feat: add neural rendering subsystem (NTC, radiance cache, neural pos…#379
Merged
Conversation
…t-processing) Implements a lightweight GPU neural inference engine and three neural rendering features built on top of it: - Neural Inference Engine: compute shader-based MLP evaluator with CPU fallback, no external ML framework dependency - Neural Texture Compression (NTC): per-block MLP encoding of textures with .ntex file format, CPU training, GPU/CPU decompression - Neural Radiance Cache: multi-resolution hash grid + MLP for cached indirect lighting (Instant NGP-inspired) - Neural Post-Processing: neural denoiser (IDenoiser backend) and neural super-resolution enhancement for SparkSR All systems integrate with existing engine infrastructure: - TextureCompressor: added Neural format variant - DenoiserInterface: added Neural backend enum - UpscalingSettings: added neuralEnhancement flag - EngineContext: NeuralInferenceEngine registered at startup - CMake: ENABLE_NEURAL_RENDERING toggle (ON by default) 22 new files (~4000 lines), 6 HLSL compute shaders, 4 test files with 40+ tests. All 3714 tests pass. https://claude.ai/code/session_01TS53MsdxLYbFPpiGkEQmoR
Contributor
Code Coverage (GCC + lcov) |
Contributor
✅ CI Errors ResolvedAll previously reported errors have been fixed. All builds passing. Last checked: 2026-04-05T08:41:30Z |
MSVC doesn't transitively include <algorithm> unlike GCC/Clang, causing a build failure on Windows CI. https://claude.ai/code/session_01TS53MsdxLYbFPpiGkEQmoR
Contributor
Code Coverage (GCC + lcov) |
Replace naive scalar CPU fallback in NeuralInferenceEngine with SSE2/AVX2-accelerated MLP forward pass via MultiISA dispatch pattern. Large batches (>=16 samples) auto-parallelize across JobSystem workers. - New CpuNeuralInference class: SIMD layer eval, weight padding, batch dispatch - NeuralInferenceEngine.EvaluateCPU now delegates to SIMD path - NeuralPostProcessing: batch all patch pixels in single inference call - NeuralRadianceCache: batch all queries in single inference call - 14 new tests covering SIMD correctness, batching, and reference matching https://claude.ai/code/session_01CVmfPaczLYeyZNiCqsvh82
Claude/cpu neural features g jnlb
Contributor
Code Coverage (GCC + lcov) |
- Fix extra whitespace in DynamicQualityTypes.h neuralEnhancement field - Replace hardcoded /tmp/ paths with std::filesystem::temp_directory_path() in neural test files (fixes ACCESS_VIOLATION on Windows CI) https://claude.ai/code/session_01TS53MsdxLYbFPpiGkEQmoR
Contributor
Code Coverage (GCC + lcov) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…t-processing)
Implements a lightweight GPU neural inference engine and three neural rendering features built on top of it:
All systems integrate with existing engine infrastructure:
22 new files (~4000 lines), 6 HLSL compute shaders, 4 test files with 40+ tests. All 3714 tests pass.
https://claude.ai/code/session_01TS53MsdxLYbFPpiGkEQmoR