| Age | Commit message (Collapse) | Author |
|---|
|
This patch implements the vector types (i.e:`float2`) by making heavy
usage of templating. All vector functions are now outside of the vector
classes (inside the `blender::math` namespace) and are not vector size
dependent for the most part.
In the ongoing effort to make shaders less GL centric, we are aiming
to share more code between GLSL and C++ to avoid code duplication.
####Motivations:
- We are aiming to share UBO and SSBO structures between GLSL and C++.
This means we will use many of the existing vector types and others
we currently don't have (uintX, intX). All these variations were
asking for many more code duplication.
- Deduplicate existing code which is duplicated for each vector size.
- We also want to share small functions. Which means that vector
functions should be static and not in the class namespace.
- Reduce friction to use these types in new projects due to their
incompleteness.
- The current state of the `BLI_(float|double|mpq)(2|3|4).hh` is a
bit of a let down. Most clases are incomplete, out of sync with each
others with different codestyles, and some functions that should be
static are not (i.e: `float3::reflect()`).
####Upsides:
- Still support `.x, .y, .z, .w` for readability.
- Compact, readable and easilly extendable.
- All of the vector functions are available for all the vectors types
and can be restricted to certain types. Also template specialization
let us define exception for special class (like mpq).
- With optimization ON, the compiler unroll the loops and performance
is the same.
####Downsides:
- Might impact debugability. Though I would arge that the bugs are
rarelly caused by the vector class itself (since the operations are
quite trivial) but by the type conversions.
- Might impact compile time. I did not saw a significant impact since
the usage is not really widespread.
- Functions needs to be rewritten to support arbitrary vector length.
For instance, one can't call `len_squared_v3v3` in
`math::length_squared()` and call it a day.
- Type cast does not work with the template version of the `math::`
vector functions. Meaning you need to manually cast `float *` and
`(float *)[3]` to `float3` for the function calls.
i.e: `math::distance_squared(float3(nearest.co), positions[i]);`
- Some parts might loose in readability:
`float3::dot(v1.normalized(), v2.normalized())`
becoming
`math::dot(math::normalize(v1), math::normalize(v2))`
But I propose, when appropriate, to use
`using namespace blender::math;` on function local or file scope to
increase readability.
`dot(normalize(v1), normalize(v2))`
####Consideration:
- Include back `.length()` method. It is quite handy and is more C++
oriented.
- I considered the GLM library as a candidate for replacement. It felt
like too much for what we need and would be difficult to extend / modify
to our needs.
- I used Macros to reduce code in operators declaration and potential
copy paste bugs. This could reduce debugability and could be reverted.
- This touches `delaunay_2d.cc` and the intersection code. I would like
to know @howardt opinion on the matter.
- The `noexcept` on the copy constructor of `mpq(2|3)` is being removed.
But according to @JacquesLucke it is not a real problem for now.
I would like to give a huge thanks to @JacquesLucke who helped during this
and pushed me to reduce the duplication further.
Reviewed By: brecht, sergey, JacquesLucke
Differential Revision: https://developer.blender.org/D13791
|
|
Reverted because the commit removes a lot of commits.
This reverts commit a2c1c368af48644fa8995ecbe7138cc0d7900c30.
|
|
This patch implements the vector types (i.e:float2) by making heavy
usage of templating. All vector functions are now outside of the vector
classes (inside the blender::math namespace) and are not vector size
dependent for the most part.
In the ongoing effort to make shaders less GL centric, we are aiming
to share more code between GLSL and C++ to avoid code duplication.
Motivations:
- We are aiming to share UBO and SSBO structures between GLSL and C++.
This means we will use many of the existing vector types and others we
currently don't have (uintX, intX). All these variations were asking
for many more code duplication.
- Deduplicate existing code which is duplicated for each vector size.
- We also want to share small functions. Which means that vector functions
should be static and not in the class namespace.
- Reduce friction to use these types in new projects due to their
incompleteness.
- The current state of the BLI_(float|double|mpq)(2|3|4).hh is a bit of a
let down. Most clases are incomplete, out of sync with each others with
different codestyles, and some functions that should be static are not
(i.e: float3::reflect()).
Upsides:
- Still support .x, .y, .z, .w for readability.
