Error when using Microchip 16 compiler

[David John Barker]

Hi Jerry

Thanks for taking the time to look at this.

> I tried to see if maybe there was some sort of optimization change that might "fix" this

Firewing16 (not Firewing32) uses -O0, so it would be a bit cheeky of Microchip if they have turned down the optimisation for the "free" version of xv16 even further. Certainly if I was a "C" coder I would not be happy with the changes in code size between the two versions. Even worse though is why 1.11 works but code generated by 1.20 can fail. I've not got to the bottom of it and like you, I don't have the time or energy to take it further.

[Jerry Messina]

Supposedly optimization level -O1 is available in the free mode, and I seem to recall adding that dropped the code down to ~17K or so.

I didn't really look at the output, so I've no clue if it actually worked or not. I know debugging can be a pain once you enable anything past -O0.

[Jerry Messina]

FWIW (if anything), there's a new version of XC16 V1.21

v1.21

This release contains fixes for some major reported issues and part-support file updates. Additionally the compiler version that a part support is released with is now commented in the support files. This will faciliate future part-support only releases.

[David John Barker]

Thanks for the heads up Jerry. Here is what I'm now seeing. Assuming "progLCDPlus.bas" with "SetRTC()" enabled, as previously discussed

XC16 1.11 20,835 Bytes [Program works as expected]
XC16 1.20 26,346 Bytes [Program *does not* run correctly]
XC16 1.21 24,534 Bytes [Program works as expected]

Clearly 1.20 is a turkey. 1.21 seems to work OK, but the program is still significantly larger that 1.11...

[Jerry Messina]

Well, at least they're going in the right direction!

V1.20 wouldn't even generate correct code for one of their own demos...

XC16-598 Explorer 16 Demo Code is broken with the v1.20 (LCD display is affected), worked fine with v1.11