3DPrinting/PrusaSlicer-NonPlainar
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Speed improvement of initial G-code preview:

Browse source 
1) Preallocating the vertex / index buffers to limit reallocation.
2) Inlining the pushing into the vertex / index buffers.
3) Running the vertex buffer generator on a limited number of threads
   as the generator does not scale well due to memory pressure.
   Not using all the threads leaves some of the threads to G-code
   generator.
This commit is contained in:
Vojtech Bubnik 2022-08-23 11:28:25 +02:00
parent 14e0cd0e96
commit 9aee934d53
4 changed files with 28 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/slic3r/GUI/GLCanvas3D.cpp
View file

@ -6873,6 +6873,11 @@ void GLCanvas3D::_load_print_object_toolpaths(const PrintObject& print_object, c
return volume;
};
const size_t volumes_cnt_initial = m_volumes.volumes.size();
// Limit the number of threads as the code below does not scale well due to memory pressure.
// (most of the time is spent in malloc / free / memmove)
// Not using all the threads leaves some of the threads to G-code generator.
tbb::task_arena limited_arena(std::min(tbb::this_task_arena::max_concurrency(), 4));
limited_arena.execute([&ctxt, grain_size, &new_volume, is_selected_separate_extruder, this]{
tbb::parallel_for(
tbb::blocked_range<size_t>(0, ctxt.layers.size(), grain_size),
[&ctxt, &new_volume, is_selected_separate_extruder, this](const tbb::blocked_range<size_t>& range) {
@ -7047,6 +7052,7 @@ void GLCanvas3D::_load_print_object_toolpaths(const PrintObject& print_object, c
vol->indexed_vertex_array.shrink_to_fit();
#endif // ENABLE_LEGACY_OPENGL_REMOVAL
});
}); // task arena
BOOST_LOG_TRIVIAL(debug) << "Loading print object toolpaths in parallel - finalizing results" << m_volumes.log_memory_info() << log_memory_info();
// Remove empty volumes from the newly added volumes.