Embercode: AI-powered software engineer course

1. The Warm program in the lesson prints batch 1 at about 6312 us, and the later batches at about 4400 us. Why is batch 1 slower?

The results array is allocated on the first callThe JVM runs the method in interpreted mode before the JIT compiles itThe CPU clock is throttled at the start of every programThe garbage collector runs once before batch 1

2. A C++ benchmark built with -O2, where the result of work(n) is not used, prints 0.000 ms for every run. What does this mean?

The loop is so fast that the timer rounds it down to zeroThe -O2 flag makes integer arithmetic freeThe clock is broken on this machine and must be replacedThe compiler removed the loop because nothing read its result, so the timer measured nothing

3. The first measured batch of a Java benchmark shows the steady speed of the code, so it is the number to report.

True, because the first batch is the only one that runs with a fresh clockFalse, because batch 1 includes the time the JVM spends interpreting and compiling, so the steady speed comes from later batches

4. Timing one run of each version is enough to compare them, because the same code gives nearly the same time on every run.

False, because the same code took 17 ms in one run and 20 ms in another run of the same launch, so one run can differ by 10 to 20 percentTrue, because the operating system pauses background tasks during a timed run

5. The timing code below has a bug. Which one is it? ``

java
long start = System.nanoTime();
int[] data = new int[n];
for (int i = 0; i < n; i++) {
    data[i] = i % 100;
}
long total = sumArray(data);
long elapsedMs = (System.nanoTime() - start) / 1_000_000;

`

sumArray` is the method from the lesson.

System.nanoTime() returns seconds, so dividing by 1,000,000 gives the wrong unitThe int[] array cannot be timed, so it must be a List<Integer>The clock starts before the array is filled, so the cost of filling it is counted as part of the timed workThe loop counter should start at n and count down to zero

Benchmarks