unsigned char* in, size_t instart, size_t inend,
unsigned short* path, size_t pathsize,
ZopfliLZ77Store* store, ZopfliHash *h) {
size_t i, j, pos = 0;
size_t windowstart = instart > ZOPFLI_WINDOW_SIZE
? instart - ZOPFLI_WINDOW_SIZE : 0;
size_t total_length_test = 0;
if (instart == inend) return;
ZopfliResetHash(ZOPFLI_WINDOW_SIZE, h);
ZopfliWarmupHash(in, windowstart, inend, h);
for (i = windowstart; i < instart; i++) {
ZopfliUpdateHash(in, i, inend, h);
}
pos = instart;
for (i = 0; i < pathsize; i++) {
unsigned short length = path[i];
unsigned short dummy_length;
unsigned short dist;
assert(pos < inend);
ZopfliUpdateHash(in, pos, inend, h);
/* Add to output. */
if (length >= ZOPFLI_MIN_MATCH) {
/* Get the distance by recalculating longest match. The found length
should match the length from the path. */
ZopfliFindLongestMatch(s, h, in, pos, inend, length, 0,
&dist, &dummy_length);
assert(!(dummy_length != length && length > 2 && dummy_length > 2));
ZopfliVerifyLenDist(in, inend, pos, dist, length);
ZopfliStoreLitLenDist(length, dist, pos, store);
total_length_test += length;
} else {
length = 1;
ZopfliStoreLitLenDist(in[pos], 0, pos, store);
total_length_test++;
}
assert(pos + length <= inend);
for (j = 1; j < length; j++) {
ZopfliUpdateHash(in, pos + j, inend, h);
}
pos += length;
}
}
/* Calculates the entropy of the statistics */
static void CalculateStatistics(SymbolStats* stats) {
ZopfliCalculateEntropy(stats->litlens, ZOPFLI_NUM_LL, stats->ll_symbols);
ZopfliCalculateEntropy(stats->dists, ZOPFLI_NUM_D, stats->d_symbols);
}
/* Appends the symbol statistics from the store. */
static void GetStatistics(const ZopfliLZ77Store* store, SymbolStats* stats) {
size_t i;
for (i = 0; i < store->size; i++) {
if (store->dists[i] == 0) {
stats->litlens[store->litlens[i]]++;
} else {
stats->litlens[ZopfliGetLengthSymbol(store->litlens[i])]++;
stats->dists[ZopfliGetDistSymbol(store->dists[i])]++;
}
}
stats->litlens[256] = 1; /* End symbol. */
CalculateStatistics(stats);
}
/*
Does a single run for ZopfliLZ77Optimal. For good compression, repeated runs
with updated statistics should be performed.
s: the block state
in: the input data array
instart: where to start
inend: where to stop (not inclusive)
path: pointer to dynamically allocated memory to store the path
pathsize: pointer to the size of the dynamic path array
length_array: array of size (inend - instart) used to store lengths
costmodel: function to use as the cost model for this squeeze run
costcontext: abstract context for the costmodel function
store: place to output the LZ77 data
returns the cost that was, according to the costmodel, needed to get to the end.
This is not the actual cost.
