Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.lang.forth > #135352 > unrolled thread
| Started by | Terry Porter <techman001@eternal-september.org> |
|---|---|
| First post | 2026-08-16 19:49 +0000 |
| Last post | 2026-08-30 01:47 +0000 |
| Articles | 5 — 3 participants |
Back to article view | Back to comp.lang.forth
Techman is back after a few years away being very busy with embedded Forth Terry Porter <techman001@eternal-september.org> - 2026-08-16 19:49 +0000
Re: Techman is back after a few years away being very busy with embedded Forth antispam@fricas.org (Waldek Hebisch) - 2026-08-18 16:54 +0000
Re: Techman is back after a few years away being very busy with embedded Forth Terry Porter <techman001@eternal-september.org> - 2026-08-19 04:01 +0000
Re: Techman is back after a few years away being very busy with embedded Forth Kragen Javier Sitaker <kragen@canonical.org> - 2026-08-28 18:12 -0300
Re: Techman is back after a few years away being very busy with embedded Forth Terry Porter <techman001@eternal-september.org> - 2026-08-30 01:47 +0000
| From | Terry Porter <techman001@eternal-september.org> |
|---|---|
| Date | 2026-08-16 19:49 +0000 |
| Subject | Techman is back after a few years away being very busy with embedded Forth |
| Message-ID | <115t48v$eq20$1@dont-email.me> |
Hi all, I can't even remember what nick I used last time ! Perhaps it's because I'm 72 years old now, who knows ? Anyway, I created and maintain the "Unofficial Mecrisp-Stellaris userdoc" https://mecrisp-stellaris-folkdoc.sourceforge.io/index.html which I know a few have seen. I retired about six years ago and have really been productive with embedded Forth and would like to announce a couple of completed MIT licensed projects for those here that do embedded. Firstly there is 'Regmon' 10 second video as the 24" monitor is in portrait mode: https://github.com/techman00172/regmon-re/blob/master/pics/regmon.mp4 Regmon is a live register monitor for STM32 microcontrollers. It reads the chip's registers over SWD while the target runs -- no stop, no re- flash -- and decodes them against the official CMSIS-SVD files. Works alongside a running Mecrisp Forth kernel on the same debug link, so you can watch RCC/GPIO/timer registers change as your Forth words execute. Registers, bitfields and an AI-assisted analysis pane. Linux, Python. Three buttons you will not find on any other register tool: Analyse Reg samples one register about twenty times over ten seconds and the AI tells you what that register is actually doing while the chip runs. Analyse Prog fingerprints the whole chip -- it reads the RCC clock gates and the dynamic registers, samples the program counter and stack pointer, and the AI gives a verdict on what the program is doing, live, without stopping the core. That is the unusual bit: conventional microprocessor tools freeze the chip at a breakpoint to inspect it, or need a logic analyser strapped to the pins. Analyse Prog works on a running target over the same SWD link the Forth kernel is using -- no halt, no re-flash, no extra hardware. Strings scans flash for readable text, so you can see what firmware is actually on the chip -- handy for identifying an unknown board. Regmon's AI reads your registers live and tells you what they mean. As the chip runs, Regmon streams the changing register values over SWD and hands them to an AI, which decodes each one against the silicon's own register definitions -- the bitfields, the flags, what each change indicates. When a value flips, Regmon explains why it matters in plain language, in real time. No more leafing through a thousand-page data book hunting for the right page to find out what a number means -- the answer arrives as the register changes. The AI is the data book that never has to be flipped. And here's the crucial part: none of that knowledge comes from the Internet. Regmon's AI answers come from inbuilt knowledge bases, built from the manufacturer's own data sheets and register definition (SVD) files -- the same source of truth the silicon's designers used. The databases ship with Regmon and live on your machine; there's no web lookup, no scraping, no reliance on a general-purpose AI that might have "read something somewhere." When Regmon tells you what a register means, it's answering from the authoritative documentation, packaged locally. The register descriptions are tightly bound to the manufacturer's own data -- no web lookup, no guessing. get it at https://github.com/techman00172/regmon-re But there is a BIG catch, you won't be able to install it manually as it's pretty complex, but there is a dead easy install proceedure. [If you don't already have the service of a frontier AI:] * Install opencode -- https://opencode.ai (free, one command) * Get an AI account -- DeepSeek works great and is cheap; or use a free model on OpenRouter's free tier (e.g. openai/gpt-oss-20b:free). A few dollars of OpenRouter credit is plenty. * Point opencode at the AI -- opencode with your API key (DeepSeek or OpenRouter). Two minutes. [If you do have an AI] * Tell the AI to install Regmon-RE from "github.com/techman00172/regmon-re" That's it :) Upcoming posts include "Fossilcrew" for sole developers, "The Anvil Gate", Mecrisp-Stellaris LSP, CMSIS-SVD LSP and more. Cheers, Terry
[toc] | [next] | [standalone]
| From | antispam@fricas.org (Waldek Hebisch) |
|---|---|
| Date | 2026-08-18 16:54 +0000 |
| Message-ID | <11622ob$v1c9$2@paganini.bofh.team> |
| In reply to | #135352 |
Terry Porter <techman001@eternal-september.org> wrote:
>
> Regmon is a live register monitor for STM32 microcontrollers. It reads
> the chip's registers over SWD while the target runs -- no stop, no re-
> flash -- and decodes them against the official CMSIS-SVD files. Works
> alongside a running Mecrisp Forth kernel on the same debug link, so you
> can watch RCC/GPIO/timer registers change as your Forth words execute.
