Polyhedra completes Expander framework optimization to improve performance and reliability

AI+ZK public chain Polyhedra said it completed the Expander framework optimization last week. Key improvements include: completing end-to-end integration of GPU Prover with the Expander framework, which now seamlessly switches CPU and GPU execution modes through environment variables to achieve transparent GPU acceleration without additional configuration; optimized FPGA arithmetic primitives for Versal HBM devices, including adders and multipliers for M31Ext3, Goldilocks, and BN254 fields, using high bandwidth memory to maximize throughput and minimize latency, laying the foundation for complete hardware acceleration on FPGAs; benchmarking the preliminary arithmetic core to verify functional correctness and throughput expansion.