- Compact, readable and easilly extendable.
- All of the vector functions are available for all the vectors types and
can be restricted to certain types. Also template specialization let us
define exception for special class (like mpq).
- With optimization ON, the compiler unroll the loops and performance is
the same.
Downsides:
- Might impact debugability. Though I would arge that the bugs are rarelly
caused by the vector class itself (since the operations are quite trivial)
but by the type conversions.
- Might impact compile time. I did not saw a significant impact since the
usage is not really widespread.
- Functions needs to be rewritten to support arbitrary vector length. For
instance, one can't call len_squared_v3v3 in math::length_squared() and
call it a day.
- Type cast does not work with the template version of the math:: vector
functions. Meaning you need to manually cast float * and (float *)[3] to
float3 for the function calls.
i.e: math::distance_squared(float3(nearest.co), positions[i]);
- Some parts might loose in readability:
float3::dot(v1.normalized(), v2.normalized())
becoming
math::dot(math::normalize(v1), math::normalize(v2))
But I propose, when appropriate, to use
using namespace blender::math; on function local or file scope to
increase readability. dot(normalize(v1), normalize(v2))
Consideration:
- Include back .length() method. It is quite handy and is more C++
oriented.
- I considered the GLM library as a candidate for replacement.
It felt like too much for what we need and would be difficult to
extend / modify to our needs.
- I used Macros to reduce code in operators declaration and potential
copy paste bugs. This could reduce debugability and could be reverted.
- This touches delaunay_2d.cc and the intersection code. I would like to
know @Howard Trickey (howardt) opinion on the matter.
- The noexcept on the copy constructor of mpq(2|3) is being removed.
But according to @Jacques Lucke (JacquesLucke) it is not a real problem
for now.
I would like to give a huge thanks to @Jacques Lucke (JacquesLucke) who
helped during this and pushed me to reduce the duplication further.
Reviewed By: brecht, sergey, JacquesLucke
Differential Revision: http://developer.blender.org/D13791
|
|
These are only used for non-apple unix systems.
|
|
Since CMake 3.16, CMake has native precompiled header (PCH) support.
This change swaps Blender's own PCH implementation with the native implementation.
Previously, PCH was only enabled on Windows however,
this new implementation works on all platforms.
For more information see https://cmake.org/cmake/help/latest/command/target_precompile_headers.html
On my system, Linux with ninja running on an i5 8250U
I saw a 60% reduction in compile times for `bf_freestyle` + linking time.
Reviewed By: LazyDodo, brecht
Differential Revision: https://developer.blender.org/D13797
|
|
Differential Revision: https://developer.blender.org/D13783
|
|
Since Optix 7.3 is required, update the default path accordingly.
|
|
|
|
This adds the remaining bits to enable Metal on macOS. There are still
performance optimizations and other improvements planned, but it should
now be ready for early testing.
This is currently only enabled on in Arm builds for M1 GPUs. It is not
yet working on AMD or Intel GPUs.
Ref T92212
Differential Revision: https://developer.blender.org/D13503
|
|
* Don't link embree / OSL when WITH_CYCLES is disabled
* Simplify lite config by disabling Cycles as a whole using this
* Remove code handling the removed WITH_CYCLES_NETWORK option
|
|
Contributed by luzpaz.
Differential Revision: https://developer.blender.org/D13264
|
|
Followup to https://developer.blender.org/D10288
Reviewed By: Blendify
Differential Revision: https://developer.blender.org/D10346
|
|
We've now done testing to confirm this works with RDNA and RDNA2 AMD GPUs
on Windows. The AMD driver needed for this will soon be released publicly.
|
|
|
|
|
|
Make building the thumbnail extraction executable optional,
disable on macOS as this was not linking, further, macOS doesn't use
this for thumbnail extraction so it could be left disabled.
|
|
|
|
And fix various broken things in the HIP kernel compilation.
|
|
Without this, each cppcheck invocation included all defines/includes
flooding the console with unhelpful information.
Also remove nonexistent directory to exclude.
|
|
Use a temporary directory for faster performance.
|
|
queue_processes() - used for some of the "make check_*" utilities,
wasn't waiting for all jobs to finish before returning.
