Commit Graph
21 Commits
Author SHA1 Message Date
Tanish Manku e3210f6288 docs: ADR 0110, roadmap, changelog, performance and licence records for stroke coverage (M3) 2026-11-17 23:34:19 -07:00
Tanish Manku 0a6c085a9d docs: ADR 0063 animated PNG and WebP export, sizes versus GIF (M4) 2026-11-11 18:03:14 -07:00
Tanish Manku 2a8dc349b3 docs: ADR 0054, .fla import numbers, roadmap and changelog (M4) 2026-11-11 00:49:36 -07:00
Tanish Manku 340512ef4f docs: ADR 0048 GPU binning and hairlines, performance, roadmap, changelog, attribution (M3) 2026-11-10 20:24:19 -07:00
Tanish Manku c276127743 docs: ADR 0047 onion skin controls and Edit Multiple Frames, roadmap, changelog, performance (M1) 2026-11-10 19:08:19 -07:00
Tanish Manku 863e8af60a fix(gpu): tests and bench build with merge drawing's Shape.merge; ADR renumbered to 0032 after merge drawing took 0031 (M3) 2026-11-09 20:45:47 -07:00
Tanish Manku 313745b204 feat(gpu): the GPU is the default stage renderer; ADR 0031 (M3)
The shell's gpu-stage feature is on by default and Preferences > Stage >
Renderer defaults to Graphics card (GPU), with the CPU drawing whenever
there is no usable card, a tile is past the GPU's limits or anything
fails. The setting is stored under a new key: every earlier build wrote
the old one as Cpu without anyone choosing it. CI installs lavapipe so the
parity tests run on a software adapter.

Measured (release, RTX 5090): 4K filtered sample 3.9 ms on Vulkan and
4.2 ms on DX12 (CPU 31.5 ms); 200 bitmap fills 4.4 / 4.3 ms (CPU 37.2 ms).
Docs: ADR 0031, performance, roadmap, architecture, CHANGELOG.
2026-11-09 19:48:49 -07:00
Tanish Manku eb85dd2411 feat(gpu): GPU stage draws into the window's textures, no readback; ADR 0029 (M3)
The GPU stage shares eframe's wgpu device (egui_wgpu::RenderState) and
draws each stage tile into a storage texture the shell registers with
egui as it is. Fallbacks: a headless device with readback (window on
OpenGL), then the CPU oracle. Every call runs in thread-local error
scopes; nothing is installed on the window's device; a work bound per
dispatch guards the driver watchdog.

The engine stays GPU-free: Raster::Native carries an opaque image, cached
under its own 384 MB budget; only the shell downcasts it.

perf, gpu_bench on RTX 5090 (Vulkan), 32 threads:
  4K, 1,000 shapes: CPU 16.2-19.6 ms, read-back 8.4, direct 2.1-2.4 ms
  4K, 10k shapes:   CPU 25.8-28.1 ms, read-back 23.2, stage tiles 10.2 ms
  4K, masked 10k:   CPU 74.8-80.0 ms, stage tiles 10.9 ms
Binning stays on the CPU (ADR 0029: not the bottleneck as stage tiles).
2026-11-09 16:58:49 -07:00
Tanish Manku 244e21527b chore: merge main (audio, plug-ins, motion tweens) into the GPU stage (M3) 2026-11-09 03:58:26 -07:00
Tanish Manku 9173796835 docs: GPU stage ADR is 0019, as assigned by the lead (M3) 2026-11-09 03:44:33 -07:00
Tanish Manku 38209b8913 docs: ADR 0014 GPU stage, measured timings, roadmap and status (M3) 2026-11-09 03:44:33 -07:00
Tanish Manku d1798ee6c5 docs: one decisions list in performance.md after the rebase (M5) 2026-11-09 03:43:36 -07:00
Tanish Manku b5f3b5463e docs: audio decision record is ADR 0020, as assigned by the lead (M5) 2026-11-09 03:43:36 -07:00
Tanish Manku cf7e39c251 docs: audio decision record renumbered to ADR 0015 after the XFL text ADR took 0014 (M5) 2026-11-09 03:43:35 -07:00
Tanish Manku 23715fc937 docs: sound timings before the decisions list in performance.md (M5) 2026-11-09 03:43:28 -07:00
Tanish Manku b7748949d1 docs: sound and audio layers (ADR 0014), roadmap, changelog, compliance, CI ALSA headers (M5)
Measured (release, 120 s 48 kHz stereo): decode WAV 129 ms, Ogg 147 ms,
MP3 99 ms; export_audio 95 ms at 48 kHz, 128 ms resampled to 44.1 kHz
(about 1000x real time).
2026-11-09 03:43:28 -07:00
Tanish Manku 83c542b11a docs: renumber the autosave ADR to 0015 after the XFL text ADR took 0014 (M8) 2026-11-09 02:19:42 -07:00
Tanish Manku b467344408 docs: crash-recovery autosave, ADR 0014 and measured cost (M8)
perf numbers (release, 32 threads), 200k-instance 65 MB stress document:
editing thread 6.6-15.4 us, serialize 124-128 ms, write+fsync+rename
52-55 ms, background total 183-184 ms. 800k instances (262 MB): 9.9 us
on the editing thread, 725 ms in the background.
2026-11-09 02:19:41 -07:00
Tanish Manku 9adaf9dd5d perf(render): 1080p back under budget (19.8 -> 15.5 ms): parallel background fill, early edge rejection, deferred frees
Measured with anicraft-cli bench (release, 32 threads), 3 runs each: render_1080p 19.4-20.2 -> 15.3-15.8 ms, render_zoom400 11.6-11.9 -> 9.0-9.3 ms. Output byte-identical.
2026-11-06 22:19:38 -07:00
Tanish Manku ada6505bcd perf: enforce machine-independent budgets in tests; bench --check for timing budgets 2026-11-06 22:10:34 -07:00
Tanish Manku e3913afa36 Initial commit v0.1.0 early alpha
AniCraft 0.1.0 (early alpha): milestones M0 and M1. Contributors are credited in AUTHORS.md.
2026-11-05 23:58:25 -07:00