xref: /freebsd/sys/contrib/zstd/CONTRIBUTING.md (revision 5ff13fbc199bdf5f0572845351c68ee5ca828e71)
1# Contributing to Zstandard
2We want to make contributing to this project as easy and transparent as
3possible.
4
5## Our Development Process
6New versions are being developed in the "dev" branch,
7or in their own feature branch.
8When they are deemed ready for a release, they are merged into "release".
9
10As a consequences, all contributions must stage first through "dev"
11or their own feature branch.
12
13## Pull Requests
14We actively welcome your pull requests.
15
161. Fork the repo and create your branch from `dev`.
172. If you've added code that should be tested, add tests.
183. If you've changed APIs, update the documentation.
194. Ensure the test suite passes.
205. Make sure your code lints.
216. If you haven't already, complete the Contributor License Agreement ("CLA").
22
23## Contributor License Agreement ("CLA")
24In order to accept your pull request, we need you to submit a CLA. You only need
25to do this once to work on any of Facebook's open source projects.
26
27Complete your CLA here: <https://code.facebook.com/cla>
28
29## Workflow
30Zstd uses a branch-based workflow for making changes to the codebase. Typically, zstd
31will use a new branch per sizable topic. For smaller changes, it is okay to lump multiple
32related changes into a branch.
33
34Our contribution process works in three main stages:
351. Local development
36    * Update:
37        * Checkout your fork of zstd if you have not already
38        ```
39        git checkout https://github.com/<username>/zstd
40        cd zstd
41        ```
42        * Update your local dev branch
43        ```
44        git pull https://github.com/facebook/zstd dev
45        git push origin dev
46        ```
47    * Topic and development:
48        * Make a new branch on your fork about the topic you're developing for
49        ```
50        # branch names should be concise but sufficiently informative
51        git checkout -b <branch-name>
52        git push origin <branch-name>
53        ```
54        * Make commits and push
55        ```
56        # make some changes =
57        git add -u && git commit -m <message>
58        git push origin <branch-name>
59        ```
60        * Note: run local tests to ensure that your changes didn't break existing functionality
61            * Quick check
62            ```
63            make shortest
64            ```
65            * Longer check
66            ```
67            make test
68            ```
692. Code Review and CI tests
70    * Ensure CI tests pass:
71        * Before sharing anything to the community, create a pull request in your own fork against the dev branch
72        and make sure that all GitHub Actions CI tests pass. See the Continuous Integration section below for more information.
73        * Ensure that static analysis passes on your development machine. See the Static Analysis section
74        below to see how to do this.
75    * Create a pull request:
76        * When you are ready to share you changes to the community, create a pull request from your branch
77        to facebook:dev. You can do this very easily by clicking 'Create Pull Request' on your fork's home
78        page.
79        * From there, select the branch where you made changes as your source branch and facebook:dev
80        as the destination.
81        * Examine the diff presented between the two branches to make sure there is nothing unexpected.
82    * Write a good pull request description:
83        * While there is no strict template that our contributors follow, we would like them to
84        sufficiently summarize and motivate the changes they are proposing. We recommend all pull requests,
85        at least indirectly, address the following points.
86            * Is this pull request important and why?
87            * Is it addressing an issue? If so, what issue? (provide links for convenience please)
88            * Is this a new feature? If so, why is it useful and/or necessary?
89            * Are there background references and documents that reviewers should be aware of to properly assess this change?
90        * Note: make sure to point out any design and architectural decisions that you made and the rationale behind them.
91        * Note: if you have been working with a specific user and would like them to review your work, make sure you mention them using (@<username>)
92    * Submit the pull request and iterate with feedback.
933. Merge and Release
94    * Getting approval:
95        * You will have to iterate on your changes with feedback from other collaborators to reach a point
96        where your pull request can be safely merged.
97        * To avoid too many comments on style and convention, make sure that you have a
98        look at our style section below before creating a pull request.
99        * Eventually, someone from the zstd team will approve your pull request and not long after merge it into
100        the dev branch.
101    * Housekeeping:
102        * Most PRs are linked with one or more Github issues. If this is the case for your PR, make sure
103        the corresponding issue is mentioned. If your change 'fixes' or completely addresses the
104        issue at hand, then please indicate this by requesting that an issue be closed by commenting.
105        * Just because your changes have been merged does not mean the topic or larger issue is complete. Remember
106        that the change must make it to an official zstd release for it to be meaningful. We recommend
107        that contributors track the activity on their pull request and corresponding issue(s) page(s) until
108        their change makes it to the next release of zstd. Users will often discover bugs in your code or
109        suggest ways to refine and improve your initial changes even after the pull request is merged.
110
111## Static Analysis
112Static analysis is a process for examining the correctness or validity of a program without actually
113executing it. It usually helps us find many simple bugs. Zstd uses clang's `scan-build` tool for
114static analysis. You can install it by following the instructions for your OS on https://clang-analyzer.llvm.org/scan-build.
115
116Once installed, you can ensure that our static analysis tests pass on your local development machine
117by running:
118```
119make staticAnalyze
120```
121
122In general, you can use `scan-build` to static analyze any build script. For example, to static analyze
123just `contrib/largeNbDicts` and nothing else, you can run:
124
125```
126scan-build make -C contrib/largeNbDicts largeNbDicts
127```
128
129### Pitfalls of static analysis
130`scan-build` is part of our regular CI suite. Other static analyzers are not.
131
132It can be useful to look at additional static analyzers once in a while (and we do), but it's not a good idea to multiply the nb of analyzers run continuously at each commit and PR. The reasons are :
133
134- Static analyzers are full of false positive. The signal to noise ratio is actually pretty low.
135- A good CI policy is "zero-warning tolerance". That means that all issues must be solved, including false positives. This quickly becomes a tedious workload.
136- Multiple static analyzers will feature multiple kind of false positives, sometimes applying to the same code but in different ways leading to :
137   + torteous code, trying to please multiple constraints, hurting readability and therefore maintenance. Sometimes, such complexity introduce other more subtle bugs, that are just out of scope of the analyzers.
138   + sometimes, these constraints are mutually exclusive : if one try to solve one, the other static analyzer will complain, they can't be both happy at the same time.
139- As if that was not enough, the list of false positives change with each version. It's hard enough to follow one static analyzer, but multiple ones with their own update agenda, this quickly becomes a massive velocity reducer.
140
141This is different from running a static analyzer once in a while, looking at the output, and __cherry picking__ a few warnings that seem helpful, either because they detected a genuine risk of bug, or because it helps expressing the code in a way which is more readable or more difficult to misuse. These kind of reports can be useful, and are accepted.
142
143## Continuous Integration
144CI tests run every time a pull request (PR) is created or updated. The exact tests
145that get run will depend on the destination branch you specify. Some tests take
146longer to run than others. Currently, our CI is set up to run a short
147series of tests when creating a PR to the dev branch and a longer series of tests
148when creating a PR to the release branch. You can look in the configuration files
149of the respective CI platform for more information on what gets run when.
150
151Most people will just want to create a PR with the destination set to their local dev
152branch of zstd. You can then find the status of the tests on the PR's page. You can also
153re-run tests and cancel running tests from the PR page or from the respective CI's dashboard.
154
155Almost all of zstd's CI runs on GitHub Actions (configured at `.github/workflows`), which will automatically run on PRs to your
156own fork. A small number of tests run on other services (e.g. Travis CI, Circle CI, Appveyor).
157These require work to set up on your local fork, and (at least for Travis CI) cost money.
158Therefore, if the PR on your local fork passes GitHub Actions, feel free to submit a PR
159against the main repo.
160
161### Third-party CI
162A small number of tests cannot run on GitHub Actions, or have yet to be migrated.
163For these, we use a variety of third-party services (listed below). It is not necessary to set
164these up on your fork in order to contribute to zstd; however, we do link to instructions for those
165who want earlier signal.
166
167| Service   | Purpose                                                                                                    | Setup Links                                                                                                                                            | Config Path            |
168|-----------|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------|
169| Travis CI | Used for testing on non-x86 architectures such as PowerPC                                                  | https://docs.travis-ci.com/user/tutorial/#to-get-started-with-travis-ci-using-github <br> https://github.com/marketplace/travis-ci                     | `.travis.yml`          |
170| AppVeyor  | Used for some Windows testing (e.g. cygwin, mingw)                                                         | https://www.appveyor.com/blog/2018/10/02/github-apps-integration/ <br> https://github.com/marketplace/appveyor                                         | `appveyor.yml`         |
171| Cirrus CI | Used for testing on FreeBSD                                                                                | https://github.com/marketplace/cirrus-ci/                                                                                                              | `.cirrus.yml`          |
172| Circle CI | Historically was used to provide faster signal,<br/> but we may be able to migrate these to Github Actions | https://circleci.com/docs/2.0/getting-started/#setting-up-circleci <br> https://youtu.be/Js3hMUsSZ2c <br> https://circleci.com/docs/2.0/enable-checks/ | `.circleci/config.yml` |
173
174Note: the instructions linked above mostly cover how to set up a repository with CI from scratch.
175The general idea should be the same for setting up CI on your fork of zstd, but you may have to
176follow slightly different steps. In particular, please ignore any instructions related to setting up
177config files (since zstd already has configs for each of these services).
178
179## Performance
180Performance is extremely important for zstd and we only merge pull requests whose performance
181landscape and corresponding trade-offs have been adequately analyzed, reproduced, and presented.
182This high bar for performance means that every PR which has the potential to
183impact performance takes a very long time for us to properly review. That being said, we
184always welcome contributions to improve performance (or worsen performance for the trade-off of
185something else). Please keep the following in mind before submitting a performance related PR:
186
1871. Zstd isn't as old as gzip but it has been around for time now and its evolution is
188very well documented via past Github issues and pull requests. It may be the case that your
189particular performance optimization has already been considered in the past. Please take some
190time to search through old issues and pull requests using keywords specific to your
191would-be PR. Of course, just because a topic has already been discussed (and perhaps rejected
192on some grounds) in the past, doesn't mean it isn't worth bringing up again. But even in that case,
193it will be helpful for you to have context from that topic's history before contributing.
1942. The distinction between noise and actual performance gains can unfortunately be very subtle
195especially when microbenchmarking extremely small wins or losses. The only remedy to getting
196something subtle merged is extensive benchmarking. You will be doing us a great favor if you
197take the time to run extensive, long-duration, and potentially cross-(os, platform, process, etc)
198benchmarks on your end before submitting a PR. Of course, you will not be able to benchmark
199your changes on every single processor and os out there (and neither will we) but do that best
200you can:) We've adding some things to think about when benchmarking below in the Benchmarking
201Performance section which might be helpful for you.
2023. Optimizing performance for a certain OS, processor vendor, compiler, or network system is a perfectly
203legitimate thing to do as long as it does not harm the overall performance health of Zstd.
204This is a hard balance to strike but please keep in mind other aspects of Zstd when
205submitting changes that are clang-specific, windows-specific, etc.
206
207## Benchmarking Performance
208Performance microbenchmarking is a tricky subject but also essential for Zstd. We value empirical
209testing over theoretical speculation. This guide it not perfect but for most scenarios, it
210is a good place to start.
211
212### Stability
213Unfortunately, the most important aspect in being able to benchmark reliably is to have a stable
214benchmarking machine. A virtual machine, a machine with shared resources, or your laptop
215will typically not be stable enough to obtain reliable benchmark results. If you can get your
216hands on a desktop, this is usually a better scenario.
217
218Of course, benchmarking can be done on non-hyper-stable machines as well. You will just have to
219do a little more work to ensure that you are in fact measuring the changes you've made not and
220noise. Here are some things you can do to make your benchmarks more stable:
221
2221. The most simple thing you can do to drastically improve the stability of your benchmark is
223to run it multiple times and then aggregate the results of those runs. As a general rule of
224thumb, the smaller the change you are trying to measure, the more samples of benchmark runs
225you will have to aggregate over to get reliable results. Here are some additional things to keep in
226mind when running multiple trials:
227    * How you aggregate your samples are important. You might be tempted to use the mean of your
228    results. While this is certainly going to be a more stable number than a raw single sample
229    benchmark number, you might have more luck by taking the median. The mean is not robust to
230    outliers whereas the median is. Better still, you could simply take the fastest speed your
231    benchmark achieved on each run since that is likely the fastest your process will be
232    capable of running your code. In our experience, this (aggregating by just taking the sample
233    with the fastest running time) has been the most stable approach.
234    * The more samples you have, the more stable your benchmarks should be. You can verify
235    your improved stability by looking at the size of your confidence intervals as you
236    increase your sample count. These should get smaller and smaller. Eventually hopefully
237    smaller than the performance win you are expecting.
