|
|
|
|
|
|
| |
| |
|
|
From: Sebastian H
Subject: Linux version is slower than windows version?
Date: 30 Jul 2002 11:07:51
Message: <3D46AC6C.3070703@web.de>
|
|
|
| |
| |
|
|
Hi there
I just compared the rendering speed of the official windows and linux
version of POV3.5 and got the very strage result that the linux version
takes the double amount of time (2.11x) the windows version needs to
render the benchmark scene (scenes/advanced).
For both OSes I used the scene from the linux version and the ini
settings given in the file (benchmark.pov).
Result:
_Windows_:
Ray->Shape Intersection Tests Succeeded Percentage
----------------------------------------------------------------------------
Box 79358745 9334253 11.76
Cone/Cylinder 78379572 6515873 8.31
CSG Intersection 162923001 55142852 33.85
CSG Merge 741865 34312 4.63
Fractal 1831644 104482 5.70
Height Field 3901191 105560 2.71
Height Field Box 3901191 707900 18.15
Height Field Triangle 3376757 108905 3.23
Height Field Block 5889587 1742423 29.58
Height Field Cell 23435856 1852284 7.90
Isosurface 11669415 721134 6.18
Isosurface Container 12154734 11669868 96.01
Isosurface Cache 176199 42446 24.09
Mesh 12295275 64396 0.52
Plane 85284171 1281858 1.50
Sphere 260332524 153124522 58.82
Superellipsoid 620681 43704 7.04
Torus 2887283 411263 14.24
Torus Bound 2887283 480972 16.66
True Type Font 838010 80727 9.63
Clipping Object 2557372 1524945 59.63
Bounding Box 653803865 124207145 19.00
Vista Buffer 21535803 11650388 54.10
----------------------------------------------------------------------------
Isosurface roots: 11664259
Function VM calls: 169416006
----------------------------------------------------------------------------
Roots tested: 480972 eliminated: 274031
Calls to Noise: 4512422082 Calls to DNoise: 2404216638
----------------------------------------------------------------------------
Media Intervals: 36350371 Media Samples: 327887642 (9.02)
Shadow Ray Tests: 118768042 Succeeded: 48914570
Reflected Rays: 223228 Total Internal: 470
Refracted Rays: 145054
Transmitted Rays: 602889
Number of photons shot: 74025
Surface photons stored: 64488
Priority queue insert: 2226874
Priority queue remove: 163988
Gather function called: 659394
----------------------------------------------------------------------------
Smallest Alloc: 25 bytes Largest: 1440024
Peak memory used: 6401076 bytes
----------------------------------------------------------------------------
Time For Parse: 0 hours 0 minutes 6.0 seconds (6 seconds)
Time For Photon: 0 hours 1 minutes 59.0 seconds (119 seconds)
Time For Trace: 1 hours 16 minutes 13.0 seconds (4573 seconds)
Total Time: 1 hours 18 minutes 18.0 seconds (4698 seconds)
POV-Ray finished
_Linux_:
Ray->Shape Intersection Tests Succeeded Percentage
----------------------------------------------------------------------------
Box 77632067 9442332 12.16
Cone/Cylinder 76120049 6536461 8.59
CSG Intersection 161344159 55414898 34.35
CSG Merge 807486 34776 4.31
Fractal 1764372 102306 5.80
Height Field 3820218 105458 2.76
Height Field Box 3820218 705167 18.46
Height Field Triangle 3369358 108722 3.23
Height Field Block 5876169 1734582 29.52
Height Field Cell 23360039 1850304 7.92
Isosurface 11757862 728681 6.20
Isosurface Container 12260392 11758389 95.91
Isosurface Cache 148786 41647 27.99
Mesh 12352449 65016 0.53
Plane 85650538 1287339 1.50
Sphere 261625223 155438827 59.41
Superellipsoid 633727 44548 7.03
Torus 3018587 424283 14.06
Torus Bound 3018587 496226 16.44
True Type Font 816332 80799 9.90
Clipping Object 2572219 1535917 59.71
Bounding Box 501031242 147127728 29.36
Vista Buffer 22076619 12692365 57.49
