Most of them are not currently used but are essential for the further work.
- CPU kernels with SSE2 support will now have sse3b, sse3f and sse3i
- Added templatedversions of min4, max4 which are handy to use with register
variables.
- Added util_swap function which gets arguments by pointers.
So hopefully it'll be a portable version of std::swap.
Using this paper: Sven Woop, Watertight Ray/Triangle Intersection
http://jcgt.org/published/0002/01/05/paper.pdf
This change is expected to address quite reasonable amount of reports from the
bug tracker, plus it might help reducing the noise in some scenes.
Unfortunately, it's currently about 7% slower than the previous solution with
pre-computed triangle plane equations, but maybe with some smart tweaks to the
code (tests reshuffle, using SIMD in a nice way or so) we can avoid the speed
regression.
But perhaps smartest thing to do here would be to change single triangle / ray
intersection with multiple triangles / ray intersections. That's how Embree does
this and it's watertight single ray intersection is not any faster that this.
Currently only triangle intersection is modified accordingly to the paper, in
the future we would also want to modify the node / ray intersection.
Reviewers: brecht, juicyfruit
Subscribers: dingto, ton
Differential Revision: https://developer.blender.org/D819
The idea is to store visibility flags for leaf nodes only since visibility check
for inner nodes costs too much for QBVH hence it is not optimal to perform.
Leaf QBVH nodes have plenty of space to store all sort of flags, so we can make
nodes one element smaller, saving noticeable amount of memory.
Previously offsets were calculated based on the BVH node size,
which is wrong and real PITA in cases when some extra data is
to be added into (or removed from) the node.
Now use offsets which are not calculated form the node size.
This solves quite an over-allocation in BVH instances packing code,
unfortunately, it's not a magic bullet to solve memory bump caused
by the recent QBVH changes.
For that we'll likely need to decouple storage for leaf and inner
nodes. However, it's not really clear for now if it's something
important since that'd still be just a fraction of memory comparing
to all the hi-res textures.
Title says it all, quite straightforward implementation.
Would only mention that there's a bit of code duplication around packing node
into pack.nodes. Trying to de-duplicate it ends up in quite hairy code (like
functions with loads of arguments some of which could be NULL in certain
circumstances etc..). Leaving solving this duplication for later.
Before all the nodes were counted and allocated, leading to situations when
bunch of allocated memory is not used because reasonable amount of nodes are
simply ignored.
Visibility flags are set to all visibility anyway, So there was no reason
to perform that test.
TODO: We need to investigate if having primitive intersection functions
which doesn't do visibility check gives any speedup here as well.
This way extending intersection routines with some pre-calculation step wouldn't
explode the single file size, hopefully keeping them all in a nice maintainable
state.
These nodes were assuming sRGB input/output which is for sure wrong for the
shader pipeline which works in the linear space.
So now conversion to/from linear space happens in these nodes which makes them
making sence in the shader context but which might change look and feel of
existing scenes.
The issue was caused by the way how RNA pointer was created for the bMain:
namely Cycles was using RNA_id_pointer_create to create the pointer, which
would then try to refine the poniter based on the ID type.
This is just wrong and worked so far just because of co-incident, with the
file path from the bug report first letters in the ID name happened to be
NT which corresponds to NodeTree, and for sure refining such pointer will
fail.
Simple solution -- use proper way to create RNA pointer for non-ID block.
As discussed in rB983c71931b1886d4, we should print a warning in case of building on non-Windows and WITH_BF_IME enabled. We also terminate build in this case, so the warning isn't scrolled away. Was worked out together with @sergey.
This was caused by some internal optimization which evaluated SSS with
size of zero as BSDF but used different ID so the evaluation result
didn't appear in regular diffuse pass.
This lead to situation when SSS data was nowhere stored if the
size was zero.
Now SSS with zero size and close-to-zero sizes will be handled in the
same way from the passes point of view.
Original patch by @random (D765) with some minor work done by @campbell
and me.
At this place, I'd like call out a number of people who were involved and
deserve a big "Thank you!":
* At the first place @randon who developed and submitted the patch
* The Blendercn community which helped a lot with testing - espacially
* @yuzukyo, @leon_cheung and @kjym3
* @campbellbarton, @mont29 and @sergey for their help and advises during
* review
* @ton who realized the importance of this early on and asked me for
* reviewing
We are still not finished, as this is only the first part of the
implementaion, but there's more to come!
SpaceMouse Wireless
SpaceMouse Pro Wireless
Device info is from user reports. I don’t yet have the new devices, so
these are untested but likely to work :D
This way CUDA errors are visible in the image info line,
which makes things to behave the same across viewport and
final rendering.
That's right, we've got error reported via reports and info
line now. This is based on the feedback from our gooseberry
team.
This way when something goes wrong in Cycles (for example out of VRAM, timelimit
launching the kernel etc) we'll have a nice report in the Info space header.
Sure it'll be nice to have mention of error in the image editor's information
line, but that's for the future.
This fixes T42747: "CUDA error" appears only momentarily, then disappears
Currently it acts the same as set_cancel(), but this way we're able to
distinguish situations when rendering was aborted by user demand (for
example pressing Esc in standalone renderer) or if something went horribly
wrong (for example out of VRAM error).
This way for example we wouldn't wait a fortune while BVH is building after
GPU run out of memory when loading images just to see the render failure
message.
Forbid OSL from polluting current conext with obscure stuff from
windows.h, it's not useful and unhealthy anyway.
Maybe we sohuld also forbid using abbreviated Glog constants as
well tho.
Since the aligned allocation of shader closures in OSL memory pool
this workaround is no longer needed.
Also put a comment which describes the desired layout of the structure
so array of shader closures is all nicely aligned.
This solves bugs like T42210 which are caused by compiler being
smart and using some SSE instructions to operate with closure
classes, which was failing because those classes are not allocated
by the regular allocator but allocated in memory pool in OSL.
With newer versions of OSL it is now possible to force closure
classes being aligned to a given boundary and this commit uses
this new functionality.
Unfortunately, it means we're no longer compatible with older
versions of OSL, only latest git version from upstream and our
branch at github are supported:
https://github.com/Nazg-Gul/OpenShadingLanguage/tree/blender-fixes
For OSX and Windows it's not an issue because libraries are
already updated there, Linux users would need to run install_deps
script.
