18f45k22 union problem: union not working exactly !!!

CYembedded · Joined: 31 Jan 2015 Posts: 6

Hi, I need a help about union usage for converting hex to floating point with the communication of GPRS modem.
I think it is simple but my code didn't show correct output.
In Visual Studio 2010 the code worked fine but in CCS the float value is always 0.0. What do you think about the problem.

ckielstra · Joined: 18 Mar 2004 Posts: 3680 Location: The Netherlands

One big difference between CCS and Visual Studio is how a float is represented internally. Just a simple search on the internet found me the following link: https://www.ccsinfo.com/forum/viewtopic.php?p=52346

Ttelmah · Joined: 11 Mar 2010 Posts: 19535

and, as an 'add on' to this, since when that post was made, CCS have added similar functions, to those supplied with the compiler. Look at the file 'ieefloat.c'.

The 'reason' relates to the hardware/history. Before IEEE float came out, there were several different float formats. Soon after the PIC was launched, Microchip published an arithmetic library 'optimised' for the instructions available on the PIC's as they then existed (earliest PIC14/16's). These laid the data out this way to save a few instructions on several of the operations.

When CCS launched, they used the work that had already been done.

Then IEEE float came along.

Now the optimisations were still worthwhile, so the format was retained. The 'origins' in Microchip's work, is why the format is sometimes also called Microchip float32 format. With the latest DsPIC chips, extra instructions, mean that there is no advantage, so PCD, uses IEEE float format.