Faster DIG algorithm

[Tabsoft]

I confirmed you observations when I reran the routines @ 16MHz and PBPL.

I did notice if you use PBP's "DIG" command, the number of instruction cycles used are not consistent for various 4 digit numbers.

Here is the data I collected.

Algorithm 1:
Pressure = 1234
'Start Stopwatch in MPLAB Simulator
tmpDig0 = Pressure dig 0
tmpDig1 = Pressure dig 1
tmpDig2 = Pressure dig 2
tmpDig3 = Pressure dig 3
'Stop Stopwatch in MPLAB Simulator

Algorithm 2:
Pressure = 1234
'Start Stopwatch in MPLAB Simulator
tmpDig0 = Pressure // 10
tmpDig1 = Pressure / 10
tmpDig2 = Pressure / 100
tmpDig3 = Pressure / 1000

tmpDig1 = tmpDig1 // 10 
tmpDig2 = tmpDig2 // 10
tmpDig3 = tmpDig3 // 10
'Stop Stopwatch in MPLAB Simulator

Algorithm 3:
Pressure = 1234
tmpDig0  = Pressure // 10 
pressure = Pressure / 10 
tmpDig1  = Pressure // 10
pressure = Pressure / 10
tmpDig2  = Pressure // 10
tmpDig3  = Pressure / 10
'Stop Stopwatch in MPLAB Simulator


PBPW @ 16MHz
Algorithm   Inst Cycles     Time
1           2958            739.50 us
2           2086            521.50 us
3           1783            445.75 us

PBPL @ 16MHz
Algorithm   Inst Cycles     Time
1           9626            2.406500 ms
2           6830            1.707500 ms
3           5839            1.459750 ms

Obviously PBPL slows down the routines considerably.

Good luck.

[pedja089]

Thanks.
I'll post current consumption when I get PCB.

[ruijc]

Hi guys,

This is an interesting topic.

I was wondering if the following line ( i dont know how fast it runs ) can also run faster with the technic shown above:

LCDOUT $FE,LINE2,"FREQ ",DEC DISPFREQ DIG 2,DEC DISPFREQ DIG 1,",",DEC DISPFREQ DIG 0, " KHz    "

Best regards
Rui

[Tabsoft]

Yes, it will reduce the number of instruction cycles in your example.
Given that you are only extracting 3 digits of your DISPFREQ variable, Algorithm #2 will be faster.

Assuming you are using PBPW not PBPL here are the instruction cycles results.

Using the DIG method: 7606 Inst Cycles
Using Algorithm #2: 7293 Inst Cycles

That is a savings of 313 instruction cycles. If you are running a 4Mhz Fosc/4 MCU then that would be a 313us savings.
Though it is a small savings on cycles and time, it does come at a cost of 3 extra byte variables to hold the values.

Here are the two code options tested.

tmpDig0 var byte
tmpDig1 var byte
tmpDig2 var byte
DISPFREQ var byte
LINE2 con $c0


DISPFREQ = 1234

'Start Stopwatch in MPLAB Simulator
LCDOUT $FE,LINE2,"FREQ ",DEC DISPFREQ DIG 2,DEC DISPFREQ DIG 1,",",DEC DISPFREQ DIG 0, " KHz    "
'Stop Stopwatch in MPLAB Simulator
'Output: FREQ 23,4 KHz

'Start Stopwatch in MPLAB Simulator
tmpDig0 = DISPFREQ // 10
tmpDig1 =DISPFREQ / 10
tmpDig1 = tmpDig1 // 10
tmpDig2 = DISPFREQ / 100
tmpDig2 = tmpDig2 // 10

LCDOUT $FE,LINE2,"FREQ ",DEC tmpDig2,DEC tmpDig1,",",DEC tmpDig0, " KHz    "
'Stop Stopwatch in MPLAB Simulator 
'Output: FREQ 23,4 KHz

[ruijc]

Greetings Tabsoft,

Thank you for the information and tip.
Very helpfull !

Regards
Rui