double type

[Jerry Messina]

Thanks for adding support for 64-bit floats!

Are there plans for adding/changing the functions in math and convert to support the new type too?

[David John Barker]

Time permitting I would like to do that. For the time being though, you can pass a double to the math routines but note the parameter and result will be scaled to 32 bit single. If anyone can help out in this regard I would be very grateful

[Jerry Messina]

Is there a way to add command-line switches when building?

In order to get a 64-bit double with XC16, I had to add '-fno-short-double' to the GCC build

dim d as double = 1.23450

with -fno-short-double:
1.23450000000000
3FF3C083126E978D

without (FW/XC16 default):
1.234500050544739
3ff3c08320000000

If anyone can help out in this regard I would be very grateful

Sure, if I can figure out what the various functions in math.bas do, that is.
Most of them are probable fine using floats.

[Jerry Messina]

Also, it looks like you could squeak out a few more digits when generating the C source

Code: Select all

dim d as double

d = 1.2345678901234560      // 0x3FF3C0CA428C59F8
d = 1.2345678901234562      // 0x3FF3C0CA428C59F9
d = 1.2345678901234564      // 0x3FF3C0CA428C59FA


generated the following (rounding the last few digits)

Code: Select all

   // d = 1.2345678901234560      // 0x3FF3C0CA428C59F8
   asm volatile("; RD_I000001_F000_000045_P000127");
   F0_F64.val = 1.23456789012346;
   // d = 1.2345678901234562      // 0x3FF3C0CA428C59F9
   asm volatile("; RD_I000002_F000_000046_P000127");
   F0_F64.val = 1.23456789012346;
   // d = 1.2345678901234564      // 0x3FF3C0CA428C59FA
   asm volatile("; RD_I000003_F000_000047_P000127");
   F0_F64.val = 1.23456789012346;


[David John Barker]

I've uploaded a new BETA (B4) which should fix both of the above. Let me know how it goes. Note that the cstr() function will not have the resolution needed to display the floats accuracy but you should see the correct values in MPLAB. The float to string routine used by CStr() can be found in:

..library\sys.imports\SysProcs.bas
__strtofp()

[David John Barker]

Just a heads up, but I will be adding a "math" module in the next update for PIC16 and PIC32 that supports double. If you need it before then, just drop me an email...