64 bit math

[kuriken]

Hi,

I need to perform 32 bit multiplications and 64 bit addition and subtractions. I realize that CCS does not support int64. I also found the following discussion on this forum that provides a way to create a unsigned 64 type.

http://www.ccsinfo.com/forum/viewtopic.php?p=72037

Unfortunately I need a signed 64 (long long in ANSI C) to port over a code library created using ANSI C. Has anybody come across sample code on how I might be able to do this?

Thanks,


Ken

[temtronic]

I have to ask the obvious, why are you asking a poor little PIC to do that kind of math. Just seems 'wrong'.

[kuriken]

My previous project used a PIC18 and I'm just adding one more feature to it. It would be convenient for me if I could use the same platform. I'm contemplating jumping over to a ATMEL chip or using a PIC24 but I would like to avoid it if I can.

[SherpaDoug]

Are there any quantities in the universe that are known to 64 bit accuracy? That is 18 x 10^18, or roughly 10,000 x the number of grains of sand in the Hoover dam! What would a lowly micro controller do with such a number?

Try calculating how big the 32x32 bit multiplication could actually get, given realistic inputs. I bet you don't really need 64 bit math. I designed a product that measures ships in 24 bit millimeters. It can handle any ship ever designed.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[PICman]

[Douglas Kennedy]

Well, all operations involving integer math can be done in base 2 (binary)notation.
All that is required is storage space for the notation. If its 64 bits then then less than 100 bytes of storage would be needed to contain the binary notation and arithmetic operations when carried out to two to the power 64.
The issue is convenience and speed. CCS's compiler provides routines to manipulate this notation in 8bit 16bit and 32bit chunks ( using standard operators). Now a typical hand held calculator does this in 4 bit binary coded decimal and handles decimal notation to the number of digits its lcd can show (often 10 or more digits). It is true it is easier to handle the carry with a 32bit MCU register and it may take fewer instructions but an 8bit PIC with several hundred bytes of storage could perform 500 bit binary notation arithmetic. The choice is speed and convenience. As said in an earlier post with the exception of pure mathematics ( like in cryptography) there is no need for accuracy beyond the 12 digit decimal precision of that of quantum mechanics (our most accurate physical law todate)...2 to the power 64 or say 20 decimal digits is overkill for the real world we live in.

[asmboy]

i gotta say - there are times that integrated
24 bit math routines would be a MIGHTY Useful to me ;-))

i'm thinking here of a customer, project
i'm looking at, using the ADS 1225 with precision load cells
and 24 bit integrated 24 bit functions would be very very sweet
for now - its all about stuffing it into an int32