This assert slows the loop down by 10x. We can get similar
coverage by asserting at the beginning & end of the loop.
We need this fix because Debian compiles zstd with asserts
enabled. Separately, we should ask them why, and if they would
consider disabling asserts in their builds. Since we don't
optimize for assert enabled builds.
Fixes Issue #3150.
The zstd symlinks, notably `zstdcat`, weren't working as expected
because only the `tests/cli-tests/bin/zstd` wrapper was symlinked. We
still invoked `zstd` with the name `zstd`. The fix is to create a
directory of zstd symlinks in `tests/cli-tests/bin/symlinks` for each
name that zstd recognizes. And when `tets/cli-tests/bin/zstd` is
invoked, it selects the correct symlink to call.
See the test `zstd-cli/zstdcat.sh` for an example of how it would work.
* It couldn't detect that the `fastCoverParams` can't be non-null, since it was just an assertion.
* It thought we were accesing `wksp->dtable` beyond the bounds because we were using it to set the `workSpace` value. Instead, compute the workspace size used in a different way.
Add cli-tests to `make test`. This adds a `python3` dependency to `make
test`, but not `make check`. We could make this dependency optional by
skipping the tests if `python3` is not present.
`datagen` was printing a `\n` even when it had no other output. Raise
the output level for the final `\n` to the minimum output level used.
This minor bug was caught by the new testing framework.
The option `--auto-threads` should still be accepted and parsed, even if
`ZSTD_MULTITHREAD` is not defined. It doesn't mean anything, but we
should still accept the option. Since we want scripts to be able to work
generically.
This bug was caught by tests I added to the new testing framework.
Passing 0 inputs to `DiB_shuffle()` caused an assertion failure where
it should just return.
A test is added in a later commit, with the initial introduction of the
new testing framework.
Fixes#3007.
Knowing the version of zlib/lz4/lzma we're linking against is very
useful for debugging issues with those libraries, so print it out in the
verbosity 4 version output.
Also print this information at the top of `playTests.sh`.
When decompressing with `-q` and an output file, the progress bar was mistakenly printed. This is a minimal fix, with a larger refactor to be stacked on top of it.
Fixes#2967.
oss-fuzz uncovered a scenario where we're evaluating the cost of litLength = 131072,
which can't be represented in the zstd format, so we accessed 1 beyond LL_bits.
Fix the issue by making it cost 1 bit more than litLength = 131071.
There are still follow ups:
1. This happened because literals_cost[0] = 0, so the optimal parser chose 36 literals
over a match. Should we bound literals_cost[literal] > 0, unless the block truly only
has one literal value?
2. When no matches are found, the cost model isn't updated. In this case no matches were
found for an entire block. So the literals cost model wasn't updated at all. That made
the optimal parser think literals_cost[0] = 0, where it is actually quite high, since
the block was entirely random noise.
Credit to OSS-Fuzz.
After merging #2951 I realized that we will want to explicitly disable
assembly when we aren't including the assembly source file. Otherwise,
if some non clang/gcc compiler does support assembly, it would fail to
build.
Move portability macros to `lib/common/portability_macros.h`. This file
only contains platform/feature detection (e.g. 0/1 macros). This file is
shared between C and ASM code, so it cannot include any C code.
Rename `HUF_` ASM macros to be `ZSTD_` prefixed, and move to the new
header.
Restrict `ZSTD_ASM_SUPPORTED` to `__GNUC__`, because we need the GAS
assembler.
Finally, only include the ASM code if we are actually going to use it.
This disables it on all Windows platforms, which should resolve the
problem brought up in Issue #2789.
Use the same trick as we did for zstd_lazy in PR #2828:
* Create one search function specialization for each (dictMode, mls).
* Select the search function pointer at the top of the match finder.
Additionally, we no longer inline `ZSTD_compressBlock_opt_generic` into
every function, since `dictMode` is no longer used as a template. Create
two specializations, for opt levels 0 and 2, and call one of the two
specializations.
Lastly, remove the hack that disabled inlining for zstd_opt for the
Linux Kernel, as we've gotten most of the benefit already.