238    * Most processors will take some time to get `hot` when running anything. The observations
239    you collect during that time period will very different from the true performance number. Having
240    a very large number of sample will help alleviate this problem slightly but you can also
241    address is directly by simply not including the first `n` iterations of your benchmark in
242    your aggregations. You can determine `n` by simply looking at the results from each iteration
243    and then hand picking a good threshold after which the variance in results seems to stabilize.
2442. You cannot really get reliable benchmarks if your host machine is simultaneously running
245another cpu/memory-intensive application in the background. If you are running benchmarks on your
246personal laptop for instance, you should close all applications (including your code editor and
247browser) before running your benchmarks. You might also have invisible background applications
248running. You can see what these are by looking at either Activity Monitor on Mac or Task Manager
249on Windows. You will get more stable benchmark results of you end those processes as well.
250    * If you have multiple cores, you can even run your benchmark on a reserved core to prevent
251    pollution from other OS and user processes. There are a number of ways to do this depending
252    on your OS:
253        * On linux boxes, you have use https://github.com/lpechacek/cpuset.
254        * On Windows, you can "Set Processor Affinity" using https://www.thewindowsclub.com/processor-affinity-windows
255        * On Mac, you can try to use their dedicated affinity API https://developer.apple.com/library/archive/releasenotes/Performance/RN-AffinityAPI/#//apple_ref/doc/uid/TP40006635-CH1-DontLinkElementID_2
2563. To benchmark, you will likely end up writing a separate c/c++ program that will link libzstd.
257Dynamically linking your library will introduce some added variation (not a large amount but
258definitely some). Statically linking libzstd will be more stable. Static libraries should
259be enabled by default when building zstd.
2604. Use a profiler with a good high resolution timer. See the section below on profiling for
261details on this.
2625. Disable frequency scaling, turbo boost and address space randomization (this will vary by OS)
2636. Try to avoid storage. On some systems you can use tmpfs. Putting the program, inputs and outputs on
264tmpfs avoids touching a real storage system, which can have a pretty big variability.
265
266Also check our LLVM's guide on benchmarking here: https://llvm.org/docs/Benchmarking.html
267
268### Zstd benchmark
269The fastest signal you can get regarding your performance changes is via the in-build zstd cli
270bench option. You can run Zstd as you typically would for your scenario using some set of options
271and then additionally also specify the `-b#` option. Doing this will run our benchmarking pipeline
272for that options you have just provided. If you want to look at the internals of how this
273benchmarking script works, you can check out programs/benchzstd.c
274
275For example: say you have made a change that you believe improves the speed of zstd level 1. The
276very first thing you should use to asses whether you actually achieved any sort of improvement
277is `zstd -b`. You might try to do something like this. Note: you can use the `-i` option to
278specify a running time for your benchmark in seconds (default is 3 seconds).
279Usually, the longer the running time, the more stable your results will be.
280
281```
282$ git checkout <commit-before-your-change>
283$ make && cp zstd zstd-old
284$ git checkout <commit-after-your-change>
285$ make && cp zstd zstd-new
286$ zstd-old -i5 -b1 <your-test-data>
287 1<your-test-data>         :      8990 ->      3992 (2.252), 302.6 MB/s , 626.4 MB/s
288$ zstd-new -i5 -b1 <your-test-data>
289 1<your-test-data>         :      8990 ->      3992 (2.252), 302.8 MB/s , 628.4 MB/s
290```
291
292Unless your performance win is large enough to be visible despite the intrinsic noise
293on your computer, benchzstd alone will likely not be enough to validate the impact of your
294changes. For example, the results of the example above indicate that effectively nothing
295changed but there could be a small <3% improvement that the noise on the host machine
296obscured. So unless you see a large performance win (10-15% consistently) using just
297this method of evaluation will not be sufficient.
298
299### Profiling
300There are a number of great profilers out there. We're going to briefly mention how you can
301profile your code using `instruments` on mac, `perf` on linux and `visual studio profiler`
302on windows.
303
304Say you have an idea for a change that you think will provide some good performance gains
305for level 1 compression on Zstd. Typically this means, you have identified a section of
306code that you think can be made to run faster.
307
308The first thing you will want to do is make sure that the piece of code is actually taking up
309a notable amount of time to run. It is usually not worth optimizing something which accounts for less than
3100.0001% of the total running time. Luckily, there are tools to help with this.
