Comparing Linux C and C++ Compilers: Benchmarks and Analysis

[Scott Robert Ladd]

Hello,

I've posted a comparison of recent GCC versions (3.3, 3.4, and the
coming 4.0) with Intel C++ 8.1, including several benchmarks and
"state-of-the-product" reviews. You can find the article at:

http://www.coyotegulch.com/reviews/linux_compilers/

The above article replaces an older article I published in late 2002. This
new comparison marks what I hope will be an ongoing series that tracks the
quality of Linux compilers.

...Scott

 

[CBFalconer]

I've posted a comparison of recent GCC versions (3.3, 3.4, and
the coming 4.0) with Intel C++ 8.1, including several benchmarks
and "state-of-the-product" reviews. You can find the article at:

http://www.coyotegulch.com/reviews/linux_compilers/

The above article replaces an older article I published in late
2002. This new comparison marks what I hope will be an ongoing
series that tracks the quality of Linux compilers.

FYI the above produces peculiar results, like an overprint, when
received here with Netscape 4.75 under W98. Has it been passed
through html verification?

It neglects to show the optimization levels etc. used in running
gcc. The complete command line would be useful. You are aware
than gcc can also be told to use various instruction sets, from
the 386 onwards?

 

[Scott Robert Ladd]

FYI the above produces peculiar results, like an overprint, when
received here with Netscape 4.75 under W98. Has it been passed through
html verification?

The web page uses XHTML formatting in the tables; it works fine with
Firefox 0.9.3 and 1.0PR, IE 6, and Konqueror 3.3, and Mozilla 1.7. I don't
have Netscape installed anywhere. The page passed a "tidy" test on in
Quanta Plus, my web editor.

It neglects to show the optimization levels etc. used in running gcc.
The complete command line would be useful. You are aware than gcc can
also be told to use various instruction sets, from the 386 onwards?

Yes, the article *does* show the switches used, in the section on "Test
Methods."

I compiled for the native instruction set of the target processor (using
-march=opteron and -march=pentium4, as appropriate). Again, this is stated
in the article. To quote from the article:

* for all GCC versions on Pentium 4:
-march=pentium4 -mfpmath=sse -fomit-frame-pointer -ffast-math -O3
* for GCC 3.4 and 4.0 on Opteron:
-march=opteron -ffast-math -O3
* for GCC 3.3 on Opteron:
-march=athlon-xp -ffast-math -O3
* for Intel C++ on Pentium 4:
icc -O3 -xN -tpp7 -ipo

 

[pete]

The web page uses XHTML formatting in the tables; it works fine with
Firefox 0.9.3 and 1.0PR, IE 6, and Konqueror 3.3,
and Mozilla 1.7. I don't have Netscape installed anywhere.
The page passed a "tidy" test on in Quanta Plus, my web editor.

On that page,
http://www.coyotegulch.com/reviews/linux_compilers/
my Netscape 3.04g has overprint,
my IE V6.0 has no overprint.

 

[Jack Klein]

FYI the above produces peculiar results, like an overprint, when
received here with Netscape 4.75 under W98. Has it been passed
through html verification?

If someone told me that a web page of mine did not render properly
under Netscape 4.75, I would ask them why they are using that creaky
antique. Mozilla 1.72, Mozilla Firefox and Netscape 7.2 are free. I
tested the page in two up-to-date browsers, Firefox and Opera.

So why are you using that creaky antique?

 

[CBFalconer]

.... snip ...

If someone told me that a web page of mine did not render properly
under Netscape 4.75, I would ask them why they are using that creaky
antique. Mozilla 1.72, Mozilla Firefox and Netscape 7.2 are free.
I tested the page in two up-to-date browsers, Firefox and Opera.

So why are you using that creaky antique?

For similar reasons to why I am running it on a 486/80. My
primary use is as a newsreader, and every time I consider
upgrading I hear about one more reason that Mozilla/Firefox/NS7.2
would create a problem for me. Besides, it is quite young, just
barely old enough to go to kindergarten. My cars are all older
too, I have a '63 Honda motorcycle in the garage, and I won't even
mention my bicycle.

 

[pete]

For similar reasons to why I am running it on a 486/80. My
primary use is as a newsreader, and every time I consider
upgrading I hear about one more reason that Mozilla/Firefox/NS7.2

If you want to get the newest viruses and worms,
then you need the newest software.

 

[Minti]

Hello,

I've posted a comparison of recent GCC versions (3.3, 3.4, and the
coming 4.0) with Intel C++ 8.1, including several benchmarks and
"state-of-the-product" reviews. You can find the article at:

http://www.coyotegulch.com/reviews/linux_compilers/

The above article replaces an older article I published in late 2002. This
new comparison marks what I hope will be an ongoing series that tracks the
quality of Linux compilers.

..Scott

I just hope you forgot to post this on your bathroom tiles.

 

[Rajeev]

Hello,

I've posted a comparison of recent GCC versions (3.3, 3.4, and the
coming 4.0) with Intel C++ 8.1, including several benchmarks and
"state-of-the-product" reviews. You can find the article at:

http://www.coyotegulch.com/reviews/linux_compilers/

The above article replaces an older article I published in late 2002. This
new comparison marks what I hope will be an ongoing series that tracks the
quality of Linux compilers.

..Scott

Scott,

Terrific work ! I've been wishing to find exactly this.

I'm looking for discussion groups, mailing lists, etc where folks discuss
experiences in achieving optimizations. Thanks for your suggestions, on-line
or off-

Regards,
-rajeev-