LuxRender 0.9-dev Windows build of 06-04-2012

[Lord Crc]

New month, new weeky builds

Primary new feature is the merging of the statistics rewrite provided by our newly promoted developer Omniflux. The goal was to provide tool tips for each statistic in the GUI, but that required a big change in the internals. The relevant thread is here. The "short" statistics printed by luxconsole and pylux has been changed somewhat, please provide feedback on this if you'd prefer it to look different.

Lots of fixes since last time, primarily to SQBVH and SPPM. I've also tried to make the slave reconnection logic better, so that it will only try to establish a new session if the slave explicitly rejects the reconnection command.

Note that there's still some issue with bidir being darker than path. Hopefully jeanphi will resolve the issue shortly one way or another, I'll provide updated builds then.

These builds contain LuxRender, LuxConsole, LuxMerger, LuxBlend25 and Pylux for Blender 2.6x, and LuxBlend for Blender 2.49.
The NoOpenCL builds does not require any OpenCL driver or capabilities, but also does not use the GPU for the "hybrid" renderer.

LuxRender_CVS_20120406_64bit_OpenCL.exe
LuxRender_CVS_20120406_64bit_NoOpenCL.exe
LuxRender_CVS_20120406_32bit_SSE2_OpenCL.exe
LuxRender_CVS_20120406_32bit_SSE2_NoOpenCL.exe
LuxRender_CVS_20120406_32bit_SSE1_NoOpenCL.exe

Built with Visual Studio 2008 SP1 + Windows 7 SDK on Win 7 64bit, using AMD APP SDK 2.5 (formerly ATI Stream SDK).

Changelog:

Code: Select all

77 minutes ago   Asbjørn Heid   Fix to slave reconnection logic.
101 minutes ago   Asbjørn Heid   Changed MagnitudeReduce and MagnitudePrefix to reduce at 1000 instead of 2000.
107 minutes ago   Asbjørn Heid   Changed position of magnitude prefix in formatted statistics to match up with unit.
119 minutes ago   Asbjørn Heid   Reuse existing statistics labels in GUI.
3 hours ago   Asbjørn Heid   Merged statistics_rewrite branch.
8 days ago   Guillaume Bouchard   SPPM: small convention issue.
8 days ago   Jason Clarke   SPPM: Fix a build issue on OS X
8 days ago   Guillaume Bouchard   SPPM: change default values
8 days ago   Guillaume Bouchard   SPPM: add one parameter *storeglossy*
8 days ago   Guillaume Bouchard   SPPM: store the bounce/store component in HitPoints class
8 days ago   Guillaume Bouchard   SPPM: fix direct light sampling issue with arealigh
8 days ago   Guillaume Bouchard   SPPM: remove the glossy threshold behavior
8 days ago   Guillaume Bouchard   SPPM: fix direct light sampling with passthrough material
8 days ago   Guillaume Bouchard   SPPM: reduce photon data footprint
8 days ago   Guillaume Bouchard   SPPM: factorise accel code
8 days ago   Guillaume Bouchard   SPPM: factorise data
13 days ago   Asbjørn Heid   Changed boost::thread::sleep to boost::this_thread:sleep.
13 days ago   David Bucciarelli   A small optimization in the way the SQBVH split axis is choosen
2 weeks ago   Guillaume Bouchard   Typo preventing compilation on non-windows with boost > 1.48.
2 weeks ago   Jens Verwiebe   When controlling exports, also the base-class must be exported ( patch by homer_s thx )
2 weeks ago   David Bucciarelli   Added a comment to explain last commit
2 weeks ago   David Bucciarelli   Merge
2 weeks ago   David Bucciarelli   Fixed a numerical stability problem in SQBVH code
2 weeks ago   Jens Verwiebe   OSX/cmake: we must use VERSION_LESS instead LESS it seems for ${XCODE_VERSION}
2 weeks ago   David Bucciarelli   A minor cleanup in SQBVH code
2 weeks ago   David Bucciarelli   Fixed a problem with Visual C++ and SQBVH code
2 weeks ago   David Bucciarelli   Fixed a SQBVH problem with tree made by a single leaf
2 weeks ago   David Bucciarelli   Fixed a numerical related problem in SQBVH object split code
2 weeks ago   David Bucciarelli   Removed "-O3" from CMake definitions in order to have CMake Release/Debug options to work
3 weeks ago   David Bucciarelli   Fixed a problem in handling no-triangle primitives with SQBVH
3 weeks ago   Jean-Philippe Grimaldi   A forgotten fix in previous commit
3 weeks ago   Jean-Philippe Grimaldi   Fix some glotches in the new bidirectional code.
3 weeks ago   Asbjørn Heid   Fixed previous cleanup to film.cpp
3 weeks ago   Asbjørn Heid   Fixed indentation in film.cpp
3 weeks ago   Asbjørn Heid   Minor code cleanup in film.cpp
3 weeks ago   Asbjørn Heid   Fixed haltspp/halttime not properly stopping the renderer when using SPPM.
3 weeks ago   Jean-Philippe Grimaldi   Fix hybrid bidir with new light sampling code.
3 weeks ago   David Bucciarelli   Changed the default value for Opencl workgroup size (from 0 to 64)
3 weeks ago   Asbjørn Heid   Only pull in the atomic functions we need, not entire namespace.
3 weeks ago   Asbjørn Heid   Fixed erronous negation of isnan on windows.
3 weeks ago   Amir Mohammadkhani-Aminabadi   Fixed two warnings when compiling with MSVC
3 weeks ago   Amir Mohammadkhani-Aminabadi   Boost changed interprocess::detail to interprocess::ipcdetail
3 weeks ago   Jens Verwiebe   Fix a division-by-zero crash when scene contains no mesh
4 weeks ago   Asbjørn Heid   Fixed issue with getting queryable string attributes.
4 weeks ago   Jean-Philippe Grimaldi   Use light strategy for bidirectional direct light sampling.
4 weeks ago   Jens Verwiebe   Fix some overloaded function warnings
4 weeks ago   Jean-Philippe Grimaldi   A few cosmetic changes.
4 weeks ago   Jean-Philippe Grimaldi   One more PartialDifferentialGeometry fix.
4 weeks ago   Asbjørn Heid   Fixed issue with erronous override of Primitive::Pdf().
4 weeks ago   Jens Verwiebe   Silence another warning by making dtor for GenericQueryableAttribute virtual
4 weeks ago   Jens Verwiebe   ome warning fixes and bugs:
4 weeks ago   Jean-Philippe Grimaldi   Reorder bidirectional code to prepare for light sampling improvements.
4 weeks ago   Jean-Philippe Grimaldi   Fix a nasty typo spotted by Rom1.


