Free, open-source 3D sculpting for Linux, macOS and Windows: dense meshes, remeshing, painting, built-in AI (image/text to 3D), Send to Unreal, game-ready LODs, Gaussian splats and OpenUSD.
C++ 3d3d-sculptingcppdear-imguidigital-sculptingfbxgame-developmentgaussian-splattinggltfimage-to-3dmcpopenglopenusdsculptingtext-to-3dunreal-engine
0.10.9: What's New section, CHANGELOG under Unreleased, TESTERS.md, REPLY.md (#33); whats_new test checks the 0.10.9 section
islandmedc24190b9e809
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex ce4b8654..87d2f2fb 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -6,6 +6,30 @@ All notable changes to SculptForge are listed here. The format follows## [Unreleased]+### Changed+- Large brushes on dense meshes no longer stall the view: from 2 million vertices a stroke's dabs are placed on the+ stroke thread (the same dabs, in the same order, as the frame loop places them; no preview), so a frame keeps+ drawing while a 240 px dab moves millions of vertices, and a light batch is still waited for within the frame.+ On the standard scene's 240 px brush the slowest frames drop from 139 ms (15M) and 306 ms (33M) to about 30 ms in+ first runs, and lifting the pen no longer waits (162 ms at 33M). `SF_BACKGROUND_DABS=0` turns it off.+- Move on dense meshes no longer stalls: it worked out its whole region's normals and bounds again for every pointer+ sample (several times a frame with a 240 Hz pen, more the slower the frame got: seconds at 33M under load). The+ positions still follow every sample; the normals and bounds now follow once a frame, and the mesh comes out the+ same, bit for bit.+- The first orbit after a long stroke no longer stalls (180-300 ms at 33M): the moving view's lighter stand-ins for+ the changed areas are rebuilt a few milliseconds a frame, the rest drawn in full detail until then.++### Fixed+- Help > Benchmark (and `--benchmark`) checks the memory free now before each scene instead of trusting the+ computer's total: a scene that doesn't fit (15M needs about 2.2 GB free, 33M about 5 GB, plus a margin) is skipped+ with a clear message ("33M skipped: needs ~5 GB free, 1.3 GB available. Close other apps and run again") rather+ than running the system out of memory. The result lists skipped scenes; the command line exits with code 2.+- Subdivide, Divide, Divide to a micro level and Voxel Remesh ask first ("Not enough free memory", Continue anyway or+ Cancel) when the mesh they would build doesn't fit the memory free now.+- Runs started by scripts and agents (`SDL_APP_ID=sculptforge-agent`, or `SF_ISOLATE=1`) use their own settings and+ log folder for every mode (renders, `--bench`, `--calibrate` too), so they no longer write into your+ sculptforge.log or move your session's log aside. Isolated runs remove their temporary folder when they pass.+## [0.10.8] - 2026-11-02### Fixed@@ -16,11 +40,6 @@ All notable changes to SculptForge are listed here. The format follows## [0.10.7] - 2026-11-01### Changed-- Large brushes on dense meshes no longer stall the view: from 2 million vertices a stroke's dabs are placed on the- stroke thread (the same dabs, in the same order, as the frame loop places them; no preview), so a frame keeps- drawing while a 240 px dab moves millions of vertices, and a light batch is still waited for within the frame.- On the standard scene's 240 px brush the slowest frames drop from 139 ms (15M) and 306 ms (33M) to about 30 ms in- first runs, and lifting the pen no longer waits (162 ms at 33M). `SF_BACKGROUND_DABS=0` turns it off.- The membership card at launch is redesigned: the cube mark and the promise up top, the three memberships side byside with their perks checked off and one bold button each (Insider highlighted), and the members' Discord in thefooter. It takes the current look's colours and opens in 150 ms. "Maybe later" now hides it for 30 days, updates@@ -39,21 +58,6 @@ All notable changes to SculptForge are listed here. The format followsInsiders repository and posts to #insiders (`.github/workflows/insider-preview.yml`).