Sergey "Shnatsel" Davidoff
Sergey "Shnatsel" Davidoff
Yeah, here's the GUI that analyzes this format: https://github.com/KDE/heaptrack
`heaptrack --analyze "/path/to/file.gz"` works, you don't even need to ungzip it.
`heaptrack rg --no-ignore AddressSanitizer`, it's that simple. By the way, it also shows a ton of short-lived allocations somewhere in colored printing even though I don't have a ton of...
GNU grep only uses 125Mb at its peak: ``` time grep -r AddressSanitizer 60.91user 208.19system 15:22.34elapsed 29%CPU (0avgtext+0avgdata 125588maxresident)k 179260304inputs+0outputs (2major+56824minor)pagefaults 0swaps ``` Here's the -uuu command: ``` time rg...
To find the many small allocations, go to bottom-up view and sort by the number of allocations, then select the first entry. This is what the bottom-up list sorted by...
`-j1` makes a huge difference. For a good part of the execution the memory use is just 1Mb! The final result is even lower than GNU grep! ``` time rg...
I have 4 cores, no SMT. `/proc/$pid/status` shows 5 threads for the `rg` process when running `rg --no-ignore AddressSanitizer`
Ah, I also have a directory structure that's very wide this time around, so that's what probably hogs all the memory. Some directories have 600,000 files in them. That's probably...
Ah, so even leaving the files to be searched aside for a moment, completing some work on the directories may result in _more_ directories being discovered that we have to...
~~I have 10 directories with a million files each.~~ I'll try the branch, thanks!