[haiku] Ang: Re: gcc4 hibryd

  • From: "holmqvist.fredrik@xxxxxxxxxxxx" <holmqvist.fredrik@xxxxxxxxxxxx>
  • To: <haiku@xxxxxxxxxxxxx>
  • Date: Sun, 8 Mar 2009 15:09:56 +0100 (CET)

>joe.prostko+haiku@xxxxxxxxx wrote:
>Datum: 08-03-2009 14:24
>Till: <haiku@xxxxxxxxxxxxx>
>Ärende: [haiku] Re: gcc4 hibryd
>The current GCC4 release of BeZillaBrowser via OptionalPackages uses
>-O1 according to about:buildconfig.  This release is done natively
>from within Haiku via GCC 4.3.3 though (I'm pretty sure), so I am
>wondering if there is a noticeable difference between -O3 and -O1 at
>all.  I'm getting the impression there isn't according to what you are

You can gain a lot from different optimizations, a properly optimized firefox 
is noticably faster. 
I actually started back in the Phoenix days with doing optimized builds, but 
what I was saying was that with the same 
optimizations gcc4 builds of firefox don't get noticable faster compared to 
gcc2 builds. Maybe it uses less cpu, but it 
still suffers from other bottlenecks which is the real problem.

For reference here is what I've landed on when compiling 'optimized builds':
'-O3 -march=i586 -mtune=i586 -fthread-jumps -fforce-addr -frerun-cse-after-loop 
-frerun-loop-opt -fexpensive-
optimizations -falign-functions=4 -falign-jumps=4'
I used to use i686, but people were using K6 cpus and it does not support i686 
it seems. Also I don't know how many 
bugs I've found just because I've been trying different optimizations. It can 
help find bugs as well, not just create 
interesting new ones...

(I know mtune and march is basically the same)
/Fredrik Holmqvist, TQH

Other related posts: