Now extra buffer space is reserved so that repeating bytes for
any single match will never need to copy from two places (both
the beginning and the end of the buffer). This simplifies
dict_repeat() and helps a little with speed.
This seems to reduce .lzma decompression time about 2 %, so
with .xz and CRC it could be slightly less. The small things
add up still.
It's not completely obvious if this is better in the decoder.
It should be good if compiler can avoid creating a branch
(like using CMOV on x86).
This also makes lzma_encoder.c use the new macros.
The new decoder resumes the first decoder loop in the Resumable mode.
Then, the code executes in Non-resumable mode until it detects that it
cannot guarantee to have enough input/output to decode another symbol.
The Resumable mode is how the decoder has always worked. Before decoding
every input bit, it checks if there is enough space and will save its
location to be resumed later. When the decoder has more input/output,
it jumps back to the correct sequence in the Resumable mode code.
When the input/output buffers are large, the Resumable mode is much
slower than the Non-resumable because it has more branches and is harder
for the compiler to optimize since it is in a large switch block.
Early benchmarking shows significant time improvement (8-10% on gcc and
clang x86) by using the Non-resumable code as much as possible.
The new "safe" range decoder mode is the same as old range decoder, but
now the default behavior of the range decoder will not check if there is
enough input or output to complete the operation. When the buffers are
close to fully consumed, the "safe" operations must be used instead. This
will improve speed because it will reduce the number of branches needed
for most of the range decoder operations.
Perhaps the generated files aren't even copyrightable but
using the same license for them as for the rest of the liblzma
keeps things more consistent for tools that look for license info.
The initial commit 5d018dc035
in 2007 had a comment in sha256.c that the code is based on
Crypto++ Library 5.5.1. In 2009 the Authors list in sha256.c
and the AUTHORS file was updated with information that the
code had come from Crypto++ but via 7-Zip. I know I had viewed
7-Zip's SHA-256 code but back then the C code has been identical
enough with Crypto++, so I don't why I thought the author info
would need that extra step via 7-Zip for this single file.
Another error is that I had mixed sha.* and shacal2.* files
when checking for author info in Crypto++. The shacal2.* files
aren't related to liblzma's sha256.c and thus Kevin Springle's
code in Crypto++ isn't either.
If liblzma is configured with --disable-clmul-crc
CFLAGS="-msse4.1 -mpclmul", then it will fail to compile because the
generic version must be used but the CRC tables were not included.
The code was using HAVE_FUNC_ATTRIBUTE_IFUNC instead of CRC_USE_IFUNC.
With ARM64, ifunc is incompatible because it requires non-inline
function calls for runtime detection.
Even though the proper name for the architecture is aarch64, this
project uses ARM64 throughout. So the rename is for consistency.
Additionally, crc32_arm64.h was slightly refactored for the following
changes:
* Added MSVC, FreeBSD, and macOS support in
is_arch_extension_supported().
* crc32_arch_optimized() now checks the size when aligning the
buffer.
* crc32_arch_optimized() loop conditions were slightly modified to
avoid both decrementing the size and incrementing the buffer
pointer.
* Use the intrinsic wrappers defined in <arm_acle.h> because GCC and
Clang name them differently.
* Minor spacing and comment changes.
The CRC_GENERIC is now split into CRC32_GENERIC and CRC64_GENERIC, since
the ARM64 optimizations will be different between CRC32 and CRC64.
For the same reason, CRC_ARCH_OPTIMIZED is split into
CRC32_ARCH_OPTIMIZED and CRC64_ARCH_OPTIMIZED.
ifunc will only be used with x86-64 CLMUL because the runtime detection
methods needed with ARM64 are not compatible with ifunc.
The CRC32 instructions in ARM64 can calculate the CRC32 result
for 8 bytes in a single operation, making the use of ARM64
instructions much faster compared to the general CRC32 algorithm.
Optimized CRC32 will be enabled if ARM64 has CRC extension
running on Linux.
Signed-off-by: Chenxi Mao <chenxi.mao2013@gmail.com>
Not all RISC-V processors support fast unaligned access so
it's better to read only one byte in the main loop. This can
be faster even on x86-64 when compared to reading 32 bits at
a time as half the time the address is only 16-bit aligned.
The downside is larger code size on archs that do support
fast unaligned access.
The new Filter ID is 0x0B.
Thanks to Chien Wong <m@xv97.com> for the initial version of the Filter,
the xz CLI updates, and the Autotools build system modifications.
