| Age | Commit message (Collapse) | Author |
|---|
|
Effectively default to vblank_mode=3 on Intel to avoid tearing by default.
Users wanting to go "as fast as possible" (despite not being able to see frames
faster than their refresh rate allows) can still set the vblank_mode manually.
|
|
|
|
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=16123 .
|
|
|
|
|
|
There are 6 subpixel units per pixel, not 16.
|
|
|
|
This fixes piglit's fp-kil and fp-generic/kil-swizzle tests.
|
|
While R500 fragment program texture instructions appear to support writemasks,
they cannot write to the output FIFO immediately, so we need to insert a MOV
for these instructions.
This fixes piglit's fp-fragment-position and fp-incomplete-tex tests.
|
|
Fixes oglconform rbGetterFuncs testcase. The span code for this mode hasn't
actually been tested.
|
|
Fixes piglit's tex1d-border test.
|
|
This fixes one type of lockup I've been seeing on my test system.
|
|
Conflicts:
src/mesa/drivers/dri/common/dri_bufmgr.c
src/mesa/drivers/dri/i965/brw_wm_surface_state.c
|
|
Previously, we would just store the data as malloced memory hanging off the
object, which would get memcpyed in at validate time. This broke an
oglconform render-to-texture test, since validate wasn't called but a miptree
was expected.
|
|
|
|
This reverts commit 94979950e8991bd44899eb4067c3ae43449ce51e.
I've fixed it instead
|
|
should fix fd.o 14441
|
|
patch from Fedora. maybe someone can fix this later but for now
lets try and release Mesa so ajax can live his life and get Xorg 7.4 out.
|
|
Accidentally pushed as part of the last commit.
|
|
This reverts commit b993d539a76e7f1446890a85e4b61deec4d4162d. The
patch was applied incorrectly. Actual fix coming soon. Sorry for the
noise.
|
|
Numbers of destination depth registers corrected (destination stencil
register was sent as depth register).
|
|
|
|
This will avoid clflushing entire buffers for small acesses, such as those
commonly used by regression tests.
|
|
|
|
Didn't hurt 915, but needed for 965.
|
|
This lets us avoid duplicated code for doing so, including the depthstencil
paths that aren't covered by SpanRenderStart/Finish. Those paths were
missing the span funcs setup, leading to a null dereference in the fbotexture
demo.
|
|
Some buffer swap intel render buffer fields (pf_num_pages & vbl_pending) are
also used for page flipping, so enable the code that sets & updates them on
965. This allows buffer swaps and page flips to work on 965 and prevents hangs
in LOCK_HARDWARE in the buffer swap case due to an uninitialized vbl_pending
field.
Fixes FDO #16118.
|
|
|
|
Bump the driver date and insert the string "GEM". When running tests,
this make it much easier to know that the right driver is being used.
|
|
|
|
Enable support for ARB_texture_env_crossbar in the master extension
list instead of in every single device-specific list.
|
|
|
|
Those functions rely on being able to treat the GET_PTR returned value as an
array indexed by x, but that's not the case for our tiling.
Bug #16387
|
|
|
|
This patch is a straightforward duplication of the R200 SetTexOffset code,
except that there is no big-endian tx_table[] array.
|
|
|
|
Share almost all code with r500_fragprog now.
This also fixes Piglit's texrect-many test, which means that the compiz
bicubic plugin should work with hardware acceleration now.
|
|
This just makes the use of mklib more consistent throughout Mesa where
we always want to pass the linker and LDFLAGS when we might be making a
shared library.
|
|
Respect the user's choice of shell when running mklib rather than always
using /bin/sh.
|
|
|
|
|
|
When an input is marked in gl_program.InputsRead but is not actually read
in the final program (due to dead-code elimination or whatever), the order
of input registers must still match gl_program.InputsRead. This is done
even more explicitly now.
|
|
This fixes the last r500 bug related to glean/fragProg1.
|
|
Use an abstracted instruction scheduling and register allocation algorithm
that we will be able to share with r300_fragprog.
Unlike the original emit code, this code tries to pair instructions that
only use the RGB part of the ALU with instructions that only use the alpha
part. However, the pairing algorithm still has some shortcomings;
for example, it doesn't generate optimal code for the emulation of LIT.
|
|
In addition, this pass fixes non-native swizzles.
|
|
|
|
|
|
based on info from hw team
|
|
|
|
pointed out and debugged by stringfellow on #dri-devel
|
