Before I post my actual timings, I'd just like to air my preliminary gut 
reaction that my properly-installed/downloaded 3.5 binary on Libranet 
Linux (chiefly based on Debian woody, with 2.4.x kernel) running on a 
Celeron 400 with 128 MB RAM, is SLOWER than the Povray 3.0.xx .deb that 
came with Debian potato.   I may have to revert, unless I can be 
convinced otherwise.  (I'm not an advanced user, and I'm struggling 
along on one particular project, so I may not even notice the 
improvements for awhile....

Also, my quality param. (Q2 for rough views just to check obj. 
placement) doesn't seem to make any difference.  Should I copy a 
povray.ini to my home dir. and tweak it for my usual settings?

TIA as always,
Mark

Post a reply to this message

Try downloading the source and compiling it.
  Due to a human mistake the official binary was compiled with no
optimizations, which makes it awfully slow. (I don't know if an updated
binary has already been made available.)

-- 
#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

Warp wrote:
>   Try downloading the source and compiling it.
>   Due to a human mistake the official binary was compiled with no
> optimizations, which makes it awfully slow. (I don't know if an updated
> binary has already been made available.)
> 

I noticed Mark Gordon posted a note about placing a new Linux binary
up to address the I/O restriction problem. I downloaded it a little while
ago and ran benchmark.pov on it and it is a lot faster than the older
non-optimized version was. It's not as fast as can be achieved with
doing your own compile using various optimization settings but it's
a definite improvement. Thanks Mark :)

Times are as follows:

Windows prebuilt         = 25m 48s (1548s)
older Linux prebuilt     = 55m 40s (3340s)
new (8/7) Linux prebuilt = 34m 10s (2050s)
gcc3 athlon optimized    = 24m 19s (1459s)

-Roz

Post a reply to this message

On Wed, 07 Aug 2002 23:40:41 -0400, Roz wrote:

> Warp wrote:
>>   Try downloading the source and compiling it. Due to a human mistake
>>   the official binary was compiled with no
>> optimizations, which makes it awfully slow. (I don't know if an updated
>> binary has already been made available.)
>> 
>> 
> I noticed Mark Gordon posted a note about placing a new Linux binary up
> to address the I/O restriction problem. I downloaded it a little while
> ago and ran benchmark.pov on it and it is a lot faster than the older
> non-optimized version was. It's not as fast as can be achieved with
> doing your own compile using various optimization settings but it's a
> definite improvement. Thanks Mark :)
> 
> Times are as follows:
> 
> Windows prebuilt         = 25m 48s (1548s) older Linux prebuilt     =
> 55m 40s (3340s) new (8/7) Linux prebuilt = 34m 10s (2050s) gcc3 athlon
> optimized    = 24m 19s (1459s)
> 
> -Roz

The error Warp notes in the original binary has been fixed in the updated
one.  Additional optimizations (e.g. Pentium II optimizations) may
sacrifice portability, so I'm not using them in the official binary at
this point.

-Mark Gordon

Post a reply to this message

Mark Gordon wrote:
> On Wed, 07 Aug 2002 23:40:41 -0400, Roz wrote:
> The error Warp notes in the original binary has been fixed in the updated
> one.  Additional optimizations (e.g. Pentium II optimizations) may
> sacrifice portability, so I'm not using them in the official binary at
> this point.
> 
> -Mark Gordon


Thought ya did that based on the speed I saw last night when rendering
benchmark.pov with the new binary (dated 8/7). I agree with the need
to keep it portable. That's probably why it's statically linked too.
Even so, your "compatible" optimizations improved the speed from
55m to 34m. I think that's a really good improvement!

You could also try the Intel compiler at some point if you're feeling
adventurous. From looking at Micha's speed postings, the Intel compile
was a bit faster than the gcc one.

-Roz

Post a reply to this message

But for those of us who don't know our way around source code, how easy
is it to find the file(s) to optimize, and make the appropriate
changes.  I'm not just a newbie to POV, I'm a total stranger in a
strange land when it comes to hacking.
Is "rolling my own" a realistic option, then?
--Mark W.

