optimisation: calculate steps per unit reciprocal at compile time

[gudnimg]

Using the reciprocal allows us to use multiplication instead of division at runtime.

This commit removes 7 divison operations at run time.

Change in memory:
Flash: -210 bytes
SRAM: 0 bytes

[github-actions]

All values in bytes. Δ Delta to base

ΔFlash	ΔSRAM	Used Flash	Used SRAM	Free Flash	Free SRAM
-210	0	27922	1667	750	893

[DRracer]

Interesting, even though you are introducing new variables, the resulting code is significantly shorter than before, good job. Please have a look at the unit tests failing .... oh, it's just the comment 🤦

Run mshick/[email protected]
Error: Resource not accessible by integration

[gudnimg]

Please have a look at the unit tests failing .... oh, it's just the comment 🤦

Yup it is causing all PRs to fail. Only someone with admin access to the repo can fix it. We had the same problem on Prusa-Firmware repo before.

[DRracer]

@3d-gussner has admin rights 😉

[gudnimg]

@3d-gussner it looks like some of the unit tests are timing out on this PR. Not sure if its random, but I don't see it in the other PRs. I haven't looked into if the unit tests need updating.

[3d-gussner]

All motion tests fail. Also testing with M404 results in IDLER CANNOT HOME
Before PR 107 of 107 tests passed.
After PR 63 of 107 tests passed.

[gudnimg]

I think I see what the problem is... its with (T)axisScale[AU::axis].stepsPerUnitReciprocal. Where T is int32_t. And since axisScale[AU::axis].stepsPerUnitReciprocal is less than 1 casting this into integer gives just 0.