GCodes for the next-generation Duet

[deckingman]

@fma I like this idea. Perhaps we could have an interpreter, a bit like when I used to write programs for my ZX81 using "Basic" (few people here will be old enough to know what I'm talking about). So "M569 Drive 0 Backwards" gets interpreted as "M569 P0 S0".

[deckingman]

@dc42 said in GCodes for the next-generation Duet:

@deckingman said in GCodes for the next-generation Duet:

I like this idea but why can't we use "B" for board or "E" for expansion board.

So e.g M569 B0 P0 means drive 0 on board zero. The default if no "B" value is present would be the main board so M569 P0 would mean the same thing and would be backwardly compatible for the majority of users with just one main board. Then M569 B1 P2 would mean drive 2 on (expansion) board 1.

That's a possibility, but it presupposes that B or E is not already used for another purpose by any of the GCode commands affected. Also it potentially makes it harder to identify the device when reading the GCode command, because the B or E parameter might be at one end of the line of GCode and the other parameter at the other end.

Are we stuck with having to use a single letter? Is something like "EB" to denote expansion board a no no?

[fma]

@deckingman said in GCodes for the next-generation Duet:

[...] I used to write programs for my ZX81 using "Basic" (few people here will be old enough to know what I'm talking about).

I am old enough! That good old ZX81, my very first computer ;o)

I would totally avoid the use of G/M-Codes, as there are so many, now, it is difficult to now what they are doing when reading the config file; using something close to linux shell commands would be better.

To make something really better, this would also require to re-think all commands, instead of translate each G/M-Code: many of them are related, and could be grouped under the same command, with modifier params. Auto bed leveling, for example, is very complicated, split in so many M-Codes that I gave up using it (well, in fact, I don't need it, as my bed is perfectly flat, and stiff enough ;o) ).

But this is a huge work :o/

[deckingman]

@fma I was just thinking that we seem to be stuck with G and M codes, the same way that we are stuck with querty keyboards. So something along the lines of users being able to use plain language to describe what they want to configure or happen, and which would shield them from having to know all the various G and M codes along with the associated parameters. The "interpreter" would take the plain language file or command, and convert it into the relevant G or M code.

[fma]

Another approach would be to use configs variables... More verbose from the user point of view (leading to a longer config file), but maybe easier to handle from the firmware point of view?

KINEMATIC_MODE = CORE_XY
X_SPEED_MAX = 10000
X_ACCEL = 3000

FAN_0_PORT = 0.4
FAN_0_PWM = 500
FAN_0_PWM_INVERT = True
...

Don't know if all commands can be set this way; some parameters may need to be set at the same time, so more difficult to deal with one by one...

[gtj0]

I like where this is going but even though the controller is 32-bit, speed and memory are still an issue.

@dc42 have you guys thought about separating the controller from the I/O boards, maybe in a hat or mother/daughter arrangement? This way you might be able to offer a basic arrangement with the SAM platform or more advanced arrangements with something like an i.MX6 SOM.

[dc42]

@fma said in GCodes for the next-generation Duet:

Another approach would be to use configs variables... More verbose from the user point of view (leading to a longer config file), but maybe easier to handle from the firmware point of view?

KINEMATIC_MODE = CORE_XY
X_SPEED_MAX = 10000
X_ACCEL = 3000

FAN_0_PORT = 0.4
FAN_0_PWM = 500
FAN_0_PWM_INVERT = True
...

Don't know if all commands can be set this way; some parameters may need to be set at the same time, so more difficult to deal with one by one...

Yes there are many cases in which it is best to set several parameters in one go.

I am looking for a solution that we can run on the Duet 2 series too, so that I can use a lot of common source code for the Duet 2 series and the next generation Duet. So whatever solution we pick shouldn't use a lot of additional RAM, but it can use more flash memory.

[dc42]

Here's a early draft proposal for new GCodes to create fans, temperature sensors, heaters and filament monitors. I have used M1000 as the base code for this, but it could instead be e.g. M594 or something else.

