As a rudimentary strategy for speeding up conversion of large images, consider parallel processing in tiles. Experiment and see if it is viable to do by slicing image separately and starting multiple phstl processes with the same scale settings. Compare to similar approach using single image and multiple phstl processes invoked with different window regions. (If that works, investigate how straightforward it would be to build a parallel processing tile mode into the script.)
As a rudimentary strategy for speeding up conversion of large images, consider parallel processing in tiles. Experiment and see if it is viable to do by slicing image separately and starting multiple phstl processes with the same scale settings. Compare to similar approach using single image and multiple phstl processes invoked with different window regions. (If that works, investigate how straightforward it would be to build a parallel processing tile mode into the script.)