3 This section describes the workflow and some best practices for developing
4 the Libre-SOC hardware. We use nmigen, yosys and symbiyosys, and this
5 page is intended not just to help you get set up, it is intended to
6 help advise you of some tricks and practices that will help you become
7 effective team contributors.
9 It is particularly important to bear in mind that we are not just
10 "developing code", here: we are creating a "lasting legacy educational
11 resource" for other people to learn from, and for businesses and students
12 alike to be able to use, learn from and augment for their own purposes.
14 It is also important to appreciate and respect that we are funded under
15 NLNet's Privacy and Enhanced Trust Programme <http://nlnet.nl/PET>. Full
16 transparency, readability, documentation, effective team communication
17 and formal mathematical proofs for all code at all levels is therefore
20 Therefore, we need not only to be "self-sufficient" (absolutely
21 under no circumstances critically reliant on somebody else's servers
22 **or protocols**) we also need to ensure that everything (including
23 all communication such as the mailing list archives) are recorded,
24 replicable, and accessible in perpetuity. Use of slack or a "forum"
25 either actively prevents or makes that much harder.
27 # Collaboration resources
29 The main message here: **use the right tool for the right job**.
31 * mailing list: general communication and discussion.
32 * irc channel #libre-soc: real(ish)-time communication.
33 * bugtracker: task-orientated, goal-orientated *focussed* discussion.
34 * ikiwiki: document store, information store, and (editable) main website
35 * git repositories: code stores (**not binary or auto-generated output store**)
36 * ftp server (<https://ftp.libre-soc.org/>): large (temporary,
37 auto-generated) file store.
39 We will add an IRC channel at some point when there are enough people
40 to warrant having one (and it will be publicly archived)
42 Note also the lack of a "forum" in the above list. this is very
43 deliberate. forums are a serious distraction when it comes to technical
44 heavily goal-orientated development. recent internet users may enjoy
45 looking up the "AOL metoo postings" meme.
47 Note also the complete lack of "social platforms". if we wanted to tell
48 everybody how much better each of us are than anyone else in the team,
49 how many times we made a commit (look at me, look at me, i'm so clever),
50 and how many times we went to the bathroom, we would have installed a
51 social media based project "management" system.
53 ## Main contact method: mailing list
55 To respect the transparency requirements, conversations need to be
56 public and archived (i.e not skype, not telegram, not discord,
57 and anyone seriously suggesting slack will be thrown to the
58 lions). Therefore we have a mailing list. Everything goes through
59 there. <https://lists.libre-soc.org/mailman/listinfo/libre-soc-dev>
60 therefore please do google "mailing list etiquette" and at the very
61 minimum look up and understand the following:
63 * This is a technical mailing list with complex topics. Top posting
64 is completely inappropriate. Don't do it unless you have mitigating
65 circumstances, and even then please apologise and explain ("hello sorry
66 using phone at airport flight soon, v. quick reply: ....")
67 * Always trim context but do not cut excessively to the point where people
68 cannot follow the discussion. Especially do not cut the attribution
69 ("On monday xxx wrote") of something that you are actually replying
71 * Use inline replies i.e. reply at the point in the relevant part of
72 the conversation, as if you were actually having a conversation.
73 * Follow standard IETF reply formatting, using ">" for cascaded
74 indentation of other people's replies. If using gmail, please: SWITCH
75 OFF RICH TEXT EDITING.
76 * Please for god's sake do not use "my replies are in a different
77 colour". Only old and highly regarded people still using AOL are allowed
78 to get away with that (such as Mitch).
79 * Start a new topic with a relevant subject line. If an existing
80 discussion changes direction, change the subject line to reflect the
81 new topic (or start a new conversation entirely, without using the
83 * DMARC is a pain on the neck. Try to avoid GPG signed messages. sigh.
84 * Don't send massive attachments. Put them online (no, not on facebook or
85 google drive or anywhere else that demands privacy violations) and provide
86 the link. Which should not require any kind of login to access. ask the
87 listadmin if you don't have anywhere suitable: FTP access can be arranged.
89 ### Actionable items from mailing list
91 If discussions result in any actionable items, it is important not to
92 lose track of them. Create a bugreport, find the discussion in the
93 archives <https://lists.libre-soc.org/pipermail/libre-soc-dev/>,
94 and put the link actually in the bugtracker as one of the comments.
96 At some point in any discussion, the sudden realisation may dawn on one
97 or more people that this is an "actionable" discussion. at that point
98 it may become better to use <https://bugs.libre-soc.org/>
99 itself to continue the discussion rather than to keep on dropping copies
100 of links into the bugtracker. The bugtracker sends copies of comments
101 *to* the list however this is 'one-way' (note from lkcl: because this
102 involves running an automated perl script from email, on every email,
103 on the server, that is a high security risk, and i'm not doing it. sorry.)
