7 This section describes the workflow and some best practices for developing
8 the Libre-SOC hardware. We use nmigen, yosys and symbiyosys, and this
9 page is intended not just to help you get set up, it is intended to
10 help advise you of some tricks and practices that will help you become
11 effective team contributors.
13 It is particularly important to bear in mind that we are not just
14 "developing code", here: we are creating a "lasting legacy educational
15 resource" for other people to learn from, and for businesses and students
16 alike to be able to use, learn from and augment for their own purposes.
18 It is also important to appreciate and respect that we are funded under
19 NLNet's Privacy and Enhanced Trust Programme <http://nlnet.nl/PET>. Full
20 transparency, readability, documentation, effective team communication
21 and formal mathematical proofs for all code at all levels is therefore
24 Therefore, we need not only to be "self-sufficient" (absolutely
25 under no circumstances critically reliant on somebody else's servers
26 **or protocols**) we also need to ensure that everything (including
27 all communication such as the mailing list archives) are recorded,
28 replicable, and accessible in perpetuity. Use of slack or a "forum"
29 either actively prevents or makes that much harder.
31 # Collaboration resources
33 The main message here: **use the right tool for the right job**.
35 * mailing list: general communication and discussion.
36 * irc channel #libre-soc on irc.libera.chat: real(ish)-time communication.
37 * bugtracker: task-orientated, goal-orientated *focussed* discussion.
38 * ikiwiki: document store, information store, and (editable) main website
39 * git repositories: code stores (**not binary or auto-generated output store**)
40 * ftp server (<https://ftp.libre-soc.org/>): large (temporary,
41 auto-generated) file store.
43 Note also the lack of a "forum" in the above list. this is very
44 deliberate. forums are a serious distraction when it comes to technical
45 heavily goal-orientated development. recent internet users may enjoy
46 looking up the "AOL metoo postings" meme.
48 Note also the complete lack of "social platforms". if we wanted to tell
49 everybody how much better each of us are than anyone else in the team,
50 how many times we made a commit (look at me, look at me, i'm so clever),
51 and how many times we went to the bathroom, we would have installed a
52 social media based project "management" system.
54 ## Main contact method: mailing list
56 To respect the transparency requirements, conversations need to be
57 public and archived (i.e not skype, not telegram, not discord,
58 and anyone seriously suggesting slack will be thrown to the
59 lions). Therefore we have a mailing list. Everything goes through
60 there. <https://lists.libre-soc.org/mailman/listinfo/libre-soc-dev>
61 therefore please do google "mailing list etiquette" and at the very
62 minimum look up and understand the following:
64 * This is a technical mailing list with complex topics. Top posting
65 is completely inappropriate. Don't do it unless you have mitigating
66 circumstances, and even then please apologise and explain ("hello sorry
67 using phone at airport flight soon, v. quick reply: ....")
68 * Always trim context but do not cut excessively to the point where people
69 cannot follow the discussion. Especially do not cut the attribution
70 ("On monday xxx wrote") of something that you are actually replying
72 * Use inline replies i.e. reply at the point in the relevant part of
73 the conversation, as if you were actually having a conversation.
74 * Follow standard IETF reply formatting, using ">" for cascaded
75 indentation of other people's replies. If using gmail, please: SWITCH
76 OFF RICH TEXT EDITING.
77 * Please for god's sake do not use "my replies are in a different
78 colour". Only old and highly regarded people still using AOL are allowed
79 to get away with that (such as Mitch).
80 * Start a new topic with a relevant subject line. If an existing
81 discussion changes direction, change the subject line to reflect the
82 new topic (or start a new conversation entirely, without using the
84 * DMARC is a pain on the neck. Try to avoid GPG signed messages. sigh.
85 * Don't send massive attachments. Put them online (no, not on facebook or
86 google drive or anywhere else that demands privacy violations) and provide
87 the link. Which should not require any kind of login to access. ask the
88 listadmin if you don't have anywhere suitable: FTP access can be arranged.
90 ### Actionable items from mailing list
92 If discussions result in any actionable items, it is important not to
93 lose track of them. Create a bugreport, find the discussion in the
94 archives <https://lists.libre-soc.org/pipermail/libre-soc-dev/>,
95 and put the link actually in the bugtracker as one of the comments.
97 At some point in any discussion, the sudden realisation may dawn on one
98 or more people that this is an "actionable" discussion. at that point
99 it may become better to use <https://bugs.libre-soc.org/>
100 itself to continue the discussion rather than to keep on dropping copies
101 of links into the bugtracker. The bugtracker sends copies of comments
102 *to* the list however this is 'one-way' (note from lkcl: because this
103 involves running an automated perl script from email, on every email,
104 on the server, that is a high security risk, and i'm not doing it. sorry.)
106 ### Mailing list != editable document store
108 Also, please do not use the mailing list as an "information or document
109 store or poor-man's editor" **including not sending large images**.
110 We have the wiki for that. Edit a page and
111 tell people what you did (summarise rather than drop the entire contents
112 at the list) and include the link to the page.
114 Or, if it is more appropriate, commit a document (or source code)
115 into the relevant git repository then look up the link in the gitweb
116 source tree browser and post that (in the bugtracker or mailing list)
117 See <https://git.libre-soc.org/>
119 ### gmail "spam"ifying the list
121 See <https://blog.kittycooper.com/2014/05/keeping-my-mailing-list-emails-out-of-gmails-spam-folder/>
123 Basically it is possible to select any message from the list, create a
124 "filter" (under "More"), and, on the 2nd dialog box, click the "never
125 send this to Spam" option.
129 bugzilla. old and highly effective. sign up in the usual way. any
130 problems, ask on the list.
132 Please do not ask for the project to be transferred to github or other
133 proprietary nonfree service "because it's soooo convenient", as the
134 lions are getting wind and gout from overfeeding on that one.
138 Runs the main libre-soc.org site (including this page). effective,
139 stunningly light on resources, and uses a git repository not a database.
140 That means it can be edited offline.
142 Usual deal: register an account and you can start editing and contributing
145 Hint: to create a new page, find a suitable page that would link to it,
146 first, then put the link in of the page you want to create, as if the
147 page already exists. Save that page, and you will find a question mark
148 next to the new link you created. click that link, and it will fire up a
149 "create new page" editor.
151 Wiki pages are formatted in [[markdown|ikiwiki/markdown]] syntax.
153 Hint again: the wiki is backed by a git repository. Don't go overboard
154 but at the same time do not be afraid that you might "damage" or "lose"
155 pages. Although it would be a minor pain, the pages can always be
156 reverted or edited by the sysadmins to restore things if you get in a tiz.
158 Assistance in creating a much better theme greatly appreciated. e.g.
159 <http://www.math.cmu.edu/~gautam/sj/blog/20140720-ikiwiki-navbar.html>
163 We use git. More on this below. We also use
164 [gitolite3](https://gitolite.com/gitolite/) running on a dedicated server.
165 again, it is extremely effective and low resource utilisation. Reminder:
166 lions are involved if github is mentioned.
168 [gitweb](https://git.wiki.kernel.org/index.php/Gitweb) is provided which
169 does a decent job. <https://git.libre-soc.org/>
171 [Git](https://en.wikipedia.org/wiki/Git) does version control, ie it
172 tracks changes to files so that previous versions can be got back or
175 Checklist page [[HDL_workflow/git_checklist]]
179 <https://ftp.libre-soc.org/> is available for storing large files
180 that do not belong in a git repository, if we have (or ever need)
181 any. Images (etc.) if small and appropriate should go into the
182 wiki, however .tgz archives (etc.) and, at some point, binaries,
183 should be on the ftp server.
185 Ask on the list if you have a file that belongs on the ftp server.
189 As an aside: all this is "old school" and run on a single core 512MB
190 VM with only a 20GB HDD allocation. it costs only 8 GBP per month from
191 mythic-beasts and means that the project is in no way dependent on anyone
192 else - not microsoft, not google, not facebook, not amazon.
194 We tried [gitlab](https://about.gitlab.com/). it didn't go well. please
195 don't ask to replace the above extremely resource-efficient services
200 RAM is the biggest requirement. Minimum 16GB, the more the better (32
201 or 64GB starts to reach "acceptable" levels. Disk space is not hugely
202 critical: 256GB SSD should be more than adequate. Simulations and
203 FPGA compilations however are where raw processing power is a must.
204 High end Graphics Cards are nonessential.
206 What is particularly useful is to have hi-res screens (curved is
207 *strongly* recommended if the LCD is over 24in wide, to avoid eyeballs
208 going "prism" through long term use), and to have several of them: the
209 more the better. Either a DisplayLink UD160A (or more modern variant)
210 or simply using a second machine (lower spec hardware because it will
211 run editors) is really effective.
213 Also it is really recommended to have a UHD monitor (4k - 3840x2160),
214 or at least 2560x1200. If given a choice, 4:3 aspect ratio is better
215 than 16:9 particularly when using several of them. However, caveat
216 (details below): please when editing do not assume that everyone will
217 have access to such high resolution screens.
221 First install and become familiar with
222 [Debian](https://www.debian.org/) ([Ubuntu](https://ubuntu.com/)
224 must) for standardisation cross-team and so that toolchain installation
225 is greatly simplified. yosys in particular warns that trying to use
226 Windows, BSD or MacOS will get you into a world of pain.
228 Only a basic GUI desktop is necessary: fvwm2, xfce4, lxde are perfectly
229 sufficient (alongside wicd-gtk for network management). Other more
230 complex desktops can be used however may consume greater resources.
232 # editors and editing
234 Whilst this is often a personal choice, the fact that many editors are
235 GUI based and run full-screen with the entire right hand side *and* middle
236 *and* the majority of the left side of the hi-res screen entirely unused
237 and bereft of text leaves experienced developers both amused and puzzled.
239 At the point where such full-screen users commit code with line lengths
240 well over 160 characters, that amusement quickly evaporates.
242 Where the problems occur with full-screen editor usage is when a project
243 is split into dozens if not hundreds of small files (as this one is). At
244 that point it becomes pretty much essential to have as many as six to
245 eight files open *and on-screen* at once, without overlaps i.e. not in
246 hidden tabs, next to at least two if not three additional free and clear
247 terminals into which commands are regularly and routinely typed (make,
248 git commit, nosetests3 etc). Illustrated with the following 3840x2160
249 screenshot (click to view full image), where *every one* of those 80x70
250 xterm windows is *relevant to the task at hand*.
252 [[!img 2020-01-24_11-56.png size=640x ]]
254 (hint/tip: fvwm2 set up with "mouse-over to raise focus, rather than
255 additionally requiring a mouse click, can save a huge amount of cumulative
256 development time here, switching between editor terminal(s) and the
259 Once this becomes necessary, it it turn implies that having greater
260 than 80 chars per line - and running editors full-screen - is a severe
261 hinderance to an essential *and highly effective* workflow technique.
