With a Titan Xp, reduces path trace local memory from 1092MB to 840MB.
Benchmark performance was within 1% with both RX 480 and Titan Xp.
Original patch was implemented by Sergey.
Differential Revision: https://developer.blender.org/D2249
Similar to what we did for area lights previously, this should help
preserve stratification when using multiple BSDFs in theory. Improvements
are not easily noticeable in practice though, because the number of BSDFs
is usually low. Still nice to eliminate one sampling dimension.
This was needed when we accessed OSL closure memory after shader evaluation,
which could get overwritten by another shader evaluation. But all closures
are immediatley converted to ShaderClosure now, so no longer needed.
Also pass by value and don't write back now that it is just a hash for seeding
and no longer an LCG state. Together this makes CUDA a tiny bit faster in my
tests, but mainly simplifies code.
Since all the shadow catchers are already assumed to be in the footage,
the shadows they cast on each other are already in the footage too. So
don't just let shadow catchers skip self, but all shadow catchers.
Another justification is that it should not matter if the shadow catcher
is modeled as one object or multiple separate objects, the resulting
render should be the same.
Differential Revision: https://developer.blender.org/D2763
It uses an idea of accumulating all possible light reachable across the
light path (without taking shadow blocked into account) and accumulating
total shaded light across the path. Dividing second figure by first one
seems to be giving good estimate of the shadow.
In fact, to my knowledge, it's something really similar to what is
happening in the denoising branch, so we are aligned here which is good.
The workflow is following:
- Create an object which matches real-life object on which shadow is
to be catched.
- Create approximate similar material on that object.
This is needed to make indirect light properly affecting CG objects
in the scene.
- Mark object as Shadow Catcher in the Object properties.
Ideally, after doing that it will be possible to render the image and
simply alpha-over it on top of real footage.
Decoupled ray marching is not supported yet.
Transparent shadows are always enabled for volume rendering.
Changes in kernel/bvh and kernel/geom are from Sergey.
This simiplifies code significantly, and prepares it for
record-all transparent shadow function in split kernel.
This does a few things at once:
- Refactors host side split kernel logic into a new device
agnostic class `DeviceSplitKernel`.
- Removes tile splitting, a new work pool implementation takes its place and
allows as many threads as will fit in memory regardless of tile size, which
can give performance gains.
- Refactors split state buffers into one buffer, as well as reduces the
number of arguments passed to kernels. Means there's less code to deal
with overall.
- Moves kernel logic out of OpenCL kernel files so they can later be used by
other device types.
- Replaced OpenCL specific APIs with new generic versions
- Tiles can now be seen updating during rendering
Seems CUDA failed to de-duplicate the array across multiple inlined
versions of the shadow_blocked(). Helped it a bit with that now.
Gives about 100MB memory improvement on a scenes after previous
commit and brings up memory "regression" to only 100MB comparing to
the master branch now.
The idea is to record all possible transparent intersections when
shooting transparent ray on GPU (similar to what we were doing on
CPU already).
This avoids need of doing whole ray-to-scene intersections queries
for each intersection and speeds up a lot cases like transparent
hair in the cost of extra memory.
This commit is a base ground for now and this feature is kept
disabled for until some further tweaks.
Now we break the traversal cycle and then perform volume attenuation
and check with zero throughput. Not sure it makes any measurable sense
at this moment, but in the future it might help de-duplicating some
extra logic here.
Mostly this is making inlining match CUDA 7.5 in a few performance critical
places. The end result is that performance is now better than before, possibly
due to less register spilling or other CUDA 8.0 compiler improvements.
On benchmarks scenes, there are 3% to 35% render time reductions. Stack memory
usage is reduced a little too.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D2269
While they prevent legit write past the array boundary error
those fixes introduced regression in behavior when having exact
max_hits transparent intersections and nothing else.
Previous code would have considered such case a totally opaque,
but it's not correct.
Fixes T48941: Some materials don't get transparent shadows anymore
It was possible to miss bounces termination criteria in this functions,
mainly when max_hits was set to 0.
Made the check more robust in traversal functions (which should not
affect performance, it's an operation of same complexity AFAIK).
