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 consise 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, make sure that all CI tests pass on your local fork. 72 See our section on setting up your CI environment for more information on how to do this. 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 contributers 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## Performance 144Performance is extremely important for zstd and we only merge pull requests whose performance 145landscape and corresponding trade-offs have been adequately analyzed, reproduced, and presented. 146This high bar for performance means that every PR which has the potential to 147impact performance takes a very long time for us to properly review. That being said, we 148always welcome contributions to improve performance (or worsen performance for the trade-off of 149something else). Please keep the following in mind before submitting a performance related PR: 150 1511. Zstd isn't as old as gzip but it has been around for time now and its evolution is 152very well documented via past Github issues and pull requests. It may be the case that your 153particular performance optimization has already been considered in the past. Please take some 154time to search through old issues and pull requests using keywords specific to your 155would-be PR. Of course, just because a topic has already been discussed (and perhaps rejected 156on some grounds) in the past, doesn't mean it isn't worth bringing up again. But even in that case, 157it will be helpful for you to have context from that topic's history before contributing. 1582. The distinction between noise and actual performance gains can unfortunately be very subtle 159especially when microbenchmarking extremely small wins or losses. The only remedy to getting 160something subtle merged is extensive benchmarking. You will be doing us a great favor if you 161take the time to run extensive, long-duration, and potentially cross-(os, platform, process, etc) 162benchmarks on your end before submitting a PR. Of course, you will not be able to benchmark 163your changes on every single processor and os out there (and neither will we) but do that best 164you can:) We've adding some things to think about when benchmarking below in the Benchmarking 165Performance section which might be helpful for you. 1663. Optimizing performance for a certain OS, processor vendor, compiler, or network system is a perfectly 167legitimate thing to do as long as it does not harm the overall performance health of Zstd. 168This is a hard balance to strike but please keep in mind other aspects of Zstd when 169submitting changes that are clang-specific, windows-specific, etc. 170 171## Benchmarking Performance 172Performance microbenchmarking is a tricky subject but also essential for Zstd. We value empirical 173testing over theoretical speculation. This guide it not perfect but for most scenarios, it 174is a good place to start. 175 176### Stability 177Unfortunately, the most important aspect in being able to benchmark reliably is to have a stable 178benchmarking machine. A virtual machine, a machine with shared resources, or your laptop 179will typically not be stable enough to obtain reliable benchmark results. If you can get your 180hands on a desktop, this is usually a better scenario. 181 182Of course, benchmarking can be done on non-hyper-stable machines as well. You will just have to 183do a little more work to ensure that you are in fact measuring the changes you've made not and 184noise. Here are some things you can do to make your benchmarks more stable: 185 1861. The most simple thing you can do to drastically improve the stability of your benchmark is 187to run it multiple times and then aggregate the results of those runs. As a general rule of 188thumb, the smaller the change you are trying to measure, the more samples of benchmark runs 189you will have to aggregate over to get reliable results. Here are some additional things to keep in 190mind when running multiple trials: 191 * How you aggregate your samples are important. You might be tempted to use the mean of your 192 results. While this is certainly going to be a more stable number than a raw single sample 193 benchmark number, you might have more luck by taking the median. The mean is not robust to 194 outliers whereas the median is. Better still, you could simply take the fastest speed your 195 benchmark achieved on each run since that is likely the fastest your process will be 196 capable of running your code. In our experience, this (aggregating by just taking the sample 197 with the fastest running time) has been the most stable approach. 198 * The more samples you have, the more stable your benchmarks should be. You can verify 199 your improved stability by looking at the size of your confidence intervals as you 200 increase your sample count. These should get smaller and smaller. Eventually hopefully 201 smaller than the performance win you are expecting. 202 * Most processors will take some time to get `hot` when running anything. The observations 203 you collect during that time period will very different from the true performance number. Having 204 a very large number of sample will help alleviate this problem slightly but you can also 205 address is directly by simply not including the first `n` iterations of your benchmark in 206 your aggregations. You can determine `n` by simply looking at the results from each iteration 207 and then hand picking a good threshold after which the variance in results seems to stabilize. 