> Registers, bitfields and an AI-assisted analysis pane. Linux, Python.
>
> Three buttons you will not find on any other register tool:
>
> Analyse Reg samples one register about twenty times over ten seconds
> and the AI tells you what that register is actually doing while the
> chip runs.
>
> Analyse Prog fingerprints the whole chip -- it reads the RCC clock
> gates and the dynamic registers, samples the program counter and
> stack pointer, and the AI gives a verdict on what the program is
> doing, live, without stopping the core. That is the unusual bit:
> conventional microprocessor tools freeze the chip at a breakpoint
> to inspect it, or need a logic analyser strapped to the pins.
> Analyse Prog works on a running target over the same SWD link the
> Forth kernel is using -- no halt, no re-flash, no extra hardware.
What target do you mean here? Arbitrary non-cooperating target
or Forth extended by your routines?
--
Waldek Hebisch
[toc] | [prev] | [next] | [standalone]
| From | Terry Porter <techman001@eternal-september.org> |
|---|---|
| Date | 2026-08-19 04:01 +0000 |
| Message-ID | <11639rb$2bqra$1@dont-email.me> |
| In reply to | #135365 |
On 2026-08-18, Waldek Hebisch <antispam@fricas.org> wrote: > Terry Porter <techman001@eternal-september.org> wrote: >> >> Regmon is a live register monitor for STM32 microcontrollers. It reads >> the chip's registers over SWD while the target runs -- no stop, no re- >> flash -- and decodes them against the official CMSIS-SVD files. Works >> alongside a running Mecrisp Forth kernel on the same debug link, so you >> can watch RCC/GPIO/timer registers change as your Forth words execute. >> Registers, bitfields and an AI-assisted analysis pane. Linux, Python. >> >> Three buttons you will not find on any other register tool: >> >> Analyse Reg samples one register about twenty times over ten seconds >> and the AI tells you what that register is actually doing while the >> chip runs. >> >> Analyse Prog fingerprints the whole chip -- it reads the RCC clock >> gates and the dynamic registers, samples the program counter and >> stack pointer, and the AI gives a verdict on what the program is >> doing, live, without stopping the core. That is the unusual bit: >> conventional microprocessor tools freeze the chip at a breakpoint >> to inspect it, or need a logic analyser strapped to the pins. >> Analyse Prog works on a running target over the same SWD link the >> Forth kernel is using -- no halt, no re-flash, no extra hardware. > > What target do you mean here? Arbitrary non-cooperating target > or Forth extended by your routines? > From: antispam@fricas.org (Waldek Hebisch) Newsgroups: comp.lang.forth Subject: Re: Techman is back after a few years away being very busy with embedded Forth Terry Porter <techman001@eternal-september.org> wrote: > > Regmon is a live register monitor for STM32 microcontrollers. It reads > the chip's registers over SWD while the target runs -- no stop, no re- > flash -- and decodes them against the official CMSIS-SVD files. Works > alongside a running Mecrisp Forth kernel on the same debug link, so you > can watch RCC/GPIO/timer registers change as your Forth words execute. > Registers, bitfields and an AI-assisted analysis pane. Linux, Python. > > Three buttons you will not find on any other register tool: > > Analyse Reg samples one register about twenty times over ten seconds > and the AI tells you what that register is actually doing while the > chip runs. > > Analyse Prog fingerprints the whole chip -- it reads the RCC clock > gates and the dynamic registers, samples the program counter and > stack pointer, and the AI gives a verdict on what the program is > doing, live, without stopping the core. That is the unusual bit: > conventional microprocessor tools freeze the chip at a breakpoint > to inspect it, or need a logic analyser strapped to the pins. > Analyse Prog works on a running target over the same SWD link the > Forth kernel is using -- no halt, no re-flash, no extra hardware. What target do you mean here? Arbitrary non-cooperating target or Forth extended by your routines? Re: Techman is back - what target does Regmon actually work on? Waldek, good question - and the honest answer is: both, depending on which part of the tool you mean. The register