This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: Reducing code size of Position Independent Executables (PIE) by shrinking the size of dynamic relocations section
- From: Cary Coutant <ccoutant at gmail dot com>
- To: Rahul Chaudhry <rahulchaudhry at google dot com>
- Cc: Roland McGrath <roland at hack dot frob dot com>, Sriraman Tallam <tmsriram at google dot com>, Florian Weimer <fw at deneb dot enyo dot de>, Rahul Chaudhry via gnu-gabi <gnu-gabi at sourceware dot org>, Suprateeka R Hegde <hegdesmailbox at gmail dot com>, Florian Weimer <fweimer at redhat dot com>, David Edelsohn <dje dot gcc at gmail dot com>, Rafael Avila de Espindola <rafael dot espindola at gmail dot com>, Binutils Development <binutils at sourceware dot org>, Alan Modra <amodra at gmail dot com>, Xinliang David Li <davidxl at google dot com>, Sterling Augustine <saugustine at google dot com>, Paul Pluzhnikov <ppluzhnikov at google dot com>, Ian Lance Taylor <iant at google dot com>, "H.J. Lu" <hjl dot tools at gmail dot com>, Luis Lozano <llozano at google dot com>, Peter Collingbourne <pcc at google dot com>, Rui Ueyama <ruiu at google dot com>, llvm-dev at lists dot llvm dot org
- Date: Thu, 14 Dec 2017 00:11:29 -0800
- Subject: Re: Reducing code size of Position Independent Executables (PIE) by shrinking the size of dynamic relocations section
- Authentication-results: sourceware.org; auth=none
- References: <CAGWvnynFwXFGLj3tAVgDatn0zmuHcWHyRNuDvR+wRZCXLnar_A@mail.gmail.com> <8737cosnym.fsf@localhost.localdomain.i-did-not-set--mail-host-address--so-tickle-me> <CAGWvnynEe3QkhDMGc=Tx8Vr44egtv3xLuh1yiVcAhv+e3GLtZg@mail.gmail.com> <a3e5c76c-8cb9-fc53-a30a-96b2c85079e1@gmail.com> <a68a5d29-09d6-e758-8680-d94f42762adf@redhat.com> <7e698a5f-32d7-6549-7e23-8850b85e6c10@gmail.com> <CAAs8Hmziqc0hebPndiGuZN=buFm=M+O+2fGCfsv_rvDro9zJZA@mail.gmail.com> <CAJRD=ooGubyUOLE6W7LHdeU2ZNDEG1A=84+P=1iOvfmD7-7GNg@mail.gmail.com> <874lozec25.fsf@mid.deneb.enyo.de> <CAAs8HmwMRTjyLjvUAbP9drkagbpedonHOGGRvoFQVr1TE7wyCQ@mail.gmail.com> <CAJRD=opP96vFuSKK-1d1jw3nOKeTDE1T_E5hDwj3Zy-VUeAnRA@mail.gmail.com> <CAORpzuMftCGpXUObOyoFY0=jorMBDWEDbQJ23DifTNW3v-WA6Q@mail.gmail.com> <CAJRD=opERJszwQMFfaKMVdOYF-YAbqqYW0iNWWMqNp3pq2njzw@mail.gmail.com>
> While adding a 'stride' field is definitely an improvement over simple
> delta+count encoding, it doesn't compare well against the bitmap based
> encoding.
>
> I took a look inside the encoding for the Vim binary. There are some instances
> in the bitmap based encoding like
> [0x3855555555555555 0x3855555555555555 0x3855555555555555 ...]
> that encode sequences of relocations applying to alternate words. The stride
> based encoding works very well on these and turns it into much more compact
> [0x0ff010ff 0x0ff010ff 0x0ff010ff ...]
> using stride==0x10 and count==0xff.
Have you looked much at where the RELATIVE relocations are coming from?
I've looked at a PIE build of gold, and they're almost all for
vtables, which mostly have consecutive entries with 8-byte strides.
There are a few for the GOT, a few for static constructors (in
.init_array), and a few for other initialized data, but vtables seem
to account for the vast majority. (Gold has almost 19,000 RELATIVE
dynamic relocs, and only about 500 non-RELATIVE dynamic relocs.)
Where do the 16-byte strides come from? Vim is plain C, right? I'm
guessing its RELATIVE relocation count is fairly low compared to big
C++ apps. I'm also guessing that the pattern comes from some large
structure or structures in the source code where initialized pointers
alternate with non-pointer values. I'm also curious about Roland's
app.
In my opinion, the data and my intuition both support your choice of a
jump + bit vector representation.
-cary