As more and more hardware is moving towards multithreading I found couple places where Blender could be improved regarding performance.
I know that not every operation can be parallelised, but nevertheless I want to share my feedback.
Firtst example is saving and opening compressed .blend files (when compression is enabled in preferences) - this issue might be somewhat overlooked. Compression is great as it saves a lot of diskspace but it’s very annoing when user needs to wait for it to do it’s job every time Ctrl+S is used during work. Currently compression is using only 1 CPU core.
Same thing can be said about official 3DPrint Addon Check All function which also is single threaded. High quality 3D models for printing can contain a lot of geometry and that creates a bottleneck when a model needs to be checked.
Last thing is Decimate modifier. I found this tool great for reducing polycount but with large meshes it’s limited again by utilising only 1CPU core. Without any caching ant the way this modifier is working there could be a lot trial and error until the final polycount-quality ratio is found.