On Wed, 31 Jul 2002 07:06:26 -0400, Sebastian H. wrote:


>> Yes, it's a very real effect.  I got 10 seconds to render Skyvase under
>> Windows, and 24 seconds under Linux.  The reason for this is the
>> different compilers used: in the Windows version, the compiler produced
>> several versions of parts of the code, each version optimized for a
>> different CPU.  This gives a larger but faster binary than the Linux
>> compiler, which appears to have optimized for a generic x86 processor.
>> 
>> 
> Nice and not nice to know...
> Would a for the specific (linux) system optimized compilation gain the
> speed of the windows bins?
 
Maybe with the Intel compiler, but probably not with gcc.  Speed is no
longer a reason to use the Unix version, and it hasn't been for quite some time.

-Mark Gordon

Post a reply to this message

Sebastian H. <seb### [at] webde> wrote:
>>   If everything was exactly identical, I don't see any reason for those
>> numbers to be different.

> Yup this it what makes me wonder too.

  Then why don't you just make sure that you are using *all* the same ini
settings?
  You can save the whole set of ini settings with a command line parameter.
You can use this ini file directly as the ini file for the file to be
rendered.

-- 
#macro N(D)#if(D>99)cylinder{M()#local D=div(D,104);M().5,2pigment{rgb M()}}
N(D)#end#end#macro M()<mod(D,13)-6mod(div(D,13)8)-3,10>#end blob{
N(11117333955)N(4254934330)N(3900569407)N(7382340)N(3358)N(970)}//  - Warp -

Post a reply to this message

Mark Gordon <mtg### [at] povrayorg> wrote:
> Maybe with the Intel compiler, but probably not with gcc.

  I thought that gcc 3.1 has pretty good optimizations for current processors
(getting quite close to intel's compiler).

-- 
#macro M(A,N,D,L)plane{-z,-9pigment{mandel L*9translate N color_map{[0rgb x]
[1rgb 9]}scale<D,D*3D>*1e3}rotate y*A*8}#end M(-3<1.206434.28623>70,7)M(
-1<.7438.1795>1,20)M(1<.77595.13699>30,20)M(3<.75923.07145>80,99)// - Warp -

Post a reply to this message

Mark Gordon wrote:
> Maybe with the Intel compiler, but probably not with gcc.  Speed is no
> longer a reason to use the Unix version, and it hasn't been for quite some time.

I'm not sure what you mean by speed no longer being the reason to use 
the Unix version. Do you mean that people using the Unix version don't 
mind it being much slower than the Windows version? Well I'd like to see 
it come close at least. The faster the render, the better ;)

I rendered the Balcony scene from the advanced sample scenes and the 
Windows version was noticeably faster: 505s for Windows, 785s for Unix 
(Linux). Both version 3.5, same machine, same settings.

I figured that you had compiled the Unix version with the mindset of 
targeting the widest possible cpu compatibility (i.e. 386 on up). I'm 
just guessing though as I didn't watch you build the binary ;)

Perhaps a 2nd binary version targeted at higher end machines could be 
provided? Just a thought. I appreciate that there's a Unix version at all!

Once the source is available, then this may become a non-issue anyway :)

-Roz

Post a reply to this message

Mark Gordon wrote:

> I may release a binary built with the Intel compiler at some point.

I've heard programs compiled with the Intel compiler aren't very stable. 
This may be annoying at long renderings.

-- 
light_source{0#macro L(K,H,W)sphere{H.5}sphere{K.5}sphere{W.5}cylinder{
H,K.5}cylinder{H,W.5}#end 3}union{L(0v*-2<2,-2>)L(y*-3z-v*5z*3-y)L(-y*3
0u*3)L(y*-3v*-5-z,z*-3-y)rotate-v*clock pigment{rgb.5}translate<0,2,9>}
// +KFF200 +KF720 +W120 +H90 -F -A -GA -P

Post a reply to this message

Warp wrote:
> Mark Gordon <mtg### [at] povrayorg> wrote:
>> Maybe with the Intel compiler, but probably not with gcc.
>
>   I thought that gcc 3.1 has pretty good optimizations for current
> processors (getting quite close to intel's compiler).

compared to the previous version, gcc 3.1 is a huge leap forward. but it not
the fastest compiler by any stretch of the imagination.

