An artefact of retaining the mmapped partial buffers is that it
magnified the effect of stealing those for readback, causing extra
writes on non-llc platforms.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Bo used for batch buffers are handled differently and not tracked
through the active cache, so we failed to notice when we might be able
to run retire and recover a suitable buffer for reuse. So simply always
run retire when we might need to create a new linear buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we change tiling on a bo, we are effectively discarding the cached
mmap so it is preferable to look for another.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We use the XF86DRI as a user configurable option to control whether to
build DRI support for i810, but it is also used internally within xorg
and there exists a public define in xorg-server.h which overrides our
configure option. So rename our define to HAVE_DRI1 to avoid the
conflict.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=46590
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If you are suffering from regular X crashes and rendering corruption
with a flood of ENOSPC or even EFILE reported in the Xorg.log, try
adding this snippet to your xorg.conf:
Section "Driver"
Option "BufferCache" "False"
EndSection
References: https://bugs.freedesktop.org/show_bug.cgi?id=39552
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As gen3 only uses the single state emission block, and uniformly calls
get_rectangles(), we can move that caller protocol into the callee.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As we prematurely end the batch if we bail on extending the vbo for CA
glyphs, we need to force the flush.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The heuristic of using the mapping only before the first use in an
execbuffer was suboptimal and broken by the change in bo initialisation.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Due to the w/a for its buggy shaders, gen4 is significantly different
that backporting the simple patch from gen5 was prone to failure. We
need to check that the vertices have not already been flushed prior to
flushing again.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Or upon actually closing the vertex buffer.
However, the underlying issue remains. That is we are failing to re-emit
the first-pass for CA text after flushing the vertex buffer (and so
emitting the second-pass for the flushed vertices).
Reported-by: lemens Eisserer <linuxhippy@gmail.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=42891
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As we may wait upon the bo having finished rendering before we can
execute the flip, flushing the render cache as early as possible is
beneficial
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Before blocking and waiting for further input, we need to make sure that
we have not developed too large a queue of outstanding rendering. As we
rendering to the front-buffer with no natural throttling and allow X
clients to render as fast as they wish, it is entirely possible for a
large queue of outstanding rendering to develop. For such an example,
watch firefox rendering the fishietank demo and notice the delay that
can build up before the tooltips appear.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we change the Screen pixmap due to a change of mode, we lose the
flag that we've attached a DRI2 buffer to it. So the next time we try to
copy from/to it, reassert its DRI2 status.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
i965_video.c: In function 'gen6_create_cc_state':
i965_video.c:1374:12: warning: passing argument 4 of
'intel_bo_alloc_for_data' discards 'const' qualifier from pointer target
type [enabled by default]
Repeated ad nauseam.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
For a singular region, we want to use a value for nboxes of 0 not 1,
fortunately if you pass in a box, it ignores the value of nboxes.
RegionInit() is a most peculiar API!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Two fixes in this commit, first we only need to check the
front left buffer, for other attachment we don't need to
check them. The second is, we should fixup the pixmap's
drawable not the original drawable.
Signed-off-by: Zhigang Gong <zhigang.gong@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
To support easy buffer exchange at glamor layer, glamor
added a new API glamor_egl_exchange_buffers() to exchange
two pixmaps' EGL image and fbos and textures without
recreating any of them. But this simple method's requirement
is that there are two pixmaps. A exceptional case is:
If we are using triple buffer when do page flipping, we
will have an extra back_buffer which doesn't have a pixmap
attached to it. Then each time we set that buffer to a
pixmap, we will have to call the create_egl_textured_pixmap
to create the corresponding EGL image and fbo and texture
for it. This is not efficient.
To fix this issue, this commit introduces a new back_pixmap
to intel structure to hold the back buffer and corresponding
glamor resources. Then we will just need to do the light
weight buffer exchanging at both DDX and glamor layer.
As the new back pixmap is similar to the screen pixmap
and need to be handled carefully when close screen. As the
glamor data structure is a per screen data, and will be
released at its close screen method. The glamor's close
screen method must cleanup the screen pixmap and back
pixmap's glamor resources. screen pixmap is easy to get,
but there is no good way to store the back pixmap.
So the glamor add a new API glamor_egl_create_textured_screen_ext
function to pass the back pixmap's pointer to glamor layer.
This commit make us depend on glamor commit: 4e58c4f.
And we increased the required glamor version from 0.3.0 to 0.3.1
Signed-off-by: Zhigang Gong <zhigang.gong@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Add a new element back_name to intel structure to track
the back bo's name then avoid flink every time.
And at function I830DRI2ExchangeBuffers, after finish
the BO exchange between info's front and back pixmap,
it set the new front bo to the screen pixmap. But the
screen pixmap should be the same as front's pixmap,
so this is a duplicate operation and can be removed.
Signed-off-by: Zhigang Gong <zhigang.gong@linux.intel.com>
In order to prevent walking upwards off the top of the pixmap when
rendering a clipped vertical edge, we need to tweak the boundary
conditions for the vertical edge walker.
Reported-by: Clemens Eisserer <linuxhippy@gmail.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=46261
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
On gen4, the tiling/fence constraints are fairly lax, only requiring
page alignment of the object and its size, and so we can switch
tiling modes without incurring a GPU stall on active bo.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The idea was that we could afford to allocate an active CPU bo for
copying to from using the GPU and later sync just before we need to
write to the shadow pixels. Having the sync inside the allocation
function potentially causes an unwanted stall.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson
Zhigang Gong