[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130420174639.GA15232@brightrain.aerifal.cx>
Date: Sat, 20 Apr 2013 13:46:39 -0400
From: Rich Felker <dalias@...ifal.cx>
To: musl@...ts.openwall.com
Subject: Automating libc comparison
Hi,
Here's a project I'd like to do but I thought I probably should spend
the time actually coding/maintaining, so I figured I'd post it in case
anyone else wants to help or do it:
Basic idea is to automate the libc comparison
(http://www.etalabs.net/compare_libcs.html) so that we can keep it up
to date and add new libcs (BSD? Bionic? etc.) and tests to it. A good
part of the challenge is doing tests in a fairly libc-agnostic way so
the same tests could be applied with little effort to other libcs.
Some ideas:
- Complete .a set size can be computed with -Wl,--whole-archive and
all the standard -l options listed in POSIX.
- Complete .so set might be possible with ldd or readelf and knowledge
of the library path.
- Most overhead/bloat measurements are easy with an appropriate test
program and /proc.
- Resource-exhaustion tests would need test programs and could
simulate exhaustion with resource limits.
- Performance comparisons all presently come from libc-bench, but
there's a lot of room for replacing/improving them. It would be nice
for some of them to be converted to units like MB/sec.
- The bottom half of the table is a lot less measurable, but at least
some things like features might be able to be partially-automated.
Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
