[gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Spider]

[-- Attachment #1: Type: text/plain, Size: 1071 bytes --]

Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
(unofficial) benchmark.

I went for galeon, I had originally intended to use mozilla, but the
time-results borked so I go for galeon instead.. smaller codebase, so
its not as great difference, but it does have both c and c++ code, so it
might be a decent choice.


gcc 2.95.3 :
real    3m38.592s
user    2m46.810s
sys     0m28.100s
CFLAGS="-march=i686 -O3 -pipe"
CXXFLAGS="-march=i686 -O3 -pipe"


gcc 3.0.4 :
real    5m6.465s
user    3m27.440s
sys     0m30.140s
CFLAGS="-march=athlon -O3 -pipe"
CXXFLAGS="-march=athlon -O3 -pipe"




if you only compare the "user" time it should be enough... as the "sys"
show, there's a few percentages difference between them, so this is not
scientific or anything.

Would be interesting to compare the results as well, since those are
quite likely rather different with the new levels of optimization...


//Spider


--
begin  happy99.exe
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Bart Verwilst]

Hmm.. so gcc 3.x is indeed slower than gcc 2.95.3...
That's it, no more gcc 3.x plans for me for the first few months! :o)

Thanks!

On Monday 08 April 2002 00:54, Spider wrote:
|| Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
|| (unofficial) benchmark.
||
|| I went for galeon, I had originally intended to use mozilla, but the
|| time-results borked so I go for galeon instead.. smaller codebase, so
|| its not as great difference, but it does have both c and c++ code, so it
|| might be a decent choice.
||
||
|| gcc 2.95.3 :
|| real    3m38.592s
|| user    2m46.810s
|| sys     0m28.100s
|| CFLAGS="-march=i686 -O3 -pipe"
|| CXXFLAGS="-march=i686 -O3 -pipe"
||
||
|| gcc 3.0.4 :
|| real    5m6.465s
|| user    3m27.440s
|| sys     0m30.140s
|| CFLAGS="-march=athlon -O3 -pipe"
|| CXXFLAGS="-march=athlon -O3 -pipe"
||
||
||
||
|| if you only compare the "user" time it should be enough... as the "sys"
|| show, there's a few percentages difference between them, so this is not
|| scientific or anything.
||
|| Would be interesting to compare the results as well, since those are
|| quite likely rather different with the new levels of optimization...
||
||
|| //Spider

-- 
Bart Verwilst
Gentoo Linux Developer, Desktop Team
Gent, Belgium

Re: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Bart Verwilst]

Oh, and euhm..
What about this?: 
http://gcc.gnu.org/faq.html#optimizing

If gcc itself is optimised with for example '-O3 -fomit-frame-pointer',
won't that make it faster than gcc 2.95.3? :o)
Just taking wild guesses here :o)

See ya

On Monday 08 April 2002 00:54, Spider wrote:
|| Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
|| (unofficial) benchmark.
||
|| I went for galeon, I had originally intended to use mozilla, but the
|| time-results borked so I go for galeon instead.. smaller codebase, so
|| its not as great difference, but it does have both c and c++ code, so it
|| might be a decent choice.
||
||
|| gcc 2.95.3 :
|| real    3m38.592s
|| user    2m46.810s
|| sys     0m28.100s
|| CFLAGS="-march=i686 -O3 -pipe"
|| CXXFLAGS="-march=i686 -O3 -pipe"
||
||
|| gcc 3.0.4 :
|| real    5m6.465s
|| user    3m27.440s
|| sys     0m30.140s
|| CFLAGS="-march=athlon -O3 -pipe"
|| CXXFLAGS="-march=athlon -O3 -pipe"
||
||
||
||
|| if you only compare the "user" time it should be enough... as the "sys"
|| show, there's a few percentages difference between them, so this is not
|| scientific or anything.
||
|| Would be interesting to compare the results as well, since those are
|| quite likely rather different with the new levels of optimization...
||
||
|| //Spider

-- 
Bart Verwilst
Gentoo Linux Developer, Desktop Team
Gent, Belgium

Re: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Spider]

[-- Attachment #1: Type: text/plain, Size: 1566 bytes --]

And some more numbers:

gcc 3.0.4 - gabber
real    14m51.179s
user    12m3.010s
sys     0m34.320s

gcc 2.95.3 - gabber
real    7m53.177s
user    6m57.450s
sys     0m29.940s



begin  quote
On Mon, 8 Apr 2002 00:54:46 +0200
Spider <spider@gentoo.org> wrote:

> Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
> (unofficial) benchmark.
> 
> I went for galeon, I had originally intended to use mozilla, but the
> time-results borked so I go for galeon instead.. smaller codebase, so
> its not as great difference, but it does have both c and c++ code, so
> it might be a decent choice.
> 
> 
> gcc 2.95.3 :
> real    3m38.592s
> user    2m46.810s
> sys     0m28.100s
> CFLAGS="-march=i686 -O3 -pipe"
> CXXFLAGS="-march=i686 -O3 -pipe"
> 
> 
> gcc 3.0.4 :
> real    5m6.465s
> user    3m27.440s
> sys     0m30.140s
> CFLAGS="-march=athlon -O3 -pipe"
> CXXFLAGS="-march=athlon -O3 -pipe"
> 
> 
> 
> 
> if you only compare the "user" time it should be enough... as the
> "sys" show, there's a few percentages difference between them, so this
> is not scientific or anything.
> 
> Would be interesting to compare the results as well, since those are
> quite likely rather different with the new levels of optimization...
> 
> 
> //Spider
> 
> 
> --
> begin  happy99.exe
> This is a .signature virus! Please copy me into your .signature!
> See Microsoft KB Article Q265230 for more information.
> end
> 


--
begin  happy99.exe
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

RE: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Ric Messier]

I think the important question is -- does gcc3 generate better/faster
code? If it generates faster code, I'm okay with spending a little more
time to create it because over the life of the system, the amount of
time I spend compiling is small in comparison.

Ric


|> -----Original Message-----
|> From: gentoo-dev-admin@gentoo.org 
|> [mailto:gentoo-dev-admin@gentoo.org] On Behalf Of Spider
|> Sent: Sunday, April 07, 2002 6:55 PM
|> To: gentoo-dev
|> Subject: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion
|> 
|> 
|> Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
|> (unofficial) benchmark.
|> 
|> I went for galeon, I had originally intended to use mozilla, 
|> but the time-results borked so I go for galeon instead.. 
|> smaller codebase, so its not as great difference, but it 
|> does have both c and c++ code, so it might be a decent choice.
|> 
|> 
|> gcc 2.95.3 :
|> real    3m38.592s
|> user    2m46.810s
|> sys     0m28.100s
|> CFLAGS="-march=i686 -O3 -pipe"
|> CXXFLAGS="-march=i686 -O3 -pipe"
|> 
|> 
|> gcc 3.0.4 :
|> real    5m6.465s
|> user    3m27.440s
|> sys     0m30.140s
|> CFLAGS="-march=athlon -O3 -pipe"
|> CXXFLAGS="-march=athlon -O3 -pipe"
|> 
|> 
|> 
|> 
|> if you only compare the "user" time it should be enough... 
|> as the "sys" show, there's a few percentages difference 
|> between them, so this is not scientific or anything.
|> 
|> Would be interesting to compare the results as well, since 
|> those are quite likely rather different with the new levels 
|> of optimization...
|> 
|> 
|> //Spider
|> 
|> 
|> --
|> begin  happy99.exe
|> This is a .signature virus! Please copy me into your 
|> .signature! See Microsoft KB Article Q265230 for more 
|> information. end
|> 

Re: [gentoo-dev] gcc 2.95.3 / 3.0.4 speed comparsion

[Daniel Robbins]

On Sun, 2002-04-07 at 16:54, Spider wrote:
> Hello, I've just upgraded my -rc6 to -1.0-gcc3 and decided to make an
> (unofficial) benchmark.

What are you benchmarking?  Compile time?  Runtime execution speed? 
It's not clear.

-- 
Daniel Robbins                                  <drobbins@gentoo.org>
Chief Architect/President                       http://www.gentoo.org 
Gentoo Technologies, Inc.