summaryrefslogtreecommitdiff
path: root/workspaces/main
AgeCommit message (Expand)Author
21 hoursrefactor(cnn): rename cnn_effect to cnn_v1_effect for clarityskal
21 hoursarchive(cnn): move CNN v1 to cnn_v1/ subdirectoryskal
21 hoursrefactor(cnn): isolate CNN v2 to cnn_v2/ subdirectoryskal
23 hoursdebug flash-cube-effectskal
24 hoursrefactor(timeline): retiming and effect consolidationskal
28 hoursmisc updatesskal
30 hoursfix(shader): add dfWithID to sdf_test.wgsl for test compatibilityskal
45 hoursfeat(gpu): add SDF camera infrastructure and effect base classskal
46 hourssome tweakingskal
2 daysrefactor(wgsl): consolidate SDF shapes into single common fileskal
2 daysrefactor(wgsl): replace inline fullscreen_vs with common includeskal
2 daysrefactor(wgsl): modularize common shader functionsskal
2 daysFix: CNN v2 compute shader validation errorskal
2 daysCNN v2: bilinear mip-level sampling and UI improvementsskal
3 daysReplace hard clamp with sigmoid activation in CNN v2skal
3 daysFix --mix option: blend prev layer with static p4-p7, not p0-p3skal
3 daysCNN v2: Alpha channel depth handling and layer visualizationskal
3 daysCNN v2: Fix WebGPU validation error in uniform buffer alignmentskal
3 daysCNN v2: Change feature #6 from sin(10*x) to sin(20*y)skal
3 daysCNN v2: Add TODO for flexible feature layout in binary format v3skal
3 daysCNN v2: Add mip-level support to runtime effectskal
3 daysAdd pop-punk drum track from documented sequenceskal
3 daysCNNv2Effect: Document per-layer kernel sizes supportskal
3 daysCNN v2: Refactor to uniform 12D→4D architectureskal
3 daysRemediation: Implement shared common/shaders/ directoryskal
3 daysUpdate workspace.cfg for new directory structureskal
3 daysAdd weights/ subdirectory to workspaces for CNN training outputsskal
3 daysRefactor: Reorganize workspaces and remove assets/ directoryskal
3 daysRefactor: Move application entry points to src/app/skal
4 daystest_demo: Add beat-synchronized CNN post-processing with version selectionskal
4 daysTimeline editor: CPU load visualization + UX improvementsskal
4 daysRefine training script output and validationskal
4 daysTODO: 8-bit weight quantization for 2× size reductionskal
4 daysCNN v2: storage buffer architecture foundationskal
4 daysCNN v2: parametric static features - Phases 1-4skal
4 dayscleanup: remove test-only files from main workspaceskal
5 daysfix: update shader files to use beat_phase instead of beatskal
5 daysfeat: implement beat-based timing systemskal
5 daysadd trained layersskal
5 daysdocs: Update CNN comments and add bias fix summaryskal
5 daysfix: CNN bias accumulation and output format improvementsskal
5 daysupdate cnn codeskal
5 daysfix: Register cnn_conv1x1 snippet and add verificationskal
5 daysfix: Move sigmoid activation to call site in CNN layer shaderskal
5 daysfix: Replace clamp with sigmoid in CNN final layerskal
5 daysformat .wgsl layer code (cosmetics)skal
6 daysopt: Move invariant in1 calculation outside CNN convolution loopsskal
6 daysopt: Vec4-optimize CNN convolution shaders for SIMDskal
6 dayschore: Update CNN architecture to 3×3×3 with new trained weightsskal
6 daysfix: Correct UV coordinate computation to match PyTorch linspaceskal