Commit Graph

36 Commits

Author SHA1 Message Date
Brotli
fe754f3459 Use a hash table header and SIMD to speed up hash table operations (similar to [Swiss Tables](https://abseil.io/about/design/swisstables)).
PiperOrigin-RevId: 638686412
2024-05-30 09:51:43 -07:00
Brotli
04388304a6 Use a hash table header and SIMD to speed up hash table operations (similar to [Swiss Tables](https://abseil.io/about/design/swisstables)).
PiperOrigin-RevId: 632238409
2024-05-09 12:42:42 -07:00
Brotli
bb809ac908 Use a hash table header and SIMD to speed up hash table operations (similar to [Swiss Tables](https://abseil.io/about/design/swisstables)).
PiperOrigin-RevId: 631982664
2024-05-08 17:59:58 -07:00
Brotli
1b3a5ccb6e Prefetch the backreference hashtable bucket.
Place the prefetch before the last distance checks, to give the prefetch enough time to work.

PiperOrigin-RevId: 626228820
2024-04-18 20:00:02 -07:00
Evgenii Kliuchnikov
d5e697b3c7 remove dependency on os-specific defines
PiperOrigin-RevId: 600449944
2024-01-22 07:24:41 -08:00
Evgenii Kliuchnikov
6b6adb7ae8 fix build for Microsoft-designed ARM64 ABI
PiperOrigin-RevId: 595334901
2024-01-03 02:01:23 -08:00
Evgenii Kliuchnikov
6ee96e291d Internal changes
PiperOrigin-RevId: 545262005
2023-07-04 07:55:25 +00:00
Evgenii Kliuchnikov
e5dba91c38 Add BROTLI_ENABLE_DUMP build option
PiperOrigin-RevId: 520047051
2023-07-04 07:54:07 +00:00
Evgenii Kliuchnikov
6db17c87f5 0.4-1.5% decoder speedup
PiperOrigin-RevId: 513248503
2023-07-04 07:53:20 +00:00
Zhang Na
1d8452b783
Add loongarch64 support (#1022) 2023-06-20 09:44:23 +02:00
Evgenii Kliuchnikov
a8f5813b84 Update
Documentation:
  - add note that brotli is a "stream" format, not an archive-like
  - regenerate .1 with Pandoc
Build:
  - drop legacy "BROTLI_BUILD_PORTABLE" option
  - drop "BROTLI_SANITIZED" definition
Code:
  - c: comb includes
  - c/enc: extract encoder state into separate header
  - c/enc: drop designated q10 codepath
  - c/enc: dealing better with flushing of empty stream
  - fix MSVC compilation
API:
  - py: use library version instead of one in version.h
  - c: add plugable API to report consumed input / produced output
  - c/java: support "lean" prepared dictionaries (without copy of source)
2022-11-17 13:03:09 +00:00
Eugene Kliuchnikov
8376f72ed6
Prepare for copybara (#939)
Co-authored-by: Eugene Kliuchnikov <eustas@chromium.org>
2021-11-10 10:34:39 +01:00
Gabriel
97006561ea
Fix VC C++ 12.0 BROTLI_MSVC_VERSION_CHECK calls (#843) 2020-09-21 13:22:56 +02:00
Eugene Kliuchnikov
223d80cfbe
Update (#826)
* IMPORTANT: decoder: fix potential overflow when input chunk is >2GiB
 * simplify max Huffman table size calculation
 * eliminate symbol duplicates (static arrays in .h files)
 * minor combing in research/ code
2020-08-26 12:32:27 +02:00
Eugene Kliuchnikov
0c5603e07b
Fix output parameter type for _BitScanReverse (#819)
Fix #811
2020-07-09 16:40:01 +02:00
fisherwky
d2ea198232
Update platform.h (#813)
fix compile error (platform.h:362: error: cast discards qualifiers from pointer target type)
2020-06-30 11:23:07 +02:00
Eugene Kliuchnikov
7f740f1308
Update (#807)
- fix formatting
 - fix type conversion
 - fix no-op arithmetic with null-pointer
 - improve performance of hash_longest_match64
 - go: detect read after close
 - java decoder: support compound dictionary
 - remove executable flag on non-scripts
2020-05-15 11:06:21 +02:00
Clinton Ingram
924b2b2b9d
Move TZCNT and BSR intrinsics to platform.h, add MSVC versions (#636) 2020-03-19 11:57:56 +01:00
Eugene Kliuchnikov
3d1767186d
Fix include for EMCC build (#765) 2019-07-30 10:01:21 +02:00
Eugene Kliuchnikov
f1124c8524
More careful sanitizer detection (#764) 2019-07-22 14:29:51 +02:00
Eugene Kliuchnikov
4b2b2d4f83
Update (#749)
Update:

 * Bazel: fix MSVC configuration
 * C: common: extended documentation and helpers around distance codes
 * C: common: enable BROTLI_DCHECK in "debug" builds
 * C: common: fix implicit trailing zero in `kPrefixSuffix`
 * C: dec: fix possible bit reader discharge for "large-window" mode
 * C: dec: simplify distance decoding via lookup table
 * C: dec: reuse decoder state members memory via union with lookup table
 * C: dec: add decoder state diagram
 * C: enc: clarify access to static dictionary
 * C: enc: improve static dictionary hash
 * C: enc: add "stream offset" parameter for parallel encoding
 * C: enc: reorganize hasher; now Q2-Q3 require exactly 256KiB
           to avoid global TCMalloc lock
 * C: enc: fix rare access to uninitialized data in ring-buffer
 * C: enc: reorganize logging / checks in `write_bits.h`
 * Java: dec: add "large-window" support
 * Java: dec: improve speed
 * Java: dec: debug and 32-bit mode are now activated via system properties
 * Java: dec: demystify some state variables (use better names)
 * Dictionary generator: add single input mode
 * Java: dec: modernize tests
 * Bazel: js: pick working commit for closure rules
2019-04-12 13:57:42 +02:00
Eugene Kliuchnikov
a1e44975a7
Fix #698 2018-10-19 17:01:54 +02:00
Stephen Kyle
cc7a74f15f decode: fix NEON inclusion (#714)
The macro that checks for NEON support should be __ARM_NEON, not
__ARM_NEON__. [1]

AArch64 compilers define __ARM_NEON but not __ARM_NEON__.
AArch32 compilers currently seem to define both, but could be within their
rights to drop __ARM_NEON__ in future versions.

This change moves the check into the common/platform.h file, checks for
both forms, and sets BROTLI_TARGET_NEON if NEON support is available.

[1] Section 6.5.4 of the ARM C Language Extensions.
    (At the time of writing, the latest version was Release 2.1.)
2018-10-08 15:40:11 +02:00
Stephen Kyle
9402ac5c08 decode: faster huffman code loading on 32-bit Arm (#703)
* platform: add macro for using the 'aligned' attribute

* decode: add accessor macros for HuffmanCode fields

Adds a constructor function for building HuffmanCode values
so they can be accessed quickly on different architectures.

Also adds macros for marking a HuffmanCode table pointer
that can be accessed quickly (BROTLI_HC_MARK_TABLE_FOR_FAST_LOAD),
adjusting the index into that table (BROTLI_HC_ADJUST_TABLE_INDEX),
and getting the .bits or .value fields out of the table at the
current index (BROTLI_HC_GET_BITS/VALUE).

For example, assuming |table| contains a HuffmanCode pointer:

  BROTLI_HC_MARK_TABLE_FOR_FAST_LOAD(table);
  BROTLI_HC_ADJUST_TABLE_INDEX(table, index_into_table);
  *bits = BROTLI_HC_GET_BITS(table);
  *value = BROTLI_HC_GET_VALUE(table);
  BROTLI_HC_ADJUST_TABLE_INDEX(table, offset);
  *bits2 = BROTLI_HC_GET_BITS(table);
  *value2 = BROTLI_HC_GET_VALUE(table);

All uses of the HuffmanCode have been updated appropriately.

* decode: add alternative accessors for HuffmanCode on Arm AArch32
2018-09-27 13:15:46 +02:00
Eugene Kliuchnikov
2216a0dd63
Update (#706)
Update
 * add ASAN/MSAN unaligned read specializations
 * add "brotli" prefix to u_uint64 type
 * increment version to 1.0.06
 * fix CoverityScan "unused assignment" warning
 * fix JDK 8<->9 incompatibility
 * add encoder optimization for empty input
 * regenerate JS decoder
 * unbreak Travis builds
2018-09-13 08:09:32 -04:00
Stephen Kyle
d4cd6cdf1c platform: fix unaligned 64-bit accesses on AArch32 (#702)
Ensures that Aarch32 Arm builds with an Armv8 compiler do not set
BROTLI_64_BITS.

This scenario is possible with ChromeOS builds, as they may use a
toolchain with the target armv7-cros-gnueabi, but with -march=armv8.
This will set __ARM_ARCH to 8 (defining BROTLI_TARGET_ARMV8), but will
also set __ARM_32BIT_STATE and not __ARM_64BIT_STATE. Without this,
illegal 64-bit non-word-aligned reads (LDRD) may be emitted.