105 ### Mailing list != editable document store
107 Also, please do not use the mailing list as an "information or document
108 store or poor-man's editor". We have the wiki for that. Edit a page and
109 tell people what you did (summarise rather than drop the entire contents
110 at the list) and include the link to the page.
112 Or, if it is more appropriate, commit a document (or source code)
113 into the relevant git repository then look up the link in the gitweb
114 source tree browser and post that (in the bugtracker or mailing list)
115 See <https://git.libre-soc.org/>
117 ### gmail "spam"ifying the list
119 See <https://blog.kittycooper.com/2014/05/keeping-my-mailing-list-emails-out-of-gmails-spam-folder/>
121 Basically it is possible to select any message from the list, create a
122 "filter" (under "More"), and, on the 2nd dialog box, click the "never
123 send this to Spam" option.
127 bugzilla. old and highly effective. sign up in the usual way. any
128 problems, ask on the list.
130 Please do not ask for the project to be transferred to github or other
131 proprietary nonfree service "because it's soooo convenient", as the
132 lions are getting wind and gout from overfeeding on that one.
136 Runs the main libre-soc.org site (including this page). effective,
137 stunningly light on resources, and uses a git repository not a database.
138 That means it can be edited offline.
140 Usual deal: register an account and you can start editing and contributing
143 Hint: to create a new page, find a suitable page that would link to it,
144 first, then put the link in of the page you want to create, as if the
145 page already exists. Save that page, and you will find a question mark
146 next to the new link you created. click that link, and it will fire up a
147 "create new page" editor.
149 Wiki pages are formatted in [[markdown|ikiwiki/markdown]] syntax.
151 Hint again: the wiki is backed by a git repository. Don't go overboard
152 but at the same time do not be afraid that you might "damage" or "lose"
153 pages. Although it would be a minor pain, the pages can always be
154 reverted or edited by the sysadmins to restore things if you get in a tiz.
156 Assistance in creating a much better theme greatly appreciated. e.g.
157 <http://www.math.cmu.edu/~gautam/sj/blog/20140720-ikiwiki-navbar.html>
161 We use git. More on this below. We also use
162 [gitolite3](https://gitolite.com/gitolite/) running on a dedicated server.
163 again, it is extremely effective and low resource utilisation. Reminder:
164 lions are involved if github is mentioned.
166 [gitweb](https://git.wiki.kernel.org/index.php/Gitweb) is provided which
167 does a decent job. <https://git.libre-soc.org/>
169 [Git](https://en.wikipedia.org/wiki/Git) does version control, ie it
170 tracks changes to files so that previous versions can be got back or
175 <https://ftp.libre-soc.org/> is available for storing large files
176 that do not belong in a git repository, if we have (or ever need)
177 any. Images (etc.) if small and appropriate should go into the
178 wiki, however .tgz archives (etc.) and, at some point, binaries,
179 should be on the ftp server.
181 Ask on the list if you have a file that belongs on the ftp server.
185 As an aside: all this is "old school" and run on a single core 512MB
186 VM with only a 20GB HDD allocation. it costs only 8 GBP per month from
187 mythic-beasts and means that the project is in no way dependent on anyone
188 else - not microsoft, not google, not facebook, not amazon.
190 We tried [gitlab](https://about.gitlab.com/). it didn't go well. please
191 don't ask to replace the above extremely resource-efficient services
196 RAM is the biggest requirement. Minimum 16GB, the more the better (32
197 or 64GB starts to reach "acceptable" levels. Disk space is not hugely
198 critical: 256GB SSD should be more than adequate. Simulations and
199 FPGA compilations however are where raw processing power is a must.
200 High end Graphics Cards are nonessential.
202 What is particularly useful is to have hi-res screens (curved is
203 *strongly* recommended if the LCD is over 24in wide, to avoid eyeballs
204 going "prism" through long term use), and to have several of them: the
205 more the better. Either a DisplayLink UD160A (or more modern variant)
206 or simply using a second machine (lower spec hardware because it will
207 run editors) is really effective.
209 Also it is really recommended to have a UHD monitor (4k - 3840x2160),
210 or at least 2560x1200. If given a choice, 4:3 aspect ratio is better
211 than 16:9 particularly when using several of them. However, caveat
212 (details below): please when editing do not assume that everyone will
213 have access to such high resolution screens.
217 First install and become familiar with
218 [Debian](https://www.debian.org/) ([Ubuntu](https://ubuntu.com/)
220 must) for standardisation cross-team and so that toolchain installation
221 is greatly simplified. yosys in particular warns that trying to use
222 Windows, BSD or MacOS will get you into a world of pain.
224 Only a basic GUI desktop is necessary: fvwm2, xfce4, lxde are perfectly
225 sufficient (alongside wicd-gtk for network management). Other more
226 complex desktops can be used however may consume greater resources.
