IBM and AMD unveil Pelagos, a GPU-accelerated tool to advance quantum error correction research, optimizing simulations for fault-tolerant quantum computing.
IBM and AMD have jointly launched Pêlagos, an open-source software tool designed to accelerate quantum error correction (QEC) research. The tool addresses the computational challenges of simulating QEC decoders, which are critical for achieving fault-tolerant quantum computing. By leveraging GPU acceleration, Pêlagos enables researchers to conduct large-scale simulations more efficiently, facilitating the development and evaluation of advanced decoding strategies.
Quantum error correction encodes logical information across multiple physical qubits, requiring decoders to interpret error syndromes and recover data. Evaluating these decoders involves simulating numerous error patterns, a process that demands significant computational resources. Pêlagos accelerates this workflow by utilizing modern GPU parallelism, enabling researchers to process more configurations and achieve greater statistical certainty in their results.
The tool’s core innovation lies in its GPU-accelerated implementation of Relay-BP, a QEC decoder adapted from belief propagation algorithms used in classical low-density parity-check (LDPC) codes. Relay-BP iteratively exchanges error information to improve decoding accuracy. Pêlagos processes multiple syndromes in parallel, achieving throughput rates up to 97 times higher than a single CPU core, as reported in a joint IBM and AMD white paper.
Pêlagos was built using the Python-based Triton framework, which simplifies GPU kernel development by combining a NumPy-like interface with automated tuning for memory layouts and workload partitioning. This approach ensures strong performance across diverse GPU architectures, including AMD Instinct™ GPUs. The software also incorporates device fingerprinting and cached tuning results, making it adaptable for future research needs.
Source link