263 Additionally, care should be taken to respect that not everyone will have
264 200+ column editor windows and the eyesight of a hawk. They may only have
265 a 1280 x 800 laptop which barely fits two 80x53 xterms side by side.
266 Consequently, having excessively long functions is also a hindrance to
267 others, as such developers with limited screen resources would need to
268 continuously page-up and page-down to read the code even of a single
271 This helps explain in part, below, why compliance with
272 [pep8](https://pep8.org/) is enforced, including its 80 character limit.
273 In short: not everyone has the same "modern" GUI workflow or has access
274 to the same computing resources as you, so please do respect that.
276 More on this concept is
277 [here](https://www.linuxjournal.com/content/line-length-limits).
278 Note *very pointedly* that Linus Torvalds *specifically* states that
279 he does not want Linux kernel development to become the exclusive
280 domain of the "wealthy". That means **no** to assumptions about
281 access to ultra-high resolution screens.
283 # Software prerequisites<a name="software-prerequisites"></a>
285 **Please make sure if you install manually that you install dependencies
286 in strict order. Failing to adhere to this will result in pip3 downloading
287 unauthorised older software versions. See
288 <http://lists.libre-soc.org/pipermail/libre-soc-dev/2021-September/003666.html>**
290 Whilst many resources online advocate "`sudo`" in front of all root-level
291 commands below, this quickly becomes tiresome. run "`sudo bash`", get a
292 root prompt, and save yourself some typing.
295 * apt-get install vim exuberant-ctags
296 * apt-get install build-essential
297 * apt-get install git python3.7 python3.7-dev python3-nose
298 * apt-get install graphviz xdot gtkwave
299 * apt-get install python3-venv
300 * apt-get install python-virtualenv # this is an alternative to python3-venv
301 * apt-get install tcl-dev libreadline-dev bison flex libffi-dev iverilog
302 * return to user prompt (ctrl-d)
304 (The above assumes that you are running Debian.)
306 This will get you python3 and other tools that are
307 needed. [graphviz](https://graphviz.org/) is essential
308 for showing the interconnections between cells, and
309 [gtkwave](http://gtkwave.sourceforge.net/) is essential for debugging.
311 If you would like to save yourself a lot more typing, check out the
312 [dev-env-setup](https://git.libre-soc.org/?p=dev-env-setup.git;a=summary)
313 repository, examine the scripts there and use them to automate much of
316 If you would like just to install only the apt dependencies use
317 [install-hdl-apt-reqs](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=install-hdl-apt-reqs;hb=HEAD) instead.
319 This page gives more details and a step by step process : [[HDL_workflow/devscripts]]
323 Look up good tutorials on how to use git effectively. There are so many
324 it is hard to recommend one. This is however essential. If you are not
325 comfortable with git, and you let things stay that way, it will seriously
326 impede development progress.
328 If working all day you should expect to be making at least two commits per
329 hour, so should become familiar with it very quickly. If you are *not*
330 doing around 2 commits per hour, something is wrong and you should read
331 the workflow instructions below more carefully, and also ask for advice
334 Worth noting: *this project does not use branches*. All code is committed
335 to master and we *require* that it be either zero-impact additions or that
336 relevant unit tests pass 100%. This ensures that people's work does not
337 get "lost" or isolated and out of touch due to major branch diversion,
338 and that people communicate and coordinate with each other.
340 This is not a hard rule: under special cirmstances branches can be useful.
341 They should not however be considered "routine".
345 Follow the source code (git clone) instructions here, do **not** use
346 the "stable" version (do not download the tarball):
347 <http://www.clifford.at/yosys/download.html>
349 Or, alternatively, use the
350 [hdl-tools-yosys](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=hdl-tools-yosys;hb=HEAD)
351 script (which also installs symbiyosys and its dependencies)
353 Do not try to use a fixed revision of yosys (currently 0.9), nmigen is
354 evolving and frequently interacts with yosys.
356 [Yosys](http://www.clifford.at/yosys/) is a framework for Verilog RTL.
357 [Verilog](https://en.wikipedia.org/wiki/Verilog) is a hardware description
359 RTL [Register Transfer
360 Level](https://en.wikipedia.org/wiki/Register-transfer_level)
361 models how data moves between
362 [registers](https://en.wikipedia.org/wiki/Hardware_register).
366 To install follow the [instructions
367 here](https://symbiyosys.readthedocs.io/en/latest/install.html)
368 Once done look at [A simple BMC
369 example](https://symbiyosys.readthedocs.io/en/latest/quickstart.html)
371 You do not have to install all of those (avy, boolector can be left
372 out if desired) however the more that are installed the more effective
373 the formal proof scripts will be (less resource utilisation in certain
376 [SymbiYosys](https://symbiyosys.readthedocs.io/en/latest/) (sby) is a
377 front-end driver program for Yosys-based formal hardware verification
382 **PLEASE NOTE: it is critical to install nmigen as the first dependency
383 prior to installing any further python-based Libre-SOC HDL repositories.
384 If "pip3 list" shows that nmigen has been auto-installed please remove it**
386 [nmigen](https://nmigen.info/) may be installed as follows:
390 * git clone https://github.com/nmigen/nmigen.git
393 * python3 setup.py develop
396 Testing can then be carried out with "python3 setup.py test"
398 nmigen is a Python toolbox for building complex digital hardware.