Also fix unaligned 64-bit reads on AArch32 - STRD was still possible to
emit.
2018-07-25 11:43:06 +02:00
Eugene Kliuchnikov
8a073bd9e2
Revert "platform: fix unaligned 64-bit accesses on AArch32 (#699)" (#701)
This reverts commit 6d027d1648.
2018-07-24 17:32:13 +02:00
Stephen Kyle
6d027d1648 platform: fix unaligned 64-bit accesses on AArch32 (#699)
Ensures that Aarch32 Arm builds with an Armv8 compiler do not set
BROTLI_64_BITS.

This scenario is possible with ChromeOS builds, as they may use a
toolchain with the target armv7-cros-gnueabi, but with -march=armv8.
This will set __ARM_ARCH to 8 (defining BROTLI_TARGET_ARMV8), but will
also set __ARM_32BIT_STATE and not __ARM_64BIT_STATE. Without this,
illegal 64-bit non-word-aligned reads (LDRD) may be emitted.

Also fix unaligned 64-bit reads on AArch32 - STRD was still possible to
emit.
2018-07-24 17:29:50 +02:00
Eugene Kliuchnikov
eb12ec04eb
Update (#688)
* add rolling-composite-hasher for large-window mode
* make API methods explicitly public
2018-06-20 15:14:10 +02:00
Eugene Kliuchnikov
8544ae858d
Update (#680)
* fix MSVC warnings
 * cleanups
2018-06-09 11:17:13 +02:00
davidlt
f9b8c02673 Add RISC-V 64-bit (riscv64) platform configuration (#669)
Signed-off-by: David Abdurachmanov <david.abdurachmanov@gmail.com>
2018-05-22 14:35:04 +02:00
Eugene Kliuchnikov
f5ed35d065
Update (#664)
* Update
 * fix ifdef style
 * get back to fine-compiler-version-based-macros (use Hedley)
 * fix q=0 histogram collection for very long copy/insert commands
2018-05-03 11:16:21 +02:00
Eugene Kliuchnikov
6000396155
Remove unprefixed macros from public headers (#662) 2018-04-20 14:10:55 +02:00
Eugene Kliuchnikov
0f3c84e745
Update (#656)
* proper fix for the "fall through" warning"
 * automatic NDIRECT/NPOSTFIX tuning (better compression)
 * fix unaligned access for `aarch64`-cross-`armhf` build
 * fix `aarch64` detection (10% decoder speedup)
 * expose `large_window` CLI option
 * make default window size 16MiB
 * ramp up version to 1.0.4
2018-03-27 22:29:22 +02:00
Eugene Kliuchnikov
35e69fc7cf
New feature: "Large Window Brotli" (#640)
* New feature: "Large Window Brotli"

By setting special encoder/decoder flag it is now possible to extend
LZ-window up to 30 bits; though produced stream will not be RFC7932
compliant.

Added new dictionary generator - "DSH". It combines speed of "Sieve"
and quality of "DM". Plus utilities to prepare train corpora
(remove unique strings).

Improved compression ratio: now two sub-blocks could be stitched:
the last copy command could be extended to span the next sub-block.

Fixed compression ineffectiveness caused by floating numbers rounding and
wrong cost heuristic.

Other C changes:
 - combined / moved `context.h` to `common`
 - moved transforms to `common`
 - unified some aspects of code formatting
 - added an abstraction for encoder (static) dictionary
 - moved default allocator/deallocator functions to `common`

brotli CLI:
 - window size is auto-adjusted if not specified explicitly

Java:
 - added "eager" decoding both to JNI wrapper and pure decoder
 - huge speed-up of `DictionaryData` initialization

* Add dictionaryless compressed dictionary

* Fix `sources.lst`

* Fix `sources.lst` and add a note that `libtool` is also required.

* Update setup.py

* Fix `EagerStreamTest`

* Fix BUILD file

* Add missing `libdivsufsort` dependency

* Fix "unused parameter" warning.
2018-02-26 09:04:36 -05:00
Eugene Kliuchnikov
da254cffdb
Update (#630)
* merge {dec|enc}/port.h into common/platform.h
 * fix one-shot q=10 1-byte input compression
 * fix some unprefixed definitions
 * make hashers host-endianness-independent
 * extract enc/params.h from enc/quality.h
 * fix API documentation / typos
 * improve `BrotliEncoderMaxCompressedSize`
2017-12-12 14:33:12 +01:00