Dear Mr. Zhuang,
I found that cp2kdata (v0.7.3) parses large AIMD runs slowly.
For example, the speed of energy or force parsing is about 100 steps/s for 4000-step long CP2K output. However, for the simulation of 50000 steps it takes about 1 s to parse a single step. Therefore, the whole output will be parsed for 50000 s (energies) + 50000 s (forces) + few minutes for other data = more than a day.
There is obviuos nonlinearity of parsing time depending on output length. I assume that the problem is the usage of the pop() function instead of e.g. "for" cycle, since with pop() the parser should manipulate with the entire array each step.
Please consider the option to accelerate this part of the code. My expertise is quite low so I failed to do it by myself.
Sincerely yours,
Dmitry
Dear Mr. Zhuang,
I found that cp2kdata (v0.7.3) parses large AIMD runs slowly.
For example, the speed of energy or force parsing is about 100 steps/s for 4000-step long CP2K output. However, for the simulation of 50000 steps it takes about 1 s to parse a single step. Therefore, the whole output will be parsed for 50000 s (energies) + 50000 s (forces) + few minutes for other data = more than a day.
There is obviuos nonlinearity of parsing time depending on output length. I assume that the problem is the usage of the pop() function instead of e.g. "for" cycle, since with pop() the parser should manipulate with the entire array each step.
Please consider the option to accelerate this part of the code. My expertise is quite low so I failed to do it by myself.
Sincerely yours,
Dmitry