Benchmarking Regexp multicore performance on Rubinius 2.0.0pre with removed onig_lock

gistfile1.txt

rbx -v
rubinius 2.0.0dev (1.8.7 03ecc4ad yyyy-mm-dd JI) [x86_64-apple-darwin10.7.4]
with patch https://gist.github.com/1020720

String#gsub

time rbx -e 's = "." * 16_000_000; n = 1; c = s.size / n; (0...n).map { |i| Thread.new { s[c * i, c].gsub(".", "-") }}.map(&:join)'
real	0m33.407s
user	0m33.437s
sys	0m0.074s

time rbx -e 's = "." * 16_000_000; n = 2; c = s.size / n; (0...n).map { |i| Thread.new { s[c * i, c].gsub(".", "-") }}.map(&:join)'
real	0m23.163s
user	0m41.998s
sys	0m2.463s

time rbx -e 's = "." * 16_000_000; n = 3; c = s.size / n; (0...n).map { |i| Thread.new { s[c * i, c].gsub(".", "-") }}.map(&:join)'
real	0m23.963s
user	0m58.010s
sys	0m7.356s

time rbx -e 's = "." * 16_000_000; n = 4; c = s.size / n; (0...n).map { |i| Thread.new { s[c * i, c].gsub(".", "-") }}.map(&:join)'
real	0m51.476s
user	1m38.169s
sys	0m49.303s