always use argb8888 instead of rgba8888, this is far more likely to hit a faster memcopy path in mesa (at least on little endian systems, possibly on big endian too)
Always mark tex state atom as dirty when the texture image is dirty, this ensures texture cache gets flushed in case the new texture has the same offset as the old one (fixes glean pixelFormat test at least on r200).
fix GL_REFLECTION_MAP texgen by not using tex matrix negation when lighting is not enabled (?). This fixes demos/cubemap as well as glean texCube test.
Removing some of r200 dependency, cleaning up code a bit,
and fixing a couple of warning. More cleanup and shuffle
to come. I have tested this change they might broke things
especialy with r300_texstate.c change (format_x doesn't
seems to be use at all by r300).
Potential fix for doom3 lockups. Seems that there is a conflict
between the vertex cache, the vertex shader and the clipping stages,
all of which are competitors for URB entries assigned to the VS unit.
This change reduces the maximum number of clip and VS threads by
enough to ensure that they cannot consume all the available URB
entries, and then reduces the number somewhat more up to an arbitary
amount I discovered by trial and error. Unfortunately trial and error
solutions don't inspire total confidence...
remove code to deal with non-normalized texture coordinates for tex rect targets (swtcl tex coord translation stage, tcl tex matrix adaption) and use the chip's native handling of such coords instead (!!!). Seems noone noticed those bits in the se_coord_fmt reg, even though it works fairly similar to r200 (except it's set per-unit and always active, so only enable it for texture rectangles).
Redefine the BYTE_TO_FLOAT() and SHORT_TO_FLOAT() macros to avoid problems
converting 0 to exactly 0.0
This goes against the OpenGL spec (see table 2.6) but solves problems
when doing byte -> float -> ubyte conversion for particular texture formats.
Basic facility for playing back captured aubfiles. Requires a small
hack to the drm to disable command verification on the cmd_buffer
ioctl. Doesn't exactly replay as commands are normally delivered as
batchbuffers but are captured and replayed as commands on the ring.
cause a vtxfmt fallback directly when hitting NewList for radeon and r200 drivers, otherwise we'll mix up tcl and vtxfmt path for some reason. This fixes a warzone2100 asssertion failure.
Cope with memory pool fragmentation by allowing a second attempt at
rendering operations to take place after evicting all resident
buffers.
Cope better with memory allocation failures throughout the driver and
improve tracking of failures.
make really sure R200_VAP_PROG_VTX_SHADER_ENABLE is never set when we're already in a tcl fallback, otherwise the chip will instantly lock up when vertex progs are enabled the next time not in a tcl fallback (fixes for instance guaranteed lockup running any program which uses vertex progs with tcl_mode=0 and then later with tcl_mode=1).
only allow VERT_ATTRIB_MAX instead of _TNL_ATTRIB_MAX for inputs of vertex programs (fixes a segfault since the result of the shift is undefined otherwise, and it may happen that _TNL_ATTRIB_POINTSIZE will be tried to read, unlike all other attribs this however may be unitialized (might be a bug in itself)).