Closed GoogleCodeExporter closed 8 years ago
I need the DXT1a support to.
(It works with older NVidia compression lib)
Original comment by kacem.be...@gmail.com
on 22 Jun 2007 at 3:11
Here is a patch that enables DXT1a fast compression. I tried using the weighted
alpha
option for squish, however, it didn't work properly for me because it doesn't
allow
zero weights. Let me know what you think.
Original comment by ae.pas...@gmail.com
on 25 Jul 2007 at 12:59
Attachments:
That looks pretty good. Thanks!
I'd improve 'compressBlock_BoundsRangeAlpha' to only use the 3 color mode when
the
given block has any transparent texel. So, when it's entirely opaque, I'd still
use
the 4 color mode, that in general produces better results.
In squish the approach should be similar, force the 3 color mode when
transparent
texels are present in the current block. Rather than setting the weights to
zero, you
could set them to 1, so that they still influence the result, but not in a
significant way.
Original comment by cast...@gmail.com
on 25 Jul 2007 at 1:11
Good point. I guess I was treating all blocks the same when blocks can alternate
between the 3 color and 4 color modes. I'll take a look at it again when I get a
chance. Thanks!
Original comment by ae.pas...@gmail.com
on 25 Jul 2007 at 5:22
I'd like to third(4th) the dxt1a support request. CUDA support would be ideal,
but
just getting a fully supported version in there soon would be quite useful.
Thanks!
Original comment by brdms...@gmail.com
on 1 Aug 2007 at 10:14
I'm working on this right now. In a couple of days I'll submit a fast compressor
based on aa.pascal's contribution and some improvements that I've done. I'll
finish
the high quality compressor and CUDA's implementation during the next two weeks.
Original comment by cast...@gmail.com
on 1 Aug 2007 at 10:23
bump! :)
Any news on this?
Thanks!
Ben
Original comment by brdms...@gmail.com
on 11 Sep 2007 at 9:28
I've added an improved DXT1a fast compressor. I still don't have the high
quality
compressor, but the improved fast compressor is much better than the old one,
although there's still room for improvement. Please, checkout the code in the
subversion repository and let me know what do you think.
I'll resume the work in the high quality compressor as soon as I finish
refactoring
the fast compressors (you can find the new code in QuickCompress.{h,cpp}).
Original comment by cast...@gmail.com
on 11 Sep 2007 at 10:13
Original comment by cast...@gmail.com
on 28 Nov 2007 at 3:14
Just checking in again - we're basically at the stage where we'd like to fully
integrate a CUDA compression solution, but high quality DXT1a is important for
us to
continue. If I understand the current state of things, I can get high quality
images for everything but DXT1a. For DXT1a, I would have to disable CUDA to
get
high quality images - is this correct? Thanks!!
Ben
Original comment by brdms...@gmail.com
on 14 Dec 2007 at 2:30
brdmster, there's a small amount of work needed to enable high quality DXT1a
compression in the CPU. For GPU compression there's a bit more work required.
So,
right now, only the quick DXT1a compressor is available.
The DXT1a format will be fully supported for the next beta, that will be
released
early next year. I'll keep you posted as soon as I make more progress.
Original comment by cast...@gmail.com
on 14 Dec 2007 at 2:45
OK, DXT1a is supported now. The CUDA accelerated compressor will be added later,
possibly after the beta.
Original comment by cast...@gmail.com
on 4 Feb 2008 at 7:55
I've opened Issue 29 to trace progress of the CUDA accelerated DXT1a compressor.
Original comment by cast...@gmail.com
on 4 Feb 2008 at 8:03
Original issue reported on code.google.com by
cast...@gmail.com
on 17 Apr 2007 at 9:00