COL729 HW 3 : Scalar evolution

Due date - 23:55, 14 Oct 2026

Weightage: 20 Marks (Actual weightage, We will decide later)

We have discussed scalar evolution in class. Consider the following C program:

int sum(int *a, int n)
{
  int s = 0;
  for (int i = 0; i < n; i++) {
    for (int j = 0; j < 4; j++) {
      s += a[4*i + j];
    }
  }
  return s;
}

Convert this program to (slightly optimized) LLVM using the following command:

clang -O1 -S -emit-llvm -Xclang -disable-llvm-passes t.c -o - | opt -passes="mem2reg,loop-rotate" -S -o t.ll

The -emit-llvm flag instructs clang to emit LLVM IR; the mem2reg optimization pass converts local variables to pseudo-registers; the loop-rotate optimization pass implements the loop inversion (also called loop rotation) transformation) discussed in class.

Use the following command to print the information related to the Scalar Evolution (SCEV) analysis:

opt -passes='print<scalar-evolution>' -disable-output t.ll

Interpret the printed scalar evolution results, and briefly explain them in a short report. Submit the report.