Thanks to Igor Pavlov for his many contributions to the design of
the filter.
Now crc_simd_body() in crc_x86_clmul.h is only called once
in a translation unit, we no longer need to be so cautious
about ensuring the always-inline behavior.
CRC_CLMUL was split to CRC_ARCH_OPTIMIZED and CRC_X86_CLMUL.
CRC_ARCH_OPTIMIZED is defined when an arch-optimized version is used.
Currently the x86 CLMUL implementations are the only arch-optimized
versions, and these also use the CRC_x86_CLMUL macro to tell when
crc_x86_clmul.h needs to be included.
is_clmul_supported() was renamed to is_arch_extension_supported().
crc32_clmul() and crc64_clmul() were renamed to
crc32_arch_optimized() and crc64_arch_optimized().
This way the names make sense with arch-specific non-CLMUL
implementations as well.
A CLMUL-only build will have the crcxx_clmul() inlined into
lzma_crcxx(). Previously a jump to the extern lzma_crcxx_clmul()
was needed. Notes about shared liblzma on ELF platforms:
- On platforms that support ifunc and -fvisibility=hidden, this
was silly because CLMUL-only build would have that single extra
jump instruction of extra overhead.
- On platforms that support neither -fvisibility=hidden nor linker
version script (liblzma*.map), jumping to lzma_crcxx_clmul()
would go via PLT so a few more instructions of overhead (still
not a big issue but silly nevertheless).
There was a downside with static liblzma too: if an application only
needs lzma_crc64(), static linking would make the linker include the
CLMUL code for both CRC32 and CRC64 from crc_x86_clmul.o even though
the CRC32 code wouldn't be needed, thus increasing code size of the
executable (assuming that -ffunction-sections isn't used).
Also, now compilers are likely to inline crc_simd_body()
even if they don't support the always_inline attribute
(or MSVC's __forceinline). Quite possibly all compilers
that build the code do support such an attribute. But now
it likely isn't a problem even if the attribute wasn't supported.
Now all x86-specific stuff is in crc_x86_clmul.h. If other archs
The other archs can then have their own headers with their own
is_clmul_supported() and crcxx_clmul().
Another bonus is that the build system doesn't need to care if
crc_clmul.c is needed.
is_clmul_supported() stays as inline function as it's not needed
when doing a CLMUL-only build (avoids a warning about unused function).
It requires fast unaligned access to 64-bit integers
and a fast instruction to count leading zeros in
a 64-bit integer (__builtin_ctzll()). This perhaps
should be enabled on some other archs too.
Thanks to Chenxi Mao for the original patch:
https://github.com/tukaani-project/xz/pull/75 (the first commit)
According to the numbers there, this may improve encoding
speed by about 3-5 %.
This enables the 8-byte method on MSVC ARM64 too which
should work but wasn't tested.
This fixes the recent change to lzma_lz_encoder that used memzero
instead of the NULL constant. On some compilers the NULL constant
(always 0) may not equal the NULL pointer (this only needs to guarentee
to not point to valid memory address).
Later code compares the pointers to the NULL pointer so we must
initialize them with the NULL pointer instead of 0 to guarentee
code correctness.
The first member of lzma_lz_encoder doesn't necessarily need to be set
to NULL since it will always be set before anything tries to use it.
However the function pointer members must be set to NULL since other
functions rely on this NULL value to determine if this behavior is
supported or not.
This fixes a somewhat serious bug, where the options_update() and
set_out_limit() function pointers are not set to NULL. This seems to
have been forgotten since these function pointers were added many years
after the original two (code() and end()).
The problem is that by not setting this to NULL we are relying on the
memory allocation to zero things out if lzma_filters_update() is called
on a LZMA1 encoder. The function pointer for set_out_limit() is less
serious because there is not an API function that could call this in an
incorrect way. set_out_limit() is only called by the MicroLZMA encoder,
which must use LZMA1 where set_out_limit() is always set. Its currently
not possible to call set_out_limit() on an LZMA2 encoder at this time.
So calling lzma_filters_update() on an LZMA1 encoder had undefined
behavior since its possible that memory could be manipulated so the
options_update member pointed to a different instruction sequence.
This is unlikely to be a bug in an existing application since it relies
on calling lzma_filters_update() on an LZMA1 encoder in the first place.
For instance, it does not affect xz because lzma_filters_update() can
only be used when encoding to the .xz format.
