This gives quite the same problems as experimental CUDA kernels
and for until it's found a root cause of the problem we'd just
explicitly uninline the function.
TODO: We might want to refactor debug passes into PASS_DEBUG and some
debug_type (similar to Blender's side passes) to avoid issue of running
out of bits.
This commit contains all the work related on the AMD megakernel split work
which was mainly done by Varun Sundar, George Kyriazis and Lenny Wang, plus
some help from Sergey Sharybin, Martijn Berger, Thomas Dinges and likely
someone else which we're forgetting to mention.
Currently only AMD cards are enabled for the new split kernel, but it is
possible to force split opencl kernel to be used by setting the following
environment variable: CYCLES_OPENCL_SPLIT_KERNEL_TEST=1.
Not all the features are supported yet, and that being said no motion blur,
camera blur, SSS and volumetrics for now. Also transparent shadows are
disabled on AMD device because of some compiler bug.
This kernel is also only implements regular path tracing and supporting
branched one will take a bit. Branched path tracing is exposed to the
interface still, which is a bit misleading and will be hidden there soon.
More feature will be enabled once they're ported to the split kernel and
tested.
Neither regular CPU nor CUDA has any difference, they're generating the
same exact code, which means no regressions/improvements there.
Based on the research paper:
https://research.nvidia.com/sites/default/files/publications/laine2013hpg_paper.pdf
Here's the documentation:
https://docs.google.com/document/d/1LuXW-CV-sVJkQaEGZlMJ86jZ8FmoPfecaMdR-oiWbUY/edit
Design discussion of the patch:
https://developer.blender.org/T44197
Differential Revision: https://developer.blender.org/D1200
This replaces sequential ray moving followed with scene intersection with
single BVH traversal, which gives us all possible intersections.
Only implemented for CPU, due to qsort and a bigger memory usage on GPU
which we rather avoid. GPU still uses the regular bvh volume intersection code, while CPU now uses the new code.
This improves render performance for scenes with:
a) Camera inside volume mesh
b) SSS mesh intersecting a volume mesh/domain
In simple volume files (not much geometry) performance is roughly the same
(slightly faster). In files with a lot of geometry, the performance
increase is larger. bmps.blend with a volume shader and camera inside the
mesh, it renders ~10% faster here.
Patch by Sergey and myself.
Differential Revision: https://developer.blender.org/D1264
This more a workaround for CUDA optimizer which can't optimize clamp(x, 0, 1)
into a single instruction and uses 4 instructions instead.
Original patch by @lockal with own modification:
Don't make changes outside of the kernel. They don't make any difference
anyway and term saturate() has a bit different meaning outside of kernel.
This gives around 2% of speedup in Barcelona file, but in more complex shader
setups with lots of math nodes with clamping speedup could be much nicer.
Subscribers: dingto
Projects: #cycles
Differential Revision: https://developer.blender.org/D1224
This way we can get rid of inefficient memory usage caused by BVH boundbox
part being unused by leaf nodes but still being allocated for them. Doing
such split allows to save 6 of float4 values for QBVH per leaf node and 3
of float4 values for regular BVH per leaf node.
This translates into following memory save using 01.01.01.G rendered
without hair:
Device memory size Device memory peak Global memory peak
Before the patch: 4957 5051 7668
With the patch: 4467 4562 7332
The measurements are done against current master. Still need to run speed tests
and it's hard to predict if it's faster or not: on the one hand leaf nodes are
now much more coherent in cache, on the other hand they're not so much coherent
with regular nodes anymore.
Reviewers: brecht, juicyfruit
Subscribers: venomgfx, eyecandy
Differential Revision: https://developer.blender.org/D1236
Issue was caused by MSVC not being able to optimize some code out in the same
way as GCC/Clang does, so now that parts of code are explicitly unfolded in
order to help compilers out.
This makes speed loss much less drastic on my laptop. That's probably as good
as we can do with MSVC without investing infinite amount of time looking trying
to workaround the optimizer.
This argument was unused and got nicely optimized out. But once it
starts to be using registers are getting stressed really crazy,
causing slow down of render.
It's not that bad because this typo could only caused not really
efficient BVH traversal, causing higher render times. Not as if
it was causing render artifacts.
This inconsistency drove me totally crazy, it's really confusing
when it's inconsistent especially when you work on both Cycles and
Blender sides.
Shouldn;t cause merge PITA, it's whitespace changes only, Git should
be able to merge it nicely.
The fix was really flacky, in terms during speed benchmarks i had
abort() in the fallback block to be sure it never runs in production
scenes, but that affected on the optimization as well. Without this
abort there's quite bad slowdown of 5-7% on the renders even tho
the Pleucker fallback was never run.
This is all weird and for now reverting the change which affects on
all the production scenes and will look into alternative fixes for
the original issue with precision loss on huge planes.
This reverts commit 9489205c5c.
The issue was caused by numerical instability whrn having ray origin close to a huge
triangle, which could have aused bad ray distance check.
Watertight Woop intersection isn't really addressing such cases, it's dealing with
small triangles far away from the ray origin instead, so it's a bit tricky yo make
it working reliably.
While we're quite close to the release it's safer to do check in Pleaucker coordinates
if ray close to a huge triangle. Likely this additional check combined with some other
tweaks to the code doesn't cause measurable slowdown in the scenes tested here.
After the release we can play a bit more with this code in order to make it more
stable without Pleucker fallback.
From more investigation of the numeric failures in the kernel it appears
the check was rather correct. But in theory it;s also needed for the motion
triangles.
This is the same issue T43475: SSE4 code is more robust to non-finite values
in the ray origin/direction. So for now added a check before doing BVH traversal
for pre-SSE4 CPUs.
For sure actual root of the issue is a bit different and much more tricky to
solve, especially without disturbing render results too much. Still looking
into this.
In any case, it's kinda fine to have such a check, we might later make it to be
a kernel_assert() instead of just a return.
Previous fix didn't quite work well. For some reason everything worked fine when
using native nvcc in 32bit environment, but cross-compiling from 64bit platform
it was still running out of memory.
For now just made it so all the kernels are slower on 32bit CUDA as a temporary
solution. Either it'll be solved in next CUDA releases (by dropped 32bit? =\) or
we'll find better workaround.
Slowdown was caused by watertight intersection commit and follow-up workaorund
for compiler crash which uninlined utility function which rotates the ray.
Now it's only uninlined for sm_50 and sm_52 experimental kernels which are the
only ones which failed to compile.
Rendering still might be a bit slower but at least shouldn't be that dramatic.
The issue was caused bu the optimization in surface attributes for cases when
there's only a volume shader used. Some attributes doesn't make sense in that
case and were skipped from calculation.
However, it is possible that kernel would still try to access them (because of
the shader setup etc). Prevented an infinite loop in the kernel now, which
should not have much affect on regular renders.
It is possible that ray distance will be zero which would make intersection
refinement return NaN as the refined position which would later lead to all
sort of mathematical issues.
Don't think there are ways to improve intersection accuracy for such rays
so just return original intersection coordinate.
This should fix T43475.
TODO: Need to look into possible issues in Ashikhmin BSDF which might return
zero-length reflected/transmitted ray?
This commit enables BVH leaf nodes split by the primitive type and makes it
so BVH traversal code is now aware and benefits from this.
As was mentioned in original commit, this change is crucial to be able to do
single ray to multiple triangle intersection. But it also appears to give
barely visible speedup in some scene.
In any case there should be no noticeable slowdown, and this change is what
we need to have anyway.
OpenCL apparently does not support templates, so the idea of generic
function for swapping is a bit of a failure. Now it is either inlined
into the code (in triangle intersection) or has specific implementation
for QBVH.
This is probably even better, because we can't create QBVH-specific
function in util_math anyway.
This way we'll be sure (in debug builds) that regular BVH traversal is not used
for QBVH tree (could happen because of mismatch of logic in kernel and render).
