Re: [Paparazzi-devel] Source code analysis of paparazzi

[Karoly Molnar]

I agree to Felix. 

@Matthieu possibly you should share the chosen compiler settings that you use so we can review them. Bear in mind this is not a 64bits PC so int and const numbers manifest in 32 bits not in 64 bits.

Regards

Karoly

---

Date: Fri, 28 Nov 2014 21:35:10 +0100
From: address@hidden
To: address@hidden
CC: address@hidden
Subject: Re: [Paparazzi-devel] Source code analysis of paparazzi

Hi Matthieu,

I can't reproduce the issue.

Seems like your parser adds the (long) casts as that is nowhere in our code:
#define INT32_RAD_OF_DEG(_deg) (int32_t)(((int64_t)(_deg) * 14964008)/857374503)

So this looks good to me as you should be able to pass up to 61637 degrees to that function...
Maximum you can pass before intermediate multiplication overflows:
pow(2,63) / 14964008 = 616370429423 in deg*1e7= 61637 in deg

Cheers, Felix

On Wed, Nov 26, 2014 at 4:53 PM, Matthieu Lemerre <address@hidden> wrote:

Hi Felix,

Thanks for the list of configurations. I have began the analysis of the
Quad_LisaM_2 target, and I found the following bug: in ins_int.c, the
ins_init function does:

llh_nav0.lat = INT32_RAD_OF_DEG(NAV_LAT0);

The macro expansion of which gives (If I configured things correctly):

llh_nav0.lat = (int)(((long)435641194 * (long)14964008) / (long)857374503);

The intermediate calculation (long) 435641194 * (long)14964008 causes an
overflow, and may not be compiled correctly.

A work-around is to produce this code instead:

int j = ((435641194ULL * 14964008ULL) / 857374503ULL);

The intermediary computation fits in 64bit, and this solve the problem
here. However, I don't know a general solution to this problem if you
had a constant _expression_ which required an intermediate computation
that does not fit in 64bit, I would be interested to know if you could
find one.

Cheers,
Matthieu

Felix Ruess <address@hidden> writes:

> Hi Matthieu,
>
> If you want only two examples, I would suggest the following aircrafts from
> the example conf:
> - rotorcraft, stm32f1: Quad_LisaM_2
> - fixedwing, lpc21: Microjet
>
> Of course these only represent a small subset of Paprazzi...
>
> Cheers, Felix
>
> On Fri, Nov 21, 2014 at 4:48 PM, Matthieu Lemerre <address@hidden>
> wrote:
>
>>
>> Hi Michal,
>>
>> Thanks. Actually, a single-threaded mainloop is simpler to analyze, so
>> I'll stick with that for the moment.
>>
>> However, I would still be interested if you could give me the name of 2
>> configurations that are different in both dimensions {fixedwing,
>> rotorcraft} and {lpc2148,stm32}.
>>
>> Thanks,
>> Matthieu
>>
>>
>> Michal Podhradsky <address@hidden> writes:
>>
>> > Hi Matthieu,
>> >
>> > I am really glad that you are analysing Paparazzi code. There is an
>> > experimental branch based on RTOS ChibiOS, current version can be found
>> > here: https://github.com/paparazzi/paparazzi/tree/rt_paparazzi
>> > It  contains working configuration for Rotorcraft & Lisa M/MX. There are
>> > around 13 threads running for various subsystems, plus mutexes etc. so
>> you
>> > might find it interesting to analyse multi-threaded version of Paparazzi.
>> >
>> > Regards
>> > Michal
>> >
>> > On Thu, Nov 6, 2014 at 1:59 AM, Alexandre Bustico <
>> > address@hidden> wrote:
>> >
>> >> Le 06/11/2014 10:38, Matthieu Lemerre a écrit :
>> >>
>> >>> it seems that the generated source code is single-threaded, but
>> >>> are there other threads?
>> >>>
>> >>
>> >> In the master branch, there  is single-threaded mainloop, but many
>> >> callbacks are interrupt driven,
>> >> from hardware timer, and all the peripherals (i2c, spi, dma, uart, etc
>> >> etc).
>> >>
>> >> there is also in another branch an experimental rtos (chibios) version
>> >> which use threads.
>> >>
>> >> --
>> >> Alexandre
>> >>
>> >>
>> >>
>> >> _______________________________________________
>> >> Paparazzi-devel mailing list
>> >> address@hidden
>> >> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>> >>
>> > _______________________________________________
>> > Paparazzi-devel mailing list
>> > address@hidden
>> > https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>
>> _______________________________________________
>> Paparazzi-devel mailing list
>> address@hidden
>> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel

_______________________________________________ Paparazzi-devel mailing list address@hidden https://lists.nongnu.org/mailman/listinfo/paparazzi-devel