400 ## Softfloat and sfpy
402 These are a test suite dependency for the
403 [ieee754fpu](https://www.gaisler.com/index.php/products/ipcores/ieee754fpu)
404 library, and will be changed in the future to use Jacob's
405 [simple-soft-float](https://crates.io/crates/simple-soft-float) library.
406 In the meantime, sfpy can be built as follows:
408 git clone --recursive https://github.com/billzorn/sfpy.git
411 git apply ../softposit_sfpy_build.patch
412 git apply /path/to/ieee754fpu/SoftPosit.patch
413 cd ../berkely-softfloat-3
414 # Note: Do not apply the patch included in sfpy for berkely-softfloat,
415 # it contains the same changes as this one
416 git apply /path/to/ieee754fpu/berkeley-softfloat.patch
419 # prepare a virtual environment for building
422 # or, if you prefer the old way:
423 # virtualenv -p python3 .env
425 # install dependencies
426 source .env/bin/activate
427 pip3 install --upgrade -r requirements.txt
432 make inplace -j$(nproc)
436 deactivate # deactivates venv, optional
437 pip3 install dist/sfpy*.whl
439 You can test your installation by doing the following:
442 >>> from sfpy import *
445 It should print out `Posit8(1.3125)`
447 ## qemu, cross-compilers, gdb
449 As we are doing POWER ISA, POWER ISA compilers, toolchains and
450 emulators are required.
451 Again, if you want to save yourself some typing, use the dev scripts.
452 [install-hdl-apt-reqs](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=install-hdl-apt-reqs;hb=HEAD)
453 script will install the qemu;
454 [ppc64-gdb-gcc](https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=ppc64-gdb-gcc;hb=HEAD)
455 script will install the toolchain and the corresponding debugger.
456 The steps are provided below only for reference; when in doubt,
457 consider checking and running the scripts.
459 Install powerpc64 gcc:
461 apt-get install gcc-8-powerpc64-linux-gnu
465 apt-get install qemu-system-ppc
467 Install gdb from source. Obtain the required tarball matching
468 the version of gcc (8.3) from here <https://ftp.gnu.org/gnu/gdb/>,
471 cd gdb-8.3 (or other location)
474 ../configure --srcdir=.. --host=x86_64-linux --target=powerpc64-linux-gnu
478 [gdb](https://en.wikipedia.org/wiki/GNU_Debugger) lets you debug running
479 programs. [qemu](https://www.qemu.org/) emulates processors, you can
480 run programs under qemu.
482 ## power-instruction-analyzer (pia)
484 We have a custom tool built in Rust by programmerjake to help analyze
485 the OpenPower instructions' execution on *actual* hardware.
489 curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
491 Make sure we have the correct and up-to-date rust compiler (rustc & cargo):
493 rustup default stable
496 Install the Python extension from git source by doing the following:
498 git clone https://salsa.debian.org/Kazan-team/power-instruction-analyzer.git pia
500 ./libre-soc-install.sh
504 As this is an actual ASIC, we do not rely on an FPGA's JTAG TAP
505 interface, instead require a full complete independent implementation
506 of JTAG. Staf Verhaegen has one, with a full test suite, and it is
507 superb and well-written. The Libre-SOC version includes DMI (Debug
510 git clone https://git.libre-soc.org/git/c4m-jtag.git/
512 python3 setup.py develop
514 Included is an IDCODE tap point, Wishbone Master (for direct memory read
515 and write, fully independent of the core), IOPad redirection and testing,
516 and general purpose shift register capability for any custom use.
518 We added a DMI to JTAG bridge in LibreSOC which is
519 directly connected to the core, to access registers and
520 to be able to start and stop the core and change the PC.
521 In combination with the JTAG Wishbone interface the test
522 [ASIC](https://en.wikipedia.org/wiki/Application-specific_integrated_circuit)
523 can have a bootloader uploaded directly into onboard
524 [SRAM](https://en.wikipedia.org/wiki/Static_random-access_memory) and
527 [Chips4Makers](https://chips4makers.io/) make it possible for makers
528 and hobbyists to make their own open source chips.
530 [JTAG](https://en.wikipedia.org/wiki/JTAG) (Joint Test Action Group) is
531 an industry standard for verifying designs and testing printed circuit
532 boards after manufacture.
535 bus](https://en.wikipedia.org/wiki/Wishbone_%28computer_bus%29) is an open
536 source hardware computer bus intended to let the parts of an integrated
537 circuit communicate with each other.
540 See [[HDL_workflow/coriolis2]] page, for those people doing layout work.
544 A portable FPGA place and route tool.
546 See [[HDL_workflow/nextpnr]] page for installation instructions of nextpnr with ECP5 support for Lattice FPGA ECP5 series. Also see
547 [[HDL_workflow/ECP5_FPGA]] for connecting up to JTAG with a ULX3S
548 and the Lattice VERSA_ECP5.
552 The fastest Verilog and SystemVerilog simulator. It compiles Verilog to C++ or SystemC.
554 Advise use only v4.106 at the moment.
556 See [[HDL_workflow/verilator]] page for installation instructions.
560 GHDL is a shorthand for G Hardware Design Language. It is a VHDL analyzer, compiler, simulator and (experimental) synthesizer that can process (nearly) any VHDL design.
