q uantitative e valuation of e mbedded s ystems
DESCRIPTION
Q uantitative E valuation of E mbedded S ystems. QUESTION DURING CLASS? Email : [email protected]. FAIL!. Thank you, Robin Wolffensperger en Ruben Lubben!. Exercise: Model a car manufacturing line. Consider a car manufacturing line consisting of. Four assembly robots: A,B,C and D - PowerPoint PPT PresentationTRANSCRIPT
Quantitative Evaluation of Embedded Systems
QUESTION DURING CLASS?Email : [email protected]
FAIL!
Thank you, Robin Wolffensperger en Ruben Lubben!
Consider a car manufacturing line consisting of...
Exercise:Model a car manufacturing line
• Four assembly robots: A,B,C and D• A production unit that needs 20 minutes to produce a chassis• A production unit that needs 10 minutes to produce a steering installation• A production unit that needs 10 minutes to produce a breaking system• A production unit that needs 20 minutes to produce a body• Three painting units that each need 30 minutes to paint a body• A production unit that needs 15 minutes to produce a radio• Robot A compiles the chassis and the steering installation in 4 min. and sends it to B• Robot B adds the breaking system in 3 min. and sends it to C• Robot C adds a painted body in 5 min. and sends it to D• Robot D adds a radio in 1 min. and sends the car out of the factory• For safety reasons, there can be at most 3 ‘cars’ between A and C, and only 2 between B and D• Every robot can only deal with one of each of the assembled components at a time
Exercise: calculate the first 3 firings of each actor
A B C D
10min
20min 10min 20min
30min20min
5min3min4min
15min
1min
Disclaimer: no actual car assembly line was studied in order to make this model.
Answer: Model a car manufacturing line
Simulate a few firings assuming sufficient input tokens.
Determine the (max,+) matrix.
Determine the max. throughput.Determine a periodic schedule for:1) µ = MCM2) µ = 2*MCM3) µ = 3*MCM4) as a function of µ
EXERCISE:
Keep your answers for next time!
Quantitative Evaluation of Embedded Systems
Recall the characteristic equations…
u(n)Dmax(n)xCy(n)u(n)Bmax(n)xA1)(nx
or for autonomous systems…
(n)xCy(n)(n)xA1)(nx
What about this one?
0 ms
y
(n)x(0)y(n)(n)x(0)1)(nx
0y(n)
Cycles with a 0 execution time cause livelocks
But when logging events, this is mathematically okay...
And this one ?
u(n)Dmax(n)xCy(n)u(n)Bmax(n)xA1)(nx
A B
C D
1ms 2ms
4ms
u y
3ms
Theorem: The number of tokens on any cycle is constant!
Therefore, every cycle must contain at least one token,otherwise a deadlock occurs.
And this one?
A B
C D
1ms 2ms
4ms
u
x3
yx1
x2
3ms u(n)max(n)x2y(n)
u(n)1
1
max(n)x
2
5810
5810
1)(nx
Reducing rows…
A B
C D
1ms 2ms
4ms
u
x3
yx1
x2
3ms
u(n)max(n)x2y(n)
u(n)1
max(n)x2
5101)(nx
...but only when assuming: x1(1) = x2(1)which is ok for self-timed execution,but not when reasoning aboutperiodic schedules
What about reducing columns?
A B
C1ms 2ms
u x3 y
x1
x2
3ms u(n)max(n)x22y(n)
u(n)
1
4max(n)x
1
545
22
1)(nx
And back to (max,+) algebra…
302
01
11
230
121
301
20
ΙAAAA
0
n
n
times