Markus Becker wrote:

> Mark Wagner wrote:
> >
> > How does GCC compare for optimizing code?
>
> As far as I know, they are about equal. (MS and GCC)
>
> Markus

One point on the PNG mailing lists came up with the non-optimizing VC++
and GCC. Someone compared some straight C code versus some
hand-optimized assembly language, and gcc on the straight code did a
better job than the hand-optimized assembly code under gcc or VC++. So,
for some things it might even beat VC++.

--
"My new computer's got the clocks, it rocks
But it was obsolete before I opened the box" - W.A.Y.

Post a reply to this message

Jon A. Cruz <jon### [at] geocitiescom> wrote:
: gcc on the straight code did a
: better job than the hand-optimized assembly code

  This only means that the person who made the assembly code didn't know
what he was doing.
  You can always make the same and often better code than a program can.
It requires LOTS of knowledge about the cpu, though.

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

On Wed, 22 Dec 1999 16:29:42 +0100, Markus Becker
<mar### [at] studentuni-siegende> wrote:

>Sander wrote:
>> 
>> What about Borland (Inprise ;-)) C++ compiler? No one mentions it. Is it not
>> of the same class as Watcom and MS?
>
>No, it isn't. It's about twice as expensive and does not optimize
>as good. Watcom has long been the compiler producing the fastest code,
>no overtaken by MS.

Does anyone have any benchmarks on the Intel Performance Toolkit C++
compiler that comes with vTune?  I have it here, and I've been
building 3.5 with it (note that that doesn't say anything about what
the final official version will be built with.)  If nobody has
benchmarks,  I might be persuaded to build an MS version and an Intel
version and compare, but I'd have to wait to do the tests because I
don't have a Pentium running Windows here at home (this box is a
K6/233.)

Post a reply to this message

On 27 Dec 1999 05:10:49 -0500, Nieminen Juha
<war### [at] punarastascstutfi> wrote:

>  You can always make the same and often better code than a program can.
>It requires LOTS of knowledge about the cpu, though.

That's knowledge most mere mortals don't have when it comes to P5/P6
optimization.

Post a reply to this message

Ron Parker <par### [at] fwicom> wrote:
:>  You can always make the same and often better code than a program can.
:>It requires LOTS of knowledge about the cpu, though.

: That's knowledge most mere mortals don't have when it comes to P5/P6
: optimization.

  You are terribly right.
  A long time ago, in the age of 286's and even 386's the architecture of
the CPU was so simple that making a superduper-optimized asm code which
was at least twice as fast as any C code, was quite easy.
  Nowadays, however, the CPU's have evolved to such an extent and they are
so extremely complicated, that the same tricks of the old asm coding do not
hold anymore. To beat a good C-compiler it would require so much knowledge
about the CPU (and often other hardware) that it's almost impossible for
the average asm-coder.
  The compiler has two main advantages over a human coder: It will remember
everything taught to it and it will compile extremely fast.

  Nowadays it doesn't make much sense to code in assembler for the big
computers. I have not written a line of asm in many years. A pitty, really.
  Fortunately we still have the embedded systems...

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

Nieminen Juha wrote:

>   Nowadays, however, the CPU's have evolved to such an extent and they are
> so extremely complicated, that the same tricks of the old asm coding do not
> hold anymore. To beat a good C-compiler it would require so much knowledge
> about the CPU (and often other hardware) that it's almost impossible for
> the average asm-coder.

The new complexity is also reason to pity the compiler writers!
 
-Mark Gordon

Post a reply to this message

Mark Gordon wrote:

> The new complexity is also reason to pity the compiler writers!

Pity or respect ?

-- 
Ken Tyler -  1300+ Povray, Graphics, 3D Rendering, and Raytracing Links:
http://home.pacbell.net/tylereng/index.html http://www.povray.org/links/

Post a reply to this message

Ken wrote:
> 
> Mark Gordon wrote:
> 
> > The new complexity is also reason to pity the compiler writers!
> 
> Pity or respect ?

I respect them if they write good compilers.  I pity them regardless.
;-)

-Mark Gordon

Post a reply to this message

> > What about Borland (Inprise ;-)) C++ compiler? No one mentions it. Is it
not
> > of the same class as Watcom and MS?
>
> No, it isn't. It's about twice as expensive and does not optimize
> as good. Watcom has long been the compiler producing the fastest code,
> no overtaken by MS.

Which version of the compiler are you referring to? Be careful, your
(Markus') above statement by itself is not objective and comes close to
being merely flame-bait. Which version of the compiler are you referring to?
Did you do an apples-to-apples comparison yourself?

I test compiled povray 3.0 a while back using Borland C++ 4.5 (IIRC) and it
*was* slightly slower than the normal binary downloaded from povray.org. I
would be interested to see the results of a comparison using the Borland
C++Builder 4.0 compiler compared with binaries produced by other compilers.

I am writing a flight dynamics model for the open source flight simulator
called FlightGear (www.flightgear.org). My C++ source code is compilable
under the CygWin compiler (egcs 2.91-xx), all Borland compilers, gcc under
Linux, cc under IRIX, and - with some hacks - under MSVC++ 6.0 (which does
not #define M_PI in its headers like every other header file set does).
Compliance to standards is important to me, and MS skirts the edge - and
sometimes steps over the edge - in my experience.

You are right when saying that Borland is not in the same class as Watcom
and MS. For my needs, they are way above the rest of the crowd. They may not
have a compiler which produces the fastest *povray* executable, but until I
see a fair comparison using the latest versions of all compilers tested,
statements like yours claiming poor optimization and slow binaries will be
taken with a grain of salt (no offense meant).

jb

Post a reply to this message

"Jon S. Berndt" wrote:
> 
> Which version of the compiler are you referring to? Be careful, your
> (Markus') above statement by itself is not objective and comes close to
> being merely flame-bait. Which version of the compiler are you referring to?
> Did you do an apples-to-apples comparison yourself?

I was comparing Borland C++-Builder 4 with MSVC++ 6 and Watcom 11.b
all (at that time) the latest.

> You are right when saying that Borland is not in the same class as Watcom
> and MS. For my needs, they are way above the rest of the crowd. They may not

When it comes to interface design and how good the class library is,
definitely yes. But I like the MS-IDE better. 

Markus

Post a reply to this message