正常结束模拟后，如果觉得模拟时间让人觉得不够长时，可以使用tpbconv写一个延长模拟的tpr文件，一般格式如下：
tpbconv -s topol.tpr -f traj.trr -e ener.edr -extend 1000 -o newtopol.tpr
1000表示延长步数
5.0.x 用的是gmx convert-tpr命令
example:
gmx convert-tpr -s step7_production1000W.tpr -f step7_production1000W.trr -e step7_production1000W.edr -extend 1000 -o 1000W+1000.tpr
trn version: GMX_trn_file (single precision)
Read trr frame 5000: step 10000000 time 200000.000
Using frame of step 10000000 time 200000
Extending remaining runtime of by 1000 ps (now 50000 steps)
Writing statusfile with starting step 10000000 and length 50000 steps...
time 200000.000 and length 1000.000 ps
然后执行新tpr的运算：
gmx mdrun -s 1000W+1000.tpr -cpnum -v
然后会进行运算了：
Command line:
gmx mdrun -s 1000W+1000.tpr -cpnum -v
Running on 1 node with total 24 cores, 48 logical cores
Hardware detected:
CPU info:
Vendor: GenuineIntel
Brand: Genuine Intel(R) CPU @ 2.20GHz
SIMD instructions most likely to fit this hardware: AVX2_256
SIMD instructions selected at GROMACS compile time: AVX2_256
Reading file 1000W+1000.tpr, VERSION 5.1.4 (single precision)
Using 48 MPI threads
NOTE: This file uses the deprecated 'group' cutoff_scheme. This will be
removed in a future release when 'verlet' supports all interaction forms.
starting mdrun 'Martini system'
10050000 steps, 201000.0 ps (continuing from step 10000000, 200000.0 ps).
step 10000000, will finish Wed Apr 26 15:22:04 2017
NOTE: Turning on dynamic load balancing
step 1000000 ,will finish Wed Apr 26 13:44:49 2017vol 0.54 imb F 5%
说明：这个其实很快的，只是我的服务器同时运行了多个项目，有的在跑2000ns的，所以比较慢了。