M1000.1 Create temperature sensor
M1000.2 Create heater
M1000.3 Create fan
M1000.4 Create endstop group
M1000.5 Create Z probe
M1000.6 Create filament monitor
M1000.7 Create general purpose output (for use by M42, M452, M452)
M1000.8 Create servo output (for use by M280)

Parameters:

Pn Index of the thing being created, e.g. 0, 1, 2...
A"name" Optional name of the thing being created, e.g. "Bed heater"
Dn Input/output terminal ID(s), e.g. 0, 2.1, ...
Fnnn PWM frequency, if applicable
In Invert/don't invert output, if applicable

Other parameters as required for the thing being created.

Thoughts? We could have a "Create axis" command too, however I think M584 does that well enough already.

[resam]

How sophisticated is the current g-code parser in RRF?
Would it be possible to make the M1000 a bit more self-describing, by adding a string identifier? for example:

M1000.fans P1 ...
M1000.heaters P1 ...
M1000.filament_monitors P1 ...

I fear that all those M1000.1 -> M1000.42 numbers become yet another random number everybody has problems remembering and identifying when reading through a config.g file...

[wilriker]

@resam Currently this will extract the number after the letter and will do a switch() { case... } over it. C/C++ can only do a switch on integers. There are workaround solutions that can be used to switch on string as well but they fall into the category of using more RAM - though not too much more RAM, so this might be a solution but I have no overview on how much RAM is left.

@dc42 I want to express how great I find the idea that you are a) asking the community and b) start to rattle off some of the rust that has settled on G-/M-Code.

EDIT: One slightly different idea for thought: how about creating a new code prefix for configuration like Cnnn or Dnnn?

[dragonn]

What about making groups of commands? Like for example:
Hnnn - anything connected with heaters
Snnn - anything connected with speeds
This should not add a big overhead in the software but would be already make it more readable.

[fcwilt]

@resam said in GCodes for the next-generation Duet:

How sophisticated is the current g-code parser in RRF?
Would it be possible to make the M1000 a bit more self-describing, by adding a string identifier? for example:

M1000.fans P1 ...
M1000.heaters P1 ...
M1000.filament_monitors P1 ...

I fear that all those M1000.1 -> M1000.42 numbers become yet another random number everybody has problems remembering and identifying when reading through a config.g file...

Remember the details can be difficult at times.

I add comments to each line so when I come back at a later date it jogs my memory.

Frederick

[wilriker]

@dragonn said in GCodes for the next-generation Duet:

What about making groups of commands? Like for example:
Hnnn - anything connected with heaters
Snnn - anything connected with speeds
This should not add a big overhead in the software but would be already make it more readable.

I like this idea! It would take some refactoring but as stated elsewhere, I offer my help for coding. I am a Java and Go developer in my day job and C++ is not my main language but I think I know it good enough to get good results as long as someone more experienced is doing code review.

[deckingman]

@dc42 said in GCodes for the next-generation Duet:

Here's a early draft proposal for new GCodes to create fans, temperature sensors, heaters and filament monitors. I have used M1000 as the base code for this, but it could instead be e.g. M594 or something else.

M1000.1 Create temperature sensor
M1000.2 Create heater
M1000.3 Create fan
M1000.4 Create endstop group
M1000.5 Create Z probe
M1000.6 Create filament monitor
M1000.7 Create general purpose output (for use by M42, M452, M452)
M1000.8 Create servo output (for use by M280)

Parameters:

Pn Index of the thing being created, e.g. 0, 1, 2...
A"name" Optional name of the thing being created, e.g. "Bed heater"
Dn Input/output terminal ID(s), e.g. 0, 2.1, ...
Fnnn PWM frequency, if applicable
In Invert/don't invert output, if applicable

Other parameters as required for the thing being created.

Thoughts? We could have a "Create axis" command too, however I think M584 does that well enough already.

