imp-for and native-live-stack
Matthew Fluet
fluet@CS.Cornell.EDU
Fri, 7 Dec 2001 12:02:07 -0500 (EST)
Just for fun, I looked at imp-for with -native-live-stack true:
> MLton0 -- mlton-stable -drop-pass localRef -native-live-stack false
> MLton1 -- mlton-stable -drop-pass localRef -native-live-stack true
> MLton2 -- mlton-stable -drop-pass foobar -native-live-stack false
> MLton3 -- mlton-stable -drop-pass foobar -native-live-stack true
> compile time
> benchmark MLton0 MLton1 MLton2 MLton3
> imp-for 0.64 0.62 0.60 0.62
> run time
> benchmark MLton0 MLton1 MLton2 MLton3
> imp-for 13.32 9.51 8.23 6.73
> run time ratio
> benchmark MLton1 MLton2 MLton3
> imp-for 0.71 0.62 0.51
> size
> benchmark MLton0 MLton1 MLton2 MLton3
> imp-for 21,909 21,941 20,485 20,341
That's pretty impressive, even if it is a contrived benchmark.