Re: [gnugo-devel] olib & xlib definition

[Trevor Morris]

At 04:41 PM 2/6/2002 +0100, Arend Bayer wrote:
>> 
>> I'd like to provide access to "approxlib" from the patterns as
>> well, and think that it's probably sufficient to use "approxlib" 
>> than "accurate_approxlib" usually, so I would propose changing
>> xlib & olib as follows, and adding accurate_[ox]lib:
>> 
>>  accurate_xlib: 
>>    accurate_approxlib(%s, OTHER_COLOR(color), MAX_LIBERTIES, NULL)
>>  accurate_olib: 
>>    accurate_approxlib(%s, color, MAX_LIBERTIES, NULL)
>>  xlib: 
>>    approxlib(%s, OTHER_COLOR(color), MAX_LIBERTIES, NULL)
>>  olib: 
>>    approxlib(%s, color, MAX_LIBERTIES, NULL)
>> 
>
>Is this really worth changing? In my nngs-test suite profiles,
>accurate_approxlib used only 0.2% of CPU time (mainly from calls from
>owl-pattern autohelpers). In contrast, approxlib used 4.8% of CPU time
>(appr. 70% for calls from order_moves, the rest for calls from other
>functions in reading.c). See the parts of the call graph below.

Ah, right, you haven't profiled the experimental pattern-based reading 
patterns (and I wouldn't recommend it until my next patch!).  They
use olib & xlib quite a lot (perhaps more than they should...)  That
was my motivation for the change.  approxlib really is a lot faster
than accurate_approxlib.