### Fixed-- Help > Benchmark (and `--benchmark`) checks the memory free now before each scene instead of trusting the- computer's total: a scene that doesn't fit (15M needs about 2.2 GB free, 33M about 5 GB, plus a margin) is skipped- with a clear message ("33M skipped: needs ~5 GB free, 1.3 GB available. Close other apps and run again") rather- than running the system out of memory. The result lists skipped scenes; the command line exits with code 2.-- Subdivide, Divide, Divide to a micro level and Voxel Remesh ask first ("Not enough free memory", Continue anyway or- Cancel) when the mesh they would build doesn't fit the memory free now.-- Move on dense meshes no longer stalls: it worked out its whole region's normals and bounds again for every pointer- sample (several times a frame with a 240 Hz pen, more the slower the frame got: seconds at 33M under load). The- positions still follow every sample; the normals and bounds now follow once a frame, and the mesh comes out the- same, bit for bit.-- The first orbit after a long stroke no longer stalls (180-300 ms at 33M): the moving view's lighter stand-ins for- the changed areas are rebuilt a few milliseconds a frame, the rest drawn in full detail until then.-- Runs started by scripts and agents (`SDL_APP_ID=sculptforge-agent`, or `SF_ISOLATE=1`) use their own settings and- log folder for every mode (renders, `--bench`, `--calibrate` too), so they no longer write into your- sculptforge.log or move your session's log aside. Isolated runs remove their temporary folder when they pass.- Custom and Community brushes (Clay Slab and other presets) show the whole alpha at Focal Shift -100, as thebuilt-in brushes do since #60: the stamp reaches its square's corners instead of being cut by the preset's roundtip (#75, reported by rodrigomoretto). At the default Focal Shift presets look as before; saved presets are unchanged.diff --git a/REPLY.md b/REPLY.mdnew file mode 100644index 00000000..90f10fba--- /dev/null+++ b/REPLY.md@@ -0,0 +1,5 @@+# Draft replies for 0.10.9 (post at release, after reading)++## #33 Make Sure performances are up to industry standard (bobbe3636)+Another step for dense meshes in 0.10.9: a big brush (240 px) on a 33 million triangle model no longer freezes the view while each dab is worked out. In our standard benchmark scene (first measurements) the slowest frames went from about 300 ms to about 35 ms, and lifting the pen no longer waits. Move keeps up on the same mesh with a fast pen, and the first orbit after a long stroke no longer hitches. Help > Benchmark now also checks free memory first and skips a scene your computer can't hold rather than running it out of memory. If you try it on your heaviest scenes, the benchmark's numbers (Help > Benchmark) would help us a lot.+-- the SculptForge developerdiff --git a/REPORT.md b/REPORT.mdindex 708cb5b4..d2897837 100644--- a/REPORT.md+++ b/REPORT.md@@ -1,4 +1,5 @@-# perf/bigbrush-p99 (0.10.8). STATUS: PAUSED (usage limit). Pushed 1aa0e20 = background dabs (green). Local commit on top = NOT pushed, full ctest NOT run.+# perf/bigbrush-p99 -> 0.10.9. STATUS: merged with origin/main (0.10.8 released, none of this branch in it), builds clean; NOT pushed: full ctest not run yet (queued run killed by the low-memory reaper at load 93). Pushed 1aa0e20 = background dabs only.+New: What's New lines under '## Version 0.10.9' (whats_new.sfui checks it), CHANGELOG lines under [Unreleased] (released sections identical to main), TESTERS.md, REPLY.md (#33 draft).Done + pushed (1aa0e20): background dabs for big brushes (provisional: 33M 240 px p99 306 -> ~33 ms, stall 325 -> ~41, pen lift 162 -> 0), bit-identical (ui_stroke_background); details in the earlier report text in git history.Done locally (unpushed): (1) core/MemoryGuard: free-memory check; benchmark skips a scene that doesn't fit ("33M skipped: needs ~5 GB free, 1.3 GB available..."), JSON "skipped", exit 2, submit.py refuses; Subdivide/Divide/micro/remesh ask "Not enough free memory" (app/MemoryConfirm.cpp). Tests: test_memory_guard (mocked free memory), ui_memory_confirm - both pass. Verified: SF_AVAILABLE_MEMORY_MB=1240 --benchmark standard skips both scenes, rc=2.