ECE 468/573 — Midterm 1

October 1, 2014

Name: ______________________________________

Purdue email: ______________________________________

Please sign the following:

I affirm that the answers given on this test are mine and mine alone. I did not receive
help from any person or material (other than those explicitly allowed).

X___________________________________________

Note: ECE 468 students do not have to complete Part 6.

Part Points Score

1 10

2 18

3 14

4 24

5 34

6 15

Total 115
Part 1: Short answers (10 points)

1) In recent years, many companies have switched from GCC-based compiler

toolchains to LLVM-based compiler toolchains even though LLVM re-used
GCC’s front ends. Given that, what reason might a company have had for
switching to LLVM? (5 points)

LLVM may have a better optimizing back end — even if the same front-end is used,
LLVM can do a better job of producing optimized code

Some people thought that LLVM was a true virtual machine with a JIT compiler; I gave
credit for reasonable answers along those lines.

2) One nice thing about maintaining a common ISA is that when a new processor
is released, you don’t have to recompile any code, or change the compiler.
Nevertheless, when Intel releases new x86 processors, companies update their
compilers anyway. Why might they want to do that? (5 points)

Even if the ISA is the same, the implementation of that ISA might be different—some
instructions may be more efficient—so it can be important to change the backend to
account for these differences. Also, newer x86 processors may support instruction set
extensions (SIMD, x86-64) and compiler backends need to be changed to generate
code targeting those extensions.
Part 2: Regular expressions, finite automata and scanners (18 points)

1) Consider the following NFA. Fill in the transition table below with its
corresponding DFA using the subset construction. You may not use all of the
rows (16 points):

2 c 4 a 6
a a c
1 b
b b
a

3 c 5 c 7

State Final? a b c
1 No 2, 3 X X

2, 3 No X 3 4, 5

3 No X X 5

4, 5 No 3, 6 4 7

5 No X 4 7

3, 6 Yes X X 5

4 No 3, 6 4 7

7 Yes X X X

2 pts/row

2) List which states (if any) should be merged when you reduce the DFA you just
derived (2 points):

{4, 5} and 4
Note that 3 and {3, 6} can’t be merged because one is final and the other isn’t.
Part 3: Grammars (14 points)

Let G be the grammar: S ! A$

A ! (AB)
A!
B ! (A)
B!x
Using this grammar, answer the following questions.

1) What are the terminals and non-terminals of this grammar? (6 points)

Terminals: (, ), x, $
Non-terminals: S, A, B

1 point per mistake (note that λ is not a terminal!)

2) Draw the parse tree for the string “((x)x)$” (8 points)

A B $

( A B ) x

λ x
Part 4: LL parsers (24 points)

Answer the questions in this part using the same grammar from Part 3.

1) Give the First sets for each non-terminal in the grammar (6 points)

First(S) : {(, $} [3 points]

First(A) : {(, λ} [2 points]
First(B) : {(, x} [1 point]

2) Give the Follow sets for each non-terminal in the grammar (6 points)

Follow(S) : { } [1 point]
Follow(A) : {x, (, $, )} [4 points]
Follow(B) : { ) } [1 point

3) Give the Predict sets for each production in the grammar (5 points)

Predict(1) : {(, $}
Predict(2) : { ( }
Predict(3) : { x, (, $, ) }
Predict(4) : { ( }
Predict(5) : { x }

4) Fill in the following parse table (6 points)

( ) x $

S 1 1

A 2, 3 3 3 3

B 4 5

5) Is the grammar LL(1)? Why or why not? (1 points)

No, there is a predict conflict when predicting an A with a lookahead of ‘(‘

Part 5: LR(0) Parsers (34 points)

Use the following grammar for the next questions: S ! A$

A ! (AB)
A ! ()
B ! (A)
B ! ()
1) Fill in the missing information for the for the following CFSM (15 points)

State 0 A State 1 $ State 2

S → • A$ S →A • $ S → A$ •
A → • (AB)
A → • () (
State 4 A State 5

State 3 ) A → ( • AB) A → (A • B)
A→(•) B → • (A)
A → () • ( A → • (AB) B → • ()
A → • ()
(
B
State 6
B → () • (
State 8
)
State 7 A → (AB • )
State 9 B → ( • A)
A )
B→(•)
B → (A • )
A → • (AB)
) A → • () State 10
A → (AB) •
State 11
B → (A) •

2) List the shift and reduce states in the above CFSM (6 points)

Shift: 0, 1, 4, 5, 7, 8, 9
Reduce: 2 (accept), 3, 6, 10, 11

3) Is this an LR(0) grammar? Why or why not? (1 point)

Yes. No S/R or R/R conflicts
4) For the following sub-questions, use the CFSM you built in the previous
question. Each question will provide the state of the parser in mid-parse, giving
the state stack (most recent state on the right) and the next token. For each
question, give the action the parser will take next, using the format “Shift X” for
shift actions (where X is the state being shifted to) and “Reduce R, goto X” for
reduce actions (where R is the rule being reduced, and X is the state the parser
winds up in after finishing the reduction). Also provide the new state stack. (3
points each)

a) State stack: 0 4 5 7. Next token: (

Shift 4
04574

b) State stack: 0 4 4 5 8 10. Next token: (

Reduce 2 goto 5
045

c) State stack: 0 4 4 5. Next token: )

Parse error

d) State stack: 0 4 5 7 4. Next token: )

Shift 3
045743
Part 6 (573 only): LR(1) Parsers (15 points):

1) For the following grammar, show what the first state of the LR(1) machine
would be (10 points):

S ! AB$
A ! xy
S → • AB$, { }
A ! Bz A → • xy, {w, $}
A → • Bz, {w, $}
B ! Cz B → • Cz, {z}
B! B → λ •, {z}
C → • w, {z}
C!w

2) Give the action table entries for the first state (i.e., for each token that could be
coming up, say whether the parser would shift or reduce). For reduce actions, say
what rule would be reduced. (5 points)

Lookahead Action
x Shift
y Error
z Reduce 5
w Shift
$ Error