Permalink
Switch branches/tags
Nothing to show
Commits on Aug 4, 2011
  1. salt parameter mistakenly got put under Required Options

    Terrence Cole committed Aug 4, 2011
  2. remove references to older algorithms

    Terrence Cole committed Aug 4, 2011
  3. update README and javadoc overview in preparation for 1.1

    Terrence Cole committed Aug 4, 2011
  4. Merge branch 'vash1.1'

    Terrence Cole committed Aug 4, 2011
Commits on Aug 3, 2011
  1. the sources for the vash multiplication blog post

    Terrence Cole committed Aug 3, 2011
  2. improve documentation coverage for Tree classes

    Terrence Cole committed Aug 3, 2011
  3. remove loudness; it was not useful

    Terrence Cole committed Aug 3, 2011
  4. document nextDouble and nextInt in seed

    Terrence Cole committed Aug 3, 2011
  5. add some helper scripts we have been using

    Terrence Cole committed Aug 3, 2011
  6. ensure that this works, even if we change the ordering

    Terrence Cole committed Aug 3, 2011
  7. document these two exceptions and add copyright header

    Terrence Cole committed Aug 3, 2011
Commits on Aug 1, 2011
  1. list all algorithms in --known-algorithms

    Terrence Cole committed Aug 1, 2011
    We add a (deprecated) tag to old algorithms, but continue
    to list them.
  2. ensure that salt and data don't get set to the same file

    Terrence Cole committed Aug 1, 2011
  3. add test for same file sent to salt and data

    Terrence Cole committed Aug 1, 2011
  4. add a debug flag to print out entropy bits used

    Terrence Cole committed Aug 1, 2011
  5. add --debug-tree to vash options

    Terrence Cole committed Aug 1, 2011
  6. support Tree.show on stdout with - as the filename

    Terrence Cole committed Aug 1, 2011
  7. test that salting for 1/1-fast are doing what we expect

    Terrence Cole committed Aug 1, 2011
  8. pass the salt through to the underlying creation

    Terrence Cole committed Aug 1, 2011
Commits on Jul 30, 2011
  1. limit our plane cache size to 64MiB (total with outstanding)

    Terrence Cole committed Jul 30, 2011
    Testing shows the cache saves some time on small images
    but is basically a memory sink on larger images.  64MiB
    seems to be reasonable place to set a cutoff.
  2. always allocate our output plane after visiting our children

    Terrence Cole committed Jul 30, 2011
    This saves us lots of memory in some cases because we
    end up being able to re-use a cached plane from our
    child's computation, and we hold less memory while
    our entire child's tree computes.
  3. remove unused from LinearGradient1, now that it works

    Terrence Cole committed Jul 30, 2011
  4. require cli salt to be base64 encoded

    Terrence Cole committed Jul 30, 2011
    This should remove all encoding ambiguity and make things
    easier to use.  It's not as if we ever want an ascii salt.
Commits on Jul 29, 2011
  1. obey channel exclusions when building the tree

    Terrence Cole committed Jul 29, 2011
  2. use salt in 1 and 1-fast by appending to data

    Terrence Cole committed Jul 29, 2011
  3. take salt when passed on command line

    Terrence Cole committed Jul 29, 2011
Commits on Jul 28, 2011
  1. add license text to top of makefile

    Terrence Cole committed Jul 28, 2011
Commits on Jul 27, 2011
  1. add 1.1 algorithm test set to tree builder

    Terrence Cole committed Jul 27, 2011
    We don't have golden output for this yet.  We will
    probably use this test with asserts disabled to
    generate it once we get our exact parameters nailed
    down.
  2. add a new, saner linear gradient operation

    Terrence Cole committed Jul 27, 2011
    This uses vector math instead of slope-intercept form.
    Much simpler and more efficient this way.  Not sure why
    I didn't do it this way the first time around.
Commits on Jul 26, 2011
  1. allow the seed to provide int values

    Terrence Cole committed Jul 26, 2011
    We need this for our new per-channel calculators.  We
    also allow this to be used with the older seed functions,
    but this will never actually call them, since we only
    use it with newer algorithms.
  2. compute total for guided random walk on the fly

    Terrence Cole committed Jul 26, 2011
    We used to pre-compute these, for performance.  This
    is not going to be possible with each channel of the
    tree using a different subset.  We may also need, at
    some point, to modify the node frequencies on the fly.
    We can revisit this later if we find a performance
    bottleneck here.
  3. mark XTR_SIZE, a static final documenation int, as unused

    Terrence Cole committed Jul 26, 2011
  4. fix an odd auto-formatting anomaly

    Terrence Cole committed Jul 26, 2011
  5. implement random channel selection/masking logic

    Terrence Cole committed Jul 26, 2011
  6. rename opRatios to ops to fit new usage

    Terrence Cole committed Jul 26, 2011