Also avoid doing ray-scene intersection from shadow_blocked when
limit of transparent bounces was already reached.
This commit adds a new distribution to the Glossy, Anisotropic and Glass BSDFs that implements the
multiple-scattering microfacet model described in the paper "Multiple-Scattering Microfacet BSDFs with the Smith Model".
Essentially, the improvement is that unlike classical GGX, which only models single scattering and assumes
the contribution of multiple bounces to be zero, this new model performs a random walk on the microsurface until
the ray leaves it again, which ensures perfect energy conservation.
In practise, this means that the "darkening problem" - GGX materials becoming darker with increasing
roughness - is solved in a physically correct and efficient way.
The downside of this model is that it has no (known) analytic expression for evalation. However, it can be
evaluated stochastically, and although the correct PDF isn't known either, the properties of MIS and the
balance heuristic guarantee an unbiased result at the cost of slightly higher noise.
Reviewers: dingto, #cycles, brecht
Reviewed By: dingto, #cycles, brecht
Subscribers: bliblubli, ace_dragon, gregzaal, brecht, harvester, dingto, marcog, swerner, jtheninja, Blendify, nutel
Differential Revision: https://developer.blender.org/D2002
This commit makes it so malloc() is only happening once per volume and
once per transparent shadow query (per thread), improving scalability of
the code to multiple CPU cores.
Hard to measure this with a low-bottom i7 here currently, but from quick
tests seems volume sampling gave about 3-5% speedup.
The idea is to store allocated memory in kernel globals, which are per
thread on CPU already.
Reviewers: dingto, juicyfruit, lukasstockner97, maiself, brecht
Reviewed By: brecht
Subscribers: Blendify, nutel
Differential Revision: https://developer.blender.org/D1996
The idea is to switch from allocating separate buffers for shader data's
structure of arrays to allocating one huge memory block and do some index
trickery to make it accessed as SOA.
This saves quite reasonable amount of lines of code in device_opencl and
also makes it possible to get rid of special declaration of ShaderData
structure.
As a side effect it also makes it easier to experiment with SOA vs. AOS
for split kernel.
Works fine here on NVidia GTX580, Intel CPU amd AMD Fiji cards.
Reviewers: #cycles, brecht, juicyfruit, dingto
Differential Revision: https://developer.blender.org/D1593
Use KernelGlobals to access all the global arrays for the intermediate
storage instead of passing all this storage things explicitly.
Tested here with Intel OpenCL, NVIDIA GTX580 and AMD Fiji, didn't see
any artifacts, so guess it's all good.
Reviewers: juicyfruit, dingto, lukasstockner97
Differential Revision: https://developer.blender.org/D1736
The goal is to make Experimental kernel closer in performance to the
official kernel, avoiding spills and such.
There should not be big impact on official kernel, own tests showed
few percent performance drop on laptop's GPU. CPU was always the
same speed on AVX, AVX2 and SSE4.1 CPUs i've been testing here.
This seems to be the last essential step before we can get rid of
Experimental kernel and enable SSS officially on GPU without causing
some major performance issues.
Surely some more tweaks are possibly required, but that we can do
for until cows go home anyway.
This commit changes the way how we pass bounce information to the Light
Path node. Instead of manualy copying the bounces into ShaderData, we now
directly pass PathState. This reduces the arguments that we need to pass
around and also makes it easier to extend the feature.
This commit also exposes the Transmission Bounce Depth to the Light Path
node. It works similar to the Transparent Depth Output: Replace a
Transmission lightpath after X bounces with another shader, e.g a Diffuse
one. This can be used to avoid black surfaces, due to low amount of max
bounces.
Reviewed by Sergey and Brecht, thanks for some hlp with this.
I tested compilation and usage on CPU (SVM and OSL), CUDA, OpenCL Split
and Mega kernel. Hopefully this covers all devices. :)
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
Volume scatter might happen before path termination, so
need to check transparent bounces and consider shadow an
opaque when max transparent bounces are reached.
TODO: CPU code seems to have different branching in conditions
which made me thinking it does different things with volume
attenuation, but from the render results it seems the same
exact things are happening there. Worth looking into making
simplifying code a bit here to improve readability.
