* Move all ZSTDMT parameter setting code to ZSTD_CCtxParams_*Parameter().
ZSTDMT now calls these functions, so we can keep all the logic in the
same place.
* Clean up `ZSTD_CCtx_setParameter()` to only add extra checks where needed.
* Clean up `ZSTDMT_initJobCCtxParams()` by copying all parameters by default,
and then zeroing the ones that need to be zeroed. We've missed adding several
parameters here, and it makes more sense to only have to update it if you
change something in ZSTDMT.
* Add `ZSTDMT_cParam_clampBounds()` to clamp a parameter into its valid
range. Use this to keep backwards compatibility when setting ZSTDMT parameters,
which clamp into the valid range.
Test a positive compression level with uncompressed literals,
and a negative compression level with compressed literals.
I double checked the `results.csv` and made sure that the compressed
sizes make sense.
* Add configs that test multithreading, LDM, and setting explicit
parameters.
* Update the `compress cctx` method to accept `ZSTD_parameters`.
* Compile against the multithreaded `libzstd.a`.
* Update `results.csv` for the new configs.
Unless you think there are more configs/methods I should test, I think
we have a fairly wide set of configs/methods, so I'll pause adding
more for now.
Compare the input and output files by their inode number and
refuse to open the output file if the input file is the same.
This doesn't work when (de)compressing multiple files to a single
file, but that is a very uncommon use case, mostly used for
benchmarking by me.
Fixes#1422.
* Fix `ZSTD_estimateCCtxSize()` with negative levels.
* Fix `ZSTD_estimateCStreamSize()` with negative levels.
* Add a unit test to test for this error.
The `--no-progress` flag disables zstd's progress bars, but leaves
the summary.
I've added simple tests to `playTests.sh` to make sure the parsing
works.
When we switched `ZSTD_SKIPPABLEHEADERSIZE` to a macro, the places where we do:
MEM_readLE32(ptr) + ZSTD_SKIPPABLEHEADERSIZE
can now overflow `(unsigned)-8` to `0` and we infinite loop. We now check
the frame size and reject sizes that overflow a U32.
Note that this bug never made it into a release, and was only in the dev branch
for a few days.
Credit to OSS-Fuzz
Dictionaries are prebuilt and saved as part of the data object.
The config decides whether or not to use the dictionary if it is
available. Configs that require dictionaries are only run with
data that have dictionaries. The method will skip configs that are
irrelevant, so for example ZSTD_compress() will skip configs with
dictionaries.
I've also trimmed the silesia source to 1MB per file (12 MB total),
and added 500 samples from the github data set with a dictionary.
I've intentionally added an extra line to the `results.csv` to make
the nightly build fail, so that we can see how CircleCI reports it.
Full list of changes:
* Add pre-built dictionaries to the data.
* Add `use_dictionary` and `no_pledged_src_size` flags to the config.
* Add a config using a dictionary for every level.
* Add a config that specifies no pledged source size.
* Support dictionaries and streaming in the `zstdcli` method.
* Add a context-reuse method using `ZSTD_compressCCtx()`.
* Clean up the formatting of the `results.csv` file to align columns.
* Add `--data`, `--config`, and `--method` flags to constrain each
to a particular value. This is useful for debugging a failure
or debugging a particular config/method/data.
The regression tests run nightly or on the `regression`
branch for convenience. The results get uploaded as the
artifacts of the job. If they change, check the diff
printed in the job. If all is well, download the new
results and commit them to the repo.
This code will only run on a UNIX like platform. It
could be made to run on Windows, but I don't think that
it is necessary. It also uses C99.
* data: This module defines the data to run tests on.
It downloads data from a URL into a cache directory,
checks it against a checksum, and unpacks it. It also
provides helpers for accessing the data.
* config: This module defines the configs to run tests
with. A config is a set of API parameters and a set of
CLI flags.
* result: This module is a helper for method that defines
the result type.
* method: This module defines the compression methods
to test. It is what runs the regression test using the
data and the config. It reports the total compressed
size, or an error/skip.
* test: This is the test binary that runs the tests for
every (data, config, method) tuple, and prints the
results to the output file and stderr.
* results.csv: The results that the current commit is
expected to produce.
We could allocate up to 2^28 bytes of memory when using 2 threads with
window log = 24. Now, we limit it to 2^26 bytes of memory when not running
big tests.
I chose max window log = 22 since that is the maximum source size when
big tests are disabled. Hopefully this will be enough to reduce or
eliminate the test failures.
* Updates CircleCI to use workflows.
We can now specify any number of test jobs to run in parallel.
* Switch the image to `buildpack-deps:trusty` which is only 500 MB
instead of 7 GB, so that saves 7 minutes to download it if it isn't
already cached on the host.
* Publish the source tarball and sha256sum as artifacts.
* If the `GITHUB_TOKEN` environment variable is set, we will also
add the tarball + sha256sum to the tagged release, after manual
approval.
We could undersize the literals buffer by up to 11 bytes,
due to a combination of 2 bugs:
* The literals buffer didn't have `WILDCOPY_OVERLENGTH` extra
space, like it is supposed to.
* We didn't check the literals buffer size in `ZSTD_sufficientBuff()`.
When the primary normalization method fails, and
`(1 << tableLog) == (maxSymbolValue + 1)`, and every symbol gets assigned
normalized weight 1 or -1 in the first loop, then the next division can
raise `SIGFPE`.
The correct parameters are used once, but once `ZSTD_resetCStream()` is
called the default parameters (level 3) are used. Fix this by setting
`requestedParams` in the `ZSTD_initCStream*()` functions.
The added tests both fail before this patch and pass after.
