Perhaps it's me that's missing something.

You have an 8 bit value (0-255), you want a range of "(about) 0..34,000 or
so".
Assuming 32,768 is close enough to "34,000 or so", why not just << 7  ?

Or is this a case of "close enough" isn't close enough ?  :-)

-Denny




On Wed, May 15, 2019 at 4:45 PM Dwayne Reid <dwayner@planet.eon.net> wrote:

> Good day to all.
>
> I'm working on a project where I need to scale an 8-bit value to the
> range of (about) 0..34,000 or so.  This is to feed the period
> register of a PWM generator - actual largest value is not
> critical.  This routine needs to execute very millisecond and system
> clock frequency is 4 MHz (1us instruction rate), so I'm opting to go
> with a simple integer-based fixed multiply routine.
>
> As always, I turn to Nikolai Golovchenko's wonderful ConstDivMul
> calculator whenever I need to scale some quantity by a fixed-value
> constant.  It consistently generates code that is at least as tight
> and small as anything that I can create myself.
>
> This time, however, the routine I came up with seems to be much
> shorter / faster than Nilolai's version.  I'm assuming that I somehow
> made a mistake but I'm not seeing it.  I'm hoping that some of the
> math wizzes still on the list can give an opinion.
>
> Input quantity is a single byte (0..255) and is to multiplied by a
> fixed constant of 132
>
> ; ACC =3D ACC * 132.000000
> ;
> ; ALGORITHM:
> ; Clear accumulator
> ; Add input * 128 to accumulator
> ; Add input * 4 to accumulator
> ; Move accumulator to result
> ;
> ; Error in constant approximation : 0.000000, %
>
>
> ; Input:  ACC0  (8 bits)
> ; Output: ACC0 .. ACC1  (16 bits)
>
>         cblock
>         ACC0
>         ACC1
>         TEMP0
>         TEMP1
>         endc
>
>         clrc
>         rlf     ACC0, f
>         clrf    ACC1
>         rlf     ACC1, f
>         rlf     ACC0, f
>         rlf     ACC1, f
>         movf    ACC0, w
>         movwf   TEMP0
>         movf    ACC1, w
>         movwf   TEMP1
>         swapf   TEMP1, f
>         swapf   TEMP0, w
>         andlw   15
>         iorwf   TEMP1, f
>         swapf   TEMP0, w
>         andlw   -16
>         movwf   TEMP0
>         clrc
>         rlf     TEMP0, f
>         rlf     TEMP1, f
>         movf    TEMP0, w
>         addwf   ACC0, f
>         movf    TEMP1, w
>         skpnc
>         incfsz  TEMP1, w
>         addwf   ACC1, f
>
>
> My version:
>
> ; Clear accumulator
> ; Add input * 128 to accumulator
> ; Add input * 4 to accumulator
> ; Move accumulator to result
>
>         clrf    TEMP0
>         clrf    ACC1
>         clrc                    ;* 128
>         rrf     ACC0,W          ;b0 -> C
>         movwf   TEMP1           ;
>         rrf     TEMP0,F         ;C -> b7
>
>         clrc                    ;* 4
>         rlf     ACC0,F
>         rlf     ACC1,F
>         rlf     ACC0,F
>         rlf     ACC1,F
>
>         movfw   TEMP0           ;ACC1,0 + TEMP1,0 -> ACC1,0
>         addwf   ACC0,F
>         movfw   TEMP1
>         skpnc
>           incfsz TEMP1,W
>             addwf ACC1,F
>
>
> Question: did I make some silly mistake or is my version reasonable?
>
> Anyone have any other optimizations to offer?
>
> Many thanks!
>
> dwayne
>
> --
> Dwayne Reid   <dwayner@planet.eon.net>
> Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
> 780-489-3199 voice   780-487-6397 fax   888-489-3199 Toll Free
> www.trinity-electronics.com
> Custom Electronics Design and Manufacturing
>
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
--=20
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.