228 # editors and editing
230 Whilst this is often a personal choice, the fact that many editors are
231 GUI based and run full-screen with the entire right hand side *and* middle
232 *and* the majority of the left side of the hi-res screen entirely unused
233 and bereft of text leaves experienced developers both amused and puzzled.
235 At the point where such full-screen users commit code with line lengths
236 well over 160 characters, that amusement quickly evaporates.
238 Where the problems occur with full-screen editor usage is when a project
239 is split into dozens if not hundreds of small files (as this one is). At
240 that point it becomes pretty much essential to have as many as six to
241 eight files open *and on-screen* at once, without overlaps i.e. not in
242 hidden tabs, next to at least two if not three additional free and clear
243 terminals into which commands are regularly and routinely typed (make,
244 git commit, nosetests3 etc). Illustrated with the following 3840x2160
245 screenshot (click to view full image), where *every one* of those 80x70
246 xterm windows is *relevant to the task at hand*.
248 [[!img 2020-01-24_11-56.png size=640x ]]
250 (hint/tip: fvwm2 set up with "mouse-over to raise focus, rather than
251 additionally requiring a mouse click, can save a huge amount of cumulative
252 development time here, switching between editor terminal(s) and the
255 Once this becomes necessary, it it turn implies that having greater
256 than 80 chars per line - and running editors full-screen - is a severe
257 hinderance to an essential *and highly effective* workflow technique.
259 Additionally, care should be taken to respect that not everyone will have
260 200+ column editor windows and the eyesight of a hawk. They may only have
261 a 1280 x 800 laptop which barely fits two 80x53 xterms side by side.
262 Consequently, having excessively long functions is also a hindrance to
263 others, as such developers with limited screen resources would need to
264 continuously page-up and page-down to read the code even of a single
267 This helps explain in part, below, why compliance with
268 [pep8](https://pep8.org/) is enforced, including its 80 character limit.
269 In short: not everyone has the same "modern" GUI workflow or has access
270 to the same computing resources as you, so please do respect that.
272 More on this concept is
273 [here](https://www.linuxjournal.com/content/line-length-limits).
274 Note *very pointedly* that Linus Torvalds *specifically* states that
275 he does not want Linux kernel development to become the exclusive
276 domain of the "wealthy". That means **no** to assumptions about
277 access to ultra-high resolution screens.
279 # Software prerequisites
281 Whilst many resources online advocate "`sudo`" in front of all root-level
282 commands below, this quickly becomes tiresome. run "`sudo bash`", get a
283 root prompt, and save yourself some typing.
286 * apt-get install vim exuberant-ctags
287 * apt-get install build-essential
288 * apt-get install git python3.7 python3.7-dev python-nosetest3
289 * apt-get install graphviz xdot gtkwave
290 * apt-get install python3-venv
291 * apt-get install python-virtualenv # this is an alternative to python3-venv
292 * apt-get install tcl-dev libreadline-dev bison flex libffi-dev iverilog
293 * return to user prompt (ctrl-d)
295 (The above assumes that you are running Debian.)
297 This will get you python3 and other tools that are
298 needed. [graphviz](https://graphviz.org/) is essential
299 for showing the interconnections between cells, and
300 [gtkwave](http://gtkwave.sourceforge.net/) is essential for debugging.
302 If you would like to save yourself a lot more typing, check out the
303 [dev-env-setup](https://git.libre-soc.org/?p=dev-env-setup.git;a=summary)
304 repository, examine the scripts there and use them to automate much of
307 If you would like just to install only the apt dependencies use
308 [install-hdl-apt-reqs](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=install-hdl-apt-reqs;hb=HEAD) instead.
312 Look up good tutorials on how to use git effectively. There are so many
313 it is hard to recommend one. This is however essential. If you are not
314 comfortable with git, and you let things stay that way, it will seriously
315 impede development progress.
317 If working all day you should expect to be making at least two commits per
318 hour, so should become familiar with it very quickly. If you are *not*
319 doing around 2 commits per hour, something is wrong and you should read
320 the workflow instructions below more carefully, and also ask for advice
323 Worth noting: *this project does not use branches*. All code is committed
324 to master and we *require* that it be either zero-impact additions or that
325 relevant unit tests pass 100%. This ensures that people's work does not
326 get "lost" or isolated and out of touch due to major branch diversion,
327 and that people communicate and coordinate with each other.
331 Follow the source code (git clone) instructions here, do **not** use
332 the "stable" version (do not download the tarball):
333 <http://www.clifford.at/yosys/download.html>
335 Or, alternatively, use the
336 [yosys-et-al](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=yosys-et-al;hb=HEAD)
337 script (which also installs symbiyosys and its dependencies)
339 Do not try to use a fixed revision of yosys (currently 0.9), nmigen is
340 evolving and frequently interacts with yosys.