----------------------------------------------------------------------------
Isosurface roots: 11752486
Function VM calls: 170441799
----------------------------------------------------------------------------
Roots tested: 496226 eliminated: 280279
Calls to Noise: 4531067372 Calls to DNoise: 2415593890
----------------------------------------------------------------------------
Media Intervals: 36494048 Media Samples: 329183857 (9.02)
Shadow Ray Tests: 119312109 Succeeded: 49156296
Reflected Rays: 224291 Total Internal: 1066
Refracted Rays: 146314
Transmitted Rays: 602466
Number of photons shot: 74025
Surface photons stored: 63352
Priority queue insert: 724352
Priority queue remove: 136937
Gather function called: 666007
----------------------------------------------------------------------------
Smallest Alloc: 9 bytes Largest: 1440008
Peak memory used: 5442716 bytes
----------------------------------------------------------------------------
Time For Parse: 0 hours 0 minutes 8.0 seconds (8 seconds)
Time For Photon: 0 hours 3 minutes 39.0 seconds (219 seconds)
Time For Trace: 2 hours 41 minutes 44.0 seconds (9704 seconds)
Total Time: 2 hours 45 minutes 31.0 seconds (9931 seconds)
I use an Athlon900Mhz with 312M Ram
Windows 98SE
SuSE Linux 8.0 (Kernel 2.4.18 is this important???)
Under Linux I ran POV as a user and shut down nearly all other tasks (X,
network, etc.) so top said pov uses 98% CPU power.
Under Windows I used the gui to load the ini and the benchmark file.
(Resolutions and AA are the same :-) () )
My question
Does anyone has similar problems so it is in the binaries or is it just
me and my sytem?
10000/4700 Sekonds is a very bad result!
Sebastian H.
Post a reply to this message
|
|
| |
| |
|
|
From: Warp
Subject: Re: Linux version is slower than windows version?
Date: 30 Jul 2002 11:41:19
Message: <3d46b39e@news.povray.org>
|
|
|
| |
| |
|
|
Sebastian H. <seb### [at] webde> wrote:
> _Windows_:
> Box 79358745 9334253 11.76
> Cone/Cylinder 78379572 6515873 8.31
> _Linux_:
> Box 77632067 9442332 12.16
> Cone/Cylinder 76120049 6536461 8.59
Are you completely and absolutely 100% sure that you are using the exact
same benchmark scene and the exact same ini file settings (and I mean *all*
settings)?
If everything was exactly identical, I don't see any reason for those
numbers to be different.
--
#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
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 30 Jul 2002 11:41:19 -0400, Warp <war### [at] tagpovrayorg> wrote:
> If everything was exactly identical, I don't see any reason for those
> numbers to be different.
Jitter for AA ?
ABX
Post a reply to this message
|
|
| |
| |
|
|
From: Felix Wiemann
Subject: Re: Linux version is slower than windows version?
Date: 30 Jul 2002 12:56:10
Message: <3d46c529@news.povray.org>
|
|
|
| |
| |
|
|
Sebastian H. wrote:
> I just compared the rendering speed of the official windows and linux
> version of POV3.5 and got the very strage result that the linux version
> takes the double amount of time (2.11x) the windows version needs to
> render the benchmark scene (scenes/advanced).
Maybe you should make sure that the general POV-Ray configuration files are
the same for both OS. What about copying povray.ini to $HOME/.povrayrc ?
Also note that they aren't compiled by the same compiler. The windows
version may be much more optimized than the linux version. When sources are
available you can compile your own version with gcc -O3 and some system or
processor specific optimizations.
--
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
|
|
| |
| |
|
|
From: Sebastian H
Subject: Re: Linux version is slower than windows version?
