estimate its memory if its BufferedImage is a subimage of another
BufferedImage.
Include the amount of memory used by colorizations in ImageManager's cache
usage calculations.
Crank ImageManager's cache size up to 32 megs since a) it's actually seeing the
memory used by colorizations and b) the new seamonsters are pretty sizable.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@291 ed5b42cb-e716-0410-a449-f6a68f950b19
for what ultimately boils down adding this method to BundledComponentRepository:
@Override // from IMImageProvider
public Mirage getTileImage (String path, Rectangle bounds, Colorization[] zations)
{
// we don't need our images prepared for screen rendering
BufferedImage src = _imgr.getImage(getImageKey(path), zations);
if (bounds != null) {
src = src.getSubimage(bounds.x, bounds.y, bounds.width, bounds.height);
}
return new BufferedMirage(src);
}
So much fun to revisit code I wrote six years ago.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@283 ed5b42cb-e716-0410-a449-f6a68f950b19
things fail & attract attentino, rather than trying to work around it and printing
a stack trace that will likely go unnoticed amidst a slew of other build messages.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@273 ed5b42cb-e716-0410-a449-f6a68f950b19
for things like scales and rotates so 0,0 still meant the same place, but that
meant I couldn't just reposition something, since it would promptly undo the
transform I passed in.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@256 ed5b42cb-e716-0410-a449-f6a68f950b19
naturalOrder() which allows media to use a more consistent "same render order"
order resolution value than Object.hashCode().
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@248 ed5b42cb-e716-0410-a449-f6a68f950b19
four-bone-per-vertex configuration for all skinned meshes. Depending on
the models we have, it may be worth compiling a separate one for two
bones per vertex. Added a hook to customize the material color (for
per-vertex colors, e.g.)
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@245 ed5b42cb-e716-0410-a449-f6a68f950b19
I tracked the ATI slowdown down to using a vertex shader in combination
with fixed function fog. Also cleaned up ShaderConfig a little.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@244 ed5b42cb-e716-0410-a449-f6a68f950b19
their drivers can't do anything useful like report an error; they just
silently screw up. Limiting it to 31 makes it work, but there's still a
strange slowdown issue (and it's not that the shaders are being run in
software, but it's worth checking for that anyway).
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@243 ed5b42cb-e716-0410-a449-f6a68f950b19
weights even if we only have one weight per vertex.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@232 ed5b42cb-e716-0410-a449-f6a68f950b19
inaccurate System.currentTimeMillis() timer. Fortuately Sun introduced
System.nanoTime() in 1.5 which provides civilized timing information, so we can
use that instead of sun.misc.Perf which is unavailable in an applet sandbox. We
will need to go in and try to cope as best we can with currentTimeMillis() if
we want to support unsigned play in JDK 1.4 or earlier. We'll do that later.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@231 ed5b42cb-e716-0410-a449-f6a68f950b19
how to handle different lighting arrangements and complex materials
(e.g., the metal effect for the Iron Plate).
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@229 ed5b42cb-e716-0410-a449-f6a68f950b19
relative position throughout all animations. Also, don't include
transforms in the animations for nodes that never move.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@218 ed5b42cb-e716-0410-a449-f6a68f950b19
Charlie Groves