*/
static double LZ77OptimalRun(ZopfliBlockState* s,
const unsigned char* in, size_t instart, size_t inend,
unsigned short** path, size_t* pathsize,
unsigned short* length_array, CostModelFun* costmodel,
void* costcontext, ZopfliLZ77Store* store,
ZopfliHash* h, float* costs) {
double cost = GetBestLengths(s, in, instart, inend, costmodel,
costcontext, length_array, h, costs);
free(*path);
*path = 0;
*pathsize = 0;
TraceBackwards(inend - instart, length_array, path, pathsize);
FollowPath(s, in, instart, inend, *path, *pathsize, store, h);
assert(cost < ZOPFLI_LARGE_FLOAT);
return cost;
}
void ZopfliLZ77Optimal(ZopfliBlockState *s,
const unsigned char* in, size_t instart, size_t inend,
int numiterations,
ZopfliLZ77Store* store) {
/* Dist to get to here with smallest cost. */
size_t blocksize = inend - instart;
unsigned short* length_array =
(unsigned short*)malloc(sizeof(unsigned short) * (blocksize + 1));
unsigned short* path = 0;
size_t pathsize = 0;
ZopfliLZ77Store currentstore;
ZopfliHash hash;
ZopfliHash* h = &hash;
SymbolStats stats, beststats, laststats;
int i;
float* costs = (float*)malloc(sizeof(float) * (blocksize + 1));
double cost;
double bestcost = ZOPFLI_LARGE_FLOAT;
double lastcost = 0;
/* Try randomizing the costs a bit once the size stabilizes. */
RanState ran_state;
int lastrandomstep = -1;
if (!costs) exit(-1); /* Allocation failed. */
if (!length_array) exit(-1); /* Allocation failed. */
InitRanState(&ran_state);
InitStats(&stats);
ZopfliInitLZ77Store(in, &currentstore);
ZopfliAllocHash(ZOPFLI_WINDOW_SIZE, h);
/* Do regular deflate, then loop multiple shortest path runs, each time using
the statistics of the previous run. */
/* Initial run. */
ZopfliLZ77Greedy(s, in, instart, inend, &currentstore, h);
GetStatistics(&currentstore, &stats);
/* Repeat statistics with each time the cost model from the previous stat
run. */
for (i = 0; i < numiterations; i++) {
ZopfliCleanLZ77Store(&currentstore);
ZopfliInitLZ77Store(in, &currentstore);
LZ77OptimalRun(s, in, instart, inend, &path, &pathsize,
length_array, GetCostStat, (void*)&stats,
&currentstore, h, costs);
cost = ZopfliCalculateBlockSize(&currentstore, 0, currentstore.size, 2);
if (s->options->verbose_more || (s->options->verbose && cost < bestcost)) {
fprintf(stderr, "Iteration %d: %d bit\n", i, (int) cost);
}
if (cost < bestcost) {
/* Copy to the output store. */
ZopfliCopyLZ77Store(&currentstore, store);
CopyStats(&stats, &beststats);
bestcost = cost;
}
CopyStats(&stats, &laststats);
ClearStatFreqs(&stats);
GetStatistics(&currentstore, &stats);
if (lastrandomstep != -1) {
/* This makes it converge slower but better. Do it only once the
randomness kicks in so that if the user does few iterations, it gives a
better result sooner. */
AddWeighedStatFreqs(&stats, 1.0, &laststats, 0.5, &stats);
CalculateStatistics(&stats);
}
if (i > 5 && cost == lastcost) {
CopyStats(&beststats, &stats);
RandomizeStatFreqs(&ran_state, &stats);
CalculateStatistics(&stats);
lastrandomstep = i;
}
lastcost = cost;
}
free(length_array);
free(path);
free(costs);
ZopfliCleanLZ77Store(&currentstore);
ZopfliCleanHash(h);
}
void ZopfliLZ77OptimalFixed(ZopfliBlockState *s,
const unsigned char* in,
size_t instart, size_t inend,
ZopfliLZ77Store* store)
{
/* Dist to get to here with smallest cost. */
size_t blocksize = inend - instart;
unsigned short* length_array =
(unsigned short*)malloc(sizeof(unsigned short) * (blocksize + 1));
unsigned short* path = 0;
size_t pathsize = 0;
ZopfliHash hash;
ZopfliHash* h = &hash;
float* costs = (float*)malloc(sizeof(float) * (blocksize + 1));
if (!costs) exit(-1); /* Allocation failed. */
if (!length_array) exit(-1); /* Allocation failed. */
ZopfliAllocHash(ZOPFLI_WINDOW_SIZE, h);
s->blockstart = instart;
s->blockend = inend;
/* Shortest path for fixed tree This one should give the shortest possible
result for fixed tree, no repeated runs are needed since the tree is known. */
LZ77OptimalRun(s, in, instart, inend, &path, &pathsize,
length_array, GetCostFixed, 0, store, h, costs);
free(length_array);
free(path);
free(costs);
ZopfliCleanHash(h);
}
tools/bible/zopfli4k/squeeze.h
+61
View File
@@ -0,0 +1,61 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
The squeeze functions do enhanced LZ77 compression by optimal parsing with a
cost model, rather than greedily choosing the longest length or using a single
step of lazy matching like regular implementations.