562 VHDL is an acronym for Very High Speed Integrated Circuit (VHSIC) Hardware Description Language (HDL), which is a programming language used to describe a logic circuit by function, data flow behavior, or structure.
564 Unlike some other simulators, GHDL is a compiler: it directly translates a VHDL file to machine code, without using an intermediary language such as C or C++. Therefore, the compiled code should be faster and the analysis time should be shorter than with a compiler using an intermediary language.
566 GHDL aims at implementing VHDL as defined by IEEE 1076. It supports the 1987, 1993 and 2002 revisions and, partially, 2008. PSL is also partially supported.
568 See [[HDL_workflow/ghdl]] page for installation instructions.
572 Icarus Verilog is a Verilog simulation and synthesis tool. It operates as a compiler, compiling source code written in Verilog (IEEE-1364) into some target format.
574 See [[HDL_workflow/iverilog]] page for installation instructions.
578 cocotb is a COroutine based COsimulation TestBench environment for verifying VHDL and SystemVerilog RTL using Python.
580 See [[HDL_workflow/cocotb]] page for installation instructions.
584 A fully open source toolchain for the development of FPGAs. Currently it targets Xilinx 7-series, Lattice iCE40 and ECP5, Quicklogic EOS S3.
586 Needed for the Arty A7 100t Digilent FPGA board.
588 See [[HDL_workflow/symbiflow]] for installation instructions
591 # Registering for git repository access<a name="gitolite3_access"></a>
593 After going through the onboarding process and having agreed to take
594 responsibility for certain tasks, ask on the mailing list for git
595 repository access, sending in a public key (`id_rsa.pub`). If you do
596 not have one then generate it with `ssh-keygen -t rsa`. You will find it
599 NEVER SEND ANYONE THE PRIVATE KEY. By contrast the public key, on
600 account of being public, is perfectly fine to make... err... public.
602 Create a file `~/.ssh/config` with the following lines:
604 Host git.libre-soc.org
607 Test that you have access with this command:
609 ssh -v -p922 gitolite3@git.libre-soc.org
611 Please note: **DO NOT TYPE A PASSWORD** - the server gets hit by a lot of
612 port-scanning, and detection of password failures are used to instantly
615 Wait for the Project Admin to confirm that the ssh key has been added
616 to the required repositories. Once confirmed, you can clone any of the
617 repos at https://git.libre-soc.org/:
619 git clone gitolite3@git.libre-soc.org:REPONAME.git
621 Alternatively, the .ssh/config can be skipped and this used:
623 git clone ssh://gitolite3@git.libre-soc.org:922/REPONAME.git
625 Note: **DO NOT ATTEMPT TO LOG IN TO THE SERVER WITH A PERSONAL ACCOUNT**.
626 fail2ban is running and, due to repeated persistent port-scanning spammers
627 is set up to instantly ban any unauthorised ssh access for up to two weeks.
628 This keeps log file sizes down on the server (which is resource-constrained).
629 If you are wondering why this is done, it's a *lot* of port-scans.
631 Therefore, *only* ssh in to server with the gitolite3 account, *only*
632 on port 922, and *only* once the systems administrator has given you
633 the all-clear that the ssh key has been added.
637 Although there are methods online which describe how (and why) these
638 settings are normally done, honestly it is simpler and easier to open
639 ~/.gitconfig and add them by hand.
641 core.autocrlf is a good idea to ensure that anyone adding DOS-formatted
642 files they don't become a pain. pull.rebase is something that is greatly
643 preferred for this project because it avoids the mess of "multiple
644 extra merge git tree entries", and branch.autosetuprebase=always will,
645 if you want it, always ensure that a new git checkout is set up with rebase.
654 autosetuprebase = always
656 # Checking out the HDL repositories
658 Before running the following, install the
659 dependencies. This is easiest done with this script
660 <https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=install-hdl-apt-reqs;hb=HEAD>
662 **It is critically important to install these in STRICT order, otherwise
663 pip3 interferes and performs unauthorised downloads without informing
664 you of what it is doing**.
668 * git clone gitolite3@git.libre-soc.org:nmigen.git
669 * git clone gitolite3@git.libre-soc.org:c4m-jtag.git
670 * git clone gitolite3@git.libre-soc.org:nmutil.git
671 * git clone gitolite3@git.libre-soc.org:openpower-isa.git
672 * git clone gitolite3@git.libre-soc.org:ieee754fpu.git
673 * git clone gitolite3@git.libre-soc.org:nmigen-soc.git
674 * git clone gitolite3@git.libre-soc.org:soc.git
676 In each of these directories, **in the order listed***, track down the
677 `setup.py` file, then, as root (`sudo bash`), run the following:
679 * python3 setup.py develop
681 The reason for using "develop" mode is that the code may be edited
682 in-place yet still imported "globally". There are variants on this theme
683 for multi-user machine use however it is often just easier to get your
684 own machine these days.
686 The reason for the order is because soc depends on ieee754fpu, and
687 ieee754fpu depends on nmutil. If you do not follow the listed order
688 pip3 will go off and download an arbitrary version without your
691 If "`python3 setup.py install`" is used it is a pain: edit, then
692 install. edit, then install. It gets extremely tedious, hence why
693 "develop" was created.
695 If you prefer you can use this script instead: of course you checked it
696 in advance and accept full responsibility.
