Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo

Close

Code Log


Commit Date  
[r11] by jkbonfield

Release 4.4

2012-08-30 09:03:24 Tree
[r10] by jkbonfield

Minor updates for publishing 4.4. ("make dist")

2012-08-30 09:02:40 Tree
[r9] by jkbonfield

Bug fix to SOLiD mode. Files larger than one block produced errors and
aborted early.

2012-06-01 16:35:18 Tree
[r8] by jkbonfield

Removal of debugging output and added -S (solid) mode to the help. It's
been there for ages, just undocumented.

Ideally this should be autodetected.

2012-06-01 14:25:23 Tree
[r7] by jkbonfield

Fixed initialisation of last2 in encode and decode seq functions. It
wasn't genuingly the reverse complement of last (although it soon will
be after enough loops and shifts).

Improved the memory prefetching in encode_seq8 with -b option
enabled. The same trick unfortunately isn't usable when decoding.

2012-06-01 11:22:11 Tree
[r6] by jkbonfield

Removed redundant code ("nop" change).

2012-05-31 16:22:17 Tree
[r5] by jkbonfield

Experimental and currently non-functioning alternative to base_model.h
that uses one uint32_t instead of 4 uint8_t counters for the
frequencies.

In theory this permits faster updating as some ops can be done in
parallel. Decoding isn't working yet (and hence encoding maybe wrong
too?), but the speed looks promising.

2012-05-25 14:23:58 Tree
[r4] by jkbonfield

Brought SVN up to date to match latest code versions.
(Not a new release, but work in progress.)

Major changes since official release are:

- Support for improved SOLiD formats. Fastq seems variable, so we cope
with either having or not having a quality for the primer base.

- Sped up seq model by improved memory prefetching and also removed
the attempts at working out the most likely base for an N. Just
encode 0 (A) and be done with it.

Maybe we should have fixed prob of 1/SummFreq for all N and assume
they're rare? That allows proper encoding of N rather than relying
on quality 0.

- Investigations into using lookup tables for divides in the range
coder. Only used for sequences currently due to 255 max for
SummFreq.

- Removed SummFreq in base_model and compute it on the fly. It reduces
max memory by 25% and also strangely is faster to compute than keep
up to date

2012-05-25 14:22:02 Tree
[r3] by jkbonfield

fqzcomp-4.3 release

2012-05-25 10:31:22 Tree
[r2] by jkbonfield

Initial import of version 4.3 (the final fqzcomp entry to
SequenceSqueeze contest.

2012-05-25 10:30:11 Tree
[r1] by allura

Initial commit

2011-11-18 12:07:00 Tree