Closed Super99Master closed 1 month ago
I haven't looked at the code yet; would it be possible to launch N downscaling processes where N is the number of logical processors on the system?
I haven't looked at the code yet; would it be possible to launch N downscaling processes where N is the number of logical processors on the system?
Yes https://github.com/Super99Master/video2x/commit/a1b2aa380d67fe423d63c113d6bc471b03c8c3ff
Results of Multythreading: Number of threads | Whole time compleation (3 run avg) | Runs 1 | 118,19566 | 117.69998 119.31545 117.57157 4 | 52,39386 | 54.2574 52.24034 50.68386 12 | 40,72011 | 40.01302 41.70235 40.44497
I think 4 threads is good enough without pinning the processor at 100%. With 4 threads it was around 80% utilization
https://github.com/Super99Master/video2x/commits/master
This is more or less how multithreading the downscaling process should work,
Can i leave the downscaling, put that commit into mine, and then continue where i left off don't wanna spend 5 hours downscaling when you spent 4 upscaling
Can i leave the downscaling, put that commit into mine, and then continue where i left off don't wanna spend 5 hours downscaling when you spent 4 upscaling
Sorry, but i don't understand what you mean. Multy threaded downscaling was added in the dev tree
also, huh, i downloaded it from git and the downscaling isn't multi threaded, i checked it on the python code and it isn't there
edit: OH GOD DAMNIT, DEV TREE? I DOWNLOADED THE MASTER
also, huh, i downloaded it from git and the downscaling isn't multi threaded, i checked it on the python code and it isn't there
edit: OH GOD DAMNIT, DEV TREE? I DOWNLOADED THE MASTER
FYI: The 5.0 beta will be streamlined, so it will also be multi-threaded...or multi-processed rather.
Description
Add Hardware Acceleration to the Downscaling process since it's barely using the CPU.
Sources
Running 2 Downscaler for a max of 32% CPU utilization.