342 [Yosys](http://www.clifford.at/yosys/) is a framework for Verilog RTL.
343 [Verilog](https://en.wikipedia.org/wiki/Verilog) is a hardware description
345 RTL [Register Transfer
346 Level](https://en.wikipedia.org/wiki/Register-transfer_level)
347 models how data moves between
348 [registers](https://en.wikipedia.org/wiki/Hardware_register).
352 To install follow the [instructions
353 here](https://symbiyosys.readthedocs.io/en/latest/install.html)
354 Once done look at [A simple BMC
355 example](https://symbiyosys.readthedocs.io/en/latest/quickstart.html)
357 You do not have to install all of those (avy, boolector can be left
358 out if desired) however the more that are installed the more effective
359 the formal proof scripts will be (less resource utilisation in certain
362 [SymbiYosys](https://symbiyosys.readthedocs.io/en/latest/) (sby) is a
363 front-end driver program for Yosys-based formal hardware verification
368 [nmigen](https://m-labs.hk/gateware/nmigen/) may be installed as follows:
372 * git clone https://github.com/nmigen/nmigen.git
375 * python3 setup.py develop
378 Testing can then be carried out with "python3 setup.py test"
380 nmigen is a Python toolbox for building complex digital hardware.
382 ## Softfloat and sfpy
384 These are a test suite dependency for the
385 [ieee754fpu](https://www.gaisler.com/index.php/products/ipcores/ieee754fpu)
386 library, and will be changed in the future to use Jacob's
387 [simple-soft-float](https://crates.io/crates/simple-soft-float) library.
388 In the meantime, sfpy can be built as follows:
390 git clone --recursive https://github.com/billzorn/sfpy.git
393 git apply ../softposit_sfpy_build.patch
394 git apply /path/to/ieee754fpu/SoftPosit.patch
395 cd ../berkely-softfloat-3
396 # Note: Do not apply the patch included in sfpy for berkely-softfloat,
397 # it contains the same changes as this one
398 git apply /path/to/ieee754fpu/berkeley-softfloat.patch
401 # prepare a virtual environment for building
404 # or, if you prefer the old way:
405 # virtualenv -p python3 .env
407 # install dependencies
408 source .env/bin/activate
409 pip3 install --upgrade -r requirements.txt
414 make inplace -j$(nproc)
418 deactivate # deactivates venv, optional
419 pip3 install dist/sfpy*.whl
421 You can test your installation by doing the following:
424 >>> from sfpy import *
427 It should print out `Posit8(1.3125)`
429 ## qemu, cross-compilers, gdb
431 As we are doing POWER ISA, POWER ISA compilers, toolchains and
432 emulators are required.
434 Install powerpc64 gcc:
436 apt-get install gcc-9-powerpc64-linux-gnu
440 apt-get install qemu-system-ppc
442 Install gdb from source. Obtain the required tarball matching
443 the version of gcc (9.1) from here <https://ftp.gnu.org/gnu/gdb/>,
446 cd gdb-9.1 (or other location)
449 ../configure --srcdir=.. --host=x86_64-linux --target=powerpc64-linux-gnu
453 [gdb](https://en.wikipedia.org/wiki/GNU_Debugger) lets you debug running
454 programs. [qemu](https://www.qemu.org/) emulates processors, you can
455 run programs under qemu.
457 ## power_instruction_analyzer (pia)
459 We have a custom tool built in rust by programmerjake to help analyze
460 the power instructions execution on *actual* hardware.
462 Note: a very recent version of pip3 is required for this to work.
466 curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
468 Make sure we have the correct and up-to-date rust compiler (rustc):
470 rustup default stable
473 Use rust's package manager *cargo* to install the rust-python build
476 cargo install maturin
478 Install from git source by doing the following:
480 git clone https://salsa.debian.org/Kazan-team/power-instruction-analyzer.git pia
482 maturin build --cargo-extra-args=--features=python-extension
483 python3 -m pip install --user target/wheels/*.whl
485 Note: an ongoing bug in maturin interferes with successful installation.
486 This can be worked around by explicitly installing only the `.whl`
487 files needed rather than installing everything (`\*.whl`).
491 As this is an actual ASIC, we do not rely on an FPGA's JTAG TAP
492 interface, instead require a full complete independent implementation
493 of JTAG. Staf Verhaegen has one, with a full test suite, and it is
494 superb and well-written. The Libre-SOC version includes DMI (Debug
497 git clone https://git.libre-soc.org/git/c4m-jtag.git/
499 python3 setup.py develop
501 Included is an IDCODE tap point, Wishbone Master (for direct memory read
502 and write, fully independent of the core), IOPad redirection and testing,
503 and general purpose shift register capability for any custom use.