697 <https://git.libre-soc.org/?p=dev-env-setup.git;a=blob;f=hdl-dev-repos;hb=HEAD>
703 * new members, add yourself to the [[about_us]] page and create yourself
704 a home page using someone else's page as a template.
705 * communicate on the mailing list or the bugtracker an intent to take
706 responsibility for a particular task.
707 * assign yourself as the bug's owner
708 * *keep in touch* about what you are doing, and why you are doing it.
709 * edit your home page regularly, particularly to track tasks so that
710 they can be paid by NLNet.
711 * if you cannot do something that you have taken responsibility for,
712 then unless it is a dire personal emergency please say so, on-list. we
713 won't mind. we'll help sort it out.
715 Regarding the above it is important that you read, understand, and agree
716 to the [[charter]] because the charter is about ensuring that we operate
717 as an effective organisation. It's *not* about "setting rules and meting
722 for actual code development
726 * plan in advance to write not just code but a full test suite for
727 that code. **this is not optional**. large python projects that do not
728 have unit tests **FAIL** (see separate section below).
729 * Prioritise writing formal proofs and a single clear unit test that is more
730 like a "worked example".
731 We receive NLNet funds for writing formal proofs, plus they
732 cover corner cases and take far less time to write
734 ### Commit tested or zero-dependent code
736 * only commit code that has been tested (or is presently unused). other
737 people will be depending on you, so do take care not to screw up.
738 not least because, as it says in the [[charter]] it will be your
739 responsibility to fix. that said, do not feel intimidated: ask for help
740 and advice, and you'll get it straight away.
744 * commit often. several times a day, and "git push" it. this is
745 collaboration. if something is left even overnight uncommitted and not
746 pushed so that other people can see it, it is a red flag.
748 yourself thinking "i'll commit it when it's finished" or "i don't want to
749 commit something that people might criticise" *this is not collaboration*,
750 it is making yourself a bottleneck. pair-programming is supposed to help
751 avoid this kind of thing however pair-programming is difficult to organise
752 for remote collaborative libre projects (suggestions welcomed here)
754 ### Enable editor auto-detection of file changes by external programs
756 This is important. "`git pull`" will merge in changes. If you then
757 arbitrarily save a file without re-loading it, you risk destroying
760 You can avoid damaging the repositories by following some simple procedures:
762 run appropriate unit tests
764 run appropriate unit tests again (checks other people's work)
765 git diff # and actually read and review the output
766 git status # check for any missing files
767 git commit # with appropriate arguments and message
768 git push # always always always do this
770 ### Absolutely no auto-generated output
772 * **do not commit autogenerated output**. write a shell script and commit
773 that, or add a `Makefile` to run the command that generates the output, but
774 **do not** add the actual output of **any** command to the repository.
775 ever. this is really important. even if it is a human-readable file
776 rather than a binary object file.
777 * it is very common to add PDFs (the result of running `latex2pdf`) or
778 configure.in (the result of running `automake`), they are an absolute
779 nuisance and interfere hugely with git diffs, as well as waste hard
780 disk space *and* network bandwidth. don't do it.
781 * do not add multi-megabyte or multi-gigabyte "test data".
782 use shell scripts and commit that, which automatically downloads the
783 "test data" from a well-known known-good reliable location instead.
785 ### Write commands that do tasks and commit those
787 * if the command needed to create any given autogenerated output is not
788 currently in the list of known project dependencies, first consult on
789 the list if it is okay to make that command become a hard dependency of
790 the project (hint: java, node.js php and .NET commands may cause delays
791 in response time due to other list participants laughing hysterically),
792 and after a decision is made, document the dependency and how its source
793 code is obtained and built (hence why it has to be discussed carefully)
794 * if you find yourself repeating commands regularly, chances are high
795 that someone else will need to run them, too. clearly this includes
796 yourself, therefore, to make everyone's lives easier including your own,
797 put them into a `.sh` shell script (and/or a `Makefile`), commit them to
798 the repository and document them at the very minimum in the README,
799 INSTALL.txt or somewhere in a docs folder as appropriate. if unsure,
800 ask on the mailing list for advice.
802 ### Keep commits single-purpose
804 * edit files making minimal *single purpose* modifications (even if
805 it involves multiple files. Good extreme example: globally changing
806 a function name across an entire codebase is one purpose, one commit,
807 yet hundreds of files. miss out one of those files, requiring multiple
808 commits, and it actually becomes a nuisance).
810 ### Run unit tests prior to commits
812 * prior to committing make sure that relevant unit tests pass, or that
813 the change is a zero-impact addition (no unit tests fail at the minimum)
815 ### Do not break existing code
817 * keep working code working **at all times**. find ways to ensure that
818 this is the case. examples include writing alternative classes that
819 replace existing functionality and adding runtime options to select
820 between old and new code.
822 ### Small commits with relevant commit message
824 * commit no more than around 5 to 10 lines at a time, with a CLEAR message
825 (no "added this" or "changed that").
826 * if as you write you find that the commit message involves a *list* of
827 changes or the word "and", then STOP. do not proceed: it is a "red flag"
828 that the commit has not been properly broken down into separate-purpose
829 commits. ask for advice on-list on how to proceed.
831 ### Exceptions to small commit: atomic single purpose commit
833 * if it is essential to commit large amounts of code, ensure that it
834 is **not** in use **anywhere** by any other code. then make a *small*
835 (single purpose) followup commit which actually puts that code into use.