Date: 30 Jul 2002 17:48:30
Message: <3D470A56.3010100@web.de>
|
|
|
| |
| |
|
|
Warp schrieb:
> Sebastian H. <seb### [at] webde> wrote:
>
>>_Windows_:
>>Box 79358745 9334253 11.76
>>Cone/Cylinder 78379572 6515873 8.31
>
>
>>_Linux_:
>>Box 77632067 9442332 12.16
>>Cone/Cylinder 76120049 6536461 8.59
>
>
> Are you completely and absolutely 100% sure that you are using the exact
> same benchmark scene and the exact same ini file settings (and I mean *all*
> settings)?
It was definitly the same .pov file I copied it into a seperate
directory as the .ini file.
The differences you mentioned above made me wonder too but the deviation
here wasn't very high.
A real huge difference is in "Priority queue insert:"
2226874 (Windows)
724352 (Linux)
But what does that mean ?:-)
The other stuff of the POV output of both renderings were quite similar
but I'am not sure about jitter (to ABX) (But since it was the same .ini
file...).
> 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.
A friend of mine who has a bit more knowledge about computer stuff in
general (whatever that means)
said that an optimization that gives you half rendering time would be
_very_ good (to Felix).
I don't know much about this stuff but I like linux and would prefer to
play with POV here (Dualboot sucks if you can't share data without
bending everything to fat).
I'll comapare it again take care of the parameters (hopefully a fault of
mine :-) ), we'll see...
Thanks for comments
Sebastian H.
Post a reply to this message
|
|
| |
| |
|
|
From: Sebastian H
Subject: Re: Linux version is slower than windows version?
Date: 30 Jul 2002 17:50:42
Message: <3D470ADA.5000101@web.de>
|
|
|
| |
| |
|
|
Woops! Should have gone as a reply to Warp (Using Mozilla is so
difficult :-) ).
Sebastian H.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Tue, 30 Jul 2002 12:58:49 -0400, Felix Wiemann wrote:
> Sebastian H. wrote:
>
>> I just compared the rendering speed of the official windows and linux
>> version of POV3.5 and got the very strage result that the linux version
>> takes the double amount of time (2.11x) the windows version needs to
>> render the benchmark scene (scenes/advanced).
>
> Maybe you should make sure that the general POV-Ray configuration files
> are the same for both OS. What about copying povray.ini to
> $HOME/.povrayrc ?
Yes, default run-time settings are different between Windows and
Linux. Maybe I should change that.
> Also note that they aren't compiled by the same compiler. The windows
> version may be much more optimized than the linux version. When sources
> are available you can compile your own version with gcc -O3 and some
> system or processor specific optimizations.
This binary was built with gcc. I may release a binary built with the
Intel compiler at some point.
-Mark Gordon
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Tue, 30 Jul 2002 11:10:36 -0400, Sebastian H. quoth:
> Hi there
>
> I just compared the rendering speed of the official windows and linux
> version of POV3.5 and got the very strage result that the linux version
> takes the double amount of time (2.11x) the windows version needs to
> render the benchmark scene (scenes/advanced).
> I use an Athlon900Mhz with 312M Ram
> Windows 98SE
> SuSE Linux 8.0 (Kernel 2.4.18 is this important???)
> My question
> Does anyone has similar problems so it is in the binaries or is it just
> me and my sytem?
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.
--
Mark
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Tue, 30 Jul 2002 23:51:18 +0200, "Sebastian H." <seb### [at] webde> wrote:
> The other stuff of the POV output of both renderings were quite similar
> but I'am not sure about jitter (to ABX) (But since it was the same .ini
> file...).
From documentation: "Note: the jittering noise is random and non-repeatable"
From benchmark.pov: Jitter_Amount=1 Jitter=On
ABX
Post a reply to this message
|
|
| |
| |
|
|
From: Sebastian H
Subject: Re: Linux version is slower than windows version?
Date: 31 Jul 2002 07:03:38
Message: <3D47C4B2.1050108@web.de>
|
|
|
| |
| |
|
|
>
> 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?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |