Add 5 batch GPU operations to the full stack: - zk_knowledge_verify_batch (Schnorr knowledge proofs) - zk_dleq_verify_batch (discrete-log equality proofs) - bulletproof_verify_batch (range-proof polynomial checks) - bip324_aead_encrypt_batch (ChaCha20-Poly1305 AEAD encrypt) - bip324_aead_decrypt_batch (ChaCha20-Poly1305 AEAD decrypt) Stack changes: - gpu_backend.hpp: 5 virtual methods (default → Unsupported) - gpu_backend_cuda.cu: full CUDA implementations using existing kernels from zk.cuh + new bip324.cuh - gpu_backend_opencl.cpp: explicit stubs with TODO(parity) - gpu_backend_metal.mm: explicit stubs with TODO(parity) - ufsecp_gpu.h: C ABI declarations with Doxygen - ufsecp_gpu_impl.cpp: C ABI implementations - cuda/include/bip324.cuh: extracted BIP-324 device code from bench_bip324_transport.cu into shared header |
||
|---|---|---|
| .. | ||
| gpu_backend.hpp | ||
| gpu_cuda_host_helpers.h | ||