This is fixed by using memzero() to set all members of lzma_lz_encoder
to NULL after it is allocated. This ensures this mistake will not occur
here in the future if any additional function pointers are added.
lzma_raw_encoder() and lzma_raw_encoder_init() used "options" as the
parameter name instead of "filters" (used by the declaration). "filters"
is more clear since the parameter represents the list of filters passed
to the raw encoder, each of which contains filter options.
lzma_encoder_init() did not check for NULL options, but
lzma2_encoder_init() did. This is more of a code style improvement than
anything else to help make lzma_encoder_init() and lzma2_encoder_init()
more similar.
The macro lzma_attr_visibility_hidden has to be defined to make
fastpos.h usable. The visibility attribute is irrelevant to
fastpos_tablegen.c so simply #define the macro to an empty value.
fastpos_tablegen.c is never built by the included build systems
and so the problem wasn't noticed earlier. It's just a standalone
program for generating fastpos_table.c.
Fixes: https://github.com/tukaani-project/xz/pull/69
Thanks to GitHub user Jamaika1.
In ELF shared libs:
-fvisibility=hidden affects definitions of symbols but not
declarations.[*] This doesn't affect direct calls to functions
inside liblzma as a linker can replace a call to lzma_foo@plt
with a call directly to lzma_foo when -fvisibility=hidden is used.
[*] It has to be like this because otherwise every installed
header file would need to explictly set the symbol visibility
to default.
When accessing extern variables that aren't defined in the
same translation unit, compiler assumes that the variable has
the default visibility and thus indirection is needed. Unlike
function calls, linker cannot optimize this.
Using __attribute__((__visibility__("hidden"))) with the extern
variable declarations tells the compiler that indirection isn't
needed because the definition is in the same shared library.
About 15+ years ago, someone told me that it would be good if
the CRC tables would be defined in the same translation unit
as the C code of the CRC functions. While I understood that it
could help a tiny amount, I didn't want to change the code because
a separate translation unit for the CRC tables was needed for the
x86 assembly code anyway. But when visibility attributes are
supported, simply marking the extern declaration with the
hidden attribute will get identical result. When there are only
a few affected variables, this is trivial to do. I wish I had
understood this back then already.
MinGW (formely a MinGW.org Project, later the MinGW.OSDN Project
at <https://osdn.net/projects/mingw/>) has GCC 9.2.0 as the
most recent GCC package (released 2021-02-02). The project might
still be alive but majority of people have switched to MinGW-w64.
Thus it seems clearer to refer to MinGW-w64 in our API headers too.
Building with MinGW is likely to still work but I haven't tested it
in the recent years.
It properly adds -DLZMA_API_STATIC when compiling code that
will be linked against static liblzma. Having it there on
systems other than Windows does no harm.
See: https://www.msys2.org/docs/pkgconfig/
This partially reverts creating crc_clmul.c
(8c0f9376f5) where is_clmul_supported()
was moved, extern'ed, and renamed to lzma_is_clmul_supported(). This
caused a problem when the function call to lzma_is_clmul_supported()
results in a call through the PLT. ifunc resolvers run very early in
the dynamic loading sequence, so the PLT may not be setup properly at
this point. Whether the PLT is used or not for
lzma_is_clmul_supported() depened upon the compiler-toolchain used and
flags.
In liblzma compiled with GCC, for instance, GCC will go through the PLT
for function calls internal to liblzma if the version scripts and
symbol visibility hiding are not used. If lazy-binding is disabled,
then it would have made any program linked with liblzma fail during
dynamic loading in the ifunc resolver.
Currently crc32 is always enabled, so COND_CHECK_CRC32 must always be
set. Because of this, it makes the recent change to conditionally
compile check/crc_clmul.c appear wrong since that file has CLMUL
implementations for both CRC32 and CRC64.
After forcing crc_simd_body() to always be inlined it caused
-fsanitize=address to fail for lzma_crc32_clmul() and
lzma_crc64_clmul(). The __no_sanitize_address__ attribute was added
to lzma_crc32_clmul() and lzma_crc64_clmul(), but not removed from
crc_simd_body(). ASAN and inline functions behavior has changed over
the years for GCC specifically, so while strictly required we will
keep __attribute__((__no_sanitize_address__)) on crc_simd_body() in
case this becomes a requirement in the future.
