[for]

NO, it would not be "ideal" or even suitable for controlling heavy machinery like a lathe.

It may be OK for 3D printers where it does not really matter if something goes wrong. Although printers usually move quite fast and this is likely to mean lack of RT will lead to errors. Since home 3D printers produce pretty crude results anyway you may not notice some of them.

Please don't use it for anything bigger.

It's not my intention to start new vs. old flame war. I'm interesting in a fair critique of Nikolay approach using DMA controller. Given large enough circular buffer and dedicated high priority channel for DMA operations you could minimize negative effects of whatever controls CPU, RT kernel or not. I presume most promising results could be achieved using closed-loop drivers. Just to be clear: I'm talking here about general approach of using DMA controllers, not this very particular found in BCM2835/BCM2836, which could be inappropriate for some other reasons.