Code Generation Questions and Answers

Answer: A

Increment and decrement instructions, multiply and accumulate, and address computation instructions are typical examples. Recognising them is part of instruction selection. A machine description lists the idioms the generator should look for.

Answer: B

Correct answer: correct but longer and slower than necessary

Machine idioms - an increment instruction, an add with shift, a multiply-accumulate - do in one instruction what would otherwise take several. Ignoring them costs performance and code size but not correctness, since the longer sequence computes the same result.

Answer: C

Everything that depends on the machine lives in the back end. The front end depends on the source language and should not need to change at all. That separation is precisely what an intermediate representation exists to provide.

Answer: D

Machine description languages let a new target be supported by writing a specification rather than code. The approach trades some code quality for much faster retargeting. A hand written back end still wins where the last few per cent matter.

Answer: A

Correct answer: needs no register allocation, since operands are on a stack

A stack machine takes its operands from the top of the stack implicitly, so there are no registers to allocate and code generation reduces to a postorder walk of the expression tree. That simplicity is why stack code is used for portable intermediate forms - and why it is slower than register code on real hardware.

Answer: B

Each leaf becomes a push and each interior node an operator that consumes the values pushed by its children. The result is exactly reverse Polish order. No register allocation is required, which is what makes the model so simple.

Answer: C

Explicit registers avoid the traffic that a stack discipline forces. Stack code is more compact, which is why it remains attractive for a portable distribution format. Just in time compilers translate such code into register code before running it.

Answer: D

Correct answer: find the storage location and type of each name

The generator needs to know where each name lives - a register, a frame offset, a static address - and what type it has, since the type determines which instructions apply. Both come from the symbol table, which is why it survives long after the parse.

Answer: A

Correct answer: a conversion instruction inserted where the semantic analyser required

The semantic analyser has already decided where a conversion is needed and recorded it in the tree; the generator’s task is to emit the machine’s conversion instruction at that point. Treating the two representations as interchangeable would produce nonsense, since their bit patterns mean quite different things.

Answer: B

Correct answer: checking that every identifier has been declared

Declaration checking belongs to semantic analysis and is complete long before code generation begins - the generator assumes a correct program. Instruction selection, register allocation and scheduling are its three tasks, and all three are about how to express a program already known to be valid.