First of all I wan to be clear, that this question is not about the collection performance, but indexing performance.
What is the performance bottleneck of the indexing step?
Some context: Action in game that should take under 1 second took over 10 minutes with TT data collection, and produced over 80 GB run file. (input conditions not part of the question)
What exactly does the indexing do?
"Debug Engine Host Process" uses 37-41 % of the CPU, ~2 GB RAM, and 12-26 MB/s disk IO It uses the all 32 CPU cores almost the same, so from that I assume it is well parallelized. RAM usage is so low that it can be ignored, I assume. Disk IO is also low 0-3% active time because it works on relatively fast NVME drive, with all 4 PCIe gen3 lanes directly from the CPU. This drive is less than 10% used capacity, so even if it would be in the part that is not always SLC, the cells would be in SLC mode, because of enough free space.
I am quite interested what are your thoughts about this, and if you think I can do something to make it faster, like some hidden option to use twice as many threads?