Best execution times in seconds (see below for details):
| Benchmark |
|
|---|---|
| >
|
|
|
|
>
> |
| total (lower is better) | >
|
| performance relative to best on host (higher is better) |
>
|
| >
|
| performance relative to best on host (higher is better) |
|
| Benchmark | best | average | worst | 2σ | n | timed command | ||
|---|---|---|---|---|---|---|---|---|
| median | trimmed | mean | ||||||
All benchmarks are automatically started via ssh. The timed command is executed in a bash for loop with all output directed to local files (the actual loop used is written on one line):
sleep 5 for b in $BENCHMARKS do for s in $sizes do echo >&2 "benchmark:$BENCHGROUP:$b:$s" tt="$THE_TIMED_COMMAND" echo >&2 "timed:$tt" for i in $BENCHITERATIONS do time -p $tt done done done >BENCHOUT 2>BENCHERR
For each execution the user time as output by the time command is taken. Statistics are calculated using the Statistics::Descriptive Perl module: