Page MenuHome

Sv. Lockal (lockal)
User

Projects

User Details

User Since
Oct 17 2010, 10:55 AM (435 w, 23 h)

Recent Activity

Fri, Jan 25

Sv. Lockal (lockal) committed rBTS4807: Update from https://www.transifex.com/translateblender/blender-ru/blender/ (87%).
Update from https://www.transifex.com/translateblender/blender-ru/blender/ (87%)
Fri, Jan 25, 9:11 PM

Jan 13 2019

Sv. Lockal (lockal) committed rBTS4783: Update from https://www.transifex.com/translateblender/blender-ru/blender/ (86%).
Update from https://www.transifex.com/translateblender/blender-ru/blender/ (86%)
Jan 13 2019, 4:43 PM

Jan 3 2019

Sv. Lockal (lockal) committed rBTS4767: Update from https://www.transifex.com/translateblender/blender-ru/blender/ (83%).
Update from https://www.transifex.com/translateblender/blender-ru/blender/ (83%)
Jan 3 2019, 9:26 PM

Jan 1 2019

Sv. Lockal (lockal) committed rBTS4763: Update from https://www.transifex.com/translateblender/blender-ru/blender/ (80%).
Update from https://www.transifex.com/translateblender/blender-ru/blender/ (80%)
Jan 1 2019, 7:47 PM

Dec 30 2018

Sv. Lockal (lockal) committed rBTS4757: Update from https://www.transifex.com/translateblender/blender-ru/blender/ (79%).
Update from https://www.transifex.com/translateblender/blender-ru/blender/ (79%)
Dec 30 2018, 11:52 AM

Dec 25 2018

Sv. Lockal (lockal) created P878 Get ru translations.
Dec 25 2018, 1:43 AM

Dec 20 2018

Sv. Lockal (lockal) closed T59488: Joining request as Resolved.

I added Michael to https://www.transifex.com/translateblender/blender-ru/ and we agreed to wait for 2.8 and discuss translation improvements meanwhile. So I guess the task is solved now.

Dec 20 2018, 10:39 PM · Translations

Dec 19 2018

Sv. Lockal (lockal) added a comment to T59488: Joining request.

@Michael Soluyanov (crantisz) , glad to see you, you are an awesome guy! I'll contact you today in vk.

Dec 19 2018, 8:28 PM · Translations

Jan 4 2018

Sv. Lockal (lockal) committed rBTS4653: Update ru.po from https://www.transifex.com/translateblender/blender-ru (92%).
Update ru.po from https://www.transifex.com/translateblender/blender-ru (92%)
Jan 4 2018, 9:44 PM

Jan 3 2018

Sv. Lockal (lockal) committed rBTS4652: Update ru.po from https://www.transifex.com/translateblender/blender-ru (91%).
Update ru.po from https://www.transifex.com/translateblender/blender-ru (91%)
Jan 3 2018, 8:21 PM

Jul 30 2017

Sv. Lockal (lockal) closed T52220: join request as Resolved.
Jul 30 2017, 11:42 PM · Translations
Sv. Lockal (lockal) added a comment to T52220: join request.

@Denis Bryukhanov (dennyishere) , check out weblate website https://translateblender.ru/ - you can register there and submit your own translations, which I'll upload to Blender SVN.

Jul 30 2017, 11:41 PM · Translations

Jun 26 2016

Sv. Lockal (lockal) committed rBTS4270: Update from https://translateblender.ru.
Update from https://translateblender.ru
Jun 26 2016, 3:37 PM

Nov 10 2015

Sv. Lockal (lockal) updated the title for P285 Select similar location from Select simolar location to Select similar location.
Nov 10 2015, 11:50 AM
Sv. Lockal (lockal) edited P285 Select similar location.
Nov 10 2015, 11:50 AM

Jun 8 2015

Sv. Lockal (lockal) committed rBTS3588: Update from https://translateblender.ru.
Update from https://translateblender.ru
Jun 8 2015, 11:20 AM

May 18 2015

Sv. Lockal (lockal) accepted rB4fc318811277: Cycles: Get rid of one more OpenGL matrix manipulation/push/pop..

Fixed in rB88acb3c5

May 18 2015, 7:58 AM

May 16 2015

Sv. Lockal (lockal) added a comment to T44727: Links for old releases, lead to current dl page.

Hi! I would like to download Blender 2.74(***), but it downloads 2.74, same for all other versions. Please check this problem. I downloaded here: http://www.blender.org/features/past-releases/

May 16 2015, 3:14 PM · Infrastructure: Websites

May 15 2015

Sv. Lockal (lockal) added a comment to T44707: viewport render in broken with border render activated.

Reproduced this problem with no ocio, added an inline comment with possible fix here: rB4fc318811

May 15 2015, 7:21 PM · Cycles, BF Blender
Sv. Lockal (lockal) raised a concern with rB4fc318811277: Cycles: Get rid of one more OpenGL matrix manipulation/push/pop..

For T44707

May 15 2015, 7:20 PM

May 13 2015

Sv. Lockal (lockal) abandoned D1294: Cycles: Simplify scale-move transform for raster to camera transformation.

Superseded by completely another more complex approach. Still experimenting, but new patch has nothing common with this patch, so closing this one.

May 13 2015, 7:00 PM

May 11 2015

Sv. Lockal (lockal) committed rBc7bccb30bf03: Cycles: check for F16C support with __cpuid, as we do for BMI and BMI2 (authored by Sv. Lockal (lockal)).
Cycles: check for F16C support with __cpuid, as we do for BMI and BMI2
May 11 2015, 5:53 PM

May 10 2015

Sv. Lockal (lockal) added a comment to D1294: Cycles: Simplify scale-move transform for raster to camera transformation.

Based in the fact, that no rotation/shear transformations are used with rastertocamera transform, and perspective coefficient is never zero (how could it be?).

May 10 2015, 11:19 PM
Sv. Lockal (lockal) retitled D1294: Cycles: Simplify scale-move transform for raster to camera transformation from to Cycles: Simplify scale-move transform for raster to camera transformation.
May 10 2015, 11:00 PM
Sv. Lockal (lockal) committed rBd55868c3b200: Cycles: And yet another compilation fix after half-float commit for clang. (authored by Sv. Lockal (lockal)).
Cycles: And yet another compilation fix after half-float commit for clang.
May 10 2015, 9:36 PM
Sv. Lockal (lockal) committed rB3ec168465d31: Cycles: fix compilation on 32-bit Windows for half-floats (authored by Sv. Lockal (lockal)).
Cycles: fix compilation on 32-bit Windows for half-floats
May 10 2015, 9:07 PM
Sv. Lockal (lockal) committed rB8db2a9a352c6: Cycles: Add -mf16c for previous commit for Scons (authored by Sv. Lockal (lockal)).
Cycles: Add -mf16c for previous commit for Scons
May 10 2015, 7:51 PM
Sv. Lockal (lockal) committed rB2ec221aa281d: Cycles: Use native float->half conversion instructions for Haswell CPUs. (authored by Sv. Lockal (lockal)).
Cycles: Use native float->half conversion instructions for Haswell CPUs.
May 10 2015, 6:45 PM

May 5 2015

Martijn Berger (juicyfruit) awarded rB7201f6d14c01: Cycles: Use curve approximation for blackbody instead of lookup table a Love token.
May 5 2015, 7:43 PM
Thomas Dinges (dingto) awarded rB7201f6d14c01: Cycles: Use curve approximation for blackbody instead of lookup table a Love token.
May 5 2015, 11:26 AM
Sv. Lockal (lockal) committed rB7201f6d14c01: Cycles: Use curve approximation for blackbody instead of lookup table (authored by Sv. Lockal (lockal)).
Cycles: Use curve approximation for blackbody instead of lookup table
May 5 2015, 8:12 AM
Sv. Lockal (lockal) closed D1280: Cycles: Use curve approximation for blackbody instead of lookup table by committing rB7201f6d14c01: Cycles: Use curve approximation for blackbody instead of lookup table.
May 5 2015, 8:12 AM
Sv. Lockal (lockal) updated D1280: Cycles: Use curve approximation for blackbody instead of lookup table.
May 5 2015, 8:11 AM
Sv. Lockal (lockal) updated the diff for D1280: Cycles: Use curve approximation for blackbody instead of lookup table.

Fix if() formatting and add pad3.

May 5 2015, 8:03 AM

May 4 2015

Sv. Lockal (lockal) added a comment to D1280: Cycles: Use curve approximation for blackbody instead of lookup table.

@Thomas Dinges (dingto), it seems to me that current implementation is affected by "off-by-one" mistake somewhere around here: https://developer.blender.org/diffusion/B/browse/master/intern/cycles/render/blackbody.cpp$96 . I've already noticed that our current implementation gives different colors from OSL and my approximation is based on unbiased function.

May 4 2015, 11:47 PM
Sv. Lockal (lockal) added a comment to D1280: Cycles: Use curve approximation for blackbody instead of lookup table.

For CUDA (7.0) the difference is smaller, 00:52.86 (lut) and 00:51:91 (approx).

May 4 2015, 8:19 PM
Sv. Lockal (lockal) added a comment to D1280: Cycles: Use curve approximation for blackbody instead of lookup table.

Quick test on CPU with plane gave 25.45s (lut) vs 20.26s (approx). Input socket is obviously connected.

May 4 2015, 5:50 PM
Sv. Lockal (lockal) added a comment to D1280: Cycles: Use curve approximation for blackbody instead of lookup table.

Difference between original (non-lut) function and approximation:

May 4 2015, 5:42 PM
Sv. Lockal (lockal) retitled D1280: Cycles: Use curve approximation for blackbody instead of lookup table from to Cycles: Use curve approximation for blackbody instead of lookup table.
May 4 2015, 5:33 PM
Sv. Lockal (lockal) edited P222 BlackbodyApproxTest.cpp.
May 4 2015, 12:32 PM · Cycles

May 1 2015

Sv. Lockal (lockal) added a comment to T44197: Cycles OpenCL kernel-splitting work.

I've checked rB5ad79b83afae again with NVidia and it doesn't compile on my computer anymore even with ADV shading disabled. The compile process looks similar to nvcc, clBuildProgram works for 1 minute or so and then eats up all available RAM and crashes. I guess this is related to @varunsundar08@gmail.com (varunsundar08) report.

May 1 2015, 9:35 PM · Cycles, BF Blender

Apr 8 2015

Sv. Lockal (lockal) added inline comments to rB808de65d9124: EdgeHash: ensure function, avoids multiple lookups.
Apr 8 2015, 8:03 PM

Apr 7 2015

Sv. Lockal (lockal) added a comment to D1224: Cycles: use native saturate function for CUDA.

Common diff in PTX (same for SASS), 92 replacements:

< 	mov.f32 	%f6781, 0f00000000;
< 	max.ftz.f32 	%f6782, %f1500, %f6781;
< 	mov.f32 	%f6783, 0f3F800000;
< 	min.ftz.f32 	%f10110, %f6782, %f6783;
---
> 	cvt.ftz.sat.f32.f32	%f10029, %f1500;
Apr 7 2015, 10:51 PM · Cycles
Sv. Lockal (lockal) retitled D1224: Cycles: use native saturate function for CUDA from to Cycles: use native saturate function for CUDA.
Apr 7 2015, 10:47 PM · Cycles

Apr 1 2015

Sv. Lockal (lockal) added a comment to T44197: Cycles OpenCL kernel-splitting work.

I've tested this patch (the original one from D1200) on Ubuntu 14.10 x86-64 and NVidia GTX 690 with MikePan BMW scene.

Apr 1 2015, 2:47 PM · Cycles, BF Blender

Mar 27 2015

Sv. Lockal (lockal) awarded D1200: Cycles OpenCL kernel-splitting work a The World Burns token.
Mar 27 2015, 3:03 PM · Cycles

Mar 26 2015

Sv. Lockal (lockal) added a comment to D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.


Procedural checker texture with Position node.

Mar 26 2015, 11:51 AM

Mar 25 2015

Sv. Lockal (lockal) added a comment to D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.

Sorry, I don't understand your proposal.

Mar 25 2015, 1:19 PM

Mar 23 2015

Sv. Lockal (lockal) updated D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.
Mar 23 2015, 1:22 PM
Thomas Dinges (dingto) awarded D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates a Like token.
Mar 23 2015, 5:48 AM

Mar 22 2015

Sv. Lockal (lockal) added a comment to D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.

