[Kinosearch] Sort::External 0.10_3

[Marvin Humphrey]

On Jul 25, 2005, at 10:29 PM, Marvin Humphrey wrote:

> I'll have to go try the benchmark test with single character lines  
> rather than 60 character lines.  Maybe tomorrow I'll throw away a  
> bunch of the work I did today.

Even when the lines are 1 byte long and delimited by 16 bytes, the  
linesep/split approach is faster than lexer/regex-capture.

The code goes!

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/

------------------------------------------------------------

cache_level_hit_disk: 4.2569 wallclock secs
     ( 3.76 usr +  0.06 sys =  3.82 CPU) @ 13.09/s (n=50)
cache_level_hit_disk_linesep: 3.89435 wallclock secs
     ( 2.73 usr +  0.07 sys =  2.80 CPU) @ 17.83/s (n=50)

mem_thresh_hit_disk: 4.65152 wallclock secs
     ( 4.16 usr +  0.05 sys =  4.21 CPU) @ 11.87/s (n=50)
mem_thresh_hit_disk_linesep: 4.2672 wallclock secs
     ( 3.06 usr +  0.09 sys =  3.15 CPU) @ 15.88/s (n=50)




_______________________________________________
Kinosearch mailing list
Kinosearch at rectangular.com
http://www.rectangular.com/mailman/listinfo/kinosearch