> rg uses a lot of memory in the OpenSubtitles test. 903M vs 29M for ugrep. Unlike the previous test, we are not told the size of the file being searched. Whic

> rg uses a lot of memory in the OpenSubtitles test. 903M vs 29M for ugrep. Unlike the previous test, we are not told the size of the file being searched. Which test exactly? That's just likely because of memory maps futzing with the RSS data. Not actually more heap memory. Try with --no-mmap. I'm not sure I understand the rest of your comment about grep -o. Grep tools usually have a flag to print the offset of each match. EDIT: Now that I have hands on a keyboard, I'll demonstrate the mmap thing. First, ugrep: $ time ugrep-4.4.1 -c '\w+\s+Sherlock\s+Holmes\s+\w+' sixteenth.txt 72 real 22.115 user 22.015 sys 0.093 maxmem 30 MB faults 0 $ time ugrep-4.4.1 -c '\w+\s+Sherlock\s+Holmes\s+\w+' sixteenth.txt --mmap 72 real 21.776 user 21.749 sys 0.020 maxmem 802 MB faults 0 And now for ripgrep: $ time rg-14.0.3 -c '\w+\s+Sherlock\s+Holmes\s+\w+' opensubtitles/2018/en/sixteenth.txt 72 real 0.076 user 0.046 sys 0.030 maxmem 779 MB faults 0 $ time rg-14.0.3 -c '\w+\s+Sherlock\s+Holmes\s+\w+' opensubtitles/2018/en/sixteenth.txt --no-mmap 72 real 0.087 user 0.033 sys 0.053 maxmem 15 MB faults 0 It looks like the difference here is that ripgrep chooses to use a memory map by default here. I don't think it makes much of a difference here. If the file were bigger than available memory, then the OS would automatically handle paging.