505 We added a DMI to JTAG bridge in LibreSOC which is
506 directly connected to the core, to access registers and
507 to be able to start and stop the core and change the PC.
508 In combination with the JTAG Wishbone interface the test
509 [ASIC](https://en.wikipedia.org/wiki/Application-specific_integrated_circuit)
510 can have a bootloader uploaded directly into onboard
511 [SRAM](https://en.wikipedia.org/wiki/Static_random-access_memory) and
514 [Chips4Makers](https://chips4makers.io/) make it possible for makers
515 and hobbyists to make their own open source chips.
517 [JTAG](https://en.wikipedia.org/wiki/JTAG) (Joint Test Action Group) is
518 an industry standard for verifying designs and testing printed circuit
519 boards after manufacture.
522 bus](https://en.wikipedia.org/wiki/Wishbone_%28computer_bus%29) is an open
523 source hardware computer bus intended to let the parts of an integrated
524 circuit communicate with each other.
528 See [[HDL_workflow/coriolis2]] page, for those people doing layout work.
532 A portable FPGA place and route tool.
534 See [[HDL_workflow/nextpnr]] page for installation instructions of nextpnr with ECP5 support for Lattice FPGA ECP5 series.
536 # Registering for git repository access
538 After going through the onboarding process and having agreed to take
539 responsibility for certain tasks, ask on the mailing list for git
540 repository access, sending in a public key (`id_rsa.pub`). If you do
541 not have one then generate it with `ssh-keygen -t rsa`. You will find it
544 NEVER SEND ANYONE THE PRIVATE KEY. By contrast the public key, on
545 account of being public, is perfectly fine to make... err... public.
547 Create a file `~/.ssh/config` with the following lines:
549 Host git.libre-soc.org
552 Wait for the Project Admin to confirm that the ssh key has been added
553 to the required repositories. Once confirmed, you can clone any of the
554 repos at https://git.libre-soc.org/:
556 git clone gitolite3@git.libre-soc.org:REPONAME.git
558 Alternatively, the .ssh/config can be skipped and this used:
560 git clone ssh://gitolite3@git.libre-soc.org:922/REPONAME.git
564 Although there are methods online which describe how (and why) these
565 settings are normally done, honestly it is simpler and easier to open
566 ~/.gitconfig and add them by hand.
568 core.autocrlf is a good idea to ensure that anyone adding DOS-formatted
569 files they don't become a pain. pull.rebase is something that is greatly
570 preferred for this project because it avoids the mess of "multiple
571 extra merge git tree entries", and branch.autosetuprebase=always will,
572 if you want it, always ensure that a new git checkout is set up with rebase.
581 autosetuprebase = always
583 # Checking out the HDL repositories
585 Before running the following, install the
586 dependencies. This is easiest done with this script
587 <https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=install-hdl-apt-reqs;hb=HEAD>
591 * git clone gitolite3@git.libre-soc.org:nmutil.git
592 * git clone gitolite3@git.libre-soc.org:ieee754fpu.git
593 * git clone gitolite3@git.libre-soc.org:nmigen-soc.git
594 * git clone gitolite3@git.libre-soc.org:soc.git
596 In each of these directories, in the order listed, track down the
597 `setup.py` file, then, as root (`sudo bash`), run the following:
599 * python3 setup.py develop
601 The reason for using "develop" mode is that the code may be edited
602 in-place yet still imported "globally". There are variants on this theme
603 for multi-user machine use however it is often just easier to get your
604 own machine these days.
606 The reason for the order is because soc depends on ieee754fpu, and
607 ieee754fpu depends on nmutil
609 If "`python3 setup.py install`" is used it is a pain: edit, then
610 install. edit, then install. It gets extremely tedious, hence why
611 "develop" was created.
617 * new members, add yourself to the [[about_us]] page and create yourself a home page using someone else's page as a template.
618 * communicate on the mailing list or the bugtracker an intent to take
619 responsibility for a particular task.
620 * assign yourself as the bug's owner
621 * *keep in touch* about what you are doing, and why you are doing it.
622 * edit your home page regularly, particularly to track tasks so that
623 they can be paid by NLNet.
624 * if you cannot do something that you have taken responsibility for,
625 then unless it is a dire personal emergency please say so, on-list. we
626 won't mind. we'll help sort it out.
628 Regarding the above it is important that you read, understand, and agree
629 to the [[charter]] because the charter is about ensuring that we operate
630 as an effective organisation. It's *not* about "setting rules and meting
635 for actual code development
639 * plan in advance to write not just code but a full test suite for
640 that code. **this is not optional**. large python projects that do not
641 have unit tests **FAIL** (see separate section below).
642 * Prioritise writing formal proofs and a single clear unit test that is more
643 like a "worked example".
