Function benchmark

test/compute-engine/compile-performance.test.ts:27–34

(fn: () => any, iterations: number)

 *
 * The execution-speed comparisons further down (compiled runner vs
 * interpreter) are still timed, because there is no counter that can stand in
 * for "the generated code runs faster than the interpreter". They compare two
 * measurements taken back to back in the same process, so load affects both
 * sides, and their observed margins are one to two orders of magnitude.
 */

describe('COMPILATION PERFORMANCE', () => {
  const verbose =
    process.env.COMPILE_PERF_VERBOSE === '1' ||
    process.env.BENCH_VERBOSE === '1' ||
    process.env.VERBOSE === '1';