demo.git - Vide-coded 64k demo system

Age	Commit message (Expand)	Author
29 hours	feat(gpu): add SDF camera infrastructure and effect base class	skal
30 hours	some tweaking	skal
33 hours	refactor(wgsl): consolidate SDF shapes into single common file	skal
33 hours	refactor(wgsl): replace inline fullscreen_vs with common include	skal
33 hours	refactor(wgsl): modularize common shader functions	skal
35 hours	Fix: CNN v2 compute shader validation error	skal
41 hours	CNN v2: bilinear mip-level sampling and UI improvements	skal
46 hours	Replace hard clamp with sigmoid activation in CNN v2	skal
2 days	CNN v2: Alpha channel depth handling and layer visualization	skal
2 days	CNN v2: Fix WebGPU validation error in uniform buffer alignment	skal
2 days	CNN v2: Change feature #6 from sin(10x) to sin(20y)	skal
2 days	CNN v2: Add TODO for flexible feature layout in binary format v3	skal
2 days	CNN v2: Add mip-level support to runtime effect	skal
2 days	CNNv2Effect: Document per-layer kernel sizes support	skal
2 days	CNN v2: Refactor to uniform 12D→4D architecture	skal
3 days	Remediation: Implement shared common/shaders/ directory	skal
3 days	test_demo: Add beat-synchronized CNN post-processing with version selection	skal
3 days	Refine training script output and validation	skal
3 days	TODO: 8-bit weight quantization for 2× size reduction	skal
3 days	CNN v2: storage buffer architecture foundation	skal
4 days	CNN v2: parametric static features - Phases 1-4	skal
4 days	cleanup: remove test-only files from main workspace	skal
4 days	fix: update shader files to use beat_phase instead of beat	skal
4 days	feat: implement beat-based timing system	skal
4 days	add trained layers	skal
4 days	docs: Update CNN comments and add bias fix summary	skal
4 days	fix: CNN bias accumulation and output format improvements	skal
4 days	update cnn code	skal
4 days	fix: Register cnn_conv1x1 snippet and add verification	skal
5 days	fix: Move sigmoid activation to call site in CNN layer shader	skal
5 days	fix: Replace clamp with sigmoid in CNN final layer	skal
5 days	format .wgsl layer code (cosmetics)	skal
5 days	opt: Move invariant in1 calculation outside CNN convolution loops	skal
5 days	opt: Vec4-optimize CNN convolution shaders for SIMD	skal
5 days	chore: Update CNN architecture to 3×3×3 with new trained weights	skal
5 days	fix: Correct UV coordinate computation to match PyTorch linspace	skal
5 days	fix: Add clamp to CNN final layer to match PyTorch training	skal
5 days	refactor: Optimize CNN grayscale computation	skal
5 days	update train_cnn.py and shader	skal
5 days	fix: CNN training normalization pipeline consistency	skal
5 days	udpate CNN shader code.	skal
5 days	refactor: Optimize CNN normalization to eliminate redundant conversions	skal
5 days	fix: Flip Y-axis to match ShaderToy coordinate convention	skal
5 days	feat: Add Scene1 effect from ShaderToy (raymarching cube & sphere)	skal
5 days	fix: Support variable kernel sizes in CNN layer generation	skal
5 days	feat: CNN RGBD→grayscale with 7-channel augmented input	skal
5 days	fix: Resolve CNN effect black screen bug (framebuffer capture + uniforms)	skal
5 days	feat: Add multi-layer CNN support with framebuffer capture and blend control	skal
6 days	feat: Add coordinate-aware CNN layer 0 for position-dependent stylization	skal
6 days	feat: Add CNN post-processing effect with modular WGSL architecture	skal