201811xx foredrag c_cpp

embedded Linux med C/C++
from a manager's
perspective
21/11-2018

Stefan Pielmeier

Content
• C

• C++

• emb. Linux

Why talk about C
embedded engineer CV:
• 1997: University of Erlangen/Nürnberg M.Sc.E.E.

• -2003: E/// 3G Basestation TI C6x DSP + System Des.

• -2003: TI Nice, 2.5G Project Leader/Senior Dev., TI DSP &
2G Chipset

• -2007: TI Aalborg 3G Protocol Stack

• -2010: Telenor Aalborg, Data Core

• -2018: Thrane & Thrane, SAILOR Radio

Why talk about C
embedded engineer CV:
• 1997: University of Erlangen/Nürnberg M.Sc.E.E.

• -2003: E/// 3G Basestation TI C6x DSP + System Des.

• -2003: TI Nice, 2.5G Project Leader/Senior Dev., TI DSP &
2G Chipset

• -2007: TI Aalborg 3G Protocol Stack

• -2010: Telenor Aalborg, Data Core

• -2018: Thrane & Thrane, SAILOR Radio
no C/C++

What do we expect from
Software Programming Lan
robust whine instead of c
extendible
self-explaining, safe to
ambiguous
reusable
motivating to make goo
test them
low production cost
ﬂexible, attract mo
workforce, easy to u
documented (google)
easy to integrate wit
short lead time
embedded: save ressources
nifty switches allowing
real time ressource

What do we expect from
Software Programming Language C
robust whine instead of crash
extendible
self-explaining, safe to use, non-
ambiguous
reusable
motivating to make good APIs and

test them
low production cost
ﬂexible, attract motivated
workforce, easy to use, well
documented (google) libraries,
easy to integrate with CM/CI
short lead time
embedded: save ressources
nifty switches allowing to tweak
real time ressource need

Flexibility
• C/C++:

• really ﬂexible

• fragile?

• not so easy to understand

• Python:

• loads of standard components

• a bit bulky sometimes

• easy to reuse

• easy to build & document

Why are we using C in
embedded systems?
• 20 years ago: embedded processor compilers: ASS, C

• Unix/Linux is based on C: C libraries already there =>
"cheaper" to use C for applications as well

• BOM: control of memory & ressources => cheap HW

• optimization working good

• Only in the last 5-10 years, processors allow for other
languages, but BOM price taken into account, C is often
the winner

The sharp knive
• if used correctly:

• more precise/direct, you see what you will get in
RAM

• more secure, actually, because you know exactly
where you cut

• eﬀective (just handle the address of a pointer directly)

• requires

• training (cut yourself with aid at hand) and

• experience

The dangers of plain old C
• direct manipulation of adress-pointers

• no mechanism to control if dynamically allocated
memory is used for multiple purposes at the same
time

• no mechanism to detect memory overwriting, out of
bounds exceptions, etc. (like in Java)

• memory leaks: no garbage collection

• no exception handling (though in C++)

• does not encourage encapsulation as such, programs
end up to be card houses/birds nests

How to detect?
• Team not able to show an actual design with clear APIs
and encapsulated data/responsibilities in the code

• Team not able to estimate remaining work eﬀort

• Team needs to ﬁx everything before next integration step

• Defect resolution of a single issue leads to new
regressions
0
25
50
75
100
April May June July

Mitigation
• Code reviews & best practices

• reuse working code, refactor bad code

• modularization, keep ﬁles small

• make it easy to access data through API

• document your code (doxygen)

• use purify and unit test, test driven design

• encourage the use of object orientation C++ where possible, as it
enhances the use of proofen design patterns

C -> C++
• encourages to major reuse (inheritance & polymorphism)

• encourages encapsulation

• very good libraries (boost, STL, mm.)

• possible to use C++ without dyn. mem. allocation

• ARDUINO on Atmega 328P (8 bit, 32kFlash, 2kB RAM)

• very motivating for ambitious programmers

motivation
• many problems solved in Open Source: SW and HW

• network, WIFI, Bluetooth, webserver, secure shell, SSL, ...

• acceptable tools for free (and the code to improve them)

• Graphic Frameworks for MMI

• no license fee

• "standard" interprocess komm. (IPC)

• very robust kernel with good documentation ressources

emb. Linux Platform
• rPi, DKK219 (dk-rsonline.com), 1 Stk.(!)

• 1GB RAM, micro SD ﬂash, Arm A53 quadcore, 64-bit,
WIFI, 40 GPIO, Lyd, H264, MPEG-4, OpenGL

emb. Linux SW Arch
source: https://bootlin.com/pub/conferences/2011/limoges-clermont/presentation.pdf
opensource

challenges
• Linux = multitasking, memory managed, cached ...

• realtime predictability is limited (requires extensions)

• 10us on an Cortex A8 is possible, but is challenging

• 1ms is more realistic

• memory management (how much RAM is free?)

• multiprocess/thread/driver debugging is not trivial

• upgrade of kernel/glibc/gcc baseline to maintain platform

After 10 years leading 10-18
using C
• good to have 1-2 really innovative architects, Code Review!

• be careful with arrays, sizeof(), pointer arithm. in general

• allow them to use OO with C++ was the best thing to happen

• use Beaglebone, Raspberry Pi, Arduino as reference design

• Embedded Linux where you need network or existing OpenSource, time to
market

• Arduino Development platform encourages object orientation with C++,
high productivity due to libraries, very good documentation: even 9-th class
pupils can handle that.

• don't forget to follow-up and plan for change, CI and maturation :-)

Albatros-Tech IVS
• se www.albatros-tech.eu

• hjælper rigtigt gerne med embedded udvikling

• har erfaring med digital kommunikationsteknologi og
systemer

• yder gerne project hjælp/management

• tilbyder specielle modeller for startups

• SW process hjælp/management/troubleshooting

201811xx foredrag c_cpp

More Related Content

Similar to 201811xx foredrag c_cpp (20)

More from InfinIT - Innovationsnetværket for it (20)

Recently uploaded (20)

201811xx foredrag c_cpp