Age | Commit message (Collapse) | Author |
|
Opted to keep includes if they are used indirectly (even if removing is possible).
|
|
|
|
|
|
|
|
|
|
Normals for each kdtree node were allocated but never used,
and search args only use in particles/boids code.
|
|
Issue of this bug is that most part of fftw is not thread safe, only compute-intensive fftw_execute & co are.
Since smoke was affected by this issue as well, a global fftw mutex was added to BLI_threads.
Audaspace also uses fftw in one of its readers (AUD_BandPassReader.cpp),
but this is not an issue currently since this code is disabled in CMake/scons files.
There was another threading issue with smoke, we need to copy dm used by emit_from_derivedmesh(),
as it is modified by this func.
Reviewers: sergey, brecht
Reviewed By: brecht
CC: brecht
Differential Revision: https://developer.blender.org/D374
|
|
updates.
This file crashes on loading with NULL pointer access to curve_cache:
{F77132}
The pdInitEffectors function was amalgamating the simple
collection of effector objects with an automatic precalculation for
curve guides and the like. This precalculation requires object data
that may not be available until the DAG has finished.
Since for DAG dependencies only the list of effectors is required,
added an argument to disable precalculation when collecting effectors.
|
|
|
|
|
|
Issue was caused by KD tree being allocated with the wrong size.
|
|
|
|
Summary:
Made objects update happening from multiple threads. It is a task-based
scheduling system which uses current dependency graph for spawning new
tasks. This means threading happens on object level, but the system is
flexible enough for higher granularity.
Technical details:
- Uses task scheduler which was recently committed to trunk
(that one which Brecht ported from Cycles).
- Added two utility functions to dependency graph:
* DAG_threaded_update_begin, which is called to initialize threaded
objects update. It will also schedule root DAG node to the queue,
hence starting evaluation process.
Initialization will calculate how much parents are to be evaluation
before current DAG node can be scheduled. This value is used by task
threads for faster detecting which nodes might be scheduled.
* DAG_threaded_update_handle_node_updated which is called from task
thread function when node was fully handled.
This function decreases num_pending_parents of node children and
schedules children with zero valency.
As it might have become clear, task thread receives DAG nodes and
decides which callback to call for it.
Currently only BKE_object_handle_update is called for object nodes.
In the future it'll call node->callback() from Ali's new DAG.
- This required adding some workarounds to the render pipeline.
Mainly to stop using get_object_dm() from modifiers' apply callback.
Such a call was only a workaround for dependency graph glitch when
rendering scene with, say, boolean modifiers before displaying
this scene.
Such change moves workaround from one place to another, so overall
hackentropy remains the same.
- Added paradigm of EvaluaitonContext. Currently it's more like just a
more reliable replacement for G.is_rendering which fails in some
circumstances.
Future idea of this context is to also store all the local data needed
for objects evaluation such as local time, Copy-on-Write data and so.
There're two types of EvaluationContext:
* Context used for viewport updated and owned by Main. In the future
this context might be easily moved to Window or Screen to allo
per-window/per-screen local time.
* Context used by render engines to evaluate objects for render purposes.
Render engine is an owner of this context.
This context is passed to all object update routines.
Reviewers: brecht, campbellbarton
Reviewed By: brecht
CC: lukastoenne
Differential Revision: https://developer.blender.org/D94
|
|
|
|
also fully remove freestyle raycasting_algorithm
|
|
|
|
|
|
Expect to be no functional changes :)
|
|
causes a crash
fuel array might be null!
|
|
|
|
|
|
|
|
|
|
character. also remove some dead code (return directly after return).
|
|
inconsistency between 2.67/2.68RC and previous versions
This was in fact really nasty bug, caused by multitex_nodes
function using global variable R (which is a copy of current
renderer). this variable is not initialized to anything
meaningful for until first rendering (preview or final)
happened.
Since multitex_nodes might be used outside of render pipeline,
made it so whether CM is on or off as an argument to functions
multitex_ext_safe and multitex_ext. Now multitex_nodes() is
only shall be used for stuff happening from render pipeline!
Also needed to make some changes to other places, so all the
usages of texture sampling knows for the fact whether CM is
on or off.
And one more change is related on behavior of dispalcement,
wave, warp, weightvg modifiers and smoke. They'll be always
using CM off since texture is used for influence, not for
color.
It's rather bigger patch, but it's mostly straightforward
changes, which we really need to be done.
Reviewed by Brecht, thanks!
|
|
many modifiers were calculating normals, when those normals were ignored by the next modifier.
now flag normals as dirty and recalculate for modifiers that set use `dependsOnNormals()` callback.
