Bug #13783
closed
Memory allocation / GC does not free unused memory
Description
100.times {x=[]; 1000000.times {|i| x.push [i,i*3]} }
100.times should use (more or less) the same memory as 3.times,
also GC.start after x=[] doesn't help!
Problem: system begins swapping if doing that many times (using large data arrays running for longer times)
Updated by shyouhei (Shyouhei Urabe) over 8 years ago
- Status changed from Open to Feedback
Hmm, I don't know exactly why so this might not fork for you but at least on my machine, the situation seems fixed in latest trunk. Can you test?
zsh % rbenv shell 2.4.1 zsh % ruby -v ruby 2.4.1p111 (2017-03-22 revision 58053) [x86_64-darwin15] zsh % for i in 1 4 16 64; do /usr/bin/time -l ruby -e $i.'times {x=[]; 1000000.times {|i| x.push [i,i*3]} }' |& fgrep resident done 78553088 maximum resident set size 195747840 maximum resident set size 311996416 maximum resident set size 341024768 maximum resident set size zsh % rbenv shell trunk@svn zsh % ruby -v ruby 2.5.0dev (2017-08-04 trunk 59496) [x86_64-darwin15] zsh % for i in 1 4 16 64; do /usr/bin/time -l ruby -e $i.'times {x=[]; 1000000.times {|i| x.push [i,i*3]} }' |& fgrep resident done 73068544 maximum resident set size 183775232 maximum resident set size 183783424 maximum resident set size 185192448 maximum resident set size 
Updated by opti (Andreas Opti) over 8 years ago
I didn't manage to compile Ruby v250 (because of other errrors), nice to see that this bug is fixed now.
(I mentioned it in a mailinglist some months ago).
Now I found in the changelog "r59074 | ko1 | 2017-06-13 11:52:33 +0900 (Tue, 13 Jun 2017) | 5 lines"
- not sure if it's that problem, maybe someone could backport that to 2.4.x ?
Seems also if updating 2.4.x -> 2.4.x+1 I have to reinstall [compile] all gems?
thanks
Opti
Updated by jeremyevans0 (Jeremy Evans) about 6 years ago
- Status changed from Feedback to Closed