837 This last rule is kinda flexible, because if you add the code *and* add
838 the unit test *and* added it into the main code *and* ran all relevant
839 unit tests on all cascade-impacted areas by that change, that's perfectly
840 fine too. however if it is the end of a day, and you need to stop and
841 do not have time to run the necessary unit tests, do *not* commit the
842 change which integrates untested code: just commit the new code (only)
843 and follow up the next day *after* running the full relevant unit tests.
845 ### Why such strict rules?
847 The reason for all the above is because python is a dynamically typed
848 language. make one tiny change at the base level of the class hierarchy
849 and the effect may be disastrous.
851 It is therefore worth reiterating: make absolutely certain that you *only*
852 commit working code or zero-impact code.
854 Therefore, if you are absolutely certain that a new addition (new file,
855 new class, new function) is not going to have any side-effects, committing
856 it (a large amount of code) is perfectly fine.
858 As a general rule, however, do not use this an an excuse to write code
859 first then write unit tests as an afterthought. write *less* code *in
860 conjunction* with its (more basic) unit tests, instead. then, folliw up with
861 additions and improvements.
863 The reason for separating out commits to single purpose only becomes
864 obvious (and regretted if not followed) when, months later, a mistake
865 has to be tracked down and reverted. if the commit does not have an
866 easy-to-find message, it cannot even be located, and once found, if the
867 commit confuses several unrelated changes, not only the diff is larger
868 than it should be, the reversion process becomes extremely painful.
872 * all code needs to conform to pep8. use either pep8checker or better
873 run autopep8. however whenever committing whitespace changes, *make a
874 separate commit* with a commit message "whitespace" or "autopep8 cleanup".
875 * pep8 REQUIRES no more than 80 chars per line. this is non-negotiable. if
876 you think you need greater than 80 chars, it *fundamentally* indicates
877 poor code design. split the code down further into smaller classes
880 ### Docstring checker
882 * TBD there is a docstring checker. at the minimum make sure to have
883 an SPD license header, module header docstring, class docstring and
884 function docstrings on at least non-obvious functions.
886 ### Clear code commenting and docstrings
888 * make liberal but not excessive use of comments. describe a group of
889 lines of code, with terse but useful comments describing the purpose,
890 documenting any side-effects, and anything that could trip you or other
891 developers up. unusual coding techniques should *definitely* contain
894 ### Only one class per module (ish)
896 * unless they are very closely related, only have one module (one class)
897 per file. a file only 25 lines long including imports and docstrings
898 is perfectly fine however don't force yourself. again, if unsure,
901 ### File and Directory hierarchy
903 * *keep files short and simple*. see below as to why
904 * create a decent directory hierarchy but do not go mad. ask for advice
909 * please do not use "from module import \*". it is extremely bad practice,
910 causes unnecessary resource utilisation, makes code readability and
911 tracking extremely difficult, and results in unintended side-effects.
913 Example: often you want to find the code from which a class was imported.
914 nirmally you go to the top of the file, check the imports, and you know
915 exactly which file has the class because of the import path. by using
916 wildcards, you have absolutely *no clue* which wildcard imported which
919 Example: sometimes you may accidentally have duplicate code maintained
920 in two or more places. editing one of them you find, puzzlingly, that
921 the code behaves in some files with the old behaviour, but in others it
922 works. after a massive amount of investigation, you find that the working
923 files happen to have a wildcard import of the newer accidental duplicate
924 class **after** the wildcard import of the older class with exactly the
925 same name. if you had used explicit imports, you would have spotted
926 the double import of the class from two separate locations, immediately.
928 Really. don't. use. wildcards.
930 ### Keep file and variables short but clear
932 * try to keep both filenames and variable names short but not ridiculously
933 obtuse. an interesting compromise on imports is "from ridiculousfilename
934 import longsillyname as lsn", and to assign variables as well: "comb =
935 m.d.comb" followed by multiple "comb += nmigen_stmt" lines is a good trick
936 that can reduce code indentation by 6 characters without reducing clarity.
938 Additionally, use comments just above an obtuse variable in order to
939 help explain what it is for. In combination with keeping the the module
940 itself short, other readers will not need to scroll back several pages
941 in order to understand the code.
943 Yes it is tempting to actually use the variables as
944 self-explanatory-comments and generally this can be extremely good
945 practice. the problem comes when the variable is so long that a function
946 with several parameters csn no longer fit on a single line, and takes
947 up five to ten lines rather than one or two. at that point, the length
948 of the code is adversely affected and thus so is readability by forcing
949 readers to scroll through reams of pages.
951 It is a tricky balance: basically use your common sense, or just ask
952 someone else, "can you understand this code?"
954 ### Reasons for code structure
956 Regarding code structure: we decided to go with small modules that are
957 both easy to analyse, as well as fit onto a single page and be readable
958 when displayed as a visual graph on a full UHD monitor. this is done
961 * using the capability of nmigen (TODO crossref to example) output the
962 module to a yosys ilang (.il) file
963 * in a separate terminal window, run yosys
964 * at the yosys prompt type "read_ilang modulename.il"
965 * type "show top" and a graphviz window should appear. note that typing
966 show, then space, then pressing the tab key twice will give a full list
967 of submodules (one of which will be "top")
969 You can now fullsize the graphviz window and scroll around. if it looks
970 reasonably obvious at 100% zoom, i.e the connections can be clearly
971 related in your mind back to the actual code (by matching the graph names
972 against signals and modules in the original nmigen code) and the words are
973 not tiny when zoomed out, and connections are not total incomprehensible
974 spaghetti, then congratulations, you have well-designed code. If not,
975 then this indicates a need to split the code further into submodules
976 and do a bit more work.