Compilation time sees a ~4x reduction:
| Compiler | Flags | Dev Time (s) | PR Time (s) | Delta |
|----------|----------------------------------|--------------|-------------|-------|
| gcc | -O3 | 10.1 | 2.3 | -77% |
| gcc | -O3 -fsanitize=address,undefined | 61.1 | 10.2 | -83% |
| clang | -O3 | 9.0 | 2.1 | -76% |
| clang | -O3 -fsanitize=address,undefined | 33.5 | 5.1 | -84% |
Build size is reduced by 150KB - 200KB:
| Compiler | Dev libzstd.a Size (B) | PR libzstd.a Size (B) | Delta |
|----------|------------------------|-----------------------|-------|
| gcc | 1327476 | 1177108 | -11% |
| clang | 1378324 | 1167780 | -15% |
There is a <2% speed loss in all cases:
| Compiler | Level | Dev Speed (MB/s) | PR Speed (MB/s) | Delta |
|----------|-------|------------------|-----------------|--------|
| gcc | 16 | 4.78 | 4.72 | -1.25% |
| gcc | 17 | 3.49 | 3.46 | -0.85% |
| gcc | 18 | 2.92 | 2.86 | -2.04% |
| gcc | 19 | 2.61 | 2.61 | 0.00% |
| clang | 16 | 4.69 | 4.80 | 2.34% |
| clang | 17 | 3.53 | 3.49 | -1.13% |
| clang | 18 | 2.86 | 2.85 | -0.34% |
| clang | 19 | 2.61 | 2.61 | 0.00% |
Fixes Issue #2862.
short-tests-0 were silently failing. I think because of the && make clean construction. Switch to ; instead.
Also fix all the test failures that were exposed.
`make all` is failing on CircleCI because it is missing Docker. Move that test
to GitHub actions, and switch the pedantic CircleCI test to `make allmost`.
gcc-5 didn't like the l-value overload for defaulted operator=. There is
no reason it needs to be l-value overloaded, so just remove it.
I'm not sure why the build broke for @mckaygerhard in Issue #2811, since
this code hasn't changed since it was added. But, there is no harm in
fixing it.
Fixes issue #2811.
Tests that libzstd.so doesn't have the exec-stack bit set using
readelf. If the stack is marked executable systemd will refuse
to link against zstd. We now test that it isn't set on every PR.
Adds a test for PR #2857
Fixes Issue #2865
Allow the `dictContentSize` to be any size. The finalized dictionary
content size must be at least as large as the maximum repcode (8). So we
add zero bytes to the dictionary to ensure that we meet that
requirement.
I've removed this restriction because its been causing us headaches when
people complain that dictionary training failed. It fails because there
isn't enough useful content to put in the dictionary. Either because
every sample is exactly the same and less than ZDICT_CONTENTSIZE_MIN bytes,
or there isn't enough content. Instead, we should succeed in creating
the dictionary, and it is up to the user to decide if it is worthwhile.
It is possible that the tables alone provide enough value.
NOTE: This allows us to produce dictionaries with finalized
`dictContentSize < ZDICT_CONTENTSIZE_MIN`. But, they are still valid
zstd dictionaries. We could remove the `ZDICT_CONTENTSIZE_MIN` macro,
but I've decided to leave that for now, so we don't break users.
* When dynamic dispatching to bmi2 add lzcnt and bmi to the
TARGET_ATTRIBUTE.
* Centralize the bmi2 TARGET_ATTRIBUTE definition to
BMI2_TARGET_ATTRIBUTE so we can change it in the future.
* Only enable bmi2 when both bmi1 & bmi2 are supported. There shouldn't
be any cases where bmi2 is supported but bmi1 isn't. But, since we are
using the instruction we should check bmi1 as well.
It is no longer necessary to get good performance, there is only a small
speed difference between -O2 and -O3, so just stick to the default of
-O2. I've measured neutral compression speed and a ~3% decompression
speed loss in userspace with clang & gcc. I've also measured neutral
compression speed and a ~1% decompression speed loss in the kernel
benchmarks.
This also fixes the stack space usage on parisc. The compiler was buggy
for -O3 and used ~3KB of stack space for several functions. With -O2 the
problem is completely resolved, and stack space is back to a few hundred
bytes.
Additionally, we get a large code size win on gcc:
| Compiler | Before (Bytes) | After (Bytes) | Delta (Bytes) |
|----------|----------------|---------------|---------------|
| gcc-11 | 952754 | 738954 | -213800 |
| clang-12 | 976290 | 938826 | -37464 |
The optimal parser is unlikely to be used in the linux kernel in
practice. There is no reason these functions should be force inlined,
since we aren't gaining anything, and are losing build size.
| Compiler | Before (Bytes) | After (Bytes) | Delta (Bytes) |
|----------|----------------|---------------|---------------|
| gcc-11 | 1142090 | 952754 | -189336 |
| clang-12 | 1228402 | 976290 | -252112 |
This is a temporary solution pending the resolution of PR #2862 in the
`dev` branch.
