There are several programs that convert LDraw models to POV-Ray. I'm 
wondering if there are differences in memory usage between the various 
types of output. How would I measure the memory usage? Would it be 
better to measure peak usage? Average usage over time? Thanks.


Mike

Post a reply to this message

On 1/30/2018 2:46 PM, Mike Horvath wrote:
> There are several programs that convert LDraw models to POV-Ray. I'm 
> wondering if there are differences in memory usage between the various 
> types of output. How would I measure the memory usage? Would it be 
> better to measure peak usage? Average usage over time? Thanks.
> 
> 
> Mike

I meant memory usage inside POV-Ray, if I was not clear.


Mike

Post a reply to this message

Le 18-01-30 à 14:46, Mike Horvath a écrit :
> There are several programs that convert LDraw models to POV-Ray. I'm 
> wondering if there are differences in memory usage between the various 
> types of output. How would I measure the memory usage? Would it be 
> better to measure peak usage? Average usage over time? Thanks.
> 
> 
> Mike

They all convert to mesh/mesh2 from the same data.
The only difference would be if you perform any subdivision step prior 
to the conversion itself.

Post a reply to this message

On 1/31/2018 6:37 PM, Alain wrote:
> Le 18-01-30 à 14:46, Mike Horvath a écrit :
>> There are several programs that convert LDraw models to POV-Ray. I'm 
>> wondering if there are differences in memory usage between the various 
>> types of output. How would I measure the memory usage? Would it be 
>> better to measure peak usage? Average usage over time? Thanks.
>>
>>
>> Mike
> 
> They all convert to mesh/mesh2 from the same data.
> The only difference would be if you perform any subdivision step prior 
> to the conversion itself.

Okay thanks.

Mike

Post a reply to this message

Am 01.02.2018 um 00:37 schrieb Alain:
> Le 18-01-30 à 14:46, Mike Horvath a écrit :
>> There are several programs that convert LDraw models to POV-Ray. I'm
>> wondering if there are differences in memory usage between the various
>> types of output. How would I measure the memory usage? Would it be
>> better to measure peak usage? Average usage over time? Thanks.
>>
>>
>> Mike
> 
> They all convert to mesh/mesh2 from the same data.
> The only difference would be if you perform any subdivision step prior
> to the conversion itself.

Some programs can make use of the LGEO parts library; in that case the
result is CSG for any part that is in the library.

For small models, or models that use a gazillion of different parts,
that may be more memory-efficient. Haven't tested it though. The main
advantage of the LGEO parts library is that it is more precise.

Post a reply to this message

Mike Horvath <mik### [at] gmailcom> wrote:
> There are several programs that convert LDraw models to POV-Ray. I'm
> wondering if there are differences in memory usage between the various
> types of output. How would I measure the memory usage? Would it be
> better to measure peak usage? Average usage over time? Thanks.

If high memory usage is your problem, I think it is mainly caused by the LGEO
library as an OpenGL mesh export only (as done by my LDCad) uses just 32MB (yes
mega not giga) during rendering in POV-Ray on my system.

This using your datsville scene (rev528, datsville_master/datsville.ldr)

But I have seen it use over 20GB in the past when doing a LGEO based export
through LDView's export).

Post a reply to this message

On 2/1/2018 4:48 PM, Roland Melkert wrote:
> Mike Horvath <mik### [at] gmailcom> wrote:
>> There are several programs that convert LDraw models to POV-Ray. I'm
>> wondering if there are differences in memory usage between the various
>> types of output. How would I measure the memory usage? Would it be
>> better to measure peak usage? Average usage over time? Thanks.
> 
> If high memory usage is your problem, I think it is mainly caused by the LGEO
> library as an OpenGL mesh export only (as done by my LDCad) uses just 32MB (yes
> mega not giga) during rendering in POV-Ray on my system.
> 
> This using your datsville scene (rev528, datsville_master/datsville.ldr)
> 
> But I have seen it use over 20GB in the past when doing a LGEO based export
> through LDView's export).
> 
> 

That's odd, because I run out of memory (8GB) unless I use LDBoxer first 
to cut down on triangles.


Mike

Post a reply to this message

Here is the top of my POV-Ray file.


Mike




// Generated by L3P v1.4 BETA 20131202 (Win32)  (C) 1998-2013 Lars C. 
Hassing SP.### [at] AMhassingsdk
// L3P website:  http://www.hassings.dk/l3/l3p.html
// LDraw.org:    http://www.ldraw.org
// Date:         Thu Jan 18 04:57:32 2018
// Input file: 
D:\Datsville\datsville_inlined\datsville_rev531_noinline_boxed_none.xmpd
// Title:        Datsville
// Author:       John VanZwieten, Michael Horvath, Tore Eriksson
// ModelDir:     D:\Datsville\datsville_inlined
// WorkDir:      D:\Datsville\helper_scripts\povray
// LDRAWDIR:     C:\Users\Public\Documents\LDraw (LDRAWDIR environment 
variable)
// SearchDirs:   (Default)
// SearchDir1:   D:\Datsville\datsville_inlined\ (0x10)
// SearchDir2:   C:\Users\Public\Documents\LDraw\P\ (0x9)
// SearchDir3:   C:\Users\Public\Documents\LDraw\PARTS\ (0x4)
// ldconfig.ldr: C:\Users\Public\Documents\LDraw\ldconfig.ldr
// Commandline:  "E:\Programs Windows\LDraw\l3p14beta\L3P20131202.EXE" 
D:\Datsville\datsville_inlined\datsville_rev531_noinline_boxed_none.xmpd 
-o l3p_datsville_rev531_noinline_boxed_none.pov -il -ic

Post a reply to this message

Mike Horvath <mik### [at] gmailcom> wrote:
> Here is the top of my POV-Ray file.
> // Generated by L3P v1.4 BETA 20131202 (Win32)  (C) 1998-2013 Lars C.

This is the one I rendered yesterday (@1920x1080 in about 12 mins using only
32MB)
http://www.melkert.net/action/download/datsville528MasterPov.7z
6MB 7zip containing 84MB pov file

Haven't used L3P in a very long time, but I think L3P generated meshes for every
single .dat, in LDCad I generate meshes per high level part (basically just
dumping the OpenGL data).

This export was done with the upcomming LDCad 1.6b, but you can also use 1.6a to
make your own export. You only might need to fix some missing object references
resulting from empty models used (I fixed that in 1.6b)

Post a reply to this message

On 2/2/2018 4:10 PM, Roland Melkert wrote:
> Mike Horvath <mik### [at] gmailcom> wrote:
>> Here is the top of my POV-Ray file.
>> // Generated by L3P v1.4 BETA 20131202 (Win32)  (C) 1998-2013 Lars C.
> 
> This is the one I rendered yesterday (@1920x1080 in about 12 mins using only
> 32MB)
> http://www.melkert.net/action/download/datsville528MasterPov.7z
> 6MB 7zip containing 84MB pov file
> 
> Haven't used L3P in a very long time, but I think L3P generated meshes for every
> single .dat, in LDCad I generate meshes per high level part (basically just
> dumping the OpenGL data).
> 
> This export was done with the upcomming LDCad 1.6b, but you can also use 1.6a to
> make your own export. You only might need to fix some missing object references
> resulting from empty models used (I fixed that in 1.6b)
> 
> 

Wow, that is so much better! I don't even have to inline the model or 
replace parts with boxes.

The render time is roughly the same. But the parse time is just a few 
seconds versus 10-15 minutes. And it barely uses any memory!

Thanks so much!


Mike

Post a reply to this message