Passing floating point parameters issue

mattht42 — Wed, 11 May 2022 08:24:11 GMT

Hello,

I am currently working on device K70FN1M0M15 with MQX OS.

FPU : fpv4-sp-d16

I have a function which takes an enumerate and a float as parameters and the problem is that sometimes the function’s float parameter takes the value 0 instead of the given value.

Here is an example of the function call, then the source code of the function:

MRTimer_SetValue(mTimerRefBracketAlarm1, 3.0);

void MRTimer_SetValue(tTimersRef TimerRef, float setTimerValue) { uint16_t *pDestValue = NULL; uint16_t rawValue = 0; test = setTimerValue; if(setTimerValue == 0) test++; if ((TimerRef >= MRTIMER_10MS_FIRSTREF) && (TimerRef <= MRTIMER_10MS_LASTREF)) { rawValue = (uint16_t) (setTimerValue * 100.0); pDestValue = &timers10ms[TimerRef - MRTIMER_10MS_FIRSTREF]; } else if (TimerRef >= MRTIMER_100MS_FIRSTREF && TimerRef <= MRTIMER_100MS_LASTREF) { rawValue = (uint16_t) (setTimerValue * 10.0); pDestValue = &timers100ms[TimerRef - MRTIMER_100MS_FIRSTREF]; } else if (TimerRef >= MRTIMER_1000MS_FIRSTREF && TimerRef <= MRTIMER_1000MS_LASTREF) { rawValue = (uint16_t) (setTimerValue * 1.0); pDestValue = &timers1000ms[TimerRef - MRTIMER_1000MS_FIRSTREF]; } else { return; } if(rawValue >65530 || rawValue < 3 || pDestValue == NULL) { tmpdebug_ref = TimerRef; tmpdebug_valeur = rawValue; test++; } else { *pDestValue = rawValue; } return; }

As you can see, I give 3.0 as parameter but when the bug occurs « setTimerValue » in the function take the value 0 instead of 3.0.

The call of the function uses this assembler code:

577 MRTimer_SetValue(mTimerRefBracketAlarm2, 3);

0001fd08: ldr r3, [r7, #4]

0001fd0a: ldrh r3, [r3, #6]

0001fd0c: addw r3, r3, #287 ; 0x11f

0001fd10: uxth r3, r3

0001fd12: mov r0, r3

0001fd14: vmov.f32 s0, #8

0001fd18: bl 0x2bae8 <MRTimer_SetValue>

In my opinion, the value of the parameter is stored in the register s0 (red line).

Then when I enter in the function, I have this assembler code:

MRTimer_SetValue:

0002bae8: push {r7, lr}

0002baea: sub sp, #16

0002baec: add r7, sp, #0

0002baee: mov r3, r0

0002baf0: vstr s0, [r7]

0002baf4: strh r3, [r7, #6]

The instruction” 0002baf0: vstr s0, [r7]” is supposed to store the content of an extension register(S0) to memory (address given by r7).

When the bug occurs, after this instruction the value in the memory is 0.

For your information, it’s not possible to watch S0 in KDS 3.0.

Do you have any clue about this odd behavior ?

Re: Passing floating point parameters issue

DanielRuvalcaba — Tue, 17 May 2022 15:38:11 GMT

Hi,

I highly recommend you to switch to MCUXpresso IDE.

About your problem, is this happening every time you run your code?

Do you have problems with floating point numbers in general or is it just in this case in particular?

Best regards, Daniel.

Re: Passing floating point parameters issue

bobpaddock — Wed, 18 May 2022 17:30:18 GMT

In general Floating Point should be avoided, it is rarely needed in Embedded Code, you are seeing why with this code.

Comparison to zero has no guarantee of being true, see:

https://stackoverflow.com/questions/19837576/comparing-floating-point-number-to-zero#19841844

and the referenced C++ FAQ entry.

Also use 0.0 not 0, however it will not make a lot of difference due to the above.

Re: Passing floating point parameters issue

mattht42 — Tue, 24 May 2022 14:21:19 GMT

Hello,

Thanks for the answer, the problem occurs randomly when the code is running. I only noticed this problem with this function but I can imagine it can appear in other functions where I use float as parameter.

I'm on a travel work for 2 week, I will give you more information when i come back.

topic Re: Passing floating point parameters issue in Kinetis Microcontrollers

Passing floating point parameters issue

Re: Passing floating point parameters issue

Re: Passing floating point parameters issue

Re: Passing floating point parameters issue