division operation with float variable mplabx v6.00 xc16 (v2.00)

RE: division operation with float variable mplabx v6.00 xc16 (v2.00)

jc2048 — Mon, 20 May 2024 09:56:51 GMT

Can the value returned by the read encoder routine be zero? If so, you're going to hit the problem of dealing with a divide by zero error at some point when doing the floating point reciprocal.

It's possible that that might be what's going on here. The watch variables going to zero hints at a reset, which might be coming from the floating routines throwing an exception (it's a long time since I worked with PICs, and I never used the float stuff much for anything, anyway - they waste too much time in an embedded system where there's a lot going on - so I can't tell you directly).

Even if it isn't the problem here, it's still a good idea to anticipate it and deal with it in your code.

You could a) ensure the read encoder never returns zero; b) test for zero just before doing the division and substitute whatever answer you want under those circumstances; or c) understand properly what happens after the float division by zero and deal with it there, either dealing with the trap, if there is one, or handling a NaN or 'infinity' value handed back, if that's what the floating routines do.

The last may not be very portable - I don't think floating point libraries are very consistent in how they deal with the outcome of a divide by zero.

RE: division operation with float variable mplabx v6.00 xc16 (v2.00)

genebren — Sat, 18 May 2024 19:33:36 GMT

You need to declare Freq_PWM1_F outside of the function. If it is inside a function (as it is now) it is declared on the stack frame, which is not guaranteed to be available after the function call. I am a little confused to how this was allowed to compile without an error.

RE: division operation with float variable mplabx v6.00 xc16 (v2.00)

genebren — Sat, 18 May 2024 17:08:50 GMT

You might want to try:

float Freq_PWM1_F = 1.0/(float)Freq_PWM1;

RE: division operation with float variable mplabx v6.00 xc16 (v2.00)

javagoza — Sat, 18 May 2024 16:12:24 GMT

I don't understand what you are asking. It seems to me that you are comparing two different statements.

float Freq_PWM1_F = (float)1/1000;

This is a constant expression, so the compiler can optimize it.

In the other case:

float Freq_PWM1_F = (float)1/Freq_PWM1;

the computed value for Freq_PWM1_F cannot be shown because Freq_PWM1 has not been evaluated yet.