Since the cost model is based on the Huffman tree that can only be calculated
after the LZ77 data is generated, there is a chicken and egg problem, and
multiple runs are done with updated cost models to converge to a better
solution.
*/
#ifndef ZOPFLI_SQUEEZE_H_
#define ZOPFLI_SQUEEZE_H_
#include "lz77.h"
/*
Calculates lit/len and dist pairs for given data.
If instart is larger than 0, it uses values before instart as starting
dictionary.
*/
void ZopfliLZ77Optimal(ZopfliBlockState *s,
const unsigned char* in, size_t instart, size_t inend,
int numiterations,
ZopfliLZ77Store* store);
/*
Does the same as ZopfliLZ77Optimal, but optimized for the fixed tree of the
deflate standard.
The fixed tree never gives the best compression. But this gives the best
possible LZ77 encoding possible with the fixed tree.
This does not create or output any fixed tree, only LZ77 data optimized for
using with a fixed tree.
If instart is larger than 0, it uses values before instart as starting
dictionary.
*/
void ZopfliLZ77OptimalFixed(ZopfliBlockState *s,
const unsigned char* in,
size_t instart, size_t inend,
ZopfliLZ77Store* store);
#endif /* ZOPFLI_SQUEEZE_H_ */
tools/bible/zopfli4k/symbols.h
+239
View File
@@ -0,0 +1,239 @@
/*
Copyright 2016 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Utilities for using the lz77 symbols of the deflate spec.
*/
#ifndef ZOPFLI_SYMBOLS_H_
#define ZOPFLI_SYMBOLS_H_
/* __has_builtin available in clang */
#ifdef __has_builtin
# if __has_builtin(__builtin_clz)
# define ZOPFLI_HAS_BUILTIN_CLZ
# endif
/* __builtin_clz available beginning with GCC 3.4 */
#elif __GNUC__ * 100 + __GNUC_MINOR__ >= 304
# define ZOPFLI_HAS_BUILTIN_CLZ
#endif
/* Gets the amount of extra bits for the given dist, cfr. the DEFLATE spec. */
static int ZopfliGetDistExtraBits(int dist) {
#ifdef ZOPFLI_HAS_BUILTIN_CLZ
if (dist < 5) return 0;
return (31 ^ __builtin_clz(dist - 1)) - 1; /* log2(dist - 1) - 1 */
#else
if (dist < 5) return 0;
else if (dist < 9) return 1;
else if (dist < 17) return 2;
else if (dist < 33) return 3;
else if (dist < 65) return 4;
else if (dist < 129) return 5;
else if (dist < 257) return 6;
else if (dist < 513) return 7;
else if (dist < 1025) return 8;
else if (dist < 2049) return 9;
else if (dist < 4097) return 10;
else if (dist < 8193) return 11;
else if (dist < 16385) return 12;
else return 13;
#endif
}
/* Gets value of the extra bits for the given dist, cfr. the DEFLATE spec. */
static int ZopfliGetDistExtraBitsValue(int dist) {
#ifdef ZOPFLI_HAS_BUILTIN_CLZ
if (dist < 5) {
return 0;
} else {
int l = 31 ^ __builtin_clz(dist - 1); /* log2(dist - 1) */
return (dist - (1 + (1 << l))) & ((1 << (l - 1)) - 1);
}
#else
if (dist < 5) return 0;
else if (dist < 9) return (dist - 5) & 1;
else if (dist < 17) return (dist - 9) & 3;
else if (dist < 33) return (dist - 17) & 7;
else if (dist < 65) return (dist - 33) & 15;
else if (dist < 129) return (dist - 65) & 31;
else if (dist < 257) return (dist - 129) & 63;
else if (dist < 513) return (dist - 257) & 127;
else if (dist < 1025) return (dist - 513) & 255;
else if (dist < 2049) return (dist - 1025) & 511;
else if (dist < 4097) return (dist - 2049) & 1023;
else if (dist < 8193) return (dist - 4097) & 2047;
else if (dist < 16385) return (dist - 8193) & 4095;
else return (dist - 16385) & 8191;
#endif
}
/* Gets the symbol for the given dist, cfr. the DEFLATE spec. */
static int ZopfliGetDistSymbol(int dist) {