Take the same approach as in PR #2828 [0] to remove functions that force
inline many function bodies and `switch`. Instead, create one function per
"template" combination, and then switch between these functions. This
allows the compiler to break the large function into many small
functions, which generally helps codegen.
Also, in the `extDict` modes when there is no ext-dict, call the top
level function instead of the force inlined one, to save on code size.
I'm specifically doing this because gcc on the parisc architecture doesn't
handle the large function body well, and ends up using a lot of excess
stack space. Outlining these functions fixes it.
Speed up compilation times by moving each specialized search function
into its own function. This is faster because compilers can handle many
smaller functions much faster than one gigantic function. The previous
approach generated one giant function with `switch` statements and
inlining to select the implementation.
| Compiler | Flags | Dev Time (s) | PR Time (s) | Delta |
|----------|-------------------------------------|--------------|-------------|-------|
| gcc | -O3 | 16.5 | 5.6 | -66% |
| gcc | -O3 -g -fsanitize=address,undefined | 158.9 | 38.2 | -75% |
| clang | -O3 | 36.5 | 5.5 | -85% |
| clang | -O3 -g -fsanitize=address,undefined | 27.8 | 17.5 | -37% |
This also reduces the binary size because the search functions are no
longer inlined into the main body.
| Compiler | Dev libzstd.a Size (B) | PR libzstd.a Size (B) | Delta |
|----------|------------------------|-----------------------|-------|
| gcc | 1563868 | 1308844 | -16% |
| clang | 1924372 | 1376020 | -28% |
Finally, the performance is not impacted significantly by this change,
in fact we generally see a small speed boost.
| Compiler | Level | Dev Speed (MB/s) | PR Speed (MB/s) | Delta |
|----------|-------|------------------|-----------------|-------|
| gcc | 5 | 110.6 | 110.0 | -0.5% |
| gcc | 7 | 70.4 | 72.2 | +2.5% |
| gcc | 9 | 53.2 | 53.5 | +0.5% |
| gcc | 13 | 12.7 | 12.9 | +1.5% |
| clang | 5 | 113.9 | 110.4 | -3.0% |
| clang | 7 | 67.7 | 70.6 | +4.2% |
| clang | 9 | 51.9 | 52.2 | +0.5% |
| clang | 13 | 12.4 | 13.3 | +7.2% |
The compression strategy is unmodified in this PR, so the compressed size
should be exactly the same. I may have a follow up PR to slightly improve
the compression ratio, if it doesn't cost too much speed.
Fix underflow of `nbCompares` by switching to an `int` and comparing
`nbCompares > 0`. This is a minimal fix, because I don't want to change
the logic. These loops seem to be doing `nbCompares + 1` comparisons.
The bug was reported by Dan Carpenter and found by Smatch static
checker.
https://lore.kernel.org/all/20211008063704.GA5370@kili/
There is no minimum value check, so the parameter could be negative.
Switch to the standard pattern of using `BOUNDCHECK()`.
The bug was reported by Dan Carpenter and found by Smatch static
checker.
https://lore.kernel.org/all/20211008063704.GA5370@kili/
PR #2784 introduced a bug in the decompressor that caused some valid
inputs to fail to decompress. The bitstream isn't reloaded after the 4X*
loop if the number of elements remaining is small enough, causing us to
read more bits than are available in the bitcontainer.
This was caught by the MSAN fuzzer in OSS-Fuzz because the assembly
implementation isn't used in the MSAN build.
Credit to OSS-Fuzz.
Test the kernel build with the standard warnings enabled so that we
don't miss issues like fixed in PR #2802.
Stacked on top of PR #2802 so CI passes, so it includes the fix. But, I
won't merge until it is merged, so @solbjorn gets the credit for the
fix.
Switch to a macro `ZSTD_FALLTHROUGH;` instead of a comment. On supported
compilers this uses an attribute, otherwise it becomes a comment.
This is necessary to be compatible with clang's `-Wfall-through`, and
gcc's `-Wfall-through=2` which don't support comments. Without this the
linux build emits a bunch of warnings.
Also add a test to CI to ensure that we don't regress.
Instead of calling `ZSTD_compress_advanced()` and
`ZSTD_initCStream_advanced()`, which each take a `ZSTD_parameters` by
value, use the new advanced API.
Stack usage went from 2024 -> 1944.
