NVIDIA 在 JUPITER 上模拟 104 量子比特容错协议
NVIDIA 于 10 月 19 日宣布,在欧洲首台百亿亿次超算 JUPITER 上使用 8,192 块 GH200 GPU 完成 104 量子比特容错量子处理器模拟,称这是迄今规模最大的此类完整模拟。该运行验证了 magic-state cultivation,所需物理量子比特比此前方法少 90%。模拟包含相干相位误差与泄漏效应,团队借助 GB200 NVL72 开发平台将原本数周的模拟缩短至数天。
The technical post describes the engineering behind the run. The simulator reorganizes operations while preserving their dependencies and measurement statistics, and it reuses storage as qubits are measured and released. It packs groups of three-level systems into a binary representation that cuQuantum’s cuStateVec library can execute: five three-level systems have 243 basis states and fit within eight binary storage bits, which provide 256 entries. In the 8,192-GPU configuration, a 45-bit enco
匹配字段正文