#ifdef ZOPFLI_HAS_BUILTIN_CLZ
if (dist < 5) {
return dist - 1;
} else {
int l = (31 ^ __builtin_clz(dist - 1)); /* log2(dist - 1) */
int r = ((dist - 1) >> (l - 1)) & 1;
return l * 2 + r;
}
#else
if (dist < 193) {
if (dist < 13) { /* dist 0..13. */
if (dist < 5) return dist - 1;
else if (dist < 7) return 4;
else if (dist < 9) return 5;
else return 6;
} else { /* dist 13..193. */
if (dist < 17) return 7;
else if (dist < 25) return 8;
else if (dist < 33) return 9;
else if (dist < 49) return 10;
else if (dist < 65) return 11;
else if (dist < 97) return 12;
else if (dist < 129) return 13;
else return 14;
}
} else {
if (dist < 2049) { /* dist 193..2049. */
if (dist < 257) return 15;
else if (dist < 385) return 16;
else if (dist < 513) return 17;
else if (dist < 769) return 18;
else if (dist < 1025) return 19;
else if (dist < 1537) return 20;
else return 21;
} else { /* dist 2049..32768. */
if (dist < 3073) return 22;
else if (dist < 4097) return 23;
else if (dist < 6145) return 24;
else if (dist < 8193) return 25;
else if (dist < 12289) return 26;
else if (dist < 16385) return 27;
else if (dist < 24577) return 28;
else return 29;
}
}
#endif
}
/* Gets the amount of extra bits for the given length, cfr. the DEFLATE spec. */
static int ZopfliGetLengthExtraBits(int l) {
static const int table[259] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 0
};
return table[l];
}
/* Gets value of the extra bits for the given length, cfr. the DEFLATE spec. */
static int ZopfliGetLengthExtraBitsValue(int l) {
static const int table[259] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 2, 3, 0,
1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5,
6, 7, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6,
7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12,
13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2,
3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28,
29, 30, 31, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 0, 1, 2, 3, 4, 5, 6,
7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26,
27, 28, 29, 30, 31, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 0
};
return table[l];
}
/*
Gets the symbol for the given length, cfr. the DEFLATE spec.
Returns the symbol in the range [257-285] (inclusive)
*/
static int ZopfliGetLengthSymbol(int l) {
static const int table[259] = {
0, 0, 0, 257, 258, 259, 260, 261, 262, 263, 264,
265, 265, 266, 266, 267, 267, 268, 268,
269, 269, 269, 269, 270, 270, 270, 270,
271, 271, 271, 271, 272, 272, 272, 272,
273, 273, 273, 273, 273, 273, 273, 273,
274, 274, 274, 274, 274, 274, 274, 274,
275, 275, 275, 275, 275, 275, 275, 275,
276, 276, 276, 276, 276, 276, 276, 276,
277, 277, 277, 277, 277, 277, 277, 277,
277, 277, 277, 277, 277, 277, 277, 277,
278, 278, 278, 278, 278, 278, 278, 278,
278, 278, 278, 278, 278, 278, 278, 278,
279, 279, 279, 279, 279, 279, 279, 279,
279, 279, 279, 279, 279, 279, 279, 279,
280, 280, 280, 280, 280, 280, 280, 280,
280, 280, 280, 280, 280, 280, 280, 280,
281, 281, 281, 281, 281, 281, 281, 281,
281, 281, 281, 281, 281, 281, 281, 281,
281, 281, 281, 281, 281, 281, 281, 281,
281, 281, 281, 281, 281, 281, 281, 281,
282, 282, 282, 282, 282, 282, 282, 282,
282, 282, 282, 282, 282, 282, 282, 282,
282, 282, 282, 282, 282, 282, 282, 282,
282, 282, 282, 282, 282, 282, 282, 282,
283, 283, 283, 283, 283, 283, 283, 283,
283, 283, 283, 283, 283, 283, 283, 283,
283, 283, 283, 283, 283, 283, 283, 283,
