On Tue, Nov 05, 2002 at 03:20:57PM +0000, Paul Makepeace wrote: > On Tue, Nov 05, 2002 at 01:20:15AM +0000, David Cantrell wrote:
> > Scaling perl CGIs up to a few hundred a second is merely hard, but not > > impossible. > > Definitely on the "difficult" side of "hard" :-) Agreed. > $ time perl -e '$a = `perl -le "print q[hello]"` for (1..200)' > real 0m1.068s > user 0m0.600s > sys 0m0.380s > $ host -t hinfo paulm.com > paulm.com HINFO "Intel-1.8GHz-P4" "Debian-GNU/Linux" > $ > > For fun, > > $ time perl -e '$a = `perl -MCGI -le "print CGI->header, CGI->h1(q[hello])"` for >(1..200)' > real 0m10.329s > user 0m8.400s > sys 0m0.990s > $ $ time perl -e '$a = `perl -le "print q[hello]"` for (1..200)' real 0m2.353s user 0m1.350s sys 0m0.960s $ time perl -e '$a = `perl -MCGI -le "print CGI->header, CGI->h1(q[hello])"` for(1..200)' real 0m22.025s user 0m19.560s sys 0m2.350s Here's the first 1%: $ time perl -e '$a = `perl -MCGI= -le "print CGI->header, CGI->h1(q[hello])"` for(1..200)' real 0m21.659s user 0m19.370s sys 0m2.280s Scarily, here's another 29%: time perl -e '$a = `perl5.00503 -MCGI= -le "print CGI->header, CGI->h1(q[hello])"` for(1..200)' real 0m15.654s user 0m13.930s sys 0m1.610s (perl was 5.8, 5.6.1 built by me, so probably the same compiler options as the other two is: real 0m20.931s user 0m18.340s sys 0m2.470s ) I think using 5.005_03 is a fair speedup, given that your example isn't using any 5.6.1 features (yet), and isn't tickling any 5.005_03 bugs. "Obviously" real optimisation problems aren't this simplistic. Nicholas Clark
