How can i make this routine smaller and efficientely

[The Puma]

void max7221_write_value( int32 num ) {
int8 i;

led_data[0]=(num%10);
led_data[1]=(num%100)/10;
led_data[2]=(num%1000)/100;
led_data[3]=(num%10000)/1000;
led_data[4]=(num%100000)/10000;
led_data[5]=(num%1000000)/100000;
led_data[6]=(num%10000000)/1000000;
led_data[7]=(num%100000000)/10000000;

// Do leading zero blanking
if (num<1) led_data[0] = SEGMENT_CHAR_BLANK;
if (num<10) led_data[1] = SEGMENT_CHAR_BLANK;
if (num<100) led_data[2] = SEGMENT_CHAR_BLANK;
if (num<1000) led_data[3] = SEGMENT_CHAR_BLANK;
if (num<10000) led_data[4] = SEGMENT_CHAR_BLANK;
if (num<100000) led_data[5] = SEGMENT_CHAR_BLANK;
if (num<1000000) led_data[6] = SEGMENT_CHAR_BLANK;
if (num<10000000) led_data[7] = SEGMENT_CHAR_BLANK;

for(i=DIGIT_0;i<=DIGIT_7;i++)
max7221_write(i,led_data[i-1]);
}

[Ttelmah]

Unfortunately, you seem to want the data in 'reverse order'. Otherwise, make LED_DATA one character larger, and use:

sprintf(led_data,"%8LU",num);

Best Wishes

[Falgellus]

Something like that:

int32 dv=10;

for (i=0; i<8 ; i++ )
{
led_data[i]=(num%(dv))/(dv/10);
if (num<(dv/10)) led_data[i] = SEGMENT_CHAR_BLANK;
dv *= 10;
}


Caution: I have not tested it so make all tests.

Bye

Paolo

[Ttemah]

Also, remember that the value%10, is really developed by performing value-((value/10*10), so you can save work by seperating this yourself.
Also if you run the result 'forward', the division is only /10 at each stage.
You should also remember that only seven such divisions are needed, and then the remainder is the last digit.
You seem to also blank the bottom digit. Normally this would not be done.

[The Puma]

Thanks i will try both codes

[Falgellus]

int32 dv=10;

for (i=0; i<8 ; i++ )
{
if (num<(dv/10))
led_data[i] = SEGMENT_CHAR_BLANK;
else
led_data[i]=(num%(dv))/(dv/10);

dv *= 10;
}


Caution: I have not tested it so make all tests.

Bye

Paolo

[Ttelmah]

I don't think this is 'more efficient'. The problem is that division, is a very slow operation (the slowest of the basic arithmetic operations). The test now involves a division, as does the sum itself. Basic 'tests' are quick (perhaps ten machine cycles for an int32), but divisions, can easily involve 1000 machine cycles. The '%' operation, actually involves the sum I have already outlined, and is even worse, involving a division, multiplication, and subtraction. So though this code is a lot smaller, in terms of machine operations , it will be worse.
This is why, though my code could do with a lot of tweaking I'm sure, one of the key things was to keep the divisions to a minimum (just one per loop).

Best Wishes

[The Puma]

I will use the code above for driving a SAA1064

My question is now how to implement this segment table in the below code
It displays now only segments, but not correct numbers (0,1,2,3,4,5,6,7,8,9)

// Segment table (In hex: d.p.=80,g=40,f=20,e=10,d=08,c=04,b=02,a=01)
byte const seg_table[10] = {
0x3F,0x06,0x5B,0x4F,0x66,0x6D,0x7D,0x07,0x7F,0x6F }; // 0,1,2,3,4,5,6,7,8,9

void saa1064_write_num(int8 addr,int8 control,int16 num) {
int1 leading=true;
int8 led_data[4];
int8 j;
int16 temp16;

for(j=0;j<3;j++) {
temp16=num/10;
led_data[j]=num-(temp16*10);
num=temp16;
}
led_data[3]=num;

for(j=3;j>0;j--) {
if (leading && led_data[j]==0)
led_data[j]=SEGMENT_CHAR_BLANK;
else
leading=false;
}

i2c_start();
i2c_write(addr);
i2c_write(SAA1064_REG_CONTROL);
i2c_write(control);

for(j=0;j<=3;j++)
i2c_write(led_data[j]);

i2c_stop();
}