Potential performance problem with new parser framework and large files
This plots some parser runs of the new and the old FHI-aims parser (compared to mainfile size).
As you see, the general performance is comparable. Maybe the new has a little steeper time-to-input relation, but thats not too bad. The green oval is a little weird. Maybe this is because I disabled the parser instance reuse.
The problematic one is the red oval. For very large files, we get these extra long run times. Don't let you fool by the logarithmic scale. This is quite drastic and raises the average exec time from 100s to 1.000s. Ten fold. Instead of 1.000.000 calcs, I only get 100.000 calcs overnight. It's kind of a deal breaker.
This does not happen with the old parser, see also:
@ladinesa Could this be memory limitations due to the memory input mapping thing? Would there be a work around?
@himanel1 Not sure if this is related to #481 (closed)