32 bit square root

[ScaleRobotics]

I used LCDOUT instead of your serial routine though.

[Charles Linquis]

I'll try timing it Monday. I eventually got everything to work using the SQR routine in PBPL. I gained some extra time by not using the ADCIN command. Instead, I start a channel, then do the computation for the previous channel. Not having to wait on the A/D bought me enough time to get everything done. I'm doing true RMS on 4 channels simultaneously using an '8723 at 40MHz.

[ScaleRobotics]

I'll try timing it Monday.

Thanks Charles,

No worries, I was just curious if pbpl was any slower compared to using pbp with the square root assembly code. Since pbpl uses a bit more space, I was curious to know if it was faster, or slower. It appears that a 20mhz chip can do about 5 or 6 of these in 1 ms using the assembly include file.

If anyone is interested, to make things easier, I have attached it as an include file. It can only be used on PIC18 chips, and according to TB040, must be modified for use with PIC17 devices that have a hardware multiplier. But if you did not have a new version of pbp (that had pbpl included), this would allow you to perform 32 bit square root. And it is much smaller than compiling in pbpl.

To use, load argh with the upper 16 bits, and argl with the lower 16 bits, then call square. Result will be in word variable RES.

INCLUDE "square.pbp"

'some defines here
main:
'and a little bit of code there....

ARGH = $0001       'load upper 16 bits into argument (any value you want)
ARGL = $ffff       'load lower 16 bits into argument (any value you want)
call square        'call square assembly function
lcdout $FE,1,#RES  'print result to lcd

Here are the results from codetimer.bas:
Time: 84.66328 usec
OSC Freq: 48 Mhz

[ScaleRobotics]

Well, I was wrong. PBPL won my crude speed comparison, and was 1.4 times faster when compared to the above TB040 assembly code. I had thought that Microchip's TB040 code would be pretty optimized, but that is pretty impressive MeLabs! Maybe I can stop being afraid to us pbpl now!

[Bruce]

You would flip if you knew some of the big manufacturers we've sold PBP to that have
produced multi-million dollar gadgets with it.

It's a lot more solid than some folks give it credit for.