Older GCC versions refuse to inline a function with ASAN if the
caller and callee do not agree on sanitization flags
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89124#c3). If the
function was forced to be inlined, it will not compile if the callee
function has __no_sanitize_address__ but the caller doesn't.
After testing a 32-bit Release build on MSVC, only lzma_crc64_clmul()
has the bug. crc_simd_body() and lzma_crc32_clmul() do not need the
optimizations disabled.
Forcing this to be inline has a significant speed improvement at the
cost of a few repeated instructions. The compilers tested on did not
inline this function since it is large and is used twice in the same
translation unit.
This macro must be used instead of the inline keyword. On MSVC, it is
a replacement for __forceinline which is an MSVC specific keyword that
should not be used with inline (it will issue a warning if it is).
It does not use a build system check to determine if
__attribute__((__always_inline__)) since all compilers that can use
CLMUL extensions (except the special case for MSVC) should support this
attribute. If this assumption is incorrect then it will result in a bug
report instead of silently producing slow code.
A detailed description of the three dispatch methods was added. Also,
duplicated comments now only appear in crc32_fast.c or were removed from
both crc32_fast.c and crc64_fast.c if they appeared in crc_clmul.c.
Both crc32_clmul() and crc64_clmul() are now exported from
crc32_clmul.c as lzma_crc32_clmul() and lzma_crc64_clmul(). This
ensures that is_clmul_supported() (now lzma_is_clmul_supported()) is
not duplicated between crc32_fast.c and crc64_fast.c.
Also, it encapsulates the complexity of the CLMUL implementations into a
single file and reduces the complexity of crc32_fast.c and crc64_fast.c.
Before, CLMUL code was present in crc32_fast.c, crc64_fast.c, and
crc_common.h.
During the conversion, various cleanups were applied to code (thanks to
Lasse Collin) including:
- Require using semicolons with MASK_/L/H/LH macros.
- Variable typing and const handling improvements.
- Improvements to comments.
- Fixes to the pragmas used.
- Removed unneeded variables.
- Whitespace improvements.
- Fixed CRC_USE_GENERIC_FOR_SMALL_INPUTS handling.
- Silenced warnings and removed the need for some #pragmas
The C standards don't allow an empty translation unit which can be
avoided by declaring something, without exporting any symbols.
When I committed f644473a21 I had
a feeling that some specific toolchain somewhere didn't like
empty object files (assembler or maybe "ar" complained) but
I cannot find anything to confirm this now. Quite likely I
remembered nonsense. I leave this here as a note to my future self. :-)
When the generic fast crc64 method is used, then we omit
lzma_crc64_table[][]. Similar to
d9166b52cf, we can avoid compiler warnings
with -Wempty-translation-unit (Clang) or -pedantic (GCC) by creating a
never used typedef instead of an extra symbol.
xrealloc() is obviously incorrect, modern GCC docs even
mention realloc() as an example where this attribute
cannot be used.
liblzma's lzma_alloc() and lzma_alloc_zero() would be
correct uses most of the time but custom allocators
may use a memory pool or otherwise hold the pointer
so aliasing issues could happen in theory.
The xstrdup() case likely was correct but I removed it anyway.
Now there are no __malloc__ attributes left in the code.
The allocations aren't in hot paths so this should make
no practical difference.
The argument to vli_ceil4() should always guarantee the return value
is also a valid lzma_vli. Thus the highest three valid lzma_vli values
are invalid arguments. All uses of the function ensure this so the
assert is updated to match this.
This was not a security bug since there was no path to overflow
UINT64_MAX in lzma_index_append() or when it calls index_file_size().
The bug was discovered by a failing assert() in vli_ceil4() when called
from index_file_size() when unpadded_sum (the sum of the compressed size
of current Stream and the unpadded_size parameter) exceeds LZMA_VLI_MAX.
Previously, the unpadded_size parameter was checked to be not greater
than UNPADDED_SIZE_MAX, but no check was done once compressed_base was
added.
This could not have caused an integer overflow in index_file_size() when
called by lzma_index_append(). The calculation for file_size breaks down
into the sum of:
- Compressed base from all previous Streams
- 2 * LZMA_STREAM_HEADER_SIZE (size of the current Streams header and
footer)
- stream_padding (can be set by lzma_index_stream_padding())
- Compressed base from the current Stream
- Unpadded size (parameter to lzma_index_append())
The sum of everything except for Unpadded size must be less than
LZMA_VLI_MAX. This is guarenteed by overflow checks in the functions
that can set these values including lzma_index_stream_padding(),
lzma_index_append(), and lzma_index_cat(). The maximum value for
Unpadded size is enforced by lzma_index_append() to be less than or
equal UNPADDED_SIZE_MAX. Thus, the sum cannot exceed UINT64_MAX since
LZMA_VLI_MAX is half of UINT64_MAX.
