Oh, forgot to include dynopt numbers under that older macosx build: ~/projects/jruby ➔ jruby -Xcompile.dynopt=true -Xinvokedynamic.invocation=false --server -J-d32 bench/bench_fib_recursive.rb 5 35 9227465 1.366000 0.000000 1.366000 ( 1.293000) 9227465 0.949000 0.000000 0.949000 ( 0.948000) 9227465 0.965000 0.000000 0.965000 ( 0.965000) 9227465 0.932000 0.000000 0.932000 ( 0.932000) 9227465 0.924000 0.000000 0.924000 ( 0.924000)
As mentioned before, the old build was just a hair slower than dynopt. - Charlie On Thu, May 26, 2011 at 6:36 PM, Charles Oliver Nutter <head...@headius.com> wrote: > On Thu, May 26, 2011 at 6:21 PM, Charles Oliver Nutter > <head...@headius.com> wrote: >> No worries...I'll repost 32-bit asm in a moment. > > Attached. This is just fib again. 64-bit is definitely faster than > 32-bit for this benchmark, but the older macosx build is also still > faster than the current mlvm: > > old macosx: > > 9227465 > 1.235000 0.000000 1.235000 ( 1.154000) > 9227465 > 1.012000 0.000000 1.012000 ( 1.012000) > 9227465 > 0.969000 0.000000 0.969000 ( 0.969000) > 9227465 > 0.962000 0.000000 0.962000 ( 0.962000) > 9227465 > 0.953000 0.000000 0.953000 ( 0.953000) > > current mlvm: > > 9227465 > 1.505000 0.000000 1.505000 ( 1.406000) > 9227465 > 1.175000 0.000000 1.175000 ( 1.174000) > 9227465 > 1.770000 0.000000 1.770000 ( 1.771000) > 9227465 > 1.255000 0.000000 1.255000 ( 1.255000) > 9227465 > 1.140000 0.000000 1.140000 ( 1.141000) > > - Charlie > _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
