Page 2

Question 1

• full credit for mentioning condition codes
• half credit for nop
• none for incorrect statements

Question 2

• full credit for nop
• none for incorrect statements

Page 3

Questions 3–7

• Full credit for getting the right answer.
• Partial for one of
  • having one extra modification
  • missing the modification
  • having the wrong value in the correct register
  • having the wrong register with the correct value
• No credit if have tow or more of above problems

extraneous %, 0x, etc OK.

Correct answers:

• Q3
  • RAX = 190
• Q4
  • RSP = 1F8 or 192 or 198 (full for any of the three: hex conversion was not the point of this question)
• Q5
  • RSP = 208
• Q6
  • RDI = 120
• Q7
  • RDX = 1C3F5678 (also accept 21C3F5678)

Page 4

> It became clear that we either failed to teach the error detection topics adequately or there was some unnoticed ambiguity in the questions about it; as a consequence, we added 0.75 to every score on each of the questions on this topic to help correct our failure, whichever one it was.

Question 8

• full credit for D
• half credit for C
• no credit for the others

Question 9

• full for none
• half for T or L

Question 10

• full for T

Question 11

• full for P
• Also accept “none” if accompanied by explanation about not being wrong in comments, strings, etc

Question 12

• full for L
• half for T

Question 13

• full for 7

Page 5

Question 14

• full for 20
• half for 8

Question 15

• full for 16
• half for 14

Question 16

The bug is the malloc is too small; it should have a sizeof(int) term. This creates a second bug: we could go off the end of the array (heap buffer overflow) both in the copying loop and potentially in every other access to s.data[index].

The leak is fixed by adding a free(s.data) after the for and before the s.data = tmp;.

Graded in four parts:

• Q16 found bug
  • full if identify either (a) the malloc or (b) every s.data[index] (there are three of them)
  • partial for identifying just one s.data[index] or the memory leak
• Q16 explained bug
  • full if description suggests they understand why what they identified is wrong
  • partial if description unclear (e.g., “missing sizeof” without why)
• Q16 fixed leak
  • full if free(s.data)
  • partial if free(tmp) – note there is no non-buggy way to do this, so they miss the next points too
• Q16 leak fix not buggy
  • full if free in correct place
  • partial if free before or in loop or after s.data = tmp

Page 6

Question 17

The set of correct solutions includes

do {
    y += 1;
    if (x&1) {
        x *= 3;
        x += 1;
    } else {
        x >>= 1;
    }
} while (x > 1);

There are other solutions; for example,

y += 1; x = (x%2==0) ? x/2 : x*3+1;
while (x>1) { y += 1; x = (x%2==0) ? x/2 : x*3+1; }

… as well as things like for(;;) with an internal break, etc.

Don’t be picky on ;, etc., unless expressly called out below.

Graded in three parts:

• Q17 goto-free C
  • full if
    • no goto or labels
    • no loop-breaking ; like(...);{ or do{ ... }; while
    • no major (more than delimiter) mistakes
  • partial if just one of the above mistakes
  • none if two or more of the above
• Q17 if case
  • full if
    • there is an if/else (or two complimentary if, etc)
    • the x-is-odd case has the *3+1 and the even has the >>1 (or equivalent)
  • partial if missing/always does one case or has the two backwards
• Q17 do-loop semantics
  • full if both
    • is a loop
    • does the body once before first check
    • terminates on the right conditions
  • partial if a loop with one of the other two cases
  • none if not a loop, or both check-first and wrong-condition