Very interesting and instructive, thanks!
( scratchpad ) [ testconcat ] gc benchmark [ testmake ] gc benchmark
[ testintersperse ] gc benchmark
--- Data stack:
48568
22680
16535
On Jun 19, 2009, at 6:26 AM, Slava Pestov wrote:
> Your code still has the inefficiency of dynamic variable lookups on
> every call to , as well as resizing the underlying array of the vector
> being constructed. Here is a more direct version:
>
> : intersperse ( seq1 seq2 -- seq )
> 2dup [ length ] bi@ + <vector>
> [ '[ [ _ push ] bi@ ] 2each ] keep
> { } like ;
>
> Slava
>
> On Thu, Jun 18, 2009 at 3:56 AM, Phil Dawes<[email protected]> wrote:
>> I tried:
>> { 1 2 3 } { 3 2 1 } [ [ swap , , ] 2each ] { } make
>>
>> to remove the extra array creation and the concat pass. AFAICS it's
>> quicker on large arrays:
>>
>> : testmake ( -- )
>> 100000 >array dup reverse
>> [ [ swap , , ] 2each ] { } make drop ;
>>
>> : testconcat ( -- )
>> 100000 >array dup reverse
>> [ 2array ] 2map concat drop ;
>>
>> ( scratchpad ) [ testconcat ] gc benchmark [ testmake ] gc benchmark
>
> ------------------------------------------------------------------------------
> Crystal Reports - New Free Runtime and 30 Day Trial
> Check out the new simplified licensing option that enables unlimited
> royalty-free distribution of the report engine for externally facing
> server and web deployment.
> http://p.sf.net/sfu/businessobjects
> _______________________________________________
> Factor-talk mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/factor-talk
------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables unlimited
royalty-free distribution of the report engine for externally facing
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Factor-talk mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/factor-talk
