diff options
| -rwxr-xr-x | mtime.sh | 14 | ||||
| -rw-r--r-- | speedup_import.md | 36 |
2 files changed, 50 insertions, 0 deletions
diff --git a/mtime.sh b/mtime.sh new file mode 100755 index 00000000..1cb5f8dc --- /dev/null +++ b/mtime.sh @@ -0,0 +1,14 @@ +#!/bin/bash + +TMPFILE=`mktemp /tmp/mtime.XXXXXX` || exit 1 + +for x in {1..100} +do + gtime -f "real %e user %U sys %S" -a -o $TMPFILE "$@" + #tail -1 $TMPFILE +done + +awk '{ et += $2; ut += $4; st += $6; count++ } END { printf "%d iterations\n", count ; printf "average: real %.3f user %.3f sys %.3f\n", et/count, ut/count, st/count }' $TMPFILE + +rm $TMPFILE + diff --git a/speedup_import.md b/speedup_import.md new file mode 100644 index 00000000..12d95bf3 --- /dev/null +++ b/speedup_import.md @@ -0,0 +1,36 @@ +# Speedup Import + +## Assumptions + +I created a simple script to run a command 20 times and calculate +the average clock time for each run of the command. This script requires +some unix tools, including the gnu flavor of the `time` command. This script +can is called `mtime.sh` and is included in this branch. + +These tests were all run on my 2015 MacBook Pro with a 3.1 GHz Intel Core i7 +and 16GB of memory. + + +## Baseline measurement + +First let's see how long it takes to start up python. The longish path here +ensures we aren't measuring the time it takes the pyenv shims to run: +``` +$./mtime.sh ~/.pyenv/versions/cmd2-3.6/bin/python -c "" +100 iterations +average: real 0.028 user 0.020 sys 0.000 +``` + +## Initial measurement + +From commit c7753352b, which has `__init.py__` importing `cmd2.cmd2.Cmd` +and a bunch of other stuff, we get: +``` +$ ./mtime.sh ~/.pyenv/versions/cmd2-3.6/bin/python -c "import cmd2" +100 iterations +average: real 0.140 user 0.100 sys 0.030 +``` + +From the baseline and this initial measurement, we infer it takes ~110 ms +to import the `cmd2` module. + |