I am pleased to announce BRC, a fast and strong bwt postcoder I've been working on.
Source code: https://github.com/loxxous/Behemoth-Rank-Coding
Exe's: https://github.com/loxxous/Behemoth-...oding/releases
(Repost since the last thread vanished)
I am pleased to announce BRC, a fast and strong bwt postcoder I've been working on.
Source code: https://github.com/loxxous/Behemoth-Rank-Coding
Exe's: https://github.com/loxxous/Behemoth-...oding/releases
(Repost since the last thread vanished)
SolidComp (7th November 2018)
Yeah, what happened to all the threads today?
I think the server crashed or lost power in the middle of the night and had to rollback to the last backup, unfortunately losing all 3 pages of this thread and a ton of other threads.
Reattaching my rans_dynamic experiments which includes a RLE0 version. Warning, this is a total rag-tag bunch of experiments with some not even having working decoders. It's simply an exploration task, but some good may come of it.
BRC4 is marginally larger, as it has RLE0 already built in so this model isn't ideal, although it's faster as it has less data to encode.Code:$ cc -O3 -g -DARITH order0.16b_rle0c.c $ ./a.out < ~/scratch/data/enwik9bwt1024.brc2 > /tmp/_ 1000983528 to 169252683 Took 7098 msec, 141 MB/s $ ./a.out -d < /tmp/_ > /tmp/_2 Took 14362 msec, 69 MB/s
Edit: actually that statement isn't correct. It's largely the same time on both brc2 and brc4 as the entropy encoder is quicker with low complexity data it seems.