[Question][Hw6, 7] Format of input programs, expected result in hw7

[p51lee]

Hello,

I have two three questions about hw6 and hw7:

1. Hw6
   1. In hw6, do we have to handle boolean values applied to the sink function, like sink ( x < y )?
2. Hw7
   1. Do all functions have only one parameter?
   2. When checking example1.ll, what my implementation prints is not identical to example1.expected:

   -|ID 11: (= (type_env \|  %w = alloca i32, align 4\|) (ptr i32))|
   -|ID 12: (= (type_env \|  %v = alloca i8, align 1\|) (ptr i8))|
   -|ID 27: (= (ptr (type_env \|  %div = sdiv i32 %2, %3\|))
   -   (type_env \|  %w = alloca i32, align 4\|))|
   -|ID 28: (= (ptr (type_env \|  %div = sdiv i32 %2, %3\|))
   -   (type_env \|  %v = alloca i8, align 1\|))|
   +|ID 13: (= (ptr i32) (type_env \|  %w = alloca i32, align 4\|))|
   +|ID 14: (= (ptr i8) (type_env \|  %v = alloca i8, align 1\|))|
   +|ID 41: (= (type_env \|  %w = alloca i32, align 4\|)
   +   (ptr (type_env \|  %div = sdiv i32 %2, %3\|)))|
   +|ID 43: (= (type_env \|  %v = alloca i8, align 1\|)

   Is it important to equalize IDs or the order of lhs and rhs for each equation between implementation result and example1.expected?

Thank you.

[goodtaeeun]

Hi,

1. Short answer, yes. Longer answer, even if the c code may look like sink ( x < y ), in the LLVM IR, x<y is pre-caculated then a single variable is passed to the sink function. In other words, sink ( x < y ) is no different from sink(x) or sink(y).

2-1. Yes, in this homework, you may assume that functions only have one parameter.
2-2. No, the satisfiability is what matters. So you don't have to match the IDs and other details.

[p51lee]

Thank you for the answer.
Question 1 -> If so, I think we should also consider instructions for converting boolean(i1) values to integer(i32) in LLVM IR.

[goodtaeeun]

Hi,

There was some misunderstanding when I replied to your question.
Programs that contain sink(x<y) are not within the scope of this assignment.
You may ignore any LLVM instructions that are not mentioned in Section 3 of the README.

Handling boolean values passed to sink requires handling LLVM instructions that should be considered nop. Thus, you don't need to support it.

Sorry for the confusion I may have caused with the previous reply.

[YoungseokCh]

Hello, I have a follow-up question for 2-1. In HW7 documentation, there is an example function type void -> i32*. However, since C does not support void type argument, I thought it was the type for the function without argument. Can we still assume that functions always have one parameter? @goodtaeeun

[goodtaeeun]

Hi,
I believe what you meant by Although C does not support void-type argument, is that using an argument within the function body that has a void type is impossible. This is true. However, we can design the analyzer to handle the functions with no arguments as if they are functions that have a single void-type argument. Example function type void -> i32* is regarding such case.

[jirheee]

Hi,

1. Short answer, yes. Longer answer, even if the c code may look like sink ( x < y ), in the LLVM IR, x<y is pre-caculated then a single variable is passed to the sink function. In other words, sink ( x < y ) is no different from sink(x) or sink(y).

2-1. Yes, in this homework, you may assume that functions only have one parameter. 2-2. No, the satisfiability is what matters. So you don't have to match the IDs and other details.

So failing CI due to ID difference doesn't matter I guess..?

[goodtaeeun]

Yes, ID difference does not matter