978 The reasons for doing a proper modularisation job are several-fold:
980 * firstly, we will not be doing a full automated layout-and-hope
981 using alliance/coriolis2, we will be doing leaf-node thru tree node
982 half-automated half-manual layout, finally getting to the floorplan,
983 then revising and iteratively adjusting.
984 * secondly, examining modules at the gate level (or close to it) is just
985 good practice. poor design creeps in by *not* knowing what the tools
986 are actually doing (word to experienced developers: yes, we know that
987 the yosys graph != final netlist).
988 * thirdly, unit testing, particularly formal proofs, is far easier on
989 small sections of code, and complete in a reasonable time.
991 ## Special warning / alert to vim users!
993 Some time around the beginning of 2019 some bright spark decided that
994 an "auto-recommend-completion-of-stuff" option would be a nice, shiny
995 idea to enable by default from that point onwards.
997 This incredibly annoying "feature" results in tabs (or spaces) being
998 inserted "on your behalf" when you press return on one line, for your
999 "convenience" of not needing to type lots of spaces/tabs just to get
1000 to the same indentation level.
1002 Of course, this "feature", if you press return on one line in edit
1003 mode and then press "escape", leaves a bundle-of-joy extraneous
1004 whitespace **exactly** where you don't want it, and didn't ask for it,
1005 pooped all over your file.
1007 Therefore, *please*: **before** running "git commit", get into the
1008 habit of always running "git diff", and at the very minimum
1009 speed-skim the entire diff, looking for tell-tale "red squares"
1010 (these show up under bash diff colour-syntax-highlighting) that
1011 inform you that, without your knowledge or consent, vim has
1012 "helpfully" inserted extraneous whitespace.
1014 Remove them **before** git committing because they are not part
1015 of the actual desired code-modifications, and committing them
1016 is a major and constant distraction for reviewers about actual
1017 important things like "the code that actually *usefully* was
1018 modified for that commit"
1020 This has the useful side-effect of ensuring that, right before
1021 the commit, you've got the actual diff right in front of you
1022 in the xterm window, on which you can base the "commit message".
1026 For further reading, see the wikipedia page on
1027 [Test-driven Development](https://en.wikipedia.org/wiki/Test-driven_development)
1029 This deserves its own special section. It is extremely important to
1030 appreciate that without unit tests, python projects are simply unviable.
1031 Python itself has over 25,000 individual tests.
1033 This can be quite overwhelming to a beginner developer, especially one
1034 used to writing scripts of only 100 lines in length.
1036 Thanks to Samuel Falvo we learned that writing unit tests as a formal
1037 proof is not only shorter, it's also far more readable and also, if
1038 written properly, provides 100% coverage of corner-cases that would
1039 otherwise be overlooked or require tens to hundreds of thousands of
1042 No this is not a joke or even remotely hypothetical, this is an actual
1045 The ieee754fpu requires several hundreds of thousands of tests to be
1046 run (currently needing several days to run them all), and even then we
1047 cannot be absolutely certain that all possible combinations of input have
1048 been tested. With 2^128 permutations to try with 2 64 bit FP numbers
1049 it is simply impossible to even try.
1051 This is where formal proofs come into play.
1053 Samuel illustrated to us that "ordinary" unit tests can then be written
1054 to *augment* the formal ones, serving the purpose of illustrating how
1055 to use the module, more than anything.
1057 However it is appreciated that writing formal proofs is a bit of a
1058 black art. This is where team collaboration particularly kicks in,
1059 so if you need help, ask on the mailing list.
1061 ## Don't comment out unit tests: add them first (as failures) and fix code later
1063 Unit tests serve an additional critical purpose of keeping track of code
1064 that needs to be written. In many cases, you write the unit test *first*,
1065 despite knowing full well that the code doesn't even exist or is completely
1066 broken. The unit test then serves as a constant and important reminder
1067 to actually fix (or write) the code.
1069 Therefore, *do not* comment out unit tests just because they "don't work".
1070 If you absolutely must stop a unit test from running, **do not delete it**.
1071 Simply mark it with an appropriate
1072 ["skip" decorator](https://docs.python.org/3/library/unittest.html#skipping-tests-and-expected-failures),
1073 preferably with a link to a URL in the [bugtracker](https://bugs.libre-soc.org/)
1074 with further details as to why the unit test should not be run.
1078 Find appropriate tutorials for nmigen and yosys, as well as symbiyosys.
1080 * Robert Baruch's nmigen tutorials look really good:
1081 <https://github.com/RobertBaruch/nmigen-tutorial>
1082 * Although a verilog example this is very useful to do
1083 <https://symbiyosys.readthedocs.io/en/latest/quickstart.html#first-step-a-simple-bmc-example>
1084 * This tutorial looks pretty good and will get you started
1085 <https://web.archive.org/web/20210123052724/http://blog.lambdaconcept.com/doku.php?id=nmigen:nmigen_install>
1086 and walks not just through simulation, it takes you through using
1088 * There exist several nmigen examples which are also executable
1089 <https://github.com/nmigen/nmigen/tree/master/examples/> exactly as
1090 described in the above tutorial (python3 filename.py -h)
1091 * More nmigen tutorials at [[learning_nmigen]]