Track performance stats (memory and time?)

[coilysiren]

[coilysiren]

I might be able to use docker runtime options to do meaningful comparisons across languages??? https://docs.docker.com/config/containers/resource_constraints/

[luser]

Just thought I'd drop a few links that you might find useful here! You can get the count of CPU instructions executed using perf stat on Linux: http://brendangregg.com/perf.html#CPUstatistics

$ perf stat ./hello
Hello world

 Performance counter stats for './hello':

              0.56 msec task-clock                #    0.632 CPUs utilized          
                 0      context-switches          #    0.000 K/sec                  
                 0      cpu-migrations            #    0.000 K/sec                  
                55      page-faults               #    0.099 M/sec                  
         1,081,405      cycles                    #    1.948 GHz                    
           727,622      stalled-cycles-frontend   #   67.28% frontend cycles idle   
           716,960      instructions              #    0.66  insn per cycle         
                                                  #    1.01  stalled cycles per insn
           148,252      branches                  #  267.062 M/sec                  
             7,975      branch-misses             #    5.38% of all branches        

       0.000878743 seconds time elapsed

       0.000930000 seconds user
       0.000000000 seconds sys

Measuring performance with time is complicated, using instructions is a more stable measure. perf stat has some useful options like -x SEP to generate CSV output and -o <file> to write the output to a file. Note that you might have to tweak some kernel options or run as root to make perf work, but it's pretty good about telling you what you need to do.

If you use the time command (not the shell builtin) you can get peak RSS (which is probably the thing you care about measuring for memory usage) in kilobytes:

$ /usr/bin/time -f '%M' ./hello
Hello world
1340

You can pass -o <file> to ask time to write its output to a file instead.