io -> nio where available. Mostly it is just an API change, maybe marginal speedups just by that. The major functional changes however have a huge impact especially on the initial asset conversion. We now read / write files through buffers that provide much more optimal performance. Asset conversion also did a lot of temp files when processing files in WAD archives, now all this is performed in-memory without the need for any temp files.
The initial asset conversion for me is now roughly just 30s with this. So further 2 mins shaved (compared to just the multithreading stuff, the earlier NIO stuff already sped this up a lot). Definitely better than the 10 minutes that we started with.
io -> nio where available. Mostly it is just an API change, maybe marginal speedups just by that. The major functional changes however have a huge impact especially on the initial asset conversion. We now read / write files through buffers that provide much more optimal performance. Asset conversion also did a lot of temp files when processing files in WAD archives, now all this is performed in-memory without the need for any temp files.
The initial asset conversion for me is now roughly just 30s with this. So further 2 mins shaved (compared to just the multithreading stuff, the earlier NIO stuff already sped this up a lot). Definitely better than the 10 minutes that we started with.