* In particular, if there is a mismatched layer registered under e.g. /etc then
we always need to elevate, even if the user wants their layer to be fixed and
registered user-local.
Introduce a method to check if a platform supports desktop GL.
Most platforms support desktop GL by default, so for them the
implementation of this method is trivial. For EGL, though, we
have to implement a proper check, since EGL support for non
GLES APIs is optional.
For Wayland support on GL we currently fall back to the GLES code path
just to get EGL. This commit uses the EGL platform directly for Wayland,
also removing the dependency on GLES.
EGL (>= 1.4) supports desktop GL, in addition to GLES. This commit makes
EGL support independently configurable at build time, to allow GL to use
it even if GLES is not needed or available.
* The histogram shouldn't be calculated as channel-wise averages, but instead by
'stacking' the channels (each channel contributes 1 to the appropriate
bucket).
* Degenerate channels (where min==max) are excluded and listed only as a single
spike, to avoid blowing out the automatic y-axis.
* If the UI was launched with a filename as a parameter to open the capture, it
will be added to the recent capture file list. Only later (relatively
speaking) if we make a capture connection will we realise that it is temporary
and potentially delete the file. If we do so, remove the capture from the
recent file list.
This ensures that ImgRefs/MemRefs exist for images/memory with initial
contents. Previously if the resource was never accessed, then no
information was saved about the resource references; this lack of
reference info is pesimistically interpreted at replay as requiring
complete reset.
This change now explicitly inserts empty ImgRefs/MemRefs for each
resource when preparing the initial contents.
Change-Id: I51930640cd22bc59b3436c920e29d9b670bd35cc
So that we can replay a draw multiple times.
Added m_EventResourceUsage that keeps track of resource usage
for an event, allows quickly figuring out if an event had any
shader RW resources, and only then doing an expensive part of
analyzing/rewriting a shader.
There are some situations that cause us to reset the memory backing an
image, but not the image itself--this puts the image in a corrupted
state.
This fix updates the image initialization logic to check the bound
memory. If the memory is written by any captured command, or if the
memory is reset, then the entire is image is assumed to be
uninitialized.
Change-Id: Ifb1ec6bfd83f93def0818dbabcb0ea57f0febe54
baldurk