Unsigned 32bx16b multiply subroutine

HydraHacker

Hello,
I thought that I would post a short 32bx16b unsigned multiply subroutine that is only 7 instructions long, not counting the return, which runs in 14 clocks.
Please note that the multiplicand is destroyed upon subroutines exit.

HydraHacker

Unsgnd32x16Mul
			
			getword		pl, mc, #1 				' get high word of mc into pl
			mul		pl, mp					' multiply high word of pl with low word of the mp
			getword		ph, pl, #1				' shift product in pl and ph left by 16 bits
			shl		pl, #16						
			mul		mc, mp					' multiply low word of mc by the low word of the mp
			add		pl, mc		wc			' add mc into pl
			addx		ph, #0					' add carry to ph
			ret	

mc			res	1						' multiplacand 
mp			res	1						' multiplier
pl			res	1						' product low
ph			res	1						' product high

Peter Jakacki

Don't forget that you can get a free return ticket by using _ret_ on the addx

evanh

Here's what I got it down to:

mul32x16
		getword	temp1, pa, #1
		mul	temp1, pb		'pah * pbl
		mul	pa, pb			'pal * pbl
		shl	temp1, #16
	_ret_	add	pa, temp1

More examples - https://forums.parallax.com/discussion/169698/substantially-faster-shorter-multiply

evanh

Small enough to be inlined.

HydraHacker

Evanh,
My routine returns a 48 bit result

evanh

Probably should add it to the other topic.

Can the Cordic use that for a QDIV/QFRAC? I think it might.

HydraHacker

Evanh, I probably should move it to the other topic. I like your smaller routine if you only need a 32 bit result.

Hydrahacker