From: Larry M. <lm...@bi...> - 2012-12-28 00:30:59
|
> lang cat grep hash loop proc fib sort wc > tcl8.5 0.40 1.33 0.59 0.04 0.36 1.63 1.42 0.68 > tcl 0.37 0.46 0.55 0.05 0.47 1.87 1.31 0.68 > > First line is tcl8.5, second line is a beta of 8.6 w/ the PCRE change. > So it's close to 1.5x for PCRE. Perl added in: lang cat grep hash loop proc fib sort wc pl 0.14 0.15 0.27 0.03 0.15 1.87 0.98 0.50 tcl8.5 0.40 1.33 0.59 0.04 0.36 1.63 1.42 0.68 tcl8.6b 0.37 0.46 0.55 0.05 0.47 1.87 1.31 0.68 slower 2.6 3.1 2.0 1.6 3.1 same 1.3 1.4 last line is tcl8.6beta+pcre compared to perl 5.14. Cat is the one to look at - if you fixed that (not fcopy, actually fixed it) and achieved parity with perl on that one, I think that tcl in general would be as fast or close as perl on all of these. Note that the loop overhead between tcl and perl is pretty similar, yeah, perl is somewhat faster but it's also such a short amount of time, the other factors will dominate. Fix cat.tcl. -- --- Larry McVoy lm at bitmover.com http://www.bitkeeper.com |