summaryrefslogtreecommitdiff
AgeCommit message (Expand)Author
24 hoursdocs: Streamline auxiliary texture init doc for handoffskal
24 hoursfix: Use actual resolution in RotatingCubeEffectskal
24 hoursdocs: Update tech doc with Renderer3D initialization fixskal
24 hoursfix: Use initialized_ flag instead of ctx_.device checkskal
24 hoursfix: Complete auxiliary texture initialization fixskal
24 hoursdocs: Add auxiliary texture initialization tech docskal
24 hoursfix: Call resize() before init() to set dimensions firstskal
24 hoursrefactor: Use lazy initialization for auxiliary texturesskal
24 hoursfix: Resolve auxiliary texture resolution mismatch bugskal
24 hoursrefactor: Simplify effect render API and fix uniform initializationskal
25 hoursadd --save-intermediates to train.py and cnn_testskal
26 hoursfix: Move sigmoid activation to call site in CNN layer shaderskal
26 hoursfix: Replace clamp with sigmoid in CNN final layerskal
26 hoursfeat: Add early stopping to CNN trainingskal
26 hoursfix: CNN training/inference to match WGSL sliding windowskal
26 hoursformat .wgsl layer code (cosmetics)skal
27 hoursfix: Guard cnn_test build with STRIP_ALL checkskal
27 hoursrefactor: Modularize CMake build system into 10 specialized modulesskal
28 hoursfix: CNN test tool GPU readback with wgpuDevicePollskal
28 hoursdebug: Add shader load verification and GPU sync to CNN test toolskal
29 hoursfix: CNN test tool ping-pong bug and RGBA16Float intermediatesskal
29 hoursfeat: Add CNN shader testing tool with GPU texture readbackskal
35 hoursfix: Use patch-based inference to match CNN training distributionskal
35 hoursopt: Move invariant in1 calculation outside CNN convolution loopsskal
36 hoursopt: Vec4-optimize CNN convolution shaders for SIMDskal
37 hourschore: Update CNN architecture to 3×3×3 with new trained weightsskal
37 hoursdocs: Update CNN training documentation with patch extractionskal
37 hoursfeat: Add salient-point patch extraction for CNN trainingskal
38 hoursfix: Use ClampToEdge sampler for CNN to avoid edge wrappingskal
38 hoursfix: Correct UV coordinate computation to match PyTorch linspaceskal
38 hoursfix: Add clamp to CNN final layer to match PyTorch trainingskal
39 hoursfix: PostProcessHelperTest signature and fixture sharingskal
39 hoursrefactor: Optimize CNN grayscale computationskal
39 hoursupdate train_cnn.py and shaderskal
40 hoursdocs: Add inference mode to training documentationskal
40 hoursfeat: Add inference mode to train_cnn.py for ground truth generationskal
40 hoursfix: CNN training normalization pipeline consistencyskal
40 hoursudpate CNN shader code.skal
41 hoursfeat: Add --shader-only option to convert_shadertoy.pyskal
41 hoursrefactor: Optimize CNN normalization to eliminate redundant conversionsskal
41 hoursupdate timeline.seqskal
41 hoursfix: Flip Y-axis to match ShaderToy coordinate conventionskal
41 hoursrefactor: Improve convert_shadertoy.py to generate compile-ready codeskal
41 hoursfeat: Add Scene1 effect from ShaderToy (raymarching cube & sphere)skal
42 hourschore: Remove incomplete CubeSphere effectskal
42 hoursdocs: Fix EFFECT keyword syntax and add automation-friendly workflowskal
42 hoursfeat: Add ShaderToy conversion toolsskal
42 hoursfix: Support variable kernel sizes in CNN layer generationskal
42 hoursdocs: Document WGPU builder refactoring in COMPLETED.mdskal
42 hoursrefactor: Factor WGPU boilerplate into builder pattern helpersskal