So do not offset it again when processing the fallback composite
operation.
Regression from commit 6921abd810
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Thu Jul 18 16:21:27 2013 +0100
sna: Add a fast path for the most common fallback for CPU-CPU blits
References: https://bugs.freedesktop.org/show_bug.cgi?id=66990
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As a first step towards working out what to do with the remaining
used-once PCI IDs, delete the used-never ones.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Typically, PutImage is not a performance critical path since ShmPutImage
uses CopyArea and so PutImage is relegated to small one off transfers.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The codename changed midcycle - along more rational lines (all the chips
within the platform are now part of the Baytrail family rather than
different codenames for each).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the macros conflict with another definition on the system, that is
actually a useful warning in this case as it means there exists a typo
somewhere.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When printing out the list of supported chipsets, remove the (many
times) repeated names as they are very confusing and obscure the product
name you may be searching for.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
For these parts we do not yet known the correct marketing name, so
replace our use of the codename with the generic term.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we recognise the PCI-ID but do not have a marketing name for it, it
means we are deliberating ignoring it - presumably because it is an
engineering sample. In this case, we do not want to print out a warning
into the logfile so replace the "Unknown chipset" with some info
about the future product.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Rather than duplicating the information we already use in the kernel, we
can reuse the pci-id tables so long as we apply a little fuzz.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Drop the GLX includes as they are not needed and obviously break if the
GLX are not present on the system (and we fail to check during
configure).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This allows us to avoid the occassional kernel bug if mmap() fails and
the upload is possible via our indirect upload paths.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Instead of checking that the DRI2Buffers match up with the current
DRI2Drawable, clip the area that is being copied to the minimum extents
of the Drawable, source and destination buffers. This ensures that we
never read nor write beyond the extents of the allocated or visible
regions.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67210
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67305
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This reverts a portion of commit 6d80bd6a73 so that we do not compare an
offset redirected window against its outer frame (e.g. glxgears in a
300x300 under unity sits within a much larger ~330x330 frame).
References: https://bugs.freedesktop.org/show_bug.cgi?id=67210
Reported-by: Joseph Yasi <joe.yasi@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This bogus comment confused both Chris Halse Rogers and myself.
Suggested-by: Christopher James Halse Rogers <raof@ubuntu.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Prompted by a suggestion by Haihao, clarify the intent behind checking
the incoming maximum blit extents against the recorded sizes of the
attached bo. Due to the asynchronous nature of DRI2 invalidation, it is
possible for the DRI2 buffer to be stale and for its bo to be smaller
than required for the client's blit.
References: https://bugs.freedesktop.org/show_bug.cgi?id=67210
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: haihao <haihao.xiang@intel.com>
Storing only the low 16-bits of the key for the pipelined state meant
that we forced an update with every new drawing op - with the side
effect of flushing the render cache.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67157
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In a Zaphod configuration, the set of CRTCs available for an output is
limited, and if that set does not match with the already established
linkage, our quick probe function will select no mode. However, since
the output is already on, the user definitely would like to continue
using it, so fallback to InitialConfiguratio to select an appropriate
mode.
Reported-by: Nick Bowler <nbowler@draconx.ca>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67176
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
QA complained that the full flush to memory was too much of a
performance hit, so let's try a lighter weight non-stalling pipe flush
without.
References: https://bugs.freedesktop.org/show_bug.cgi?id=67157
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Mostly for symmetry with GetImage, this in theory just bypasses a lot of
intermedate state checking.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Often we may find that the BLT operations we wish to do after a
XY_BLT_SETUP are either completely clipped out or transparent and so do
not generate any subsequent commands. This leaves us with many unused
XY_SETUP_BLT.
References: https://bugs.freedesktop.org/show_bug.cgi?id=61477
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Yet another missing chunk from
commit 6921abd810
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Thu Jul 18 16:21:27 2013 +0100
sna: Add a fast path for the most common fallback for CPU-CPU blits
I had the composite offsets, but not the normal window offsets!
Reported-by: F.Brown <francisbrwn9@gmail.com>
Reported-by: Jiri Slaby <jirislaby@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67124
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67064
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Along some paths, we use a CPU mapping on a temporary object that never
gets a user ref - so the assertion is bogus. A lesson I keep forgetting.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
A regression in
commit 6921abd810
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Thu Jul 18 16:21:27 2013 +0100
sna: Add a fast path for the most common fallback for CPU-CPU blits
And add the full set of assertions to validate the memcpy.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Egbert Eich
Chris Wilson