search

hzeller / beagleg

G-code interpreter and stepmotor controller for crazy fast coordinated moves of up to 8 steppers. Uses the Programmable Realtime Unit (PRU) of the Beaglebone.
http://beagleg.org/
GNU General Public License v3.0
122 stars 51 forks source link

Update absolute position #22

Closed lromor closed 7 years ago

hzeller commented 7 years ago

Thanks, that looks pretty good already. I have some concerns regarding unnecessary computation in the critical phases when we enqueue things in the history list, so made some suggestions there.

Essentially: if we always record in the history what at the end of that the positions should look like, we can make the whole UpdateAbsolutePosition() implementation a lot cheaper, and we only ever would need to call it when we call GetMotorLoops() (which should be called GetMotorPosition()), not at enqueue time.

lromor commented 7 years ago

wow. Nice solution

lromor commented 7 years ago

Now the code should be simpler, shorter and faster. Thanks. (Next time, if you want, don't tell me the solution =P)

p.s. could you add to .gitignore simple-lexer_test?

hzeller commented 7 years ago

Looks much nicer. There are still a couple of smallish comments left, but I think we're mostly ready to merge.

hzeller commented 7 years ago

I now added all *_test in .gitignore

hzeller commented 7 years ago

So I tried this patch on my machine, but it starts hanging when it does a G28 and reaches the endstop. I have not dug into details yet why this is happening. (this happens on the z-axis with home-pos = max in case that is relevant). Can you have a look ?

BTW, I changed little things so that it passes my test with various compilers and valgrind. Attached a patch relative to your change (and I had to name it .txt instead of .patch as github then complains that it can't deal with that file type. D'oh).

valgrind-proof.txt

lromor commented 7 years ago

it seems the subtract thing is making problems, from what I saw with the debugger there's a no wait loop, causing him do underflow.. or something like that, I will work on it this evening

lromor commented 7 years ago

So, for X and Y works right?

hzeller commented 7 years ago

I don't know, the Z axis is the first that was homed when it got stuck :)

lromor commented 7 years ago

Ok, now should be fixed :)

hzeller commented 7 years ago

Alright, seems to work. Pushed as submit https://github.com/hzeller/beagleg/commit/03950be17eec2490909d67048824e2cf50d7ed55 (with only minor changes to make it compile with various compilers and with uninitalized memory fixed as found by make valgrind-test).

Thanks!