| Age | Commit message (Collapse) | Author |
|---|
|
`targetIsDirectory` slipped through the code review of
{D6679}/{rBafb1a64ccb81}.
`BLI_is_dir` exists to check for directory status of a file.
Remove some `else-after-return`s.
Use `r_` prefix for return value arguments, and move it to the end
in the list of arguments.
|
|
The routine to find dissolvable vertices had a check to ensure
that the vertex was exactly in line with the two neighbors.
I have convinced myself that this check is unneccesary (it was
failing with only a 1e-9 difference from 0 on a cross check),
so have removed it.
|
|
`BKE_mesh_calc_edges` was the main performance bottleneck in D9141.
While openvdb only needed ~115ms, calculating the edges afterwards
took ~960ms. Now with some parallelization this is reduced to ~210ms.
Parallelizing `BKE_mesh_calc_edges` is not entirely trivial, because it
has to perform deduplication and some other things that have to happen
in a certain order. Even though the multithreading improves performance
with more threads, there are diminishing returns when too many threads
are used in this function.
The speedup is mainly achieved by having multiple hash tables that are
filled in parallel. The distribution of the edges to hash tables is based on
a hash (that is different from the hash used in the actual hash tables).
I moved the function to C++, because that made it easier for me to
optimize it. Furthermore, I added `BLI_task.hh` which contains some
light tbb wrappers for parallelization.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D9151
|
|
The underlying type of the enum cannot be fixed here due to its usage
in C code.
All the values possible in the width of the underlying type are not
valid for an enum.
Only 0 to (2*max - 1) if all enumerators are unsigned.
So the macro asks for the biggest value among the //listed// ones.
If any enumerator C is set to say `A|B`, then C would be the maximum.
(2*max-1) is used as the mask.
The warnings (for each enum modified in this commit):
GPU_vertex_buffer.h:43:1: runtime error: load of value 4294967291
which is not a valid value for type 'GPUVertBufStatus'
https://github.com/llvm/llvm-project/commit/1c2c9867
Ref T81340
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D9067
|
|
|
|
We have our own assert implementation, `BLI_assert()` that is prefered over the
C standard library one. Its output is more consistent across compilers and
makes termination on assert failure optional (through `WITH_ASSERT_ABORT`).
In many places we'd include the C library header without ever accessing it.
|
|
|
|
Cleanup old tracker task format to the new. e.g: [#34039] to T34039
Ref D8718
|
|
This is already the most widely used convention.
Use this so `make check_spelling_c` will ignore all email addresses.
|
|
|
|
The problem is due to lack of precision with small and coplanar triangles.
Also, triangles that have vertices with the same coordinate are at the
threshold of the intersection.
We could add an epsilon to consider the minimum distance for intersection.
But that would add a lot of overhead to the code.
The solution used is to increase precision using doubles.
|
|
the same time
The issue was two fold.
First something sets the loop element tag and doesn't clear it before
the UV code in question tries to use the tags. Added a sanity clear to
make sure that it operates on a clean tag state.
The next one was that the UV maps in question had quite a few points
that had zero length UV loop edges. This would lead to division by
zero.
Reviewed By: Jeroen Bakker, Brecht
Differential Revision: http://developer.blender.org/D8967
|
|
Since at least MSVC2010 there is a `SSIZE_T` available for windows, use
it to typedef `ssize_t` on this platform.
|
|
|
|
|
|
|
|
The ffmpeg, guardedalloc and blenlib are quite isolated and putting them in
their own executable separate from blender_test is faster for development than
linking the entire blender_tests executable.
For Cycles, this also bundles all the unit tests into one executable.
Ref T79958
Differential Revision: https://developer.blender.org/D8714
|
|
Set flags directly on the target, and use common function for all cases.
This refactoring helps with the next commit for test executables.
Ref D8714
|
|
Since this buffer is used as an array of 12 32bits integers, and C++
string expect a NULL-terminated C-string, we need an extra char to
ensure last one is always NULL.
See D8906. Thanks to @brecht for noting this one too.
|
|
Also added code so that exact solver does the whole collection at once.
This patch allows users to use a collection (as an alternative to Object)
for the boolean modifier operand, and therefore get rid of a long modifier stack.
|
|
|
|
|
|
|
|
|
|
Thanks for @deadpin for noticing!
|
|
The reason for this functions existence was poorly documented
|
|
|
|
This adds a generic string search library in `BLI_string_search.h`.
The library has a simple to use C api that allows it's users to
filter and sort a set of possible search results based on some
query string.
Reviewers: Severin
Differential Revision: https://developer.blender.org/D8825
|
|
On systems with 512kb stack this happened at around 13k points.
This happened at times with grease-pencil, although callers that
frequently use complex polygons should be using BLI_polyfill_calc_arena.
|
|
For more information see rB2aff45146f1464ba8899368ad004522cb6a1a98c.
|
|
The semantic of those is the same as for std::string_view.
|
|
|
|
No functional changes.
|
|
Also correct wrapped lines of example code in threads.cc.
|
|
Well, this only fixes the example if one uses the new Exact mode,
but since that is available, seems fair to call this fixed.
Since these were not closed-volume operands, the Exact mode needed
some adjustment to the threshold used for "inside-outside" tests
for the case of deciding if the cutter is inside the other shape
for a Difference.
|
|
|
|
|
|
|
|
The code to detect non-trivial coplanar intersection sometimes
falsely said there wasn't one. This caused some coplanar intersections
to be missed. Also added a test for this case.
|
|
No functional changes
|
|
|
|
Building with `WITH_GMP=ON` revealed more const-return-type issues.
No functional changes.
|
|
This was reported for the "Add Node" search functionality, but is
relevant in other searches as well.
So e.g. when searching for "Separate XYZ", typing "sep", then " " (with
the intention to type "X" next) would clear the search field. Now use
the same method (matching against all search words) as in F3 searching
('menu_search_update_fn') in other searches as well [searching IDs,
property objects, finding nodes,...]
This should give a much nicer search experience in general.
Note: this does not touch other searches in the Dopesheet, Outliner,
Filebrowser or User Preferences that have other search implementations.
Maniphest Tasks: T78084
Differential Revision: https://developer.blender.org/D8232
|
|
|
|
|
|
I somehow undid these changes again before committing, sorry for the
noise...
|
|
Names were changed in 66b12ef4ab94, but the comment wasn't updated.
|
|
The animated objects was not updated for each internal substep for the rigidbody sim.
This would lead to unstable simulations or very annoying clipping artifacts.
Updated the code to use explicit substeps and tie it to the scene frame rate.
Fix T47402: Properly updating the animated objects fixes the reported issue.
Reviewed By: Brecht, Jacques
Differential Revision: http://developer.blender.org/D8762
|
|
|
|
This error happened only with O2 or O3 in my tests.
Casting to uintptr_t and back seems to quiet the compiler.
|