283, 283, 283, 283, 283, 283, 283, 283,
284, 284, 284, 284, 284, 284, 284, 284,
284, 284, 284, 284, 284, 284, 284, 284,
284, 284, 284, 284, 284, 284, 284, 284,
284, 284, 284, 284, 284, 284, 284, 285
};
return table[l];
}
/* Gets the amount of extra bits for the given length symbol. */
static int ZopfliGetLengthSymbolExtraBits(int s) {
static const int table[29] = {
0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0
};
return table[s - 257];
}
/* Gets the amount of extra bits for the given distance symbol. */
static int ZopfliGetDistSymbolExtraBits(int s) {
static const int table[30] = {
0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
9, 9, 10, 10, 11, 11, 12, 12, 13, 13
};
return table[s];
}
#endif /* ZOPFLI_SYMBOLS_H_ */
tools/bible/zopfli4k/tree.c
+101
View File
@@ -0,0 +1,101 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "tree.h"
#include <assert.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "katajainen.h"
#include "util.h"
void ZopfliLengthsToSymbols(const unsigned* lengths, size_t n, unsigned maxbits,
unsigned* symbols) {
size_t* bl_count = (size_t*)malloc(sizeof(size_t) * (maxbits + 1));
size_t* next_code = (size_t*)malloc(sizeof(size_t) * (maxbits + 1));
unsigned bits, i;
unsigned code;
for (i = 0; i < n; i++) {
symbols[i] = 0;
}
/* 1) Count the number of codes for each code length. Let bl_count[N] be the
number of codes of length N, N >= 1. */
for (bits = 0; bits <= maxbits; bits++) {
bl_count[bits] = 0;
}
for (i = 0; i < n; i++) {
assert(lengths[i] <= maxbits);
bl_count[lengths[i]]++;
}
/* 2) Find the numerical value of the smallest code for each code length. */
code = 0;
bl_count[0] = 0;
for (bits = 1; bits <= maxbits; bits++) {
code = (code + bl_count[bits-1]) << 1;
next_code[bits] = code;
}
/* 3) Assign numerical values to all codes, using consecutive values for all
codes of the same length with the base values determined at step 2. */
for (i = 0; i < n; i++) {
unsigned len = lengths[i];
if (len != 0) {
symbols[i] = next_code[len];
next_code[len]++;
}
}
free(bl_count);
free(next_code);
}
void ZopfliCalculateEntropy(const size_t* count, size_t n, double* bitlengths) {
static const double kInvLog2 = 1.4426950408889; /* 1.0 / log(2.0) */
unsigned sum = 0;
unsigned i;
double log2sum;
for (i = 0; i < n; ++i) {
sum += count[i];
}
log2sum = (sum == 0 ? log(n) : log(sum)) * kInvLog2;
for (i = 0; i < n; ++i) {
/* When the count of the symbol is 0, but its cost is requested anyway, it
means the symbol will appear at least once anyway, so give it the cost as if
its count is 1.*/
if (count[i] == 0) bitlengths[i] = log2sum;
else bitlengths[i] = log2sum - log(count[i]) * kInvLog2;
/* Depending on compiler and architecture, the above subtraction of two
floating point numbers may give a negative result very close to zero
instead of zero (e.g. -5.973954e-17 with gcc 4.1.2 on Ubuntu 11.4). Clamp
it to zero. These floating point imprecisions do not affect the cost model
significantly so this is ok. */
if (bitlengths[i] < 0 && bitlengths[i] > -1e-5) bitlengths[i] = 0;
assert(bitlengths[i] >= 0);
}
}
void ZopfliCalculateBitLengths(const size_t* count, size_t n, int maxbits,
unsigned* bitlengths) {
int error = ZopfliLengthLimitedCodeLengths(count, n, maxbits, bitlengths);
(void) error;
assert(!error);
}
tools/bible/zopfli4k/tree.h
+51
View File
@@ -0,0 +1,51 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Utilities for creating and using Huffman trees.
