TFHE-rs v0.6: Zero-Knowledge Support and Signed Integer Operations on GPU
Blog post from Zama
TFHE-rs v0.6 introduces significant updates, notably enhancing cryptographic capabilities by incorporating Zero-Knowledge Proofs to verify the correctness of public key encryption without revealing the encrypted message. This update also improves GPU support, particularly for signed integer operations and parallel processing, allowing faster computations compared to CPUs. The release includes a new C API enabling cross-language support and provides performance improvements such as stabilized multi-bit Programmable Bootstrapping (PBS). The advanced use of H100 GPUs, which offer better performance over previous GPUs like V100, is highlighted, showcasing their capability in large-scale machine learning tasks. Additional enhancements involve operations like checked division, improved multiplication speed, and new noise distributions, with future plans to reduce ciphertext sizes and support multi-GPU computations, all aimed at bolstering efficiency and performance in cryptographic applications.