2082. You cannot really get reliable benchmarks if your host machine is simultaneously running 209another cpu/memory-intensive application in the background. If you are running benchmarks on your 210personal laptop for instance, you should close all applications (including your code editor and 211browser) before running your benchmarks. You might also have invisible background applications 212running. You can see what these are by looking at either Activity Monitor on Mac or Task Manager 213on Windows. You will get more stable benchmark results of you end those processes as well. 214 * If you have multiple cores, you can even run your benchmark on a reserved core to prevent 215 pollution from other OS and user processes. There are a number of ways to do this depending 216 on your OS: 217 * On linux boxes, you have use https://github.com/lpechacek/cpuset. 218 * On Windows, you can "Set Processor Affinity" using https://www.thewindowsclub.com/processor-affinity-windows 219 * 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 2203. To benchmark, you will likely end up writing a separate c/c++ program that will link libzstd. 221Dynamically linking your library will introduce some added variation (not a large amount but 222definitely some). Statically linking libzstd will be more stable. Static libraries should 223be enabled by default when building zstd. 2244. Use a profiler with a good high resolution timer. See the section below on profiling for 225details on this. 2265. Disable frequency scaling, turbo boost and address space randomization (this will vary by OS) 2276. Try to avoid storage. On some systems you can use tmpfs. Putting the program, inputs and outputs on 228tmpfs avoids touching a real storage system, which can have a pretty big variability. 229 230Also check our LLVM's guide on benchmarking here: https://llvm.org/docs/Benchmarking.html 231 232### Zstd benchmark 233The fastest signal you can get regarding your performance changes is via the in-build zstd cli 234bench option. You can run Zstd as you typically would for your scenario using some set of options 235and then additionally also specify the `-b#` option. Doing this will run our benchmarking pipeline 236for that options you have just provided. If you want to look at the internals of how this 237benchmarking script works, you can check out programs/benchzstd.c 238 239For example: say you have made a change that you believe improves the speed of zstd level 1. The 240very first thing you should use to asses whether you actually achieved any sort of improvement 241is `zstd -b`. You might try to do something like this. Note: you can use the `-i` option to 242specify a running time for your benchmark in seconds (default is 3 seconds). 243Usually, the longer the running time, the more stable your results will be. 244 245``` 246$ git checkout <commit-before-your-change> 247$ make && cp zstd zstd-old 248$ git checkout <commit-after-your-change> 249$ make && cp zstd zstd-new 250$ zstd-old -i5 -b1 <your-test-data> 251 1<your-test-data> : 8990 -> 3992 (2.252), 302.6 MB/s , 626.4 MB/s 252$ zstd-new -i5 -b1 <your-test-data> 253 1<your-test-data> : 8990 -> 3992 (2.252), 302.8 MB/s , 628.4 MB/s 254``` 255 256Unless your performance win is large enough to be visible despite the intrinsic noise 257on your computer, benchzstd alone will likely not be enough to validate the impact of your 258changes. For example, the results of the example above indicate that effectively nothing 259changed but there could be a small <3% improvement that the noise on the host machine 260obscured. So unless you see a large performance win (10-15% consistently) using just 261this method of evaluation will not be sufficient. 262 263### Profiling 264There are a number of great profilers out there. We're going to briefly mention how you can 265profile your code using `instruments` on mac, `perf` on linux and `visual studio profiler` 266on windows. 267 268Say you have an idea for a change that you think will provide some good performance gains 269for level 1 compression on Zstd. Typically this means, you have identified a section of 270code that you think can be made to run faster. 271 272The first thing you will want to do is make sure that the piece of code is actually taking up 273a notable amount of time to run. It is usually not worth optimzing something which accounts for less than 2740.0001% of the total running time. Luckily, there are tools to help with this. 275Profilers will let you see how much time your code spends inside a particular function. 276If your target code snippit is only part of a function, it might be worth trying to 277isolate that snippit by moving it to its own function (this is usually not necessary but 278might be). 279 280Most profilers (including the profilers dicusssed below) will generate a call graph of 281functions for you. Your goal will be to find your function of interest in this call grapch 282and then inspect the time spent inside of it. You might also want to to look at the 283annotated assembly which most profilers will provide you with. 284 285#### Instruments 286We will once again consider the scenario where you think you've identified a piece of code 287whose performance can be improved upon. Follow these steps to profile your code using 288Instruments. 289 2901. Open Instruments 2912. Select `Time Profiler` from the list of standard templates 2923. Close all other applications except for your instruments window and your terminal 2934. Run your benchmarking script from your terminal window 294 * You will want a benchmark that runs for at least a few seconds (5 seconds will 295 usually be long enough). This way the profiler will have something to work with 296 and you will have ample time to attach your profiler to this process:) 297 * I will just use benchzstd as my bencharmking script for this example: 298``` 299$ zstd -b1 -i5 <my-data> # this will run for 5 seconds 300``` 3015. Once you run your benchmarking script, switch back over to instruments and attach your 302process to the time profiler. You can do this by: 303 * Clicking on the `All Processes` drop down in the top left of the toolbar. 