demo.git - Vide-coded 64k demo system

Age	Commit message (Expand)	Author
10 hours	fix(cli): enable WAV dump without DEMO_HEADLESS requirement	skal
10 hours	fix(cli): change --dump_wav to --dump-wav and add error handling	skal
10 hours	feat(gpu): add shader labels for better error identification	skal
10 hours	feat(spectral-editor): add waveform intensity viewer for sample offset	skal
24 hours	feat(gpu): add two-pass raymarching infrastructure	skal
25 hours	feat(gpu): add SDF camera infrastructure and effect base class	skal
26 hours	docs: add audio system enhancement tasks	skal
26 hours	docs: mark tracker humanization task as IMPLEMENTED	skal
26 hours	feat(tracker): add sample offset and humanization	skal
26 hours	some tweaking	skal
28 hours	refactor(wgsl): consolidate SDF shapes into single common file	skal
28 hours	refactor(wgsl): replace inline fullscreen_vs with common include	skal
28 hours	style: Apply clang-format to codebase	skal
29 hours	refactor(wgsl): modularize common shader functions	skal
29 hours	refactor(gpu): Relocate effects to src/effects and streamline includes	skal
29 hours	feat: Move test_platform.cc to util subdirectory	skal
29 hours	docs: Update agent state from project documentation	skal
29 hours	Refactor: extract setup_standard_textures helper	skal
29 hours	Refactor: add gpu_create_post_process_texture helper	skal
30 hours	Refactor: add gpu_create_texture_view_2d helper	skal
31 hours	Docs: WGPU helper functions reference	skal
31 hours	Refactor: factorize WGPU init code for 3D tests	skal
31 hours	Fix: CNN v2 compute shader validation error	skal
31 hours	Refactor: factorize common WGPU patterns into helper functions	skal
36 hours	CNN v2: bilinear mip-level sampling and UI improvements	skal
41 hours	Factor common test patterns into reusable utilities	skal
41 hours	Remove non-functional Gantt test scripts	skal
41 hours	Update docs: CNN v2 sigmoid activation summary	skal
42 hours	Replace hard clamp with sigmoid activation in CNN v2	skal
42 hours	Add --lr parameter to CNN v2 training pipeline	skal
42 hours	Document CNN v2 training pipeline improvements	skal
42 hours	Streamline CNN v2 training pipeline output	skal
42 hours	Fix CNN v2 training: always save final checkpoint, derive num_layers	skal
43 hours	Add --output-weights option to CNN v2 training pipeline	skal
43 hours	Fix --mix option: blend prev layer with static p4-p7, not p0-p3	skal
43 hours	Fix CNN v2 static feature channel mapping (p4-p7 → channels 8-11)	skal
43 hours	CNN v2 web tool: Remove vizScale, always show raw layer values	skal
43 hours	CNN v2 web tool: Fix vizScale for final layer preview	skal
43 hours	gen_identity_weights: Change --mix to 50-50 blend	skal
43 hours	gen_identity_weights: Add --p47 option for static feature visualization	skal
43 hours	gen_identity_weights: Add --mix option for static feature blending	skal
43 hours	CNN v2 tool: Fix off-by-one in composited layer filenames	skal
43 hours	CNN v2: Fix weight buffer offset bug	skal
44 hours	cnn_test: --weights now overrides layer config from .bin file	skal
44 hours	cnn_test: Add --weights option for runtime weight loading	skal
44 hours	CNN v2: Remove vizScale, always clip to [0,1]	skal
44 hours	CNN v2: Fix Layer 0 visualization scale (was 0.5, now 1.0)	skal
44 hours	CNN v2: Add debugging tools for mismatch investigation	skal
44 hours	CNN v2 training: Fix float64/float32 dtype mismatch in depth feature	skal
44 hours	CNN v2: Alpha channel depth handling and layer visualization	skal