Quick test on mesh with 12 modifiers (mostly build type), calculated normals 6 times, now it only runs once - so this will give some speedup too.
|
|
|
|
velocities never apply additional margin ahead the fluid.
Also skip new flow velocity because thats not accurate enough to be of any use in there.
|
|
often write protected on modern systems.
This caused high resolution smoke to always regenerate new tile when domain was reinitialized, slowing down especially adaptive domain simulations. Now noise tile is saved in Blender temp directory instead.
|
|
|
|
Previously it was nearly impossible to have fast moving objects emitting smoke or they would just leave behind a row of smoke poofs instead of continious stream of smoke. Now it's possible to set number of subframes for each smoke flow.
Another new thing is ability to set size of smoke flow particles instead of using closest smoke cell. This also works with my earlier "full sample" commit, so no more blocky particles either. :)
For more info check my blog post: http://www.miikahweb.com/en/blog/2013/05/17/blender-smoke-subframes
This commit also includes couple of fixes I spotted while testing:
* Fix: dissolve was applied at different time for low res and high res simulations.
* Fix: full sample setting didn't get copied with domain.
|
|
- In some cases smoke flow could have 1 cell clipped off its sides.
- Forgot to re-enable openmp for emission code.
|
|
This is hopefully the ultimate solution against smoke blockiness near emitter.
Previously high resolution flow/emitter voxels were generated based on the low resolution ones. So if you had 32 resolution and 4 division high resolution, it still used smoke flow generated from those 32 resolution voxels. Now I introduced a new sampling method called "Full Sample" that generates full resolution flow for for high resolution domain as well.
Read more about it in my blog post: https://www.miikahweb.com/en/blog/2013/05/10/getting-rid-of-smoke-blockiness
Also changed "quick smoke" operator default voxel data interpolation mode to "Cubic B-Spline" to smoothen out it even more.
|
|
Issue was introduced by svn rev56273 and was caused by
both smokeModifier_do and smokeModifier_reset (which is
called from smokeModifier_do) do mutex lock/unlock.
This lead to quite undefined behavior caused by the same
thread released mutex twice.
Solved by not locking mutex from inside routines calling
from smokeModifier_process -- mutex is locked in parent
function anyway.
|
|
Added a mutex lock for smoke data access. The render was already working with a
copy of the volume data, so it's just a short lock to copy things and should not
block the UI much.
|
|
MATHUTILS_TOT_CB and reduce float->double conversions.
|
|
generator with a local one. It's not thread safe and will not give repeatable
results, so in most cases it should not be used.
Also fixes #34992 where the noise texture of a displacement modifier was not
properly random in opengl animation render, because the seed got reset to a
fixed value by an unrelated function while for final render it changed each
frame.
|
|
|
|
In some cases high resolution domain could have higher density than the low resolution counterpart, causing adaptive domain to clip off areas where density is still above domain threshold. Now also high resolution data is used to determine domain bounds.
|
|
adaptive domain "threshold".
|
|
|
|
produced way too high values. Now values are also clamped within correct range.
|
|
disappear from the simulation.
|
|
Also as minor change skip base_res update on domain step if adaptive domain isn't enabled.
|
|
It was caused by image threading safe commit and it was noticeable
only on really multi-core CPU (like dual-socket Xeon stations), was
not visible on core i7 machine.
The reason of slowdown was spinlock around image buffer referencing,
which lead to lots of cores waiting for single core and using image
buffer after it was referenced was not so much longer than doing
reference itself.
The most clear solution here seemed to be introducing Image Pool
which will contain list of loaded and referenced image buffers, so
all threads could skip lock if the pool is used for reading only.
Lock only needed in cases when buffer for requested image user is
missing in the pool. This lock will happen only once per image so
overall amount of locks is much less that it was before.
To operate with pool:
- BKE_image_pool_new() creates new pool
- BKE_image_pool_free() destroys pool and dereferences all image
buffers which were loaded to it
- BKE_image_pool_acquire_ibuf() returns image buffer for given
image and user. Pool could be NULL and in this case fallback to
BKE_image_acquire_ibuf will happen.
This helps to avoid lots to if(poll) checks in image sampling
code.
- BKE_image_pool_release_ibuf releases image buffer. In fact, it
will only do something if pool is NULL, in all other case it'll
equal to DoNothing operation.
|
|
|
|
|
|
|
|
High resolution cache didn't always read all written data fields which caused colors to get messed up.
|
|
without audaspace.
|