Thanks to Joona Kannisto for reporting this.
To workaround Automake lacking Windows resource compiler support, an
empty source file is compiled to overwrite the resource files for static
library builds. Translation units without an external declaration are
not allowed by the C standard and result in a warning when used with
-Wempty-translation-unit (Clang) or -pedantic (GCC).
Clang 16.0.0 and earlier have a bug that the ifunc resolver function
triggers the -Wunused-function warning. The resolver function is static
and only "used" by the __attribute__((__ifunc()__)).
At this time, the bug is still unresolved, but has been reported:
https://github.com/llvm/llvm-project/issues/63957
This is not a problem in GCC.
This further improves the documentation from commit
f36ca7982f. The previous wording of
"supported options" was slightly misleading since the options that are
printed are the ones that are relevant for encoding/decoding. It is not
about which options can or must be specified.
Commit 78704f36e7 added an empty
initializer {} to prevent a warning. The empty initializer is a GNU
extension and results in a build failure on MSVC. The -wpedantic flag
warns about empty initializers.
This change only impacts the compiler warning since it was impossible
for the wait_abs struct in stream_encode_mt() to be used before it was
initialized since mythread_condtime_set() will always be called before
mythread_cond_timedwait().
Since the mythread.h code is different between the POSIX and
Windows versions, this warning was only present on Windows builds.
Thanks to Arthur S for reporting the warning and providing an initial
patch.
In lzma_memcmplen(), the <intrin.h> header file is only included if
_MSC_VER and _M_X64 are both defined but _BitScanForward64() was
previously used if _M_X64 was defined. GCC for MSYS2 defines _M_X64 but
not _MSC_VER so _BitScanForward64() was used without including
<intrin.h>.
Now, lzma_memcmplen() will use __builtin_ctzll() for MSYS2 GCC builds as
expected.
The ifunc method avoids indirection via the function pointer
crc64_func. This works on GNU/Linux and probably on FreeBSD too.
The previous __attribute((__constructor__)) method is kept for
compatibility with ELF platforms which do support ifunc.
The ifunc method has some limitations, for example, building
liblzma with -fsanitize=address will result in segfaults.
The configure option --disable-ifunc must be used for such builds.
Thanks to Hans Jansen for the original patch.
Closes: https://github.com/tukaani-project/xz/pull/53
Reword "options required" to "supported options". The previous may have
suggested that the options listed were all required anytime a filter is
used for encoding or decoding. The reword makes this more clear that
adjusting the options is optional.
The lzma_mt_block_size() was previously just an internal function for
the multithreaded .xz encoder. It is used to provide a recommended Block
size for a given filter chain.
This function is helpful to determine the maximum Block size for the
multithreaded .xz encoder when one wants to change the filters between
blocks. Then, this determined Block size can be provided to
lzma_stream_encoder_mt() in the lzma_mt options parameter when
intializing the coder. This requires one to know all the filter chains
they are using before starting to encode (or at least the filter chain
that will need the largest Block size), but that isn't a bad limitation.
Legacy Windows did not need to #include <intrin.h> to use the MSVC
intrinsics. Newer versions likely just issue a warning, but the MSVC
documentation says to include the header file for the intrinsics we use.
GCC and Clang can "pretend" to be MSVC on Windows, so extra checks are
needed in tuklib_integer.h to only include <intrin.h> when it will is
actually needed.
The \mainpage command is used in the first block of comments in lzma.h.
This changes the previously nearly empty index.html to use the first
comment block in lzma.h for its contents.
lzma.h is no longer documented separately, but this is for the better
since lzma.h only defined a few macros that users do not need to use.
The individual API header files all have a disclaimer that they should
not be #included directly, so there should be no confusion on the fact
that lzma.h should be the only header used by applications.
Additionally, the note "See ../lzma.h for information about liblzma as
a whole." was removed since lzma.h is now the main page of the
generated HTML and does not have its own page anymore. So it would be
confusing in the HTML version and was only a "nice to have" when
browsing the source files.
Lasse Collin
Jia Tan
Chenxi Mao
Hans Jansen
Dimitri Papadopoulos Orfanos