VASP多节点并⾏参数设置
VASP currently offers parallelization and data distribution over bands and/or over plane wave coefficients (e and ), and parallelization over k-points (no data distribution, e ).
干香菇的做法To obtain high efficiency on massively parallel systems or modern multi-core machines, it is strongly recommended to u all at the same time.
以下假设计算总核⼼数为N，单节点计算核⼼数为n潮汕菜粿
杨坚的皇后NPAR与NCORE
两者相关：NCORE*NPAR=N, 取其⼀设置即可，vaspwiki上建议优先设置NPAR，但NCORE更为⽅便 ()
NPAR决定能带并⾏策略（NPAR determines the number of bands that are treated in parallel）
NCORE决定轨道并⾏策略（NCORE determines the number of compute cores that work on an individual orbital）
⼤规模并⾏，vaspwiki上建议NPAR≈Sqrt(N) or NCORE = n. 若N开⽅⾮整数，则取开⽅结果附近的整数，这⾥需要注
意：NPAR需要适当选取，使得，以减⼩节点间通信的overhead，否则VASP会采⽤默认设置，即NPAR=N or NCORE=1，默认值适⽤于⼩核数（8核以内）和⼩的通信带宽
，对于100个原⼦左右的unit cell, NCORE∼4；对于⼤的unit cell (more than 400 atoms), NCORE∼12-16 KPAR
The t of k-points is distributed over KPAR groups
KPAR决定K点并⾏策略（KPAR determines the number of k-points that are to be treated in parallel）
choo KPAR such that it is an integer divisor of N
体表
the data is not distributed additionally over k-points
用既也造句总结：
以上源于vaspwiki的⼀般性总结，具体请以实际测试为准！
特别在超算上，⼀定要花时间测试！否则可能吃（算）⼒不讨好！
个⼈实践测试经验
VASP针对k点和能带做了并⾏计算处理，可以从vasp刚开始输出的结果看出
running on N total cores倍的意思
distrk: each k-point on N/KPAR cores, KPAR groups怎么锻炼腰部力量
distr: one band NCORE cores, NPAR groups
其中NCORE需要为n的factor. 相对于NPAR，设置NCORE更为⽅便，因为直接设NPAR还要验证NCORE=N/NPAR是否是n的factor. vaspwiki建议对于100个原⼦左右的unit cell, NCORE∼4；对于⼤的unit cell (more than 400 atoms), NCORE∼12-16. 因此，可根据⾃⼰的体系和资源合理选取.
王者荣耀改头像
