This is the mail archive of the libc-hacker@sourceware.cygnus.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

comment on today's changes


Hi,

I just want to explain the changes I checked in today.  I mean these
hp-timing files.  They implement as least for the i686 high-resolution
timers which I needed in the rtld to gather some numbers.  Since the
overhead is so low I'll leave the code in.

First, if your platform does not support low-overhead timers don't
bother implementing the hp-timing stuff.  It'll never be critical.  If
there is an easy way to implement it, go ahead.

Why I needed this is because I'm looking into rtld improvements and
the data I got is quite intersting.  It showed that the symbol
relocation at startup time (even for lazy binding!) requires 70-80% of
the startup time.  For non-lazy loading it's of course more (and
therefore for MIPS it's always that high).  This was just the
acknowledgment I needed to justify what will come sometime soon.

Richard and I came up with two improvements for the dynamic linking
process which hopefully will cut the relocation time
dramatically. Well, the ideas originally come from SGI and Sun but we
had to reverse-engineer them and them we combined them (neither Sun
nor SGI has the complete solution).  In a two-step development we'll
hopefully be able to cut the time for the relocation by 50 to 100%.
Well, 100% will be seldomly reached but it is possible.

I'll keep you informed when something is available.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]