- PME long-ranged interactions can now run on a single GPU, which means many fewer CPU cores are needed for good performance.
- Optimized SIMD support for recent CPU architectures: AMD Zen, Intel Skylake-X and Skylake Xeon-SP.
也就是為了GPU與CPU加速都有所更新。另外還有一些新的功能:
- AWH(Accelerated Weight Histogram),在計算自由能的時候可以克服free energy barriers的方法
- Add a new dual-list dynamic-pruning algorithm for the short-ranged interactions
- Add a physical validation suite:在安裝的時候可以測試有沒有問題,注意這東西可能要跑幾個小時
- Include all Berendsen and Parrinello-Rahman schemes
- Removed hybrid GPU+CPU nonbonded mode
- QM/MM: removed optimization and transition-state search
- Updated application clock handling on Pascal+ GPUs
- Removed continuation from gmx convert-tpr:續跑不再用gmx convert-tpr來做,而是直接用checkpointing來做
沒有留言:
張貼留言