[daala] Why not upgrading Daala range coding with much faster rANS or tANS?
Jarek Duda
dudajar at gmail.com
Fri Aug 14 07:41:42 PDT 2015
Hi,
You are probably aware of new large alphabet accurate entropy coders:
tANS: Huffman-like (tabled, one symbol/lookup: ~500MS/s/core decoding),
but using nearly accurate probabilities (compression ratio like range
coding), e.g. https://github.com/Cyan4973/FiniteStateEntropy
rANS: similar to range coding, but many times faster: using single
multiplication/symbol, more convenient frequency handling, better for
SIMD, e.g. https://github.com/rygorous/ryg_rans
They are used e.g. in Apple LZFSE (tANS), CRAM 3.0 of European
Bioinformatics Institute (rANS), LZNA of RAD Game Tools (rANS) and many
others:
http://encode.ru/threads/2078-List-of-Asymmetric-Numeral-Systems-implementations
Paper:
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7170048&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D7170048
I wanted to ask why it is not considered for Daala?
Best,
Jarek
--
dr Jarosław Duda
Institute of Computer Science and Computer Mathematics,
Jagiellonian University, Cracow, Poland
http://th.if.uj.edu.pl/~dudaj/
More information about the daala
mailing list