benchmark:
  vm_eval: |
    eval("1")
loop_count: 6000000