reads themselves are non-cooperating. Reading RCC/GPIO/timer registers goes over the ST-Link's SWD debug port - a mem read is a debug-port read, full stop. It works on any target, running or halted, whether it's Mecrisp, C, or a bare chip with nothing flashed. No cooperation needed. There's a C-mode of the swdd daemon that skips all the Forth machinery and just serves mem/reg to an arbitrary target. Where it does depend on the kernel is the no-halt live channel - the part that makes it interesting. The "watch a register change while your Forth word runs, no breakpoint, no re-flash" trick is built on the SWD ring buffer that Mecrisp-Stellaris maintains. The kernel actively services that ring as part of its inner loop, so swdd can exchange data with it without halting the core. That's genuine cooperation - it only works because Mecrisp is designed that way. A C firmware that doesn't service a ring buffer won't give you that zero-intrusion live channel. So the truthful split: - Low-level register reads (the mem path): any target, non-cooperating, via the debug port. - Analyse Reg / Analyse Prog live sampling (the "watch it run" path): needs a Mecrisp kernel servicing the ring buffer - that's the cooperating case, and honestly it's the case the tool is built around. On "no halt" specifically: reading PC and SP via the debug port's core-register access does work while the core runs, but on a non-cooperating target it can stall the core briefly and it's nowhere near as clean. The continuous, genuinely zero-intrusion sampling is the Mecrisp ring-buffer path. I'd rather say that plainly than pretend it's magic. It's an STM32-specific tool (CMSIS-SVD register descriptions, roughly sixty families), Linux, Python, and it shares the debug link with the running Forth kernel - same ST-Link, no contention.
[toc] | [prev] | [next] | [standalone]
| From | Kragen Javier Sitaker <kragen@canonical.org> |
|---|---|
| Date | 2026-08-28 18:12 -0300 |
| Message-ID | <877blac5k5.fsf@debian> |
| In reply to | #135352 |
Terry Porter <techman001@eternal-september.org> writes: > I retired about six years ago and have really been productive with > embedded Forth and would like to announce a couple of completed MIT > licensed projects for those here that do embedded. It’s great to see you back! I’ve missed talking with you on ##electronics and ##forth on Libera.chat, but I understand that IRC is a pretty flawed medium. Kragen
[toc] | [prev] | [next] | [standalone]
| From | Terry Porter <techman001@eternal-september.org> |
|---|---|
| Date | 2026-08-30 01:47 +0000 |
| Message-ID | <1170240$3n6a0$1@dont-email.me> |
| In reply to | #135468 |
On 2026-08-28, Kragen Javier Sitaker <kragen@canonical.org> wrote: > Terry Porter <techman001@eternal-september.org> writes: >> I retired about six years ago and have really been productive with >> embedded Forth and would like to announce a couple of completed MIT >> licensed projects for those here that do embedded. > > It’s great to see you back! I’ve missed talking with you on > ##electronics and ##forth on Libera.chat, but I understand that IRC is a > pretty flawed medium. > > Kragen Hi Xentrac! I also have missed our chats and your insightful deep knowledge. After I fell out with the irc.Liberia #electronics OP, I later fell out with the Redit/Forth Moderator and decided to install SLRN again and revisit Usenet, my old stomping ground for decades from 1994. It's the true home of free speech in my opinion. I was just getting started with AI while at Liberia.irc #electronics and have been using it daily for 4 - 5 months now, for all my software needs. As you know, I'm a electronics guy, not a programmer, so AI was the perfect fit for me and I've been making hay with it while the sun shines. In other words I've been using Deepseek-V4-Flash a LOT as it's so cheap right now, spending on average about $3 USD a day for around 10 hours a day. I don't believe this will last so I've spent a lot of that time using Deeseek to build a local Agent system for when the prices rise to those of the current American AI token pricing and it's working but at a much reduced capacity given my meagre computing resources. But we do what we can. Cheers Terry
[toc] | [prev] | [standalone]
Back to top | Article view | comp.lang.forth
csiph-web