Digital filter and assembler code examples

[pythoncoder]

It would be great to have this implemented. I gather subroutine call works as follows:

Code: Select all

	bl(foo) # jump to subroutine, return address goes into r14
	...
	label(foo)
	add(r0, r0, r0) # do something
	mov(pc, r14) # return

[Damien]

That's a neat idea: calling a subroutine within the actual function using a label. Are you requesting this feature, or proper subroutine calls to external entities?

[pythoncoder]

This feature would be fine for my purposes.

[Damien]

Bx and bl now implemented. Along with ite and all its variations. Examples in the tests/inlineaam dir.

[pythoncoder]

Excellent. Can I suggest that the assembler tutorial document includes a link to the relevant section of the wiki as this is where much of the relevant detail is located?

I think you've implemented a very usable subset. One thought for the future is implementing interrupt service routines in assembler. This may be a matter of figuring out how to do it and documenting it rather than changing MicroPython. I wonder if anyone has achieved this already?

[Damien]

Registering an assembler function as a callback should work just fine. Eg timer.callback(asmfunc). It will have a small bit of overhead setting up the call.

I can link to the wiki, but even better might be to put the wiki stuff in the docs, since the docs are lacking with respect to the inline assembler.

[PieMan]

Reading about "Accepting arguments" here: https://micropython.org/doc/tut-asm and here: http://wiki.micropython.org/platforms/b ... /assembler showed me how to write an ISR/callback in asm. The following code example works for me and I measure the call overhead to be about 3.5us. Calling a "normal" function which in turn calls an asm functions incurs a 9us call overhead (by my measurements).

Code: Select all

from pyb import Timer, Pin
import stm

# Waggle pin A2 (aka X3) on then off.
# A timer callback receives the timer obj as a parameter which is not used here.
@micropython.asm_thumb
def asm_pin_waggle(r0):
    movwt(r1, stm.GPIOA) # Port A
    movw(r2, 1<<2)       # A2
    strh(r2, [r1, stm.GPIO_BSRRL]) # Set high...
    strh(r2, [r1, stm.GPIO_BSRRH]) # then set low.

# Configure pin X3 (aka A2)
PinX3 = Pin(Pin.board.X3, Pin.OUT_PP)

# Setup a 100us 50:50 PWM on pin X2.
tim2 = Timer(2,prescaler=83,period=99,mode=Timer.DOWN)
ch2 = tim2.channel(2,Timer.PWM,pin=Pin('X2'))
ch2.pulse_width_percent(50)

# Have an ISR written in assembler called.
ch2.callback(asm_pin_waggle)

[pythoncoder]

Very neat and much simpler than I'd imagined! This might amuse:

I'd been puzzling over how to enable an assembler function to handle more than three arrays, given the limit of three register arguments. The solution was obviously to use indirection but I couldn't see how. It proved to be trivial:

Code: Select all

# Return the address of an array (to put in another array)
@micropython.asm_thumb
def getaddress(r0):
    nop()

This allows you to populate an array with the addresses of other arrays. I've put this in the Wiki in case anyone else encounters this issue.