-O3 contains what?

Message

eyeL · Post by **eyeL** » Mon Jan 16, 2006 9:27 pm

hrm, I was searching for a link to see what all -O3 contains, because I don't want to put something else in there that is already there.. but I couldn't find one. Anyone have any links?

truc · Post by **truc** » Mon Jan 16, 2006 9:30 pm

this can be interesting(read all the posts:) )

HydroSan · Post by **HydroSan** » Mon Jan 16, 2006 9:31 pm

http://gcc.gnu.org/onlinedocs/gcc-3.4.5 ... ze-Options

Code: Select all

-O
-O1
    Optimize. Optimizing compilation takes somewhat more time, and a lot more memory for a large function.

    With -O, the compiler tries to reduce code size and execution time, without performing any optimizations that take a great deal of compilation time.

    -O turns on the following optimization flags:

              -fdefer-pop 
              -fmerge-constants 
              -fthread-jumps 
              -floop-optimize 
              -fif-conversion 
              -fif-conversion2 
              -fdelayed-branch 
              -fguess-branch-probability 
              -fcprop-registers
         

    -O also turns on -fomit-frame-pointer on machines where doing so does not interfere with debugging.
-O2
    Optimize even more. GCC performs nearly all supported optimizations that do not involve a space-speed tradeoff. The compiler does not perform loop unrolling or function inlining when you specify -O2. As compared to -O, this option increases both compilation time and the performance of the generated code.

    -O2 turns on all optimization flags specified by -O. It also turns on the following optimization flags:

              -fforce-mem 
              -foptimize-sibling-calls 
              -fstrength-reduce 
              -fcse-follow-jumps  -fcse-skip-blocks 
              -frerun-cse-after-loop  -frerun-loop-opt 
              -fgcse  -fgcse-lm  -fgcse-sm  -fgcse-las 
              -fdelete-null-pointer-checks 
              -fexpensive-optimizations 
              -fregmove 
              -fschedule-insns  -fschedule-insns2 
              -fsched-interblock  -fsched-spec 
              -fcaller-saves 
              -fpeephole2 
              -freorder-blocks  -freorder-functions 
              -fstrict-aliasing 
              -funit-at-a-time 
              -falign-functions  -falign-jumps 
              -falign-loops  -falign-labels 
              -fcrossjumping
         

    Please note the warning under -fgcse about invoking -O2 on programs that use computed gotos.
-O3
    Optimize yet more. -O3 turns on all optimizations specified by -O2 and also turns on the -finline-functions, -fweb, -frename-registers and -funswitch-loops options.

eyeL · Post by **eyeL** » Mon Jan 16, 2006 9:32 pm

Thanks alot guys. This'll help me

eyeL · Post by **eyeL** » Mon Jan 16, 2006 9:37 pm

err, wait, I thought -fomit-frame-pointer was in there as well?

EDIT:

Ok, here's my cflags now.

CFLAGS="-O3 -march=athlon64 -mmmx -msse -msse2 -m3dnow -funroll-all-loops -fpeel -loops -ftracer -fforce-addr -pipe"

Thoughts, comments, suggestions, ideas?

Post by **Halcy0n** » Mon Jan 16, 2006 10:12 pm

eyeL wrote:err, wait, I thought -fomit-frame-pointer was in there as well?

EDIT:

Ok, here's my cflags now.

CFLAGS="-O3 -march=athlon64 -mmmx -msse -msse2 -m3dnow -funroll-all-loops -fpeel -loops -ftracer -fforce-addr -pipe"

Thoughts, comments, suggestions, ideas?

-mmmx, -msse, -msse2, -m3dnow are all redundant, get rid of them.

-funroll-all-loops and -fpeel-loops - ricer flags, will probably break things, get rid of them
-fforce-addr - may or may not produce better code, get rid of it
-O3 - broken in some cases, use -O2

So, what we are left with is sane:

CFLAGS="-O2 -march=athlon64 -ftracer -pipe"

More flags does not == better in this case.

eyeL · Post by **eyeL** » Mon Jan 16, 2006 10:13 pm

Thanks alot. and should I remove

CXXFLAGS="${CFLAGS} -fvisibility-inlines-hidden"

that extra part as well?

truc · Post by **truc** » Mon Jan 16, 2006 10:32 pm

Not sure but I don't think so (read the link I gave above:

slycordinator wrote:One small thing with that...

There might be a good reason to leave redundant CFLAGS in.

Lets say one ebuild automatically strips CFLAGS and changes O3 to O2 (since O2 is "safer"). Lets then say that one of the options that O3 provides is perfectly safe. If you add it in redundantly, then you'll still get the effect of it after the ebuild does the change.

Food for thought.

Vann · Post by **Vann** » Mon Jan 16, 2006 10:32 pm

I generally find that my system runs slightly better with -Os than with -O2 or -O3. I imagine it is because it allows for better use of the CPU cache. My CFLAGS are very simple: CFLAGS="-march=athlon-xp -pipe -Os"

Vann · Post by **Vann** » Mon Jan 16, 2006 10:33 pm

If an ebuild is dropping -O3 to -O2, then manually specifying flags set by -O3 is probably a bad idea. ebuilds only do this when the more aggressive flags are known to break things.

eyeL · Post by **eyeL** » Mon Jan 16, 2006 10:34 pm

Vann wrote:I generally find that my system runs slightly better with -Os than with -O2 or -O3. I imagine it is because it allows for better use of the CPU cache. My CFLAGS are very simple: CFLAGS="-march=athlon-xp -pipe -Os"

Where do you notice the performance in? Run times, compiling code, or just emerging?

AaronPPC · Post by **AaronPPC** » Tue Jan 17, 2006 12:40 am

I have been using -O3 when I re-installed my sytem a couple of months ago. I am very disappointed by my system performance since switching. I swear it ran faster with -O2. The next time I upgrade gcc, I think I'll switch to -O2 or -Os since I'll have to do an emerge -e world.

eyeL · Post by **eyeL** » Tue Jan 17, 2006 2:16 am

I'm trying to decide the advantages of -O2 over -Os, or the opposite way, that way I know which one to pick.