Performance difference between striped and non-striped ELFs

**borromini** · 10-12-2005

I have a simple question

. Today I installed privoxy, to get some decent ad-blocking in Opera (my filter.ini file and stuff was giving mixed results - ugly layout and stuff...). Visual improvement all the way, no problemo. However, I did not succeed in compiling it myself (make install exited with error[1] and it wasn't clear to me what the cause was), so I installed a premade .tgz package (the only one I could find on slackware packages sites).

I installed privoxy, configured my browser, so far so good, it worked. But loading sites took ages! Now I use a Perl script to check packages (my own and other's) if they are created well (stripped ELFs, right directories, permissions, ...), and all seemed fine with this one. However, I ran the 'file' command on /usr/sbin/privoxy, and it said to me 'ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), not stripped'.

Ok. So I remove the installed package, extract the tgz, and strip the executable, and I make a new .tgz. Now everything goes a lot faster - I nearly don't notice the difference between surfing without or with privoxy!?

Now my question is: Can striping really make such a difference?!

**borromini** · 10-14-2005

Noone with similar experiences? ...

**bidi** · 10-14-2005

As far as I know, stripping an executable just removes all the comments (lines starting with #) out of the file. I guess it could make it slightly faster, but not enough for you to notice. When I did the Pocket Linux Guide, I had to stripe every program so that it would not take up as much space, so I can see it speeding up things since there's not as much stuff that needs to go into memory when you load it.

**bigtomrodney** · 10-14-2005

I was under the impression that comments were ignored completely by compilers - is this not the case?

**borromini** · 10-14-2005

Originally Posted by bigtomrodney

I was under the impression that comments were ignored completely by compilers - is this not the case?

What do you mean? There's debugging information and lots of other stuff present after compiling.

**techieMoe** · 10-14-2005

The manpage for strip (that was a difficult at-work Google, let me tell you) gives a little more information:

http://web.cse.msu.edu/cgi-bin/man2h...rip?1?/usr/man

My professors in college said that stripping did in fact help with performance. I haven't done enough C++ development professionally to say.

**borromini** · 10-14-2005

Right now, Opera with privoxy seems to be as quick as without before, so... And surely a lot quicker then with an unstripped privoxy binary. Of course I haven't timed it, it could just have been me proud about being able to strip a stupid binary

.

**bidi** · 10-14-2005

Just goes to show my memory sucks. I did remember it took unneeded info though.

You can do the same to libraries, but you will need the "--strip-unneeded" parameter or else it'll just remove way too much stuff, some of which you will want.

**borromini** · 10-14-2005

Are there switches you need for executables? I just did

Code:

# strip privoxy

Thread Tools

Display

Performance difference between striped and non-striped ELFs

Posting Permissions