Move UBL mesh_index_to_x/y from sram to progmem

[bgort]

Tested and working well here. Saves 120 bytes of SRAM (in the case of a 15x15 mesh).

Discussed here: #6424

[Roxy-3D]

 constexpr float unified_bed_leveling::mesh_index_to_xpos[16],
                 unified_bed_leveling::mesh_index_to_ypos[16];

I thought this would have to have PROGMEM in it? But I guess the compiler likes what you have or Travis would have complained.

[bgort]

Yeah -- it works and there are no errors or warnings. The actual definition where PROGMEM matters is in the header.

[thinkyhead]

So we've decided not to implement variable grid size for UBL…?

I was hoping to extend leveling to allow the number of grid points and the probe region to be variable with other forms of leveling (as they are currently with AUTO_BED_LEVELING_LINEAR). I actually need this capability for my client, so I'm adding it to AUTO_BED_LEVELING_BILINEAR.

But, it would be good to be able to make it fixed and stored in PROGMEM if there's a need for extra SRAM.

[bgort]

What do you mean by variable, specifically? Without recompiling?

[thinkyhead]

It means that you can specify a different region of the bed or a different number of probe points when running G29… without recompiling.

[bgort]

Ah, yeah. Not sure what it'd take to make UBL variable in that way, but obviously the recent progmem-related changes don't help.

[thinkyhead]

Well, we've got one week to patch up stuff for release, so we'll revisit this for 1.1.1 / 1.2.

[bgort]

Makes sense to me.

Out of curiosity - and if you don't mind sharing - what's the use case requiring variable leveling?

[Roxy-3D]

Well, we've got one week to patch up stuff for release,

@thinkyhead @bgort @oldmcg
Check out: #6241 (comment)

Seriously.... We can have UBL working on Deltas if we wait an extra week. If a few people with Deltas pitch in... It is a no-brainer. Right now... I'm OK with pushing the release out to the first week of April. The Marlin user base gets a big bonus if we do that. All of the Delta users will have a 'Stable Release' with UBL in it working for them. And I'll have the Grid Based leveling in it along with the 'Smart Fill' command.

So we've decided not to implement variable grid size for UBL…?

Actually... That is almost done. It is a fall out of @Bob-the-Kuhn 's grid code he gave me to start with. You can specify the probe matrix except it is with a J number and not a P number. All of the P numbers are used for the various 'Phases' you go through to bring up a UBL Mesh Leveling System. You will be able to tilt your mesh based on a 2x2 to 9x9 grid. That grid is fed into an incremental Least Squares Fit algorithm that @edwilliams16 designed to keep memory usage to a minimum. We don't even have to hold onto the full data set to do a 9x9 LSF probing of the bed!

That gets merged in at the same time the Smart Fill gets merged in and that is already developed, debugged AND tested.

[Roxy-3D]

@edwilliams16 Your algorithm is 'Real'. It is in the main line code base now!!!

THANK YOU FOR THE HELP!!!!

[thinkyhead]

if we wait an extra week

That's not going to happen. I already have everything lined up and have been announcing Sunday's release in various high-visibility locations. Please hold any new work on UBL until 1.1.1 and keep focusing on basic spit and polish for tomorrow's release.