*/
#ifndef ZOPFLI_TREE_H_
#define ZOPFLI_TREE_H_
#include <string.h>
/*
Calculates the bitlengths for the Huffman tree, based on the counts of each
symbol.
*/
void ZopfliCalculateBitLengths(const size_t* count, size_t n, int maxbits,
unsigned *bitlengths);
/*
Converts a series of Huffman tree bitlengths, to the bit values of the symbols.
*/
void ZopfliLengthsToSymbols(const unsigned* lengths, size_t n, unsigned maxbits,
unsigned* symbols);
/*
Calculates the entropy of each symbol, based on the counts of each symbol. The
result is similar to the result of ZopfliCalculateBitLengths, but with the
actual theoritical bit lengths according to the entropy. Since the resulting
values are fractional, they cannot be used to encode the tree specified by
DEFLATE.
*/
void ZopfliCalculateEntropy(const size_t* count, size_t n, double* bitlengths);
#endif /* ZOPFLI_TREE_H_ */
tools/bible/zopfli4k/util.c
+35
View File
@@ -0,0 +1,35 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "util.h"
#include "zopfli.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
void ZopfliInitOptions(ZopfliOptions* options) {
options->verbose = 0;
options->verbose_more = 0;
options->numiterations = 15;
options->blocksplitting = 1;
options->blocksplittinglast = 0;
options->blocksplittingmax = 15;
}
tools/bible/zopfli4k/util.h
+159
View File
@@ -0,0 +1,159 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
/*
Several utilities, including: #defines to try different compression results,
basic deflate specification values and generic program options.
*/
#ifndef ZOPFLI_UTIL_H_
#define ZOPFLI_UTIL_H_
#include <string.h>
#include <stdlib.h>
/* Minimum and maximum length that can be encoded in deflate. */
#define ZOPFLI_MAX_MATCH 258
#define ZOPFLI_MIN_MATCH 3
/* Number of distinct literal/length and distance symbols in DEFLATE */
#define ZOPFLI_NUM_LL 288
#define ZOPFLI_NUM_D 32
/*
The window size for deflate. Must be a power of two. This should be 32768, the
maximum possible by the deflate spec. Anything less hurts compression more than
speed.
*/
/* MeshCore host-only modification: match the firmware's existing 4 KiB history. */
#define ZOPFLI_WINDOW_SIZE 4096
/*
The window mask used to wrap indices into the window. This is why the
window size must be a power of two.
*/
#define ZOPFLI_WINDOW_MASK (ZOPFLI_WINDOW_SIZE - 1)
/*
A block structure of huge, non-smart, blocks to divide the input into, to allow
operating on huge files without exceeding memory, such as the 1GB wiki9 corpus.
The whole compression algorithm, including the smarter block splitting, will
be executed independently on each huge block.
Dividing into huge blocks hurts compression, but not much relative to the size.
Set it to 0 to disable master blocks.
*/
#define ZOPFLI_MASTER_BLOCK_SIZE 1000000
/*
Used to initialize costs for example
*/
#define ZOPFLI_LARGE_FLOAT 1e30
/*
For longest match cache. max 256. Uses huge amounts of memory but makes it
faster. Uses this many times three bytes per single byte of the input data.
This is so because longest match finding has to find the exact distance
that belongs to each length for the best lz77 strategy.
Good values: e.g. 5, 8.
*/
#define ZOPFLI_CACHE_LENGTH 8
/*
limit the max hash chain hits for this hash value. This has an effect only
on files where the hash value is the same very often. On these files, this
gives worse compression (the value should ideally be 32768, which is the
ZOPFLI_WINDOW_SIZE, while zlib uses 4096 even for best level), but makes it
faster on some specific files.
Good value: e.g. 8192.
*/
#define ZOPFLI_MAX_CHAIN_HITS 8192
/*
Whether to use the longest match cache for ZopfliFindLongestMatch. This cache
consumes a lot of memory but speeds it up. No effect on compression size.
*/
#define ZOPFLI_LONGEST_MATCH_CACHE
/*
Enable to remember amount of successive identical bytes in the hash chain for
finding longest match
required for ZOPFLI_HASH_SAME_HASH and ZOPFLI_SHORTCUT_LONG_REPETITIONS
This has no effect on the compression result, and enabling it increases speed.