Roz wrote:
> 
> Warp wrote:
> >   Try downloading the source and compiling it.
> >   Due to a human mistake the official binary was compiled with no
> > optimizations, which makes it awfully slow. (I don't know if an updated
> > binary has already been made available.)
> >
> 
> I noticed Mark Gordon posted a note about placing a new Linux binary
> up to address the I/O restriction problem. I downloaded it a little while
> ago and ran benchmark.pov on it and it is a lot faster than the older
> non-optimized version was. It's not as fast as can be achieved with
> doing your own compile using various optimization settings but it's
> a definite improvement. Thanks Mark :)
> 
> Times are as follows:
> 
> Windows prebuilt         = 25m 48s (1548s)
> older Linux prebuilt     = 55m 40s (3340s)
> new (8/7) Linux prebuilt = 34m 10s (2050s)
> gcc3 athlon optimized    = 24m 19s (1459s)
> 
> -Roz

Post a reply to this message

Mark M. Wilson wrote:
> But for those of us who don't know our way around source code, how easy
> is it to find the file(s) to optimize, and make the appropriate
> changes.  I'm not just a newbie to POV, I'm a total stranger in a
> strange land when it comes to hacking.
> Is "rolling my own" a realistic option, then?
> --Mark W.

If you're unfamiliar with Makefiles and editing source files, such as
optout.h, then it would probably be easier for you to download and
use one of the prebuilt optimized binaries others have provided rather
than roll your own. Or at the very least, try downloading and installing
the new Linux binary from:

http://www.povray.org/download

It will be much faster than what you currently have and installation
should be straightforward (instructions are on the download page).

Micha Riser has provided some prebuilt binaries at the following url:

http://www.povworld.org/povray/binaries.html

Unfortunately, I cannot pull this page up right now to see what cpus
have been targeted. You have a Celeron and I'm not sure if the Pentium
optimized ones will work. It won't hurt to try though (except for
bandwidth). Just extract a prebuilt binary into its own directory and
run it from there with './povray [options] [scene-file]' (no quotes).

But at the very least, try the new one from POV-Ray's site and save
yourself some rendering time.

-Roz

Post a reply to this message

Roz wrote:
> 
> Micha Riser has provided some prebuilt binaries at the following url:
> 
> http://www.povworld.org/povray/binaries.html
> 
> Unfortunately, I cannot pull this page up right now to see what cpus
> have been targeted. 

Unfortunately the server's connection is broken and I have no physical 
access to it till Monday :( So be a bit patient.

> You have a Celeron and I'm not sure if the Pentium
> optimized ones will work. It won't hurt to try though (except for
> bandwidth). Just extract a prebuilt binary into its own directory and
> run it from there with './povray [options] [scene-file]' (no quotes).

I am not sure if the Pentium III-optimized binary runs on Celerons but I 
suppose it does (at least on some, AFAIK there are quite some different 
sub-types of Celeron CPUs). Otherwise there is also a version for generic 
pentium compatible cpus (which probably is still quite a bit faster than 
the new official).

- Micha

-- 
http://objects.povworld.org - the POV-Ray Objects Collection

Post a reply to this message

Thanks! But now the power supply on my box is dead, so it will be
awhile.... >:-(
--Mark W.

Roz wrote:
> 
> Mark M. Wilson wrote:
> > But for those of us who don't know our way around source code, how easy
> > is it to find the file(s) to optimize, and make the appropriate
> > changes.  I'm not just a newbie to POV, I'm a total stranger in a
> > strange land when it comes to hacking.
> > Is "rolling my own" a realistic option, then?
> > --Mark W.
> 
> If you're unfamiliar with Makefiles and editing source files, such as
> optout.h, then it would probably be easier for you to download and
> use one of the prebuilt optimized binaries others have provided rather
> than roll your own. Or at the very least, try downloading and installing
> the new Linux binary from:
> 
> http://www.povray.org/download
> 
> It will be much faster than what you currently have and installation
> should be straightforward (instructions are on the download page).
> 
> Micha Riser has provided some prebuilt binaries at the following url:
> 
> http://www.povworld.org/povray/binaries.html
> 
> Unfortunately, I cannot pull this page up right now to see what cpus
> have been targeted. You have a Celeron and I'm not sure if the Pentium
> optimized ones will work. It won't hurt to try though (except for
> bandwidth). Just extract a prebuilt binary into its own directory and
> run it from there with './povray [options] [scene-file]' (no quotes).
> 
> But at the very least, try the new one from POV-Ray's site and save
> yourself some rendering time.
> 
> -Roz

Post a reply to this message

Roz wrote:
> 
 Just extract a prebuilt binary into its own directory and
> run it from there with './povray [options] [scene-file]' (no quotes).
> 

Ugh, would I always have to run it from its own directory?  I suppose
adding said dir. would be too confusing, assuming I hadn't uninstalled
the other binary, and that's the reason?

> But at the very least, try the new one from POV-Ray's site and save
> yourself some rendering time.

Am I correct in thinking that the new on the POV-Ray site is a "revised"
3.5?

--Mark W.

Post a reply to this message