This might be useful to avoid breaking the current compute state when
monitoring MP perf counters because we use a compute kernel to read out
those counters. This has been initially suggested by Ilia Mirkin.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
It's not doing anything according to shader-db now that we're using NIR.
It would have had to be reworked significantly anyway, to handle control
flow.
We were generating piles of FRAG_W for interpolation, only to CSE them
away immediately. Since this is the only thing that CSE is doing for us
any more, just avoid making the CSE work necessary.
This is one of two uses of the current QIR CSE pass according to
shader-db. The NIR pass means that we'll end up doing Newton-Raphson on
our RCP, which we weren't doing before, but that's probably actually a
good thing.
That format has first_non_void < 0. This fixes a regression in piglit
arb_shader_image_load_store-semantics that was introduced by commit 76b8c5cc60,
while hopefully still shutting Coverity up (and failing in a more obvious way
if a similar error should re-appear).
Reviewed-by: Jakob Sinclair <sinclair.jakob@openmailbox.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The GL4.1 spec bumps this to 2048, so we should do so.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The file(s) within are already picked thanks to the build rule of the
respective test. No need to have the folder in EXTRA_DIST.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
From Section 7.3.1.1 (Naming Active Resources) of the OpenGL 4.5 spec:
"For the property LOCATION_COMPONENT, a single integer indicating the first
component of the location assigned to an active input or output variable is
written to params. For input and output variables with a component specified
by a layout qualifier, the specified component is written. For all other
input and output variables, the value zero is written."
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
I don't think this will do much as it's a compiler error
to use component without location which is already in the
table but its good to be consistent.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
This is needed so we don't optimise away the varying when more than
one shares the same location.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This change checks for component overlap, including handling overlap of
locations and components by doubles. Previously there was no validation
for assigning explicit locations to a location used by the second half
of a double.
V3: simplify handling of doubles and fix double component aliasing
detection
V2: fix component matching for matricies
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We make use of the existing IR field location_frac used for tracking
component locations.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
align_malloc() is used to allocate dri_sw_dt->data, thus we should not
be using FREE() but align_free().
Cc: "11.2 11.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
[Emil Velikov: tweak commit summary/shortlog]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Add support for creating images from Android native buffers with dma-buf
fd. As dma-buf support also requires DRI image loader extension, add
that as well.
This is based on several originally patches written by Varad Gautam.
I've collapsed them into logical changes and done a bit of reformatting.
Using dma-bufs vs. GEM handles is now a runtime decision similar to the
wayland EGL instead of being compile time selection. The dma-buf support
is also re-written to use common dri2_create_image_dma_buf function in
egl_dri2.c.
Cc: Varad Gautam <varadgautam@gmail.com>
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
In preparation to use the same code for dma-bufs, factor out the code to a
separate function.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use of __DRI_DRI2_LOADER extension is only supported for card nodes. In
order to support dmabufs, Android will be moving to using render nodes and
we need to disable the DRI2 loader extension.
This is based on the Wayland EGL code.
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Different versions of make behave differently in whether $(wildcard) sorts
the results or not. The Android build now explicitly sorts
all-named-subdir-makefiles which breaks the build because src/gallium
must be included after src/mesa/drivers/dri.
The Android build system doesn't support doing "include $(call
all-named-subdir-makefiles,...)" twice, so rework things by generating
the included makefile list and including them in 2 steps.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
OpenGL 4.5 Core Profile section 7.1, in the documentation for
CompileShader, says: "Changing the source code of a shader object with
ShaderSource does not change its compile status or the compiled shader
code."
According to Karol Herbst, the game "Divinity: Original Sin - Enhanced
Edition" depends on this odd quirk of the spec.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93551
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Missed with commit 100796c15c "gallium/radeon: drop support for LLVM
3.5"
v2: s/LLVN/LLVM/ in shortlog (Nicolai)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v1)
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Namely the python script, the ICD header and private headers. We could
get the system version of the ICD ones, although there is no .pc file to
easily locate and/or manage them.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Rather than using cat + cpp feed the file(s) directly into the latter.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Otherwise we risk picking the possibly outdated file in the source dir
over the fresh one in the builddir.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