*/
#define ZOPFLI_HASH_SAME
/*
Switch to a faster hash based on the info from ZOPFLI_HASH_SAME once the
best length so far is long enough. This is way faster for files with lots of
identical bytes, on which the compressor is otherwise too slow. Regular files
are unaffected or maybe a tiny bit slower.
This has no effect on the compression result, only on speed.
*/
#define ZOPFLI_HASH_SAME_HASH
/*
Enable this, to avoid slowness for files which are a repetition of the same
character more than a multiple of ZOPFLI_MAX_MATCH times. This should not affect
the compression result.
*/
#define ZOPFLI_SHORTCUT_LONG_REPETITIONS
/*
Whether to use lazy matching in the greedy LZ77 implementation. This gives a
better result of ZopfliLZ77Greedy, but the effect this has on the optimal LZ77
varies from file to file.
*/
#define ZOPFLI_LAZY_MATCHING
/*
Appends value to dynamically allocated memory, doubling its allocation size
whenever needed.
value: the value to append, type T
data: pointer to the dynamic array to append to, type T**
size: pointer to the size of the array to append to, type size_t*. This is the
size that you consider the array to be, not the internal allocation size.
Precondition: allocated size of data is at least a power of two greater than or
equal than *size.
*/
#ifdef __cplusplus /* C++ cannot assign void* from malloc to *data */
#define ZOPFLI_APPEND_DATA(/* T */ value, /* T** */ data, /* size_t* */ size) {\
if (!((*size) & ((*size) - 1))) {\
/*double alloc size if it's a power of two*/\
void** data_void = reinterpret_cast<void**>(data);\
*data_void = (*size) == 0 ? malloc(sizeof(**data))\
: realloc((*data), (*size) * 2 * sizeof(**data));\
}\
(*data)[(*size)] = (value);\
(*size)++;\
}
#else /* C gives problems with strict-aliasing rules for (void**) cast */
#define ZOPFLI_APPEND_DATA(/* T */ value, /* T** */ data, /* size_t* */ size) {\
if (!((*size) & ((*size) - 1))) {\
/*double alloc size if it's a power of two*/\
(*data) = (*size) == 0 ? malloc(sizeof(**data))\
: realloc((*data), (*size) * 2 * sizeof(**data));\
}\
(*data)[(*size)] = (value);\
(*size)++;\
}
#endif
#endif /* ZOPFLI_UTIL_H_ */
tools/bible/zopfli4k/zlib_container.c
+79
View File
@@ -0,0 +1,79 @@
/*
Copyright 2013 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "zlib_container.h"
#include "util.h"
#include <stdio.h>
#include "deflate.h"
/* Calculates the adler32 checksum of the data */
static unsigned adler32(const unsigned char* data, size_t size)
{
static const unsigned sums_overflow = 5550;
unsigned s1 = 1;
unsigned s2 = 1 >> 16;
while (size > 0) {
size_t amount = size > sums_overflow ? sums_overflow : size;
size -= amount;
while (amount > 0) {
s1 += (*data++);
s2 += s1;
amount--;
}
s1 %= 65521;
s2 %= 65521;
}
return (s2 << 16) | s1;
}
void ZopfliZlibCompress(const ZopfliOptions* options,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize) {
unsigned char bitpointer = 0;
unsigned checksum = adler32(in, (unsigned)insize);
unsigned cmf = 120; /* CM 8, CINFO 7. See zlib spec.*/
unsigned flevel = 3;
unsigned fdict = 0;
unsigned cmfflg = 256 * cmf + fdict * 32 + flevel * 64;
unsigned fcheck = 31 - cmfflg % 31;
cmfflg += fcheck;
ZOPFLI_APPEND_DATA(cmfflg / 256, out, outsize);
ZOPFLI_APPEND_DATA(cmfflg % 256, out, outsize);
ZopfliDeflate(options, 2 /* dynamic block */, 1 /* final */,
in, insize, &bitpointer, out, outsize);
ZOPFLI_APPEND_DATA((checksum >> 24) % 256, out, outsize);
ZOPFLI_APPEND_DATA((checksum >> 16) % 256, out, outsize);
ZOPFLI_APPEND_DATA((checksum >> 8) % 256, out, outsize);
ZOPFLI_APPEND_DATA(checksum % 256, out, outsize);
if (options->verbose) {
fprintf(stderr,
"Original Size: %d, Zlib: %d, Compression: %f%% Removed\n",
(int)insize, (int)*outsize,
100.0 * (double)(insize - *outsize) / (double)insize);
}
}
tools/bible/zopfli4k/zlib_container.h
+50
View File
@@ -0,0 +1,50 @@
/*
Copyright 2013 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#ifndef ZOPFLI_ZLIB_H_
#define ZOPFLI_ZLIB_H_
/*
Functions to compress according to the Zlib specification.
