3MF/OrcaSlicer output, placement algorithms (-A), consolidation (-C) - #38
Open
weasel0x00 wants to merge 9 commits into
Open
weasel0x00 wants to merge 9 commits into
weasel0x00 wants to merge 9 commits into
Conversation
…g generator
- CLI 3MF export (-m): emits a single OrcaSlicer/BambuStudio multi-plate
project; each placed part is its own object named after the original part
file, and each plate is restored as a separate plate in the slicer
(model_settings.config + project_settings.config + Application tag).
- Fit search (-i ideal, -g step, -N target): grow the plate size from an
ideal up to the bed size (-W/-H), adding a plate only when parts won't fit,
always preferring the fewest plates then the smallest size.
- Dependency-free ZIP writer (Zip.{h,cpp}) and 3MF writer (ThreeMF.{h,cpp});
no new build dependencies (CLI still needs only pthreads).
- gen_plater_conf.py: generate a plater.conf from a directory of STL parts,
with [a]-prefix "accent parts" and _xN quantity rules.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the byte-at-a-time Bitmap::overlaps scan with a lazily-built, mutable bit-packed occupancy mask (64 pixels/word) and a shifted-AND collision test. The mask is invalidated on setPoint and rebuilt on demand, so the plate mask is rebuilt once per placement and reused across the whole position sweep; part masks are built once. Keeps the original scan as overlapsScalar for validation/benchmarking, and adds bench_overlap (microbenchmark + correctness check). Microbenchmark (300mm@0.5mm plate, 80mm disk part): validation: 50000 checks, 0 mismatches empty plate (miss): 71.7x faster filled plate (mixed): 45.1x faster Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…RIFY_OVERLAP) Refactor overlaps() to a single return and add an optional assertion that cross-checks the bit-packed result against overlapsScalar on every call. Verified: a full -S solve on real geometry runs clean (no mismatch). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…e size Split Part::load (size-independent geometry: disk read, pixelize, rotate, trim) from Part::applyPlateSize (cheap per-plate-size rotation filtering). allBmp owns all rotation bitmaps; bmp is a per-size view. processFit now loads parts once and binary-searches the smallest feasible plate size instead of reloading and linearly scanning every step. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Maintain a per-column skyline on each rectangular plate (Plate::colHeight, updated in place()). placePartSkyline drops each part onto the skyline: the resting row is computed in O(part width) from the part's per-column bottom profile, and is overlap-free by construction (everything above the skyline is empty), so no collision tests are needed in the search at all. Opt-in via -k; falls back to the brute-force grid for circular plates. A/B vs brute force: plate count: equal on jobs A/B, fewer on job C (2 vs 3) -- never worse speed: ~1.5x (-j 0.5) to ~1.9x (-j 0.3) faster, grows with precision Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
sortMode was never set in the constructor, so a run without -S read indeterminate stack memory and could nondeterministically take the multiple-sort (time-seeded shuffle) path. Default it to REQUEST_SINGLE_SORT so single-sort runs are reproducible. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add PLACER_SCORE_CONTACT: rank skyline candidates by how much of the part's outline touches walls/placed parts (denser packing), gravity as tie-break. solve() adds contact-scored placers alongside the gravity ones when -K is set, so the solver keeps whichever packs better (never a regression). Note: like skyline, this drops onto the top profile and cannot fill interior holes/cavities -- best for hole-free parts. Opt-in via -K (implies -k). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same algorithm and result as the default brute force (full collision test, so holes/cavities are filled identically), but prunes the position search using score monotonicity: scan y upward per column and stop at the first feasible cell (the column's lowest-score spot), and skip columns once even their y=0 score can't beat the current best. Quality-identical to brute (verified byte-identical output, including hole nesting), ~1.3x faster on top of the bit-packed collision. Opt-in via -b; works for any plate shape. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Flags: - Replace -k/-K/-b with a single -A <algorithm>: brute (default) | pruned | skyline | contact. Absent -A keeps the original brute force. - Add -C: post-pass that tries to drop a plate by re-packing all parts into one fewer plate (with rearrangement, exploring largest/smallest/shuffled orderings), accepted only when it strictly reduces the plate count. Supporting changes: - Placer gains setMaxPlates(); place() returns NULL if parts don't fit within the cap (used by the consolidation repack). Uncapped behavior unchanged. - Fix uninitialized Placer::myThread (crashed when a Placer was destroyed without ever starting a thread, e.g. the consolidation placers). Validated: 2/50 random suboptimal jobs dropped a plate via -C (never worse); all -A values run with/without -C and compose with -m and the fit search. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
First in a series that splits a large set of CLI/placement improvements into smaller, reviewable PRs (rather than one big request). Each subsequent PR is stacked on this one and will be opened/rebased as this merges.
This PR (9 commits):
Bitmapcollision (64-bit mask overlap) for speed-A:skyline(bottom-left drop),contact(max-contact),pruned(hole-aware pruned brute force)-Cconsolidation pass (drop a plate by repacking; only kept if it strictly reduces plate count)Request::sortMode(was uninitialized)Follow-ups in the series: CLI ergonomics + fit search (
-i) + tall-part centering (-T); simulated-annealing placement (-A anneal) + volume balancing (-B) + shrink fit (-z); and a correctness/security hardening pass.