On 08/13/2013 01:27 PM, Roland Mainz wrote:
On Tue, Aug 13, 2013 at 10:20 PM, Ian Romanick <i...@freedesktop.org> wrote:
On 08/13/2013 11:53 AM, Uwe Schmidt wrote:
[snip]
SO MY PROPOSAL:

If 'format' is one of the S3TC types, and format!=internalFormat is true,
then set internalFormat:=format.

'format' cannot be a compressed type.  Compressed data can only be supplied
using glCompressedTexImage2D, and that function only has an internalFormat
parameter.


Else, if 'internalFormat' is one of the S3TC types, but the 'format'
isn't,
set internalFormat:=format (or any other format, Mesa3D can encode).


The only format that Mesa can encode is FXT1.  Only Intel hardware supports
FXT1, and the quality (of Mesa's compressor) is not very good.  Picking that
format would result in bug reports of "game XYZ looks horrible on Intel
graphix you suck."  So that leaves us with the only option of leaving the
data uncompressed.

Until S3 grants it's IP to OIN or the patents expire, this is going to be
the situation.  We've been through this mental exercise of the last 5 years
more times than I can count, and we always come back to the same place.

Please don't hit me with a stick if this has been asked
|powl(INT64_MAX, INT64_MAX)| times... but... erm... adding the code
(sample implementation, "... do not use without a license from S3 and
the ritual scarification of at least one software engineer...") but
having it off in the default build won't work... right ?

That is more difficult for end-users than the current situation. In the current situation, your distro can build Mesa (no S3TC), and, if you live in a country without software patents, you can just "drop in" the libtxc_dxtn library to get compression. Putting it in Mesa, along with making the distros really uncomfortable, would mean you'd have to rebuild Mesa.

Did I mention that we've been through this mental exercise a few times?

----

Bye,
Roland


_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to