*/
#include "zopfli.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
Compresses according to the zlib specification and append the compressed
result to the output.
options: global program options
out: pointer to the dynamic output array to which the result is appended. Must
be freed after use.
outsize: pointer to the dynamic output array size.
*/
void ZopfliZlibCompress(const ZopfliOptions* options,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* ZOPFLI_ZLIB_H_ */
tools/bible/zopfli4k/zopfli.h
+94
View File
@@ -0,0 +1,94 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#ifndef ZOPFLI_ZOPFLI_H_
#define ZOPFLI_ZOPFLI_H_
#include <stddef.h>
#include <stdlib.h> /* for size_t */
#ifdef __cplusplus
extern "C" {
#endif
/*
Options used throughout the program.
*/
typedef struct ZopfliOptions {
/* Whether to print output */
int verbose;
/* Whether to print more detailed output */
int verbose_more;
/*
Maximum amount of times to rerun forward and backward pass to optimize LZ77
compression cost. Good values: 10, 15 for small files, 5 for files over
several MB in size or it will be too slow.
*/
int numiterations;
/*
If true, splits the data in multiple deflate blocks with optimal choice
for the block boundaries. Block splitting gives better compression. Default:
true (1).
*/
int blocksplitting;
/*
No longer used, left for compatibility.
*/
int blocksplittinglast;
/*
Maximum amount of blocks to split into (0 for unlimited, but this can give
extreme results that hurt compression on some files). Default value: 15.
*/
int blocksplittingmax;
} ZopfliOptions;
/* Initializes options with default values. */
void ZopfliInitOptions(ZopfliOptions* options);
/* Output format */
typedef enum {
ZOPFLI_FORMAT_GZIP,
ZOPFLI_FORMAT_ZLIB,
ZOPFLI_FORMAT_DEFLATE
} ZopfliFormat;
/*
Compresses according to the given output format and appends the result to the
output.
options: global program options
output_type: the output format to use
out: pointer to the dynamic output array to which the result is appended. Must
be freed after use
outsize: pointer to the dynamic output array size
*/
void ZopfliCompress(const ZopfliOptions* options, ZopfliFormat output_type,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* ZOPFLI_ZOPFLI_H_ */
tools/bible/zopfli4k/zopfli_lib.c
+42
View File
@@ -0,0 +1,42 @@
/*
Copyright 2011 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Author: lode.vandevenne@gmail.com (Lode Vandevenne)
Author: jyrki.alakuijala@gmail.com (Jyrki Alakuijala)
*/
#include "zopfli.h"
#include "deflate.h"
#include "gzip_container.h"
#include "zlib_container.h"
#include <assert.h>
void ZopfliCompress(const ZopfliOptions* options, ZopfliFormat output_type,
const unsigned char* in, size_t insize,
unsigned char** out, size_t* outsize) {
if (output_type == ZOPFLI_FORMAT_GZIP) {
ZopfliGzipCompress(options, in, insize, out, outsize);
} else if (output_type == ZOPFLI_FORMAT_ZLIB) {
ZopfliZlibCompress(options, in, insize, out, outsize);
} else if (output_type == ZOPFLI_FORMAT_DEFLATE) {
unsigned char bp = 0;
ZopfliDeflate(options, 2 /* Dynamic block */, 1,
in, insize, &bp, out, outsize);
} else {
assert(0);
}
}