Animation works too:

Mar 22 2015, 8:35 PM
Sv. Lockal (lockal) added a reviewer for D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates: Compositing.
Mar 22 2015, 6:33 PM
Sv. Lockal (lockal) added a comment to D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.

Added sergey as reviewer.

Mar 22 2015, 5:41 PM
Sv. Lockal (lockal) added a reviewer for D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates: Sergey Sharybin (sergey).
Mar 22 2015, 4:47 PM
Sv. Lockal (lockal) updated D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.
Mar 22 2015, 4:36 PM
Sv. Lockal (lockal) retitled D1196: Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates from to Compositor: new Position node, which returns numbers from 0 to 1 for each axe depending on coordinates.
Mar 22 2015, 4:33 PM

Mar 10 2015

Sv. Lockal (lockal) added a comment to T41066: Particle Carpet renders different in 2.71.

@Thomas Dinges (dingto) yes, the report is here: https://github.com/embree/embree/issues/30

Mar 10 2015, 7:06 AM · BF Blender: Unconfirmed

Mar 7 2015

Sv. Lockal (lockal) edited P209 avx_intersector.patch.
Mar 7 2015, 7:04 PM · Cycles
Sv. Lockal (lockal) committed rBc32ded36548b: Cycles: add better specializations for SSE shuffle function and few more… (authored by Sv. Lockal (lockal)).
Cycles: add better specializations for SSE shuffle function and few more…
Mar 7 2015, 6:30 PM
Sv. Lockal (lockal) added a comment to T41066: Particle Carpet renders different in 2.71.

Committed an actual fix. This problem comes from embree, both nmadd and nmsub functions in avxf and ssef are incorrect there.

Mar 7 2015, 5:51 PM · BF Blender: Unconfirmed
Sv. Lockal (lockal) committed rBc8fb488b087f: Fix T41066: An actual fix for curve intersection on FMA-enabled CPUs (authored by Sv. Lockal (lockal)).
Fix T41066: An actual fix for curve intersection on FMA-enabled CPUs
Mar 7 2015, 5:23 PM
Sv. Lockal (lockal) added a commit to T41066: Particle Carpet renders different in 2.71: rBc8fb488b087f: Fix T41066: An actual fix for curve intersection on FMA-enabled CPUs.
Mar 7 2015, 5:23 PM · BF Blender: Unconfirmed

Dec 24 2014

Sv. Lockal (lockal) committed rBTS3234: Update from https://translateblender.ru.
Update from https://translateblender.ru
Dec 24 2014, 11:56 AM

Dec 1 2014

Sv. Lockal (lockal) committed rBTS3196: Update from https://translateblender.ru.
Update from https://translateblender.ru
Dec 1 2014, 9:52 AM

Oct 13 2014

Sv. Lockal (lockal) updated subscribers of D812: Rename "Addons" to "Add-ons".

This patch affects BTS repo, adding @Bastien Montagne (mont29).

Oct 13 2014, 2:25 PM

Aug 5 2014

Sv. Lockal (lockal) added a comment to D710: WIP: Cycles: merge primitive information to improve memory access.

Tables for common benchmark files:

koro_final
patched2:32.70
vanilla2:43.04
Aug 5 2014, 8:27 PM
Sv. Lockal (lockal) added inline comments to D710: WIP: Cycles: merge primitive information to improve memory access.
Aug 5 2014, 8:17 PM
Sv. Lockal (lockal) retitled D710: WIP: Cycles: merge primitive information to improve memory access from Cycles: merge primitive information to improve memory access to WIP: Cycles: merge primitive information to improve memory access.
Aug 5 2014, 6:55 PM
Sv. Lockal (lockal) updated D710: WIP: Cycles: merge primitive information to improve memory access.
Aug 5 2014, 5:54 PM
Sv. Lockal (lockal) added inline comments to D710: WIP: Cycles: merge primitive information to improve memory access.
Aug 5 2014, 5:53 PM
Sv. Lockal (lockal) retitled D710: WIP: Cycles: merge primitive information to improve memory access from to Cycles: merge primitive information to improve memory access.
Aug 5 2014, 5:43 PM

May 15 2014

Sv. Lockal (lockal) added a comment to T40192: Stars in Blender 2.7.

For future references: this procedural texture could be used for recreating stars. Unfortunately, it is not so easy to use, but this approach is more flexible and can be applied in cycles too.

May 15 2014, 3:35 PM · BF Blender

May 12 2014

Sv. Lockal (lockal) edited this Maniphest Task.
May 12 2014, 4:39 PM · Cycles, BF Blender, Blender 2.70
Sv. Lockal (lockal) edited this Maniphest Task.
May 12 2014, 4:29 PM · Compositing, BF Blender

May 5 2014

Sv. Lockal (lockal) awarded T38401: Metropolis Sampling a Love token.
May 5 2014, 8:57 PM · BF Blender, Cycles

Apr 27 2014

Sv. Lockal (lockal) updated subscribers of T39692: Text fields in datablocks editor are broken.
Apr 27 2014, 8:30 PM · BF Blender

Apr 22 2014

Sv. Lockal (lockal) awarded rB9ab259f55b67: Cycles: shadow function optimization for transparent shadows (CPU only). a Love token.
Apr 22 2014, 6:54 AM

Apr 12 2014

Sv. Lockal (lockal) added a comment to T39692: Text fields in datablocks editor are broken.

I forgot to mention that this assert is visible only with debug builds. Asan crashlog:

Apr 12 2014, 10:06 PM · BF Blender
Sv. Lockal (lockal) updated the task description for T39692: Text fields in datablocks editor are broken.
Apr 12 2014, 9:07 PM · BF Blender
Sv. Lockal (lockal) raised the priority of T39692: Text fields in datablocks editor are broken from to Needs Triage by Developer.
Apr 12 2014, 9:06 PM · BF Blender

Mar 30 2014

Sv. Lockal (lockal) added a comment to T38034: Cycles optimization: move srgb/alpha conversion out of cycles kernel.

jrp, are you sure that 2 iterations of Halley's method is faster than 3 iterations of Newton-Raphson method? A single iteration of Halley's method has 6*, 4+ and 1/, while Newton-Raphson has only 4*, 1+ and 1/. Halley's method is less robust because it calculates approx^5, so the working domain will be smaller.

Mar 30 2014, 7:29 PM · BF Blender, Cycles

Mar 25 2014

Sv. Lockal (lockal) edited P26 exp2f.
Mar 25 2014, 11:57 PM

Mar 24 2014

Sv. Lockal (lockal) added a comment to T38551: Minor code cleanups for Visual Studio 2013, which now has a better standard C++ library.

Cycles provides multiple entry points (a.k.a. kernel functions in OpenCL or global functions in CUDA) for different CPUs and platforms. These entry points are located in files:

  1. kernel.cl (+ kernel_compat_opencl.h)
  2. kernel.cu (+ kernel_compat_cuda.h)
  3. kernel_avx.cpp (+ kernel_compat_cpu.h)
  4. kernel_sse2.cpp (+ kernel_compat_cpu.h)
  5. kernel_sse3.cpp (+ kernel_compat_cpu.h)
  6. kernel_sse41.cpp (+ kernel_compat_cpu.h)
Mar 24 2014, 3:09 AM · BF Blender, Platform: Windows

Mar 23 2014

Sv. Lockal (lockal) added a comment to T38551: Minor code cleanups for Visual Studio 2013, which now has a better standard C++ library.
  1. The slowdown from enabling KERNEL_SSE occurs mostly because compilers are unable to reassociate and simplify math expressions, written with intrinsics (except for icc). However it does not prevents us from adding sse_float4 or sse_float3 classes with overloaded operators. In this way we could control places, where compiler uses intrinsics, and where it tries to optimize automatically with autovectorization.
Mar 23 2014, 1:49 PM · BF Blender, Platform: Windows

Mar 22 2014

Martijn Berger (juicyfruit) awarded rBc45c472e1b37: Cycles: SSE optimization for line segments/ribbons hair a Like token.
Mar 22 2014, 11:00 PM
Pablo Vazquez (pablovazquez) awarded rBc45c472e1b37: Cycles: SSE optimization for line segments/ribbons hair a Love token.
Mar 22 2014, 10:25 PM
Sv. Lockal (lockal) added a comment to T39089: Double render times with Cycles 2.69 vs 2.70RC .

Confirming bisect by @Martijn Berger (juicyfruit).

Mar 22 2014, 12:40 AM · BF Blender, Cycles

Mar 6 2014

Sv. Lockal (lockal) awarded D17: Blender Internal, remove "Stars" feature a Pterodactyl token.
Mar 6 2014, 9:14 PM

Mar 5 2014

Sv. Lockal (lockal) added a comment to T38401: Metropolis Sampling.

CMake compiles after adding this extra patch:

Mar 5 2014, 7:43 PM · BF Blender, Cycles

Feb 12 2014

Sv. Lockal (lockal) added a comment to T38615: Blender Cycles Render results looking different at the BMW MikePan Benchmark File. .

Also, both renders have almost no noise, so I guess this modification has some kind of clamping.

Feb 12 2014, 9:51 PM · BF Blender, Cycles
Sv. Lockal (lockal) updated subscribers of T38615: Blender Cycles Render results looking different at the BMW MikePan Benchmark File. .

I also can't reproduce this problem. I don't see any connection with SSE optimizations. Front headlights are just simple mesh objects with no textures.

Feb 12 2014, 9:48 PM · BF Blender, Cycles

Feb 11 2014

Sv. Lockal (lockal) closed T38501: blender crashes right after adding image texture to material in cycles as Resolved.

Closed by commit rBbd44dcb63229.

Feb 11 2014, 2:57 PM · Blender 2.70, BF Blender, Cycles
Sv. Lockal (lockal) reopened T38501: blender crashes right after adding image texture to material in cycles as "Open".

MSVC 2008 builds still crash, investigating...

Feb 11 2014, 7:59 AM · Blender 2.70, BF Blender, Cycles

Feb 10 2014

Sv. Lockal (lockal) added a comment to T38501: blender crashes right after adding image texture to material in cycles.

@Thomas Dinges (dingto): SSE kernels are enabled on win32 (except for SSE41 and AVX).

Feb 10 2014, 2:34 PM · Blender 2.70, BF Blender, Cycles
Sv. Lockal (lockal) closed T38501: blender crashes right after adding image texture to material in cycles as Resolved.

Closed by commit rB709652970498.

Feb 10 2014, 2:25 PM · Blender 2.70, BF Blender, Cycles
Sv. Lockal (lockal) edited this Maniphest Task.
Feb 10 2014, 2:25 PM · Blender 2.70, BF Blender, Cycles
Sv. Lockal (lockal) claimed T38501: blender crashes right after adding image texture to material in cycles.

Crash occurs in SSE code (either _mm_load_ps or cast to (__m128&) for unaligned data -- remember, that float3 and float4 are not aligned in win32). Assigned to myself.

Feb 10 2014, 1:01 PM · Blender 2.70, BF Blender, Cycles

Feb 9 2014

Sv. Lockal (lockal) added a comment to T38401: Metropolis Sampling.

Oops, with F12 render everything is ok (except for tiles, of course). But "rendered" viewport mode has very obvious problems. I did a full recompilation with metropolis_7 patch (gcc 4.8, linux x86-64), but the problem is still there.

Feb 9 2014, 6:30 PM · BF Blender, Cycles
Sv. Lockal (lockal) added a comment to T38401: Metropolis Sampling.

Don't know what I did wrong, but with metropolis_6.diff and 3d8c106 all shadows and reflections are wrong for me. But previous versions of this patch gave plausible results (except for dark rectangles).


Feb 9 2014, 2:15 PM · BF Blender, Cycles

Feb 7 2014

Sv. Lockal (lockal) added a comment to T38304: cycles hair (curve segments) problem.

@Brecht Van Lommel (brecht), can you look at this? I'm not very confident in raytracing. One thing I can say is that nan's behavior depends on compiler. max(1.0, NaN) can return either NaN or 1.0, depending on how compiler allocates registers. GCC mostly eliminates nans in kernel_bvh_traversal.h (both in ray->P and ray->D), but MSVC exposes them. When ray has NaNs, msvc renders transparent pixels. That's how Koro's alpha map looks in GCC:


And that's how it looks with msvc 11:

Feb 7 2014, 6:40 PM · Cycles, BF Blender