644 We receive NLNet funds for writing formal proofs, plus they
645 cover corner cases and take far less time to write
647 ### Commit tested or zero-dependent code
649 * only commit code that has been tested (or is presently unused). other
650 people will be depending on you, so do take care not to screw up.
651 not least because, as it says in the [[charter]] it will be your
652 responsibility to fix. that said, do not feel intimidated: ask for help
653 and advice, and you'll get it straight away.
657 * commit often. several times a day, and "git push" it. this is
658 collaboration. if something is left even overnight uncommitted and not
659 pushed so that other people can see it, it is a red flag. if you find
660 yourself thinking "i'll commit it when it's finished" or "i don't want to
661 commit something that people might criticise" *this is not collaboration*,
662 it is making yourself a bottleneck. pair-programming is supposed to help
663 avoid this kind of thing however pair-programming is difficult to organise
664 for remote collaborative libre projects (suggestions welcomed here)
666 ### Enable editor auto-detection of file changes by external programs
668 This is important. "`git pull`" will merge in changes. If you then
669 arbitrarily save a file without re-loading it, you risk destroying
672 ### Absolutely no auto-generated output
674 * **do not commit autogenerated output**. write a shell script and commit
675 that, or add a `Makefile` to run the command that generates the output, but
676 **do not** add the actual output of **any** command to the repository.
677 ever. this is really important. even if it is a human-readable file
678 rather than a binary object file.
679 it is very common to add pdfs (the result of running `latex2pdf`) or
680 configure.in (the result of running `automake`), they are an absolute
681 nuisance and interfere hugely with git diffs, as well as waste hard
682 disk space *and* network bandwidth. don't do it.
684 ### Write commands that do tasks and commit those
686 * if the command needed to create any given autogenerated output is not
687 currently in the list of known project dependencies, first consult on
688 the list if it is okay to make that command become a hard dependency of
689 the project (hint: java, node.js php and .NET commands may cause delays
690 in response time due to other list participants laughing hysterically),
691 and after a decision is made, document the dependency and how its source
692 code is obtained and built (hence why it has to be discussed carefully)
693 * if you find yourself repeating commands regularly, chances are high
694 that someone else will need to run them, too. clearly this includes
695 yourself, therefore, to make everyone's lives easier including your own,
696 put them into a `.sh` shell script (and/or a `Makefile`), commit them to
697 the repository and document them at the very minimum in the README,
698 INSTALL.txt or somewhere in a docs folder as appropriate. if unsure,
699 ask on the mailing list for advice.
701 ### Keep commits single-purpose
703 * edit files making minimal *single purpose* modifications (even if
704 it involves multiple files. Good extreme example: globally changing
705 a function name across an entire codebase is one purpose, one commit,
706 yet hundreds of files. miss out one of those files, requiring multiple
707 commits, and it actually becomes a nuisance).
709 ### Run unit tests prior to commits
711 * prior to committing make sure that relevant unit tests pass, or that
712 the change is a zero-impact addition (no unit tests fail at the minimum)
714 ### Do not break existing code
716 * keep working code working **at all times**. find ways to ensure that
717 this is the case. examples include writing alternative classes that
718 replace existing functionality and adding runtime options to select
719 between old and new code.
721 ### Small commits with relevant commit message
723 * commit no more than around 5 to 10 lines at a time, with a CLEAR message
724 (no "added this" or "changed that").
725 * if as you write you find that the commit message involves a *list* of
726 changes or the word "and", then STOP. do not proceed: it is a "red flag"
727 that the commit has not been properly broken down into separate-purpose
728 commits. ask for advice on-list on how to proceed.
730 ### Exceptions to small commit: atomic single purpose commit
732 * if it is essential to commit large amounts of code, ensure that it
733 is **not** in use **anywhere** by any other code. then make a *small*
734 (single purpose) followup commit which actually puts that code into use.
736 This last rule is kinda flexible, because if you add the code *and* add
737 the unit test *and* added it into the main code *and* ran all relevant
738 unit tests on all cascade-impacted areas by that change, that's perfectly
739 fine too. however if it is the end of a day, and you need to stop and
740 do not have time to run the necessary unit tests, do *not* commit the
741 change which integrates untested code: just commit the new code (only)
742 and follow up the next day *after* running the full relevant unit tests.
744 ### Why such strict rules?
746 The reason for all the above is because python is a dynamically typed
747 language. make one tiny change at the base level of the class hierarchy
748 and the effect may be disastrous.
750 It is therefore worth reiterating: make absolutely certain that you *only*
751 commit working code or zero-impact code.
753 Therefore, if you are absolutely certain that a new addition (new file,
754 new class, new function) is not going to have any side-effects, committing
755 it (a large amount of code) is perfectly fine.
757 As a general rule, however, do not use this an an excuse to write code
758 first then write unit tests as an afterthought. write *less* code *in
759 conjunction* with its (more basic) unit tests, instead. then, folliw up with
760 additions and improvements.
