Mathematics

[nailuy]

Hi,
I use PIC16F886 with 3 digits on 7 seg for output display and for input keys multiplexed 10 in total from 0-9.
I want to know why my number made with little math is not working well.
I found solution for helping in math but in my opinion is improvisation.

[guy]

Hi!
Here are some ideas:
1. use your debugger or simulator and step through the code to see at what point the code is not working well.
2.

[nailuy]

dear Guy
you are right
VAR=(int16)TST[0]*100; is solution. Compiler is working like is int8 because variable is declared as int8.
may be if I can declare an variable like int4 I use because that is necesay for me.
Now is working without "improvisation"
Thank you GUY
have nice day.

[Ttelmah]

You can make a 4bit 'component part' using bitfields, but remember that the
chip fundamentally works in bytes, so two would have to be declared together
in a byte, and accessing these will be less efficient than accessing the byte
itself.

The C 'rule' on types, is that when any maths operation is used, the maths
'type' will be that of the highest type in the operation. So:

int8*int8 -> int8 maths
int8*int16 -> int16 maths
int16*int16 -> int16 maths.
int32*int8 -> int32 maths

etc. etc..

Some compilers/chips will explicitly support overflows. So on the PC for
example, if the maths is performed using the maths co-processor it'll
automatically be using a larger type always for the maths.

You could have made the operation work, with simply:

VAR=TST[0]*100L;

The 'L' here says that this '100' constant, is a 'long' (int16) constant, not
an int8.
Explicitly casting up is the other way to go.

[nailuy]

yes Ttelmah.
your solution:
VAR=TST[0]*100L;
is also working and is using less RAM than:
VAR=(int16)TST[0]*100;

Thank you both for your little help.
For me both solutions are good and I understand why the compiler and the processor think so.

[Ttelmah]

Yes. The reason it saves a little memory, is that the 100L, is stored as a long constant in the code, so doesn't have to be converted up to be 16bit. With the cast, both values have to be converted at runtime. Saves a little time and RAM.