311Profilers will let you see how much time your code spends inside a particular function.
312If your target code snippet is only part of a function, it might be worth trying to
313isolate that snippet by moving it to its own function (this is usually not necessary but
314might be).
315
316Most profilers (including the profilers discussed below) will generate a call graph of
317functions for you. Your goal will be to find your function of interest in this call graph
318and then inspect the time spent inside of it. You might also want to to look at the
319annotated assembly which most profilers will provide you with.
320
321#### Instruments
322We will once again consider the scenario where you think you've identified a piece of code
323whose performance can be improved upon. Follow these steps to profile your code using
324Instruments.
325
3261. Open Instruments
3272. Select `Time Profiler` from the list of standard templates
3283. Close all other applications except for your instruments window and your terminal
3294. Run your benchmarking script from your terminal window
330    * You will want a benchmark that runs for at least a few seconds (5 seconds will
331    usually be long enough). This way the profiler will have something to work with
332    and you will have ample time to attach your profiler to this process:)
333    * I will just use benchzstd as my bencharmking script for this example:
334```
335$ zstd -b1 -i5 <my-data> # this will run for 5 seconds
336```
3375. Once you run your benchmarking script, switch back over to instruments and attach your
338process to the time profiler. You can do this by:
339    * Clicking on the `All Processes` drop down in the top left of the toolbar.
340    * Selecting your process from the dropdown. In my case, it is just going to be labeled
341    `zstd`
342    * Hitting the bright red record circle button on the top left of the toolbar
3436. You profiler will now start collecting metrics from your benchmarking script. Once
344you think you have collected enough samples (usually this is the case after 3 seconds of
345recording), stop your profiler.
3467. Make sure that in toolbar of the bottom window, `profile` is selected.
3478. You should be able to see your call graph.
348    * If you don't see the call graph or an incomplete call graph, make sure you have compiled
349    zstd and your benchmarking script using debug flags. On mac and linux, this just means
350    you will have to supply the `-g` flag alone with your build script. You might also
351    have to provide the `-fno-omit-frame-pointer` flag
3529. Dig down the graph to find your function call and then inspect it by double clicking
353the list item. You will be able to see the annotated source code and the assembly side by
354side.
355
356#### Perf
357
358This wiki has a pretty detailed tutorial on getting started working with perf so we'll
359leave you to check that out of you're getting started:
360
361https://perf.wiki.kernel.org/index.php/Tutorial
362
363Some general notes on perf:
364* Use `perf stat -r # <bench-program>` to quickly get some relevant timing and
365counter statistics. Perf uses a high resolution timer and this is likely one
366of the first things your team will run when assessing your PR.
367* Perf has a long list of hardware counters that can be viewed with `perf --list`.
368When measuring optimizations, something worth trying is to make sure the hardware
369counters you expect to be impacted by your change are in fact being so. For example,
370if you expect the L1 cache misses to decrease with your change, you can look at the
371counter `L1-dcache-load-misses`
372* Perf hardware counters will not work on a virtual machine.
373
374#### Visual Studio
375
376TODO
377
378## Issues
379We use GitHub issues to track public bugs. Please ensure your description is
380clear and has sufficient instructions to be able to reproduce the issue.
381
382Facebook has a [bounty program](https://www.facebook.com/whitehat/) for the safe
383disclosure of security bugs. In those cases, please go through the process
384outlined on that page and do not file a public issue.
385
386## Coding Style
387It's a pretty long topic, which is difficult to summarize in a single paragraph.
388As a rule of thumbs, try to imitate the coding style of
389similar lines of codes around your contribution.
390The following is a non-exhaustive list of rules employed in zstd code base:
391
392### C90
393This code base is following strict C90 standard,
394with 2 extensions : 64-bit `long long` types, and variadic macros.
395This rule is applied strictly to code within `lib/` and `programs/`.
396Sub-project in `contrib/` are allowed to use other conventions.
397
398### C++ direct compatibility : symbol mangling
399All public symbol declarations must be wrapped in `extern “C” { … }`,
400so that this project can be compiled as C++98 code,
401and linked into C++ applications.
402
403### Minimal Frugal
404This design requirement is fundamental to preserve the portability of the code base.
405#### Dependencies
406- Reduce dependencies to the minimum possible level.
