Save on reallocations

Simpler way to add an eleent not already present

Move search into its own function

Ignore memo databases

Don't try and remove an index if there aren't any

Memoize bit pattern generation

record the maxterms we find

Constant isn't good enough

Actually search for maxterms

Remove debugging prints

Allow cube parity code to use fewer than nine indices

Reorganise documentation

no longer used

Separate out memoization,and use it to support multiple output bit queries in one run

Give cubeattack.py more meaningful name

Make memoize sync the deatabase on every write

Make cache object directly accessible.

Memoize cube parities

break cube attack into own module

Add disk-based memoizer

update is your friend :-)

Latest maxterms from Itai

Add memoizer

Fixes to linearity tests

use new framework for writing results

break out the work of reading and writing item files into a separate
module

Test a binary function for linearity

separate out linearity test from key bit finding

add file containing only the maxterms we believe to be good

Fix a maxterm based on my own tests

Make linearity test find out what kbits should be

Remove the ones now known to be good after fixes

Add the maxterms I sent to Dinur as a separate file

Fix and allow maxterms file to come from command line

Corrections received by email from Itai Dinur

Convert to using dict-orinted format most places

Bring README up to date with repo

Use new interface for benchmark

Sort bits first

Move towards new way of listing maxterms

Modify linearity test to look for key bits

Changes Itai Dinur recommends by email

add linearity tester

Improve the way we present results

make bit reversal part of maxterm object, and separate out parser so results parser can use it

Only list non-working maxterms

Simple results parser

Loop forever logging results

Update README for current state

Do away with _getSize, use TD not _TapDescriptor

Simplify indexing

Move key generation out of test_maxterm

list does this more neatly

Repeatedly run maxterm tests

use maxterm list

Handle smaller cubes

Let's copy-paste all maxterms from paper

Working cube attack with most of stuff in assembly

Partway through cube attack implementation

no need for exception handler in context manager

Use a context manager to manage loops

Simplify address calculations

add address calculations

Use array adapters to simplify parallel trivium

add little endian adapter

Add commented-out check on zip - can't do it that way

Add transforming adapter

Generalize ArrayAdapter

Rename to slice adaptor, generalize and add group adapter

Support slices

Array adapter works

Move some stuff into a buffer object

Neater way to generate code

Separate out Trivium from copy

doGetParam

move copy into its own thing

ParallelTrivium asks for size

Move Trivium code generation into Scode subclass

Try getting the output as ulongs - doesn't work

Time call overhead

faster without extarray

more tests

There's the problem...

Merge

Starting on work to make it faster

Add comment warning about Python overhead

Code has moved - fix printcode

cut out repetition from benchmark

Correctly handle variable length key/IVs

code generation not properly part of tap description, move it out

Use _TapDescriptor more consistently

minor tweaks to code generation

Fix line wrapping; remove half-begun sentence fragment

Link to blog post in README

Add my email and a link to the code

reduce repetition in copy code

line wrap README

simplify parity sum

Add a README and LICENCE file

Switch back to faster maxterm for testing

make key bits a variable

Remove now unnecessary exercising of corepy functions

Move looping into code generation

Add code generation helper object

add total_len to all offsets

Create a 'tap descriptor'

add a benchmark

Simplify interface to ParallelTrivium object

do test differently

Paper numbers bits backwards

Change the way we calculate parity

No IV to iterate over

Test that it really does do 128 at once

start work on testing cube attack. Doesn't work yet...

Test using random numbers

Add stuff for directly setting state words

make array longs not chars

Redesign for multi-shot use

Fiddle with instruction order

Print out assembly

Make parallel implementation object-oriented

rename misnamed variable

Loop inside assembly - still works

Start preparing to loop inside

add a .hgignore

Move parallel implementation into own module

It can't just be working can it?

corepy-impl proof of concept works

make existing impl into a module

Start on Trivium

Remove unused stuff

OK, it loops now - we're practically done :-)

it works, rah!