Community Tip - You can subscribe to a forum, label or individual post and receive email notifications when someone posts a new topic or reply. Learn more! X
I need to increase solve speed for a large worksheet. I click the multithreading button. First, it is not clear how to detoggle the pick. Second, the CPU time went up. Under what conditions does this option work. Or is this just a vaporware feature of P8?
thank you
@DS_10314450 wrote:
I need to increase solve speed for a large worksheet. I click the multithreading button. First, it is not clear how to detoggle the pick. Second, the CPU time went up. Under what conditions does this option work. Or is this just a vaporware feature of P8?
thank you
Multithreading sure works in Prime, but as far as the quality is concerned, I can't make a conclusive statement. If I conclude from the rest of the quality of Prime to that of the implementation of multithreading, I would cautiously say that there could still be room for improvement. 😉
Keep in mind that multithreading may speed up calculation of some worksheets that contain a number of independent calculations,
There are also linear calculations which cannot be parallelized efficiently and which have even less favorable runtimes due to the increased overhead in multithreading mode.
So you could just give it a try yourself and chose the best settings for your sheet (even though multithreading is not a worksheet setting but an application setting).
This article
https://www.ptc.com/en/support/article/CS56790
cryptically says "Multithreading forces parallel processing: slow processes are getting faster, fast processes are getting slower". Hmmm ???
There also may be more efficient ways to speed up a worksheet apart from a more capable speedier hardware and usage of multi cores. But it involves digging into the calculations and maybe find more efficient ways to achieve the same results. But this is cumbersome and time consuming and no success is guaranteed.
As an example it is found that using vectorization is a bit speedier than using program loops (but of course not every loop can be replaced by vectorization).
Or applying a more efficient algorithm can have more impact on execution time than a couple of hardware generations (compare Quicksort to Bubble sort as a simple example).