Wireshark · Ethereal-dev: Re: [Ethereal-dev] packet-parlay.c takes a very long time to compile in Fedora Core 5

[Jeff Morriss <jeff.morriss@xxxxxxxxxxx>]

Guy Harris wrote:
> Jeff Morriss wrote:
>
> > Has anyone else noticed that packet-parlay.c (in plugins/giop/) takes 
> > a _very_ long time to compile?  I recently upgraded from Fedora Core 4 
> > to Fedora Core 5 and now it takes around 10 minutes to compile it (on 
> > my 1.6 GHz Pentium 4); I seriously thought the compiler was hung it 
> > was taking so long...
> >
> > Of course I have no clue how long it used to take and I don't have an 
> > (easy) way to test this on other versions or platforms at the 
> > moment--is anyone else seeing anything like this?
>
> Not quite that bad, but it does take a significant amount of time to 
> compile on OS X 10.4.6 (1.5 GHz PPC G4).  (Did the FC4 -> FC5 upgrade 
> either GCC or the GCC options?  The compiler might be doing more 
> aggressive optimization.)

It looks like I (should have had) gcc 4.0.2 (+ whatever Redhat patches) 
and now I have 4.1.0 .

http://gcc.gnu.org/gcc-4.1/changes.html

has a rather large section on optimization improvements, so I suspect 
that may be the difference...

I don't particularly feel like struggling with downgrading my compiler, 
but here are various compile times using gcc-4.1 on my system:

Without optimization:

real    0m17.812s
user    0m16.125s
sys     0m0.836s

-O1:

real    8m28.577s
user    8m12.139s
sys     0m1.852s

-O2:

real    9m50.071s
user    9m39.720s
sys     0m2.604s

-O3:

real    2m43.235s
user    2m31.161s
sys     0m3.008s


I'm a little shocked by the jump from -O0 to -O1 and am puzzled by the 
-O3 result.  I guess one of the (new?) optimizations is left out in -O3? 
 Seems odd...