My opinion probably doesn't count for much but personally, I think this is awful. It's a whole other bunch of codes to have to use and learn and none of them make much intuitive sense. So for us non-writers of code and especially for new users of Duet, it's going to add to what is already becoming too complex. I understand that with increased functionality comes increased complexity but there has to be a better way.

If we are stuck with using existing codes and parameters, do we have to use all alphanumeric characters? How about using existing codes to denote the "what" but with the "@" character to denote the "where"? e.g M305 @ n.n P0.... etc

[dc42]

Here is some background.

Currently, heaters and fans exist already when the firmware boots up because there is a fixed number of them. Filament monitors have to be created (M591 both creates and configures them), so do axes beyond X, Y and Z (M584).

Configuring additional temperatures to display in DWC or to control fans requires the somewhat messy "virtual heaters" concept. This is because currently, configuration of a temperature sensor is done by configuring a heater that uses that sensor. So in the new firmware I want temperature sensors to be created and configured independently of heaters.

With the next-gen Duet, creation of heaters, fans, sensors and GPIO pins will have to be done explicitly, because some configurations might need a large number of them and other configurations only a few. This means we need a mechanism to tell the firmware to create these items, either using new GCodes or using some other mechanism. From a firmware perspective, GCodes is easiest because the code to parse them is there already; but we could use some other mechanism.

My draft proposal was aimed at making the new codes easier to remember or at least to look up (by making them all sub-codes of M1000) and to use a consistent set of parameter letters within them.

There are ways we could avoid introducing new GCodes. Fans could be created by adding yet another parameter to M106. Heaters could be created by the M307 command, if I added a parameter to specify which temperature sensor to use (replacing the X parameter in M305) . M305 would be used to create temperature sensors, not heaters. But I feel that M106 already has too many parameters, and changing what M307 and M305 do would be more confusing that adding new rationalised GCodes to create items and leaving the old commands doing most of what they do currently (i.e. configuring things that already exist).

[fma]

Whatever syntax is used, I think it is a good idea to split creation of devices and their usage. M106 should only set the the fan speed. All M-Codes which do both are very confusing to use, and params not easy to remember.

If from a firmware point of view it is not possible to use a more advanced parser with plain english commands/params, then we should stay close to the current G-Code syntax. Using the . (dot) to split (or group) commands/params is a nice idea.

So, using M1000 to create stuff is OK. A question: will it be possible to dynamically remap a device? With M1000? Or another M-Code?

[dc42]

@fma said in GCodes for the next-generation Duet:

Whatever syntax is used, I think it is a good idea to split creation of devices and their usage. M106 should only set the the fan speed. All M-Codes which do both are very confusing to use, and params not easy to remember.

Thanks, it seems that you agree with me.

If from a firmware point of view it is not possible to use a more advanced parser with plain english commands/params, then we should stay close to the current G-Code syntax. Using the . (dot) to split (or group) commands/params is a nice idea.

The problem with a "plain english" parser is, what exactly constitutes plain English? Whatever definition I adopt, user s are bound to think of different ways of expressing the same thing that "ought" to be accepted, then we have the need to accept both English and American spellings. So I might end up spending a lot of time improving the parser, when it is more important to implement new firmware features. And how much would it help those for whom some form of English is not their native language?

So, using M1000 to create stuff is OK. A question: will it be possible to dynamically remap a device? With M1000? Or another M-Code?

My initial thought was that dynamic remapping would not be allowed; although it could be if it turns out to be useful. There will be some concepts that can be remapped, such as the "print cooling fan" (i.e. the fan whose speed is set by M106 with no P parameter) that can be mapped to different fan numbers at different times.

[deckingman]

How would we go about defining tools and stuff like mixing ratios?

[dc42]

I don't see any need to change how tools and mixing ratios are defined. It's things that relate directly to I/O pins that need to change.

[deckingman]

@dc42 said in GCodes for the next-generation Duet:

I don't see any need to change how tools and mixing ratios are defined. It's things that relate directly to I/O pins that need to change.

Ahh OK. Light is beginning to dawn.......
BTW I like what you are planning to do with temperature sensors being independent of heaters.