GitHub-Proxy/PrusaSlicer
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/prusa3d/PrusaSlicer.git synced 2025-07-30 20:32:01 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

SPE-2397: Increased size of time blocks batches used to calculate time estimate and actual speed

Browse Source
This commit is contained in:
enricoturri1966 2024-07-30 09:32:04 +02:00 committed by Lukas Matena
parent 88c94180f7
commit 40dbbd1c35
1 changed files with 10 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/libslic3r/GCode/GCodeProcessor.hpp
View File

@ -367,8 +367,16 @@ namespace Slic3r {
struct Planner
{
// Size of the firmware planner queue. The old 8-bit Marlins usually just managed 16 trapezoidal blocks.
// Let's be conservative and plan for newer boards with more memory.
static constexpr size_t queue_size = 64;
// ! WARNING !
// The previous implementation:
// Let's be conservative and plan for newer boards with more memory.
// static constexpr size_t queue_size = 64;
// was generating artifacts, see: https://dev.prusa3d.com/browse/SPE-2397 because the time blocks shared by two consecutive batches
// may end up being processed multiple times, giving rise to discontinuities.
// Keeping all the time blocks in memory may result in a huge buffer (i.e. greater than 1GB for huge slices), so we set an arbitrary
// batch size of around 128MB.
// This should result in a reduced number of artifacts visible only for objects whose buffers exceed this size.
static constexpr size_t queue_size = 32 * 1024 * 1024 / sizeof(TimeBlock);
// The firmware recalculates last planner_queue_size trapezoidal blocks each time a new block is added.
// We are not simulating the firmware exactly, we calculate a sequence of blocks once a reasonable number of blocks accumulate.
static constexpr size_t refresh_threshold = queue_size * 4;