In the new advanced API, adjust the parameters even if they are explicitly
set. This mainly applies to the `windowLog`, and accordingly the `hashLog`
and `chainLog`, when the source size is known.
This edge case is only possible with the new optimal encoding selector,
since before zstd would always choose `set_basic` for small numbers of
sequences.
Fix `FSE_readNCount()` to support buffers < 4 bytes.
Credit to OSS-Fuzz
Estimate the cost for using FSE modes `set_basic`, `set_compressed`, and
`set_repeat`, and select the one with the lowest cost.
* The cost of `set_basic` is computed using the cross-entropy cost
function `ZSTD_crossEntropyCost()`, using the normalized default count
and the count.
* The cost of `set_repeat` is computed using `FSE_bitCost()`. We check the
previous table to see if it is able to represent the distribution.
* The cost of `set_compressed` is computed with the entropy cost function
`ZSTD_entropyCost()`, together with the cost of writing the normalized
count `ZSTD_NCountCost()`.
The cover algorithm selects one segment per epoch, and it selects the epoch
size such that `epochs * segmentSize ~= dictSize`. Selecting less epochs
gives the algorithm more candidates to choose from for each segment it
selects, and then it will loop back to the first epoch when it hits the
last one.
The trade off is that now it takes longer to select each segment, since it
has to look at more data before making a choice.
I benchmarked on the following data sets using this command:
```sh
$ZSTD -T0 -3 --train-cover=d=8,steps=256 $DIR -r -o dict && $ZSTD -3 -D dict -rc $DIR | wc -c
```
| Data set | k (approx) | Before | After | % difference |
|--------------|------------|----------|----------|--------------|
| GitHub | ~1000 | 738138 | 746610 | +1.14% |
| hg-changelog | ~90 | 4295156 | 4285336 | -0.23% |
| hg-commands | ~500 | 1095580 | 1079814 | -1.44% |
| hg-manifest | ~400 | 16559892 | 16504346 | -0.34% |
There is some noise in the measurements, since small changes to `k` can
have large differences, which is why I'm using `steps=256`, to try to
minimize the noise. However, the GitHub data set still has some noise.
If I run the GitHub data set on my Mac, which presumably lists directory
entries in a different order, so the dictionary builder sees the files in
a different order, or I use `steps=1024` I see these results.
| Run | Before | After | % difference |
|------------|--------|--------|--------------|
| steps=1024 | 738138 | 734470 | -0.50% |
| MacBook | 738451 | 737132 | -0.18% |
Question: Should we expose this as a parameter? I don't think it is
necessary. Someone might want to turn it up to exchange a much longer
dictionary building time in exchange for a slightly better dictionary.
I tested `2`, `4`, and `16`, and `4` got most of the benefit of `16`
with a faster running time.
The new advanced API basically set `requestedParams = appliedParams` when
using a dictionary. This halted all parameter adjustment, which can hurt
compression ratio if, for example, the window log is small for the first
call, but the rest of the files are large.
This patch fixes the bug, and checks that the `requestedParams` don't change
in the new advanced API when using a dictionary, and generally in the fuzzer.
Zstdmt uses prefixes to load the overlap between segments. Loading extra
positions makes compression non-deterministic, depending on the previous
job the context was used for. Since loading extra position takes extra
time as well, only do it when creating a `ZSTD_CDict`.
Fixes#1077.
The `avgJobSize` must not be lower than 256 KB for single-pass mode.
In `zstd.h` we say the minimum value for `ZSTD_p_jobSize` is 1 MB,
so ensure that we always pick a size >= 1 MB.
Found by libFuzzer fuzzer tests with large input limits.
Integrate ldm into zstdmt by running it in serial and in order in the first
step of each job, in the same place as the hash gets updated. The input
buffer is sized to fit the whole LDM window and 2 full buffers of slack.
Input buffers cannot be reused until the LDM step is done with them.
After the LDM step is finished, the jobs don't actually have access to the
full window, only the overlap.
Tested on a few different multi-GB files with and without sanitizers,
and with different numbers of threads.
* Computes the XXH hash in the worker threads.
* Workers get a sequence number and wait until ther number shows up. On
error, ensures that its sequence is finished, so future threads don't
get blocked.
* Sets up for ldm integration, which will go in the same spot.
Setting `loadedDictEnd` was accidently removed from `ZSTD_loadDictionaryContent()`,
which means that dictionary compression will only be able to reference the parts of
the dictionary within the window. The spec allows us to reference the entire
dictionary so long as even one byte is in the window.
`ZSTD_enforceMaxDist()` incorrectly always allowed offsets up to `loadedDictEnd`
beyond the window, even once the dictionary was out of range.
When overflow protection kicked in, the check `current > loadedDictEnd + maxDist`
is incorrect if `loadedDictEnd` isn't reset back to zero. `current` could be reset
below the value, which would incorrectly allow references beyond the window. This
bug is present in `master`, but is very hard to trigger, since it requires both
dictionaries and data which triggers overflow correction.
Summary:
Allocate a single input buffer large enough to house each job, as well as
enough space for the IO thread to write 2 extra buffers. One goes in the
`POOL` queue, and one to fill, and then block on a full `POOL` queue.
Since we can't overlap with the prefix, we allocate space for 3 extra
input buffers.
Test Plan:
* CI
* With and without ASAN/UBSAN run zstdmt with different number of threads
on two large binaries, and verify that their checksums match.
* Test on the tip of the zstdmt ldm integration.
Reviewers: cyan
Differential Revision: https://phabricator.intern.facebook.com/D7284007
Tasks: T25664120
Nick Terrell