Hope you enjoy

[A-man]

Y U NO start RC cycle yet?
Haha just kidding, I've been looking forward to new weeklies for a while

[Lord Crc]

Y U NO start RC cycle yet?

Waiting for jeanphi to resolve The Darkening in bidir

[SLKabaker]

Thank you very much for posting a new weekly. I greatly appreciate your time, effort and work on this. I look forward to getting back to rendering!

[nenad73]

Hej, I just wonder is there 64bit opencl and sse2 supported version? (is it faster?)
I also wonder why support sse1, its very old (just a thought)

[Lord Crc]

Hej, I just wonder is there 64bit opencl and sse2 supported version? (is it faster?)
I also wonder why support sse1, its very old (just a thought)

The x64 platform has SSE1 and SSE2 as "native" instructions, in fact you must use SSE, the regular FPU instructions are not available in x64.

I asked if people were still interested in SSE1 support some time ago, and a few guys said yes. It costs me nothing extra to make them so

[Meelis]

Hi

Is there changes for SPPM like network rendering support (last time SPPM was only for sigle computer).
SSS materials are not yet supported i assume?

[SATtva]

SPPM more or less works in network mode, although radius reduction and photons per pass are inconsistent between the nodes. No participating media yet.

[HoloFan]

First (I can't say it often enough ) Luxrender is just awesome!

Sadly I still have to report to bugs/problems I have.
The first one is that I currently render an animation with a bunch of scenes. And on Windows 8 (hybrid path, ati) after some scenes in queuing mode luxrender crashes with a MS Viusal C++ error ("The application has tried to terminate the runtime in an unsual way") or Windows just states Lux wouldn't be responding anymore.

I encountered the second problem on three different configurations. When I start a rendering with hybrid bidir, luxrender allocates so many new integrator states that the memory isn't enough for luxrender. The scenes I tested don't come anywhere near memory limits with hybrid path tracing.
On my Windows 7 laptop (with quite old drivers) the complete OS crashes because (as far as I see) Windows isn't able to manage memory demands correctly and starts swapping extensively.
On my desktop PC with Windows 8 (newest preview drivers) luxrender quits after about 10 seconds. On Windows XP it crashes with a runtime error ("The application has tried to terminate the runtime in an unsual way"). One time it started rendering with XP successfully, but the cpu was at about 0-6% and there were only very few rays traced.

My Systems:
Laptop:
Windows 7
Intel Core i5 2520M
AMD Radeon HD 6630M
Driver Version 8.862.4.3-110926a-126759C-Sony
4 GB Ram

Desktop:
Windows 7 (latest Preview Driver)/ Windows XP (ATI Catalyst 12.3)
2 GB Ram
Amd Athlon X2
HD 6670

I hope this post fits here.

[SATtva]

The first one is that I currently render an animation with a bunch of scenes. And on Windows 8 (hybrid path, ati) after some scenes in queuing mode luxrender crashes with a MS Viusal C++ error ("The application has tried to terminate the runtime in an unsual way") or Windows just states Lux wouldn't be responding anymore.

Have you observed Lux process memory consumption with Windows System Monitor during these tests? I was under impression Dade has successfully fixed a memory leak in QBVH code which lead to this problem in the past. (I'm assuming you've used this particular build, not some former one.)