LuxRender 0.9-dev Windows build of 05-12-2011

[Lord Crc]

Sorry about the delayed "weekly". I had actually made and uploaded a build some 10 days ago, but forgot to make a post about it! Shame on me!

Some bigger changes this time around then.

Jeanphi has made some changes to light sampling, which means that for now the shadow ray count parameter isn't currently used. He's also fixed a long-standing weighting error which affects the look of glossy and metal materials.

edit: updated builds with light sampling bug fix from jeanphi

I have added some new code which allows for complex paths for motion blur, as well as a unified way of specifying motion data for camera and instances.

In addition some general fixes and improvements have been made. Layered and glossycoating material have received bugfixes. Normal mapping should work much more robustly now on long, thin geometry, as well as be faster. Fixed a race condition in Hybrid renderer, causing crashes. Pylux can now report "percent complete" in Blender.


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_20111205_64bit_OpenCL.exe
LuxRender_CVS_20111205_64bit_NoOpenCL.exe
LuxRender_CVS_20111205_32bit_SSE2_OpenCL.exe
LuxRender_CVS_20111205_32bit_SSE2_NoOpenCL.exe
LuxRender_CVS_20111205_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

45 minutes ago   Jean-Philippe Grimaldi   Hopefully definitive fix for the new light sampling code.
25 hours ago   Asbjørn Heid   Fixed network support for new motion block code.
36 hours ago   Asbjørn Heid   Introduced motion transforms, specified using new motion block.
41 hours ago   Jean-Philippe Grimaldi   Wrong patch commited, fix typos.
42 hours ago   Jean-Philippe Grimaldi   New light sampling interfaces:
42 hours ago   Jean-Philippe Grimaldi   Add a new Pdf member to Distribution1D to get the probability of a given element.
42 hours ago   Jean-Philippe Grimaldi   Improve BxDF/BSDF comments consistency.
5 days ago   Asbjørn Heid   Use actual walltime to determine network samples/sec.
6 days ago   Guillaume Bouchard   SPPM: better struct organisation to reduce memory usage
8 days ago   Jean-Philippe Grimaldi   Update bxdf.h documentation.
9 days ago   Asbjørn Heid   Fixed race condition in Path::RequestSamples, causing heap corruption with hybrid renderer.
10 days ago   Asbjørn Heid   Cleaned up Quaternion class.
10 days ago   Asbjørn Heid   Fixed nasty bug in MotionSystem::Bound() and cleaned up MotionSystem code a bit.
2 weeks ago   dougal2   Add percentComplete statistic - will be chosen from haltspp or halttime by logic in StatsData
2 weeks ago   Jean-Philippe Grimaldi   Fix various Schlick model pdf computations and update glossy translucent to
2 weeks ago   dougal2   lxs_to_lbm2: old changes, forgotten commit; starting implementation of graph
2 weeks ago   Asbjørn Heid   Added proper implementation of SchlickBSDF::rho()
2 weeks ago   Asbjørn Heid   Renamed DEBUG define in epsilon.h
2 weeks ago   Asbjørn Heid   Removed obsolete BxDF SchlickGlossyBRDF.
2 weeks ago   Asbjørn Heid   Refactored SchlickBSDF to take a MicrofacetDistribution instead of a BxDF for the coating layer.
2 weeks ago   Jens Verwiebe   Give Networkicon a slightly smaller size to fit other icons
2 weeks ago   David Washington   added JtheNinja's icons
2 weeks ago   Jens Verwiebe   OSX: take back crash unrelated last commit, needs more investigation
2 weeks ago   Jens Verwiebe   OSX: adaption to Blender gcc-4.6 builds, does not play with LTO'ed pylux
2 weeks ago   Asbjørn Heid   Templated matrix3x3.h routines, made singularity detection in Invert3x3 more robust.
2 weeks ago   Asbjørn Heid   New and more robust method for recovering derivatives from normal map.
2 weeks ago   Jens Verwiebe   OSX: add ${LUX_LIBRARY_DEPENDS} to all binaries, needed for boost >= 1.46
2 weeks ago   Jean-Philippe Grimaldi   Update layered material for coding style and try to improve some of the algorithms.
2 weeks ago   Jean-Philippe Grimaldi   Some enhancement to the new glossycoating material.
3 weeks ago   Jean-Philippe Grimaldi   Merge
3 weeks ago   Jean-Philippe Grimaldi   Clean up the BSDF mess.


Hope you enjoy!

[Pilchard123]

What's the difference between the SSE1 and SSE2 builds?

[SATtva]

If your CPU has support for SSE2 instruction set (which is most likely the case -- a CPU must be more than 7 years old to not have it), then rendering speed with the latter build will be slightly better.

[Pilchard123]

Aha. I only just have SSE2 support. Athlon 64 just sneaks in under the wire. Does anyone here still have a non-SSE2 CPU?

[Lord Crc]

Does anyone here still have a non-SSE2 CPU?

Last time I asked, a few did. It's not much hassle supporting it so why not

[Pilchard123]

What does the "percentage complete" actually mean? Does it only show that if you give halting criteria (haltspp/time), or does it show anyway?

[Lord Crc]

Does it only show that if you give halting criteria (haltspp/time)

That's my understanding.

[Jack-Jack]

If I try to run luxrender win64 NoOpenCL I get message that my system doesn't have winhoard.dll. So I cannot run luxrender.exe

[MaRo]

If I try to run luxrender win64 NoOpenCL I get message that my system doesn't have winhoard.dll. So I cannot run luxrender.exe

Hi, same here. Just to let you know. Previous build of same flavour was fine (i.e. win64 no OpenCL, can't tell for other builds).

[Lord Crc]

So sorry guys, I was trying out Hoard memory allocator, which supposedly is much faster, and I forgot to remove it from one of the build configurations. I'm making an updated build now.

edit: for the record, the dll just crashes lux, so it was a failed experiment.

edit2: should only affect x64 NonOpenCL build.