This conflicted with running cleanup operations.
|
|
This includes much improved GPU rendering performance, viewport interactivity,
new shadow catcher, revamped sampling settings, subsurface scattering anisotropy,
new GPU volume sampling, improved PMJ sampling pattern, and more.
Some features have also been removed or changed, breaking backwards compatibility.
Including the removal of the OpenCL backend, for which alternatives are under
development.
Release notes and code docs:
https://wiki.blender.org/wiki/Reference/Release_Notes/3.0/Cycles
https://wiki.blender.org/wiki/Source/Render/Cycles
Credits:
* Sergey Sharybin
* Brecht Van Lommel
* Patrick Mours (OptiX backend)
* Christophe Hery (subsurface scattering anisotropy)
* William Leeson (PMJ sampling pattern)
* Alaska (various fixes and tweaks)
* Thomas Dinges (various fixes)
For the full commit history, see the cycles-x branch. This squashes together
all the changes since intermediate changes would often fail building or tests.
Ref T87839, T87837, T87836
Fixes T90734, T89353, T80267, T80267, T77185, T69800
|
|
VS2019 had a compiler update moving it into the
range that was used to detect VS2022. This patch
updates the detection to the current VS2022
preview compiler version.
Reported by Jesse Y on chat.
|
|
The /Zc:inline flag is by default off in the MSVC
compiler however when you build with msbuild it adds
it to the build flags on its own.
Ninja however does not decide on its own to add
flags you didn't ask for and was building without
this flag.
This change explicitly adds the compiler flag so
msbuild and ninja builds are once more building
with the same build flags leading to smaller .obj
files when building with ninja and lightening the
workload for the linker.
This flag is available starting MSVC 2013 update 2
so does not need to be guarded with version checks.
|
|
Compressing blendfiles can help save a lot of disk space, but the slowdown
while loading and saving is a major annoyance.
Currently Blender uses Zlib (aka gzip aka Deflate) for compression, but there
are now several more modern algorithms that outperform it in every way.
In this patch, I decided for Zstandard aka Zstd for several reasons:
- It is widely supported, both in other programs and libraries as well as in
general-purpose compression utilities on Unix
- It is extremely flexible - spanning several orders of magnitude of
compression speeds depending on the level setting.
- It is pretty much on the Pareto frontier for all of its configurations
(meaning that no other algorithm is both faster and more efficient).
One downside of course is that older versions of Blender will not be able to
read these files, but one can always just re-save them without compression or
decompress the file manually with an external tool.
The implementation here saves additional metadata into the compressed file in
order to allow for efficient seeking when loading. This is standard-compliant
and will be ignored by other tools that support Zstd.
If the metadata is not present (e.g. because you manually compressed a .blend
file with another tool), Blender will fall back to sequential reading.
Saving is multithreaded to improve performance. Loading is currently not
multithreaded since it's not easy to predict the access patterns of the
loading code when seeking is supported.
In the future, we might want to look into making this more predictable or
disabling seeking for the main .blend file, which would then allow for
multiple background threads that decompress data ahead of time.
The compression level was chosen to get sizes comparable to previous versions
at much higher speeds. In the future, this could be exposed as an option.
Reviewed By: campbellbarton, brecht, mont29
Differential Revision: https://developer.blender.org/D5799
|
|
Change the dylib folder relative to `Blender` executable to be
the same as before rB652fbc200500497a67bd11d18b786587ba34e3d9 and same
as bpy.so : `@loader_path/../Resources/${BLENDER_VERSION}/lib`
|
|
The Xcode IDE can also benefit from the options:
- WINDOWS_USE_VISUAL_STUDIO_SOURCE_FOLDERS
- WINDOWS_USE_VISUAL_STUDIO_PROJECT_FOLDERS
So add suport to these options and also renames them as they are no
longer limited to just Windows and Visual Studio.
Reviewed By: brecht, ankitm
Differential Revision: https://developer.blender.org/D12132
|
|
|
|
|
|
For Blender.app: dropping libomp.dylib next to Blender executable is
enough for it getting picked up since `@executable_path` is an rpath.
For non-distributed binaries datatoc, makesdna, tests etc, code for
copying libomp.dylib to build folder is removed and replaced by
CMake's rpath option for *build* tree.
For bpy.so, the post build rpath change has also been replaced by CMake
rpath option for *install* tree.