762 The reason for separating out commits to single purpose only becomes
763 obvious (and regretted if not followed) when, months later, a mistake
764 has to be tracked down and reverted. if the commit does not have an
765 easy-to-find message, it cannot even be located, and once found, if the
766 commit confuses several unrelated changes, not only the diff is larger
767 than it should be, the reversion process becomes extremely painful.
771 * all code needs to conform to pep8. use either pep8checker or better
772 run autopep8. however whenever committing whitespace changes, *make a
773 separate commit* with a commit message "whitespace" or "autopep8 cleanup".
774 * pep8 REQUIRES no more than 80 chars per line. this is non-negotiable. if
775 you think you need greater than 80 chars, it *fundamentally* indicates
776 poor code design. split the code down further into smaller classes
779 ### Docstring checker
781 * TBD there is a docstring checker. at the minimum make sure to have
782 an SPD license header, module header docstring, class docstring and
783 function docstrings on at least non-obvious functions.
785 ### Clear code commenting and docstrings
787 * make liberal but not excessive use of comments. describe a group of
788 lines of code, with terse but useful comments describing the purpose,
789 documenting any side-effects, and anything that could trip you or other
790 developers up. unusual coding techniques should *definitely* contain
793 ### Only one class per module (ish)
795 * unless they are very closely related, only have one module (one class)
796 per file. a file only 25 lines long including imports and docstrings
797 is perfectly fine however don't force yourself. again, if unsure,
800 ### File and Directory hierarchy
802 * *keep files short and simple*. see below as to why
803 * create a decent directory hierarchy but do not go mad. ask for advice
808 * please do not use "from module import \*". it is extremely bad practice,
809 causes unnecessary resource utilisation, makes code readability and
810 tracking extremely difficult, and results in unintended side-effects.
812 Example: often you want to find the code from which a class was imported.
813 nirmally you go to the top of the file, check the imports, and you know
814 exactly which file has the class because of the import path. by using
815 wildcards, you have absolutely *no clue* which wildcard imported which
818 Example: sometimes you may accidentally have duplicate code maintained
819 in two or more places. editing one of them you find, puzzlingly, that
820 the code behaves in some files with the old behaviour, but in others it
821 works. after a massive amount of investigation, you find that the working
822 files happen to have a wildcard import of the newer accidental duplicate
823 class **after** the wildcard import of the older class with exactly the
824 same name. if you had used explicit imports, you would have spotted
825 the double import of the class from two separate locations, immediately.
827 Really. don't. use. wildcards.
829 ### Keep file and variables short but clear
831 * try to keep both filenames and variable names short but not ridiculously
832 obtuse. an interesting compromise on imports is "from ridiculousfilename
833 import longsillyname as lsn", and to assign variables as well: "comb =
834 m.d.comb" followed by multiple "comb += nmigen_stmt" lines is a good trick
835 that can reduce code indentation by 6 characters without reducing clarity.
837 Additionally, use comments just above an obtuse variable in order to
838 help explain what it is for. In combination with keeping the the module
839 itself short, other readers will not need to scroll back several pages
840 in order to understand the code.
842 Yes it is tempting to actually use the variables as
843 self-explanatory-comments and generally this can be extremely good
844 practice. the problem comes when the variable is so long that a function
845 with several parameters csn no longer fit on a single line, and takes
846 up five to ten lines rather than one or two. at that point, the length
847 of the code is adversely affected and thus so is readability by forcing
848 readers to scroll through reams of pages.
850 It is a tricky balance: basically use your common sense, or just ask
851 someone else, "can you understand this code?"
853 ### Reasons for code structure
855 Regarding code structure: we decided to go with small modules that are
856 both easy to analyse, as well as fit onto a single page and be readable
857 when displayed as a visual graph on a full UHD monitor. this is done
860 * using the capability of nmigen (TODO crossref to example) output the
861 module to a yosys ilang (.il) file
862 * in a separate terminal window, run yosys
863 * at the yosys prompt type "read_ilang modulename.il"
864 * type "show top" and a graphviz window should appear. note that typing
865 show, then space, then pressing the tab key twice will give a full list
866 of submodules (one of which will be "top")
868 You can now fullsize the graphviz window and scroll around. if it looks
869 reasonably obvious at 100% zoom, i.e the connections can be clearly
870 related in your mind back to the actual code (by matching the graph names
871 against signals and modules in the original nmigen code) and the words are
872 not tiny when zoomed out, and connections are not total incomprehensible
873 spaghetti, then congratulations, you have well-designed code. If not,
874 then this indicates a need to split the code further into submodules
875 and do a bit more work.
877 The reasons for doing a proper modularisation job are several-fold:
879 * firstly, we will not be doing a full automated layout-and-hope
880 using alliance/coriolis2, we will be doing leaf-node thru tree node
881 half-automated half-manual layout, finally getting to the floorplan,
882 then revising and iteratively adjusting.