question the source of any statistics as much as the results they clam. gcc
is an institution


--
Rick

Kitty5 WebDesign - http://Kitty5.com
POV-Ray News & Resources - http://Povray.co.uk
TEL : +44 (01270) 501101 - FAX : +44 (01270) 251105 - ICQ : 15776037

PGP Public Key
http://pgpkeys.mit.edu:11371/pks/lookup?op=get&search=0x231E1CEA

Post a reply to this message

In article <3D4### [at] netscapenet>, Roz <Rzl### [at] netscapenet> 
wrote:

> Mark Gordon wrote:
> > Maybe with the Intel compiler, but probably not with gcc.  Speed is no
> > longer a reason to use the Unix version, and it hasn't been for quite some 
> > time.
> 
> I'm not sure what you mean by speed no longer being the reason to use 
> the Unix version. Do you mean that people using the Unix version don't 
> mind it being much slower than the Windows version? Well I'd like to see 
> it come close at least. The faster the render, the better ;)

You can no longer get faster rendering by using Unix. The compiler used 
makes more of a difference than the OS, so rendering speed is no longer 
a reason to use Unix instead of Windows.

-- 
Christopher James Huff <chr### [at] maccom>
POV-Ray TAG e-mail: chr### [at] tagpovrayorg
TAG web site: http://tag.povray.org/

Post a reply to this message

>   Then why don't you just make sure that you are using *all* the same ini
> settings?
>   You can save the whole set of ini settings with a command line parameter.
> You can use this ini file directly as the ini file for the file to be
> rendered.
> 



I copied the ini settings from the benchmark scene into a benchmark.ini 
file and used it for both rendering so this should not be the problem.

As Christopher James Huff said it is an optimization issue of the 
compilers so not in my hands to change ;-)

Hello Windows!

Post a reply to this message

Christopher James Huff wrote:
> In article <3D4### [at] netscapenet>, Roz <Rzl### [at] netscapenet> 
> wrote:
> 
> 
>>Mark Gordon wrote:
>>
>>>Maybe with the Intel compiler, but probably not with gcc.  Speed is no
>>>longer a reason to use the Unix version, and it hasn't been for quite some 
>>>time.
>>
>>I'm not sure what you mean by speed no longer being the reason to use 
>>the Unix version. Do you mean that people using the Unix version don't 
>>mind it being much slower than the Windows version? Well I'd like to see 
>>it come close at least. The faster the render, the better ;)
> 
> 
> You can no longer get faster rendering by using Unix. The compiler used 
> makes more of a difference than the OS, so rendering speed is no longer 
> a reason to use Unix instead of Windows.
> 

So what you're really saying is that the OS doesn't matter anymore for 
rendering speed or if it did, the difference is negligible compared to 
what the different compilers generate. Of course, the compiler flags 
used would possbily have an impact.

I've only custom compiled the sources for 3.1g and MegaPOV 0.7 on Linux. 
Both showed very good rendering speed increases, faster than the 
respective precompiled Windows versions even. But a custom compiled 
version 3.5 using gcc may not end up faster than the Windows version 
anymore. I'm hoping it'll at least get closer ;) The newer version of 
gcc, as Warp suggested, or the Intel compiler may also produce better 
results. I just have to bring myself to download the 58meg file :P

I think what bothered me about Mark's statement is that it sounded like 
there's no point trying to provide an optimized Unix binary because 
it'll never be faster than the Windows version anyway. I realize Mark's 
very busy and it's been really nice to have a Unix binary version to use 
prior to the source code being released. Hopefully compiling from source 
will boost that rendering speed up!

-Roz

Post a reply to this message

On Thu, 01 Aug 2002 04:08:46 -0400, Roz wrote:

> I think what bothered me about Mark's statement is that it sounded like
> there's no point trying to provide an optimized Unix binary because
> it'll never be faster than the Windows version anyway. I realize Mark's
> very busy and it's been really nice to have a Unix binary version to use
> prior to the source code being released. Hopefully compiling from source
> will boost that rendering speed up!
> 
> -Roz
 
Based on the numbers I've seen, I'm starting to rethink this.  I'm still
running Red Hat 7.2, though, so building with gcc 3.1 isn't going to be
trivial for me.  I'll look into it, though.

-Mark Gordon

Post a reply to this message