We only support the VariablePointersStorageBuffer feature for now,
which is the only one that is mandatory, and for which we seem to
be passing all the relevant tests already.
Exposing the optional VariablePointers feature would require that
we support non-constant indexing on UBO/SSBO first.
Relevant CTS tests:
dEQP-VK.*pointer*
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11257>
In multi-context scenarios, one context writing to a resource can race
with a pctx->flush_resource() on another context/thread. Which means
that by the end of flush_resource() we can have a new write_batch.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11200>
We need to remove the batch cache entry before marking as flushed.
Note that we are already holding the batch lock, to prevent flushing
something that another context is emitting cmdstream to, but there is
a window between batch cache lookup (under screen lock) and acquiring
the batch lock that could previously result in batch cache lookup
returning a flushed batch.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11200>
For the 3D RT read case on BDW, we can use the ISL helper. For the 1D
array case, we don't actually need to flatten; we can just smash it to a
2D array surface because the layout for 1D and 2D is the same on BDW.
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11235>
A number of the piglit glx tests use multiple contexts on a single
thread, and previously the flush in MakeCurrent() was enforcing the
ordering between draws on those different contexts. When that flush
made ASYNC, now there is nothing ordering the draws because we have
two (or more) driver threads for a single frontend thread which is
using nothing more than glxMakeCurrent() to enforce the ordering.
This reverts commit 057a702a3f.
Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4903
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11278>
using the stride to calculate the buffer map size here is not correct, as
the stride is not necessarily equal to the attribute size, it's only the distance
between elements. for the case of overlapping elements (cts does this), the result
is that the attribute conversion will read past the end of the mapped src region
this is usually fine for drivers when they directly map the vertex buffer, as the memory
past the requested region is usually made available, but in the case where the readback
occurs using a staging resource sized exactly to the map region, this overflows and fails
silently, not even triggering a valgrind error because gpu memory lol
instead, add the size of the largest possible element, which will automatically be clamped
and ensure any staging buffers are correctly sized
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10963>
The execmem option can be set to false to disable the dynamic
dispatch patching that requires mmap(PROT_WRITE | PROT_EXEC),
which is undesirable on some platforms.
Signed-off-by: John Bates <jbates@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10987>
Using the system/window.h header can potentially cause AHB breakage
because the system header is reserved for platform internal use.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11206>
Android WSI api contract requires to allocate min_undequeued_buffers + 2
to achieve "triple buffering" effect, which is when the composer backend
acquired the allowed max numbder of buffers, the producer still has 2
buffers to rotate.
ANativeWindow either belongs to SurfaceView which presents directly to
SurfaceFlinger or belongs to other surfaces from the UI framework. For
the former, SurfaceFlinger hardcodes triple buffering for EGL. For the
latter, the surface caps decide the buffer limits or HWUI intercepts and
adjusts the min_undequeued_buffers to hint the EGL implementation to
prepare enough buffer cache slots while HWUI sets the max dequeued
buffer count accordingly.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11206>
A bunch of tests are currently taking ~60 seconds, dangerously close to
the current 60 seconds timeout. Add some congesion or just bad luck to
a runner, or some other background-task, and we can very easily be
pushed above that.
Since we're rendering on LLVMpipe rather than HW, this isn't as
unreasonable as it might sound at first. So let's copy a trick from
VirGL's CI playbook, and bump the timeout to 180 seconds, three times
of what we currently use.
That might seem a bit excessive, but it's better to make it *very*
unlikely to cause trouble than a little bit unlikely.
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11272>
I've seen this timing out on Zink, v3d, radeonsi, and Panfrost
periodically, but more and more lately. Try to minimise the damage by
just universally skipping it until someone comes up with a better idea.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11267>
And also add the required bits to the x86_64 kernel.
syslogd is needed by Crosvm.
iptables is needed to route packets in and out the VM.
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10862>
this works by flagging the next barrier to use the current sample locations
so that everything works as expected during decompression
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11189>
The v_mbcnt instructions can take an extra source that they add to
the result. This is not exposed in SPIR-V but we now expose it in NIR.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
According to feedback, the terminology with "exclusive scan"
and "reduction" is difficult. Change it to use "repack" instead,
which better fits what this sequence is actually used for.
The new sequence stores only 1 byte / wave to LDS, and uses packed
instructions to produce the results. This has lower latency and
fewer instructions than what we previously had.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
When the result of these loads is only used by cross-lane instructions,
it is beneficial to use a VGPR destination. This is because this allows
to put the s_waitcnt further down, which decreases latency.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
These map directly to v_perm_b32 and v_permlane_b32.
Unfortunately there is no corresponding NIR opcode or
intrinsics, and it's too tedious to puzzle these things
together from the existing NIR instructions.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
Previously there hasn't been any validation for these instructions,
but after shooting myself in the leg with it a few times, I decided
to add the validation now.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
Fix up the operand size for v_sad instructions, and implement
the new NIR horizontal add. There is no viable way to do this
in SALU, so let's always use a VGPR destination.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
NIR currently doesn't have any intrinsics for a horizontal packed add,
so this one is modeled after AMD's v_sad_u8.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
We tried to use this instruction for a more optimal sequence,
but it turned out that it doesn't exactly work as it was
supposed to. This note is to help others who want to use it.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11072>
As a result of some previous changes, it is now possible to expose ES
3.1 without having the ARB_shader_image_load_store enable set. However
we still want those other extensions. When there's a driver that's
capable of exposing ES 3.1 but not these extensions, more caps can be
added.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11204>
