| Age | Commit message (Collapse) | Author |
|---|
|
|
|
This moves the bo mapping outside the DMA layer and makes it explicit,
this should in theory make it simpler to split the clean up the dma/cmdbuf
linkage that I created before that is broken.
Tested on: r600, rv380 (tcl/no-tcl), rv200 (tcl/no-tcl)
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
|
|
Only very few places where realy converted so there isa lot of to do.
|
|
Problem was to find the correct place to run prediction. Only place that is
called for every primitive is ALLOC_VERTS so we have to do prediction there
before allocation.
|
|
Swtcl calls flush everytime primitive changes so prediction has to made again
after flushing.
|
|
We were check command buffer sizes too alte so allocated dma regions
were freed before relocations so space checking failed.
|
|
|
|
Trying to make understanding code easier with small refactoring and renaming.
|
|
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
|
|
We keep dma buffer objects in list untill they have been unused for many
draw operations. Current limit of having 100 flushes is just guess for
good performance/memory trade off.
Moving WARN_ONCE macro to common context because it is used in multiple drivers.
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
|
|
The duplication of state data caused a crash due to double-free on destruction
of context, because a variable wasn't correctly null'ed out.
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
Context destruction was nearly the same over all the drivers,
so collapse it down.
|
|
|
|
This merges lots of the hw state atom emission and firevertices code.
it also removes a lot of the extra radeon crap from r300
and merge scissor
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Makefile.template
|
|
|
|
|
|
nearly the same as outlined in bug #4707, except it disables perspective correction for point sprites to make them actually work. And, separate the state atom into two as the tcl parameters would overwrite vertex program parameters when active. Also implement the GL_VERTEX_PROGRAM_POINT_SIZE_ARB option to make vertex programs outputting a point size work correctly (untested). Smooth points will still always be size 1. While here, enable gouraud shading for fog when using fog coord.
|
|
- use macros to access and modify render inputs bit-field;
- un-alias generic vertex attributes for ARB vertex calls;
- use MAX_VERTEX_PROGRAM_ATTRIBS (NV code) or MAX_VERTEX_ATTRIBS
(ARB code) in place of VERT_ATTRIB_MAX;
- define VERT_ATTRIB_GENERIC0..15 for un-aliased vertex
attributes for ARB_vertex_shader;
- fix generic attribute index range check in arbprogparse.c;
- interface GLSL varyings between vertex and fragment shader;
- use 64-bit optimised bitset (bitset.h) for render inputs;
|
|
radeon driver when a tcl fallback is active, fixes a blender issue with non-tcl hw (bug #5601)
|
|
|
|
point sprite primitive.
|
|
|
|
during fallbacks. In one case, _swsetup_Wakeup had just been called, covering
the need there, and in the other case, we can simply exit the entire
radeonChooseVertexState function, knowing that it will be called again once we
leave the fallback.
Bugzilla #: 2516
Submitted by: sroland
|
|
remove those descriptions and the corresponding bits on r200, where they're no
longer used.
|
|
|
|
|
|
coordinate of the vertex (bugzilla #1648)
|
|
since they'll get reduced to quads sometimes if hw quads are enabled. But this needs more thought, since it looks like clipped primitives will always be emitted as polys, so the reduced primitives for triangles, quads, quads_strips, polys all need to be the same, otherwise get lockups with for instance the olympic test. Render templates would probably need to be changed for this to work.
|
|
|
|
and ignored for 1D and 2D images.
Need to pass in image dimensions (1,2,3) to the _mesa_image_address() function.
This change gets propogated to some other routines.
Also added new _mesa_image_address[123]d() convenience functions.
|
|
fog coords (it seems the chip cannot do fog factor computation when not using fragment depth as fog coord source). vtxfmt uses fallback for now (most code present but some magic would be needed if replaying vertices is necessary later on).
|
|
|
|
coordinate generation. Original code by Roland Schiedegger, with changes by
myself. While here, ensure that the swtcl path does tnl_install_attrs enough
when fog/specular are being (en/dis)abled.
Notable effects:
- projtex test works with TCL and is closer with swtcl (Bugzilla #1461)
- 8/9 squares work in texgenmix instead of 3.
- texcyl "reflect" mode works (GL_SPHERE_MAP is now a fallback -- unclear if the
hardware can actually support it).
- flickering in doom3 replaced by just plain darkness.
- blocktube fixed (Bugzilla #984)
- fixes stex3d
|
|
errors on r100 and rendering errors and hangs on r200 (same for R100 without
OLD_PACKETS).
If a command buffer filled after some state (EmitState or a VBPNTR write) was
emitted, the lock was grabbed, the buffer flushed, a new buffer prepared, and
the lock dropped. Another client could come in, set its own state as part of
rendering, and when the first client flushed the rendering commands depending
on the previous state, it got the 2nd client's state. This is fixed by checking
for enough space before beginning a set of state emits and rendering, and
flushing the buffer first if so. This guarantees that the buffer won't wrap.
Also, move the "lost_context = 1" from the end of cmdbuf flushing to
UNLOCK_HARDWARE for clarity (at a minimum) that any time the lock is dropped,
state may get overwritten. We don't have enough information at the point of the
LOCK_HARDWARE to reset our state to the last UNLOCK_HARDWARE point in the case
that we did lose our context, but saving the information to rebuild that state
may be a useful optimization (ipers data suggests up to 5%).
|
|
|
|
|
|
workarounds only for r200, not derivatives.
|