Since -id has been changed in D11748, remove the
`install_name_tool -change ...` command.
Any dylib can just be dropped at `MAC_BLENDER_TARGET_DYLIBS_DIR`
hereafter. Appending dylib path to `CMAKE_BUILD_RPATH` will be needed
for datatoc etc if linked against one (instead of copying the
dylibs around).
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D11997
|
|
This is causing issues for some users launching Blender, because EGL indirectly
requires GLVND, which is not installed by default on e.g. Ubuntu.
This reverts commit 0b18a618b88b22663e05eca0f4d976875710e7cc.
Fixes T90374
Ref D12034
|
|
This will replace GLX with EGL for X11. GLEW does not support GLX and EGL
at the same time. Most distributions build GLEW with GLX support, so we
have to use the externally provided GLEW and build with EGL support.
This effectively sets WITH_SYSTEM_GLEW to OFF for all Linux configurations.
Differential Revision: https://developer.blender.org/D12034
|
|
WITH_CYCLES_DEBUG was used for rendering BVH debugging passes. But since we
mainly use Embree an OptiX now, this information is no longer important.
WITH_CYCLES_DEBUG_NAN will enable additional checks for NaNs and invalid values
in the kernel, for Cycles developers. Previously these asserts where enabled in
all debug builds, but this is too likely to crash Blender in scenes that render
fine regardless of the NaNs. So this is behind a CMake option now.
Fixes T90240
|
|
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D12021
|
|
|
|
|
|
Blender did not support to input East Asian characters (Chinese, Japanese,
Korean) on macOS. This patch adds support for Japanese input, by implementing
the appropriate processing for the NSTextInputClient protocol.
Technical notes:
* The conversion candidate window is drawn by the input method program calling
`firstRectForCharacterRange`.
* The string before confirmation (called `composite` in blender) is handled in
the `setMarkedText` method called by the input method program.
* The string after confirmation (called `result` in the blender) is processed
in the `insertText` method called by the input method program.
Ref T51283
Differential Revision: https://developer.blender.org/D11695
|
|
In certain CMake configurations it was possible
that OCIO gave linker errors due to it thinking
it was using the shared library rather than the
static library we ship.
|
|
This patch fixes many minor spelling mistakes, all in comments or
console output. Mostly contractions like can't, won't, don't, its/it's,
etc.
Differential Revision: https://developer.blender.org/D11663
Reviewed by Harley Acheson
|
|
This adds preliminary VS 2022 support, since
there currently is no CMake version that
supports the VS2022 IDE only ninja support
was tested.
IDE support should work without any additional
changes as soon as an updated CMake becomes
available.
As VS2022 appears to keep binary compatibility
with earlier MSVC versions, the current SVN
libraries will work for this version.
|
|
This convention is no longer used for Blender's CMake files.
|
|
|
|
|
|
rB4f81b4b4ce29 mistakenly left out the changes
to platform_win32.cmake causing a linker error
when WITH_GMP and WITH_TBB_MALLOC_PROXY were on.
|
|
In Debian e.g. Clang is part of LLVM, so we need to also check its root
directory sometimes to find Clang files.
|
|
Due to the way we ship the CRT on windows TBB's
malloc proxy was unable to attach it self to
the memory management functions on windows 10.
This change moves ucrtbase.dll out of the blender.crt
folder and back into the main blender folder to side
step some undesirable behaviour on win10 making TBB
once more able to attach it self.
Having this work again, should give a speed
boost in memory allocation heavy workloads
such as mantaflow.
For details on how this only failed on Win10
see T88813
|
|
rB847579b42250 updated the TBB build script
which had some unintended consequences for
windows as the directory layout slightly
changed.
This change adjusts the builder to the new
structure, there are no version/functional
changes.
|
|
|
|
|
|
For 2.93 we bumped the minimum windows requirement
to windows 8.1, but did not do any clean-up of any
win 8/8.1 API usage we dynamically accessed though
LoadLibrary/GetProcAddress.
This patch bumps _WIN32_WINNT to 0x0603 (win 8.1)
and cleans up any API use that was accessed in a
more convoluted way than necessary
Differential Revision: https://developer.blender.org/D11331
Reviewed by: harley, nicholas_rishel
|
