Always print iterations, then elapsed time, then ips for benchmark-interface backends

Author: eregon

bench/benchmark-interface/lib/benchmark-interface/backends/simple.rb

@@ -61,8 +61,8 @@ def self.run(benchmark_set, names, options)
               # can print an ips and adjust for the next round to try to make
               # it take a second.
               ips = iterations / round_time
-              puts get_time - start_time if elapsed
               puts iterations if print_iterations
+              puts get_time - start_time if elapsed
               puts ips * inner_iterations
               iterations = (ips * freq).to_i
             end

bench/benchmark-interface/lib/benchmark-interface/backends/stable.rb

@@ -30,8 +30,8 @@ def self.run(benchmark_set, names, options)
             round_time = Time.now - start_round_time
 
             ips = 1 / round_time
-            puts Time.now - start_time if elapsed
             puts 1 if print_iterations
+            puts Time.now - start_time if elapsed
             puts ips * inner_iterations
           end
         end

mx.truffleruby/mx_truffleruby_benchmark.py

@@ -287,7 +287,7 @@ def runBenchmark(self, benchmark, bmSuiteArgs):
             lines = out.data.split('\n')[1:-1]
 
             data = self.filterLines(lines)
-            elapsed = [d for n, d in enumerate(data) if n % 3 == 0]
+            elapsed = [d for n, d in enumerate(data) if n % 3 == 1]
             samples = [d for n, d in enumerate(data) if n % 3 == 2]
 
             if lines[-1] == 'optimised away':