[Xapian-devel] Optimized VSEncoding

Hurricane Tong zhangshangtong.cpp at qq.com
Wed Mar 12 04:30:09 GMT 2014


Hi,


I optimized the code of VSEncoder,
and encode/decode the whole list, linux.postlist.


Encoding time :  
VSEncoder 103429ms
InterpolativeEncoder 684ms


Decoding time:
VSDecoder 756ms
InterpolativeDecoder 925ms


( compile with -o2, and the Interpolative Encoder/Decoder is just from the Xapian's code, bitstream.cc )


VS algorithm performs better in decoding, at the price of encoding time.


I analyse the VSEncoder, most time is spent on getting the optimal split, I try to do some optimization, but it seems very difficult to make significant progress.


I don't think the position list of a term will be frequently modified, so I think it is acceptable to sacrifice some encoding time for faster decoding.




------------------
Shangtong Zhang,Second Year Undergraduate,
School of Computer Science,
Fudan University, China.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xapian.org/pipermail/xapian-devel/attachments/20140312/dc9d17b1/attachment.html>


More information about the Xapian-devel mailing list