From: Eric Wong <normalperson@...>
Date: 2015-12-08T08:18:35+00:00
Subject: [ruby-core:71941] Re: [Ruby trunk - Feature #11786] [Open] [PATCH] micro-optimize case dispatch even harder

SASADA Koichi <ko1@atdot.net> wrote:
> On 2015/12/08 13:53, Eric Wong wrote:
> >> > how to mark literal objects?
> > Take ideas from nobu (r52708 :)
> 
> I meant marking, not for freeing a table.

Oops :x   I added rb_mark_set to the Data_Wrap_Struct calls

http://80x24.org/spew/20151208080840.24265-1-e@80x24.org/raw

benchmark results:
minimum results in each 10 measurements.
Execution time (sec)
name	a	b
loop_whileloop2	0.104	0.104
vm2_case*	0.072	0.062
vm2_case_lit*	0.540	0.539
vm2_case_small*	0.060	0.053

Speedup ratio: compare with the result of `a' (greater is better)
name	b
loop_whileloop2	0.994
vm2_case*	1.153
vm2_case_lit*	1.002
vm2_case_small*	1.136

> I'm not sure it is worth or not. IMO keeping simple with Hash is better
> on this case with your measured improvements.

Yeah, it's pretty minor.  I'm not even sure if opt_case_dispatch
makes any difference at all for most real code.  st_table still
has a lot of space overhead and id_table is not usable here.