develooper Front page | perl.perl5.porters | Postings from November 2014

some tests are slower than others

Thread Next
From:
Jarkko Hietaniemi
Date:
November 29, 2014 02:03
Subject:
some tests are slower than others
Message ID:
5479294B.3010504@iki.fi
I've noticed that some of the tests are slower than others.

No, really.

Especially if one runs the test suite in not-multi-cpu-or-core 
not-lots-of-memory not-fast-disk system.  At certain tests things seem 
to go...
slow.  Sometimes, really, really, slow.  Like 42 minutes slow.
Yes, a single test.

So I got me some numbers.  I ran the tests with the new HARNESS_TIMER=1
option, across eight different systems.  Some faster, some slower.
Wall-time difference magnitude of 34: from 11 minutes to 7 hours.

Since the numbers (wall or cpu secs) varied so lot, I normalized.
All tests sorted on their wall time, and normalized by the median.
Then I counted over the eight systems the average (of the normalized
medians), the stddev, and stddev/average.

What tests stand out doesn't surprise me because I see many "old
friends" that seem to always be ... slow.  Below is the list with
cut-off point at 100: that is, those tests take x100 or more time
then the median tests, across several systems.

(I'll try to clean up my "system" of scripts, they are not quite
presentable yet...)

The logs were all reasonably recent, within couple last days.

The 'count' is not eight in all the below because I only counted
the test if it was "ok" and I could parse the log, and some tests
are conditionally skipped.  The stddev is often quite high, I haven't
looked into why: maybe some systems have significantly faster I/O,
maybe?

type|test|count|avg of median-normalized walltimes|stddev|stddev/avg
---
core porting/regen.t 7 862.8 565.3 0.655
core re/fold_grind.t 8 644.4 431.8 0.670
core re/uniprops.t 7 391.7 431.6 1.102
core porting/podcheck.t 8 365.2 177.8 0.487
core lib/Unicode/UCD.t 7 254.5 116.9 0.459
ext ext/XS-APItest/t/handy.t 8 241.0 116.8 0.485
core lib/charnames.t 8 203.5 64.4 0.316
cpan cpan/Memoize/t/expmod_t.t 8 186.3 184.2 0.989
core re/pat_psycho.t 8 165.5 71.8 0.434
core lib/Benchmark.t 8 158.8 156.9 0.988
core lib/perl5db.t 8 155.5 19.4 0.124
dist dist/IO/t/io_sock.t 8 148.0 145.9 0.985
core uni/variables.t 8 147.8 127.5 0.862
core lib/locale.t 6 147.8 142.3 0.963
core lib/warnings.t 6 146.3 42.6 0.291
cpan cpan/IO-Compress/t/105oneshot-zip-store-only.t 8 144.0 58.9 0.409
cpan cpan/Memoize/t/speed.t 8 137.0 130.1 0.950
core lib/B/Deparse-core.t 8 135.0 68.2 0.505
cpan cpan/ExtUtils-Constant/t/Constant.t 7 133.5 25.2 0.189
ext ext/B/t/concise-xs.t 8 130.9 56.3 0.430
cpan cpan/AutoLoader/t/02AutoSplit.t 8 130.6 123.5 0.945
cpan cpan/File-Temp/t/fork.t 8 113.7 120.7 1.061
cpan cpan/ExtUtils-MakeMaker/t/basic.t 8 110.8 21.2 0.191
cpan cpan/IO-Compress/t/101truncate-zip.t 7 103.3 74.1 0.717
cpan cpan/IO-Compress/t/106prime-zip.t 8 102.7 41.5 0.404
---

Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About