(2) Move at 33M: positions per pointer sample (Sculptor::grabMove), normals+BVH once a frame (grabFinish via App::flushGrab); bit-identical (test_grab_coalesce passes). 33M Move at load 8-12: p50 28, p99 53, stall 62 ms (was p50 57, multi-second stalls under load). At load 22 swap-bound (GPU contention), unconfirmed quiet.diff --git a/TESTERS.md b/TESTERS.mdnew file mode 100644index 00000000..a2601159--- /dev/null+++ b/TESTERS.md@@ -0,0 +1,6 @@+# 0.10.9 testers: big brushes, Move and memory on dense meshes+- Divide a model to 15-35M triangles (or Help > Benchmark > Standard 33M) and sculpt with a big Clay brush (200+ px, symmetry on): the view should stay smooth and lifting the pen instant; the surface may trail the pen a little on huge brushes. Tell us if any frame freezes for more than a moment.+- Same mesh, Move brush, fast strokes with a pen: should keep up (no multi-second freezes), then orbit right after a long stroke: no hitch.+- With other heavy apps open (little free RAM), run Help > Benchmark: a scene that doesn't fit should be skipped with a "needs ~X GB free" message instead of the system running out of memory.+- On a big mesh, try Subdivide / Divide / Voxel Remesh at a high setting with little free memory: a "Not enough free memory" dialog should ask first; Cancel leaves the mesh, Continue does it.+- Please include your RAM, GPU and OS, and Help > Benchmark's numbers if you ran it.diff --git a/docs/guide/whats-new.md b/docs/guide/whats-new.mdindex fa262573..f80ce842 100644--- a/docs/guide/whats-new.md+++ b/docs/guide/whats-new.md@@ -5,12 +5,20 @@ chapter opens by itself the first time you start a new version; you can always f**Help > What's New**. Every detail is in the[changelog](https://github.com/islandmed/SculptForge/blob/main/CHANGELOG.md).-## Version 0.10.8+## Version 0.10.9- **Big brushes stay smooth on dense meshes**: on models from about 4 million triangles a stroke's dabs are workedout beside the view, so a 240 px brush on 33 million triangles no longer freezes the view for a third of asecond at a time (slowest frames about 35 ms instead of 300), and letting go of the pen is instant. The strokeitself comes out the same.+- **Move keeps up on dense meshes**: Move no longer slows to a crawl on 33 million triangles with a fast pen, and the+ first orbit after a long stroke no longer hitches.+- **No more running out of memory**: **Help > Benchmark** skips a scene your computer hasn't got the free memory+ for (and says how much to free), and Subdivide, Divide, Micro and Voxel Remesh ask first when the result+ wouldn't fit in the memory free now.++## Version 0.10.8+- **Tablets in mouse mode work again**: on some Windows tablets with **Mouse mode** on (Preferences > Tablet) thebrush stopped following the pen in 0.10.7. When the system can't say where its cursor is, the pen now moves thebrush by how far it moves, from where it was (#72).diff --git a/docs/performance.md b/docs/performance.mdindex ab2b32e3..43bed072 100644--- a/docs/performance.md+++ b/docs/performance.md@@ -12,7 +12,7 @@ sculptforge-cli --bench ... # the same without the GPU (CI,Machine for all figures: Apple M1 Pro (8 performance + 2 efficiency cores, 32 GB unified memory, 30 GB usableunder Asahi Linux), integrated GPU, Mesa 26.1 (OpenGL 4.6). How the benchmark works is described under v0.5 below.-## 0.10.8: background dabs (perf/bigbrush-p99)+## 0.10.9: background dabs, Move once a frame (perf/bigbrush-p99)`--benchmark standard` placed every stroke in the frame loop (predictive strokes are off by default), and on the33M head one 240 px dab moves 3-11M vertices (the symmetric brush covers a large part of the face). Where one suchdiff --git a/tests/ui/whats_new.sfui b/tests/ui/whats_new.sfuiindex 0e5cf4b3..ad8e78a6 100644--- a/tests/ui/whats_new.sfui+++ b/tests/ui/whats_new.sfui@@ -5,8 +5,9 @@ click "Help##palette.help"click "What's New##help.whats_new"expect_value guide_open == 1expect_guide whats-new.md+expect_text "Version 0.10.9"+expect_text "Big brushes stay smooth on dense meshes"expect_text "Version 0.10.8"-expect_text "Tablets in mouse mode work again"expect_text "Version 0.10.7"expect_text "Version 0.10.6"expect_text "Version 0.10.5"