304 * Selecting your process from the dropdown. In my case, it is just going to be labled 305 `zstd` 306 * Hitting the bright red record circle button on the top left of the toolbar 3076. You profiler will now start collecting metrics from your bencharking script. Once 308you think you have collected enough samples (usually this is the case after 3 seconds of 309recording), stop your profiler. 3107. Make sure that in toolbar of the bottom window, `profile` is selected. 3118. You should be able to see your call graph. 312 * If you don't see the call graph or an incomplete call graph, make sure you have compiled 313 zstd and your benchmarking scripg using debug flags. On mac and linux, this just means 314 you will have to supply the `-g` flag alone with your build script. You might also 315 have to provide the `-fno-omit-frame-pointer` flag 3169. Dig down the graph to find your function call and then inspect it by double clicking 317the list item. You will be able to see the annotated source code and the assembly side by 318side. 319 320#### Perf 321 322This wiki has a pretty detailed tutorial on getting started working with perf so we'll 323leave you to check that out of you're getting started: 324 325https://perf.wiki.kernel.org/index.php/Tutorial 326 327Some general notes on perf: 328* Use `perf stat -r # <bench-program>` to quickly get some relevant timing and 329counter statistics. Perf uses a high resolution timer and this is likely one 330of the first things your team will run when assessing your PR. 331* Perf has a long list of hardware counters that can be viewed with `perf --list`. 332When measuring optimizations, something worth trying is to make sure the handware 333counters you expect to be impacted by your change are in fact being so. For example, 334if you expect the L1 cache misses to decrease with your change, you can look at the 335counter `L1-dcache-load-misses` 336* Perf hardware counters will not work on a virtual machine. 337 338#### Visual Studio 339 340TODO 341 342 343## Setting up continuous integration (CI) on your fork 344Zstd uses a number of different continuous integration (CI) tools to ensure that new changes 345are well tested before they make it to an official release. Specifically, we use the platforms 346travis-ci, circle-ci, and appveyor. 347 348Changes cannot be merged into the main dev branch unless they pass all of our CI tests. 349The easiest way to run these CI tests on your own before submitting a PR to our dev branch 350is to configure your personal fork of zstd with each of the CI platforms. Below, you'll find 351instructions for doing this. 352 353### travis-ci 354Follow these steps to link travis-ci with your github fork of zstd 355 3561. Make sure you are logged into your github account 3572. Go to https://travis-ci.org/ 3583. Click 'Sign in with Github' on the top right 3594. Click 'Authorize travis-ci' 3605. Click 'Activate all repositories using Github Apps' 3616. Select 'Only select repositories' and select your fork of zstd from the drop down 3627. Click 'Approve and Install' 3638. Click 'Sign in with Github' again. This time, it will be for travis-pro (which will let you view your tests on the web dashboard) 3649. Click 'Authorize travis-pro' 36510. You should have travis set up on your fork now. 366 367### circle-ci 368TODO 369 370### appveyor 371Follow these steps to link circle-ci with your girhub fork of zstd 372 3731. Make sure you are logged into your github account 3742. Go to https://www.appveyor.com/ 3753. Click 'Sign in' on the top right 3764. Select 'Github' on the left panel 3775. Click 'Authorize appveyor' 3786. You might be asked to select which repositories you want to give appveyor permission to. Select your fork of zstd if you're prompted 3797. You should have appveyor set up on your fork now. 380 381### General notes on CI 382CI tests run every time a pull request (PR) is created or updated. The exact tests 383that get run will depend on the destination branch you specify. Some tests take 384longer to run than others. Currently, our CI is set up to run a short 385series of tests when creating a PR to the dev branch and a longer series of tests 386when creating a PR to the release branch. You can look in the configuration files 387of the respective CI platform for more information on what gets run when. 388 389Most people will just want to create a PR with the destination set to their local dev 390branch of zstd. You can then find the status of the tests on the PR's page. You can also 391re-run tests and cancel running tests from the PR page or from the respective CI's dashboard. 392 393## Issues 394We use GitHub issues to track public bugs. Please ensure your description is 395clear and has sufficient instructions to be able to reproduce the issue. 396 397Facebook has a [bounty program](https://www.facebook.com/whitehat/) for the safe 398disclosure of security bugs. In those cases, please go through the process 399outlined on that page and do not file a public issue. 400 401## Coding Style 402* 4 spaces for indentation rather than tabs 403 404## License 405By contributing to Zstandard, you agree that your contributions will be licensed 406under both the [LICENSE](LICENSE) file and the [COPYING](COPYING) file in the root directory of this source tree. 407