demo.git - Vide-coded 64k demo system

Age	Commit message (Expand)	Author
16 hours	fix: Resolve WGPU threading crash in cnn_test	skal
17 hours	update cnn code	skal
18 hours	refactor: Use linspace(-1,1) directly for coords	skal
18 hours	fix: Compute gray from [0,1] RGB in CNN shader generator	skal
19 hours	fix: Register cnn_conv1x1 snippet and add verification	skal
23 hours	docs: Sync GEMINI.md with CLAUDE.md structure	skal
23 hours	docs: Streamline auxiliary texture init doc for handoff	skal
23 hours	fix: Use actual resolution in RotatingCubeEffect	skal
23 hours	docs: Update tech doc with Renderer3D initialization fix	skal
23 hours	fix: Use initialized_ flag instead of ctx_.device check	skal
23 hours	fix: Complete auxiliary texture initialization fix	skal
23 hours	docs: Add auxiliary texture initialization tech doc	skal
23 hours	fix: Call resize() before init() to set dimensions first	skal
23 hours	refactor: Use lazy initialization for auxiliary textures	skal
23 hours	fix: Resolve auxiliary texture resolution mismatch bug	skal
23 hours	refactor: Simplify effect render API and fix uniform initialization	skal
24 hours	add --save-intermediates to train.py and cnn_test	skal
25 hours	fix: Move sigmoid activation to call site in CNN layer shader	skal
25 hours	fix: Replace clamp with sigmoid in CNN final layer	skal
25 hours	feat: Add early stopping to CNN training	skal
25 hours	fix: CNN training/inference to match WGSL sliding window	skal
26 hours	format .wgsl layer code (cosmetics)	skal
26 hours	fix: Guard cnn_test build with STRIP_ALL check	skal
27 hours	refactor: Modularize CMake build system into 10 specialized modules	skal
27 hours	fix: CNN test tool GPU readback with wgpuDevicePoll	skal
28 hours	debug: Add shader load verification and GPU sync to CNN test tool	skal
28 hours	fix: CNN test tool ping-pong bug and RGBA16Float intermediates	skal
28 hours	feat: Add CNN shader testing tool with GPU texture readback	skal
34 hours	fix: Use patch-based inference to match CNN training distribution	skal
34 hours	opt: Move invariant in1 calculation outside CNN convolution loops	skal
36 hours	opt: Vec4-optimize CNN convolution shaders for SIMD	skal
36 hours	chore: Update CNN architecture to 3×3×3 with new trained weights	skal
36 hours	docs: Update CNN training documentation with patch extraction	skal
36 hours	feat: Add salient-point patch extraction for CNN training	skal
37 hours	fix: Use ClampToEdge sampler for CNN to avoid edge wrapping	skal
37 hours	fix: Correct UV coordinate computation to match PyTorch linspace	skal
37 hours	fix: Add clamp to CNN final layer to match PyTorch training	skal
38 hours	fix: PostProcessHelperTest signature and fixture sharing	skal
38 hours	refactor: Optimize CNN grayscale computation	skal
38 hours	update train_cnn.py and shader	skal
39 hours	docs: Add inference mode to training documentation	skal
39 hours	feat: Add inference mode to train_cnn.py for ground truth generation	skal
39 hours	fix: CNN training normalization pipeline consistency	skal
39 hours	udpate CNN shader code.	skal
40 hours	feat: Add --shader-only option to convert_shadertoy.py	skal
40 hours	refactor: Optimize CNN normalization to eliminate redundant conversions	skal
40 hours	update timeline.seq	skal
40 hours	fix: Flip Y-axis to match ShaderToy coordinate convention	skal
40 hours	refactor: Improve convert_shadertoy.py to generate compile-ready code	skal
41 hours	feat: Add Scene1 effect from ShaderToy (raymarching cube & sphere)	skal