i added some more tests in later versions. current version ist 0.380.39. i think the interesting part is new+param+output, and with that and a 18k template i get:

 all_ht: 1 wallclock secs ( 1.00 usr +  0.00 sys =  1.00 CPU) @ 20.00/
+s (n=20)
all_htc: 1 wallclock secs ( 1.07 usr +  0.00 sys =  1.07 CPU) @ 120.56
+/s (n=129)
all_htj: 1 wallclock secs ( 1.08 usr +  0.02 sys =  1.10 CPU) @ 763.64
+/s (n=840)
[download]