407  Any dependency should be considered “bad” by default,
408  and only tolerated because it provides a service in a better way than can be achieved locally.
409  The only external dependencies this repository tolerates are
410  standard C libraries, and in rare cases, system level headers.
411- Within `lib/`, this policy is even more drastic.
412  The only external dependencies allowed are `<assert.h>`, `<stdlib.h>`, `<string.h>`,
413  and even then, not directly.
414  In particular, no function shall ever allocate on heap directly,
415  and must use instead `ZSTD_malloc()` and equivalent.
416  Other accepted non-symbol headers are `<stddef.h>` and `<limits.h>`.
417- Within the project, there is a strict hierarchy of dependencies that must be respected.
418  `programs/` is allowed to depend on `lib/`, but only its public API.
419  Within `lib/`, `lib/common` doesn't depend on any other directory.
420  `lib/compress` and `lib/decompress` shall not depend on each other.
421  `lib/dictBuilder` can depend on `lib/common` and `lib/compress`, but not `lib/decompress`.
422#### Resources
423- Functions in `lib/` must use very little stack space,
424  several dozens of bytes max.
425  Everything larger must use the heap allocator,
426  or require a scratch buffer to be emplaced manually.
427
428### Naming
429* All public symbols are prefixed with `ZSTD_`
430  + private symbols, with a scope limited to their own unit, are free of this restriction.
431    However, since `libzstd` source code can be amalgamated,
432    each symbol name must attempt to be (and remain) unique.
433    Avoid too generic names that could become ground for future collisions.
434    This generally implies usage of some form of prefix.
435* For symbols (functions and variables), naming convention is `PREFIX_camelCase`.
436  + In some advanced cases, one can also find :
437    - `PREFIX_prefix2_camelCase`
438    - `PREFIX_camelCase_extendedQualifier`
439* Multi-words names generally consist of an action followed by object:
440  - for example : `ZSTD_createCCtx()`
441* Prefer positive actions
442  - `goBackward` rather than `notGoForward`
443* Type names (`struct`, etc.) follow similar convention,
444  except that they are allowed and even invited to start by an Uppercase letter.
445  Example : `ZSTD_CCtx`, `ZSTD_CDict`
446* Macro names are all Capital letters.
447  The same composition rules (`PREFIX_NAME_QUALIFIER`) apply.
448* File names are all lowercase letters.
449  The convention is `snake_case`.
450  File names **must** be unique across the entire code base,
451  even when they stand in clearly separated directories.
452
453### Qualifiers
454* This code base is `const` friendly, if not `const` fanatical.
455  Any variable that can be `const` (aka. read-only) **must** be `const`.
456  Any pointer which content will not be modified must be `const`.
457  This property is then controlled at compiler level.
458  `const` variables are an important signal to readers that this variable isn’t modified.
459  Conversely, non-const variables are a signal to readers to watch out for modifications later on in the function.
460* If a function must be inlined, mention it explicitly,
461  using project's own portable macros, such as `FORCE_INLINE_ATTR`,
462  defined in `lib/common/compiler.h`.
463
464### Debugging
465* **Assertions** are welcome, and should be used very liberally,
466  to control any condition the code expects for its correct execution.
467  These assertion checks will be run in debug builds, and disabled in production.
468* For traces, this project provides its own debug macros,
469  in particular `DEBUGLOG(level, ...)`, defined in `lib/common/debug.h`.
470
471### Code documentation
472* Avoid code documentation that merely repeats what the code is already stating.
473  Whenever applicable, prefer employing the code as the primary way to convey explanations.
474  Example 1 : `int nbTokens = n;` instead of `int i = n; /* i is a nb of tokens *./`.
475  Example 2 : `assert(size > 0);` instead of `/* here, size should be positive */`.
476* At declaration level, the documentation explains how to use the function or variable
477  and when applicable why it's needed, of the scenarios where it can be useful.
478* At implementation level, the documentation explains the general outline of the algorithm employed,
479  and when applicable why this specific choice was preferred.
480
481### General layout
482* 4 spaces for indentation rather than tabs
483* Code documentation shall directly precede function declaration or implementation
484* Function implementations and its code documentation should be preceded and followed by an empty line
485
486
487## License
488By contributing to Zstandard, you agree that your contributions will be licensed
489under both the [LICENSE](LICENSE) file and the [COPYING](COPYING) file in the root directory of this source tree.
490