CCS C double variable

[ilker07]

Hello friends, in CCS C double data=data/5.0; I saw that this process takes approximately 65 us. I use the system at 64 mhz with PLL. This 65 us is too much for the work I do. How can I reduce this time?

[PrinceNai]

It might help if you tell what you are doing. Maybe it can be done with integers. Integer division by 5 takes about 25us.

[ilker07]

[PrinceNai]

data = 0.2*data;

This cuts time down a lot.

[ilker07]

[PrinceNai]

If you only have those two divisors, use an if - else to check which of the two it is and then multiply by the correct reciprocal value.

[ilker07]

[Ttelmah]

Division is always the slowest basic arithmetic operation.
As has already been pointed out multiplication by the reciprocal is faster.

Some further comments:

First what makes you think you must use 'double'???
Do you understand just how unnecessary in general this is?.

Comment, unless you are Using a 16bit PIC, it will not actually support
a 'double'. The standard PIC maths library does not have a double
type. It'll 'accept' it, but does not have the functions to support it.
So your code is probably just using float.
No physical process actually 'needs' this. Using it is pretty much always
a sign that you do not understand the costs of this extra 'accuracy',
and how pointless this is. For 99.999% of applications this is a waste. No
measurement you can make actually 'needs' this accuracy. Read up
about precision and accuracy, and understand that this is a sign you
probably do not understand what you are really dealing with. You can
get a better result, and save a lot of processing time by using a scaled
integer representation.
Arithmetic is slow. Basic fact of life.

You are not using a double. Simple fact if you are using a 64MHz processor
you are certainly using a PIC18. This does not support 'double'. So you
are using a 32bit float. Even this is taking 75uSec for this division. Using
a PIC24, the division will go down to about 40uSec.

[ilker07]

[Ttelmah]

Unless you are using PCD, you are not using a 'double'.
What PIC are you using???

[ilker07]

[Ttelmah]

So, you are not using a double.

Read the manual:

"double - A reserved word but is not a supported data type."

Your chip _does not_ support a 'double'.

You must have changed something about how you were using the value
when you switched to using double instead of float, which made it work.

Can't you simply reciprocate your division factor when you read it in, and
then the divisions become multiplications throughout. Multiplication is
about twice as fast as division.

[ilker07]

[quote="Ttelmah"]So, you are not using a double.

Read the manual:

"double - A reserved word but is not a supported data type."

Your chip _does not_ support a 'double'.

You must have changed something about how you were using the value
when you switched to using double instead of float, which made it work.



I didn't do any change.I just wrote double.

I just wrote simple code

unsigned int16 i=0;
for(i=0;i<30000;i++)
{
total += i;
}

when I specified total variable as float ,it gave me wrong value but when I specified as double , it gave me right value.

[ilker07]

[PrinceNai]

My version of compiler swallows double and it treats it as a float. The result of the calculation is the same in both cases, almost 450 millions. I checked the result here:

https://miniwebtool.com/sum-of-positive-integers-calculator/?n1=1&n2=29999

and it is correct for the code sample provided. It adds all the consecutive numbers from 1 to 29.999 (mind the meaning of the dot here )