|
Hi, anyone have a good PIC17/PIC18 (mulwf) signed or unsigned 24by16 multiply routine? I've tried some of those in AN617 but they are flawed. -- ******************************************* VISIT MY HOME PAGE: <http://home.online.no/~eikarlse/index.htm> LAST UPDATED: 23/08/2003 ******************************************* Best Regards Eirik Karlsen |
|
|
Re: mul 24by16
Started by ●December 28, 2003
Reply by ●December 28, 20032003-12-28
|
I've used a few of the routines from AN617 with great success though
I must admit that I've not needed to do 24x16. Just curious what flaws you ran into. --Scott --- In , Eirik Karlsen <eikarlse@o...> wrote: > Hi, > anyone have a good PIC17/PIC18 (mulwf) signed or unsigned 24by16 multiply routine? > I've tried some of those in AN617 but they are flawed. > -- > ******************************************* > VISIT MY HOME PAGE: > <http://home.online.no/~eikarlse/index.htm> > LAST UPDATED: 23/08/2003 > ******************************************* > Best Regards > Eirik Karlsen |
Reply by ●December 29, 20032003-12-29
|
Scott,
The routines I tried was the FXM2416S and FXM2416U PIC17CXXX MULTIPLY ROUTINES Revision: 2.2 from AN617. I believe I tried some of the 1616 ones too....no joy. I used these with some minor alterations on a PIC18. FXM2416S: essentially produces just garbage. The FXM2416U works better,
but the upper byte in the product is garbage....
Apart from all this I don't like these routines because input/output
is not 'clean'....
Last night I started rolling my own configurable Signed/Unsigned mul2416
and got it half finished, should be ready for use later tonight.
Good math routines are not easy to find...many of those in the PICLIST
math archive is also flawed.
I've used a few of the routines from AN617 with great success though
|
Reply by ●December 30, 20032003-12-30
|
The routine is now finished and tested, it is attached and FOC to anyone who
wants it. It does a signed 24by16 multiplication in about 99 clocks and an unsigned one in about 79 clocks. About 20 clocks could have been saved by changing some of the MOVFF to MOVF, but then it will be banked. As it is now it runs unbanked. ;************************************************************************************************* ; PIC18 ; 24x16 Bit Signed or Unsigned Fixed Point Multiply 24 x 16 -> 40 ,by Eirik Karlsen 29.12.03 ; ; Input:24 bit signed or unsigned fixed point multiplicand in AARGB2, AARGB1, AARGB0 ; 16 bit signed or unsigned fixed point multiplier in BARGB1, BARGB0 ; ; Use: CALL MUL2416SU ; ; Output: 40 bit signed or unsigned fixed point product in AARGB4, AARGB3, AARGB2, AARGB1, AARGB0 ; ; Input and output is right justified and ARGB0 is LSB ; ; AARGB4,AARGB3,AARGB2,AARGB1,AARGB0,BARGB1,BARGB0,TEMP3,TEMP2,TEMP1,TEMP0,SIGNMODE ; ; Result: AARG <-- AARG * BARG ; ; Timing: 99 clks ; ; MUL2416SU MOVFF AARGB0,TEMP0 ;Store AARGB, each byte is moved to TEMP3 for processing. MOVFF AARGB1,TEMP1 MOVFF AARGB2,TEMP2 CLRF AARGB3 ;Clear high order bytes CLRF AARGB4 ; SETF SIGNMODE ;Set Signed operation mode ;CLRF SIGNMODE ;Clear SIGNMODE here and the routine will operate unsigned. MOVFF BARGB0,WREG ;Multiply B0*A MULWF AARGB0 ;B0*A0 MOVFF PRODH,AARGB1 ; MOVFF PRODL,AARGB0 ;->A1,A0 MOVFF TEMP1,TEMP3 ;Get A1 MULWF TEMP3 ;B0*A1 MOVFF PRODL,WREG ; ADDWF AARGB1 ;Add to partial sum MOVFF PRODH,AARGB2 ; MOVFF BARGB0,WREG ; MOVFF TEMP2,TEMP3 ;A2 MULWF TEMP3 ;B0*A2 MOVFF PRODL,WREG ; ADDWFC AARGB2 ;Propagate carry from previous addition MOVFF PRODH,WREG ; ADDWFC AARGB3 ;Done B0*A ->A3,A0 MOVFF BARGB1,WREG ;Multiply B1*A MOVFF TEMP0,TEMP3 ;Get A0 MULWF TEMP3 ;B1*A0 MOVFF PRODL,WREG ; ADDWF AARGB1 ;B1 additions are shifted 1 byte to MSB MOVFF PRODH,WREG ; ADDWFC AARGB2 ; CLRF WREG ; ADDWFC AARGB3 ;Propagate carry from previous additions ADDWFC AARGB4 ;Done A+256*(B1*A0)->A4,A0 MOVFF BARGB1,WREG ; MOVFF TEMP1,TEMP3 ;Get A1 MULWF TEMP3 ;B1*A1 MOVFF PRODL,WREG ; ADDWF AARGB2 ; MOVFF PRODH,WREG ; ADDWFC AARGB3 ; CLRF WREG ; ADDWFC AARGB4 ;Done B1*A1 MOVFF BARGB1,WREG ; MOVFF TEMP2,TEMP3 ;Get A2 MULWF TEMP3 ;B1*A2 MOVFF PRODL,WREG ; ADDWFC AARGB3 ; MOVFF PRODH,WREG ; ADDWFC AARGB4 ;Done B1*A2 BTFSS SIGNMODE,0 ;Operatin in signed mode? RETURN ;No! BTFSS BARGB1,7 ;Yes! B negative? GOTO TSIGN2416 ;No! MOVFF TEMP0,WREG ; SUBWF AARGB2 ; MOVFF TEMP1,WREG ; SUBWFB AARGB3 ; MOVFF TEMP2,WREG ; SUBWFB AARGB4 ; TSIGN2416 BTFSS TEMP2,7 ;A negative? RETURN ;No! MOVFF BARGB0,WREG ; SUBWF AARGB3 ; MOVFF BARGB1,WREG ; SUBWFB AARGB4 ; RETURN ; ;************************************************************************************************* -- ******************************************* VISIT MY HOME PAGE: <http://home.online.no/~eikarlse/index.htm> LAST UPDATED: 23/08/2003 ******************************************* Best Regards Eirik Karlsen | |||
|
