Compiler optimizations: Within the compilation of a PL/I program, what does “machine-independent optimization” refer to?

Introduction / Context:
Optimization in compilers occurs both before and after target-specific decisions are made. Machine-independent (target-agnostic) optimizations operate on the intermediate representation (IR) without relying on particular instruction sets, focusing on program semantics to improve efficiency.

Given Data / Assumptions:

• The compiler has produced an IR after parsing and semantic analysis.
• We have not yet committed to a specific target CPU's instruction set.
• Transformations must preserve program meaning.

Concept / Approach:
Machine-independent optimizations include constant folding, dead-code elimination, common subexpression elimination, copy propagation, strength reduction (algebraic), loop-invariant code motion, and simple loop unrolling. These improve the IR regardless of the final machine. Later, machine-dependent optimizations tune register allocation, instruction scheduling, and addressing modes for a specific architecture.

Step-by-Step Solution:

Verification / Alternative check:
Modern compilers (e.g., LLVM, GCC) maintain extensive mid-level, machine-independent optimization passes that apply across architectures.

Why Other Options Are Wrong:
Reductions / syntactic constructs: parsing, not optimization. Uniform symbols: lexical analysis. Macro processor to produce better assembly: macro expansion is orthogonal and often target-specific.

Common Pitfalls:
Confusing mid-level IR optimizations with backend (machine-dependent) register allocation and scheduling.

Final Answer:
creation of more optimal intermediate representation independent of target machine.