883 * secondly, examining modules at the gate level (or close to it) is just
884 good practice. poor design creeps in by *not* knowing what the tools
885 are actually doing (word to experienced developers: yes, we know that
886 the yosys graph != final netlist).
887 * thirdly, unit testing, particularly formal proofs, is far easier on
888 small sections of code, and complete in a reasonable time.
890 ## Special warning / alert to vim users!
892 Some time around the beginning of 2019 some bright spark decided that
893 an "auto-recommend-completion-of-stuff" option would be a nice, shiny
894 idea to enable by default from that point onwards.
896 This incredibly annoying "feature" results in tabs (or spaces) being
897 inserted "on your behalf" when you press return on one line, for your
898 "convenience" of not needing to type lots of spaces/tabs just to get
899 to the same indentation level.
901 Of course, this "feature", if you press return on one line in edit
902 mode and then press "escape", leaves a bundle-of-joy extraneous
903 whitespace **exactly** where you don't want it, and didn't ask for it,
904 pooped all over your file.
906 Therefore, *please*: **before** running "git commit", get into the
907 habit of always running "git diff", and at the very minimum
908 speed-skim the entire diff, looking for tell-tale "red squares"
909 (these show up under bash diff colour-syntax-highlighting) that
910 inform you that, without your knowledge or consent, vim has
911 "helpfully" inserted extraneous whitespace.
913 Remove them **before** git committing because they are not part
914 of the actual desired code-modifications, and committing them
915 is a major and constant distraction for reviewers about actual
916 important things like "the code that actually *usefully* was
917 modified for that commit"
919 This has the useful side-effect of ensuring that, right before
920 the commit, you've got the actual diff right in front of you
921 in the xterm window, on which you can base the "commit message".
925 For further reading, see the wikipedia page on
926 [Test-driven Development](https://en.wikipedia.org/wiki/Test-driven_development)
928 This deserves its own special section. It is extremely important to
929 appreciate that without unit tests, python projects are simply unviable.
930 Python itself has over 25,000 individual tests.
932 This can be quite overwhelming to a beginner developer, especially one
933 used to writing scripts of only 100 lines in length.
935 Thanks to Samuel Falvo we learned that writing unit tests as a formal
936 proof is not only shorter, it's also far more readable and also, if
937 written properly, provides 100% coverage of corner-cases that would
938 otherwise be overlooked or require tens to hundreds of thousands of
941 No this is not a joke or even remotely hypothetical, this is an actual
944 The ieee754fpu requires several hundreds of thousands of tests to be
945 run (currently needing several days to run them all), and even then we
946 cannot be absolutely certain that all possible combinations of input have
947 been tested. With 2^128 permutations to try with 2 64 bit FP numbers
948 it is simply impossible to even try.
950 This is where formal proofs come into play.
952 Samuel illustrated to us that "ordinary" unit tests can then be written
953 to *augment* the formal ones, serving the purpose of illustrating how
954 to use the module, more than anything.
956 However it is appreciated that writing formal proofs is a bit of a
957 black art. This is where team collaboration particularly kicks in,
958 so if you need help, ask on the mailing list.
960 ## Don't comment out unit tests: add them first (as failures) and fix code later
962 Unit tests serve an additional critical purpose of keeping track of code
963 that needs to be written. In many cases, you write the unit test *first*,
964 despite knowing full well that the code doesn't even exist or is completely
965 broken. The unit test then serves as a constant and important reminder
966 to actually fix (or write) the code.
968 Therefore, *do not* comment out unit tests just because they "don't work".
969 If you absolutely must stop a unit test from running, **do not delete it**.
970 Simply mark it with an appropriate
971 ["skip" decorator](https://docs.python.org/3/library/unittest.html#skipping-tests-and-expected-failures),
972 preferably with a link to a URL in the [bugtracker](https://bugs.libre-soc.org/)
973 with further details as to why the unit test should not be run.
977 Find appropriate tutorials for nmigen and yosys, as well as symbiyosys.
979 * Robert Baruch's nmigen tutorials look really good:
980 <https://github.com/RobertBaruch/nmigen-tutorial>
981 * Although a verilog example this is very useful to do
982 <https://symbiyosys.readthedocs.io/en/latest/quickstart.html#first-step-a-simple-bmc-example>
983 * This tutorial looks pretty good and will get you started
984 <http://blog.lambdaconcept.com/doku.php?id=nmigen:nmigen_install> and
985 walks not just through simulation, it takes you through using gtkwave
987 * There exist several nmigen examples which are also executable
988 <https://github.com/m-labs/nmigen/tree/master/examples/> exactly as
989 described in the above tutorial (python3 filename.py -h)